Causing Human Actions

Causing Human Actions

New Perspectives on the Causal Theory of Action

edited by Jesús H. Aguilar and Andrei A. Buckareff

A Bradford Book
The MIT Press
Cambridge, Massachusetts
London, England
© 2010 Massachusetts Institute of Technology

All rights reserved. No part of this book may be reproduced in any form by any
electronic or mechanical means (including photocopying, recording, or information
storage and retrieval) without permission in writing from the publisher.

MIT Press books may be purchased at special quantity discounts for business or sales
promotional use. For information, please email [email protected] or
write to Special Sales Department, The MIT Press, 55 Hayward Street, Cambridge,
MA 02142.

This book was set in Stone Sans and Stone Serif by Toppan Best-set Premidia Limited.
Printed and bound in the United States of America.

“Omissions and Causalism” by Carolina Sartorio. Copyright 2009 Wiley-Blackwell.

Reproduced with Permission of Blackwell Publishing, Ltd.

“Intentional Omissions” by Randolphe Clarke. Copyright 2010 Wiley-Blackwell.

Reproduced with Permission of Blackwell Publishing, Ltd.

Library of Congress Cataloging-in-Publication Data

Causing human actions : new perspectives on the causal theory of action / edited
by Jesús H. Aguilar and Andrei A. Buckareff.
p. cm.
“A Bradford Book.”
Includes bibliographical references and index.
ISBN 978-0-262-01456-4 (hardcover : alk. paper)—ISBN 978-0-262-51476-7 (pbk. :
alk. paper)
1. Act (Philosophy). 2. Action theory. 3. Intentionality (Philosophy).
I. Aguilar, Jesús H. (Jesús Humberto), 1962–. II. Buckareff, Andrei A., 1971–.
B105.A35C38 2010
128′.4—dc22
2010003187

10 9 8 7 6 5 4 3 2 1
Contents

Preface vii

1 The Causal Theory of Action: Origins and Issues 1

Jesús H. Aguilar and Andrei A. Buckareff

2 Renewed Questions about the Causal Theory of Action 27

Michael S. Moore

3 The Standard Story of Action: An Exchange (1) 45

Michael Smith

4 The Standard Story of Action: An Exchange (2) 57

Jennifer Hornsby

5 Skepticism about Natural Agency and the Causal Theory of

Action 69
John Bishop

6 Agential Systems, Causal Deviance, and Reliability 85

Jesús H. Aguilar

7 What Are You Causing in Acting? 101

Rowland Stout

8 Omissions and Causalism 115

Carolina Sartorio

9 Intentional Omissions 135

Randolph Clarke

10 Comments on Clarke’s “Intentional Omissions” 157

Carolina Sartorio

11 Reply to Sartorio 161

Randolph Clarke
vi Contents

12 Causal and Deliberative Strength of Reasons for Action: The Case of

Con-Reasons 167
David-Hillel Ruben

13 Teleological Explanations of Actions: Anticausalism versus

Causalism 183
Alfred R. Mele

14 Teleology and Causal Understanding in Children’s Theory of

Mind 199
Josef Perner and Johannes Roessler

15 Action Theory Meets Embodied Cognition 229

Fred Adams

16 Intentions as Complex Dynamical Attractors 253

Alicia Juarrero

17 The Causal Theory of Action and the Still Puzzling Knobe

Effect 277
Thomas Nadelhoffer

References 297
Contributors 323
Index 325
Preface

This volume brings together essays by some of the leading figures working
in action theory today. What unifies all of the essays is that they either
directly engage in debates over some aspect of the causal theory of action
(CTA) or they indirectly engage with the CTA by focusing on issues that
have significant consequences for the shape of a working CTA or the ten-
ability of any version of the CTA. Some of the authors defend this theory,
while others criticize it. What they all agree on is that the CTA occupies a
central place in the philosophy of action and philosophy of mind as the
“standard story of action.” Two of the essays in this volume have appeared
elsewhere recently. (Chapters 8 and 9 by Carolina Sartorio and Randolph
Clarke, respectively, previously appeared in Noûs. They appear with the
permission of Wiley-Blackwell, and have been lightly edited for consis-
tency.) The remaining essays appear in this volume for the first time.
Editing this volume, though not an easy task, has been a labor of love
for us. We are convinced that foundational issues in the philosophy of
action, such as the issues explored in this volume, deserve greater atten-
tion. It is our hope that the publication of this collection of essays will
serve to elevate the prominence of the debates the essays range over in
future research on human action and agency. This volume, then, is in part
an effort to promote exploration of foundational issues in action theory
and especially to encourage further work on the CTA by defenders and
critics alike.
Work on this volume would have been more difficult if not impossible
without the support of a number of people and institutions. First, Philip
Laughlin, Marc Lowenthal, and Thomas Stone from MIT Press deserve a
special debt of gratitude for supporting this project. Also from MIT Press,
we would like to thank Judy Feldmann for her fantastic editorial work.
Thanks to Wiley-Blackwell for giving us the permission to publish the
essays by Carolina Sartorio and Randolph Clarke which originally appeared
viii Preface

in Noûs, and to Ernest Sosa for providing some much-needed help with
acquiring the permission to put these essays in our book. Second, we would
like to thank the authors who contributed to this volume. This volume
would not exist were it not for their efforts. Third, we would like to thank
Joshua Knobe for his help in the editing process by reviewing the essays
by Thomas Nadelhoffer, Josef Perner, and Johannes Roessler. His expertise
in experimental philosophy and psychology far outstrips ours. His philo-
sophical acumen with respect to all things action theoretic made him an
obvious person to go to for help in reviewing these essays. Fourth, some
of the work for this volume was carried out while Andrei Buckareff was a
participant in the 2009 National Endowment for the Humanities Seminar
on Metaphysics and Mind led by John Heil. Andrei wishes to thank the
NEH for the financial support and John Heil for creating a seminar envi-
ronment that afforded him the opportunity to complete some of the work
on this and other projects. Fifth, thanks are due to the institutions we work
at, Marist College and Rochester Institute of Technology, for their support
of our work on this and other research projects. Andrei is especially thank-
ful to Martin Shaffer, Dean of the School of Liberal Arts at Marist, and
Thomas Wermuth, Vice President for Academic Affairs at Marist, for the
course releases that gave him extra time to work on his research, including
editing this volume. Jesús was awarded the Paul A. and Francena L. Miller
Faculty Fellowship from the Rochester Institute of Technology to support
part of the work involved in this volume, also in the form of course
releases, something for which he is very grateful. Finally, extra special
thanks are due to our families and friends for their tolerance and their
support as we worked on this project. Andrei would especially like to thank
his spouse, Lara Kasper-Buckareff, for her encouragement and patience
with him, especially in the final weeks of working on this project. Likewise,
Jesús is full of gratitude to Amy Wolf for her constant support during his
work on this volume.

Jesús H. Aguilar
Rochester, New York

Andrei A. Buckareff
Poughkeepsie, New York
1 The Causal Theory of Action: Origins and Issues

Jesús H. Aguilar and Andrei A. Buckareff

Philosophy of action is often construed either broadly as including all of

the problems in philosophy dealing with human action and agency or
more narrowly as concerned with merely the cluster of issues that deal
directly with the nature of intentional action and the explanation of
action. However we characterize the philosophy of action, one theory has
recently enjoyed the title of “the standard story” of human action and
agency in the literature, namely, the causal theory of action (CTA).1
Strictly speaking, it is misleading to think of the CTA as a single theory
of action. A better way to think about the CTA is in terms of a set of theo-
ries that bear a family resemblance by accepting the following schema
about what makes something an action and what explains an action:

(CTA) Any behavioral event A of an agent S is an action if and only if S’s

A-ing is caused in the right way and causally explained by some appropri-
ate nonactional mental item(s) that mediate or constitute S’s reasons for
A-ing.2

If we focus on the ontological implications of this schema we can iden-

tify the source of the most significant differences among the several ver-
sions of the CTA, namely, differences over what it means to be “caused in
the right way” and over what the appropriate nonactional mental item(s)
is/are that should cause S’s A-ing in order for S’s A-ing to be an action.
Similarly, if we focus on the epistemological implications of this schema,
we can identify a significant difference between the CTA and other com-
peting theories of action regarding the specific role of reasons in the
explanation of action. In this way, we can use the CTA schema as a point
of reference to identify the main features of the CTA, including both its
ontological commitments and epistemological commitments, trace back
its historical origins, and recognize the key areas of contention and devel-
opment associated with this general view on action.
2 J. H. Aguilar and A. A. Buckareff

In this introduction we offer a brief historical examination of three key

stages in the development of the CTA, namely, the ancient classical period
represented by the work of Aristotle, the early modern period represented
by the work of Thomas Hobbes, and the contemporary period represented
by the work of Donald Davidson. This is to be followed by a general pre-
sentation of the main areas of debate related to contemporary versions of
the CTA and a concise presentation of the contents of the chapters in this
volume.

1 The Roots of the Contemporary CTA: Aristotle, Hobbes, and

Davidson

All theories of action aim at providing answers to a set of foundational

questions about human action and agency. These foundational questions
have been articulated in the form of whether we can distinguish inten-
tional action (such as winking) from mere behavior (such as blinking), or,
as Wittgenstein famously put it, whether there is anything left over if one
subtracts the fact of an arm going up from the fact of raising one’s arm
(Wittgenstein 1972, §621). Furthermore, the conclusions we reach in theo-
rizing about the nature of action and agency will provide us with extra
resources to address other nearby philosophical questions such as the
problem of free agency, the mind–body problem, the problem of akrasia,
and other related problems in metaphysics, moral philosophy, philosophy
of law, and philosophy of mind, among other philosophical subdisciplines.
Ideally, our results in action theory should also provide researchers in the
behavioral and brain sciences with a broad conceptual framework for their
research into human behavior.
But although the motivations for doing action theory and the benefits
of working out one’s ontological commitments about the nature of action
are many, it has historically been a rather narrow set of concerns that have
chiefly motivated philosophers to think about action. It should come as
no surprise that at many points in the history of philosophy the primary
motivation to theorize about the springs of action was a desire to under-
stand the nature of moral agency. This is Alan Donagan’s compelling way
of formulating this crucial relationship between moral philosophy and
theorizing about human action:

Since ought implies can, writings about morality presuppose much about human
action. Yet although conclusions about action can defensibly be drawn from estab-
lished moral theory, no moral theory can become established unless its presupposi-
tions about action can be defended independently. (Donagan 1987, viii)3
The CTA: Origins and Issues 3

When we examine the history of the philosophy of action, we find that

Donagan is echoing sentiments about the importance of action theory
found among many of the leading theorists in the history of philosophy;
for the idea of intimately linking the possibility of morality to the possibil-
ity of action and agency has a long and prestigious genealogy, going back
at least to the work of Aristotle.

1.1 Aristotle
Although Aristotle was not the first major philosopher to write about
action—Plato wrote about action before him (see, e.g., the Phaedo
98c–99a)—to our knowledge he was the first one to think seriously about
the springs of action. Furthermore, the story he told about the role of the
mental in the production and explanation of action was a causal story
much along the lines of the above schema of the CTA. Such a proposal
may be seen as anachronistic. After all, the CTA as a proper theory of action
has only been identified under that title since the 1960s.4 Nonetheless, to
the extent that Aristotle had a theory of action, his theory is clearly a
progenitor of the CTA.5
Aristotle’s commitment to a proto-CTA theory of action can be pieced
together from portions of his De anima (DA), De motu animalum (Mot.),
and Nicomachean Ethics (NE). The origin of action lies in the agent accord-
ing to Aristotle (NE, Bk. III, ch. 1.20, 1111a, 23–24). Specifically, the springs
of action are what we now identify as pro-attitudes.6 For instance, he writes
that, “the proximate reason for movement is desire [ourexeos]”7 (Mot.
701a35; cf. DA 433a10–434a20). In NE, Aristotle distinguishes between
various types of desires. Of these types of desires two are of special interest
in understanding his account of the springs of action. They are the intrinsic
desires for what are deemed worthwhile ends (boulêsis)8 and the instrumen-
tal proximal action-triggering desires for the means to achieving the ends
(prohairesis).9 A simple statement of the etiology of action is found in
chapter 2 of Book VI of NE:

The origin of an action—the efficient cause, not the final cause—is prohairesis. The
origin of prohairesis is another desire [orexis] and goal-directed reason [eneka tinos].
(NE, 1139a31–34)10

Thus, on the Aristotelian story of intentional action, (i) an agent S

desires some end and believes that by A-ing she can satisfy her desire for
that end; (ii) this gives rise to a desire (prohairesis) to A; and (iii) the imme-
diate source (the efficient cause) of S’s subsequent A-ing is S’s desire (pro-
hairesis) to A.
4 J. H. Aguilar and A. A. Buckareff

The foregoing is a rough and not uncontroversial summary of the core

notions of Aristotle’s account of the etiology of action inherited by later
causal theories of action. Although this account is certainly a teleological
one, it is also fundamentally a causal story in which the desires of an agent
are doing causal work along the lines of the CTA schema.11 Also, whether
or not Aristotle is offering a complete theory of action or a proper part of
a theory of action is of secondary importance compared to the fact that
his account provided a causal framework for the next two millennia for
thinking about the etiology and explanation of human action. It is possible
to find the core elements of Aristotle’s causal story about human action
throughout the subsequent history of philosophical work on human
agency, particularly in Great Britain after the seventeenth century and
more recently among analytic philosophers and cognitive scientists after
the middle of the twentieth century.12

1.2 Hobbes
The early modern period is very significant in the history of the CTA.
Several accounts of action emerged that incorporated the new explanatory
framework coming from the Scientific Revolution to the study of human
beings. In particular, an analysis of the internal mental causes that lead to
an action was developed within such a framework. Some accounts placed
the mental causes of action squarely in the natural realm, while others
placed them completely outside the realm of nature.
Among the naturalistic-oriented thinkers of this period the most influ-
ential figure is Thomas Hobbes, who as much as Aristotle is responsible for
what would be christened “the causal theory of action” in the twentieth
century. Although later thinkers such as John Locke, David Hume, Jeremy
Bentham, John Austin, and J. S. Mill articulated accounts of human action
and agency that reflect central tenets present in contemporary versions of
the CTA, Hobbes’s work on action captures so much of the fundamental
tenets of the CTA that nowadays this theory is sometimes referred to as
the “Hobbesian picture of agency” (Pink 1997; Schroeter 2004).
Despite its fundamental contribution to the CTA, Hobbes’s theory of
action gets very little treatment in the secondary literature on his philoso-
phy, often mentioned as a side note in discussions of his psychology and
theory of free will.13 Although this is not the place to mend such an omis-
sion, given the centrality of Hobbes’s contribution to the development of
the CTA we want to at least introduce some of its key features, in particular
those that still serve to identify the CTA as a distinctive theory vis-à-vis its
present-day competitors.
The CTA: Origins and Issues 5

Hobbes’s theory of action is best seen in contrast with the theories of

action developed by the Scholastics. Their differences notwithstanding,
three of the great Scholastics, Thomas Aquinas, Duns Scotus, and Francisco
Suarez, had a fair amount to say about the springs of action. Central to
their theories of action was a role for an irreducible faculty of will. For
instance, Aquinas when analyzing the powers of the will has this to say:
“the will as an agent moves all the powers of the soul to their respective
acts, except the natural powers of the vegetative part, which are not subject
to our choice” (Summa theologica I, q. 82, a. 4 resp).14 Thus, according to
this view espoused by what Hobbes called “the Schools,” the will is an
irreducible rational faculty that is responsive to normative reasons for
actions whereby an agent exercises executive control.15
In stark contrast to such a Scholastic view, Hobbes reduces the will.
Agents are responsive to motivating reasons for action. The reasons to
which they respond, the ones that move them to act, are what we nowa-
days identify as pro-attitudes. In fact, for Hobbes the will just is the proxi-
mal action-triggering motivational state that causes an agent to act:

In deliberation, the last appetite or aversion immediately adhering to the action, or

to the omission thereof, is what we call the WILL, the act (not the faculty) of willing.
And beasts that have deliberation must necessarily also have will. The definition of
the will given commonly by the Schools, that it is a rational appetite, is not good.
For if it were, there could there be no voluntary act against reason. For a voluntary
act is that which proceedeth from the will and no other. But if instead of a rational
appetite, we shall say an appetite resulting from a precedent deliberation, then the
definition is the same that I have given here. Will therefore is the last appetite in
deliberating. (Hobbes 1651/1994, 33)16

For Hobbes, “appetite” is roughly synonymous with “desire” and these

seem like generic titles for pro-attitudes that motivate agents (Leviathan,
Part I.2). A succinct statement of what Hobbes means by “deliberation” is
found in his Of Liberty and Necessity, where deliberation amounts to the
“alternate succession of contrary appetites” (Hobbes 1654/1999, 37).17
The foregoing seems to suggest that action only follows from delibera-
tion. But Hobbes makes it clear that not all actions follow from delibera-
tion. Some actions require no deliberation. He writes that:

I conceive that when it comes into a man’s mind to do or not to do some certain
action, if he have no time to deliberate, the doing it or abstaining necessarily follows
the present thought he has of the good or evil consequence thereof to himself. . . .
Also when a man has time to deliberate but deliberates not, because never anything
appeared that could make him doubt of the consequence, the action follows his
opinion of the goodness or harm of it. These actions I call voluntary. . . . (Ibid.)
6 J. H. Aguilar and A. A. Buckareff

So an agent’s actions can be voluntary even in the absence of deliberation.

No deliberation and no special acts of will are necessary for action. This
makes the range of actions that count as voluntary much broader than
others before him would allow. In fact, for Hobbes, human action just is
one species of animal action, and as such it is a thoroughly natural
occurrence.
Furthermore, as if there is any question about what sort of role the will
plays in the etiology of action, Hobbes makes it clear that it plays a causal
role in the production of action: “of voluntary actions the will is the neces-
sary cause” (ibid., 38). Indeed, he seems to regard any event to be uncaused
to be conceptually impossible or at least problematic (ibid., 39).
So, in brief, the story of action we get from Hobbes is quite straightfor-
ward. An agent S does A voluntarily if her A-ing is what she willed to do.
And S’s A-ing occurs in virtue of her A-ing being caused by her will to A.
S’s will is simply her final desire that terminates deliberation or, in the
absence of deliberation, the strongest desire she has at the time she acts.
In this way, for Hobbes, the locus of executive control over action is an
agent’s desire that causes her to act as she does.

1.3 Davidson
By the middle of the twentieth century and after the anticausalist hiatus
motivated by the work of philosophers such as Ludwig Wittgenstein and
Gilbert Ryle, the CTA reached its full maturity and emerged as the recog-
nized “standard story of action” among contemporary action theorists. The
locus classicus for the presentation and defense of this contemporary
version of the CTA is found in Donald Davidson’s work on action and
agency, particularly, in his groundbreaking essay “Actions, Reasons, and
Causes” (Davidson 1963/1980). We will focus on this essay, given its cen-
trality in discussions of contemporary versions of the CTA.
Davidson makes two claims about reasons-explanations of action in
“Actions, Reasons, and Causes.” The first is a claim about what constitutes
an agent’s primary reason for action, and the second claim is about the
causal role of an agent’s primary reason for action. Regarding the first
claim, Davidson echoes both Aristotle and Hobbes in affording pro-atti-
tudes a central role in his account of practical reasons. His view is particu-
larly Aristotelian in rendering beliefs an important role as well:

C1. R is a primary reason why an agent performed the action A under the descrip-
tion d only if R consists of a pro attitude of the agent towards actions with a certain
property, and a belief that A, under the description d, has that property. (Davidson
1963/1980, 5)
The CTA: Origins and Issues 7

According to this view, a reason for action is a belief and pro-attitude pair,
and explanations in terms of reasons will mention one or both of these
items.
Regarding the explanatory role of reasons, Davidson argues that the
relationship between reasons and actions displays the same pattern we
discern in causal explanations. Unless the relationship is causal, we are at
a loss to distinguish those cases in which an agent has a reason for acting
that fails to explain why she acts as she does, from cases where the agent’s
reason for acting explains why she acts as she does. Davidson notes that if
we dispense with a reason’s causal role:

Something essential has certainly been left out, for a person can have a reason for
an action, and perform the action, and yet this reason not be the reason why he
did it. Central to the relation between a reason and an action it explains is the idea
that the agent performed the action because he had the reason. (Ibid., 9)

So, to borrow an example from Alfred Mele, Sebastian may have a pair of
reasons for mowing his lawn this afternoon. He wants to mow the lawn
when the grass is dry and he also wants his spouse, Fred, to see him
mowing the lawn when Fred gets home from work in order to impress
him. It turns out that Sebastian only acts for one of these reasons. Mele
asks, “In virtue of what is it true that he mowed his lawn for this reason
and not for the other, if not that this reason (or his having it), and not
the other, played a suitable causal role in his mowing the lawn?” (Mele
1997, 240).
This explanatory problem for the noncausalist has been christened
“Davidson’s Challenge” in the action-theoretic literature (Ginet 2002; Mele
2003), and it is a direct challenge to anyone who rejects causalism about
reasons-explanations. Mele notes that the challenge is to “provide an
account of the reasons for which we act that does not treat . . . [them] as
figuring in the causation of the relevant behavior” (Mele 2003, 39).
Thus, Davidson’s second claim about reasons for action and their role
in the explanation of action has been the main source of contention
between causalists and noncausalists about reasons-explanations:

C2. A primary reason for an action is its cause. (Davidson 1963/1980, 12)

The remainder of “Actions, Reasons, and Causes” is devoted to defending

C2 from possible attacks coming from noncausalists.
One may wonder where and how Davidson gets around to making any
further direct claims about the nature of action itself. After all, a commit-
ment to causalism about reasons-explanations is at most a necessary condi-
tion for a version of the CTA.
8 J. H. Aguilar and A. A. Buckareff

Detecting Davidson’s further commitment to a version of the CTA in

“Actions, Reasons, and Causes” requires that one dig around to find any-
thing close to a statement regarding the difference that causation by
reasons makes for the actional status of some event. It is not until the very
end of the essay that Davidson says something about the ontological sig-
nificance of his claims about reasons as causes of action.
In the last couple of pages of “Actions, Reasons, and Causes,” Davidson
discusses some worries expressed by A. I. Melden (1961). Melden’s worries
are about mental causation and causal theories of action. In fact, they
prefigure more recent criticisms of the CTA based on the idea that in this
theory’s picture of the production of an action, the agent is essentially
absent.18 This is what Melden has to say:

It is futile to attempt to explain conduct through the causal efficacy of desire—all

that can explain is further happenings, not actions performed by agents. The agent
confronting the causal nexus in which such happenings occur is a helpless victim
of all that occurs in and to him. There is no place in this picture of the proceedings
either for rational appetite or desires, or even for the conduct that was to have been
explained by reference to them. (Melden 1961, 128–129)

In response, Davidson notes that Melden’s view is contradictory. Specifi-

cally, Melden identifies some actions with bodily movements, allows those
bodily movements to be caused, but denies that the causes of these bodily
movements qua movements are causes of the bodily movements qua
actions (Davidson 1963/1980, 18–19). Davidson notes that Melden attempts
to remove actions from the realm of causality altogether, getting the odd
result that an action is identical with a bodily movement that is caused,
but the action is not itself caused (ibid., 19).
Nevertheless, something more metaphysically substantive is contained
in Davidson’s reply to Melden’s criticism than merely pointing out a fla-
grant inconsistency. As part of his rejection of Melden’s views, Davidson
rejects a false dilemma that is presupposed in such anticausalist views,
namely, that we must either be agent-causes of our actions or be mere
“patients” or “victims” with respect to our actions. A third alternative is
simply this: “Some causes have no agents. Among these agentless causes
are the states and changes of state in persons which, because they are
reasons as well as causes, constitute certain events free and intentional
actions” (ibid.). We are agents in virtue of parts of ourselves causing our
behavior, and those parts are our reasons. Correspondingly, our behavior
is actional in virtue of being caused by our reasons.
Two further central features of Davidson’s version of the CTA are present
in two later seminal essays on his theory of action, “Agency” (Davidson
The CTA: Origins and Issues 9

1971/1980) and “Intending” (Davidson 1978/1980). In “Agency” he adds

that someone counts as the agent of an action if what this person does
“can be described under an aspect that makes it intentional” (1971/1980,
46). In “Intending” he revises his previous view (C1) that reduced inten-
tions to belief and pro-attitude pairs, proposing instead that intentions are
irreducible executive states that figure in the etiology and control of inten-
tional action in addition to beliefs and pro-attitudes.19
Clearly, Davidson had more to say about the theory of action and its
explanation. But what is important for our purposes is that he, along with
the likes of Aristotle and Hobbes, laid the foundations for the articulation
of a CTA that has emerged as the “standard story of action” not only in
action theory, but in the philosophy of mind and moral philosophy. We
now turn to some key problems raised for the CTA and some proposed
solutions to these problems.

2 The Contemporary Debate over the CTA

From the foregoing, we hope it is evident that there are at least two desid-
erata a satisfactory theory of action should satisfy. The first is a metaphysi-
cal desideratum requiring that a theory of action should provide us with
a way to distinguish behavior that is actional from behavior that is not
actional. This desideratum includes the need for a theory of action to tell
us a story about the role of agents in controlling their actions. The second
desideratum is epistemological and requires that a theory of action should
provide us with the resources to explain the occurrence of an action.
Addressing the metaphysical desideratum, all versions of the CTA
propose that a behavior is an action only if it is caused in the right way by
some appropriate nonactional mental items. Correspondingly, an agent
exercises her causal powers through the occurrence of nonactional mental
events and states, that is, agent causation is reducible to causation by non-
actional mental events and states. So the story of agency on the CTA is a
story about some part(s) of an agent causing some intentional behavior.
With respect to the epistemological desideratum, all versions of the CTA
propose that an action A’s occurrence can be explained by the reasons the
agent has for A-ing, and the reasons explain an agent’s A-ing only if the
reasons played a causal role in the etiology of the agent’s A-ing. Opponents
of the CTA have questioned just how well the CTA satisfies the two desid-
erata. In this section of the introduction we present the main metaphysical
and epistemological challenges for the CTA arising from the two
desiderata.
10 J. H. Aguilar and A. A. Buckareff

2.1 Basic Causal Deviance

Consider a much-cited example of basic causal deviance, namely, Donald
Davidson’s nervous climber example:

A climber might want to rid himself of the weight and danger of holding another
man on a rope, and he might know that by loosening his hold on the rope he could
rid himself of the weight and danger. This belief and want might so unnerve him
as to cause him to loosen his hold. (Davidson 1973/1980, 79)

The example is supposed to show that it is possible to satisfy the causal

conditions for an action in accordance with the CTA without the ensuing
event counting as an action. The problem of basic causal deviance is
thought by some to be the Achilles’ heel of the CTA. Whether it proves
fatal or not, what the problem shows is that any version of the CTA that
merely emphasizes a ballistic role for the relevant mental causes of action
is inadequate. That is to say, whether the proximal cause of action is a
belief-desire pair, an intention, or some event associated with either, the
causal role of the relevant mental item(s) cannot merely trigger behavior.
Exactly what must be added in order for an action to occur has been one
of the central issues in the debates over the CTA.
The most prominent strategies offered by CTA defenders to date have
afforded intentions a central role in their responses to the problem of basic
deviance. Most have emphasized the following. Either the agent’s action-
triggering intention must play a sustaining and guiding causal role through
to the completion of the action, or an agent must acquire a separate action-
guiding intention once she begins to execute her action-triggering inten-
tion that plays a sustaining and guiding role.20 How it is that an agent
guides her action via her intention that plays the relevant executive causal
role has been difficult to articulate. Some prominent defenses of the CTA
have argued that behavior that is actional must be responsive to the con-
tents of a set of the agent’s reasons that both cause and explain it.21 This
condition has been called the sensitivity condition in the literature. While
many agree that meeting the sensitivity condition is necessary for action,
it is not sufficient. To identify the sources of such skepticism, consider the
following type of example offered by Christopher Peacocke (1979b, 87).22
Ichiro has had a device planted in his brain that allows a neuroscientist
to detect whenever Ichiro has acquired an action-triggering intention to
perform an overt action. Unbeknownst to Ichiro, two things are true of
him. First, his motor areas and efferent pathways are unresponsive to his
proximal intentions to perform bodily movements. Second, the neurosci-
entist, upon detecting the presence of an action-triggering intention, stim-
The CTA: Origins and Issues 11

ulates Ichiro’s motor areas and efferent pathways, thereby satisfying

Ichiro’s intentions to execute bodily movements. When Ichiro’s body
moves, his movement is at least in part a causal response to the content
of his reasons for action.
If Ichiro succeeds in satisfying one of his intentions in this way, at least
two things prevent the outcome to count as his intentional action. First,
the causal intermediary is not the normal intermediary between the mental
cause of an action and its occurrence. Specifically, the causal path from
the action-triggering intention to its execution involves the activities of a
causal intermediary that is another agent. That is, Ichiro’s bodily move-
ment is the result of the actions of an intermediate agent who responds
to the presence of Ichiro’s intention, but the intermediate agent does so
for her own reasons. Second, the causal pathway is less than reliable. Ichiro
has enjoyed good luck since the neuroscientist happens to agree to help
him in this way. This is not like having one’s normal bodily pathways from
an intention to its execution “cooperate” (or even having a prosthetic
device serve as a causal medium). For as long as the neuroscientist does
not object to the content of Ichiro’s action-triggering intention, she will
stimulate his motor areas and efferent pathways accordingly. However, if
her reasons for action and Ichiro’s reasons for action are in conflict, the
neuroscientist can exercise executive control over Ichiro’s bodily move-
ments. So Ichiro is subject to the whims of the neuroscientist intervener.
Peacocke-type examples are a serious challenge to the CTA because they
provide us with cases in which bodily movements are explained by the
content of the mental events that cause them without such movements
counting as intentional actions. In Ichiro’s case, whenever the neuroscien-
tist cooperates, he will produce bodily movements that match up with the
contents of Ichiro’s reasons for performing them that explain why that
particular type of movement occurred, rather than another. But the fact
that Ichiro’s mental state content plays this causal and explanatory role is
not enough to show that Ichiro is acting.
Where the difficult work needs to be done is on being clear about the
types of mechanisms and states that must cause a movement in order for
it to be truly actional. This means going beyond merely assigning respon-
siveness to the content of reasons a role in the etiology of a movement.
What is needed is the shifting of attention away from what causes a move-
ment and instead concentrating attention on how it is caused.23
In the present volume, the problem of basic causal deviance is taken up
in the contributions from Jesús H. Aguilar, John Bishop, Michael S. Moore,
and Michael Smith. The differences in their proposals notwithstanding,
12 J. H. Aguilar and A. A. Buckareff

they agree that basic causal deviance poses a serious challenge to the CTA
and deserves the attention of proponents of the CTA.

2.2 The Problem of the Absent Agent

A second challenge to the CTA is related to the theory’s approach to
agency. If, as the CTA proposes, a behavior is actional in virtue of how it
is caused by some mental state or event, and no mental state or event
seems to be identical with an agent, then it is hard to see how such behav-
ior involves an agent who is in charge of producing and controlling it. In
fact, it appears that the agent has been erased from the picture that the
CTA offers about the production and control of an action, effectively
making the agent an absentee in such process. This is the problem of the
absent agent.
The problem of the absent agent comes in two flavors, depending on
the alleged agential deficit involved in the CTA. A first version of this
problem charges the CTA with making the agent a passive participant in
the production and control of an action. The agent is still part of the
picture, albeit in a very minimal way, a sort of receptacle where the relevant
mental events and states that lead to an action take place.24 A good
example of this version of the problem of the absent agent is found in
Melden’s previously mentioned criticism of the CTA. Recall that, according
to Melden, in the picture offered by this theory, “the agent confronting
the causal nexus in which such happenings occur is a helpless victim of
all that occurs in and to him” (Melden 1961, 128–29). The agent then
becomes more like a patient or victim with respect to the flow of causally
related events and states that ensue in an action.
The foregoing version of the problem of the absent agent has led some
authors to propose an alternative view of action and causation, which
turns out to be one of the main competitors of the CTA. This alternative
view of action that gives the agent herself the capacity to directly cause
the events corresponding to her actions is known as the “agent causal
theory of action” (ACTA)25 and is associated with the work of Roderick
Chisholm (1966) and Richard Taylor (1966), and, more recently, Maria
Alvarez and John Hyman (1998).26
The second version of the problem of the absent agent is more radical
in its stance, charging CTA not only with turning the agent into a passive
participant in the production of an action but of actually eliminating the
agent altogether from the picture. As critics of the CTA like Jennifer
Hornsby have put it, the upshot of the CTA “in any of its versions . . . is
not a story of agency at all” (Hornsby 2004, 2). If this criticism is correct,
The CTA: Origins and Issues 13

the price paid by the reductive strategy embraced by the CTA is literally
the abandoning of anything resembling an account of agency understood
as the capacity that subjects posses to give rise to and control an action.27
Defenders of the CTA have offered different ways of addressing the
problem of the absent agent. The two most prominent strategies involve
either embellishing the standard story given by the CTA or exploiting the
available resources of our best current versions of the CTA without making
any substantive additions. These two strategies can be labeled “embellish-
ment” and “revisionist” strategies, respectively.
Typically, versions of the embellishment strategy involve hierarchical
or “real-self” theories of agency inspired by the work of Harry Frankfurt.
Although Frankfurt is no defender of the CTA,28 his influence on the embel-
lishment strategy can be traced back to his work on autonomous agency.
The initial statement of this view of autonomy by Frankfurt (1971/1988)
was in his paper “Freedom of the Will and the Concept of a Person” and
was further developed in later papers of his.29 On this view, an agent acts
autonomously when he acts from a pro-attitude that the agent endorses
or identifies with. As Frankfurt notes, the sort of endorsement of or iden-
tification with a motivational state X that is indicative of autonomous
agency requires that the agent wants X to be the motivational state that
“moves him effectively to act” (Frankfurt 1971/1988, 10).
Michael Bratman (2000, 2001) and J. David Velleman (1992/2000) are
two prominent CTA advocates of the embellishment strategy grounded on
versions of hierarchical theories of agency. In the case of Bratman, the
emphasis is on the inclusion of self-governing policies “that say which
desires to treat in one’s effective deliberation as associated with justifying
reasons for action” (Bratman 2001, 309). For Velleman, the corresponding
emphasis consists in explaining the main functional role played by the
agent in the production of an action, namely, “that of a single party pre-
pared to reflect on, and take sides with, potential determinants of behavior
at any level in the hierarchy of attitudes” (Velleman 1992/2000, 139).
According to Velleman, this functional role is found in a propositional
attitude consisting in the desire to act in accordance with reasons.30
The differences notwithstanding, the embellishment strategies end up
suggesting that what is missing in the CTA’s reductionist approach to
agency is some species of reflective endorsement by an agent of the mental
causes of her action. That is, what the embellishment strategy offers is a
version of the CTA that is more complicated in terms of the number of
items involved in the production of actions expressing true agency. But
the ontological upshot remains essentially the same. Surely the CTA’s
14 J. H. Aguilar and A. A. Buckareff

picture of agency has become more complex; but complexity does not
imply ontological novelty. No new powers or unique phenomena that
affect the executive control of agents are conferred on them with this
extended metaphysical framework. Once we strip away the complex frame-
work added to embellish this theory in the interest of giving an account
of full-blooded agency, one is left with a “causal structure involving events,
states, and processes of a sort we might appeal to within a broadly natu-
ralistic psychology” (Bratman 2001, 312).31
It is important to note that the CTA’s commitment to a metaphysically
minimalist view of the items involved in the production of an action
contrasts starkly not only with the mentioned effort by defenders of the
ACTA to introduce a different type of causal relationship involving agents
who directly cause events, but also with the CTA’s other serious competi-
tor, namely, the volitionist theory of action (VTA). According to the VTA,
the way to correctly characterize action and fully incorporate the agent in
the picture requires the introduction of a unique type of irreducible mental
event that is essentially actional, typically a willing or a trying.32 Because
this type of actional mental event engages the agent directly, it may seem
that, if anything, the VTA is ontologically more economical than the
CTA particularly in the light of the modifications suggested by the embel-
lishment strategy. And yet, the positing of an event that is intrinsically
actional, unanalyzable, and whose single distinguishing feature is the
“actish” phenomenological feeling that comes with it33 introduces a
different and unique type of event that is supposed to do the hard
agential work. From the perspective of the CTA, the introduction by the
VTA of such a unique event amounts to an unnecessary ontological
enlargement, which by stipulation is identified as the place where agency
takes place.
Nevertheless, some defenders of the CTA recognize that at least with
respect to the commonsensical conceptual framework associated with
agency, the ACTA and the VTA appear to be responding to basic assump-
tions involved in the adoption of reactive attitudes toward agents, particu-
larly when agents are seen from a moral perspective. For unless the agent
herself is an active participant in the production and control of the actions
for which she is going to be held morally responsible and which serve as
the legitimate objects of reactive attitudes, the very possibility of these
fundamental practices is undermined. From this perspective, the problem
of the absent agent is the challenge of making sense of an agent who
occupies the central role in the story of morality and responsibility by
using exclusively the sparse ontology assumed by the CTA. This type of
The CTA: Origins and Issues 15

challenge motivates a second CTA strategy to deal with the problem of the
absent agent grounded on the recognition that the commonsensical con-
ceptual framework associated with agency needs to be accommodated
inside a reductive naturalist metaphysics of agency.34 The reason this
second strategy is revisionist is that it continues to exploit the available
metaphysical resources of the CTA while promising to deliver the goods
requested by the commonsensical conceptual framework.
Donald Davison inaugurated the revisionist effort within the CTA by
accepting the conceptual irreducibility of agency while maintaining his
well-known metaphysical commitment to monism, in this case taking the
form of the basic ontological items proposed by this theory of action. A
more recent and fully developed effort along these revisionist lines is found
in the work of John Bishop, particularly in his Natural Agency (1989). In
this book, Bishop tries to reconcile what he considers to be a fundamental
tension between two competing perspectives:

We seem committed to two perspectives on human behaviour—the ethical and the

natural—yet the two can be put in tension with one another—so seriously in
tension, in fact, as to convince some philosophers either that the acting person is
not part of the natural order open to scientific inquiry or that morally responsible
natural agency is an illusion. (Bishop 1989, 15)

Arguably, the conceptual framework presupposed by the ethical perspec-

tive arises from the conceptual framework provided by what often and
somewhat loosely is identified as common sense. However, Bishop is
careful in making sure that, whether or not a commonsensical conceptual
framework generates a tension with a reductive strategy espoused by a
philosophical theory such as the CTA,35 the roots of the tension he is react-
ing to lie in “a genuine philosophical puzzlement” (ibid., 11) related to
the effort to understand “how morally responsible agents can be part of
the natural order as our natural scientific worldview conceives of it” (ibid.,
15).
According to Bishop, what makes the challenge of reconciling the
ethical perspective with the natural perspective particularly acute is that
actions from the ethical perspective involve essentially a type of agent-
causation that reveal the agent’s exercise of control. Bishop’s proposal is
based on the idea that the tension will dissipate if agent-causalism is not
fully warranted, and this is what he tries to show. His strategy is to insist
that, from a purely metaphysical point of view, “the notion of action
that we need for our ethical point of view does not involve ontological
commitment to agent-causal relations of a kind foreign to our naturalist
16 J. H. Aguilar and A. A. Buckareff

ontology” (ibid., 42). Nevertheless, in denying that actions are constituted

by agent-causal relations, Bishop does not make claims about what it means
to perform an action. Although agent-causation may in fact be part of our
primitive concepts about agency, a successful CTA shows that agent-
causation is not ontologically basic since precisely what the CTA offers are
conditions that are “both necessary and sufficient for a basic intentional
action to occur” (ibid., 178). That is, providing necessary and sufficient
conditions for the occurrence of a basic intentional action is not the same
thing as providing necessary and sufficient conditions for the concept of a
basic intentional action. If this revisionary response to the problem of the
absent agent is correct, drawing out ontological conclusions that are alleg-
edly problematic for CTA based on the concepts of “action” and “agency”
is controversial at best and confused at worst.36

3 Reasons and the Explanation of Action

To its defenders, one of the main theoretical virtues of the CTA consists
in the way it captures the motivating contribution of reasons by identify-
ing them as what cause an agent to act. An agent may have lots of reasons
for acting, but the reason why she does it is the one that causes her to do
it. This simple idea is at the center of the most serious epistemological
debate between causalists and noncausalists about the explanatory role of
reasons concerning action. Perhaps the best way to make sense of this
debate is in terms of the above-mentioned Davidson’s Challenge, that is,
the challenge to provide an account of reasons for which someone acts
that does not treat such reasons among the causes of the action. Conse-
quently, anticausalists often see it as part of their theoretical job to show
that Davidson’s Challenge can be effectively answered, thereby seriously
undermining theories like the CTA.37
One philosopher whose work stands out as a representative of the type
of attack on the CTA grounded on an effort to deal with Davidson’s Chal-
lenge is George Wilson.38 The following schema articulates Wilson’s non-
causalist teleological view (Wilson 1989, 172):

An agent S performs some action A with the conscious purpose of satisfying a desire
to achieve some goal E, if and only if, S A’s because he desired to achieve E and S’s
A-ing occurred because S thereby intended to satisfy D.39

The notable feature of Wilson’s account is that it allegedly captures the

teleological role of reasons to explain the performance of an action without
relying on their causal roles. Regarding Davidson’s Challenge, Wilson ends
The CTA: Origins and Issues 17

up claiming that it actually “offers no particular support for a causalist

account of explanations of intentional action” (ibid., 174). Although we
cannot consider here the details of Wilson’s positive case for his teleologi-
cal alternative, it is useful to consider how a causalist may respond to
Wilson by appealing to the resources of the CTA. This response is due to
Alfred Mele. (In this volume, Mele continues his effort to show the limita-
tions of teleological accounts of reasons-explanations in contrast to the
virtues of causalist ones.40)
Mele (2003) offers the following counterexample to Wilson’s teleologi-
cal approach. Norm has left his hat, tools, and bricks on the roof. He wants
to fetch all three but is undecided about which to get when he commences
climbing. Halfway up he decides to get the bricks and forms the intention
to get the bricks in order to (thereby) satisfy his desire to get the bricks.
The question at this point that is central to the debate over reasons-
explanations is this: In virtue of what does Norm form the intention to
get the bricks? Wilson’s answer is that Norm intended thereby to satisfy
his desire. Moreover, the role of Norm’s desire is irreducibly teleological
and, hence, noncausal.
Now imagine that Norm has the intention to get his stuff by climbing
the ladder, and (as Wilson would have it) neither the intention nor a desire
plays a causal role in the production of Norm’s movements up the ladder.
At the moment Norm decides to get the bricks, random Q signals from
Mars cause his movements. But it seems to Norm that he is moving himself
up the ladder. According to Mele:

In that event, it is false that Norm climbed the rest of the way up the ladder. Although
his body continued to move up the ladder as it had been, and although he intended
of his movements that they “promote the satisfaction of [his] desire,” Norm was no
longer the agent of the movements. (Mele 2003, 46)

The problem for Wilson’s teleological account is that Norm meets Wilson’s
conditions for his reasons to explain his movements. Mele argues that
“[The thought experiment] lays bare a point that might otherwise
be hidden. Our bodily motions might coincide with our desires or inten-
tions, and even result in our getting what we want or what we intend to
get . . . , without those motions being explained by the desires or inten-
tions” (ibid., 46). Thus, it appears that one can meet all of the conditions
Wilson advances for acting for a particular reason and yet the reasons can
fail to explain one’s action (ibid., 45–51). The upshot, if Mele is correct,
is that a noncausal teleological account such as Wilson’s cannot meet
Davidson’s Challenge.
18 J. H. Aguilar and A. A. Buckareff

However, the challenges to the CTA with respect to reasons-

explanations do not come only from the direction of noncausal teleologi-
cal theories of reasons-explanations.41 A more recent challenge, to both
causalism and teleological noncausalism, comes from philosophers who
argue that reasons and explanations for action are not necessarily mental-
istic. The idea is that the reasons that explain and justify an action are not
always the internal mental states of an agent. Rather, reasons for action
are often “items that objectively favour courses of action” (Mele 2007, 91).
A useful way to make sense of this challenge is by considering it as reject-
ing psychologism, understood as the view that considers the following
thesis true:

For every explanation E of an action A of an agent S, E is a reasons-

explanation of A only if some hidden, inner mental states of S’s constitute
S’s reasons for action and are included in E.

So, for instance, an opponent of psychologism, Jonathan Dancy, asserts

that “even the most cursory glance at the sorts of reason we actually give,
in explaining our own actions or those of others, reveals that while some
certainly seem to be characterized in terms of a psychological state of the
agent, others equally certainly do not” (Dancy 2000, 15).42
A couple of quick comments are in order regarding the rejection of
psychologism as a challenge to the CTA. First, to the extent that antipsy-
chologism alleges to describe the folk practice of explaining actions by
reasons, it is not obvious that it gets the folk practice right. In fact, the
evidence from current work on attribution theory in social psychology
suggests that the folk practice is psychologistic.43 Second, as Mele (2003,
2007) and Smith (2003b) have separately noted, as a philosophical theory
of reasons-explanations, antipsychologism fails to account for cases when
agents do not assess whether what they are doing or the ends they are
pursuing are objectively desirable or favored.
Of course, more needs to be said about why antipsychologism should
not be preferred over psychologism, particularly causalist versions of psy-
chologism. Although this task cannot be carried out here, defenders of the
CTA will need to do more in the future to address the recent challenges
coming from this direction. Like noncausal teleological approaches, anti-
psychologistic approaches to reasons-explanations are not going away.44
We cannot pretend to have provided an exhaustive survey of the present
issues surrounding the CTA in this section. But we hope the foregoing gives
those who find themselves in new territory a sense of both the history of
the CTA and the state of play about core issues related to the CTA. In the
The CTA: Origins and Issues 19

remainder of this introductory chapter we will briefly summarize the con-

tents of the other chapters in this volume.

4 A Summary of the Chapters in This Volume

Some of the authors in this volume are defenders of the CTA and others
are critics. Some inform their proposals by reaching into the past and
drawing from historical sources while others concentrate solely on the
present state of the art and its future possibilities. Lastly, some authors rely
mainly on the tools of standard analytic philosophy of action to support
their claims while others base their conclusions on the latest empirical
research on human action. What all the authors of this volume have in
common is the recognition that given the centrality of the CTA, this is a
theory that for better or for worse one cannot ignore.
The first eleven chapters of the volume focus on metaphysical issues
about the CTA. Chapters 12 and 13 address the debate over reasons-
explanations of action. The final four chapters focus on assorted new
directions for thinking about CTA.
Michael S. Moore’s essay, “Renewed Questions about the Causal Theory
of Action,” starts off the portion of essays that focus on metaphysical
issues. Moore not only responds to some challenges to his own formulation
of the CTA, but addresses some more general metaphysical challenges for
the CTA, including, among other problems, the absent agent problem, how
the CTA can account for mental actions and omissions, and the nature
of the causal relation between mental items and events that count as
actions. Moore goes straight to the heart of many of the debates over the
CTA, defending the ontological credentials of the CTA while questioning
the tenability of the ontological commitments of some alternatives.
Michael Smith’s essay, “The Standard Story of Action: An Exchange (1),”
is the first part of an exchange between Smith and Jennifer Hornsby.
Hornsby’s essay, “The Standard Story of Action: An Exchange (2),” consti-
tutes the second half of the exchange. Smith defends the CTA. His paper
is a response to some of the arguments leveled against the CTA by Hornsby
(2004), particularly her arguments related to the role of agents in the etiol-
ogy of action on the CTA. Hornsby responds to Smith and argues that no
version of the CTA should be accepted.
John Bishop argues that action theorists must keep their theorizing
about action and agency in contact with the motivations for such work.
Bishop discusses the motivations for his own work on action, particularly
his desire to reconcile the ethical perspective of ourselves as moral agents
20 J. H. Aguilar and A. A. Buckareff

with the natural perspective according to which intentional action can be

the sort of thing open to scientific investigation. Bishop contrasts his own
work with that of Donald Davidson by focusing his attention on the
problem of causal deviance and addresses recent critical work of David-
Hillel Ruben (2003) on the Davidsonian formulation of the CTA.
The problem of basic causal deviance is the focus of the essay by
Jesús H. Aguilar. Aguilar discusses sensitivity strategies, focusing on John
Bishop’s (1989) proposal. Aguilar provides a sympathetic critique of
Bishop’s work. He finally recommends a way of fixing Bishop’s proposal
that brings reliability considerations into a working version of the CTA.
Rowland Stout focuses on the causal relation itself that obtains between
an agent and her action. Specifically, he takes issue with the view in the
ontology of action that takes actions to consist in the causing of events
such as bodily movements. It is not the fact that the events are of a certain
type to which Stout objects. Rather, he objects to theories of action that
take the causing of an event to be the source of the problem. Stout explores
the metaphysics of causation and gestures toward an account of the etiol-
ogy of action that affords agents a more intimate causal connection with
their actions.
The essays by Carolina Sartorio and Randolph Clarke focus on the chal-
lenge of omissions for the CTA. Sartorio argues that the CTA qua theory
of intentional behavior lacks the resources to account for intentional omis-
sions since omissions are not events and cannot be caused in the way
intentional behavior is described as being caused by the CTA. Clarke dis-
tinguishes different ways of understanding what occurs in putative cases
of intentionally omitting. He argues that the threat to the CTA posed by
omissions is a chimera. Sartorio and Clarke directly respond to one another
in follow-up essays, extending the dialectic further.
Whether reasons against acting contrary to how one typically acts can
be causes is the subject of David-Hillel Ruben’s contribution to the present
volume. Ruben treats “con-reasons,” as he refers to them, as an insuperable
problem for causalism about reasons. On the other hand, Alfred R. Mele’s
essay defends causalism. Mele critiques Scott Sehon’s (2005) defense of
noncausal teleological accounts of reasons-explanations of action. He
focuses on Sehon’s response to Mele’s (2003) objection to George Wilson’s
(1989) teleological proposal. Mele additionally assesses Sehon’s reply to
Davidson’s challenge.
In their essay, Josef Perner and Johannes Roessler examine the develop-
ment of the conception of action and its explanation in children. They
consider the implications for how adults think about action and call for a
The CTA: Origins and Issues 21

more expansive understanding of intentional action and its explanation

that incorporates teleological elements and a wider range of attitudes to
account for the perspective of agents toward their actions.
Some regard embodied cognition as a challenge for views of action like
the CTA. In his essay, Fred Adams addresses the objections to the features
of such theories of action that are the alleged root of the problem. He
argues that versions of the CTA are compatible with the demands placed
on them by some accounts of embodied cognition.
Alicia Juarrero’s essay advances aspects of her project in Dynamics in
Action (1999). She makes the case that complex dynamical systems have
emergent properties that enhance the autonomy of a system (e.g., an
agent), making room for the sort of control necessary for moral agency.
Finally, Thomas Nadelhoffer takes up the much-discussed Knobe effect
and its implications for the CTA in his contribution. He argues both for
the importance of experimental philosophy of action and that the current
data pose difficulties for the CTA.

Notes

1. The locus classicus for a contemporary presentation and defense of the CTA is
Davidson 1963/1980. More recent attempts at defending the CTA have offered a
more fine-tuned account of the role of mental items in the causation of intentional
behavior. For more recent book-length defenses of versions of the CTA, see Brand
1984; Bishop 1989; Audi 1993; Enç 2003; and Mele 1992, 2003. Prominent alterna-
tives to the CTA include various noncausal theories of action (including versions of
volitionism) and agent-causal theories of action. For noncausal theories of action,
see Ginet 1990; Goetz 1988; and McCann 1998. For defenses of agent-causal theories
of action, see Chisholm 1966; Taylor 1966; and Alvarez and Hyman 1998.

2. Although we believe this schema captures the essential features of the CTA, this
is by no means a settled view, particularly with respect to the additional epistemic
clause to what otherwise is a metaphysical view on actions. In fact, one of the
contributors to this volume has argued precisely against such a formulation, propos-
ing instead that the components of the traditional schema be divided into two
different theories: a causal theory of action (CTA) that captures the metaphysical
part of the schema and a causal theory of action explanation (CTAE) that captures
the epistemological part of the schema. See Ruben 2003, 90.

3. A notable example of this philosophical inclination is found in the work of John

Bishop, one of the contributors to this volume. In his book Natural Agency (1989)
and in chapter 5 of this volume, Bishop emphasizes the crucial role that moral
considerations have in our quest for the nature of action and agency.
22 J. H. Aguilar and A. A. Buckareff

4. John Ladd’s paper, “The Ethical Dimensions of the Concept of Action” (1965),
may be the first place where the term “the causal theory of action” is used in the
action-theoretic literature. But Ladd does not use “the causal theory of action” just
to refer to what we would now identify as the CTA. Rather, any theory of action
that “attempts to analyze ‘action’ in terms of the category of cause and effect” counts
as a causal theory of action for Ladd (1965, 640). This differs from current usage
since “causal theory of action” is used to refer to an event-causal theory of action.
Ladd’s usage allows for agent-causal theories of action to be causal theories of action.
In the same year, Daniel Bennett identifies “the causal thesis,” which is just what
we would now call the causal theory of action (Bennett 1965, 90).

5. The claim that Aristotle even had a theory of action is controversial. For instance,
John L. Ackrill argued that although Aristotle says much about what interests action
theorists (e.g., about the etiology of action and our responsibility for action), “he
does not direct his gaze steadily upon the questions ‘What is an action?’ and ‘What
is an action?’” (Ackrill 1978, 601). For a response to Ackrill, see Freeland 1985. For
a partial defense of the claim that Aristotle provides a causal theory of action, see
Mele 1981.

6. “Under [pro-attitudes] are to be included desires, wanting, urges, promptings,

and a great variety of moral views, aesthetic principles, economic prejudices, social
conventions, and public and private goals and values in so far as these can be inter-
preted as attitudes of an agent directed toward actions of a certain kind” (Davidson
1963/1980, 4).

7. The translation is Martha Nussbaum’s (Aristotle 1978, 44). Also, in what follows
in our discussion of Aristotle, we will simply use “desire” to denote a conative state
generally. We will specify what type of desire when relevant.

8. Boulêsis is often rendered as “wish.” One could just as easily refer to intrinsic
conative states aimed at ends. “Wish” is nice shorthand, since wishes are pro-atti-
tudes. But for ease we stick to using “desire.” Nothing much hangs on this since
Aristotle’s story of motivation still has it that conative states are central as the
springs of action.

9. Rendering prohairesis as “instrumental action-triggering desire” is not uncontro-

versial. Richard Sorabji translates prohairesis as a desire for the means that will lead
to one’s ends (Sorabji 1979, 57; 2004, 11). H. Rackham in a footnote to the Greek
text in the Loeb Classical Library Edition of NE suggests that “preference” along with
“choice” are suitable translations of prohairesis (Aristotle 1934, 128, note a). A. W.
Price notes that prohairesis literally means “taking in preference” (Price 2004, 30).
Given the role of prohairesis as a pro-attitude directed toward executing the means
to achieving one’s ends in acting, “instrumental action-triggering desire” seems like
a better translation than “decision” or “choice” (which are common translations of
prohairesis). This is so because “preference” is not very informative, and “choice”
may lead one to the mistaken impression that Aristotle was a volitionist about
The CTA: Origins and Issues 23

actions (treating efforts of will as somehow necessary conditions for action). Admit-
tedly, “instrumental action-triggering desire” is infelicitous for reasons we cannot
explore here. But such a translation seems warranted given the functional role of
prohairesis in the execution of the means to satisfy an agent’s ends.

10. The translation is based on Irwin’s (1999, 87), with some significant changes
made based on consulting the Greek text.

11. We thus echo the general understanding of Aristotle’s account of intentional

action as a causalist account offered by David Charles (1984, chapter 2), and Alfred
Mele (1981). Against the understanding of Aristotle as offering a version of the CTA
or partial-CTA, see Coope 2004, 2007. Ursula Coope offers a causal powers account
of Aristotle’s theory of action that is largely based on what Aristotle says about
causation in Physics III.3 and VIII.5. But it is not obvious that such an approach is
incompatible with a CTA. In fact, we think that a complete, ontologically serious
version of the CTA should have something to say about the causal powers of agents.
But Coope seems to think that any such story we tell cannot be compatible with a
CTA (what she labels “the standard story” on which actions are caused by the beliefs
and desires of agents). This is not obvious. Rowland Stout in his essay in this volume
favorably discusses some of Coope’s work in his discussion of the causation of
action.

12. In this volume two chapters that explicitly exhibit the influence of Aristotle’s
ideas are the ones by Stout (chapter 7) and Juarrero (chapter 16).

13. Perhaps the only notable exception is Van Mill’s (2001) monograph Liberty,
Rationality, and Agency in Hobbes’s “Leviathan,” who nonetheless avoids drawing the
important connection to modern versions of the CTA and Hobbes’s theory of action.
Other significant discussions of Hobbes’s theory of action, though still embedded
within a larger account of his philosophy in general, are Peters 1967; Sorell 1986;
Tuck 1989; and Martinich 2005.

14. The quotation is taken from Aquinas 1999, 198.

15. See Pink 2004 for a discussion of the relationship of Hobbes’s work on action
to that of the Scholastics, particularly Suarez.

16. The text is from Leviathan, Part I.53. It turns out that all voluntary actions are
free actions on his view. He writes in The Questions Concerning Liberty, Necessity, and
Chance that “I do indeed take all voluntary acts to be free, and all free acts to be
voluntary” (1656/1999, 82–83). See Of Liberty and Necessity (1654/1999, §28).

17. To give some context, in the sentence in which the definition appears, he writes
that “I conceive that in all deliberations, that is to say, in all alternate succession
of contrary appetites, the last is that which we call the will, and is immediately next
before the doing of the action, or next before the doing of it become impossible”
(Hobbes 1654/1999, 37).
24 J. H. Aguilar and A. A. Buckareff

18. In the next section of this introduction we discuss the problem of the absent
agent.

19. In addressing the problem of basic causal deviance in his later paper, “Problems
in the Explanation of Action,” Davidson admits that he is “convinced that the
concepts of event, cause, and intention are inadequate to account for intentional
action” (Davidson 2004, 106). John Bishop, in his essay in this volume, notes that
Davidson is here making a conceptual claim. Because Davidson identifies actions
with events, Bishop claims that for Davidson the problem of basic causal deviance
does not pose a problem for the ontology of action.

20. For a defense of action-triggering intentions as playing a causal role through

the completion of an action, see Mele 1992. For a defense of the acquisition of a
separate action-guiding intention, see Searle 1983.

21. See Bishop 1989 and Mele 1992, 2003 for explicit statements of the causal role
of reasons and intentions that requires that actions be sensitive to the contents of
reasons for action and intentions.

22. See Bishop 1989 for extended discussion of this example. In turn, Aguilar’s
chapter in this volume discusses Bishop’s proposal to deal with such cases.

23. There is widespread agreement on the inadequacy of the old ballistic model of
mental causation in the etiology of action. The differences can be quite fine-grained
at times. Some strategies include, among others, those that emphasize causation by
self-referential intentions (Harman 1976/1997; Searle 1983), those that highlight the
immediacy of causal relations between action-triggering intentions—taking actions
to start in the brain (Brand 1984; Adams and Mele 1992; Mele 1992), and those that
afford feedback loops a prominent role in making sense of the guiding function of
intentions (Bishop 1989).

24. The most extensive analysis of this version of the absent agent problem and its
consequences for the CTA is by Velleman (1992/2000).

25. We can express the core tenets of the ACTA as follows: Any behavioral event A
of an agent S is an action if and only if (1) S causes A (either directly or by causing
some event); and (2) S’s causation of A is not ontologically reducible to causation
by some mental event or state of A.

26. Such a view is at least as old as Thomas Reid’s (1788/1983) theory of action.

27. In chapter 4 of this volume, Hornsby retakes this line of attack, while Michael
Smith in chapter 3 offers a reply. Another recent criticism of the CTA along these
lines, although based on a revised understanding of the role of consciousness, is
found in Schroeter 2004; for a CTA response, see Aguilar and Buckareff 2009.

28. For an illustration of his misgivings about the CTA, see Frankfurt 1978/1988,
70–72.
The CTA: Origins and Issues 25

29. Most of these are collected in Frankfurt 1988, 1998.

30. “What really produces the bodily movements that you are said to produce, then,
is a part of you that performs the characteristic functions of agency. That part, I
claim, is your desire to act in accordance with reasons, a desire that produces behav-
ior, in your name, by adding its motivational force to that of whichever motives
appear to provide the strongest reasons for acting, just as you are said to throw your
weight behind them” Velleman (1992/2000, 141).

31. Precisely for this reason some critics of the CTA argue that the embellishment
strategy amounts to a sort of window-dressing, failing to address the source of their
concerns; see Hornsby 2004 and Schroeter 2004.

32. So, for instance, where H. A. Prichard proposed that “to act is really to will
something” (Prichard 1945, 190), Jennifer Hornsby has defended the view that
“every action is an event of trying or attempting to act, and every attempt that is
an action precedes and causes a contraction of muscles and a movement of the
body” (Hornsby 1980, 33). We can express the core tenets of the VTA as follows:
Any behavioral event A of an agent S is an action if and only if S’s A-ing is either
identical with or is the consequence of an instrumental mental action M (a trying,
willing, volition).

33. See, e.g., Ginet 1990, 11–14.

34. Velleman’s embellished CTA is also seen by him as directly addressing the com-
monsensical conceptual framework presupposed in ordinary attributions of agency,
which goes to show that the two CTA strategies to deal with the problem of the
absent agent discussed here are not exclusive but potentially complementary.

35. Writing about the motivation for any theory of action, this is what Bishop has
to say about the role of common sense as the basis for generating philosophical
theories and distinctions: “For one thing, the fact that common sense draws a dis-
tinction does not automatically warrant the need for a philosophical theory of its
basis: To provide motivation for such a theory we need to find a source of genuine
philosophical puzzlement in the wider context in which the distinction functions”
(Bishop 1989, 11).

36. It is worth noting that in this respect Bishop’s theory of action is like J. J. C.
Smart’s (1959) formulation of the identity theory of mind. Smart’s identity theory
is an ontological theory and not a theory of the meaning of mental terms. Just as
the intension of “pain” need not be the same as its extension, so also the intension
of “action” need not be the same as its extension. See also Davidson’s effort to draw
a similar distinction concerning agency: “If we can say, as I am urging, that a person
does, as agent, whatever he does intentionally under some description, then,
although the criterion of agency is, in the semantic sense, intentional, the expres-
sion of agency is itself purely extensional. The relation that holds between a person
26 J. H. Aguilar and A. A. Buckareff

and an event, when the event is an action performed by the person, holds regardless
of how the terms are described” (Davidson 1971/1980, 46–47).

37. For simplicity we are only focusing on the debate over whether reasons or
mental events associated with them are triggering causes of action, explaining action
in virtue of their triggering role. Dretske (1988) argues that reasons are structuring
causes of action. This is an interesting and provocative theory that is compatible
with a standard causalist account of reasons as triggering causes. A fully developed
account of the etiology and explanation of human action should have something
to say of the different sorts of causal roles played by mental items. Paying close
attention to Dretske’s work should play a part in such work.

38. In a recent essay, Wilson (2009) indicates that his interest in action theory
comes from his bafflement with the widespread acceptance of Davidson’s causalism
about reasons-explanations of action. This interest led Wilson to offer a full-blown
teleological theory of action in his book The Intentionality of Human Action (1989),
where among other things he tries to take on Davidson’s Challenge.

39. For other recent defenses of teleological noncausalism about reasons explana-
tions, see Ginet 1990, 2002; McCann 1998; Schueler 2003; and Sehon 2005.

40. Mele 1992 and 2003 both include responses to significant challenges to causal-
ism and defenses of noncausal accounts of reasons-explanations more generally.

41. Some of what follows is based on Buckareff and Zhu 2009.

42. T. M. Scanlon endorses a similar view. He writes that “A reason is a consideration

that counts in favor of some judgment-sensitive attitude, and the content of that
attitude must provide some guidance in identifying the kinds of considerations that
could count in favor of it” (Scanlon 1998, 67). See also Rudiger Bittner 2001. For a
recent critique of Dancy, see Mele 2003, chapter 3, and Mele 2007. For a recent
critique of Scanlon, see Mele 2003, 76–79.

43. See, e.g., the experimental work of Bertram Malle (2004) that indicates that the
folk practice is more in line with psychologism and also supports causalism. For a
discussion of Malle’s work in making the case against the antipsychologism of Julia
Tanney (2005), see Buckareff and Zhu 2009.

44. E. J. Lowe (2008) defends an antipsychologistic theory in his most recent book
on the metaphysics of mind and action.
2 Renewed Questions about the Causal Theory of Action

Michael S. Moore

1 Introduction

The causal theory of action (CTA) has long been the standard account of
human action in both philosophy and jurisprudence. The CTA essentially
asserts that human actions are particulars of a certain kind, namely, events.
Within the genus of events, human actions are differentiated by three
essential properties: (1) such actions are (at least partially) identical to
movements of the human body; (2) those movements are done in response
to certain representational states of belief, desire, intention, volition,
willing, choice, decision, deliberation, or the like; and (3) the “doing” in
(2) is analyzed in causal terms. Put simply, according to the CTA, human
actions are bodily movements caused by representational mental states
such as beliefs, desires, and so on.
The CTA is sometimes put as a conceptual thesis—this is what we mean
by the verbs of action; sometimes as an ontological thesis—human actions
just are certain mental states causing bodily movements; and sometimes
as a mere extensional equivalence regardless of any claims of synonymy
or identity—wherever there is a human action, there is bodily movement
caused by a certain mental state, and vice versa. Sometimes the CTA is
advanced as a combination of two or more of these claims. But in any case,
these are, broadly speaking, descriptive theses asserted by the CTA.
The CTA also is widely thought to have normative implications. This is
because human action is intimately linked to the agency that makes us
both morally responsible and legally liable for certain states of affairs. The
CTA thus becomes an analysis of the nature of one of the central condi-
tions of responsibility.
In an earlier work I described and defended a version of the CTA (Moore
1993). My theses were recognizable versions of the three theses distinctive
of the CTA in general: (1) each human action is partially identical to some
28 M. S. Moore

bodily movement (and fully identical only to the sequence, volition-caus-

ing-bodily movement); (2) the distinctive mental state causing bodily
movements that are actions is volition (a kind of intention not reducible
to beliefs, desires, or belief–desire sets); and (3) the relationship between
volitions and the bodily movements that are their objects is a causal one,
in the standard sense of “cause” (no sui generis “agent-causation” allowed).
Both the CTA in general and my particular version and use of it have
come under scrutiny in the intervening years.1 It thus seems an opportune
moment to reexamine the essential tenets of the CTA in light of those
intervening developments. I shall largely organize the discussion of these
developments in three clumps, corresponding to the three essential tenets
of the CTA. That is, I divide criticisms of the CTA between: (1) those that
refuse to identify actions, even in part, with bodily movements; (2) those
that refuse to identify the causes of those bodily movements that are
actions as mental states of desire, belief, intention, volition, or the like;
and (3) those that deny that the relationship between volitions and the
bodily movements that are their objects is causal. I thus organize the dis-
cussion around the causal relata (1 and 2) and the causal relation (3)
asserted to exist between such causal relata by the CTA.
Preliminarily, however, before beginning this three-part discussion, I
shall briefly consider a much broader attack on the CTA, an attack that
goes to the naturalist impulse behind the CTA more than to any one of
the CTA’s particular theses. Generally speaking, this is the view that human
agency cannot be captured by any kind of causal analysis. Such a blunder-
buss objection has been more particularly phrased as the objection: (1) that
a bodily movement is the wrong kind of thing to be even a part of human
action, because such a movement “is not an agent’s doing anything” and
that therefore when “bodily movements are identified as actions . . .
no-one ever does anything” (Hornsby 2004, 20–21); (2) that mental states
like belief, desire, intention, etc., are the wrong kinds of thing to be causes
of actions, because the only “event that merits the title ‘action’ is a person’s
intentionally doing something” (ibid., 19); or (3) that causation is the
wrong relationship here, because even persons (as irreducible agents) do
not cause actions, even if persons can be said to cause the “results in terms
of which their actions may be described” (ibid., 18). One can thus phrase
the blunderbuss objection in terms of either the relata or the relation
posited to exist by the CTA. But the objection should not be force-fit into
one of these three pigeonholes. It is much more general. It is that human
agency is a kind of primitive that cannot be reduced to causation of any-
thing by anything.
Renewed Questions about the CTA 29

2 The Mysterians of Human Agency

The heart of the blunderbuss objection is that a person’s doing something

cannot be “equated with” (analyzed as, identified as, held equivalent to)
an event or state causing something. “Agents and their irreducible agency
get left out of such an analysis,” is the objection. I (and others) call this
the “mysterian” view of human action. The label comes from the mystery
engendered the moment one holds human agency to be special in the
sense that it stands outside the normal causal-explanatory matrix of
science. Explanations in normal science proceed by citing causal rela-
tions (or probabilistic or counterfactual relations, if these are held to be
distinct) between states of affairs as well as events; and this is just what
the mysterians of human agency deny about human actions and their
explanations.
The first thing to do with this objection is to cabin its reach. Surely the
“agency is special” view is limited to human agency. Surely the claim is
implausible about the agency of items like trees, as in the sentence, “the
tree shed its leaves.”2 Unlike talk of people doing things like shedding their
clothes, talk of trees doing things like shedding their leaves seems easily
reducible to ordinary event/states-of-affairs causal talk, as in “The leaves
fell off the tree.”
The mysterians’ objection builds its case by observing that people shed
their clothes in a sense that trees do not shed their leaves. The agency of
persons seems different, more robust, than the “agency” of trees. The CTA
would capture this difference in terms of mental state causation; the mys-
terians object that that leaves the agent out.
The CTA of course rejoins that agency and agents are not left out by the
CTA—for according to the theory, the exercise of human agency that we
call a human action just is mental-state causation of a certain sort. Trees
lack such mental-state causation of their behavior, which is why trees do
not have the sort of agency that persons possess. Agents and their agency
are no more left out of CTA analyses than temperature is left out in the
ideal gas theory (where only mass, velocity, and kinetic energy of mole-
cules is explicitly talked about), or water is left out in chemical discussions
carried on explicitly only in terms of H2O. The CTA makes an identity
claim about what human actions are, and it roundly begs the question to
simply assert that there is no such identity and so agency is left out when
one talks of mental-state causation.
The mysterians need to defend a necessity claim here, namely, the claim
that necessarily, the agency of trying and acting cannot be identical to
30 M. S. Moore

events or states of any kind or in any relationship—not to mental states

of belief, desires, intention, or volition; and not to the brain states that
correlate with these mental states, either. For the mysterians, human
agency is what it is, apparently, and not any other thing.
This is an expensive ontology. The mysterians’ irreducible agency is
about as desperate a maneuver as was Descartes’s refuge in dualism, or,
closer to our own times, as was Richard Taylor’s (1965) “discovery” of a
distinct kind of causation, the “agent-causation” supposedly distinctive of
persons bringing about changes in the world through their actions. In all
cases something is introduced that is inexplicable in the basic terms with
which we understand the world. Primitive, irreducible agency fits with the
causal-explanatory matrix by which we understand the world no better
than did Taylor’s distinct kind of causing, or Descartes’s distinct kind of
mind stuff. In all cases it is mysterious how we are to think about these
items and their relations to the world of physical objects, events, states of
affairs, and their relations and properties, with which we are familiar.
To be sure, what one finds mysterious depends on what one thinks
antecedently to be familiar and unproblematic. Stare at the ordinary causal
relationship very long and it too may look a bit mysterious. Analyses of it
in terms of primitive causal relations between particulars (Tooley 1987), or
primitive causal-law relations between universals (Armstrong 1983), or
single-case chance raisings (Mellor 1995), or the possibilia of counterfactual
dependence (Lewis 1973), or nomically sufficient causal laws (Mackie
1974), or primitive causal powers and dispositional abilities (Mumford
2009), and so on, may not dispel the mystery much. In light of this, one
might urge that irreducible agency is no stranger than is the causal rela-
tionship with which it is at odds. (As David Armstrong 1999 remarked
some years ago, those who live in ontological glass houses should throw
few stones.)
Despite these concessions to ontological charity, irreducible agency is
too strange to be accepted. We should agree with Michael Bratman (and
Davidson before him) when he finds it “difficult to know what it means
to say that the agent, as distinct from relevant psychological events, pro-
cesses, and states, plays . . . a basic role in the aetiology and explanation
of action” (Bratman 2000, 39). Causal relationships, or counterfactual or
probabilistic dependencies between events and states of affairs, we think
we understand, difficult as those notions are. Indeed, those relations and
their relata are the bedrock of any naturalistic understanding of our world.
Talk of objects standing in these relations may be idiomatic English before
our talk grows seriously ontological, but it is always elliptical for events
Renewed Questions about the CTA 31

and states of affairs involving such objects.3 Talk of one such object (human
agents) standing in such relationships where the talk is explicitly not
elliptical—that is too mysterious to be taken seriously.

3 Are Bodily Movements Essential to Human Actions?

I now wish to examine the first of three more particular skepticisms about
the CTA. This is the skepticism denying that bodily movements are neces-
sary for human actions. Initially, we should distinguish the CTA thesis
discussed here from a different thesis, namely, the thesis that when there
are bodily movements involved in some human action, it is to those move-
ments that reference is made when using verbs of action. The latter thesis
could well be true while the thesis of interest here—that action essentially
involves bodily movements—was false.
My own critics, and those more generally critical of the CTA, often
proceed by way of counterexample. Surely, the argument goes, something
(x) is an action, and yet just as surely x involves no bodily movements. It
is useful to group the kinds of counterexamples proposed by different
values for the “x” in this formula. I shall consider four sorts of such coun-
terexamples. First, there are willed stillnesses by some agent. These include
the guardsman at Buckingham Palace trying (successfully) to remain per-
fectly still (Fletcher 1994, 1445), or the bird-watcher prone to twitching
who keeps still to avoid alarming the bird she is watching (Duff 2004, 83),
or the chocolate lover who is tempted to reach for the chocolates next to
her but stays still, resisting the temptation (Hornsby 2004, 5), or the hero
standing steadfast on the burning deck of some ship when every fiber of
his being urges flight (Annas 1978). These are surely actions, the objection
continues, yet they involve no movement of the bodies of the actors
involved.
A second set of counterexamples is provided by the very recent con-
struction of “mind–brain interface machines.”4 I imagined one of these in
a recent article (Moore 2009b) (before I knew they actually existed): Suppose
the American patriot who desires to warn Dawes and Revere whether the
British are coming by land or by sea is so wounded that he can’t move;
yet he is hooked up to a device on his scalp that measures the readiness
potential in the supplementary motor area of his brain when he is about
to perform a voluntary motor movement, and that device in turn is hooked
up to the light in the tower of the Old North Church in Boston. He wills
one movement of his finger. This causes not the finger to move, but the
device to register his attempt, and the light to shine but once; Revere and
32 M. S. Moore

Dawes get the message, alert their fellow rebels, and so on. Surely the para-
lyzed patriot has performed the action of alerting the rebels even though
he didn’t (because he couldn’t) move his body.
It is not very plausible to deny that there are actions in each of these
first two sorts of examples. More plausible is to deny that there are no
bodily movements in such cases. Consider, in descending order of obvious-
ness, a series of cases where this might plausibly be asserted.

(a) Displacement refraining.5 We might keep ourselves from twitching one

arm, for example, by holding it still with the other arm. There are obvious
bodily movements here.
(b) Resistings.6 When outside forces threaten to move us, we sometimes
resist them by tensing the appropriate muscles. It is not much of a stretch
to include such muscle tensings as bodily movements.
(c) Willed but indirectly caused bodily changes.7 When we know that having
a certain mental state can cause some bodily change, and we create such
a state in order to effect some further change in the world, there is a bodily
movement in the sense of a change in bodily state. In the Paul Revere
example, this is the change in the scalp picked up by the readiness poten-
tial reading device.

Accommodating all three of these as the bodily movements required by

the CTA is to carry out Donald Davidson’s early suggestion that the CTA
must “interpret the idea of a bodily movement generously” (Davidson
1980, 49). But the CTA theorist must be careful lest this kind of generosity
give away the theory. Particularly in the third kind of case, one cannot
allow any bodily changes caused by mental states to qualify as actions. An
amputee with phantom limb experience may try to move his phantom
limb, and this may cause a facial tic, yet the tic is no action of his; we thus
need to exclude cases where the bodily change is not adopted by the
subject as the means to causing some further effect in the world.8 If the
patriot in the Old North Church does not know that his trying to move
will cause changes in his scalp (and that these will set off the brain pro-
cesses changing his scalp’s readiness potential and thereby light the desired
light), setting off the light is no act of his; for he did not intend those
bodily changes, nor did he intend to produce some effect in the world by
those changes.
Nonetheless, as so qualified, these first two kinds of counterexamples
are easily dealt with. There are bodily movements in such cases, so long
as one is generous in one’s interpretation of “bodily movements.”
Renewed Questions about the CTA 33

A third sort of counterexample is not so easily dealt with. This is the

example of mental acts. As Tony Duff has recently observed, “we . . . dis-
tinguish within the realm of thought those kinds of thought in which we
are actively purposeful from those in which we are passive” (Duff 2004,
81). A memory, for example, can just come to us, or it can be retrieved by
conscious effort because it is needed for some purpose. The objection, then,
juxtaposes two facts: Mental acts like retrieving a memory are actions too,
and yet they involve no movement of the body.
Davidson was inclined to extend his generous interpretation of bodily
movements to include mental action. As he said, “the generosity must be
openhanded enough to encompass such ‘movements’ as standing fast, and
mental acts like deciding and computing” (Davidson 1980, 49). Yet I have
never seen how the CTA theorists can be this generous and still retain any
content to the thesis (a part of the CTA) that “mere movements of the
body . . . are all the actions there are” (ibid., 59).
One might say that mental actions like computing or deciding do cause
changes in bodily states, namely, changes in the brain. Yet for some such
changes, the relation is not that of causation but of identity: These mental
acts are those brain states. And for those brain states not identical to the
mental acts in question and that are caused by mental acts, these are not
intended by the subject as means to effecting some further end, and thus
do not satisfy the restriction I appended to the generosity of interpretation
of the CTA.
Mental actions are not of interest to contemporary Anglo-American
criminal law, for the law famously does not “punish for thoughts alone.”9
Mental actions are also not of central concern to morality, for one’s obliga-
tions are not to cause certain states of affairs in the real world. We have
few (if any) obligations not to think of (wish for, hope for, etc.) such states
of affairs.10 Still, legal systems and moral theories that do not respect
these limitations are not unknown to us. And in any case, the CTA is a
descriptive thesis: The claim is that action as such—not just responsibility-
enhancing action—essentially involves bodily movements.
The existence of mental actions shows that there is a kind of human
agency that exists that does not involve bodily movements. Such agency
can be exercised without there being an intended causing of any bodily
changes, as when one engages in an act of imagination or computation
for the sheer pleasure of doing so. One can of course simply stipulate that
we are not very often interested—morally, legally, or otherwise—in such
flights of fancy or other instances of such agency. Yet a fourth class of
34 M. S. Moore

counterexamples shows us that we are often interested in precisely this

form of agency, and it is thus to those kinds of examples to which I now
turn.
The fourth species of counterexamples (to the CTA’s bodily movement
thesis) is that of omission. Consider some of Tony Duff’s examples: (1) “I
can . . . insult someone . . . by not rising from my chair when she enters
the room, if politeness or deference requires me to rise; or by not acknowl-
edging an acquaintance when she greets me on the street” (Duff 2004, 81).
(2) I can breach a promise by not showing up when promised (ibid.). (3)
In the right circumstances I can “vote for a motion by not raising my
hand” (ibid.). Duff concludes of these examples both that they are instances
of omitting where no bodily movement takes place, and that these are
actions the various persons performed. Duff concludes of such cases that
“I do something without moving, or by not moving.” Duff calls these
“commissions by omission” (ibid., 82).
Duff unfortunately equates acting with his idea of doing something,
actions with things done. This is an identification to be resisted. Resisting
it allows one to say truthfully that there are no actions to be found in such
cases—in which case, of course, they are not counterexamples to the causal
theory of action. What is more accurate to say about such cases is this:
Sometimes we can insult someone, break a promise, or vote, precisely by
not acting. Adopting Jennifer Hornsby’s (2004) characterization of such
cases, in all of these we do something intentionally and yet we perform
no action.
My earlier conclusions about omissions are thus secure. One can
concede that there are no causally relevant bodily movements in cases of
omissions, but since such omissions are not actions they are not counter-
examples to the CTA. Yet a critic of the CTA is entitled to reply that
the subject of the CTA—actions—loses interest if so construed; rather, the
reply continues, the (explanatorily and morally) interesting category is the
larger notion of agency illustrated in intentional omissions and in mental
actions, captured (along with actions) under the general rubric of “things
done.”
Conceptual disagreements about how best to carve up the world can
rarely be conclusively resolved. There is something to be said for lumping
willed omissions, mental actions, and actions together under some general
rubric such as “things done,” for these items do all express some kind of
human agency. Yet cutting the other way are two considerations, one
moral and the other metaphysical. Morally, there is an important distinc-
tion most of us recognize between the often stringent negative duties of
Renewed Questions about the CTA 35

morality, on the one hand, and our less stringent positive duties (plus
nonobligatory acts of super- or suberogation), on the other.11 We have a
stringent moral duty not to kill strangers, for example, but only a much
less stringent (or nonexistent, depending on the circumstances) duty to
prevent their death. A distinction between things done/things not done
does not recommend itself as serving to mark these moral differences. For
all there would be in the “things not done” category would be noninten-
tional omissions; intentional omissions would be on the “things done”
side of the conceptual divide, despite these being breaches of less stringent
positive duties (if breaches of obligation at all). We need a distinction that
tracks the positive/negative duty distinction in morality, which the act/
omission does well and which the things done/things not done distinction
does poorly.
Second, there is an important metaphysical distinction between items
that can serve as the relata of singular causal relations and items that
cannot. Willed bodily movements that are human actions (or on some
views of relata,12 the states of affairs constituted by such actions having a
certain property) can serve as such relata; the absence of such actions (i.e.,
omissions) cannot. The cause/failure to prevent line is not happily marked
by the things done/things not done line, for omissions are failures to
prevent, no matter whether intentional, negligent, or merely nonnegli-
gently inadvertent. Some of the items in the things done category can thus
stand in singular causal relations, but others (viz., intentional omissions)
cannot.
We thus have two good reasons to prefer an act/omission line to the
broader conception of agency implicit in the idea of things done. And by
the narrower notion of agency contained in the concept of a human
action, omissions and mental acts remain on the nonactional side of the
distinction. They thus are not counterexamples to the most desirable
concept of human action, that of a willed bodily movement.

4 Are Mental States Eligible to Be Relata in a Causal Relationship with

Bodily Movements?

I turn now to the second of the theses constitutive of the CTA. This is the
thesis that it is mental states like intention that are the distinctive source
of human agency and action. It might well be, though, that the least
troublesome subthesis of the CTA is the mental-causation thesis. For the
dominant tradition, denying that mental states like intention are the right
sort of things to be causes of physical events like bodily movements is the
36 M. S. Moore

metaphysically dualist tradition; yet non-interactionist-dualist theories of

mind are implausible on their face in these denials.
More plausible are recent critiques like that of Helen Steward (1997).
Steward attacks the CTA by denying that there are particular mental states
that can stand in the requisite causal relation. Steward concedes that facts
about persons, such as the fact that a person intends to move his finger,
can have “causal relevance” to the fact that his finger moved. But “causal
relevance” for Steward is a counterfactual notion, not to be confused with
actual causation between particulars. What the CTA holds to be causal
relations between particular mental states and particular bodily move-
ments, Steward thinks to be only a counterfactual relation between facts
about persons and facts about the movements of their bodies.
Steward joins a lengthening list of metaphysicians who distinguish rela-
tions of counterfactual dependence from true causal relations between
particulars.13 She also joins a well-known list of those who take facts to be
the relata of some cause-related relation such as her “causal relevance.”14
Peculiar to Steward, however, is her insistence that the facts that are the
relata of her relation of causal relevance are just true propositions, not the
states of affairs that make true propositions true.15 Most of us would insist
that for the relations of causation or of causal relevance to be in the world,
one must at some point abandon propositions for their truthmakers. I
(Moore 2009a, 341–348), David Armstrong (1997), and Kit Fine (1982) call
such truthmakers “states of affairs”; Peter Menzies (1989) terms them “real
situations”; Hugh Mellor (1995) coins the term “facta” for them. But what-
ever the label, Steward’s proposed replacement for the CTA—in terms of
counterfactual relations between facts—constitutes its own kind of “lin-
guistic dualism”16 because of its explicit severance from particulars that
exist in the natural world.
However stands Steward’s attempt to make sense of seeming causal talk
about minds in noncausal terms, her actual critique of states of mind like
intentions standing in genuine causal relations rests on her assumption
that there is no such thing as an intention-token. Since there are no such
things as intention-tokens (or belief-tokens, or desire-tokens), there are no
particulars to stand in a genuine (singular) causal relation. So the CTA must
be false.
Steward’s reasons for skepticism about there being token states of inten-
tion are multiple and elusive. She argues that we don’t need token states
of mind that can figure in causal relations, because we (allegedly—see
above) can make do with facts about persons that stand in the relation she
calls “causal relevance.” She argues that the identity of mental states with
Renewed Questions about the CTA 37

brain states depends on there being token mental states, but that such
commitments commit us a priori to the view that folk psychology will be
borne out by neurophysiology, an a priori commitment she finds “absurd”
(Steward 1997, 134). She urges that representational states are (almost)
always individuated by their contents and not by the subjects who hold
them or the time at which they are held; so we can make do with talk of
representational states in the abstract, with no commitment to there being
tokens (ibid., 131–133).17 She argues specifically against there being tropes
(or “abstract particulars,” or “property instances”), recognizing (correctly
enough) that these are distinct from states of affairs (the having of a prop-
erty by a particular at a time).
These and other arguments raise very broad issues of general ontology,
of causal relata, and of mind–brain relations. We are not going to settle
such large-scale issues here. My own long-term, naturalistic commitments,
argued for elsewhere, are different. Unlike Steward (ibid., 38), I think that
there is a right answer, in general, as to what causal relata are, and they
do not include whole objects or persons; that there are token states of
intention and belief; that these can stand in causal relationships; that such
mental states are identical with certain brain states, even if it turns out
there are no universal type identities here but only a disjunction of local
ones; and that, accordingly, mental states of volition cause the bodily
movements that are their objects (i.e., I’m committed to the CTA). Having
argued for most of these basic commitments elsewhere,18 I shall return to
more manageable (because more specific) objections to the CTA.

5 Must Mental States of Intention or Volition Cause the Bodily

Movements That Are Their Immediate Objects, for Those Movements to
Be Human Actions?

The final worry about the CTA focuses on the causal relation between
intentions and the bodily movements that are the objects of such inten-
tions. Such worry can concede, arguendo at least, that human actions are
(at least partially) identical to bodily movements and that there are token
mental states of desire, intention, and belief. The worry here specifically is
that human action cannot be analyzed in terms of causation.
There is a raft of old worries here that should be mentioned only if to
put them aside. These were the worries of the post-Rylean ordinary lan-
guage philosophers that preceded (and largely motivated) the consensus
in contemporary philosophy about the CTA. These included the claims
that: analytically, an event could not be a human action if it was caused
38 M. S. Moore

because such actions had to be the product of free will to be actions at

all;19 reasons are the only appropriate explanation of human actions, and
reason-giving explanations are not causal in nature;20 reasons for action
give justificatory warrants for the actions they are cited to explain, not
causal explanations;21 the relation of a reason to its action is logical and
thus cannot be causal;22 actions form the objects of reasons, not the effects
of reasons;23 actions are meaningful phenomena, not the colorless, dumb
events required by mechanistic causal accounts;24 actions are instances of
rule-following behavior, whereas caused bodily movements are only
instances of rule-governed behavior; and so on.25
The CTA found its present status as the standard philosophical theory
by the rejection of these old, linguistically dualist arguments.26 Yet even
among those rejecting these arguments surfaced a new version of the worry
about equating action with a caused something. This was the “deviant
causal chain” worry.27 This was the worry that “causation of a bodily move-
ment by a mental state” was not a sufficient condition of action—because,
for example, I could desire my old enemy’s demise, believe that if I pushed
my foot down on the accelerator of the car that was aimed at my old
enemy, I would kill him, intend to push down my foot, and these beliefs,
desires, and intentions could cause my foot hitting the accelerator—and
yet, my foot moving could be no action of mine. It just slipped because I
got so excited by my opportunity to satisfy my desire and my intention,
for example.
The answer to the deviant causal chain worry that I and others have
long accepted is a naturalistic one. It is that not just any causal relation
(between intention and bodily movement) is meant by the CTA as a cri-
terion of action. Rather, as Alvin Goldman put it long ago, beliefs/desires/
intentions must cause bodily movements “in a certain characteristic way”
for these movements to be actions (Goldman 1970, 61). Moreover, although
“we are aware, intuitively, of a characteristic manner in which desires and
beliefs flow into intentional acts” (ibid., 62), the nature of this “character-
istic manner” is not fixed phenomenologically. Rather, the nature of the
kind of causal relationship involved in the execution of intentions is a
matter for science, specifically, the neuroscience of voluntary motor
movement.
This answer may seem to offer comfort to the mysterians of human
agency discussed earlier. This, because the answer concedes that not just
any old kind of belief-desire-intention (BDI) causation fits human action;
only the “agentist” kind of causation does that. This might be thought to
be a return to a kind of primitivism about agency and agent-causation. Yet
Renewed Questions about the CTA 39

this would be a mistake. There is no second kind of causation being

invoked here, nor is there a question-begging reference to irreducible
agency. Rather, it is ordinary, garden-variety causation to which agency is
to be reduced—only, it is that causal relation as it specifically exists between
the supplementary motor area of the brain (and other regions) and the
mechanisms that initiate muscle movements that fleshes out the charac-
teristic relation distinctive of human agency. As Goldman said decades ago,
this is a matter for scientific discovery, and over the intervening years we
have discovered much about the neuroscience of voluntary motor
movement.
Ironically, the very neuroscience that has fleshed out the answer to the
old “deviant causal chain” worry has also spawned a different worry for
the CTA. This worry goes like this. The neuroscience since the mid-1980s
has discovered that well prior to a subject’s conscious awareness of any
intention to move his body, there exist certain brain events that are caus-
ally connected to the bodily movements in question.28 The objection is
that the intention to move the body thus comes too late to have the causal
role the CTA says it has—the intention is a co-effect of the earlier brain
events, along with the bodily movements, but itself without effect on those
movements.
I and others have dealt with this worry at some length elsewhere,29 and
there is now an extensive literature. Briefly, and in summary of those
longer discussions, there are two sorts of answers to neuroscience’s epiphe-
nomenal challenge to the CTA. One is to refuse to identify the intention
to move one’s finger with the conscious experience of that intention;
rather, one identifies such intentions with the brain events that precede
such conscious experience.30 With such an identification, the CTA is secure:
for then the intentions to move are not the effects of the earlier brain events
that are evidenced by the readiness potential, but are identical with such
brain events and thus share the causal role of such events in the produc-
tion of voluntary bodily movements.
There are considerable variations of this kind of defense of the CTA in
the literature,31 but exploring them and choosing between them does not
answer my present purposes. Rather, I want here to explore a second kind
of reconciliation of responsibility with the facts of contemporary neurosci-
ence. This answer concedes (arguendo, to be sure) that intentions do not
occur prior to our conscious experience of them, and that therefore, given
the evidence from neuroscience, such intentions are indeed epiphenom-
enal with the bodily movements that they putatively cause. Even so, the
argument is that there is both human action and moral responsibility for
40 M. S. Moore

it, despite the lack of causation of the relevant bodily movements by the
intention to do them.
Revert to my earlier imagined patriot trying to alert his fellow rebels
that the British are coming by land. Suppose he knows that, because of his
wounds, he cannot move his finger. But suppose he also knows enough
contemporary neuroscience to know that if he tries to move his finger, 300
milliseconds prior to such effort there will be activated in his brain pro-
cesses that will cause a “readiness potential” to begin, detectable in his
scalp over the supplementary motor area of his brain. Luckily for him, a
patriot neuroscientist has hooked him up to a “mind–brain interface”
machine, and this machine will read the slow negative shift in his readi-
ness potential, which reading will cause the light to go off in the tower of
the Old North Church in Boston. Knowing all this, the patriot tries to move
his finger but once, the light goes off but once, Revere and Dawes begin
their famous ride to alert the citizens of Lexington and Concord that the
British are coming by land, and the rest is history. I take it that the patriot
is morally responsible for alerting his fellow rebels and, if caught, is fairly
hung for treason by the British. For he has performed the actions of alert-
ing the rebels, sending the signal, lighting the light—even though his
intent to do any of these things caused only an intention to try to move
his fingers, and that intention, ex hypothesi, did not cause any of these
things. Rather, the intention to try to move his finger was the co-effect of
those brain events that also caused the change in scalp readings, which
caused the light to be lit, which caused the rebels to be alerted.
These conclusions are troublesome for the CTA. I put aside two easy
responses to them. One is to “outsmart”32 the argument by denying that
there would be either action or responsibility in such cases. As Gideon
Yaffe (2009) put it in response to this argument against the CTA, “if I have
to choose between the CTA and the conclusions (of action and responsibil-
ity) in such cases, I choose to adhere to the CTA.” The intuition that the
patriot performed the action of alerting the rebels, and that he is respon-
sible for doing so, is too strong for this willing of belief, is it not?
The second response is to rely on the earlier identification of intention
with the unconscious brain events measured by the shift in readiness
potential. Then in the actual world we live in—as opposed to a merely
possible world incompatible with the laws of this world—there are no
stories like the one I just told. In the actual world, the imagined patriot
would indeed perform the action of alerting the rebels, but he would do
so in virtue of his intentions (to alert, to light, to try to move) causing the
bodily change that caused the rebels to be alerted. In the actual world, the
world in which we live, there is no epiphenomenal problem.
Renewed Questions about the CTA 41

I by and large eschew this response because it is insufficiently ambitious

for the CTA. The CTA could not be an analysis of human action, on this
response, nor could the CTA be giving the essential properties of actions.33
So I prefer a third response, which is to amend the CTA. What is essential
to human actions is that the bodily movements that in part constitute
them be the means by which some agent chooses to make changes in the
world. Almost always the means–end relationship will be a causal one:
usually the states we choose as our means to bring about some state that
is our end will be a cause of that end-state. But not always, and not neces-
sarily. If the way I know how to achieve some end-state is by bringing
about a means-state that succeeds in time my end-state, the relationship
cannot be causal. But I nonetheless do the one by doing the other. Two
time-honored examples: I may get a square hit on a golf ball by getting a
good follow-through on the swing; similarly, I may move certain of my
muscles by moving my arm or leg. In each case what I know how to do—get
a good follow-through, move my limb—will involve an earlier state, which
is what I am interested in achieving. I may not be able to achieve this
earlier state directly, but I can achieve it indirectly, because the doing of
what I know how to do will take place only if this earlier state also has
taken place—in which case I have performed an action even though my
BDI states didn’t cause it.
The upshot is that we should amend the CTA with respect to causation
(just as we amended it with respect to bodily movements to include any
physical changes in the body, whenever those changes were adopted as
means to some further end). Epiphenomenal relations between intentions
and movements will suffice to make such movements actions so long as
such intentions are the means chosen to having such movements occur.
In such cases there must still be some kinds of causal relationships, one
between certain brain events and such intentions, and the other between
those brain events and the bodily movements intended. These are what
makes possible the use of the intentions as a means to having had such
brain events occur. But there would not need to be, in the limited class of
cases imagined, a causal relationship between the intentions and the
bodily movements intended.

Notes

1. For criticism of my own version of the CTA, see the twelve articles collected in
a symposium on my Act and Crime book (Moore 1993). See also Mathis 2003 and
Duff 2004.
42 M. S. Moore

2. This is Richard Taylor’s example; see Taylor 1965, 59.

3. As I (and many before me) have argued. See Moore 2009a 333–334.

4. See the discussion of these in Farahany 2009.

5. See Moore 1993, 87–88. The concept is Bruce Vermazen’s; see Vermazen 1985.

6. See Moore 1993, 88.

7. See Moore 1993, 106–108, 263–267.

8. See Moore 1993, 106–108.

9. See the discussion in Moore 1993, 17–19.

10. Morality is not indifferent to how we feel and think, however; the virtues in
part concern themselves with such matters. See Williams 1973.

11. See Kamm 1994.

12. See Moore 2009a, chap. 15.

13. See, e.g., Moore 2009a; Dowe 2000; and Hall 2004.

14. See, e.g., Mackie 1974; Mellor 1995; and Bennett 1988.

15. “A fact exists, in my usage, when any proposition is true—there is no more than
this to the existence of facts” (Steward 1997, 104).

16. The apt phrase used to describe ordinary language philosophy’s attempt to
preserve the specialness of the mental without committing to metaphysical dualism.
See Landesman 1965.

17. One place Steward might have looked for a commitment to there being inten-
tion-tokens is in the law. In contract law, for example, there is no contract unless
the contracting parties have “the same intention,” i.e., two intention-tokens having
the same content but held by different persons. In criminal law, to take another
example, when we blame an accused for intentionally hitting one or more persons
that he in fact did hit but did so while intending to hit someone else (the “trans-
ferred intent doctrine”), we count the intention-tokens held by the accused to see
when they are “exhausted” such that no more can be transferred.

18. On functionalism, see Moore 1988; on causal relata, see Moore 2009a, chapters
14–15; on intentions specifically, see Moore 1993, chapter 6, , 1997, and 2010.

19. See, e.g., Moya 1990.

20. See, e.g., Peters 1958.

21. See, e.g., Louch 1966.

22. See, e.g., Melden 1961.

Renewed Questions about the CTA 43

23. See, e.g., Anscombe 1963.

24. See, e.g., Peters 1958.

25. See, e.g., Winch 1958.

26. Most notably in Davidson 1963.

27. Chisholm 1964.

28. This version of the much older epiphenomenal worry surfaced initially in the
work of Benjamin Libet. See, Libet et al. 1983 and Libet 1985. Libet’s work, and his
skeptical conclusions from it, have been carried on by Haggard and Elmer 1999,
Wegner 2002, and most recently, Haynes et al. 2007, and Soon et al. 2008.

29. See Moore 2006, forthcoming; and Mele 2006, 2009. There is an extensive early
symposium on Libet in 1985 Behavioral and Brain Sciences 8 (1985).

30. My main answer is in Moore forthcoming.

31. Mele (2003), for example, identifies these brain events with urges to move rather
than intentions.

32. The verb defined (out of Jack Smart’s name) in Dennett and Lambert 1978.

33. An example of the limited ambitions of this kind of claim is provided by the
theory of causation advanced in Dowe 2000.
3 The Standard Story of Action: An Exchange (1)

Michael Smith

Suppose an agent acts in some way. What makes it the case that he acted,
as distinct from his having been involved in some mere happening or
other? What makes him an agent, rather than a patient? According to the
standard story of action that gets told by philosophers, the answer lies in
the causal etiology of what happened (Hume 1777/1975; Hempel 1961;
Davidson 1963).
We begin by identifying some putative action that the agent performed
by tracing its effects back to some bodily movement. This bodily move-
ment has to be one that the agent knows how to perform, and it further
has to be the case that his knowledge how to perform it isn’t explained by
his knowledge how to do something else: in other words, it must be one
that could be a basic action (Danto 1963; Davidson 1971). We then estab-
lish whether the agent acted by seeing whether this bodily movement was
caused and rationalized in the right kind of way by some desire the agent
had that things be a certain way and a belief he had that something he
can just do, namely, move his body in the relevant way, has some suitable
chance of making things the way he desired them to be. If so, then that
bodily movement is an action; if not, then it is not.
It is easy to imagine someone objecting to this standard story right from
the outset: “If the standard story of action says that I act only if I move
my body, then it entails that there are no actions like those performed by
children who stand absolutely motionless when given the direction to do
so in a game of Freeze, or actions like sitting still in a chair or lying on a
bed. In these and a host of similar cases we plainly act, but we do so
without moving our bodies.” Despite its rhetorical force, however, this
objection rests on an uncharitable interpretation of what the standard view
has in mind when it talks about bodily movements.
When a defender of the standard view says that actions are bodily move-
ments, this has to be interpreted so that any orientation of the body counts
46 M. Smith

as a bodily movement. Think of sitting in a chair. Sometimes the orienta-

tion of your body when you sit in a chair is under your control. Even if
you’re sitting still, your doing what you’re doing is sustained by some
desire you have and the belief that you can get what you desire by simply
sitting still. It is this feature of the orientation of your body—the fact that
it is under your control in the sense of being sensitive to what you desire
and believe—that the standard story says makes it an action, whether or
not you happen to be actually moving. Sometimes, however, when you’re
sitting perfectly still the orientation of your body isn’t under your control.
This is the case when (say) you’re sitting in a chair but you’re fast asleep.
You’re sitting perfectly still, but you’re not acting. As I understand it, the
standard story offers an account of what this difference consists in. In the
former case, the orientation of your body is sensitive to what you desire
and believe, whereas in the latter it isn’t. From here on I will therefore
simply assume that any way in which an agent might orient his body
counts as a bodily movement.
To see how all of this works in practice, consider a simple example.
Suppose that John flicks a switch. What makes it the case that he acted?
According to the standard story, we answer this question by first of all
identifying what we are imagining to be an action, John’s flicking the
switch, with some relevant bodily movement of his by tracing back from
the imagined action’s effects. Let’s suppose that the bodily movement we
discover when we trace back is John’s moving his finger in some way. If
John’s flicking the switch is to be an action then this bodily movement
has to be one that John knows how to perform, and his knowledge
how to perform it mustn’t be explained by his knowledge how to do
something else. It must be the sort of bodily movement that could be a
basic action.
Note that this already provides us with grounds on which to disqualify
certain doings from being actions. Imagine that John flicked the switch
when he brushed against it with his finger, but that this in turn happened
because he was hit by a truck and, in the collision, his finger was bent
backward. In that case, what we were imagining to be an action, John’s
flicking the switch, turns out to be no such thing. It is rather a mere hap-
pening in which John is involved, albeit not as an agent. For though we
can trace the flick of the switch back to a bodily movement, the bodily
movement in question—his finger’s being bent backward—isn’t a bodily
movement of a kind that John knows how to perform without performing
some other action. Bending a finger backward isn’t something that could
be a basic action of John’s.
The Standard Story of Action (1) 47

Cases like this underscore how very loose the connection is between
doings and actions. For though it would be perfectly acceptable to describe
John as having done something in this case—he did flick the switch, after
all—his doing that isn’t an action. Moreover, even though John’s finger
did indeed bend backward, bending his finger backward isn’t something
that he did, and nor did his finger bend backward as a result of anything
else he did. So, even though John did indeed flick the switch, he didn’t
act when he did that, and he didn’t do it by doing anything else, either.
When you dwell on it, it can seem puzzling that we are permitted to
describe things in these terms. But I doubt that it is worth dwelling on for
too long. Given how loose the connection is between doings and actions,
not much in the way of philosophical illumination about action is going
to be gained by attending to those occasions on which we do and don’t
describe people as doing things.
Let’s return to our original example. Given that the bodily movement
that John performs is one that he knows how to perform without perform-
ing some other action—his moving his finger, in our original example—the
standard story tells us that we can determine whether it was an action by
investigating its causal antecedents. Was that movement caused and ratio-
nalized by a desire John had that things be a certain way and a belief he
had that his moving his finger had some suitable chance of making things
the way he desired them to be? Did he (say) desire the illumination of the
room and believe that he could illuminate the room by moving his finger
against the switch? If so, did his desire and belief cause his finger move-
ment in the right way? If so, then that finger movement is an action; if
not, then we once again have to conclude that John was involved in a
mere happening in which he wasn’t an agent.
Standard though this story of action is, it is not universally accepted.
Jennifer Hornsby, for example, has recently provided several arguments
aimed at demonstrating the story’s crucial flaws (Hornsby 2004). Defenders
of the standard story should welcome such objections, as one of the best
ways they have available to them to demonstrate the plausibility of their
view is to work through and respond to objections. In what follows I will
consider four of the objections Hornsby puts forward. To anticipate, three
of her four objections seem to me to be simply misplaced. The fourth
objection is more worrying, but it turns out that it isn’t so much an objec-
tion to the standard story itself as an expression of dissatisfaction with a
certain way of developing that story. Once the dust settles, the standard
story of action thus seems to me to remain pretty much intact, notwith-
standing Hornsby’s four objections.
48 M. Smith

1 First Objection: Actions and Omissions

Hornsby’s first objection is that there is a mismatch between a central

concern of the standard story, which is to explain agency quite generally,
and the focus of the standard story on action. This is because, as she puts
it,

someone can do something intentionally without there being any action that is
their doing that thing. Consider A who decides she shouldn’t take a chocolate, and
refrains from moving her arm towards the box; or B who doesn’t want to be dis-
turbed by answering calls, and lets the telephone carry on ringing; or C who, being
irritated by someone, pays that person no attention. Imagining that each of these
things is intentionally done ensures that we have examples of agency. . . . But since
in these cases, A, B and C don’t move their bodies, we have examples which the
standard story doesn’t speak to. (Hornsby 2004, 5)

But is Hornsby right that the standard story does not speak to what
happens in these cases?
Hornsby is surely right that we ordinarily distinguish between actions
and omissions. The question, however, is whether the standard story of
action is intended to be a story about actions, in the sense of “action” in
which actions are distinct from omissions. Since I agree with her that the
standard story aims to explain agency quite generally, I doubt that this is
so. It seems to me much more plausible to suppose that it is intended to
be a story about actions in a quite general sense in which the distinction
between actions and omissions is invisible. From here on I will assume that
this is so. It might be thought that this just makes Hornsby’s criticism even
more acute. For we have already seen that the standard story identifies
actions with bodily movements, so isn’t it going to be impossible for it to
be a story about the agency involved in omissions, given that omissions
involve failures of bodily movement? I do not think so.
Focus on the case of A, who decides that she shouldn’t take a chocolate
and so refrains from moving her arm toward the chocolate box. Is it true
that A doesn’t move her body? Put like that, the question is likely to
mislead. When A refrains from moving her arm toward the chocolate box,
there is no bodily movement of that kind: she does not move her arm
toward the chocolate box. But it hardly follows from the fact that there is
no bodily movement of that kind that A doesn’t move her body at all.
Remember what was said earlier. The standard story tells us that whenever
an agent acts, her action can be identified with some bodily movement or
other, a bodily movement that she knows how to perform, where her
knowledge how to perform that bodily movement is not explained by her
The Standard Story of Action (1) 49

knowledge how to do something else. The question is whether these condi-

tions are met in the case of the agent who refrains from moving her arm
toward the chocolate box. The mere fact that a certain sort of bodily move-
ment isn’t performed leaves that question unanswered.
When A refrains from moving her arm toward the chocolate box, she
clearly does exercise control over the way her body moves, because she
makes sure that it doesn’t move toward the chocolate box. Moreover,
she has no alternative but to do that in one of the ways available to her.
Perhaps she keeps her arm at her side, or she raises her arm in the air to
wave to a friend, or . . . and here follows a list of every possible way that
she knows how to control the location of her arm that was available to
her at the time she refrained from moving her arm toward the chocolate
box, except of course for those movements that result in her arm’s being
near the chocolate box. The reason she has no alternative but to refrain
from moving her arm in one of these ways is, moreover, obvious. For the
only way that A can keep her arm from moving toward the chocolate box
is by ensuring that it is somewhere else. Applying the tracing back idea,
we can therefore identify A’s refraining from moving her arm to the choco-
late box with whatever it was that she did with her body to ensure that it
was somewhere else.
The upshot is that when A refrains from moving her arm toward the
chocolate box, even though she does not perform that bodily movement,
she has no alternative but to refrain from doing so by moving her body
in some other way. The standard story thus applies to the case of omissions
after all. Hornsby is just wrong to suppose that there is a mismatch between
the standard story’s concern with agency quite generally and its concern
with action.

2 Second Objection: The Causal Account of Agency

Hornsby’s second objection targets the causal account of agency itself:

One may wonder . . . why casual claims like (SS) [Her desire . . . caused an event
which was her bodily movement], which are part of the standard story, should ever
have been made. For even where there is an event of the agent’s doing something,
its occurrence is surely not what gets explained. An action-explanation tells one
about the agent: one learns something about her that makes it understandable that
she should have done what she did. We don’t want to know (for example) why
there was an event of X’s offering aspirins to Y, nor why there was the actual event
of X’s offering aspirins to Y that there was. What we want to know is why X did
the thing she did—offer aspirins to Y, or whatever. When we are told that she did
50 M. Smith

it because she wanted to help in relieving Y’s headache, we learn what we wanted
to know. (Hornsby 2004, 8)

Hornsby’s objection seems to be that the standard story is an answer to a

question that never arises when we seek an action-explanation. It is an
answer to a question about the occurrence of an event, whereas an action-
explanation is never the answer to such a question.
But our requests for an explanation of why someone did something
plainly are sometimes requests for the cause of her doing that thing, where
her doing of that thing is conceived of as the occurrence of some event,
as should be clear from the earlier discussion of the two explanations we
might give of someone’s sitting in a chair. Let’s, however, consider a dif-
ferent example, one that I’m sure will strike a familiar chord with many,
just to drive the point home. When we lie in bed at night, I often feel my
wife kick me. There are typically three possible explanations: she’s asked
me to do something that I haven’t yet done, so she’s reminding me to do
it; I’m snoring and she wants me to stop; or she’s just fallen asleep and
she has kicked me as a result of some sort of bodily spasm. When I feel
her kick me, I ask myself why she did what she did. But am I seeking an
action-explanation?
The answer is that, in the first instance at any rate, I am asking a ques-
tion that is neutral with respect to whether or not what my wife did is an
action, and here it is irresistible to use the language of actions and events.
For though there certainly was an event of my wife’s kicking me, what I
want to know is whether that event was an action. This is, moreover, a
question to which the standard story supplies an answer. It tells us that
whether or not what my wife did was an action depends on what the cause
was of her doing what she did. She kicked me, but if she was asleep and
the cause of her doing what she did was a bodily spasm, then there is no
action, merely an event. However, if what prompted her doing what she
did was some desire she had, then the event in question is an action. The
reason we make claims like (SS) is thus because they allow us to distinguish
between these two cases, cases in which my wife’s kicking me was an action
and cases in which it was not.
Assuming that my wife’s kicking me was an action, I then want to know
which of the things she typically wants me to do explains why she kicked
me on this occasion. Hornsby is surely right that in answering this ques-
tion I am seeking information about my wife. But that should not now
seem inconsistent with the standard story’s suggestion that the answer will
also tell me about the causal etiology of my wife’s doing what she did. For
the answer will settle what it was that prompted her doing what she did
The Standard Story of Action (1) 51

among a narrower range of options. Was it her desire to remind me of

something, or was it her desire to stop my snoring? Once again, we seem
to be assuming that claims like (SS) are true.

3 Third Objection: The Irrelevance of History

Hornsby’s third objection is that it is a mistake to suppose, as those who

accept the standard story must, that the fact that an agent acted could be
constituted by facts about the states and events that figure in the causal
history of an action:

Bratman and Smith, when they raised questions about what it is for an agent of a
certain sort to be at work, turned these into questions about what sort of psycho-
logical cause is in operation. Like others who tell the standard story, they suppose
that citing states and events that cause a bodily movement carries the explanatory
force that might have been carried by mentioning the agent. But unless there is an
agent, who causes whatever it is that her action does, questions about action-
explanation do not even arise. An agent’s place in the story is apparent even before
anyone enquires into the history of the occasion. (Hornsby 2004, 19)

But what we just said in response to the third objection shows why
Hornsby is wrong to suppose that this is so.
Consider again my wife’s kicking me. Her place in this story as doer is
of course apparent before we inquire into the history of the occasion. But
this doesn’t suffice to secure her place in the story as the agent of anything
because, depending on how the story gets filled out, what she did may or
may not be an action. Moreover, the crucial information that fixes whether
or not what my wife did is an action is historical information. What
prompted her kicking me? This is a question about causation. If she was
asleep and her kicking me was caused by a bodily spasm, then she wasn’t
the agent of anything: she did something, but she didn’t act. But if her
kicking me was prompted not by some bodily spasm, but by a suitable
desire and belief, then she may well have been an agent. We simply cannot
ignore the history of what an agent did in determining whether she was
an agent. Her status as an agent is constituted by historical facts.

4 Fourth Objection: Agency as the Exercise of Capacities

Hornsby’s fourth objection is that a story about the causal history of an

action that mentions only states and events leaves out the crucial causal
factor:
52 M. Smith

[W]hen an account of a causal transaction in the case of agency is given in the claim
that a person’s believing something and a person’s desiring something causes that
person’s doing something, it is assumed that the whole of the causal story is told
in an action-explanation. The fact that the person exercised a capacity to bring
something about is then suppressed. It is forgotten that the agent’s causal part is
taken for granted as soon as she is said to have done something. The species of
causality that belongs with the relevant idea of a person’s exercising her capacities
is concealed. (Hornsby 2004, 22)

Hornsby begins with the charge that the standard story is incomplete, but
then ends with the charge that when you add to the standard story what
needs to be added to it to make it complete, you discover that the causality
required to make sense of agency—the agent’s exercise of her capacity to
do things—is different from anything that’s on offer in the standard story.
Let’s begin with the charge of incompleteness. As I understand it, the
standard story of action is offered as an account of necessary conditions
for agency, not an account of necessary and sufficient conditions. Indeed,
it is well known that defenders of the standard story have a very difficult
time of it when they attempt to say not just what’s necessary, but also
what is sufficient for agency (Davidson 1973; Peacocke 1979a,b; Sehon
2005). The following example is typical of those that give rise to the
problem. Imagine a piano player who wants to appear extremely nervous
when he plays the piano and who believes that he can do so by hitting a
C# when he should hit a C at a certain point in a performance. However,
when he gets to that part of the performance, the fact that he has that
desire and belief so unnerves him that he is overcome and involuntarily
hits a C#. In this case the piano player has a suitable desire and belief, and
these do indeed cause his hitting a C#, but his doing so is not an action.
The piano player is a patient, not an agent.
Defenders of the standard story who wish to provide necessary and suf-
ficient conditions for agency thus need to rule out the possibility of such
internal wayward causal chains, and this turns out to be no easy task. But
of course, defenders of the standard story aren’t obliged to provide neces-
sary and sufficient conditions for agency. They can rest content with the
more modest project of providing necessary conditions—or anyway, they
can do so provided those necessary conditions illuminate what it is to be
an agent. When the standard story is interpreted in this more modest
way, the charge of incompleteness as such just doesn’t seem to be an
objection.
So what is Hornsby’s objection? Her view seems to be that, if we are to
have any illumination of what it is to be an agent, we have to add to the
The Standard Story of Action (1) 53

standard story, so understood, the idea that agents who act exercise their
capacity to do things, an idea that cannot be made sense of in the standard
story’s own terms. There is, however, a certain irony in her offering this
as an objection to the standard story, because defenders of the story dis-
agree among themselves about the need to add an agent’s exercise of her
rational capacities as a distinct causal factor in an action explanation.
Hempel thought it is absolutely crucial that we mention the agent’s exer-
cise of her capacities; Davidson thought that Hempel was wrong and that
it is completely unnecessary (Hempel 1961; Davidson 1976). We must
therefore ask which of these two views is correct, and, if the correct view
is Hempel’s, we must ask whether we can make sense of the idea of an
agent’s exercise of her capacities within the resources available to a defender
of the standard story.
In this connection, it is instructive to consider how defenders of the
standard story attempt to rule out internal wayward causal chains. What
do they themselves think they need to add to it in order to provide a more
complete account of agency (note that they still needn’t think that it is
sufficient)? Davidson was pessimistic that internal wayward causal chains
could be ruled out in anything other than a completely uninformative
way. The best that we could say to rule them out, he thought, was that
the attitudes in question must cause actions in the right way (Davidson
1973). If this were the best we could say, then I would have sympathy with
the view that the standard story doesn’t shed too much illumination on
what it is to be an agent. For what Davidson thinks of as the right kind of
causation is presumably simply whatever it takes to underwrite agency.
On the other hand, others think it is clear what is required. They think
that the problem in cases of internal wayward causal chains is that the
match between what the agent does and the content of her desires and
beliefs is entirely fluky. It is, for example, entirely fluky that the piano
player wanted to hit just the note on the piano that his nerves subse-
quently caused him to hit. For a doing to be an action, they suggest, what
the agent does must be differentially sensitive to the contents of his desires
and beliefs (Peacocke 1979). The movement of an agent’s body is an action
only if, in addition to having been caused by a suitable belief-desire pair,
if the agent had had a range of desires and beliefs that differed ever so
slightly in their content from those he actually has, he would still have
acted appropriately.
In order to see how this differential sensitivity condition is supposed to
rule out internal wayward causal chains, consider once again the example
of the piano player. Suppose he had desired to play the piano as if
54 M. Smith

nervously and believed that he could do so by hitting a B at a certain point

in the performance, instead of a C. His actually hitting a C# is an action
only if, in this counterfactual scenario, he would have hit a B at that point
in the performance. Or suppose that he had desired to play the piano as
if nervously and believed that he could do so by maintaining a certain
rigidity as he hit the C. His actually hitting a C# is an action only if, in
this counterfactual scenario, he would have maintained a certain rigidity
while hitting the C. And so we could go on. This further condition of
nonflukiness, understood in terms of differential sensitivity, is plainly
violated in cases of internal wayward causal chains. For even if the actor
had had such ever so slightly different desires and beliefs, his nerves would
still have caused him to hit a C#.
But now consider the differential sensitivity requirement itself. What
does it guarantee? It guarantees that when an agent acts intentionally, he
doesn’t just try to realize the desires he actually has, given the means–end
beliefs he actually has, but would have tried to realize his desires, given
his means–end beliefs, in a range of nearby possible worlds in which he
had desires and means–end beliefs with ever so slightly different contents.
The requirement thus seems to guarantee nothing less than that someone
who acts intentionally does indeed have and exercise the capacity to be
instrumentally rational in at least a local domain (Smith forthcoming). The
possession and exercise of this capacity is, after all, exactly what’s required
to underwrite the truth of these counterfactuals.
This casts Hornsby’s fourth objection to the standard story in a rather
different light. Remember, Hornsby suggests that if we are to have an
account of agency at all, we have to add to the standard story the idea that
agents who act exercise their capacity to do things, an idea that cannot be
made sense of in the standard story’s own terms. It turns out that there is
a grain of truth in this. For in order to solve the problem of internal
wayward causal chains, the modest version of the standard story needs to
be supplemented with the differential sensitivity condition, a condition
that, when met, guarantees that those who act possess and exercise the
capacity to be instrumentally rational in a local domain. Hornsby is thus
right that those who act must possess and exercise this capacity, but she
is wrong that we are unable to make sense of this within the resources
available to a defender of the standard story. For the differential sensitivity
condition simply requires a modally rich pattern of regular causation by
desire and belief (see also Smith 2003a).
Of course, Hornbsy might think that the account of what it is to possess
and exercise the capacity for instrumental rationality that we glean from
The Standard Story of Action (1) 55

the differential sensitivity condition is mistaken. Or she might think that

something over and above the possession and exercise of this sort of capac-
ity is required to be an agent. But if that is her view, she will need to tell
us what is mistaken about it, or what else is required. Moreover, she will
need to explain why that capacity too can’t be made sense of with resources
available to a defender of the standard story. I myself am pessimistic about
her ability to do this.

Conclusion

Defenders of the standard story should welcome objections, as one of the

best ways they have available to them to demonstrate the plausibility of
their view is to work through and respond to such misgivings as people
have. I have considered four such objections put forward by Jennifer
Hornsby.
Three of Hornsby’s objections seem to me to be simply misplaced. Con-
trary to what she suggests, the standard story of action is a fully general
story of agency, as it applies equally to actions and omissions; the standard
story has a principled reason for talking about events in the context of
action-explanation, namely, that we can always raise the question whether
something that we do is an action or a mere event; and the standard story
therefore has a principled reason for inquiring into the history of our
doings, namely, that the status of our doings as actions or mere events
turns on facts about their causal history.
Hornsby’s fourth objection—that the standard story needs to be enriched
with an account of an agent’s exercising her capacity to do things—turns
out to be not so much an objection to the standard story itself as an expres-
sion of dissatisfaction with a modest telling of that story. Some defenders
of the standard story, such as Hempel, agree with Hornsby that an agent’s
exercise of her rational capacities plays a crucial causal role in action.
Moreover, that conclusion also seems to be forced on us when a defender
of the standard story attempts to solve the vexing problem of internal
wayward causal chains. Nothing about the standard story itself thus seems
to prevent a defender from agreeing with Hornsby that an agent’s exercise
of her capacities plays a crucial role in an action explanation.
4 The Standard Story of Action: An Exchange (2)

Jennifer Hornsby

In this second part of this exchange, I respond to Michael Smith by saying

why I think that no one should endorse the standard story of action in
any of its versions. I hope to show that there is an alternative to it.
In the first part of the exchange, Smith replies to Hornsby 2004. My
criticisms there were directed against philosophers (Smith among them)
who tell us that deficiencies in the standard story of action are to be rem-
edied by adding states of different sorts from beliefs and desires to the
causes of bodily movements (ibid., 2).1 I said that these philosophers fail
to address the question whether their story contains the causal notions
that belong in an account of human agency (23). Smith now says that I
charged the standard story with incompleteness. This strikes me as mis-
leading. True, I maintained that the causal role of agents goes missing in
an events-based conception of the causal order (16). But I never suggested
that the story stands in need of completion. I could hardly have suggested
this, given that my criticisms were directed against the idea that a correct
story could be got by adding to the standard story further conditions,
beyond those that it standardly specifies, for a case of agency. Indeed I said
that “supplements to the standard story . . . inherit its crucial flaw” (2).
The standard story of action is “events-based,” I said, and “events-based
accounts introduce a conception of the causal order in which agents have
no place” (3).
Smith grants me a “grain of truth.” In explicating the causal role of
agents, I said that in a case of human agency, a human being exercises a
capacity to bring something about. And Smith thinks that I am “right that
those who act must possess and exercise [such a] capacity.” Where I go
wrong, Smith says, is in thinking “that we are unable to make sense of this
within the resources available to a defender of the standard story.” In what
follows I hope to show that the game is up for the standard story when it
is allowed that a person’s acting is a matter of her exercising a capacity she
58 J. Hornsby

has as agent. I attempt this in section 3. Section 2 will be concerned with

the approach to an account of human agency taken by Smith and fellow
defenders of the standard story. If I can explain why it seems to me that
Smith’s understanding of his leading question “What makes it the case
that [someone] acted?” is out of place, then I shall have succeeded in
showing how one might think that the standard story should be aban-
doned, not “completed.”

1 Responses to Replies to Two Objections

Smith discerns four different objections that I made to the standard story.
I’ll make brief responses to Smith’s replies to the first two of these as a way
of introducing the main business. (Sections 2 and 3 are in effect responses
to his replies to the third and fourth objections.)
(1) When I said that the standard story fails to accommodate omissions
of certain sorts, I was thinking of the story, as Smith thinks of it, as pur-
porting to give an account of what it is for A to do something intentionally
by saying what conditions a movement of A’s body must satisfy if it is to be
A’s Φ-intentionally doing some particular thing. My point was that
someone may do something intentionally although no movement of her
body occurs. Now Smith tells us that the standard story makes use of a
notion of moving the body according to which “any orientation of the body
counts as a bodily movement,” and “move the body” has application when
someone refrains from moving. I am sympathetic to an idea of Smith’s which
is in play here—namely, that there can be an exercise of a piece of bodily
agential know-how even when there is no overt movement. So I allow that
there can be a rationale for a capacious notion of moving the body, a bit
like Smith’s. Still, it isn’t clear that the know-how that is exercised by an
agent who intentionally does something can always be brought within the
scope of the bodily, or that an agent’s doing what she does always belongs
in the category of event. One of the examples I gave was of C, who delib-
erately paid someone no attention. Suppose that there was a period of an
hour during which C paid no attention to X (it might be a period during
which X was in the same room as C). Smith, it seems, will have to say that
throughout the period, there was an exercise of a piece of bodily agential
know-how affecting the way the agent’s body was oriented.2 That doesn’t
sound quite right to me.
However this may be, I had a particular reason for introducing examples
in which an agent’s intentionally doing something appears not to be a
movement of her body. I wanted to show the impossibility of accommo-
The Standard Story of Action (2) 59

dating agency to the operation of causality as that is conceived by those

who tell the standard story. I don’t deny that a causal explanation is given
when someone does something for a reason and the question why she did
it says what she thought and/or wanted. My quarrel was with the idea that
such an explanation can always be recorded as the obtaining of a causal
relation between states and an event.
(2) My second objection then was to the general idea that explanations
that reveal a reason why someone did something can be recast simply as
causal accounts of why events occurred. I drew attention to the distinctive
character of reasons-explanations and the nature of what they convey. I
did so in order to weaken the impression, given by those who tell the
standard story, that the constituents of reasons-explanations are just those
of explanations of other sorts. Smith says that causal explanations of any
sort speak to “the causal etiology” of what is explained. I can agree with
that. My claim was that when one is told of an agent’s reason, that which
gets explained is not the occurrence of an event merely, but the fact of an
agent’s playing a certain causal role. In my own view, an agent’s playing
such a role should not be assimilated to an event of any old sort, and nor
therefore should it be taken to be an event that has the sort of etiology
that just any event has.

2 The Approach of Those Who Tell the Standard Story

2.1
On the standard approach, which gives rise to the standard story,3 actions
are taken to be events and treated then as having equal standing with
anything else that belongs in the event-causal order, in which causation
relates things in the category of events and states. On this approach, the
task of a philosophical account of bodily agency is to uncover the condi-
tions that bodily movements satisfy if and only if they are actions.
Inasmuch as these conditions are thought to be a matter of how the move-
ments are caused, “what makes it the case that” or “fixes” it that an
event is “an action rather than a mere happening in which a person is
involved” is treated as “historical information.” I think that the standard
approach is based in a false assumption. And I hope to make it clear
why I think the standard story should be abandoned by locating that
assumption now. (Smith points out that some who tell the standard story
are content if they can carry out only a more modest task. I shall come to
this. The present point is that I reject the assumption that governs the
approach.)
60 J. Hornsby

When Smith says that “the crucial information that fixes whether or
not [a bodily movement] is an action is historical information,” he is
objecting to something I had said. I had said that “unless there is an agent,
who causes whatever it is that her action causes, questions about action-
explanation do not even arise.” What I intended to convey was that his-
torical causal information cannot fix whether a movement is an action.
When causal information can be given in an action explanation, it is then,
as it were, already fixed that there is an action; whereas when a movement
of someone’s person’s body is their involvement in some mere happening,
the movement is then not so much as a candidate for an action-explana-
tion, and nothing could fix that it was an action.
Smith’s counter to this is his example of his wife’s kicking him. The
example is meant to show how historical information fixes things. Accord-
ing to Smith, a particular event of his wife’s kicking him may, or may not,
be an action, depending on its causal history. Smith is surely right in think-
ing that if he asks why she kicked him, not knowing whether her kick was
an action or not, he remains neutral on the question of whether it was.
But in order to see why this is not definitive in settling that an event may,
or may not, be an action, one can consider a case where Smith assumes
that his wife meant to kick him, and gets it wrong. So imagine that Smith
asks “Why did you kick me?”—thinking that his wife’s answer will give
her reason. He expects her to say “I wanted to ——” or “In order to ——.”
Instead she says “Oh, I didn’t mean to kick you: my kick must have been
a result of some sort of bodily spasm.” In this case, we may think that what
Smith was apt to treat as a candidate for an action-explanation wasn’t
really such a candidate; Smith’s mistake was to suppose that the event of
kicking was of a different kind from what it was. (It was of the generic kind
“bodily movement.” But if it was actually not an action, then at least it
was of some kind that he assumed it wasn’t.) So although there can be an
example of the sort Smith describes, in which he fails to know whether or
not his wife meant to kick him, this can hardly show that the event about
which he is then ignorant might itself equally well be an action or not be
an action.4 If it is not an action, then of course it fails to satisfy such con-
ditions as actions do, and Smith may come to know that it was not an
action by learning that such conditions weren’t satisfied. But it can still be
true that the event could not really satisfy such conditions—that nothing
could make it an action. And it can also be true that if his wife had meant
to kick him—had kicked him for a reason—then the kick that there would
then have been could not itself have been the product of a spasm. If these
The Standard Story of Action (2) 61

things are indeed true, then actions and movements that aren’t actions are
of fundamentally different kinds.5
On the standard approach, it is assumed that actions and other bodily
movements are of the same fundamental kind, differing from one another
in their relational properties. Taking the standard approach, one thinks it
possible to find in the event causal order a movement that might or might
not be an action, and then, by making use of a notion of causation that
relates things in the categories of event and state, to say how the move-
ment differs, in respect of its relational properties, from other things in
that order if it is an action. But if actions and other movements are of
fundamentally different kinds, then an action in its nature is not an event
about which it is genuinely an open question whether it is an action or a
movement of some other sort. (Remember that ignorance can ensure that
there seems to be an open question, even if, in the nature of things, one
answer is actually ruled out.)

2.2
We can see now that one may find fault with the standard approach to
action even if it leads only to the relatively unambitious account that
Davidson gave. Davidson asked “What events in the life of a person reveal
agency; what are his deeds and his doings in contrast to mere happenings
in his history; what is the mark that distinguishes his action?” (1971/1980,
43). And it is no wonder that the standard story is widely credited to
Davidson. He always thought that the key to understanding agency was
to take actions to be events, caused by states and other events and states,
and described in terms of states and further events that they in turn cause.
Although Davidson despaired of giving a set of conditions sufficient for an
event’s revealing agency, he never doubted that some events are actions
only because some of the bodily movements that belong in chains in the
event-causal order have a causal history of a certain sort.6
What Davidson despaired of doing was solving “the problem of wayward
causal chains” (see Smith’s chapter in this volume for the details)—a
problem that needs to be solved by someone who takes the standard
approach and aspires to give sufficient conditions for an event’s being an
action. Smith for his part thinks that he has a solution to the problem. I
shall come to this (section 3). But what I would draw attention to now is
an aspect of Smith’s own argument which might be thought to suggest
that there really need be no problem here in need of a solution. When
Smith speaks of the answer to the question of why his wife kicked him as
62 J. Hornsby

fixing whether there was an action, he appears to take susceptibility to a

certain sort of explanation to guarantee that a movement is an action. At
this point in Smith’s argument, the idea that there might be wayward
causation of his wife’s movement (that it might be caused by a belief and
desire having the sort of contents that could but actually don’t rationalize
it) is not in the picture. At this point, Smith would seem to be content to
allow that his wife’s kicking him is an action if and only if she kicked him
for a reason. But then Smith would seem to have helped himself to the
idea that actions and action-explanations go hand in hand—that all and
only actions are events the facts of whose occurrences can be explained by
giving the agent’s reasons.7
Of course, there can be no harm in helping oneself to this idea. Indeed
when action is connected with a particular sort of explanation, one has an
answer to Davidson’s question “What is the mark of an action?” For the
mark of a kind may be given simply by saying what is distinctive of its
members. Thus one might ask, about animals, “Among those that are
either stoats or weasels, what is the mark of a stoat?” An acceptable answer
could be “Stoats have black tails.” One can give this answer, without think-
ing that stoats might be weasels, or that any animal that is a stoat rather
than a weasel is only such because of the blackness of its tail. In a rather
similar way, a philosopher can ask, about events, “Among those that are
either actions or mere movements, what is the mark of an action?” The
answer may be: “Actions are susceptible to a certain sort of explanation.”
The answer tells of a respect in which actions are distinctive. But in giving
this answer, one is not committed to thinking that actions might be move-
ments of some other kind than actions, or that anything that is an action
rather than a movement of some other kind is only such because it has
certain sort of explanation.
The fact, then, is that one can give a mark of an action (in a sense just
roughly indicated) without buying into the assumption of the standard
approach. And that means that one can abandon the standard approach
without abandoning the idea which seems to inform Smith’s treatment of
his example, and whose endorsement informs the standard story as it is
told in any of its versions. The idea is that when and only when there is
an action, an agent does something for a reason he or she has. This was
Davidson’s idea, and he stuck with it. When he said that a movement is
caused by relevant attitudes in the right way if it is an action, what he needs
to have meant by “caused in the right way” is this: being such that the
fact of its occurrence can be causally explained by saying what the agent’s
reason was for making the movement.8
The Standard Story of Action (2) 63

Smith for his part thinks that the standard story in Davidson’s version
“doesn’t shed too much illumination on what it is to be an agent.” One
can agree with that, and think that further illumination can be shed—that
there is more to be said. What we are now in a position to see is that one
may have one or another of two very different reasons for thinking that
there is more to be said. (A) One may tell the standard story, and think,
as Smith does, that it fails of sufficiency because it lacks a specifiable neces-
sary condition of a case of agency. (B) One may reject the standard
approach, as I do, but not stop at giving the mark of actions, thinking that
there is more to be said about the sorts of causal notions that are in play
when there are actions. I suspect that Smith does not countenance alterna-
tive (B) because he cannot credit that someone might put the standard
approach into question. (This would explain why he should have thought
that I must have been charging the standard story with incompleteness
when I claimed that something goes missing when it is told.)

3 Exercises of Capacities

Smith’s alternative is (A). He wants to improve on Davidson’s telling of

the standard story: he thinks that the standard story will be complete, and
more illuminating, if another ingredient is introduced into it. So I turn
now to what Smith takes to be the missing ingredient: I come to agents’
capacities.
It was because I spoke myself of agent’s capacities as belonging in an
account of agency that Smith found a grain of truth in what I said. He
writes: “Nothing about the standard story itself . . . seems to prevent a
defender from agreeing with Hornsby that an agent’s exercise of her capaci-
ties plays a crucial role in an action-explanation.” But I did not say that
agents’ exercises of capacities play a crucial role in action-explanation. I
contested the assumption of the standard story that “the causal story is
told in an action-explanation” (Hornsby 2004, 22), maintaining that there
is more to a causal story of agency than action-explanations themselves
can reveal (20). What I said was that when an agential capacity is exercised,
an agent, rather than a state of an agent, plays a causal role; and that “our
actions [should] be thought of as exercises of our capacities” (21). In my
alternative story, the exercises of capacities have a crucial place because I
hold that an agent’s doing something that she intentionally does is her
exercising a capacity she has as a rational agent.
Smith thinks that exercises of capacities, so far from being actions,
are among their causes. It can seem that Smith and I have different
64 J. Hornsby

conceptions of a capacity. Certainly I am suspicious of Smith’s conception.

Smith speaks of a movement “that A knows how to perform, where her
knowledge how to perform it is not explained by her knowledge how to
do something else.” Well, I take it that when one has knowledge how to
do something and suffers no relevant impediment, one has the capacity
to do it. And a capacity is surely something general: even a simple capacity,
such as the capacity to raise one’s arm, is something whose exercise might
be constituted by movements of various different, specific sorts. But any
movement that an agent “performs” is a particular movement, of some
utterly specific soft. The problem then is that possession of the capacity to
raise one’s arm doesn’t seem to explain a movement in the way Smith
supposes that it does. When knowledge how to raise one’s arm belongs in
an account of someone’s having raised his arm ten inches on some occa-
sion, the knowledge in question could explain his doing something else—
raise his arm fourteen inches, as it might be.
However this may be, it is when Smith says that “an agent’s exercise of
her rational capacities is a distinct causal factor in an action-explanation”
that I find I take exception. My own (and I think the ordinary) conception
of a capacity of X is a sort of causal power of X, something, then, whose
exercise consists in X’s bringing something about. Thus, if you have the
capacity to raise your arm for a reason, you are so equipped that you can
bring it about that your arm is up when you have a reason to raise it. And
when, on occasion, you exercise this capacity, you bring it about on that
occasion that your arm is up. Your capacity to raise your arm provides for
your playing the role of cause, and any exercise of the capacity on any
occasion will be your raising your arm, will be your bringing it about that
your arm is up. Your possession of the capacity is a condition of its exercise,
and it might in some circumstances explain an actual exercise. But it is
difficult to see how your exercise of the capacity to raise your arm could be
a “distinct causal factor” in an explanation of your raising it.
It may be, however, that Smith and I disagree not so much about the
nature of capacities and their relation to their exercises, as about the kinds
of capacities that need to be brought into account. Smith may be con-
cerned exclusively with capacities “to be instrumentally rational in a local
domain,” which capacities he thinks are sure to be present so long as the
differential sensitivity condition is met (again see Smith’s chapter for the
details). But if we are allowed to think of the differential sensitivity condi-
tion as something that is met whenever an agent’s doing something is
explicable by reference to her reasons, then the condition would seem to
be satisfied so long as the agent has rational control over what she is doing.
The Standard Story of Action (2) 65

But then, once again, it would appear to be the sort of capacity whose
exercise just is her acting intentionally.
Suppose that this is right. And suppose that Smith, persisting with the
standard approach, says that the idea of exercising a capacity possessed by
a rational agent belongs in the standard story. He will be taken round in
an evident circle when he tells the story. For if possession of a rational
capacity is required for there to be an action, then exercises of rational
capacities will be the actions of rational beings. But then something’s being
an exercise of such a capacity will be matter of its being an action. Making
use of the idea of such a capacity could not enable one to give a necessary
condition that combines with the old standard story’s other conditions to
say what else is causally required for there to be an action.
I hope that this starts to explain why I should think that the game is
up for the standard story when capacities are introduced. Just as actions
and action-explanations go hand in hand, so it would seem that actions
and exercises of certain capacities go hand in hand. An idea of an agential
capacity, like the idea of a certain sort of explanation, is presupposed to
the idea of an action. Being the exercise of a capacity cannot then be one
among a set of conditions of a movement’s being an action. It cannot
“make” an event that might not have been an action into an action.

4 Summary and Unfinished Business

Those who take the standard approach in philosophy of action seek condi-
tions of a movement’s being an action. They think that it is proper to ask,
about a movement that, according to them, may or may not be an action:
“What makes it an action?” Being aware that actions are susceptible to a
certain sort of explanation, and that this is causal explanation, they treat
causation by rationalizing beliefs and desires as one condition of a move-
ment’s being an action. But then they face the problem of wayward causal
chains. They react either by settling for speaking of such causation as
occurring in the right way (as Davidson did), or by seeking a further condi-
tion of an event’s being an action (as Smith does). What I have suggested
is that if one disallows their assumption that the question “What fixes it
that an event an action?” is in good order, then one will say that when
there is an action, there is an event the fact of whose occurrence is expli-
cable in a certain way and does not just so happen to be explicable in that
way. Actions are not then thought of as movements that count as actions
by virtue of having relational characteristics. Actions may be thought to
involve intentions essentially.
66 J. Hornsby

The question “What makes it the case [or fixes it] that an event is some-
one’s acting and not her involvement in some mere happening?” has
become very familiar. And the fact that it is possible to say what distin-
guishes actions from other events can help it to seem to be a perfectly good
question. (We saw that there is a perfectly good question about what dis-
tinguishes actions, which makes use of one idea of a mark.) It can then
become hard to believe that the question intended by those who tell the
standard story is misguided. But it is not outlandish to suggest that phi-
losophers have sometimes asked the wrong questions.
I’m aware that I have not done much more than quarrel with Smith’s
interpretation of Hornsby 2004. But my aim in this part of our exchange
has been not so much to argue for an alternative to the standard story, as
to demonstrate that there can be one. So I have done little here in the way
of showing why one should accept an alternative. To persuade someone
of an alternative of the sort I favor would require much more work. One
needs to demonstrate the counterintuitive character of the conclusions
reached if one follows Hume on the subject of causation. (That was a task
I did attempt in Hornsby 2004.) One needs also to say something positive
about the kind of causation by agents that is irreducible to causation found
in the event causal order. Inasmuch as such causation may be causation
by agents other than rational agents, one must also say something about
what is distinctive of the capacities that rational beings possess. The fact
that these tasks can be undertaken helps to show that even if one has no
truck with the project set by those who take the standard approach, still
one has plenty to do to try to cast light on the phenomenon of human
agency.

Acknowledgments

I thank Naomi Goulder for her helpful comments on a draft of this chapter.

Notes

1. Otherwise unattributed page references are to Hornsby 2004. My objections there

were directed specifically against philosophers who had said that the standard story
needs to be supplemented if it is to include the agency of ethical beings. But I blamed
its failure to include ethical agency on its failure to include human agents.

2. Smith might suggest that in the example of C, we find, as well as the bodily
actions that the standard story purports to provide an account of, mental actions;
and then that there can be intervals during the period when C pays no attention
The Standard Story of Action (2) 67

to X at which even the orientation of C’s body is not to the point. My response
would be to say that although there may be a recognizable category of mental
actions, I doubt that we can plausibly decompose intuitively physical actions into
the mental and the bodily in the manner that such a suggestion would seem to
require. Consider, say, drinking a cup of coffee, or speaking, which admit of pauses
between bits of bodily activity.

3. I don’t know who coined the term “standard story”: perhaps it was David Velle-
man (who uses it in Velleman 1992). “Standard approach” is used in Anscombe
1989/2005 (a paper she wrote in 1974), and I want to mean by it what I think
Anscombe meant. At any rate, I think it safe to assume that the standard story, as
it is now commonly understood, is a product of the approach that Anscombe called
standard (Anscombe 2005, 111), and which she thought was hopeless.

4. I realize that I could seem to have denied the possibility of such examples in
Hornsby 2004. Following the sentence above which I’ve followed Smith in quoting,
and whose force I am now attempting to explain, came another sentence, which
Smith also quotes: “An agent’s place in the story is apparent even before anyone
enquires into the history of the occasion.” This was careless. I should not have sug-
gested that the agent’s place must always be apparent, at least if what is apparent
is known. Nevertheless, I take it that the agent’s place very often is apparent—that
very often someone is evidently doing something intentionally, even if one knows
not exactly what—and then a request for explanation takes for granted what is then
apparent.

5. This is the position of someone who endorses a so-called disjunctivism about

bodily movements: see Hornsby 1997, and Haddock 2005. Haddock is the better
source for present purposes, because he understands bodily movements to be (not
effects of actions but) actions themselves; and this understanding is one that I now
agree with Smith in endorsing (though didn’t in Hornsby 2004). In Hornsby 1997,
I distinguished epistemological and metaphysical senses of “could be a mere [i.e.,
nonactional] movement.” Here I have taken “fix” to mean what Smith needs it to
mean, and pointed out that one may be ignorant of something that actually is fixed.
If “fix” has an epistemic sense, then my point is that something fixed metaphysically
may not be fixed epistemically.

6. Davidson is the target of Anscombe’s criticism of ‘the standard approach’ (see

note 3 above). She says that Davidson “can do no more than postulate a ‘right’
[= nonwayward] causal connexion [between causes of actions and actions] in the
happy security that none such can be found” (Anscombe 1989/2005, 110). Those
who tell the standard story are apt to respond “Does Anscombe not know that
Davidson had said that ‘the request for an analysis of action goes unanswered,’ so
that he speaks of ‘right’ causal connections because he thinks it impossible to give
sufficient conditions?” I think that this response to Anscombe fails to take account
of the fact that in her view a question based on a false assumption is asked when
68 J. Hornsby

an analysis of action is requested (see further note 8). So I think that the critics of
Anscombe who take the standard approach interpret her rather as Smith has inter-
preted me—as if I thought that the standard story were told in response to a good
question.

7. I say “facts of whose occurrence can be explained” here because if I were to say
simply “can be explained” it might give the impression that the explanandum of a
reasons-explanation is an event—a false impression, but one that will be welcome
to those who take the standard approach (see para. (2) in section 1 above). Still, at
various points, I forsake accuracy in order to avoid prolixity, and sometimes speak
of actions as being explicable in a certain way. Treat this as a sort of shorthand.

8. When Anscombe speaks of Davidson as able to “do no more than postulate a

right [= nonwayward] causal connection” (see note 6 above), we might take her to
be saying that insofar as Davidson has found a mark of agency (in the sense roughly
explained in the text), the only kind of causal connection that could be present is
that which is present when an event has a certain sort of explanation. Hence her
saying “Something I do is not made into an intentional action by being caused by
a belief and desire” (Anscombe 1989/2005, 111; my italics).
5 Skepticism about Natural Agency and the Causal Theory
of Action

John Bishop

Action theorists often proceed with little or no motivational prolegomena,

as if the question of what it is for something to count as an action is just
there, most strikingly posed, perhaps, in Wittgenstein’s terms: “What is left
over if I subtract the fact that my arm goes up from the fact that I raise
my arm?” (Wittgenstein 1972, 161).1 Alfred Mele, for example, in the
introduction to his edited Oxford Readings in Philosophy collection in the
philosophy of action, begins with the twin questions (a) what are actions?
and (b) what is involved in explaining actions? And he says that “inquiring
minds . . . would want to know what actions are, what it is to act intention-
ally, what constitutes acting for a reason, whether proper explanations are
causal explanations or explanations of some other kind, and how the
psychological or mental items (states, events and processes) that are sup-
posed to be explanatory of action are to be understood” (Mele 1997b, 1,
my emphasis). Admittedly, Mele does remark that “one hopes that a full-
blown philosophy of action will solve part of the mind–body problem and
illuminate the issues of free will, moral responsibility, and practical ratio-
nality” (ibid.). So it is, of course, clear that inquiring into the nature of
actions and action explanation does have wider motivation. In my view,
however, it is helpful for action theorists to be more explicit about what
their wider motivations are, and to keep their theorizing consciously in
contact with wider philosophical goals. Debates about action can become
unfocused unless they are consciously answerable to some specific motiva-
tion or motivations for philosophical interest in action.

1 The Problem of Natural Agency

One such motivation is concern with the problem of natural agency. Witt-
genstein’s question may indeed be a focal one: what is it for behavior to
count as action, rather than just “mere” behavior? But this question already
70 J. Bishop

deploys a technical notion of action—a technical notion that has its source
in our ethical perspective on ourselves. For we hold an agent morally
responsible for a given outcome only if it came about or persisted through
that agent’s own action. Even if an agent’s behavior contributed to a certain
outcome, the agent would not be morally responsible for it unless the
behavior involved or constituted the agent’s own action.2,3 Now, if our
ethical perspective applies to the world (if agents really are sometimes
morally responsible for outcomes) then actions must be a feature of the
world. But how is it possible for actions in this sense to be part of the
natural world as our natural scientific worldview conceives it? Actions
either are or necessarily involve physical events—in particular, bodily
movements—and those events are open to natural scientific explanation.
So there is room for skepticism about how the very same outcome can
both result from morally responsible agency and also be in principle expli-
cable scientifically. In previous work, I put the problem of natural agency
thus: “We seem committed to two perspectives on human behavior—the
ethical and the natural—yet the two can be put in tension with one
another—so seriously in tension, in fact, as to convince some philosophers
either that the acting person is not part of the natural order open to sci-
entific inquiry or that morally responsible natural agency is an illusion”
(Bishop 1989, 15). One important motivation, then, for seeking to under-
stand what it is for something to be an action is to seek to resolve the
problem of natural agency. Does the concept of action that we need for
our ethical perspective have features that require going beyond the con-
fines of our prevailing natural scientific metaphysics, or may actions be
wholly accommodated within a naturalist ontology?
I shall call the claim that actions can be fitted into a naturalist ontology
reconciliatory naturalism. One way to defend reconciliatory naturalism is to
advance a causal theory of action. I suspect that this is the only route to
reconciliatory naturalism—but will here claim just that a certain sort of
causal theory of action, if successful, overcomes skepticism about natural
agency. To appreciate why a successful causal theory of action would
achieve this, consider how best to state the nature of the tension between
our ethical and natural scientific perspectives. That tension has often been
expressed as an apparent incompatibility between free will and determin-
ism—but the real problem is not really about determinism, as Gary Watson,
for one, makes very clear:

We would lack free will in a deterministic world, incompatibilists think, because

there we would not determine our own behaviour—we would lack self-determina-
tion. But the denial of determinism does not ensure, by itself, that we determine
Skepticism about Natural Agency and the CTA 71

anything. So an incompatibilist who affirms freedom—conventionally called a lib-

ertarian—must say what more is needed besides the absence of causal determination
to get “self-determination.” The . . . skeptic . . . suspects that the freedom demanded
by the [libertarian]—a “self-determination” that could not obtain in a deterministic
world—obtains in no possible world. (Watson 1982, 9)

So the real problem is that actions are understood as essentially involving

agent-causation (where I use that term simply to mean the determination
of events or outcomes by the agent through the exercise of that agent’s own
control), yet this is a kind of substance-causation excluded by the fun-
damental ontology of the naturalist perspective on the world, whether
it takes the world to be deterministic or, as is now more plausible,
indeterministic.4
If a causal theory of action is correct, however, then this problem is
resolved, because a causal theory of action (or, to name it more precisely,
an event-causal theory of action) maintains that episodes of agent-causa-
tion can properly be understood as constituted wholly by events and states of
affairs within the natural causal order, as the prevailing natural scientific
metaphysics interprets it. For a causal theory of action (CTA) makes the
positive ontological claim that actions consist in behavior caused by
mental states and events that constitute the agent’s having his or her reasons
for performing behavior of that very kind. A CTA thus implies that inten-
tional explanations (explanations in terms of an agent’s reasons for acting),
though they are to be given a realist interpretation,5 do not involve irre-
ducible ontological commitment to agent-causation, but only to event-
causation of behavior by the agent’s relevant mental states and events. If
a CTA is true there can be no mystery about how actions belong to the
natural causal order, provided—as I will here simply assume—there is no
mystery about how mental states and events may do so.6 But the CTA will
have to be defended against skeptics of the kind Watson mentions, and
this will require defeating the arguments of agent-causationists who insist
that genuine actions must involve an ontologically irreducible special kind
of causal relation where the cause is a substance, namely, the agent.
Before turning to consider the defense of a CTA, let me say something
on the question of motivation. A successful CTA will resolve the problem
of natural agency: but why be so concerned to resolve that problem—why
should it matter if the presuppositions of our ethical perspective turn out
not to fit with the prevailing natural scientific metaphysics? What would
be so bad about having to conclude that the commitments of our ethical
perspective outrun anything that a naturalist metaphysics contemplates?
There is a value judgment here—that, somehow, it would be “better for
72 J. Bishop

reconciliatory naturalism to be true than for it to turn out either that our
belief in [the reality of] agency is mistaken, or that, as agents, we belong
mysteriously beyond the natural universe that is open to scientific inquiry”
(Bishop 1989, 5). That it would be good for reconciliatory naturalism to
be true, while not, of course, counting as any kind of evidence that it is
true, might nevertheless justify our taking it to be true—though only if our
best assessment of the relevant arguments and evidence leaves its truth
open.7 But the fact that agency apparently involves a special kind of
agent-causation not admissible in fundamental naturalist ontology sug-
gests that its truth is not left open—unless a CTA can be successfully
defended, that is.
But why might one think it good that reconciliatory naturalism be
true—and so have a motivation for defending a CTA against agent-causa-
tionist skeptics about natural agency? One would need to be committed
both to the reality of human agency and to the idea that human existence
belongs wholly within the natural order. Such commitment might itself
be just intrinsic to a naturalist worldview. It might, however, be derived,
as in my own case, from the theistic religious traditions that emphasize
human freedom and responsibility but at the same time affirm human
creatureliness. We may indeed be creatures in the image of God, but we
remain creatures nonetheless. There is thus a tension between our crea-
turely dependency and our power to act, which I think is properly resolved
by reconciliatory naturalism (by contrast with the philosophical libertari-
anism to which many theists are attracted, under which, it seems to me,
it is hard to avoid the view that the human self acts from outside the
natural causal order).8 So I actually have a religious motive for defending
a reconciliatory, or “compatibilist,” naturalism, and, therefore, a suitably
formulated causal theory of action. But such a nonevidential motivation
is subject to philosophical correction, and would need to be set aside if a
reconciling account of natural agency ran into insuperable difficulty.

2 Davidson’s Defense of a Causal Theory of Action

Donald Davidson is the major figure in the defense of a CTA, and he effec-
tively endorses the view that a successful CTA resolves the problem of
natural agency. In his essay “Intending,” he remarks that “the ontological
reduction [implied by CTA], if it succeeds, is enough to answer many
puzzles about the relation between the mind and the body, and to explain
the possibility of autonomous action in a world of causality” (Davidson
1978/1980, 88, my emphasis). Indeed, this ontological reduction is just
Skepticism about Natural Agency and the CTA 73

what Davidson proposed in his most famous article, “Actions, Reasons,

and Causes,” where he says, in conclusion: “Some causes have no agents.
Among these agentless causes are the states and changes of state in persons
which, because they are reasons as well as causes, constitute certain events
free and intentional actions” (Davidson 1963/1980, 19).
Davidson did, of course, develop his version of CTA in papers following
“Actions, Reasons, and Causes.” Under his CTA, actions have to be inten-
tional under some description, and to count as such they must have
“rational” causes, in the sense of causes that also make reasonable what
they cause, at least minimally from the agent’s point of view. But, plainly
enough, agents sometimes act in ways that are irrational, and that they
themselves recognize as irrational, while yet acting intentionally. Akratic,
or weak-willed, actions are a major case in point, and Davidson is con-
cerned to defuse this potential counterexample to his CTA. Reading “How
Is Weakness of the Will Possible?” (first published in 1970) in the retrospec-
tive light of “Intending” (first published in 1978) allows one to see that it
is crucial to Davidson’s strategy to acknowledge that intention is sui generis
(rather than syncategorematic, as argued in “Actions, Reasons, and
Causes”). An akratic intentional action is reasonable with respect to the
proximate intention that causes it, but unreasonable in relation to the
agent’s “all things considered” judgment about how to act.9 Vulnerability
to weakness of will is thus the tendency to be caused to have intentions
contrary to one’s deliberative conclusions—for example, through “interfer-
ing” emotions, or through certain considerations turning out to have more
causal power than rational force. This solution to the problem of akrasia
does involve a development of the CTA from the account first offered in
“Actions, Reasons, and Causes.” It allows that the minimally rational cause
of an action may be no more than an intention to perform an action of
that kind. And that development may also be used to meet other chal-
lenges—for example, Rosalind Hursthouse’s argument from “arational
actions” (Hursthouse 1991). Someone who hurls a recalcitrant tin opener
out of the window in frustration is not, of course, acting to serve some
further end: but she is acting intentionally, and a CTA can accommodate
this by observing that, in such a case, emotion directly causes an intention,
which is then carried out. The CTA’s ontological reduction is preserved in
such cases, as in cases of weakness of will: events are constituted as actions
by having mental causes that are rational causes, but the mental causes
here are no more than intentions to perform the relevant type of action.
But there is another, potentially serious, threat to the CTA that emerges
in Davidson’s “Freedom to Act” (first published in 1973): the problem of
74 J. Bishop

causal deviance. An agent’s behavior can have mental causes that make it
reasonable and yet not count as the relevant intentional action—or,
indeed, as any kind of intentional action at all. Davidson’s nervous climber
illustrates the point: “A climber might want to rid himself of the weight
and danger of holding another man on a rope, and he might know that
by loosening his hold on the rope he could rid himself of the weight and
danger. This belief and want might so unnerve him as to cause him to
loosen his hold, and yet it might be the case that he never chose to loosen
his hold, nor did he do it intentionally” (Davidson 1973/1980, 79). For
intentional action, behavior has to be caused in the right sort of way by
mental states and events that make it reasonable, yet, in “Freedom to Act,”
Davidson reports that he “despair[s] of spelling out” what that right sort
of way is (ibid.). But surely, not being able to spell this out will be embar-
rassing for a CTA defender, if the CTA is to secure the possibility of natural
agency. Roderick Chisholm was right, I think, to see in the possibility of
deviant counterexamples a major challenge to a CTA-based analysis of
action, and hence a potential argument in favor of agent-causationism
(Chisholm 1966). No wonder, the agent-causationists will say, that you
run into problems in trying to reduce an agent’s exercise of control to
causation by that agent’s mental states. For, arguably, all that can secure
“nondeviant” causation is to bring the agent back into the picture: the
climber would have acted intentionally only if he had brought it about that
he relaxed his grip, and that has to be understood as a relation between
him, as agent, and the relevant bodily movement.
When he returned to consider problems in the explanation of action in
an essay with that title first published in 1987, Davidson had this to say
on the problem of deviance, and I know of no evidence that he ever
changed his mind: “Several clever philosophers [he mentions Armstrong
and Peacocke in a footnote] have tried to show how to eliminate the
deviant causal chains, but I remain convinced that the concepts of event,
cause and intention are inadequate to account for intentional action”
(Davidson 1987/2004, 106).10 Yet the CTA, surely, is precisely the thesis
that “the concepts of event, cause and intention” are adequate “to account
for intentional action”? Or, at least—and this is a vital point—the CTA is
the thesis that the concepts of event, cause, and intention are adequate to
provide a suitably naturalistic ontological account of what constitutes inten-
tional action even if they do not provide a conceptual definition of what it
is to act with an intention. Did Davidson in effect concede, then, that the
ontological reduction offered by the CTA could not ultimately be defended
against the challenge posed by the argument from the possibility of causal
deviance? No. I do not think so. It looks rather as if Davidson thought
Skepticism about Natural Agency and the CTA 75

that the ontological reduction would be secure, even though the deviant
cases were enough to put paid to any hope of conceptual, definitional,
analysis.11
But is that correct? Agent-causationists will, of course, agree that the
deviant cases show the impossibility of defining an agent’s acting inten-
tionally as her being caused to behave reasonably by her own relevant
mental states, but surely they may argue that the deviant cases show
further that event-causation by her own mental states could not even
constitute the agent’s intentional action.
So there is something of a puzzle as to why Davidson was not more
concerned about resolving the problem of causal deviance and appears to
have continued to believe in the CTA’s ontological reduction (and its force
in resolving the problem of natural agency) even though he himself
believed that, despite sophisticated efforts by some philosophers, the devi-
ance problem could not be resolved. It seems as if he thought that this
problem was somehow peripheral, and some philosophers have apparently
agreed. Berent Enç, for instance, speaks of the “somewhat technical
problem” posed by causal deviance (Enç 2003, 3). Enç thinks there is a
deeper difficulty with the CTA, namely, the concern that its account of
action as constituted wholly by event-causal relations seems to leave the
active agent out of the picture.12 While there is indeed such a concern, the
problem of excluding deviance is not an independent technical issue but
is, rather, expressive of that very concern, since the claim that the CTA
cannot exclude deviance is a way of drawing attention to its alleged failure
to account for the agent’s own activity. Any proponent of the CTA who
accepts Enç’s deeper concern, then, ought to care about resolving the
problem of causal deviance.
There is, I think, a simple explanation for Davidson’s unconcern over
what he thought was the irresolvability of the deviance problem. In “Prob-
lems in the Explanation of Action,” he says: “Let me begin by answering
Wittgenstein’s famous question: what must be added to my arm going up
to make it my raising my arm? The answer is, I think, nothing. In those
cases where I do raise my arm and my arm therefore goes up, nothing has
been added to the event of my arm going up that makes it a case of my
raising my arm” (Davidson 1987/2004, 101). This is, at first, very surpris-
ing. One would expect a proponent of the CTA to say that what must be
added to my arm’s going up to make it a case of my raising my arm is just
the right kind of causal history. Davidson clarifies his answer thus:

When I say nothing has to be added to my arm going up to make it a case of my

raising my arm, I don’t mean no further conditions have to be satisfied to insure
that the rising of my arm is a particular case of my raising my arm. . . . But this
76 J. Bishop

addition is an addition to the description we give of the event, not an addition to

the event itself. So what my claim comes to is this: of the many individual events
that are risings of my arm, some are cases of my raising my arm, and none of the
cases of my raising my arm are events that include more than my arm going up.
Nothing is added to the event itself that makes it into an action. (Ibid., 101–102)

What Davidson is relying on here is something that was clear from the
start in “Actions, Reasons, and Causes”: his ontology of actions is an ontol-
ogy of events. So, on Davidson’s view, actions are a subclass of events, so
that any particular action is identical with some particular event. Events
are happenings, occurrences. So actions, doings, would appear to be a
species or subclass of happenings, occurrences. To identify any particular
event as an action we do indeed need to satisfy conditions that relate to
the event’s causal context: but, though that yields a description of the
event as an action of a certain type, it does not add anything to the event
itself. Given this, the problem of natural agency just reduces to the problem
of explaining how mental states and events can belong to the natural
causal order. For, if actions are events, to be identified as such by having
the right sort of psychological causal history, then, once one has admitted
mental states and events to one’s ontology, there is no further problem
about admitting actions. The problem of causal deviance—even if it proves
irresolvable—can have no impact on this: actions are securely within any
naturalist ontology that admits psychological states and events, even if we
cannot complete in ultimate detail the conditions required for events to
count as actions.

3 Ruben on the Ontology of Action

I suggest, however, that Davidson’s event ontology of actions begs the

question against the agent-causationists, and is thus powerless, by itself,
to deal with skepticism about natural agency. The bottom line for David-
son’s ontology is that doings are a kind of happening, and that is exactly
what the skeptic doubts and the agent-causationist denies. Proponents of
the CTA will, I think, have to do better than this if they are to secure
reconciliatory naturalism, and show how actions can belong within the
ontology of a natural scientific metaphysics. I have thus found very inter-
esting David-Hillel Ruben’s recent revisiting of the fundamentals of the
ontology of action (Ruben 2003, 2008). I will here focus on the more recent
of these works. Although I do not share Ruben’s own conclusion—and
indeed will argue that accepting it would leave the problem of natural
agency quite unresolved—I believe that his canvassing of options for
Skepticism about Natural Agency and the CTA 77

understanding the ontology of action is useful in clarifying what position

proponents of a CTA ought to adopt if they are to secure the possibility of
natural agency.
Ruben observes that it is widely accepted that whenever someone acts
there is an event intrinsic to her action, such that there could have been
no such action without it, yet an event of just that type could have
occurred without any action taking place. To return to Wittgenstein’s
example: whenever someone raises her arm, there is the event of her arm
going up. She could not have raised her arm without her arm going up;
but her arm could have gone up without her raising it.13 How should we
understand the relationship between an action and its intrinsic event?
One possibility is that the action causes, and is therefore a distinct
event from, its intrinsic event. To make sense of this, it seems that actions
have to be identified as “internal tryings” or “willings”:14 my raising my
arm is my causing my arm to go up, and what could that be other than
some antecedent mental event that causes the relevant muscular contrac-
tions? On this “volitionist” view, all our actions are really mental actions—
which goes contrary to the phenomenology of action under which we
experience ourselves as having direct control over bodily movements and
are not generally conscious of “internal tryings.” Furthermore, on pain of
infinite regress, internal tryings or willings cannot themselves have intrinsic
events, and will have to be accepted as having the status of actions essen-
tially. So, if our ontology of action has to include such items, skepticism
about how action can belong in the natural causal order may justifiably
persist.
But, if our actions are not the causes of their intrinsic events, does it
then follow that they are simply identical with them? The view that it does
Ruben takes to be typical of CTAs: “If one holds that actions just are events
non-deviantly caused by prior rationalizing mental events, then the action,
his bending of his finger, just is the event, the bending of the finger, if the
latter is so caused” (Ruben 2008, 232). This is, indeed, Davidson’s posi-
tion—but my present aim is to suggest that this is not the only view tenable
for a proponent of a CTA. Ruben has several objections to a Davidsonian
CTA. He thinks it leads to a “mental overpopulation” problem when
applied to intentional mental actions; and he thinks it fails to deal ade-
quately with habitual and skilled actions (Ruben 2003, chapter 4). But the
objection he emphasizes in relation to a CTA’s presumed commitment to
the identity of an action with its intrinsic event is the very same one that
Enç and Velleman press: “the identification of actions with bodily events
robs us of the very activity of actions; it is hard to see how action can be
78 J. Bishop

constructed from the passivity of what happens” (Ruben 2008, 238). That
objection does have some force. But, in any case, as I have argued, identify-
ing an action with its intrinsic event (so that, as Davidson puts it, nothing
must be added to my arm’s going up to make it my raising my arm) leaves
the skeptic about natural agency unsatisfied, and the question begged
against the agent-causationist.
If actions are neither causes of, nor identical with, their intrinsic events,
how else could they be related to them? Are actions perhaps some kind of
complex that includes their intrinsic events as proper parts? Ruben thinks
that none of the available possibilities is attractive, and he makes the bold
move of denying the widespread assumption that actions have intrinsic
events—or, at least, of denying this assumption for the case of basic physi-
cal actions such as arm-raisings. So Ruben’s answer to Wittgenstein’s ques-
tion—what is left over if I subtract the fact that my arm goes up from the
fact that I raise my arm?—must be that typically there can be no such
subtraction. On Ruben’s view, there is no commonality among the follow-
ing three cases: (1) the mere event of an arm going up; (2) the event of its
going up where it is intrinsic to a (nonbasic) action (e.g., where I use a
pulley with my right arm in order to raise my immobilized left arm); and
(3) the basic action of my raising my arm to which no event is intrinsic.
Nevertheless, in a broad generic sense of “event” all these do count as
events. But we need a “disjunctive” theory of events in this broad sense,
under which there is no essential feature that mere event, intrinsic event,
and basic action “event” have in common. (This is to be compared with
disjunctive theories of perception, under which veridical perception and
hallucination do not have any single kind of “appearing” in common.)
This suggestion will strike many as implausible. Judgments of plausibility
are hardly decisive, however—besides, what is salient here is that Ruben’s
account, if correct, contributes nothing toward resolving the problem of
natural agency, for it is, in one respect, in the same position as the voli-
tionism he rejects, since it requires including in our ontology a sui generis
class of items that are essentially actions.
Nevertheless, Ruben has done the signal service of showing that, if we
are to avoid his own (perhaps somewhat desperate) move, we have to clean
up our account of how actions are related to their intrinsic events. And, if
we want to resolve the problem of natural agency—which is not a goal
that Ruben takes on, by the way, but it is a goal that provides sound moti-
vation for these inquiries—we will need to give an account of actions and
their intrinsic events that shows that it is reasonable to accept that they
are realized wholly within an ontology of the kind consistent with prevail-
Skepticism about Natural Agency and the CTA 79

ing natural scientific metaphysics. This can be done, I think, by defending

a CTA, but in a way that avoids identifying actions with their intrinsic
events and accepts that Davidson’s pessimism about solving the problem
of causal deviance will have to be shown to be unfounded.
If actions are not identical with, nor the causes of, their intrinsic events,
the only alternative is that they somehow include their intrinsic events
as strictly proper parts. The concept of agent-causation explains how this
can be: an action is a relation between the agent and the event intrinsic
to the action. This relation is the agent’s bringing about or making happen
the event intrinsic to the action, and this is, at least conceptually, a special
relation that is aptly labeled “agent-causation.” The agent’s agent-causing
a certain event (the event intrinsic to the action) is not itself a distinct
event that could be the cause of that intrinsic event. But neither is the
agent’s agent-causing the intrinsic event just identical with that intrinsic
event, since it is a relation that has that event as one relatum. Strictly,
then, actions are not events—although (inspired by Ruben’s disjunctiv-
ism) perhaps we can allow that there is a broad generic sense of “event”
under which actions count as events, provided we realize that an ulti-
mately disjunctive account must be given of that most inclusive notion
of an event.
I see no reason why appeal to agent-causation thus understood should
be controversial. Agent-causationism, however, is another matter. For
agent-causationists claim that the way we naturally think of actions in
terms of the special relation of agent-causation must be the way actions
are realized in the world: that is, for actions to be instantiated there need
to be ontologically irreducible relations of agent-causation. Since prevail-
ing naturalist metaphysics exclude irreducible agent-causation (and,
indeed, irreducible substance-causation of any kind), agent-causationism
rejects the possibility of natural agency and so repudiates reconciliatory
naturalism. But reconciliatory naturalism may be secured, and doubt about
the possibility of natural agency removed, by a CTA that does not simply
identify actions with their intrinsic events (under the right causal condi-
tions), but, rather, holds that actions, which are conceptually agent-causal
relations, may be wholly realized in event-causal relations of the kind
admissible in naturalist ontology. On this account of the matter, what
ontologically realizes an action is its intrinsic event in a causal context of
the right kind. Such a view will be in trouble, though, if it seems that we
are unable to specify what that “right kind” of causal context is. And the
claim that such a specification cannot be given is, of course, the burden
of the argument from the possibility of causal deviance. Without a good
80 J. Bishop

reply to that argument, agent-causationism remains persuasive: what is

conceptually primitive and irreducible will then seem to be ontologically
primitive and irreducible also.

4 Conclusion: Some Desiderata for Defending a Causal Theory of Action

Quite a lot rests, then, on whether or not an ontological CTA-analysis that

excludes deviant counterexamples can be provided, and I have previously
tried to defend just such an analysis, relying extensively on the work of
others, especially Christopher Peacocke and David Lewis.15 In brief, the
view I proposed maintains that, where—as in cases of relatively “basic”
action—deviance cannot be excluded by a condition of match with the
agent’s implicit plan, there are two conditions on the causal link between
the agent’s intention to A and the agent’s A-ing behavior that are necessary
and sufficient to ensure that it constitutes a genuine intentional action of
A-ing. First, that causal link must be sensitive to the content of the inten-
tion, in the sense that, over a sufficiently wide range of differences in
content, had the agent’s intention differed, the resulting behavior would
have differed correspondingly. Then, in order to exclude counterexamples
where the sensitivity condition is satisfied but the link between intention
and behavior passes through another agent’s practical reasoning, a further
condition is needed. If the causal mechanism linking intention to behavior
involves feedback (as typically it does) then the feedback signal must be
routed back to the agent’s central mental processes, if to anyone’s. Now is
not the time to consider directly whether an account of this kind can
succeed (nor to elaborate it). My present interest has been with the more
important and prior question of whether the development and critique of
such accounts is of any more than technical philosophical importance. I
believe I have shown that it is. I believe I have shown that, if one aims to
resolve doubts about how responsible agents could be part of the natural
causal order, one may do so by defending a causal theory of action that,
though it draws gratefully on most features of Davidson’s account, departs
from it on two related very significant points: it does not identify actions
simply with their intrinsic events, but rather with a complex of related
events that constitute the agent’s exercise of control; and it remains com-
mitted to the project of providing necessary and sufficient conditions that
(though they cannot be expected to amount to a real definition of agency)
make it reasonable to accept that episodes of agent-causation may be
wholly ontologically realized in causal relations of the kind admissible in
the prevailing naturalist metaphysics.
Skepticism about Natural Agency and the CTA 81

Acknowledgments

I am grateful to Folke Tersman and the Swedish Collegium for Advanced

Study, Uppsala, for the opportunity to read an earlier draft of this essay at
a workshop entitled “Acting for a Reason—Normativity and Mentality in
a World of Causality,” September 25, 2008, and to Torbjörn Tännsjö for
his commentary.

Notes

1. Note that I do not wish to imply that Wittgenstein himself intended to pose this
question as the foundational one for the philosophy of action.

2. This is, of course, only a necessary condition for moral responsibility: further
conditions are required, relating to behavers’ capacities to understand the moral
significance of their behavior and to offer explanations of it in terms of their own
reasons for acting. Infants and nonhuman animals may arguably perform actions,
yet without being morally responsible for related outcomes.

3. Note that agents may sometimes be responsible for an outcome by failing to

act to prevent it. For example: I sit and do nothing while the kettle boils dry. I may
be responsible for the damage—but only if my behavior is genuinely my own (in)
action rather than (for example) the result of being paralyzed or in a trance. Not
doing something may thus be, in a certain sense, something I do rather than some-
thing that merely happens to me. So it is important to recognize two different
distinctions: (a) the distinction between active and passive behavior (doing
versus omitting, letting happen, refraining, etc.) and (b) the distinction between
behavior that is my own doing (through which, ceteris paribus, my moral responsi-
bility may be transmitted) and behavior that just happens to me (and blocks my
responsibility).

4. This problem would disappear if a scientific revolution took place that presup-
posed a metaphysics of irreducible substance-causation. I shall argue that the pros-
pects for defending reconciliatory naturalism do not depend on having to expect
so apparently unlikely a development.

5. I here set aside Daniel Dennett’s “intentional stance” approach to overcoming

skepticism about natural agency, since, on such an antirealist view, intentional
explanation will obviously involve no ontological commitment that could clash
with naturalism. As Dennett explains, “the success of [the intentional stance] is of
course a matter to be settled pragmatically, without reference to whether the object
really has beliefs, intentions and so forth” (Dennett 1979, 238).

6. A CTA can resolve the problem of natural agency only if there is a generally sat-
isfactory naturalist solution to the mind–body problem: but that does not make the
82 J. Bishop

CTA otiose, of course, since skepticism about natural agency might coherently
persist even if physicalism about mental states and events is accepted.

7. I have attempted a defense of believing on such a “nonevidential” basis—under

certain conditions, anyway—in Bishop 2007. My defense is a development of
William James’s “justification of faith” in his famous, though ill-titled, lecture “The
Will to Believe” (James 1956).

8. It is important to note that some libertarians think they can resist this conclu-
sion. Robert Kane (1996) seeks to defend a naturalist libertarianism without com-
mitment to agent-causation, and Timothy O’Connor (2002) affirms the ontological
irreducibility of agent-causation yet nevertheless hopes to save naturalism by appeal
to emergentism. I have discussed these issues more fully in Bishop 2003.

9. An all things considered judgment, as Davidson points out, is still a conditional

judgment—i.e., it has the form “given considerations C, the thing to do is φ,” where
C are all the considerations the agent takes into account. In “How Is Weakness of
the Will Possible,” Davidson contrasts all things considered judgments with uncon-
ditional judgments about how to act. But, in the light of “Intending” it is clear that
these “unconditional judgments” are better understood as intentions. Both these
papers are included in Davidson 1980.

10. The references are to Armstrong 1973, 1975, and Peacocke 1979a.

11. Here Davidson’s discussion in “Intending” is important. Davidson says: “We

end up, then, with this incomplete and unsatisfactory account of acting with an
intention: an action is performed with a certain intention if it is caused in the right
way by attitudes and beliefs that rationalize it” (Davidson 1978/1980, 87). Davidson
then adds the following: “This is where Essay 1 [“Actions, Reasons, and Causes”]
left things. At the time I wrote it I believed it would be possible to characterize ‘the
right way’ in non-circular terms”(ibid., note 3). And the implicature is clear: by the
time he was preparing “Intending” for republication in Essays on Actions and Events,
he no longer thought that would be possible. His discussion then continues as
follows:

If this account [of acting with an intention as caused in the right way by attitudes and beliefs
that rationalize it] is correct, then acting with an intention does not require that there be any
mysterious act of the will or special attitude or episode of willing. For the account needs only
desires (or other pro attitudes), beliefs, and the actions themselves. There is indeed the relation
between these, causal or otherwise, to be analysed, but it is not an embarrassing entity that has
to be added to the world’s furniture. We would not, it is true, have shown how to define the
concept of acting with an intention: the reduction is not definitional but ontological. But the
ontological reduction, if it succeeds, is enough to answer many puzzles about the relation
between the mind and the body, and to explain the possibility of autonomous action in a world
of causality. (Ibid., 87–88)

12. Enç quotes J. David Velleman’s claim that the CTA cannot capture what it is
for an agent to be active: “reasons cause an intention, and an intention causes bodily
Skepticism about Natural Agency and the CTA 83

movements, but [in this picture] nobody—that is no person—does anything” (Vel-

leman 1992, 461; quoted in Enç 2003, 134).

13. It might well be empirically true that arm-risings intrinsic to arm-raisings have
distinctive empirical features that distinguish them from all physically feasible
“mere” arm-risings (such as might occur in a nervous tic or when a paralyzed arm
is lifted, etc.). The possibility that an arm-rising of the very same highly specified
type could have occurred without an arm-raising might then be merely logical.
Nevertheless, the issues about to be canvassed about how actions are related to their
intrinsic events will still need to be dealt with, if we are to be clear about what an
action is.

14. Ruben attributes a view of this general kind to H. A. Prichard in “Acting, Willing,
and Desiring” in Prichard 1949; to Jennifer Hornsby (1980); and to Paul Pietroski
(2002).

15. See Bishop 1989, chapter 5, where I draw on Peacocke 1979b and Lewis 1980.
In Lewis’s paper the relevant elaboration of causal nondeviance is applied to the
case of perception.
6 Agential Systems, Causal Deviance, and Reliability

Jesús H. Aguilar

According to the causal theory of action (CTA) an action is an event

caused by a mental state or event that rationalizes its execution. Actions
are typically exemplified by bodily movements, and their internal causes
by intentions. The CTA has traditionally been saddled with problems
emerging from so-called deviant causal chains, namely, chains of events
that satisfy the CTA’s conditions for the production of an action but
whose product is intuitively not an action. A plausible strategy for defend-
ing this theory against the possibility of deviant causal chains is grounded
on the proposal that the bodily movement corresponding to an action
must be sensitive to the content of the mental state that causes it.1 Let
us call this the sensitivity condition. Sensitivity here is understood as
the specific responsiveness that a bodily movement can have to the par-
ticular content of a motivating mental state. However, this strategy faces
a serious challenge from cases where bodily movements are produced by
means of causal chains of events that involve intermediate actions per-
formed by another agent. These cases are quite challenging because they
apparently generate deviant causal chains that satisfy the sensitivity
condition.
Consider Christopher Peacocke’s (1979b) example of a neurophysiolo-
gist who reads an action-triggering intention directly from a subject’s brain
and then stimulates the corresponding efferent nerves. The stimulation in
turn causes a bodily movement that matches the subject’s intention.
According to Peacocke, although such bodily movement is transitively
caused by an intention that rationalizes its execution, and although such
bodily movement is sensitive to the specific content of this intention, it is
not one of the subject’s actions. The reason such bodily movement fails
to count as an action is that its production contravenes a fundamental
assumption concerning the nature of action and agency, namely, that the
agent must be the originator of her own actions.2 Given the strange causal
86 J. H. Aguilar

path involved in these type of scenarios, they are usually considered

deviant and presented as counterexamples to the CTA.
If indeed such scenarios involving what we may call “prosthetic agents”
generate causal deviance, we are facing a serious challenge to the CTA’s
best effort to avoid causal deviance based on the appeal to sensitivity.
Oddly enough, if one believes that Peacocke’s neurophysiologist scenario
is deviant, then one also needs to explain in what sense this particular case
is different from other scenarios involving prosthetic agents that share
with it all their salient features without themselves being obviously prob-
lematic. Consider a similar fictional scenario where an assistant’s job is
simply to hold some wires that connect the efferent nerves of a subject,
allowing the production of the subject’s bodily movements.3 This case of
prosthetic agency is analogous to the neurophysiologist’s since in both
cases there is an intermediate agent who intervenes in the causal chain
that ends with the subject’s bodily movement. And yet, in contrast to the
neurophysiologist’s case, it is hard to say that the subject’s ensuing bodily
movement is not an action performed by the subject. If anything, the
causal role of the assistant is very similar to the causal role played by the
subject’s own efferent nerves, namely, that of a causal bridge connecting
the subject’s intentions to the subject’s bodily movements.
Peacocke’s answer to the challenge arising from deviant cases involving
prosthetic agents consists in requiring that intentional behavior explicitly
exclude the possibility of two agents interacting in this prosthetic way. He
believes that by stipulating that a causal chain leading to intentional
behavior “should not run through the intentions of another person”
(Peacocke 1979b, 88), one can safely rely on the sensitivity condition to
take care of deviance. If by definition no cases involving two agents inter-
acting in this way are possible, then the puzzling resemblance between the
cases of the neurophysiologist and the assistant is not an issue.
But this is hardly a satisfactory reply if one is a supporter of the CTA.
Not only is it an arbitrary exclusion of cases that are clearly conceivable
like the assistant’s, but this reply is also in tension with the CTA’s general
approach to action. In particular, Peacocke’s stipulation is in tension with
the CTA’s acceptance of transitive causal relations involving actions as
much as wires and nerves.4 Moreover, if there is a promising line of attack
open to the defender of the CTA to deal with general cases of deviance, it
is one based on sensitivity. But the sensitivity condition is useless in pros-
thetic cases since they all involve outcomes that are sensitive to their causal
antecedents. Thus, it is important to know exactly what is wrong or unac-
Agential Systems, Causal Deviance, and Reliability 87

ceptable about a causal chain of events that goes through someone else’s
action, produces an event that satisfies the conditions to count as an
action, and yet fails to be an action.5 At stake is nothing less than the viabil-
ity of the CTA.
Similar considerations have motivated defenders of the CTA to confront
the challenge arising from prosthetic agency and in doing so reap the
benefits of a plausible answer in the form of extra conditions that identify
an intentional action. The most ambitious and promising of all these
efforts by defenders of the CTA is due to John Bishop. Not only does Bishop
offer a causalist answer to the challenge arising from prosthetic agency,
but contained in his answer he also offers a set of necessary and sufficient
causal conditions for a basic intentional action. Not surprisingly he calls
this a “final breakthrough” in the search for the elusive and much-sought-
for causal conditions for an intentional action.6 Furthermore, Bishop’s
necessary and sufficient conditions are the result of an analysis of inten-
tional action that comes from a rich systemic perspective in which the
agent and her contributions to the world of events are at the center of
attention. All these reasons justify our taking a careful look at Bishop’s
proposal and assessing his claim to have found the necessary and sufficient
conditions for a basic intentional action. In the rest of this chapter I first
examine Bishop’s proposal, stressing the way in which he uses the notion
of agential control to deal with deviance arising from prosthetic deviance.
Then, I raise some problems for the specific way in which feedback is sup-
posed to enter in this systemic picture, and I end up by suggesting a move
in the direction of reliability to complement Bishop’s otherwise attractive
strategy to tackle basic deviance.

1 Bishop’s “Final Breakthrough”

Bishop’s version of the CTA belongs to a family of causal theories that

consider agents as functional systems capable of producing intentional
behavior by entering into internal states that besides causing such behavior
also rationalize it.7 A crucial aspect of this way of understanding agency is
that it sees the relevant agential systems as dynamic centers of behavioral
control extending their agential reach by functional mechanisms. In turn,
an action is seen as the causal product of agential systems involving ante-
cedents such as intentions and/or belief-desire sets of internal states. In
this way, the set of causal conditions that give rise to an event legitimately
identified as an action is framed within an agential systemic setting.
88 J. H. Aguilar

Correspondingly, the efforts to deal with the challenge posed by causal

deviance involve the offering of systemic conditions that would prevent
such cases.
Bishop’s strategy to establish the systemic conditions that would iden-
tify an action and thereby prevent causal deviance is grounded on the most
fundamental attribute that an agential system must satisfy, namely, control
over the actions generated by such a system. In particular, this is the strat-
egy that he offers to account for the deviant cases involving prosthetic
agency. Take the puzzling different ways in which we tend to react to
the prosthetic case involving the neurophysiologist and the prosthetic
case involving the assistant. According to Bishop, it is possible to make
sense of these two different reactions by looking into the corresponding
ways in which each of the two prosthetic agents carries out his or her
interventions:

Sometimes the second agent’s [i.e., the neurophysiologist’s] involvement in the

causal chain preempts or blocks the agent’s exercise of direct control over his or her bodily
movements. Then the second agent is no mere cog in the mechanisms that realize
the first agent’s direct control. Rather, the second agent is part of a system that
provides the first agent with, at best, only indirect control over the movements of
his or her own body. (Bishop 1989, 159)

Hence, cases (e.g., the neurophysiologist’s) where the intervention of the

prosthetic agent disrupts the first agent’s direct control over her bodily
movements are deviant, whereas cases that do not disrupt her control are
nondeviant. Consequently, to distinguish deviant from nondeviant cases
it is necessary to identify the specific features of the prosthetic agent’s
intervention capable of undermining the first agent’s control over her
bodily movements. Bishop identifies these disruptive features by focusing
on the “sustained” nature of most actions.8
An action is sustained when it involves a process whereby its agent
regulates its execution by monitoring whether the ensuing behavior has
been completed or whether it has satisfied the intended goal. This process
necessarily involves feedback information arising from the behavior under
execution. Bishop proposes that under normal conditions the information
is fed to the brain of the subject who triggers the chain of events leading
to the behavior. However, under abnormal conditions the information
may never reach the subject’s brain. Bishop contends that this last thing
is exactly what happens in deviant cases when the prosthetic agent disrupts
the control of the main agent over her bodily movements, for in such
cases:
Agential Systems, Causal Deviance, and Reliability 89

There is a servo-system functioning to match the agent’s intention all right; but
given its detailed architecture, it can hardly count as realizing the agent’s controlled
regulation of his or her bodily movements since the feedback information about
orientation and muscular states does not get carried back to the agent’s central
processing system. (Bishop 1989, 170)

To illustrate Bishop’s proposal let us try it on Peacocke’s neurophysiologist

case. If indeed the neurophysiologist is capable of reading the action trig-
gering intentions from a subject’s brain in order to produce behavior that
exactly matches their content, then the feedback information involved in
the execution of this behavior will miss the subject’s brain. That is, if
anyone is getting this feedback information, it would be the neurophysi-
ologist. This results in the subject’s lack of behavioral control, revealing its
deviant nature. In contrast, when the assistant simply holds the wire,
nothing prevents the feedback information from reaching the subject’s
brain, allowing her to control her bodily movement. This would explain
why the assistant’s case is not deviant.
Presumably this imaginary scenario does not involve the possibility of
obtaining feedback information through some external way, say, checking
whether indeed the neurophysiologist is satisfying the action-triggering
intention by watching the presence or absence of the corresponding bodily
movement. Bishop is only concerned with the internal feedback informa-
tion that is normally involved in the production and control of an inten-
tional bodily movement. However, the possibility of external feedback
does raise interesting questions concerning Bishop’s proposal to deal with
deviant cases involving prosthetic agents. For instance, one can imagine a
perceptually deviant scenario where the external feedback information is
obtained through a deployment of mirrors and fake images that produce
some data capable of being used by the subject to control one of her bodily
movements. We can then imagine that the subject’s body moves exactly
as she “sees” it moving. So she has a true belief concerning her bodily
movement that in fact is influencing and partly rationalizing her actual
bodily movement, and yet, she does not seem to be producing an inten-
tional bodily movement. This appears to be a case in which the subject
makes her body move without moving it. This type of case suggests that
Bishop is always assuming that the relevant scenarios for his proposal do
not involve external feedback.9
Thus, Bishop proposes the inclusion of feedback as a necessary condi-
tion to identify an event as a basic action. In fact, he believes that the
addition of this condition leads to the “final breakthrough” in the CTA’s
90 J. H. Aguilar

traditional effort to provide necessary and sufficient conditions to identify

an action. Hence, according to Bishop:

M performs the basic intentional action of a-ing if and only if,

(1) M has a basic intention to do a; and,

(2) M’s having this basic intention causes M to produce behavior, b, which instanti-
ates the types of state or event intrinsic to the action of a-ing; where the causal
mechanism from M’s basic intention to b satisfies the sensitivity condition; and if
this causal mechanism involves feedback, then the feedback signal is routed back
to M’s central mental processes if to anyone’s. (Bishop 1989, 172)10

Thus, besides sensitivity, which Bishop requires to preclude deviance in

typical causal chains, his proposal adds the feedback condition to deal with
the peculiar deviance that emerges in some cases involving prosthetic
agents. This move involves the recognition of a feedback signal as a con-
stitutive element of many actions. That is, the relevant causal chain of
events producing an action involves in most cases a two-way causal
sequence of events: one that goes from the agent’s mental processes of
control to the bodily movement, and a second one that sends back infor-
mation about the bodily movement that is performed to the agent’s mental
processes of control.

2 The Problem with Bishop’s Move

Unfortunately, Bishop’s reply is insufficient to deal with deviant cases

involving prosthetic agents like Peacocke’s neurophysiologist. Even the
addition of the feedback condition to the sensitivity condition does not
eliminate the possibility of having cases where these conditions are met
and yet deviance seems to occur. The source of difficulties comes from the
systemic approach favored by Bishop’s version of the CTA, together with
this theory’s commitment to causal transitivity and the way in which
Bishop himself conceives the role of the prosthetic agent.
Let us remember that in nondeviant prosthetic cases Bishop is willing
to grant that a prosthetic agent’s action can be placed between the mental
states of the main agent and her bodily movements without this under-
mining the main agent’s control. Thus, Bishop accepts that cases involving
prosthetic agents by themselves are not the source of deviance. Rather, he
proposes that deviance occurs only when the prosthetic agent blocks the
control of the main agent by stopping the feedback information from
reaching the main agent. However, let us note that in order for this pro-
posal to even start making sense the causal chain going from the main
agent to the movement of her body needs to go through the prosthetic
Agential Systems, Causal Deviance, and Reliability 91

agent’s action. This in itself does not yet distinguish this causal chain from
a nondeviant one, for it is only in the process of receiving information
from the bodily movement that the desired distinction supposedly arises.
Bishop then asks whose brain is getting the feedback information: if it is
the prosthetic agent, then we are dealing with a deviant case; if it is the
main agent, then this is not a deviant case.
The problem with this suggestion is that nothing prevents enriching
the neurophysiologist scenario with the possibility of a further link going
this time from the neurophysiologist back to the subject’s brain—in other
words, an extra link that sends the information received from the subject’s
bodily movement back to the subject’s brain with the help of the neuro-
physiologist. If causal transitivity is sufficient to go in one direction, causal
transitivity should be sufficient to permit the flow of information to go in
the other direction. If this occurs, then both brains receive feedback infor-
mation, and hence this sole feature cannot be what distinguishes cases that
are deviant from cases that are not deviant.
There are different ways in which Bishop could reply to this objection.
One way is to stipulate that it is unacceptable to have a second agent who
sends back the information to the first agent in the way suggested, much
like Peacocke’s stipulation rejecting the possibility of prosthetic agency.
But this answer is clearly at odds with Bishop’s acceptance of nondeviant
cases like that of the assistant. If with Bishop we can conceive that the
assistant is capable of bridging the subject’s brain with the subject’s behav-
ior, then we can enrich this thought experiment and conceive that the
assistant is capable of bridging the subject’s behavior with the subject’s
brain. This of course will not pose a problem for Bishop’s proposal.
However, if with Bishop we can also conceive that the neurophysiologist
is capable of bridging the subject’s brain with the subject’s behavior, then
nothing stops us from similarly enriching this thought experiment and
conceiving that the neurophysiologist is capable of bridging the subject’s
behavior with the subject’s brain. This does create a serious problem for
Bishop’s proposal.
Alternatively, Bishop can answer this objection by accepting that if the
feedback information that allows the first agent to exercise his control
reaches his brain in the suggested way, then strictly speaking we do not
have a case of deviance. The enriched neurophysiologist scenario would
involve a rather strange and circuitous causal path, but nonetheless a
nondeviant one insofar as the relevant information is reaching the subject.
In fact, this alternative possibility, where the neurophysiologist acts as a
functional bridge of feedback information, is consistent with a systemic
92 J. H. Aguilar

approach to agency. For what seems to emerge in this enriched neuro-

physiologist scenario is a larger agential system that makes use of two
subsystems in the form of two cooperative agents.
As attractive as this option may seem, it also faces substantial difficul-
ties. Even if on the surface we are apparently dealing with a larger agential
system composed of two subsystems made up of two agents working in
coordination, as soon as we analyze the details of such an arrangement,
problems emerge for this picture, particularly with the key notion of agen-
tial control. To continue with our thought experiment, let us note that
although the subject’s action-triggering intentions are apparently guiding
the enlarged agential system, the reality is quite different. The functional
role played by the subject’s action-triggering intentions is completely para-
sitic on the active functional participation of the neurophysiologist’s own
intentions, whose functional role is among other things to produce the
relevant bodily movements and send back the relevant information about
their execution. Moreover, whereas it is up to the neurophysiologist to
satisfy the subject’s action-triggering intentions and provide the corre-
sponding feedback information, it is not up to the subject to satisfy his
own action-triggering intentions or those of the neurophysiologist. This
asymmetry is crucial to establish things like agential control and attribu-
tion of responsibility for the outcome of this enlarged agential system. For
once it is clear who is doing what in this agential system, then it is incor-
rect to attribute any significant agential control to the subject. His role
resembles more the one played by a subject who advises another about
what to do than an agent who collectively participates in the execution of
the proposed action.
If this analysis of the very limited participation of the subject inside the
enlarged agential system is correct, then it appears that strictly speaking
we are dealing with only a single agent in the form of a neurophysiologist
who uses someone else’s intentions as guidance and someone else’s body
as means to produce actions. Notice that no amount of feedback informa-
tion reaching the subject is going to change this situation. This interpreta-
tion of the case at hand effectively undermines the causal efficacy of the
subject’s action-triggering intentions. So, under this interpretation of the
enlarged agential system, Bishop’s feedback condition can be satisfied and
still the performance of a basic intentional action by the subject does not
occur.
Nevertheless, if one insists that given the subject’s admittedly very
limited participation in the overall system there is a sense in which he is
causally contributing to the behavioral outcome, then we are back to a
Agential Systems, Causal Deviance, and Reliability 93

more traditional case of causal deviance: The subject’s action-triggering

intentions cause a matching behavior that is not under his control. So,
under this second interpretation of the enlarged agential system, Bishop’s
feedback condition can be satisfied and still the performance of a basic
intentional action by the subject does not occur.
What emerges from this enlarged systemic picture is that in terms of
agential control the subject’s contribution to the whole system is com-
pletely parasitic on the contribution of the neurophysiologist. It is ulti-
mately with this second agent that the decision to fulfill the relevant
action-triggering intention and/or inform via feedback about its execution
takes place. Let us also note that it is this crucial part in the exercise of
agential control that is absent in the case of the assistant since he is not
even aware of the action-triggering intentions that by holding some wires
he is helping to realize. Correspondingly, were we to imagine that not only
is he holding some wires but also fully aware of the action-triggering inten-
tions that he is helping realize, then we would treat his intervention in
much the same way as the neurophysiologist’s, that is, as giving rise to
prosthetic deviance. It is thus unclear how the satisfaction of Bishop’s
feedback condition can make any difference in terms of the agential
control that needs to be in place if the subject is going to produce inten-
tional behavior in this way as part of a larger agential system. And, if it is
unclear how the feedback condition is going to give weight to the required
agential control, it is also unclear how the satisfaction of such condition
is going to deal with the challenge coming from causally deviant cases
involving prosthetic agency.

3 Reliability and Differential Explanation

Despite Bishop’s failure in dealing with deviant cases involving a prosthetic

agent via the feedback condition, and hence his failing to produce the
“final breakthrough” by offering the necessary and sufficient conditions
for a basic intentional action, important points arise directly from his
effort. Indeed, they suggest a plausible answer to this unique type of devi-
ance essentially grounded in reliability and in Peacocke’s account of
the sensitivity condition. Moreover, this answer is fully consistent with
Bishop’s systemic approach to agency.
I propose that what gives rise to problems concerning deviance in cases
involving a prosthetic agent like that of the neurophysiologist is basically
the unreliability accompanying the prosthetic agent’s intervention. So, as
correctly diagnosed by Bishop, in cases of deviance the prosthetic agent’s
94 J. H. Aguilar

intrusion is seen as undermining the control of the first agent over the
movements of his body. However, against Bishop’s diagnosis, the lack of
control arises not because some feedback information is misdirected or
unavailable to the first agent, but rather because typically the intervention
of an agent involves the breaking of the relevant causal chain, bringing
with it causal unreliability.
In fact, this is the type of scenario that essentially troubles Peacocke
with respect to cases involving a prosthetic agent. For instance, in the case
of the neurophysiologist, he suggests that:

When we say that an event is, under a given description, intentional of a person,
we normally imply that that person was the originator of that event. It is not clear
whether there is such a person as the originator of the bodily movement in our
example, but if there is, it is certainly not the person whose brain the neurophysi-
ologist is inspecting. (Peacocke 1979b, 88)

Nevertheless, Peacocke incorrectly locates the precise source of deviance.

For we have seen how there can be cases involving a prosthetic agent where
the intervention by a second agent does not undermine our attribution of
control to the first agent. These nondeviant cases involving a prosthetic
agent effectively mimic what normally occurs when nerves connect brain
states with bodily movements. So, the mere intervention of a prosthetic
agent in these cases is not what is creating the problem alluded to by
Peacocke’s qualms about the originator.
Rather, and as suggested by Bishop, the source of deviance is linked to
the undermining of agential control. More precisely, the actual source of
deviance is the way in which the prosthetic agent is thought to be inter-
vening, namely, a way that strongly diminishes our confidence that his
intervention leaves intact the reliability of the causal connection leading
to the first agent’s bodily movement. Hence, it is because interventions
like the neurophysiologist’s turn out to be far less reliable than, say, wire
holders, real wires, or nerves, that deviance occurs. Correspondingly, if
somehow the connection were to be reliable, for instance, if the neuro-
physiologist turns out to be a compulsive satisfier of the subject’s inten-
tions, then it is no longer clear that we are dealing with a deviant case.
For then the situation would be very similar to the reliable and nondeviant
intervention of the wire holder.11
Moreover, despite Peacocke’s stipulative reaction to cases involving
prosthetic external agents, he has also provided a conceptual tool that
effectively identifies the presence of reliable causal chains, and that can
help us in dealing with deviance in such cases, namely, through the
Agential Systems, Causal Deviance, and Reliability 95

sensitivity condition understood in terms of his differential explanation.12

This is how Peacocke roughly defines this type of explanation:

x’s being ϕ differentially explains y’s being ψ iff x’s being ϕ is a non-redundant part
of the explanation of y’s being ψ, and according to the principles of explanation
(laws) invoked in this explanation, there are functions . . . specified in these laws
such that y’s being ψ is fixed by these functions from x’s being ϕ. (Peacocke 1979b,
66)

Although differential explanation strictly speaking deals with sentences,

the laws that appear in a differential explanation refer to regularities in the
world where items having certain properties functionally fix the properties
of other items. Peacocke also stresses that the functions specified in the
laws invoked in a differential explanation “may be defined not only over
numbers but also over colours, chemical elements, compounds, shapes,
and so forth” (ibid., 67), and, of course, over behavior.
For our purposes of articulating a way to distinguish reliable causal con-
nections, two general features of differential explanation are worth noting.
The first feature is that the covering laws appearing in a differential expla-
nation are nonredundant statistical causal laws, as befits an account that
is linked to reliability. Even when these causal laws capture relationships
among properties where some of them functionally “fix” the presence of
others, the laws are still statistical.13 So, only the continuous occurrence of
these relationships ensures that indeed we are facing a reliable connection,
which incidentally is something that we should also expect when dealing
with a connection that normally comes in degrees.14 The second feature is
that “differentially explains” is a transitive relation. Thus, if there are
intermediate stages of a causal chain that can be differentially explained,
every one of them is differentially explained by properties of the event
initiating the chain. This means that properties of the first event function-
ally fix the causally relevant properties of all the members in the chain.
When this occurs, according to Peacocke, “we may say that the chain is
sensitive, relative to an initial object’s being in a certain state” (ibid., 69).
Coming back to deviant cases involving prosthetic agents, we can make
the following two points that emerge from the proposal linking reliability
to differential explanation. The first point is that it should be clear why
cases like that of the neurophysiologist are considered deviant, namely,
because they involve nonreliable causal chains. The neurophysiologist is
usually portrayed as someone whose intervention starts a new causal chain
that breaks the one initiated by the subject, hence allegedly making it
impossible for this last chain to ensue in a true intentional behavior. But
96 J. H. Aguilar

we know that this cannot be the reason for considering this case deviant,
as the assistant case shows. Rather, what is crucial is the level of reliability
accompanying the intervention of the neurophysiologist. In particular,
what is crucial is the role played in this intervention by that central agen-
tial feature which can easily increase the unreliability of any system pos-
sessing it, namely, the “up-to-ness” implicit in autonomous agency. Thus,
if the neurophysiologist intervenes in such a way that it is up to her to
decide to materialize the subject’s intention, then the chances that this
involves a reliable connection diminish and the chances of its being
deviant increase, for it is easy to imagine all sorts of considerations that
may lead her not to satisfy the subject’s intention. However, if her inter-
vention is very similar to the assistant’s, namely, essentially blind to the
content of the subject’s action triggering intention, then the chances that
this involves a reliable connection increase and the chances it involves
deviance diminish.
Furthermore, a second point concerning deviant cases involving pros-
thetic agents is that making use of reliability shows that there is an inevi-
table misleading oversimplification in the neurophysiologist case as it is
normally presented in the literature. For if indeed this case can be differ-
entially explained then this is the best proof that it is reliable, and hence,
that it is not deviant. What is misleading, of course, is that we are asked
to consider a single successful case obviating the statistical evidence that
would show it to be a reliable one. Presumably, this is a fair move when
trying to pull the intuitions associated with the obstacle of an originator.
But as soon as we recognize that there are nondeviant cases involving
prosthetic agents this move loses much ground. Thus, as it stands, we
strictly speaking lack the relevant information that would settle the issue
as to whether indeed the neurophysiologist case is deviant or not. However,
this does not amount to a proof that cases involving prosthetic agents are
a real source of deviance. All it shows is that one can, with a little imagina-
tion, construct cases where the reliability of a normal causal connection
diminishes, and, hence, where the presence of deviance proportionally
increases. But that is all.
Nevertheless, it is important to note that although appealing to reli-
ability seems to take care of the sources of deviance involving prosthetic
agents, the larger question remains as to whether indeed reliability is an
objective feature of causal chains of events that separates those that are
not deviant from those that are. My view on this is that reliability together
with sensitivity and perhaps some condition involving feedback à la Bishop
provides the CTA with the conditions to identify a basic action, and hence
Agential Systems, Causal Deviance, and Reliability 97

to eliminate the possibility of causal deviance. Although a full develop-

ment of these ideas would take us away from the main objective of this
essay, let me conclude with a sketch of the way in which reliability plays
a constitutive role in our understanding of action and agency.

4 The Constitutive Role of Reliability

Even assuming that it is possible to deal with deviance in cases that involve
prosthetic agents by appealing to the presence or absence of reliable causal
chains, we still need to know why an unreliable causal chain undermines
agential control and provides the basis for deviance. This question is even
more pressing when it is not hard to imagine cases where an agent per-
forms an intentional bodily movement despite the unreliability of the
causal chain involved in the production of such movement. For this seems
to show that agential control can be preserved even when the relevant
causal chain is not reliable.
For example, it is conceivable that after trying many times a subject
whose arm is paralyzed succeeds only once in moving her arm. Although
generated through an unreliable causal chain, there is no obvious reason
why in this case the relevant movement is not intentional. It appears, then,
that the reliability or unreliability of the relevant causal chain is indepen-
dent of the intentional status of a bodily movement. But, if this is the case,
then it is unclear how reliability can be seen as a necessary element in the
production of an action.
Nevertheless, a more careful analysis of, say, the case of the single action
performed by the paralyzed subject reveals that reliability does play a con-
stitutive role in the recognition of this subject as an agent and the accep-
tance of her single fortuitous arm movement as an action. Mutatis mutandis,
the same considerations apply to every case involving an intentional
action that result from an apparently unreliable causal chain.
As has been suggested earlier, the paralyzed subject would count as an
agential system insofar as she is capable of producing specific types of
bodily movements that correspond to the content of some specific type of
internal states capable of causing such bodily movements. However, in
order to make sense of these different types of events and states into which
an agential system can enter, more than a causal relationship among par-
ticular events and states is required. This extra requirement is that the
relevant causal connections among the particular events and states of the
system are reliable enough to establish a distinctive type of event or state
that is exclusively related to another distinctive type of event or state. That
98 J. H. Aguilar

is, the particular events and states of the agential system are grouped into
relevant types of events and states insofar as they are captured by reliable
connections. Only then do we have the required types of events and states
presupposed in our description of an agential system and its functions.
If this is correct, then the very types of bodily movements available to
an agential system turn out to be a function of the reliable connections
that link such types of intended events with their corresponding internal
types of states, typically, with types of intentions. In fact, the production
of such types of intentions is grounded on some further cognitive state of
the system that conveys the information that indeed the intended type of
bodily movement is potentially executable, again, because a reliable con-
nection is assumed to exist between the state of having a specific type of
intention and the production of a specific type of bodily movement.15
Therefore and despite appearances, the single successful movement of
the paralyzed subject is an action insofar as it is an instantiation of a type
of behavior that is reliably connected to a specific type of intention. Her
agential effort is to be understood as producing an intention to move her
arm, hoping that it will in turn give rise to what it normally produces,
namely, the movement of her arm. It just happens that in her abnormal
situation the reliable connection will not likely be instantiated because of
her physiological problem. But her intention remains and it is the rational
one to have if indeed she wants to move her arm, since under normal
conditions this is the most reliable way to give rise to such movement.
Note how the very basis of our rationalizing not just her single bodily
movement but her forming the intention to produce it against all odds is
precisely the assumption that under normal conditions that is what it takes
to move one’s arm. The main point here is that normalcy can only be
cashed out in terms of reliability.

Acknowledgments

I am very grateful to Andrei Buckareff, John Bishop, Fred Adams, Paul

Pietroski, David Davies, Sarah Stroud, and Mary Clayton Coleman for
helpful discussions on these issues.

Notes

1. Peacocke 1979b, Lewis 1980, and Bishop 1989, although differing with respect
to the details of their respective accounts of the sensitivity condition (in the case
Agential Systems, Causal Deviance, and Reliability 99

of Lewis, actually dealing with perception), are the key sources for an analysis of
this condition.

2. Here is Myles Brand articulating a very similar assumption: “A person must

perform his own action: no one can perform someone else’s action. A person can
be guided, cajoled, commanded, coerced, even hypnotized, into acting; but never-
theless, if he acts, it is his action” (Brand 1984, 22).

3. John Bishop presents this example in Bishop 1989, 159.

4. The most common cases where an agent transitively causes another agent’s
action are so-called interpersonal interactions (Hart and Honoré 1959). They involve
a first agent causing a second agent’s action and hence producing a causal chain of
events that, as in the present cases, also goes through an agent’s mental events.
However, the main difference between the present cases involving prosthetic agents
and those of interpersonal interaction is that in cases involving prosthetic agents
the relevant chain produces a basic action with the help of someone else’s action,
whereas in interpersonal interactions a causal chain starts with an agent’s basic
action and ends with another agent’s basic action. Nonetheless, these two cases
exhibit the versatility of the CTA with respect to its use of causal transitivity. I have
explored some of the features of such interpersonal interactions in relation to the
exercise and attribution of agency in Aguilar 2007.

5. At this juncture one may be willing to bite the bullet and propose that cases like
the neurophysiologist’s are not deviant. That is, one might propose that the causal
chain satisfies in the relevant way the content of the motivating internal events of
the first agent and that despite its going through a second agent’s action and delib-
eration it nonetheless counts as one of the first agent’s actions. An enriched version
of this move will be considered later and shown to be a rather unappealing
strategy.

6. See Bishop 1989, 171.

7. Other versions inspired on similar systemic approaches to action and agency

include: Enç and Adams 1992; Jeannerod 1997; Clark 1997; Juarrero 1999; and Enç
2003.

8. The inclusion of such sustained character in the production of an action goes

back at least to Irving Thalberg (1984), who in turn was inspired by a critic of the
CTA, namely, Harry Frankfurt (1978/1988). The extensive literature on guiding
intentions directly captures some features of such sustained aspect of many actions.
See, e.g., Searle 1983 and Mele 1992.

9. However, as we will shortly see, internal feedback is problematic. Note also that
feedback is essentially a cognitive feature of a system. Hence, this opens the door
to an epistemic analysis of deviance and reliability that apparently has an impact
on this type of issue in action theory.
100 J. H. Aguilar

10. An interesting question here has to do with the specific mental states associated
with the “mental processes” that Bishop speaks about that are fed by the bodily
movement. Are they the same mental states that started the causing of the bodily
movement, or are they different? If they are the same mental states, then Bishop
needs to defend the idea that such mental states are sustained. If they are not the
same mental states, then he needs to defend the idea that there is some way in
which different mental states (say, different intentions) are able to work in tandem
and respond to feedback information. It seems that the most plausible proposal is
the first one. However, let us note that the questions arise as to how a state is capable
of having the required sustaining nature to count as “the same” intention and how
this intention is supposed to monitor the ensuing behavior. These questions need
to be answered if one is to have a complete account of the elements involved in the
production of a basic intentional action.

11. Note that this move effectively undermines the agential control that the neu-
rophysiologist has, for now she is seen as a neutral satisfier of whatever is intended
by the patient, that is, the “up-to-ness” of her intervention has been reduced to a
bare minimum.

12. Peacocke himself disagrees with this connection between differential explana-
tion and reliability, thinking that reliability is not a way to complement differential
explanation but rather is a rival theory. See, e.g., Peacocke 1979b, 91–95.

13. Hence, Peacocke, when clarifying the nature of the “fixing” relationship
involved in differential explanation, states that “‘fixed’ adverts only to the unique-
ness of determination by the function (as a matter of mathematics in numerical
cases): it does not imply that the laws are not statistical” (Peacocke 1979b, 67). The
laws that Peacocke has in mind are of this general form: (∀x)(∀n)(∀t)((Fxt & Gxnt)
⊃ Hxk (n)(t + δt)), where n ranges over numbers, t over times, and k is a numerical
functor. This does seem to take care of the main concern raised by Sehon (1997),
who concentrates his criticism of Peacocke’s proposal on the possibility of satisfying
particular causal chains of events as opposed to statistical types of causal chains of
events.

14. Here is Peacocke again alluding to this feature: “There need to be many cases
in which conditions producing sensitive chains actually obtain and produce a bodily
movement believed to be a ϕ-ing . . . in order for us to be able to discern an under-
lying pattern of beliefs and desires” (Peacocke 1979b, 109).

15. This picture does not preclude accepting that agential systems can be much
more complex than this simple view, which seems correct for most intentional
bodily movements like walking, eating, or moving an arm. Things get more
complicated when the type of intended action and hence the accompanying inten-
tion require from the agential system things like guidance or wholehearted
commitment.
7 What Are You Causing in Acting?

Rowland Stout

My target for attack in this essay is the fairly widespread view in the phi-
losophy of action that what an agent is doing in acting in a certain kind
of way is causing an event of some corresponding type. On this view
agency is characterized by the agent’s causing of events. To pick one of
many manifestations of this view, here are Maria Alvarez and John Hyman:

We can describe an agent as something or someone that makes things happen. And
we can add that to make something happen is to cause an event of some kind.
(Alvarez and Hyman 1998, 221)

And a particular instance of this view might be the following:

In raising your arm you are causing the event of your arm’s rising.

Such claims about the causal nature of action are sometimes presented as
conceptual claims: claims about when it is correct to describe someone as
performing such an action. But I am interested here in the possibility of
making a constitutive claim: a claim about what such an action is. Actions
seem to be causings in some sense yet to be worked out. The causal theories
that I am questioning take someone’s action of raising their arm to consist
in that person (or perhaps some of his or her mental states or events)
causing the event of his or her arm’s rising.
Despite its widespread philosophical currency, there is something puz-
zling about the idea of causing an event. The relation of causing, like the
property of acting, is not a timeless relation. By this I mean that when we
attribute this relation or property to things we must specify or presume a
time for the attribution. Gavrilo Princip was assassinating the Archduke
Ferdinand at one time but not a year earlier or a year later. But events are
things that are usually predicated timelessly in that sense; standardly,
when philosophers of causation talk about a causal relation between
events, they take it to be a timeless relation. Saying that the assassination
102 R. Stout

of the Archduke Ferdinand caused the First World War, even though we
employ the past tense of the verb “to cause,” is to attribute a relation time-
lessly. It is not that it caused it then and continues to cause it now. Rather,
we can attribute this relation between the events without having to specify
a time for that attribution.
So a standard approach to the philosophy of causation focuses on the
timeless relation of causality holding between events. And that is why it
seems appropriate to think of this in terms of the relation of counterfactual
dependence, for example, which holds timelessly in the same way. But
when we address the constitutive question about actions we are concerned
with processes that happen at one time and not at others. And if what
Princip was doing was causing something, then he was doing it then but
not at other times.
Given this, how are we to understand the claim that he was causing at
some particular time an event—the event of the death of the Archduke or
perhaps the event of the First World War? One way to understand his
causing these events is in terms of his initiating processes, the completion
of which constituted these events.1 We can say that as Princip was squeez-
ing the trigger he was initiating a process in the gun, giving momentum
to the bullet. This process in turn initiated the process of the bullet moving
under its own momentum, which initiated a process of the bullet causing
a perforation of the jugular vein of the Archduke, and then the process of
the Archduke dying as a result of this damage. Perhaps this initiated an
international relations process leading to war being declared and pursued.
Princip set the ball rolling, as it were, by squeezing his finger. Once he had
done this, various mechanisms outside of him took over one after the
other, resulting eventually in the Archduke’s being dead. In this way we
can say that in squeezing the trigger he was causing the event that was the
death of the Archduke. He was initiating these things at one time and not
at others. And his initiating the dying of the Archduke was his action of
killing the Archduke.
The idea under attack in this essay is that all actions are like this.
The target idea is that in acting I am causing an event by initiating (or
perhaps sustaining) a process whose completion is that event. Given this
idea, my role as agent is separate from the process that is initiated. Even
if we start off by identifying my action with me and some event in a causal
relation, the bit that is really associated with my agency does not include
that event itself. We are forced to accept the model of action in which I
do my stuff and then as a result the world does its—a model that forces
agency inward.
What Are You Causing in Acting? 103

But the example of killing someone, which leads to this idea, may have
peculiarities that mean that its treatment cannot be generalized to all
actions. Philosophy of action has an unhealthy obsession with murder. It
also needs to have something to say about phoning someone up, saying
something, going for a walk, eating a healthy lunch, writing a paper,
buying a train ticket, and so on. It is not at all clear that what we should
say about killing people will generalize to these other cases. In particular,
it is a peculiarity of Princip’s action of killing the Archduke that it is an
initiation of a series of processes.
On the face of it, this aspect of doing something and then waiting to
let nature take its course is not shared by all actions. My writing a paper,
buying a train ticket, going for a walk, or saying something are not obvi-
ously cases of initiating processes; in these cases I do not do my bit and
then sit back and let nature take its course. In none of these cases is there
a plausible candidate for being the event that is caused by me as I act. For
example, the event of the paper being written is not the completion of
some process initiated by me as I exercise my agency. It is the completion
of the process of my exercising my agency. It is my action, not some further
event caused as I act. Indeed, even in the assassination case, Princip did
not sit back and let nature take its course. For in reality he did not just
take a shot at the Archduke. If he had missed he would have shot at him
again. Seeing that he had hit him, he went on to shoot the Archduke’s
wife instead.
So on the face of it, many actions are not causings of events. But
this initial rejection of the target idea would be too quick if you thought
that every action is really a moving of parts of one’s body and that
every such moving is a causing of the event of those parts of one’s body
moving. Although there is no plausible candidate for being the event
caused in writing a paper, there seems to be a plausible candidate for being
the event caused when I move my body—namely, the event of my body
moving.
Donald Davidson famously argued for the conclusion that “we never
do more than move our bodies: the rest is up to nature” (Davidson 1980,
59). His argument has two premises. The first premise is that whatever we
do, we do by moving our bodies. The second premise (following Anscombe
1963) identifies our actions with the things by which we do them. So, if
the Queen killed the King by emptying the vial into his ear, her action of
killing the King is the same event as her action of emptying the vial into
his ear. And if she did that by moving her hand in a particular way, then
it is the same event as that movement of her hand.
104 R. Stout

One might deny either premise. In particular, it is not clear that what-
ever we do, we do by moving our body. Think of the action of checking
whether the baby is asleep. There may be some moving of bodies involved
in this. But there is also plenty of watching and listening. And watching
and listening are not done by moving your body. Arguably all action
involves some perceptual feedback of this sort. Or think of the action of
walking. Only when you are relearning to walk after a major injury do you
do it by moving your legs in certain ways. And even then what you do is
more than just change the relative positions of bits of your body; you have
to employ the friction of the surface you are walking on to propel the
weight of your body forward. This is not just putting one foot in front of
the other. Equally, it seems to be the wrong answer to the question, “How
do you write a philosophy paper?” to say “You do it by moving your fingers
in a certain very complicated way.”
But even if we accepted that every action is a moving of parts of one’s
body, to get to the target idea under attack in this essay we would also
have to accept the claim that moving part of one’s body is causing the
event of that body part’s moving. And I want to reject that too. In particu-
lar I want to reject the following claim:

In raising your arm you are causing the event of your arm’s rising.

If such a claim is to stand a chance, the event of your arm’s rising better
be distinct from your action of raising your arm. An action cannot be
identical with the causing of itself. In the final section of this essay I will
challenge the idea that the event of your arm’s rising is usually distinct
from that of your raising your arm. Although there may be odd examples
where we can identify a distinct event of your arm’s rising that is caused
by you as you raise your arm, I will argue that this is not the normal case.
But first I want to question the approach to causality and processes that
drives one to this sort of theory.
It does seem clear that in raising your arm you do cause your arm to
rise. But we can resist the further step to saying that in raising your arm
you cause the event of your arm’s rising. The phrase “your arm to rise” is
not really a noun phrase at all and certainly does not encode some implicit
reference to an entity that is the event of your arm’s rising.
To echo Zeno Vendler’s useful treatment of results and effects, results
are fact-like rather than event-like. Vendler gives the example of a pro-
longed frost in which the water under the pavement turned to ice, which
caused the ground to swell, which caused the pavement to crack (Vendler
1962, 13). The phrase “the ground to swell” can be nominalized to “the
What Are You Causing in Acting? 105

swelling of the ground.” And once nominalized in this way we can describe
it as a result of the water turning to ice. But here we can talk interchange-
ably of the result of the water turning to ice being the fact that the ground
swelled and its being the swelling of the ground. Other ways of understand-
ing the phrase “the swelling of the ground” are more event-like, however.
For example, if we say that the swelling of the ground was gradual, it is
clearly the event, or perhaps process, rather than the fact that is being
described as gradual. With this distinction in mind, the result of your
raising your arm looks like it must be taken to be the fact that your arm
rises, not a particular event or process of rising.
The need to locate an event as the result of a causal process characterizes
what we might think of as a Humean approach to the relationship between
causation and particular happenings in nature. According to this approach,
causation is not to be found in real things but between them. The cause
and effect are taken to be real things, and are usually described as events
in modern Humeanism. But the causing is not taken to be another thing.
In the Humean model, a basic happening is usually understood very
simply as something being in one state at one time and then in a different
state at a subsequent time—so-called Cambridge change. Happenings are
sequences of states. This model perhaps reached its classic formulation in
Russell’s conception of motion. Russell wrote: “Motion is the occupation
by one entity of a continuous series of places at a continuous series of
times” (Russell 1903, section 442). This claim can be extended to processes
generally, so that we have the claim that a process is a series of states of
affairs. For each kind of process there is a characteristic type of series of
states. The obtaining of a succession of such states is the Russellian concep-
tion of a process.
In this model causality is a relation external to happenings—not itself
something that happens but lying instead between those things that do
happen. Causality can only be part of a happening in this Humean model
if the happening consists of a sequence of lesser happenings linked by the
causal relation. But on this model, if what happens is taken to be the sum
of the component happenings, then causality is not really part of what
happens; hence Humean skepticism about causation.
Opposed to the Humean model of causality as a relation between real
things is an Aristotelian approach, which allows causings to be real things.
Causings—or causal processes—are basic constituents of our dynamic
world; causality is internal to happenings. And unlike the Humean model,
since this model takes the causing itself to be an identifiable particular in
the world, there is no need to take the result to be one as well.
106 R. Stout

Applying this to the case of human agency, the answer to the question
of what you cause when you act need not be that you cause some constitu-
ent of the dynamic world—some event or process. What you cause when
you raise your arm is not the process or event of your arm’s rising. What
you cause is your arm to rise, and that need not be taken to be an entity
itself.
If causings or causal processes are identifiable elements of nature, then
they are things that we can identify at one time but which have implica-
tions, conditional on nothing interfering, for what will happen at a later
time. So these things incorporate natural necessity of a sort; when the
causal process is happening, what is present is the conditional necessity
for certain results. This means that there are two aspects to its nature: what
makes it identifiable at the time, and what its existence at that time
requires to be the case afterward.
If you identify one of these dual natures in an object, O, you can see
that O has the property that results, R, will follow if nothing interferes.
You are identifying a conditional necessity in O for R. To put it another
way, you are identifying the actualization of a potentiality in O for R. The
reason for calling it the actualization of a potentiality is that there is often
a need to distinguish between more or less stable intrinsic properties of
the object that contribute to this dual nature and those features that can
be introduced from outside but which also contribute to this dual nature.
We can call O’s having these relatively stable intrinsic features O’s having
a potentiality for R, and O’s having all these features the actualization of
O’s potentiality for R. A car has the potentiality to accelerate under pressure
to the gas pedal, even when the engine is not switched on. That potential-
ity is actualized only if the engine is also running, the gears are engaged,
and so on. But since I am not concerned here with unactualized potentiali-
ties I do not need to pursue this distinction now.
We do not have to limit results to single end-points of causal processes.
A causal process typically has a characteristic sequence of stages as its
result. For example, we might identify the process of an object moving in
a straight line at a certain velocity with the process of that object’s momen-
tum causing it to continue traveling in that line at that velocity. It would
not make sense to identify what is caused here with the process of the
object traveling under its own momentum, since that is the causal process
itself and so cannot also be its result. But nor should we just identify what
is caused with the state of the object being at a certain end-point. The
result is that it continues traveling in a certain direction at a certain veloc-
ity, and this requires that it be not just at the end-point at a later time but
What Are You Causing in Acting? 107

also that it be at intermediate points at intermediate times according to

the standard equations for motion. In other words, the result of the process
of an object traveling under its own momentum is precisely the Russellian
conception of what its motion consists in. So Russell has described not
what motion is but the results of motion—what must obtain for motion
to have happened.
This is why the whole issue can seem so confusing. The object’s traveling
under its own momentum results in its traveling in a certain direction at a
certain velocity. The first instance of the word “traveling” in this sentence
picks out a causal process—an identifiable particular. The second picks out
the sequence of stages that characterizes this process. These stages are not
the process itself but the results it necessitates—what must be the case
for the process to have happened.
Aristotle’s definition of motion in the Physics, Book 3, has seemed con-
fusing to some commentators in just this way: “Motion is the actualisation
of what exists potentially, as such” (1983, 201a10–11). So the process of
building is the actualization of the buildable as buildable. Here I take the
actualization of a potentiality to be nothing more than the complete real-
ization of the conditions that constitute that potentiality. To say that
something is buildable is to say that it has the potentiality for something.
What something? Could it be the process of building itself? This would be
at the cost of making the definition very thin. It would be equivalent to
saying that the process of building was the actuality of something that
when actual was the process of building. Should it then be taken to be the
end-point of the process of building—the state of a house existing, or
perhaps the house itself? No, because the characteristic results of the
process of building are not limited to this. Building a house is more than
just bringing a house into existence. The intermediate stages also belong
to the process.
So we get the claim that a particular process of building is the complete
actualization of the conditions that constitute the potentiality for that
structure of stages that characterizes the result of building. An Aristotelian
process of F-ing is the presence of the potential for the Russellian concep-
tion of the process of F-ing.
To summarize, a process is the realization of a potentiality for certain
results in certain circumstances. You have to describe a structure of stages
to specify the potentiality. These are its characteristic results; they corre-
spond to the Russellian conception of the process itself. But calling this a
process is misleading; it is just the set of things that must obtain for the
process to have happened. What is required for the Aristotelian process to
108 R. Stout

be happening is not just that that structure of stages obtains, but that there
is present a potentiality for such a structure. When a potentiality is fully
present (or actualized), the Russellian process that it is a potentiality for is
not yet complete. The Aristotelian process is fully present; the conditions
for the potentiality are fully realized. But what it is a potentiality for is still
incomplete.2
So a potentiality has two sets of conditions. It has underlying conditions
whose satisfaction means that the potentiality is actualized. And it has the
conditions that characterize its results, whose satisfaction follows from its
being actualized. This essentially dual nature of Aristotelian processes can
be easily missed with certain readings of the idea of the actualization of a
potentiality. If we take a potentiality to be merely a possibility and its
actualization to be nothing more than the thing the possibility is a pos-
sibility of, then this dual nature is lost. The possibility of there being a cup
on the table in front of me is now actualized; there is a cup on the table
in front of me. But nothing is happening.
In the same way if we think of the actualization of a potentiality as the
exercise of a power, then the notion is trivialized. My power to be frighten-
ing is exercised as I reveal my most hideous facial expression; but again
nothing is happening. My exercising my power to be frightening is not
separate from my being frightening. There is nothing gained by describing
my acting as my exercising my power to act. What is crucial for the Aris-
totelian idea to yield a proper notion of a process is that the actualization
of the potentiality be distinct from whatever that potentiality is a poten-
tiality for.
Why has the Aristotelian conception of processes been so unpopular?
Empiricists like Hume assumed that one could not experience potentiali-
ties. The perceivable qualities that were the building blocks of experience
for these early modern empiricists were supposed to be present in indi-
vidual flashes of experience. Potentialities, if they are perceivable at all,
can only be perceived through a process of engagement with the thing
that has the potentialities. A purely passive model of perception, in which
a quality in an object transfers itself to the mind of the perceiver and
imprints itself on that mind, can make no sense of the perception of poten-
tialities. And if it follows that we can have no idea of a potentiality, then
we can have no more idea of the realization of a potentiality.
But if that is the objection to the Aristotelian model it should be
dropped, since the passive model of perception as a kind of imprinting
from world to mind no longer has any currency. You perceive the world
by engaging with it, exploring, interrogating, experimenting, tracking.
What Are You Causing in Acting? 109

There is no reason for doubting that potentialities and their realizations

can be discerned in these ways.
Davidson was opposed to introducing realizations of potentialities into
his metaphysics. The real things for him were the bits at either end of the
causal relation, not the causing itself. He argued that an action like my
raising my arm was an identifiable particular—an event—and he identified
it with the event of my arm rising.3 This was not to identify the raising of
my arm with the causing of the raising of my arm, since Davidson was not
trying for a constitutive causal account of action. For Davidson, the causal
history of an arm’s rising is what makes it correctly describable as an
(intentional) action of raising an arm, but that causal history is not
somehow contained in the action of raising an arm.
But since Jennifer Hornsby’s Actions (1980), we have learned to distin-
guish between arm raisings and arm risings. Hornsby argued that actions
are transitive movements of the body, events distinct from the events that
are intransitive movements of the body. In moving one’s body one causes
one’s body to move, and in doing so one causes the event that is the
intransitive movement of one’s body.

It is a necessary condition of the truth of “a φTs b” that a cause b to φI. In that case
movementsT of the body are events that cause body movementsI. (Hornsby 1980, 13)

Alvarez and Hyman (1998) argue that I am the cause of the event of my
arm’s rising. Since my action of raising my arm is identified with my
causing that event, it cannot be identified with that event itself. They go
on to say that it is not an event at all. In this respect they hold on to part
of the Humean conception of causal processes: events are the effects of
causal processes and sometimes the causes, but they are not the bits in the
middle—the causings. If causings are not allowed any metaphysical iden-
tity, we are faced with the choice of denying that actions are causings and
denying that actions have any metaphysical identity.
If the event of my arm’s rising is distinct from the event of my raising
of my arm, then it might appear to be a good candidate for being the thing
that is caused when I raise my arm. But as Ursula Coope (2007) has recently
argued, my arm rising, if it is taken to be an Aristotelian process, should
be taken to be the very same process as the process of my raising my arm.
And although this claim may seem to fly in the face of Hornsby’s account
of action, we can see that it is very close to something she has been arguing
for too. For Hornsby rejects the idea that the event of an arm rising is
something we could describe as physical rather than as mental. The event
of the arm rising is, then, to be identified not with a series of changes in
110 R. Stout

position of the arm, but as something that may essentially involve the
agent.
This is expressed clearly in the postscript to “Bodily Movements, Actions,
and Epistemology” (Hornsby 1997, 102 ff.), where she considers a disjunc-
tive approach to bodily movements. She raises the question of whether a
bodily movement that is just a reflex or the result of some external manipu-
lation could have been the sort of movement that is associated with action,
and she answers that it is fairly evident that it could not (ibid., 103).
Despite the fact that one might be ignorant as to whether a movement is
or is not associated with the action of an agent, whether it is or is not
associated with an action is essential to its identity.
Hornsby cannot here be identifying an arm rising with a series of posi-
tions of the arm through the air. She is not thinking of the Russellian
conception of a process of arm-rising. For there is nothing essential to the
series of states that an arm is in when it rises that links the rising with an
action. The natural alternative is that an arm rising is being considered as
an Aristotelian process; it is the realization of an arm-rising potentiality.
It is a different potentiality that is realized when an arm rises as a result
of the agency of the owner of that arm from the potentiality that is real-
ized when an arm rises as a reflex or because of external manipulation.
And it seems reasonable to say that the realizations of these different
potentialities are also different.
It does not follow that there is no process in common between these
processes. There might be a highest common factor between an arm rising
as a result of external manipulation and an arm rising as a result of the
arm owner’s agency. If there were, then this might count as a neutral arm-
rising. Where might this potentiality be? Do one’s muscles have the poten-
tiality to raise one’s arm? No. They have the potentiality to shorten the
distance between the points at each end of the muscles, but that is some-
thing else.
One might try to argue that the system of muscles in the arm and
shoulder has the potentiality to raise the arm inasmuch as under certain
circumstances of electronic nerve inputs to these muscles the arm will rise.
But in fact this is not the case. Those nerve inputs only result in the arm
going up if the arm is oriented in exactly one way at the start of the
process—both with respect to the body and with respect to gravity—and
has the precise weight it has, the muscles have precisely the responsiveness
they have, and so on. If any of these factors is different then that set of
nerve inputs applied to the muscle mechanism will result in something
quite different from the arm going up.
What Are You Causing in Acting? 111

What makes the arm go up is a mechanism that guides that movement

in response to feedback from proprioception. It is a control mechanism.
So might there be a “subpersonal” control mechanism for arm raising?
When my arm rises, might there be something less than me that is control-
ling its rise?
I think it would be very difficult to defend the idea that there was such
a subpersonal mechanism raising my arm when I raise my arm. What
would my relationship with that mechanism be? Would I be giving it
instructions? But in that case we would say that my action was to instruct
my arm-raising mechanism to go into action. And this does not seem right.
This is not to deny that the process of raising one’s arm involves other
automatic or subpersonal processes. For example, it seems that there are
stages in my arm’s rising in which my arm just continues with a certain
trajectory with no active involvement by me. The point is that the feedback
required for the overall process to be properly guided happens at the level
of the person (whether he or she is thinking about it or not). It is me who
adjusts the movements in accordance with this feedback, rather than any-
thing less than me.4
So, we should accept Hornsby’s disjunctive conception of a bodily
movement. My arm’s rising, at least in normal cases, might be either an
arm-rising of an agent or be a movement that is not associated with action,
and there is no process in common between them. Hornsby, however,
rejects a way that the disjunctive conception of bodily movement might
be more strictly analogous with the disjunctive conception of perceptual
appearance that John McDowell is associated with. For McDowell, “an
appearance that such-and-such is the case can be either a mere appearance
or the fact that such-and-such is the case making itself perceptually mani-
fest to someone” (McDowell 1982, 472). So a strictly analogous disjunctive
approach to bodily movements would say that a bodily movement can be
either a mere bodily movement or the agent doing something—an action.
For Hornsby the second disjunct is not the bodily movement being an
action but the bodily movement being an action-associated movement;
hence the disanalogy.
Coope (2007) has argued for the stronger claim—that the event of an
arm’s rising can itself be the event of an agent raising their arm. She
endorses what she takes to be Aristotle’s view that the process of raising
my arm is the very same process as that of my arm rising. If we accept,
with Hornsby, that the agent may be essentially involved in the process
of his or her arm rising, then at least one objection to Coope’s identifica-
tion is lost. The process of my arm rising is for neither of these writers a
112 R. Stout

series of stages of my arm’s position in space. So, why not identify it with
my action of raising my arm?5
It might be thought that the process of my arm rising must be different
from the process of my raising my arm since they have different agents.
One is the process of me doing something; the other is a process of my
arm doing something. But my arm is the patient rather than the agent in
this process. It is not raising itself; it is being raised by me. My arm rising
under my agency is the same process as my raising my arm, just as the
butter melting under the sun’s agency is the same process as the sun
melting the butter. This appears to be Aristotle’s view (Physics, Book 3,
chapter 3; and see Coope 2007, 123–124). So the process of my arm rising
is the process of my arm rising under some agency. There are not two
processes occurring here; there are not two potentialities being realized or
two mechanisms working.
If this is right then what is caused when I raise my arm is not normally
the process of my arm’s rising. Although I cause my arm to rise I do not
normally cause the process of my arm’s rising. And if my raising my arm
is correctly construed as the realization of some potentiality in me, then
this potentiality is not the potentiality for the process of my arm rising,
also construed as the realization of some potentiality.
So, what is the potentiality whose realization is my raising my arm (or
my arm rising) a potentiality for? Coope (2007, 114) argues that it is not
the potentiality for another process, but the potentiality for a state to
obtain. In particular, it is the potentiality for my arm to be up. But this is
too simple. For it is essential to my raising my arm that my arm pass
through all the appropriate stages of a rising. The complete realization of
the potentiality cannot be characterized just by an end-state. Suppose that
I get my arm to be up by pulling it in close to my body and then shooting
it out again at a higher angle. I haven’t in this case raised my arm though
I have done something that results in it being up.
What seems to be the natural candidate for the job of being the thing
that my arm-raising potentiality is a potentiality for is the Russellian con-
ception of the process of my arm rising. It is not just the end-state of my
arm being up but a particular kind of structure of stages between my arm
being down and my arm being up. There is a structure of stages character-
istic of an arm-rising, and my raising my arm is the realization of a poten-
tiality for a series of states that match that characteristic structure.
So I propose to extend Coope’s account by saying that the process of
my raising my arm or my arm rising is the realization of a potentiality for
the arm to be in a series of states characteristic of arm-rising, rather than
What Are You Causing in Acting? 113

just being the realization of a potentiality for the arm to be up. In raising
my arm and realizing that potentiality, I am causing my arm to be in that
characteristic series of stages. But I am not initiating a separate process of
my arm’s rising, nor am I causing the event of my arm’s rising.

Notes

1. Another way to understand this might be in terms of sustaining processes—by

removing obstacles to these processes or by ensuring that the necessary underlying
conditions of the processes are in place. What I have to say about initiating processes
would by and large transfer to sustaining processes.

2. This might make sense of Aristotle’s claim that processes are incomplete actual-
izations of potentialities (1983, 201b31–33). What is incomplete is not the degree
to which the potentiality is actualized but the structure of stages that characterizes
what the potentiality is for.

3. In “Agency” (Davidson 1980, chapter 3), he makes no distinction between my

raising my arm and my arm rising. And, in “Problems in the Explanation of Action”
(Davidson 2004, chapter 7), he explicitly identifies them: “If I raise my arm, then
my raising my arm and my arm rising are one and the same event” (Davidson 2004,
103).

4. Even if there are deliberate bodily movements that do just consist in instructing
subpersonal mechanisms to do their stuff, it would be absurd to generalize this to
all bodily movements, including controlled movements like raising one’s arm. I may
be able to make my arm move, though not in a very controlled way, by initiating
some process of arm-moving. But this, like the example of Princip’s assassination
of the Archduke, would be a rather special case.

5. Adrian Haddock (2005) also argues in this way and recommends that Hornsby
adopt the stronger disjunctive approach to body movements.
8 Omissions and Causalism

Carolina Sartorio

1 Introduction

Omissions are puzzling—so puzzling that people tend to say puzzling

things about them and give up otherwise attractive philosophical theories
in order to accommodate them.1 In this essay I suggest that omissions make
trouble—serious trouble, and trouble of a new, sui generis kind—for “causal-
ism,” the standard view or family of views of agency. In particular, I am
interested in causalism as an attempt to explain what it is for an agent to
behave intentionally. I will argue that causalism cannot accommodate
intentional omissions—or, at least, it cannot account for them in the same
way it accounts for (positive) actions. As a result, causalism is incomplete—
or, at best, highly disjunctive—as a theory of what it is to behave
intentionally.
I will bypass the question whether omissions can be, properly speaking,
actions—“negative actions” or “active nondoings,” as they have been
called (see, e.g., Kleinig 1976). For some people (notably, Thomson 1977),
actions are a subclass of events, where events are particulars with specific
spatiotemporal locations, intrinsic properties, and so on. On this kind of
view, it’s hard to count omissions as actions, for omissions don’t appear
to have specific spatiotemporal locations, intrinsic properties, and so on.
Nevertheless, even if omissions aren’t actions, it seems that agents can still
fail to do things intentionally, and it makes sense to ask under what condi-
tions an agent’s not doing something is intentional (see, e.g., Ginet 2004).
Thus, even if omissions aren’t actions, a theory of what it is to behave
intentionally should be able to accommodate omissions. (Note that, if
omissions aren’t actions, a theory of what it is to behave intentionally is
not the same thing as a theory of what it is to perform an intentional
action, and it might not even be the same thing as a theory of what it is
to act intentionally.)2
116 C. Sartorio

Causalism, as a theory of what it is to behave intentionally, is the view

that an agent behaves intentionally when certain events/states involving
the agent’s body (such as the agent’s moving in a certain way) are appro-
priately caused (nondeviantly caused, or caused in the “normal” way) by
certain mental events or states of the agent, in particular, the agent’s inten-
tions, belief-desire pairs, decisions, and so on. The causal link between
those mental events or states and the bodily events or states singles out
the specific reasons for which the agent behaves in the relevant way from
the possibly more inclusive set of reasons that he had for behaving in
that way. Causalism is traditionally attributed to Davidson (Davidson
1963/1980), and it is the most commonly held view of agency nowadays.
Among the mental items that cause the relevant bodily movements/states,
causalists seem to agree, intentions are special in that they play the most
central role. For intentions are those mental states by which the agent
settles on a particular course of action: they initiate and guide behavior.3
Hence my focus will be on intentions as the relevant mental items: I will
take causalism to be the view that, when an agent behaves intentionally,
the agent’s intending to behave in a certain way, or the agent’s forming
an intention to behave in a certain way, appropriately causes the relevant
bodily movement/state.
As noted, just as agents can do things intentionally (these are the agents’
“positive” intentional actions), they can also fail to do things intentionally
(these are the agents’ intentional omissions). In fact, it seems that there
are many things that we fail to do intentionally. This is so even though,
as Ginet (2004, 95) points out, there seems to be an important asymmetry
between actions and omissions in that, whereas most things we do are
things we do intentionally, most things we don’t do are not things we
don’t do intentionally (this is, presumably, because, for anything we do,
there are several things we don’t do). As a paradigm example of intention-
ally omitting to do something, consider the following case:

Drowning Child A child is drowning in a nearby pond. I could jump in

and save him. However, after deliberating about it for a bit, I choose not
to jump in and to eat an ice cream instead.
In this case, I intentionally omit to jump into the water to save the drown-
ing child. Hence causalism should tell us in virtue of what this is so. In
general, causalism should tell us what makes an agent’s omission inten-
tional, when it is intentional.
Surprisingly, very little has been said about omissions in connection
with causalism.4 Davidson, in particular, confesses to have omitted address-
Omissions and Causalism 117

ing this issue (presumably, intentionally!) in a reply to Vermazen—which

I discuss briefly below (Davidson 1985, 217).5 It is particularly surprising
that so little has been said about omissions and causalism, for the causal
status of omissions and other absences is a highly debated issue in the
metaphysics of causation. As I have pointed out, on a natural view of
events, omissions (and absences in general) are not events, but absences
thereof. However, on a familiar view of causation, only events can be
causes and effects.6 If so, it seems that causalism cannot account for inten-
tional omissions, in particular, it cannot account for my omission in
Drowning Child.
How can a causalist try to address this problem? Even if omissions aren’t
events, there at least three different things the causalist could say.
First, the causalist can say that other things besides events can be causes
and effects—notably, facts—and that, moreover, causal talk involving facts
is the most “primitive” kind of causal talk: any other kind of causal talk,
such as causal talk involving events, is made true by causal talk involving
facts.7 For example, an event can be said to cause another event because
the fact that the first event occurred caused the fact that the second event
occurred. On the basis of this view, a causalist could say that my failure
to jump into the water in Drowning Child is an intentional omission
because the fact that I formed the intention not to jump in appropriately
caused the fact that I didn’t jump in. More generally, a causalist could say
that an agent intentionally omits to do something just in case the fact that
he formed an intention with the relevant content appropriately caused the
fact that his body didn’t move in a certain way.
Second, a causalist could claim that other things besides events can be
causes and effects but causal talk involving events is still the most basic
kind of causal talk. In particular, causal talk involving omissions and other
absences can be true, but it is made true, ultimately, by causal talk involv-
ing events. This is Vermazen’s suggestion (Vermazen 1985), which David-
son explicitly embraces in his reply to Vermazen (Davidson 1985). How
can a causalist do this? Roughly, Vermazen’s idea is the following. Imagine
that I am tempted to eat some fattening morsels, but I refrain. Then my
passing on the morsels is an intentional omission because the relevant
mental states/events (pro-attitudes, intentions, etc.) cause my not eating
the morsels, and this is, in turn, because, had those mental states been
absent, then some other mental states/events (competing pro-attitudes,
intentions, etc.) would have caused my eating the morsels. In other words,
actual causal talk involving omissions is made true by counterfactual causal
talk involving positive occurrences or events.
118 C. Sartorio

Third, a causalist can claim that there are two (or maybe more) concepts
of causation, and that omissions and other absences can only be causes
and effects in the sense captured by only one (or some) of those concepts.
For example, it could be argued that there is a “productive” concept of
cause and a “counterfactual” concept of cause (as in Hall 2004), and that
omissions can be causes and effects in the counterfactual sense but not in
the productive sense. Still, to the extent that both concepts are genuine
concepts of causation, it is open to the causalist to say that an agent behaves
intentionally when his moving in a certain way, or his not moving in a
certain way, is caused by the agent’s intentions in the normal way.
On any of these views, then, what makes an omission intentional is
similar to what makes an action (a “positive” action) intentional: the fact
that a relevant piece of behavior (positive or negative) is caused by the
agent’s intentions in the normal way. For example, my failure to jump
into the water in Drowning Child is an intentional omission because I
formed the relevant intention not to jump in and such intention caused
my not jumping in, in the normal way. This is parallel to the way in which,
if I had intentionally jumped into the water to save the child, my forming
the opposite intention (the intention to jump in) would have caused the
bodily movement consisting in my jumping in, in the normal way. As we
have seen, there are different ways in which a causalist can resolve the
issue of how omissions can be causes and effects. But, to the extent that
omissions can be causes and effects, it might seem that causalism has the
resources to account for intentional omissions in basically the same way
it accounts for intentional (positive) actions.8
In what follows I argue that omissions pose a recalcitrant problem for
causalism, that is to say, a problem that persists even under the assumption
that omissions can be causes and effects in any of the ways outlined above.
Interestingly, it is a problem that bears some similarities to what can be
construed as a different challenge to the view: the challenge of the causal
exclusion of the mental by the physical (Kim 1993). This is because the
recalcitrant problem of omissions can be seen as an exclusion problem.9
Briefly, the exclusion problem for the mental and the physical is this.
According to nonreductive physicalism, a widely held view in the philoso-
phy of mind, mental states are realized by, but not identical to, physical
states. For any piece of behavior that a mental state allegedly causes, there
is an alternative explanation that appeals only to the underlying physical
state. We want to say that the physical world is “causally closed,” and thus
that the physical state is a cause of the behavior. Hence, it is tempting to
conclude that the mental states don’t really do any causal work. And, if
Omissions and Causalism 119

so, causalism doesn’t seem to get off the ground. Many people think that
this problem is not intractable.10 But what I will suggest is that the problem
that omissions pose for causalism is an exclusion problem of its own: one
that threatens not to show that mental states in general are causally inef-
ficacious, but only that, in the specific case of omissions, the relevant
mental states (in particular, intentions) cannot do the causal work that the
causalist would want them to do. For there is an alternative, and arguably
better, explanation that doesn’t appeal to those mental states, even if
mental states in general are causally efficacious, and even if omissions in
general are causes and effects.

2 The Exclusion Problem for Omissions

As we have seen, in Drowning Child, the causalist seems to be committed

to (roughly) the truth of the following claim:

(Claim 1) My forming the intention not to jump in causes my failure to

jump in.

I say “roughly” because many causalists would reject the idea that inten-
tionally φ-ing requires forming an intention to φ. Still, the consensus is that
a closely related intention is required.11 For simplicity, I will assume that
the intention in question is the intention not to jump in.
At first sight (again, assuming that there is no problem with absences
being causes and effects, or with mental events and states in general being
causes and effects), Claim 1 seems very plausible: it seems natural to say
that I didn’t jump into the water because I formed the intention not to do
so. On the face of it, intentions (and other mental events or states) can
cause people not to do things just as they can cause them to do things.
For instance, it seems that my abstaining from voting in an election can
be the result of a careful process of deliberation ending in my forming the
intention not to vote, just like my voting for a certain candidate can be
the result of a careful process of deliberation ending in my forming the
intention to vote for that particular candidate. Thus it might seem that,
once we resolve the issue of how omissions can be causes and effects, and
the issue of how mental events and states can be causes and effects, the
claim that causalism can account for intentional omissions in the same
way it accounts for (positive) actions is very plausible. I will argue, however,
that this view is misguided and that Claim 1 should be rejected.
I said that I would bypass the question of whether omissions should be
regarded as actions in their own right, on a par with “positive” actions. By
120 C. Sartorio

this I meant the question of whether we should take nondoings of certain

sorts to be actions (“negative” actions). But what I have been assuming so
far is that omissions are not just identical to positive actions. In other words,
I have been assuming that, even if omissions were actions, they wouldn’t
be actions because nondoings just are doings of certain kinds (rather,
because certain kinds of nondoings are also actions). Of course, if omissions
were simply identical with positive actions, then the question of whether
omissions can be causes and effects wouldn’t arise: it would be uncontro-
versially true that they can, for positive actions are positive occurrences
and positive occurrences can clearly be causes and effects. In what follows,
I reserve the word “action” for positive actions.
The assumption that omissions are not identical to actions requires, in
particular, that we distinguish an agent’s omission from anything that the
agent might have done instead of the action omitted. For instance, in
Drowning Child, my failing to jump in should be distinguished from my
eating ice cream on the shore at the time when I could have been jumping
in to save the child. On the face of it, this is a reasonable assumption: at
least generally, my failing to do something doesn’t seem to be identical to
my doing something. In particular, although I failed to jump in by eating
ice cream, my failure to jump in isn’t my eating ice cream.12 In support of
this idea, note that it seems that I could have failed to jump in by doing
something other than eating ice cream on the shore, for example, by
reading a book. This is a reason not to identify the omission with the
action. Another reason not to identify them is that they seem to have dif-
ferent causal powers. For instance, it seems that my failure to jump in
didn’t cause my stomachache later that day, but my eating ice cream did.
Finally, sometimes there seems to be no action with which to identify the
omission—or, in general, no positive occurrence with which to identify an
absence that appears to be causally efficacious. In those cases it seems that
the causal story would be incomplete without reference to an omission, or
an absence of some sort. Imagine that the zookeeper promised to get an
elephant for the local zoo but he failed. This made Jimmy sad. It seems
that there isn’t anything that the zookeeper did or anything that actually
happened that made Jimmy sad. We don’t want to say, for instance, that,
when Jimmy visited the zoo, the presence of a rhinoceros made him sad.
The presence of a rhinoceros didn’t make him sad; the absence of an ele-
phant did.13 At any rate, this will be an assumption of this essay: that
omissions aren’t identical with actions, or at least not generally. In particu-
lar, my omitting to jump in is not identical with my action of eating ice
cream on the shore in Drowning Child.14
Omissions and Causalism 121

Now, it seems that, if we should distinguish between my eating ice

cream and my omitting to jump in, then we should also distinguish
between my forming the intention not to jump in and my omitting to form the
opposite intention (the intention to jump in). In other words, just as there is
something I did and something I didn’t do at the level of overt or bodily
acts (I ate ice cream, and I omitted to jump in), there is also something I
did and something I didn’t do mentally (I formed the intention not to
jump in, and I omitted to form the intention to jump in). Call my forming
the intention not to jump in “A1,” my omitting to form the intention to
jump in “O1,” and my omitting to jump in “O2.” As we have seen, the
causalist would want to suggest that A1 causes O2 (this was Claim 1). But
consider, as an alternative:

(Claim 2) O1 causes O2.

Whereas Claim 1 says that the cause of O2 is what I did (mentally), Claim
2 says that it is what I omitted to do (mentally). Which one is more likely
to be true? Or can both of them be true simultaneously? In the next section
I argue for the truth of Claim 2 and for the idea that Claim 2’s truth
threatens to undermine Claim 1’s truth. I will call this thesis the thesis of
“causal exclusion for omissions” (CEO).

3 Argument for CEO

Start by focusing on bodily actions and omissions. As I have pointed out,

it is natural to draw a distinction between O2 (my omitting to jump into
the water, a bodily omission) and what I did instead of jumping in, for
example, my eating ice cream on the shore (a bodily action, call it “A2”).
But then consider the question: What caused the child’s death? Did O2 cause
it? Did A2 cause it? On the assumption that omissions can be causes and
effects, it seems clear that O2 was a cause of the child’s death: the child
died because I omitted to jump into the water to save him. Should we
think that A2 also caused it? Presumably not. For, intuitively, the child
died because of what I didn’t do, not because of what I did in its place. It
seems, in fact, irrelevant that I was eating ice cream on the shore (as
opposed to, say, reading a book, or doing anything else but jumping in):
all that matters is that I failed to jump in to save him.
In other words, consider the following claims:

(Claim 3) A2 caused the child’s death.

(Claim 4) O2 caused the child’s death.

122 C. Sartorio

The first premise of the argument reads:

(P1) Claim 4 is true and its truth undermines the truth of Claim 3.

Now, the argument continues, if the truth of Claim 4 is enough to cast

doubt on Claim 3, then, by the same token, the truth of Claim 2 should
be enough to cast doubt on Claim 1. For, again, on the assumption that
omissions can be causes, Claim 2 seems clearly true: O1 caused O2. I
omitted to jump in because I omitted to intend to jump in. And it seems
that we shouldn’t say that A1 (my forming the intention not to jump in)
also caused O2. For, again, I failed to jump in because of what I omitted
to intend to do, not because of what I intended to do. It seems, in fact,
irrelevant that I actually formed the opposite intention: all that seems
relevant is that I omitted to form the intention to jump in.15
Thus the second premise of the argument reads:

(P2) If P1 is true, then Claim 2 is true and its truth undermines the truth
of Claim 1.

From which the conclusion follows:

(C) CEO is true.

In other words, the argument suggests that the best way of conceiving
my relationship to the outcome of the child’s death is as a negative rela-
tionship throughout the causal chain. This includes my mental behavior:
the child died because of what I omitted to do, including what I omitted
to intend to do. Even if I also formed a positive intention not to be involved
in certain ways, the fact that I formed that intention seems causally irrel-
evant; all that was causally relevant is the fact that I omitted to intend to
be involved in certain ways. The argument relies heavily on an analogy
between bodily acts and mental acts. The main claim is that, if what
accounts for the outcome of the child’s death is what I didn’t do “extra-
mentally,” then what accounts for what I didn’t do extramentally is, in
turn, what I didn’t do—this time, mentally.
An important clarification is in order. I don’t mean to suggest that omis-
sions can only have other omissions as causes—or, in general, that absences
can only be caused by other absences. All I want to suggest is that this is
true of the type of situation that is our focus here. It is certainly possible
for omissions—and for absences in general—to have positive occurrences
as causes. Imagine that, besides not jumping in myself, I talked the life-
guard into thinking that it is not worth risking one’s own life to save other
people’s lives, and, as a result, the lifeguard also failed to jump in. In this
case my talking to the lifeguard (an action) caused his omission. Or imagine
Omissions and Causalism 123

that yesterday I wrote a note to myself reminding me how much I hate

water. Had I not seen the note today, I would have decided to jump in to
save the child, but seeing the note today stopped me from doing that. In
this case my writing the note (an action) caused my omission.
Why is it that in these versions of the drowning child case, but not in
the original version, an omission is caused by a positive occurrence or an
action? The answer is that in these versions of the case a positive interven-
tion is needed to “counteract” the current train of events. In the lifeguard
version, the lifeguard would not have intentionally omitted to save the
child had it not been for what I said to him: what I failed to do isn’t suf-
ficient to account for his failure to jump in. And, in the self-addressed note
version, I would not have intentionally omitted to jump in had it not been
for the note: again, what I failed to do isn’t sufficient to account for my
failure to jump in. By contrast, in the original version of Drowning Child
(and, more generally, in paradigmatic or “ordinary” omission cases) the
agent’s omission simply seems to “flow from” other things the agent omits
to do—in a similar way, I take it, that the absence of elephants from a
room at a given time is accounted for by the absence of elephants from
the room an instant earlier. So it is certainly possible for an omission to
be caused by something other than an omission; all I am claiming is that
this is not true of, for example, Drowning Child and other paradigmatic
cases of intentional omission.16
If the argument is sound, then the causalist faces an exclusion problem
for omissions. An enlightening way to put the problem is the following.
Whereas, in Drowning Child, the causalist would want to say that my
omitting to jump in stems from my forming certain malevolent (or other-
wise morally deficient) intentions, I have argued that we should regard it
as flowing from my omitting to form certain benevolent (or otherwise
morally virtuous) intentions. Importantly, there is no similar problem for
actions, on the face of it. Whereas my forming the malevolent intention
and my omitting to form the benevolent intention seem to compete for
their causal role in the case of my omission, there is no such competition
in the case of an ordinary action. Suppose I form the intention to shoot
my enemy and this leads me to pulling the trigger. Here, clearly, my
forming the malevolent intention plays a key role: what I do extramentally
flows from what I do mentally.
In what sense is the problem for omissions an exclusion problem? In the
sense that, once one recognizes the distinction between actions and omis-
sions and everything that it entails, the mental items singled out by the
causalist as causes of the relevant bodily states (i.e., the relevant intentions)
124 C. Sartorio

are excluded by other items. Those other items are better suited to play
the relevant causal role than the candidates identified by the causalist.
Crucially, the problem for omissions doesn’t rest on a general “exclusion
principle” according to which no phenomenon can have more than one
sufficient cause, or on the claim that there is no widespread overdetermina-
tion, or on any other claim in the vicinity. In this sense the exclusion
problem for omissions is very much unlike the traditional exclusion
problem for the mental and the physical, as it is typically laid out in the
literature.17
What does the argument for CEO rely on, if not a general exclusion
principle? As I pointed out, it relies on an important analogy between
bodily and mental items. The claim is that, given what we want to say
about the causal powers of the bodily items, we should say something
similar about the causal powers of the mental items. In particular, given
that my eating ice cream isn’t a cause of the death (my failing to jump in
is), my intending not to jump in also isn’t a cause of my omitting to jump
in (my omitting to intend to jump in is). This is so even if, at first sight,
the claim that the intention had those causal powers seemed plausible.
What justifies the claim about the causal powers of the bodily items, to
begin with? That is, what justifies the claim that my eating ice cream didn’t
cause the child’s death, but, instead, my failure to jump in did? There are
several things one could say to answer this question. But, on the face of
it, it seems enough to point out that, on the assumption that omissions
can be causes, the view that my failure to jump in is a cause of the death
and my eating ice cream isn’t is very intuitively plausible (as suggested
above). Again, on the face of it, there are certain things that I cause in
virtue of eating ice cream and there are other things that I cause in virtue
of not jumping into the water. Perhaps there are also other things that I
cause in virtue of both eating ice cream and failing to jump in (maybe my
remaining above my ideal weight, if I would have weighed less by dieting
or exercising?). But certainly not everything I cause in virtue of eating ice
cream is something that I cause in virtue of failing to jump in, or vice
versa. In particular, just as it seems that I cause myself to feel sick to my
stomach by eating ice cream, and not by failing to jump in, conversely, it
seems that I cause the child to die by failing to jump in, and not by eating
ice cream. Again, this is not motivated by a general exclusion principle of
any sort: it’s just a claim that seems very plausible on its own.18 (More on
the causal powers of bodily actions and omissions in section 5.)
This concludes my discussion of the argument for CEO. How could the
causalist try to respond to the argument? In the following sections I discuss
Omissions and Causalism 125

two possible responses by the causalist. The first response is an attempt to

disarm the analogy between bodily and mental acts; the second response
is an attack on the claim about bodily acts.

4 First Response: Cause Essentialism

First, the causalist might want to reply in the following way. An event
consisting in my arm moving is not an action if it was the result of
someone else’s grabbing my arm and making it move in a certain way; in
that case it is a “nonactional” event, a mere bodily movement (something
that merely “happens” to the agent, as opposed to something that the
agent does). To borrow an analogy by Mele,19 an intrinsic duplicate of a
US dollar bill fails to be a genuine bill if it is not the output of a certain
causal process involving the US Treasury Department (e.g., if it is counter-
feit); similarly, an event fails to be an action if it is not the output of a
causal process involving mental items of a particular kind. In particular,
the causalist would want to say, it is not an action unless it is the output
of a causal process involving intentions of the relevant kind. And the same
goes for (intentional) omissions, the causalist might claim: my failing to
jump into the water in Drowning Child would not be intentional unless
it were caused by a relevant intention in the relevant way. Imagine that I
didn’t jump in because someone restrained me when I was about to do so.
In that case, the causalist would say, I didn’t intentionally fail to jump in.
Although it is true that I didn’t jump in, my not jumping in isn’t an inten-
tional omission but a nonactional state (a mere “bodily state,” something
that “happens” to me, but not something I intentionally omit to do).
In other words, the objection is that the analogy on which the argument
for CEO rests breaks down: although we don’t have reason to believe that
my eating ice cream causes the child’s death (all the work is plausibly done
by my failing to jump in), we do have reason to believe that an intention
with a relevant content causes my failure to jump in. For this failure is not
any failure: it is an intentional failure, and it would not have been inten-
tional unless it was caused by a relevant intention in the relevant way.
However, this objection fails. I agree that my not jumping in wouldn’t
have been intentional if someone had been restraining me the whole time,
just like I wouldn’t have intentionally raised my arm if someone had forced
my arm upward. But this isn’t enough to show that I wouldn’t have inten-
tionally failed to jump in unless A1 (my forming the intention not to jump
in), or my forming a similar intention, had caused it. Why not? Because
it is very plausible to think that my failure to jump would be intentional
126 C. Sartorio

if O1 (my omitting to intend to jump in) caused it. For O1 is itself an

intentional omission: I voluntarily failed to form that intention, after
deliberating about whether to do so, after considering reasons for and
against doing so, and so on. And if I fail to jump in as a result of my
intentionally omitting to intend to jump in, then, presumably, my failing
to jump in is intentional too. In other words, if I am right and O1 causes
my failure to jump in, then this by itself helps explain why that failure was
intentional; we don’t need to say, in addition, that my intention not to
jump in caused it.
The causalist could protest that this isn’t a satisfying answer. For he
could say that the same question arises in connection with O1: what makes
it intentional, if not the presence of an intention?
In response, note that there are two different claims that the causalist
wants to make in the case of intentional omissions. First, the causalist
wants to say that some intention has to exist in order for an agent to omit
to do something intentionally (call this the existential claim). Second, the
causalist wants to say that such an intention has to cause the relevant
bodily nonmovement (call this the causal claim). Clearly, unless the exis-
tential claim is true, the causal claim cannot be true. But the existential
claim can be true and the causal claim still fail to be true. The argument
for CEO from the last section is an argument against the causal claim only.
For all the argument says, it might be that the existential claim is true:
perhaps some intention needs to exist in order for my omission to jump
in to be intentional.20 Imagine, for instance, that my failure to intend to
jump in would not be intentional unless I actually formed the opposite
intention, the intention not to jump in. If that were so, then I would claim
that, although the relevant intention needs to exist for my omission to be
intentional, the argument still shows that it doesn’t do the causal work
that the causalist says it does. There might still be a sense in which it would
be true to say, in that case, that the relevant intention is part of what
“makes” my omission to jump in intentional. But this wouldn’t be because
the intention causes the nonmovement, as the causalist claims; it would
only be because the nonmovement wouldn’t have been intentional in the
absence of such an intention.21
I conclude that the objection fails to establish that an omission is inten-
tional unless it is caused by an intention. It seems, in fact, plausible that
an omission could be intentional even if it were not caused by an inten-
tion. The question of whether an intention with the relevant content needs
to exist in order for an omission to be intentional is a separate question,
which we may set aside here. Naturally, if no such intention were even
Omissions and Causalism 127

needed, then this would be an independent problem for causalism. But,

even if an intention of that type had to exist, it still wouldn’t follow that
it causes the nonmovement, as the causalist claims.22

5 Second Response: Happy Coexistence

Alternatively, the causalist might want to object to the claim about bodily
acts on which the argument for CEO rests: the claim that my eating ice
cream (A2) isn’t a cause of the child’s death. One way in which the causal-
ist could try to make this reply is this. As I have suggested, the child died
because I didn’t jump in to save him. However, I didn’t jump in to save
him, in turn, because I was eating ice cream on the shore (since, given that
I was eating ice cream on the shore, I couldn’t have been jumping in).
Therefore, by transitivity, the child died because I was eating ice cream on
the shore.
The main problem with this suggestion is that, even if all of this were
right, it still wouldn’t follow that A2 caused the child’s death. For consider
the claim that I didn’t jump in to save the child (at t) because I was eating
ice cream on the shore (at t). If this claim is true, there is an explanatory
connection between A2 and O2.23 But this explanatory connection is non-
causal. (For one thing, A2 and O2 obtain simultaneously, whereas it is
generally thought that causes precede their effects.) So, even if it were true
that the child died because I was eating ice cream on the shore, it still
wouldn’t follow that A2 caused the child’s death.
Alternatively, the causalist might want to suggest that A2 caused the
child’s death, although it did so “directly” (i.e., not by way of causing O2).
However, I find this reply unmotivated. Anscombe dismissed a similar view
in a two-sentence paper.24 But I am going to try to do (a bit) more to con-
vince you that this view is not very plausible.
Why would anyone be tempted by this view? One might think that
there is some intuitive support for it. Imagine that Jim spent the night
previous to the exam partying instead of studying, and then he flunked
the exam on the following day. We are tempted to say: “Jim’s partying the
night before the exam caused him to flunk it” (instead of, in my view, the
more appropriate claim: “His failing to study the night before the exam caused
him to flunk it”). But, should we take this literally? Should we think, on
this basis, that Jim’s partying was also a cause of his flunking the exam?
Or should we think that we are speaking loosely in claiming that it was?
Here is an argument that we should think the latter. As I am imagining
the example, to the extent that we judge that Jim’s partying caused his
128 C. Sartorio

flunking the exam, it’s because he was parting instead of studying (not
because, say, too much partying impaired his writing or thinking capaci-
ties, which were a necessary requirement for doing well on the exam). But
then, by the same token, anything else that he could have done instead
of studying would be a cause too, in the corresponding scenario. In par-
ticular, had Jim been caring for convalescent Grandma all night long
instead of partying, his caring for Grandma would have caused him to
flunk the exam. Also, had he been reading a book on how to pass exams,
his reading such a book would have caused him to flunk the exam. And
so on. But these results are implausible (again, unless the book’s advice
was really bad!). Instead, it seems preferable to hold that it wasn’t really
Jim’s partying, but what that entailed (namely, the fact that he didn’t touch
the books) that caused him to flunk the exam.
Why does it seem so appealing, then, to mention Jim’s partying in con-
nection with his flunking the exam? Presumably, because it’s a vivid way
of implicating that he didn’t study for the exam, when he should have
been studying for the exam. We mention his partying because it is a more
colorful way to describe what happened, not because the partying is a cause
of the flunking of the exam per se. Again, unless there was something
about the partying itself that accounts for Jim’s doing badly on the exam,
it seems that he flunked because he didn’t study, not because of what he
did instead of studying.25
Finally, the causalist might want to argue that, although O2 was the
“main” cause of the child’s death in Drowning Child, A2 still played a
causal role in some “secondary” or “derivative” sense. Consider an example
by Yablo (1992): a pigeon, Sophie, is conditioned to peck at (all and only)
red objects; one day she is presented with a scarlet triangle and she pecks.
According to Yablo, although the triangle’s being red plays the major
causal role (it plays the role of being the cause of Sophie’s pecking, in
Yablo’s terminology), the triangle’s being scarlet (a determinate of the
determinable red) is still causally relevant to Sophie’s pecking. The idea, I
take it, is this: something’s being scarlet is a way of being red; thus the
triangle is red, on this occasion, by being scarlet. So on this occasion the
triangle has the causal powers that it has, in some sense, thanks to its being
scarlet. This role is “derivative” or “secondary” in that being scarlet only
gets to play that role in virtue of the causal powers that being red has;
however, one might argue that it still is an important role. Similarly, the
causalist could say, although O2 plays the major causal role in the drown-
ing child case, A2 is still causally relevant to the child’s death. For my
eating ice cream on the shore is, also, a way of failing to jump in (I fail to
jump in, on this occasion, by eating ice cream).
Omissions and Causalism 129

Now, imagine that this were right, that is, imagine that it were right to
say that A2 played a derivative causal role with respect to the child’s death.
Then the causalist could say that A1 plays a similar derivative role: one
that depends on the role played by O1. Would this help the causalist? I
don’t think so. For presumably, the causalist wants to say that mental items
like intentions play a primary role in giving rise to intentional acts, not
one that is parasitic on the role that something else plays. At least, this is
what the causalist wants to say about intentional actions. So, if intentions
played a primary role in the case of actions but not omissions, this would
still make for an important asymmetry between actions and omissions,
and thus it would present a problem for causalism as a general theory of
intentional behavior.26

6 Conclusion

I conclude that omissions pose a serious problem for causalism. Briefly, the
problem is that, whereas omissions can be intentional, causalism cannot
account for them in the same way that it accounts for intentional actions.
This is not so because omissions cannot be causes and effects, for it is quite
plausible to think that they can. The problem is, rather, that omissions are
not caused (at least ordinarily) by those mental items that the causalist
identifies as causes in the case of actions. As a result, causalism, conceived
as a theory of what it is for agents to behave intentionally, threatens to be
an either incomplete or highly disjunctive theory.

Acknowledgments

Special thanks to Randolph Clarke, Juan Comesaña, Michael Fara, John

Gibbons, Christopher Hitchcock, Richard Holton, Rebekah Rice, Daniel
Speak, and a referee for Noûs. Thanks also to audiences at the University
of Nebraska-Lincoln, the 2006 Inland Northwest Philosophy Conference,
the 2007 Arizona Ontology Conference, the 2007 MITing of the Minds,
and the 2007 Pacific APA.

Notes

1. For example, omissions were responsible for Lewis’s claiming that causation is
not a relation (Lewis 2004), and for Thomson’s and McGrath’s claiming that it is a
normative notion (Thomson 2003; McGrath 2005). My focus here is also omissions
and causation, in particular, on the question of whether omissions can be accom-
modated by causal theories of agency.
130 C. Sartorio

2. On this point, see Vermazen 1985, 104, and also Mele 2003, 151.

3. Different philosophers have different views of intentions: some believe that they
are reducible to belief-desire pairs, others believe that they are irreducible mental
states. But causalists seem to agree about the key role that intentions play in the
etiology of intentional action.

4. Alvarez notes this in Alvarez 2005.

5. In Davidson’s original work, there are only two brief references to omissions:
Davidson 1963, n. 2, and Davidson ,1971, 49,. In those places Davidson seems to
want to make room for omissions, but he is not very explicit about how.

6. See, e.g., Dowe 2000 and Beebee 2004. Davidson’s own view of causation in
Davidson 1967 appears to be of this kind (although he seems to take it back in his
discussion of Vermazen’s proposal, which I discuss below).

7. See, e.g., Bennett 1988 and Mellor 1995.

8. There are several questions that I’ll bypass here. For example, if we think that
there are two concepts of causation, what makes them both concepts of causation,
as opposed to concepts of something else? The two-concepts proposal only helps
the causalist to the extent that the nonproductive concept is genuinely a concept
of causation. Also, about Vermazen’s proposal: it’s unclear that the proposal explains
why my failure in Drowning Child is intentional. Imagine that, had I not formed
the intention not to jump in, I would have remained undecided. In that case it’s
not true that, had I not formed the intention not to jump in, I would have formed
the opposite intention, which would have caused my jumping into the water. So,
then, in what sense did my forming the intention not to jump in cause my not
jumping in?

9. However, as I will note in due course, there are also very important differences
between the two challenges. Notably, the strongest formulation of the problem of
omissions doesn’t appeal to a general exclusion principle. To my mind, this makes
the problem of omissions much more powerful than the traditional exclusion
problem (more on this later).

10. There are two main options: to insist that mental states are still causally effica-
cious, or to restate causalism as the claim that the physical realizers of mental states
are the causes of actions.

11. See Mele 1992 and Mele and Moser 1994. For arguments that intentionally φ-ing
doesn’t require an intention to φ, see Harman 1976 and Bratman 1984.

12. Davidson famously embraced a coarse-grained conception of events according

to which some “by statements” involving events are identity statements. For
example, if I flip the switch by moving my finger in a certain way, then my flipping
the switch is my moving my finger (Davidson 1971). On this view, the only actions
Omissions and Causalism 131

that exist are “primitive” or “basic” actions, or mere bodily movements (the actions
that take place “inside the agent’s skin”). Now, the sense in which I flip the switch
by moving my finger is not the same sense in which I fail to jump in by eating ice
cream. I flip the switch by moving my finger because the moving of my finger causes
the switch to be flipped; by contrast, I don’t fail to jump in by eating ice cream in
this sense: the eating of my ice cream doesn’t cause my not being in the water (more
on this later). The class of omissions that is of interest to us is that of primitive
bodily nonmovements (see Vermazen 1985, 102–103). Davidson acknowledges this
in his reply to Vermazen (Davidson 1985).

13. Or consider Ginet’s example (in Ginet 2004, 105): S intentionally did not mow
the grass in her backyard this summer because she wanted it to revert to a wild state.
As Ginet claims, it would be very implausible to suggest that there is something S
intended to do this summer in virtue of which she intentionally not mowed the
grass. For related arguments, see Weinryb 1980, Higginbotham 2000, and Vihvelin
and Tomkow 2005.

14. Note that this assumption is consistent with different views of omissions. In
particular, it’s consistent with views according to which some, but not all, omissions
are identical with actions.

15. Note that my omission to intend to jump in is also an intentional omission

(this will play a key role in my response to an objection in section 4 below). I argued
for a similar claim, although in a different context, in Sartorio 2005, 464–465.
However, I then (unintentionally) failed to draw attention to the significance of the
fact that my failure to form the relevant intention was also intentional.

16. By calling these cases “ordinary” and “paradigmatic” I do not mean to suggest
that there aren’t many cases of intentional omission of a different sort, say, cases
where the agent has to take active measures to counteract an existing trend or habit.
All I mean to imply is that the cases that are my focus here are the ones with the
simplest structure, given that the nonmovement simply flows from another omis-
sion. Thanks to Richard Holton for discussion of this point.

17. Kim famously grounded his exclusion argument in a general exclusion princi-
ple. For discussion of this principle, see Kim 1989.

18. In particular, note that this claim is consistent with the existence of cases
where both an agent’s action and an omission by the same agent are sufficient causes
of an outcome. Imagine that a sick patient will die at t unless his doctor gives
him a certain drug before that time. Imagine that, besides not giving him the drug,
he injects him with a poisonous drug that takes effect at t. In that case, arguably,
both the doctor’s failure to inject the patient with the medicine and his poisoning
him cause the patient’s death. Now, I think it is clear that the Drowning Child
case doesn’t have a relevantly similar structure: whereas here there is a good
reason to think that both the action and the omission are causes, there isn’t such
132 C. Sartorio

a reason in Drowning Child. Thanks to an anonymous referee for discussion of this

point.

19. Mele 1997b, 3–4.

20. Zimmerman (1981), Ginet (2004), and Clarke (forthcoming) believe this. But
what if I had remained undecided about what to do until the child died? In that
case, you’d still want to blame me for not jumping in; I was aware of the presence
of the child in the water, I knew that I could save him, etc. Could one argue that
my omission is still intentional in this case, even if I don’t form an intention one
way or the other? I think that the causalist can plausibly argue that my omission
isn’t intentional in this case. Maybe it’s not unintentional either. But even if it’s not
unintentional, some philosophers see a middle ground between intentional and
unintentional behavior (see, e.g., Mele and Moser 1994), and it is plausible to suggest
that my omission in this case falls in that middle ground. Another potential coun-
terexample to the claim that intentionally failing to jump in requires an intention
with the relevant content is this: a neuroscientist has been closely monitoring my
brain; he lets me fail to intend to jump in (which I do intentionally), but he prevents
me from forming the intention not to jump in (or any other intention with a similar
content). Is this scenario possible? I don’t know; fortunately, we don’t need to decide
this issue here.

21. The following objection might be raised: if I couldn’t intentionally omit to

intend to jump in without forming the intention not to jump in, then it is plausible
to think that the following counterfactual holds: had A1 not occurred, O2 wouldn’t
have occurred. But counterfactual dependence is sufficient for causation. Therefore,
it follows that A1 causes O2. In response, I think that counterfactual dependence
isn’t sufficient for causation, and it is illuminating to see why. Change the Drowning
Child case slightly: imagine that the two things I most love in the world are eating
ice cream and swimming. In that case, we may suppose, had I not eaten ice cream,
I would have jumped into the water and I would have rescued the child. So the
child’s death would counterfactually depend on my eating ice cream. Still, my eating
ice cream would not cause the child’s death. For, again, the child dies because of
what I don’t do, not because of what I do; this is so even if, if I hadn’t done what
I did, the child would have lived. By the same token, it seems to me that A1 still
wouldn’t cause O2, even if O2 counterfactually depended on A1.

22. Another potential challenge that I have chosen to set aside is the challenge that
negative intentions are impossible. According to some views of intentions, forming
an intention requires settling on a plan of action (Bratman 1984; Mele 1992; Enç
2003). This view creates some pressure to reject negative intentions. For it’s hard to
say what the plan might be in the case of omissions (for an argument that omissions
don’t involve “plans,” or “methods,” see Thomson 1996).

23. Although, is it really true that I didn’t jump in because I was eating ice cream?
Let’s assume that, if I was eating ice cream on the shore, then I couldn’t have been
Omissions and Causalism 133

jumping into the water at the same time, maybe in the sense that it was physically
impossible for me to do both at once. Does this mean that A2 explains O2? Compare:
I couldn’t have been a professional philosopher and a professional basketball player.
Does my being a philosopher explain my not being a basketball player? Or is this
explained by my lacking the relevant qualities for being a basketball player?

24. That’s right: a two-sentence paper (in Analysis). Here is the full text of the paper:
“The nerve of Mr. Bennett’s argument is that if A results from your not doing B,
then A results from whatever you do instead of B. While there may be much to be
said for this view, still it does not seem right on the face of it” (Anscombe 1966).

25. It might be argued that our judgments whether Jim’s partying caused his flunk-
ing the exam depend on the contrast class with respect to which we are making the
assertion: whereas it’s not the case that his partying rather than his caring for
Grandma caused him to flunk, his parting rather than studying did cause him to
flunk. (For a recent defense of a contrastive view of causation, see Schaffer 2005.) If
causation were a contrastive relation instead of a two-place relation, maybe the
causalist could make a similar claim about the intention not to jump in: whereas
it’s not the case that my intending not to jump in rather than my merely omitting
to intend to jump in caused my omitting to jump in, my intending not to jump in
rather than my intending to jump in did cause my omitting to jump in. I cannot
do full justice to this view here. But let me just note two things. First, causalism
would have to be revised accordingly, as the claim that intentions of a certain type
rather than intentions of another type cause the relevant bodily states in the relevant
way. Second, whereas the claim that explanation is not a two-place relation (but a
three-place relation, or even a four-place relation) bears some initial plausibility, the
corresponding claim about causation is very counterintuitive.

26. On similar grounds, Kim argues that the nonreductive physicalist shouldn’t
settle for the claim that the mental is causally efficacious but the causal powers of
the mental are parasitic on the causal powers of the physical (Kim 1998, 45).
9 Intentional Omissions

Randolph Clarke

Often when one omits to do a certain thing, one’s omission is due to one’s
simply not having considered, or one’s having forgotten, to do that thing.
When this is so, one does not intentionally omit to do that thing. But
sometimes one intentionally omits to do something. For example, Ann
was asked by Bob to pick him up at the airport at 2:30 AM, after his arrival
at 2:00. Feeling tired and knowing that Bob can take a taxi, Ann decides
at midnight not to pick him up at 2:30, and she intentionally omits to do
so. Other examples of intentional omissions include instances of abstain-
ing, boycotting, and fasting.1
Intentional omissions would seem to have much in common with
intentional actions. But the extent of the similarity is not immediately
obvious. Intentional omission has been recognized as a problem for theo-
ries of agency, but it is one on which, especially lately, little effort has been
expended. My aim here is to advance a conception of intentional omission,
address a number of claims that have been made about it, and examine
the extent to which an account of it should parallel an account of inten-
tional action. I’ll argue that although there might indeed be interesting
differences, there are nevertheless important similarities, and similarities
that support a causal approach to agency.
Although much of our interest in omissions concerns responsibility for
omitting, my focus is on the metaphysical and mental dimensions of inten-
tional omission. What sort of thing (if it is a thing at all) is an omission?
What, if any, mental states or events must figure in cases of intentional
omission, and how must they figure? Answers to these questions have some
bearing on the moral issue, but the questions are interesting in their own
right. And they stand in some degree of mutual independence from the
moral issue, as there can be intentional omissions for which no one is
responsible, and (on the assumption that we can be responsible for any-
thing at all) we can be responsible for omissions that aren’t intentional.2
136 R. Clarke

A preliminary distinction might help clarify the object of my attention.

My focus is on cases about which it is correct to say that someone inten-
tionally omits to do something. There are cases of another sort in which
we might say that it is intentional of some individual that she doesn’t do
a certain thing, but she doesn’t intentionally omit to do it. For example,
wanting to ensure that he wouldn’t leap into the sea when he heard the
Siren song, Ulysses had himself bound to the mast of his ship. As planned,
he didn’t jump into the sea; but he nevertheless didn’t intentionally omit
to leap in, for while he could hear the song, he tried his best to free himself
and jump into the water.3 Intentionally omitting to A at t would seem to
require, at least, that one is not at t trying to A.4

1 Omissions and Actions

Let us call actions of a familiar sort, such as raising one’s arm, walking, or
speaking, “positive actions.” It is on positive actions that action theory
has, understandably, largely focused. How are omissions related to positive
actions? For one thing, when one intentionally omits to A, is one’s omis-
sion identical with some intentional positive action that one then
performs?
Perhaps sometimes it is. Imagine a child crouching behind a chair and
holding still for several minutes while playing hide and seek.5 The child’s
holding still is arguably an intentional action; it requires the sending of a
pattern of motor signals to certain muscles, perhaps the inhibition of other
motor signals, the maintenance of balance, with fine adjustments made in
response to feedback, at least much of which arguably results from the
child’s intending to hold still. The child’s not moving is an intentional
omission. And perhaps the child’s not moving in this case is just her
holding still.6
It might be objected that the child might have not moved even if she
hadn’t intentionally held still—she could have been frozen stiff. But we
may grant this possibility without accepting that the child’s holding still
(that particular event) is distinct from her not moving, just as we may
grant that on some occasion when I walked slowly, I might have walked
without walking slowing, without thereby committing ourselves to the
implausible view that I performed two acts of walking when I walked
slowly.
On a minimizing view of act individuation, when one flips the switch,
turns on the light, illuminates the room, and startles the burglar, one
might perform only one action, which might be intentional under some
Intentional Omissions 137

of these descriptions and unintentional under others (Davidson 1980, 4–5;

Anscombe 1963, 46). Having flipped the switch, it might be said, there is
nothing further that one must do to turn on the light, and so on.
One competing view takes actions to be instantiations of act-properties,
and holds that actions are nonidentical if they are instantiations of differ-
ent act-properties (Goldman 1970, 10). Such an account will take one’s
turning on the light and one’s startling the burglar to be different actions.
But even this second view may count the walking and the walking slowly
as one action, for an instantiation of a given act-property can be designated
by each of several different expressions (cf. Bennett 1988, 93). And, simi-
larly, either of these two views may count the child’s holding still as identi-
cal with her not moving.
An extreme maximizing view combines the second account of what an
action is with the claim that we have the same act-property only when we
have the same act-description. On such an extreme view, it will be the case
that no omission (nothing described as an omission) is identical with any
positive action. Setting aside such a view, we may allow that in some cases
there is such an identity.
A further consideration favors this allowance. One might for a time plan
not to A, but then change one’s mind and decide to A. An agent in such
a case who realizes that she has a tendency to absent-mindedly revert to
abandoned plans might form an intention not to omit to A as well as the
intention to A. She might then not just intentionally A but also intention-
ally omit to omit to A. In such a case, her omission would seem to be just
her action of A-ing.7
But in many cases—perhaps in most—one’s intentionally omitting to
A at a certain time isn’t identical with any positive action that one then
performs. Ann intentionally omits to pick up Bob at 2:30. Suppose that,
though feeling too tired to go out, Ann is at home playing piano from 2:00
to 3:00. Her playing piano, let’s imagine, keeps her neighbor awake. It isn’t
Ann’s omitting to pick up Bob that keeps the neighbor awake (cf. Lewis
2004, 282; Sartorio 2009, 518). Her omitting to pick up Bob isn’t her
playing piano, nor does it seem to be any other positive action that Ann
performs. And there are similar reasons in many cases of intentional omis-
sion for denying the identity of the omission with any positive action that
one then intentionally performs.
In a case in which one’s intentional omission isn’t itself some positive
action that one performs, must one intentionally perform some other
positive action during some relevant time period? Suppose that it is A-
ing during the interval t that one intentionally omits. Must one have
138 R. Clarke

intentionally performed during t some positive action other than A-

ing?
We are almost always, during our waking lives, intentionally doing
something. Besides performing overt bodily actions, we attend to things,
concentrate, try to think of or remember something, and so on. So it might
come close to a trivial truth that when one intentionally omits to A at t,
one is, at t, intentionally doing something else. But it isn’t a truth. Ann
intentionally doesn’t pick up Bob at 2:30. This might be so even if she
went to bed at 1:00, is sound asleep at 2:30, and isn’t intentionally per-
forming any positive actions at that later time.
A related claim is that refraining from A-ing requires performing some
other action to prevent oneself from A-ing (Brand 1971, 49). One might
take refraining to be something different from intentionally omitting,
though I doubt that ordinary language precludes an overlap. In any case,
the claim is mistaken about refraining. I might comply with instructions
to refrain from touching a freshly painted object until it’s dry. There might
be several things I do while refraining, but there need be nothing that I do
in order to prevent myself from touching the object (cf. Vermazen 1985,
103; Walton 1980, 322).
If no positive action is required at t, and none need be performed in
order to prevent oneself from A-ing, does intentionally omitting to A at t
nevertheless require that one intentionally perform some positive action
at t or earlier, such that one believes one can’t both intentionally do that
then and A at t (as Zimmerman 1981, 547 claims)? When I comply with
the instructions not to touch the painted object, I might remain standing
where I am—within easy reach of the object—my arms hanging freely at
my sides, whistling a tune. I needn’t believe that my standing there whis-
tling is incompatible with my touching the object (obviously it isn’t). And
having my arms at my sides during this period need not be any more an
intentional action than it was before I read the instructions and came to
intend not to touch the object. I might have no temptation that needs to
be resisted, and I need not intentionally hold my arms at my sides.
If intentional omissions are not always identical with positive actions,
and if one need not, at any relevant time, perform any other positive action
when one intentionally omits to A, are intentional omissions nevertheless
themselves acts, even if “negative acts”? One might take them to be such
because, one holds, they fulfill intentions and forming an intention is an
act (McIntyre 1985, 93). But intentions can be nonactively acquired as well
as actively formed in making decisions. (See, e.g., Audi 1993, 64, and Mele
2003, 200–201.) One might wish to call even nonactively coming to have
Intentional Omissions 139

an intention a “mental act,” but we should not lose sight of the difference
between such an occurrence and an intentional action.
We might decide to call intentional omissions “acts of omission” simply
because they’re intentional, or on the grounds that (we think) they express
intentions. (Whether intentionally omitting requires having a pertinent
intention is a question I’ll address below.) There is warrant for this choice
of terminology, as what is done intentionally is, in some sense, a manifes-
tation of agency. (I’ll return to this point in section 6.)
Still, we ought not assume that intentional positive actions and such
negative acts are thoroughly alike. The question of whether there are in
fact significant differences arises at several points in the discussion to
follow. Since it will be convenient to have an economic way to refer spe-
cifically to positive actions, henceforth when I use “action” or “act”
without qualification, I intend positive action. Using the terms this way
is, of course, meant to be consistent with what I observed in the preceding
paragraph.

2 Absences

When an omission isn’t itself an action, what is it? It would seem to be

an absence, an absence of an action. Ann’s intentional omission is the
absence of an action by her of picking up Bob at the airport at 2:30.
What is such an absence? One might take it to be a negative event, or
a negative state of affairs, or the instantiation of a negative property. Any
such view is problematic, for there are good reasons to deny the existence
of such negative entities. (See, e.g., Armstrong 1978, 23–29.)
One alternative view holds that when one omits to do a certain thing,
one stands in the not-causing relation to a certain (positive) event. For
example, “when I refrain from shooting a child, it might be appropriate
to say that I stand in the ‘not-causing’ relation to the event, ‘the child’s
being shot by me’” (Fischer 1985–1986, 265). Aside from quarrels about a
relation of not-causing, the suggestion is objectionable on the grounds that
no one stands in any genuine relation to any nonexistent thing; relations
require relata (Lewis 2004, 283).8 Typically, if one refrains from shooting
a child, there is no event that is the child’s being shot by oneself.
Omissions aren’t, of course, the only sort of absence. The world is
missing certain events, states of affairs, objects, and properties. Arguably,
the absences of these things aren’t queer sorts of entities; they aren’t beings
or things at all, simply absences of things (Kukso 2006, 29; Lewis 2004,
282). The view is available, and with much to recommend it, that when
140 R. Clarke

Ann doesn’t pick up Bob, her omission is such an absence, and nothing
more, even if more is required for it to be an intentional omission.
It can seem puzzling just when and where omissions occur. Does Ann
omit to pick up Bob at midnight, when she decides not to pick him up,
or at 2:30, when she isn’t at the airport to pick him up, or during some
portion, or all, of the interval from that earlier time to the later one? Does
her omission take place at her house, where Ann is located throughout
that interval, or at the airport, or along the route that Ann would have
taken had she gone to pick up Bob? If omissions (those that aren’t actions)
are absences, and absences aren’t things, then (these) omissions don’t
occur anytime or anywhere. There isn’t an action by Ann at 2:30 of picking
up Bob at the airport. The time and place in question are some pertinent
time and place at which there isn’t such an action. That there isn’t such
an action at that time and place is what it is for there to be such an absence.
Which absences of actions are omissions? Some philosophers (e.g.,
Fischer 1985–1986, 264–265) take it that there is an omission anytime an
agent does not perform a certain action. Somewhat less generously, others
(e.g., Zimmerman 1981, 545) hold that there is an omission whenever (and
only when) an agent is able to perform some action A and does not A.9
Whether omitting to A requires that one be able to A is a complicated
matter, for there are several different sorts of thing each of which may
fairly be called an ability to act. Arguably, some type of ability to do other
than what one actually does is ruled out if determinism is true. Some other
types of unmanifested abilities, such as talents or skills, general capacities,
or powers to do certain things, are plainly compatible with determinism.
Similarly, it seems that in cases of preemptive overdetermination, in which
an agent does a certain thing on her own, but would have been made to
do it anyway had she not done it on her own, some type of ability to do
otherwise is precluded, while the agent might nonetheless retain a capacity
or power to act that, it is ensured, she won’t exercise.10
It hardly seems to follow from the truth of determinism that no one
ever omits to send holiday greetings, wear their seat belts, and so forth, or
that we never abstain, boycott, or fast. It doesn’t seem credible that omit-
ting to A, or that intentionally omitting to A, requires that one have any
sort of ability to A that would be ruled out by determinism. And agents in
cases of preemptive overdetermination might omit to do things that, in
some sense, they’re unable to do.11
On the other hand, lacking an ability of another sort can seem to pre-
clude one’s intentionally omitting to do a certain thing. We might plau-
sibly judge that an agent who intended not to get out of bed, and who
Intentional Omissions 141

didn’t so act, didn’t intentionally omit to get out of bed if, unbeknownst
to her, she was paralyzed and wouldn’t have risen from bed even if she
had tried (Ginet 2004, 108).12 I suspect that it would be a delicate matter
to say exactly what type of ability to act is required for omission, or for
intentional omission, and I’ll not attempt that project here.
Although for some purposes we might wish to say that there is an omis-
sion whenever an agent doesn’t perform a certain action that she is, in a
relevant sense, able to perform, we don’t commonly use the term so
broadly. Setting aside cases of intentional omission and those in which it
is intentional of some agent that she doesn’t do a certain thing, in ordinary
contexts we tend to take “omission” to be applicable only when an action
isn’t performed despite being recommended or required by some norm
(not necessarily a moral norm; cf. Feinberg 1984, 161; Smith 1990; Wil-
liams 1995, 337). We may sensibly count as omissions only those absences
of actions that satisfy some such restriction (as well as whatever ability
requirement is appropriate).
In any case, since the focus here is on intentional omissions, the
absences that count will be restricted in a different way. Only absences of
actions in cases in which the agents have certain mental states are inten-
tional omissions.

3 Intentions

At least generally, in cases of intentional action, the agent has some inten-
tion with relevant content. Typically, when I intentionally walk, I intend
to walk. However, arguably, even if intentionally A-ing requires having an
intention, it doesn’t require intending to A (or having an intention to A).
While walking, I might intentionally take a certain step, without intending
specifically to take that step. It might suffice that while taking that step I
intend to walk then, I’m a competent walker fully capable at the moment
of exercising that competence, there’s no obstacle requiring any special
adjustment of my walking, and my taking that step results in a normal
way from my intending to walk then. (On this type of case, see Mele 1997a,
242–243; I’ll describe below some further cases in which, apparently, one
can intentionally A without intending to A.)
Does intentionally omitting to A require having some intention with
relevant content? If so, what content must the intention have? And when
must one have the intention?
Suppose that one intentionally omits to A during t. Must one (as Ginet
2004 maintains) intend throughout the interval t not to A?13
142 R. Clarke

No. Some actions require preparation, and preparatory steps must some-
times be taken by a certain time prior to the action in question. If I am to
attend a meeting in a distant city on Monday afternoon, I must earlier
book a flight, get to the airport, and so forth. Suppose that having decided
not to attend, I intentionally don’t perform such preparatory actions.
Having forgone the preparations, I have no further need of the intention
not to attend, and with other things on my mind, I may dispense with it.
(The claim isn’t that the intention couldn’t be retained, only that it need
not be.) Nevertheless, when I don’t show up at the meeting, I might inten-
tionally omit to do so.14
Does intentionally omitting to A during t require having, at some rel-
evant time, an intention not to A? Several writers (e.g., Ginet 2004 and
Zimmerman 1981) have claimed that it does, but again the claim appears
mistaken. An intention with some other content might do.
Suppose that Charles wants to abstain from smoking for a week, but he
thinks it unlikely that he’ll succeed. Cautious fellow that he is, Charles
forms only an intention to try not to smoke. He plans to spend time with
friends who don’t smoke, to chew gum to diminish his desire to smoke,
and so forth, which he hopes will enable him to resist the temptation.
Suppose that Charles then makes the effort and succeeds, and there’s
nothing magical or fluky about his success: his plan works just as he hoped
it would. Charles omits to smoke because he tries not to smoke. He inten-
tionally omits to smoke for a week, even though he didn’t have an inten-
tion not to smoke for a week.15
The case parallels one of action in which, though thinking success
unlikely, one intends to try to A, one makes the effort, and one unexpect-
edly succeeds. If the success isn’t a fluke, one might then have intention-
ally A-ed without having intended to A (cf. Mele 1992, 131–133).
It might be objected that Charles has it as his aim or goal that he not
smoke, and to take something as an aim or goal is to intend that thing.16
But one can have something as a hoped-for goal without intending that
thing. Arguably, there is a negative belief constraint on rationally intend-
ing, such that it isn’t rational to intend to A while believing that one
probably won’t succeed.17 Given his expectations, Charles might take
abstaining from smoking for a week only as a hoped-for goal, intending
no more than to try his best.
Some different cases also suggest that one might intentionally omit to
A without intending to omit to A. Suppose that while walking in the
countryside you come to a fork in the path. You’re aware that the path on
the left is more pleasant, and you realize that should you take the path on
Intentional Omissions 143

the right your walk will be less enjoyable. Suppose that you nevertheless
decide to take the path on the right (perhaps believing that path shorter),
and you then do so, aware that in so doing you aren’t taking the left path.
It seems that you needn’t intend not to take the left path in order for it
to be the case that you intentionally don’t take (omit to take) that path.
As several theorists see it, one can carry out an intention to A and be
aware that by A-ing one will do something B, without then intending to
B, and yet intentionally B. This might be so when one is aware of a reason
not to B and one decides to A despite that consideration. For example, I
might intentionally start my car in the morning despite being aware that
(since my car is very noisy) by so doing I’ll disturb my neighbors’ sleep. I
might then intentionally disturb them without having intended to do so
(Ginet 1990, 76; cf. Harman 1976/1997, 151–152).18 Examples such as that
in the preceding paragraph make an equally strong case for the view that
one can carry out an intention to A and be aware that, in intentionally
A-ing, one will not do something B, without then intending not to B, and
yet intentionally omit to B.19
Must any intention with relevant content figure in the history of an
intentional omission? Suppose that I see a child struggling in a pond but
intentionally omit to jump into the water to save the child. I don’t intend
to jump in. Might it suffice for my omitting to jump in to be intentional
that this omission results from my intentionally omitting to intend to
jump in (as Sartorio 2009, 523 suggests)?20
What would make it the case that my not intending to jump in is itself
an intentional omission? One might say: “I voluntarily failed to form that
intention, after deliberating about whether to do so, after considering
reasons for and against doing so, etc.” (Sartorio 2009, 523). But the claim
that my failure to intend was voluntary seems to presuppose that, rather
than explain how, it was intentional. And that my not intending to jump
in comes after deliberation about whether to do so, and after consideration
of reasons for and against jumping in, does not suffice to make my not so
intending intentional. I might have simply failed to make up my mind.
Suppose the case unfolds this way. I deliberate about whether to jump
in or not, never making up my mind. As I continue deliberating, I realize
that the child is drowning and I’m doing nothing to save her. In this
version of the case, do I intentionally omit to jump in to save the child
despite having no intention with pertinent content?
Deliberating is itself activity, and typically it is intentional activity. (We
can intentionally try to think of relevant considerations, intentionally turn
attention to one thing or another, intentionally try to make up our minds.)
144 R. Clarke

I am engaged in that activity as I stand on the shore. Presumably, I intend

to deliberate, or something of that sort, and as I watch the child drowning,
I carry out an intention to continue deliberating, all the while aware that
in doing so I am failing to rescue the child. My mental state includes an
intention that is relevant to my omission. Just as, in the earlier case of the
forking paths, your not taking the left path might be intentional even
though you don’t intend not to take that path, so here my not jumping
in might be intentional even though I don’t form an intention not to jump
in. In both cases, one’s mental state includes some relevant intention,
together with an awareness that, in acting as one intends, one won’t
perform a certain other action that one has reason to perform.
Might some combination of wanting not to A and awareness that one
is not A-ing suffice, as far as having the required attitudes goes, for inten-
tionally omitting to A? Just as one can desire to A without being committed
to A-ing, so one can desire not to A without being committed to not A-ing.
And just as desiring to A, together with an awareness that one is A-ing,
does not suffice, as far as having the required attitudes goes, for intention-
ally A-ing, so it seems having the desire and the awareness is not enough
for one’s omission to be intentional. Someone who sees the child in trouble
might want not to jump in to help, and she might be aware that she’s not
jumping in, but she might nevertheless not intentionally omit to jump in.
She might not be committed to not jumping in, or to anything else at that
moment.
If, as appears to be the case, intentionally omitting requires having an
intention with relevant content, what content counts as relevant? The
concept of intentional omission, like that of intentional action, is vague.
There are cases in which the agent clearly has a relevant intention, cases
in which the agent clearly does not, and cases in which it is unclear
whether the content of any intention possessed by the agent is suitably
relevant. In the last of these cases, it might be unclear whether the agent
intentionally omits to act. Besides saying that, for intentionally omitting
to A, intentions other than one not to A can have relevant content, I doubt
that there is much more to say on this point.
Several further observations can be made about intentions in cases of
intentional omission. On a widely held view, the representational content
of an intention to act is an action plan, and the attitudinal mode is that
of being settled or committed: intending to perform an action of A-ing is
being settled on or committed to A-ing (see, e.g., Mele 1992, chs. 8 and
9). When one intends not to A, the content of one’s plan is not-to-A. There
Intentional Omissions 145

is no need to see the attitudinal mode as any different from that of intend-
ing to act.
Sometimes, in intentionally omitting something, one performs some
action as a means to that omission (e.g., chewing gum so that one won’t
smoke). An intention to omit to A can include a more or less elaborate
plan for so omitting. But there are cases in which one’s intention not to
A need not include any plan at all about how not to A. One need not
always, in order to intentionally omit to A, perform or intend to perform
any action at all as a means to not A-ing.21
Moreover (contrary to Wilson 1989, 137–142), no intention that one
has in a case of intentional omission need refer to anything that the agent
in fact does; the agent need not intend of some positive behavior in which
she engages that it not be (or not include, or be done instead of, or be
allowed not to be, or not allow her to perform) the omitted action (cf.
Ginet 2004, 104–106). Ann intends not to pick up Bob at the airport. She
need not also intend of her piano playing, or of her going to bed, or of
anything else, any of the suggested things. (I might boycott veal for the
remainder of my life, without intending of any particular thing I do—and
certainly without intending of all of it—that it not be the purchasing of
veal.)
At least generally, when one intentionally omits to A, for some relevant
time period, one does not intend to A. Certainly one might have earlier
intended to A and have since changed one’s mind. And one might cease
to intend to A and come to intend not to A without ever revoking one’s
earlier intention. One might forget that one so intends, cease to so intend
because one so forgets, and not remember the earlier intention when one
later acquires the intention not to A. Finally, just as one can have a nonoc-
current intention to act of which one is unaware and intentionally do
something contrary to that intention, so one can have a nonoccurrent
intention to A and yet intentionally omit to A. Intentionally omitting to
A does not strictly require (for any time period) not intending to A.

4 Causes

On a widely held view of intentional action, mental states or events (or

their neural realizers)22 play a significant causal role in each case of inten-
tional action. It might seem that if many intentional omissions are
absences, then such omissions aren’t caused by the agents’ mental states
(or by events involving those agents).
146 R. Clarke

The conditional might be affirmed on the grounds that absences can’t

be causes or effects.23 The issue is a contentious one, with many writers on
causation holding that absences can indeed cause and be caused. Even
some (e.g., Kukso 2006 and Lewis 2004) who take absences not to be enti-
ties at all hold that there can be causation of and by absences.
However, it has been argued (by Sartorio 2009) that even if absences
can be causes and effects, in a case of intentional omission, it isn’t the
agent’s intending something that causes her omission; it is, rather, her not
intending something that causes the omission.
Suppose that Diana is standing by a pond eating ice cream when she
sees a child struggling in the water. Diana decides not to jump in to save
the child, she intentionally omits to jump in, and the child drowns. One
might think that if absences can be caused, then Diana’s intending not to
jump in is a cause of her not jumping in. However, the argument goes, it
is instead her not intending to jump in that causes her omission.
The argument appeals to an analogy between what causes the child’s
death and what causes Diana’s omitting to jump into the water. Suppose
that, instead of jumping in, Diana remains on the shore and continues
eating ice cream. Consider the following two candidates for causes of the
child’s death: Diana’s eating ice cream on the shore, and her not jumping
into the water. Assuming that absences can be causes, clearly the latter
causes the death; the child dies because Diana doesn’t jump in to save her.
Moreover, Diana’s eating ice cream isn’t an additional cause of the death.
The child dies because of what Diana doesn’t do, and not because of what
she does instead. In fact, it’s irrelevant that Diana goes on eating ice cream;
all that matters is that she fails to jump in to save the child.
Now, the argument continues, the question of what causes Diana’s not
jumping in should be answered in an analogous way. Consider these two
candidates: her intending not to jump in, and her not intending to jump
in. Again, assuming that absences can be causes, the latter clearly causes
Diana’s not jumping into the water; she omits to jump in because she
doesn’t intend to jump in. And again, we shouldn’t say that her intending
not to jump in also causes her omitting to jump in. She omits to jump in
because of what she doesn’t intend, and not because of what she does
intend. In fact, it’s irrelevant that she intends not to jump in; all that
matters is that she doesn’t intend to jump in.
I don’t know whether absences can be causes or effects, and although
I find the argument just presented less than fully convincing,24 I’m far from
sure that it’s mistaken. I’ll suppose, for the sake of argument, that (many
or all) intentional omissions that are absences aren’t caused by the agents’
Intentional Omissions 147

mental states (or by mental events involving those agents), either because
absences can’t be caused, or because absences of mental states or events
cause these omissions. Still, I contend, the agents’ mental states (or mental
events involving those agents) play a causal role in such cases, one that
parallels, in interesting respects, the role they are required by causal theo-
ries of action to play in cases of intentional action. Relevant mental states
(or events) must cause the agent’s subsequent thought or action, even if
they needn’t cause the absence of some action.
If Diana’s omission to jump into the water is caused by her not intend-
ing to jump in, the omission’s being so-caused evidently isn’t what renders
it an intentional omission.25 There might be other folks on the shore who
also don’t intend to jump in, who also don’t jump in, whose not intending
to jump in is as good a candidate for a cause of their omission as Diana’s
is of hers, but who nevertheless don’t intentionally omit to jump in.
Perhaps the lifeguard fails to notice that the child is in trouble; perhaps
someone else, though noticing the trouble, doesn’t think of jumping in to
help.
We’ve seen reason to think that in order to intentionally omit to A, one
must have an intention with relevant content. As the story goes, Diana
has such an intention: in deciding not to jump into the water, she forms
an intention not to jump in.
Is it enough for Diana’s omission to be intentional simply that she have
this intention? Need the intention do anything at all? Suppose that, upon
deciding not to jump in, Diana immediately wonders what to do instead,
forms an intention to walk over to get a better view of the impending
tragedy, and then does so (eating her ice cream all the while). In the
normal case, we would take it that the intention not to jump in is among
the causes of this subsequent sequence of thought and action.26 Moreover,
this causal role seems more than incidental; if we suppose that the inten-
tion causes no such things, then it no longer seems that we have a case of
intentional omission.
Suppose that Diana’s intention not to jump in, whatever it is—some
distributed state of her brain, perhaps—comes to exist with the usual causal
powers of such a state but is from its start prevented from causally influ-
encing anything—prevented, that is, from manifesting its powers. Just after
deciding not to jump in, Diana happens to wonder what to do instead,
decides to walk over for a better view, then does so. Something causes this
stream of thought and action—perhaps a chip implanted earlier in Diana’s
brain by a team of neuroscientists, who just happen to have picked the
present moment to test their device—but her intention not to jump in
148 R. Clarke

isn’t a cause of any of what Diana does. In this case, does she intentionally
omit to jump into the water?
It seems clear that she does not. Her not jumping in is intended—and
she’s guilty of so intending—but she doesn’t intentionally omit to jump
in, because her intention doesn’t in any way influence her subsequent
thought or action. It’s pure happenstance that what she does accords in
any way with her intention. For all her intention has to do with things,
what she was caused to do might just as well have been to jump into the
water and save the child.
The case is analogous to one in which an agent intends to perform a
certain action, the appropriate bodily movement occurs, but the agent
doesn’t perform the intended action, because his intention is ineffective.
Unaware that my arm has become paralyzed, I might intend to raise it
now. Unaware that I currently so intend, my doctor might test the new
motor-control device that she implanted in me during recent surgery. It’s
sheer coincidence that the movement caused by the doctor accords with
my intention, and I don’t intentionally raise my arm.
Lest one think that Diana’s not jumping in fails to be intentional
because, it might now seem, her omission doesn’t counterfactually depend
on her not intending to jump in,27 suppose that the implanted chip has
an unforeseen flaw: it will remain inert if, when the activation signal is
sent to it, the agent in whose brain it’s implanted has just decided to jump
into water. In any case, an omission’s being intentional doesn’t require
that the omission depend counterfactually on the absence of an intention
to perform the action in question. Diana might intentionally omit to jump
into the water even if, had she acquired an intention to jump in, she might
have changed her mind.
In some cases, the role played by an intention not to act might be less
pronounced. Suppose that having decided not to jump in to help, Diana
simply stays where she is and continues eating her ice cream. Arguably,
an intention she already had—to eat the ice cream—causes her continuing
activity. Nevertheless, there would seem also to be a causal role played by
her newly formed intention not to jump in. That intention might play a
sustaining role, contributing to her continuing to intend to stand there
eating the ice cream, and thus to her continuing activity. As intentions
usually do, this one could be expected to inhibit further consideration of
the question that it settles—the question of whether to help the child. It
thereby causally influences Diana’s flow of thought. And it might play a
causal role that isn’t just a matter of what it in fact causes, that of standing
Intentional Omissions 149

ready to cause behavior in accord with its content, should that be

necessary.
Generally, there is much less that an intention not to act must do in
order to succeed than is required of an intention to act. Where A-ing is
performing some positive action, A-ing generally requires that one’s behav-
ior fall within some fairly narrow range of possible behaviors, while not
A-ing generally requires only that one’s behavior remain outside—any-
where outside—some quite narrow range.28 Hence, the work demands on
intentions not to act are usually comparatively light. They can generally
succeed by doing relatively little. But if they had no work to do at all—or
if in fact they typically did no work at all—it would be a wonder that we
ever bother to form such intentions.
The similarity between intentional action and intentional omission
appears to go further. In both cases, if one’s intention causes what it causes
in some peculiar way, so that one has no control over what follows it, then
we don’t have a case in which something is done intentionally. It’s a
common observation in action theory that an intention might cause the
intended behavior and yet that behavior not constitute intentional action,
if the causal pathway is wayward or deviant.29 Something similar would
seem to be so in the case of intentional omissions. Ann might intentionally
omit to pick up Bob at the airport at 2:30 even if she’s asleep at 2:30, but
we won’t have a case of intentional omission if her falling asleep happens
this way: her forming the intention not to pick up Bob immediately (and
unexpectedly) triggers a prolonged episode of narcolepsy.30
If, in a case of intentional omission, an intention with relevant
content must indeed play a causal role in the course of what does occur,
then, in many cases at least, there is an important sense in which inten-
tional omissions result from intentions not to act, even if such intentions
don’t cause the omissions. Suppose that Diana’s standing on the shore
eating ice cream at time t is caused by (among other things) her intention
(formed prior to t) not to jump in. It can’t be that at t she’s standing on
the shore and at t she’s jumping into the water; the one precludes the
other. Her not jumping in is a consequence of (it results from) her intend-
ing not to jump in, even if it isn’t a causal consequence. (Note that this
holds as well if the omission, because it is an absence, can’t be caused by
anything.)31
A standard causal theory of action holds that, in a case of intentional
action, an intention with relevant content must appropriately cause certain
agent-involving events. It seems that in a case of intentional omission,
150 R. Clarke

likewise, an intention with relevant content must play an appropriate

causal role with respect to what happens.
In the latter case, no specific intention is required, at no particular time,
and what intention there is need not cause anything in particular—it need
not, for example, directly cause any action. Still, there are imaginable cases
in which no intention with relevant content is present at any pertinent
time, or in which, though present, such an intention causes nothing, or
causes what it does in such a way that control is lost. And in such cases,
it does not seem that the omission is intentional. The apparent need for
the appropriate causal effectiveness of a pertinent intention is nontrivial,
and it constitutes an interesting and important similarity between inten-
tional action and intentional omission.

5 Omitting for Reasons

Generally, intentional actions are actions performed for reasons. Likewise,

typically when one intentionally omits to A, one omits for a reason. Ann
omits to pick up Bob at the airport because she’s tired and (she knows) he
can take a taxi. What, if anything, is the causal role of the reasons for
which one intentionally omits?
According to causal theories of action, when one acts for reasons, the
reasons for which one acts (or mental states with those reasons as their
objects—for brevity I’ll henceforth ignore this alternative) are causes of
one’s action. On the supposition that causal theories are correct about the
role of reasons in cases of intentional action, in cases of intentional omis-
sion in which the omission is itself an action, the role of reasons is the
same. On that same supposition, in cases of intentional omission in which
the omission is not an action, reasons still seem to play a causal role.
Intentionally omitting to A, it appears, requires having at some perti-
nent time an intention with relevant content. Reasons for which one
intentionally omits to A would play the same role in causing such an
intention as do reasons in causing the intention to act that one has when
one intentionally As. (This is so in Diana’s case as well as in cases like that
of the forking paths.)
I’ve argued, further, that if in fact an intention with relevant content
is required, then when one intentionally omits to A, such an intention is
a cause of one’s subsequent thought and action—it plays a causal role with
respect to events that do occur. Reasons for which one omits to A would,
on our supposition, play a similar role, by way of causing one’s intention.
Even if one’s intention not to A (or other relevant intention) doesn’t cause
Intentional Omissions 151

one’s omission, there is causal work for the reasons for which one omits
to A.

6 Agency, Actions, and Omissions

Still, if for either of the reasons identified at the start of section 4, inten-
tional omissions that are absences aren’t caused by the agents’ intentions,
does this fact constitute trouble for “causalism as an attempt to explain
what it is for an agent to behave intentionally” (Sartorio 2009, 513)? It
presents no problem for causal theories of action. Such theories are not
theories of omissions that aren’t actions, and they strictly imply nothing
about such things.
We might sensibly construe agency more broadly as encompassing all
that is done intentionally, and so as including intentional omissions. At
least typically, things done intentionally fulfill intentions and are done for
reasons. Such things may fairly be said to be manifestations of our agency.
We might then wonder about the prospects for a causal theory of this
broader phenomenon.
But with agency so construed, it should not be expected that it must
be possible to construct a uniform theory of it, for the phenomenon
itself lacks uniformity, including, as it does, actions as well as things that
aren’t actions. It will be no fault of any theory of intentional action if it
does not apply, in a straightforward way, to all of what is then counted
as intentional agency. And a comprehensive theory of agency (if any
such thing is possible) might play out one way in the case of action and
another in the case of omission. If such an account has to have a disjunc-
tive character, that need might accurately reflect the diversity of its subject
matter.
We might nevertheless expect that the right account of intentional
omission will resemble in important respects the right account of inten-
tional action. If what I’ve said here is correct, the resemblance to causal
theories of action is significant. But to the extent that omissions that aren’t
actions are unlike actions, it should not be surprising that the similarity is
imperfect.

Acknowledgments

For helpful comments on earlier versions of this essay, I wish to thank Carl
Ginet, Alison McIntyre, Al Mele, Carolina Sartorio, Kadri Vihvelin, Michael
Zimmerman, and an anonymous referee for Noûs.
152 R. Clarke

Notes

1. The last two of these, and more, are mentioned by McIntyre (1985).

2. Some theorists (e.g., Bennett 2008, 49) hold that one is responsible for something
only if one is either blameworthy or praiseworthy for that thing. Plainly on this
view there can be intentional omissions—those that are morally neutral or
indifferent—for which no one is responsible. But there can be such omissions even
if there can be moral responsibility for morally neutral things. Some agents (e.g.,
young children, or people suffering from certain mental illnesses) lacking some of
the capacities required for responsibility nevertheless engage in intentional action
and intentionally omit to do certain things. And while many theorists hold that
responsibility for unintentional omissions must stem from something done inten-
tionally, others (e.g., Smith 2005) deny this claim. In sum, the relation between
responsibility and what is done intentionally is both complex and contested. This
fact constitutes one reason for taking a direct approach to the topic of intentional
omission.

3. Cf. Mele 2003, 152, and McIntyre 1985, 47–48. McIntyre draws the distinction
as one between intentionally omitting and its being intentional on one’s part that
one omits. Note that we might say that Ulysses intentionally prevented himself
from jumping into the sea. We should then recognize that one can intentionally
prevent oneself from doing a certain thing and yet not intentionally omit to do that
thing. Finally, one might take this case to support the view that intentionally omit-
ting to A requires that one be able to A. I briefly discuss such a requirement in
section 2 below.

4. “To A” is used throughout as a stand-in for expressions indicating types of posi-

tive action, such as to raise one’s arm or to speak.

5. The example is from Mele (1997a, 232), though he employs it for a different
purpose.

6. Note that the child doesn’t simply prevent herself from moving; she both inten-
tionally holds still and intentionally omits to move. Unlike Ulysses, at the time in
question she isn’t trying to do what she does not do.

7. This kind of case was suggested in conversation by Al Mele. Note that one can
omit to intentionally omit to A, and yet not A. I might plan not to A, forget my
plan, but also not think to A (and thus not A).

8. But can’t we think about nonexistent things? Sure, but the intentionality or
directedness of a thought isn’t a genuine relation (Brentano 1995, 271–274; cf.
Molnar 2003, 62).

9. Fischer and Zimmerman make these claims with respect to a “broad” conception
of omission; both recognize that there are narrower conceptions.
Intentional Omissions 153

10. Cases of this sort are offered by Frankfurt (1969) to rebut the thesis that one
can be responsible for what one has done only if one could have done otherwise.

11. There is a sizable literature addressing the question of whether agents in Frank-
furt-type cases might be responsible for omitting to do certain things even though
they’re unable to do those things. See, e.g., Byrd 2007; Clarke 1994; Fischer 1985–
1986; Fischer and Ravizza 1998, ch. 5; McIntyre 1994; and Sartorio 2005. Partici-
pants in this debate evidently take there to be some type of ability to A that isn’t
required for omitting to A.

12. Ginet (2004) takes the case to support his claim that intentionally not A-ing at
t requires that one could have A-ed at t (or at least could have done something by
which one might have A-ed at t). I’ve suggested that whether the requirement holds
depends on which type of ability to act is being invoked.

13. To be precise, it is “intentionally not doing” that is the target of Ginet’s analysis.
However, his examples are cases of what I’m calling intentional omissions.

14. McIntryre (1985, 79–80) discusses a similar case, though for a different purpose.

15. Al Mele suggested this case in conversation.

16. The objection was raised by a referee for Noûs.

17. For discussion of such a requirement, see Mele 1992, ch. 8. Bratman’s video-
game case (1987, 113–115) supports a different line of argument for the claim that
one can take something as a hoped-for goal, and try to achieve it, without intending
it. The case also supports the view that one can intentionally A without intending
to A.

18. Some writers distinguish between “direct intention” and “oblique intention,”
holding that although one might lack a direct intention to bring about certain
consequences, one has an oblique intention to bring them about if one’s so doing
is foreseen (or considered likely or certain). (The distinction stems from Bentham
1996, 86.) An oblique intention is said to be “a kind of knowledge or realisation”
(Williams 1987, 421).
Such a state differs importantly from what action theorists commonly call inten-
tion. The latter is typically distinguished by its characteristic functional role. Having
an intention to perform a certain action at some point in the nonimmediate future
(a future-directed intention) tends to inhibit subsequent deliberation about whether
to do that thing (though having such an intention doesn’t altogether preclude
reconsideration); it tends to inhibit consideration of actions obviously incompatible
with what is intended; and it tends to promote further reasoning about means to
what is intended. When one becomes aware that the time for action has arrived, a
future-directed intention to act tends to cause (or become) an intention to act
straightaway. Such a present-directed intention tends to cause an attempt to perform
the intended action. When carried out, a present-directed intention typically
154 R. Clarke

triggers, sustains, and guides action, often in response to feedback. (Bratman 1987
and Mele 1992, part 2, develop this conception of intention.) So-called oblique
intentions don’t play such a role in practical reasoning or action.
One might suggest that actions that are only obliquely intended can only be
obliquely intentional. It is hard to see more to the claim than an acknowledgment
that an action can be intentional even though the agent lacked an intention (of the
sort just characterized) to perform it.

19. I’ve described several cases (that of an individual step taken while walking, that
of trying to do something when one expects not to succeed, and that of foreseen
consequences that one has reason to avoid) that have been taken to undermine a
certain thesis about intentional action, viz., that intentionally A-ing requires intend-
ing to A. As a referee for Noûs observed, one might seek to defend that thesis by
appealing to a distinction between an action’s being intentional (under some
description or other) and its being intentional under a certain specified description.
One might claim, for example, that while what I do when I take the step is inten-
tional under some description (such as “walking”), it isn’t intentional under the
description “taking that step.” I don’t myself find this claim convincing. However,
it isn’t my aim here to refute the indicated thesis. What I’ve aimed to do is present
some forceful considerations that have been brought against it and show that there
is an equally strong, largely parallel case to be made against the view that intention-
ally omitting to A requires intending not to A.

20. Sartorio is mainly concerned to argue that no intention need cause an inten-
tional omission. Though she seems doubtful that any relevant intention need even
be possessed, she doesn’t commit herself on this latter question.

21. Sartorio (2009, 528, n. 22) suggests that “negative intentions” might have to be
rejected, for “it’s hard to say what the plan might be in the case of omissions.” In
some cases, one’s plan for not A-ing might be just not-to-A.

22. Some causalists (e.g., Mele 1992, ch. 2), in response to the problem of mental
causation, hold that (roughly) it is enough for intentional action if the neural real-
izers of the agent’s mental states play the appropriate causal role, provided that what
the agent does counterfactually depends, in a certain way, on her mental states. I’ll
henceforth simplify the causalist view by omitting this variation.

23. Weinryb (1980) argues that omissions have no causal effects. Beebee (2004)
argues that absences aren’t causes or effects.

24. For one thing, the argument relies heavily on explanatory claims. But true
explanatory claims don’t always cite causes, even when the explanations are causal.
On this point, see Beebee 2004.

25. As discussed earlier, Sartorio takes the omission to be intentional because, she
says, it is caused by the agent’s intentionally omitting to intend to jump into the
Intentional Omissions 155

water. I’ve raised questions about how it can be made out that the agent’s not so
intending is itself intentional.

26. Might an argument like the one sketched earlier in this section show that, in
the normal case, it would be the absence of an intention to jump in, and not the
intention not to jump in, that causes not just the omission but also the subsequent
thought and action? Diana’s decision and her subsequent thinking might be brain
occurrences, and (in our original version of the case) the production of the latter
by the former might consist in just the sort of transfer of energy that we have in
standard cases of direct causal production. Moreover, her decision seems as plainly
a cause of her positive behavior in this version of the case as our intentions to act
seem to be causes of our intentional actions. One might try pushing a standard
argument against mental causation at this juncture, but the special argument
focused on omissions seems inapplicable.

27. One might take a lack of counterfactual dependence as at least a defeasible

reason for denying a causal relation. Then, if like Sartorio, one thinks that an inten-
tional omission to A must be caused by the absence of a certain intention, one might
take a lack of counterfactual dependence to count against the claim that the omis-
sion is intentional.

28. Bennett (1995, ch. 6) makes a similar point, albeit about what I take to be a
different distinction, that between making happen and allowing to happen.

29. For discussion of the problem of causal deviance, and for some proposed solu-
tions, see Bishop 1989, ch. 5; Brand 1984, 17–30; Davidson 1980, 79; Enç 2003, ch.
4; Goldman 1970, 61–63; and Mele 2003, 51–63.

30. It’s an interesting question, and one that I can’t answer here, whether the
required nondeviance in the case of intentional omission is susceptible to concep-
tual analysis. Given how little causal work might be required of one’s intention in
such a case, there’s reason to doubt that causation “in the right way” is analyzable
here in the same way that it is in the case of intentional action. Whether the impos-
sibility of conceptual analysis would be fatal to a causal theory of intentional omis-
sion depends on what is to be expected of such a theory. (I’m grateful to a referee
for raising this issue.)

31. One might (as a referee suggested) take Diana’s not jumping in to be a causal
consequence of her intending not to jump in on the following grounds: her so
intending causes her standing on the shore, and (one might hold) her not jumping
in is a logical consequence of her standing on the shore. It would seem that, barring
time travel, Diana’s jumping into the water at t is logically incompossible with her
standing on the shore at t. However, it is far from obvious that an event causes the
absence of whatever is logically incompossible with what that event causes.
10 Comments on Clarke’s “Intentional Omissions”

Carolina Sartorio

Clarke argues for two main claims in his essay. The first is:

(i) In order for an agent’s omitting to A to be intentional, some intention

with the appropriate content (e.g., the intention not to A or a related
intention) must play a causal role in the situation.

This is the main source of disagreement between Clarke’s proposal and my

proposal. I argued that, at least ordinarily, the agent’s omitting to intend to
A (which is itself an intentional omission) causes his omitting to A, and
that this is enough to explain why the agent’s omitting to A is intentional.
There is no further need to say that the agent’s having formed a certain
intention also plays a causal role (in particular, by causing his omitting to
A).
Clarke also argues for:

(ii) The causal role the relevant intention plays in each case is that (or
includes the fact that) it causes the agent’s subsequent thought and action.

I’ll briefly comment on each of these claims.

1 Clarke’s Argument for (i)

This comment is partly a request for clarification, since I’m not sure I
understand exactly what Clarke’s argument for (i) is. Clarke doesn’t seem
to have objections to my claim that, when an agent intentionally omits
to A, the agent’s omission to intend to A causes his omitting to A. But,
presumably, he thinks that this isn’t enough to explain why the agent’s
omitting to A is intentional. For, he seems to think, such a fact would only
explain why the omission to A is intentional if it were clear that the omis-
sion to intend to A is also intentional; however, it is hard to say what it is
for an agent’s omitting to intend to act to be intentional (Clarke, this vol.,
chap. 9, 143–144 and n. 24).1
158 C. Sartorio

I agree that it might be hard to specify precisely the conditions under

which an omission to intend to A is intentional. But, surely, we can make
sense of the concept of intentionally omitting to intend to act, and there
seem to be some clear applications of such a concept. I will consider a
couple. Take, first, an agent who is unaware of the presence of the child
in the water. According to a broad conception of omissions that Clarke
thinks is theoretically fruitful (one on which “there is an omission when-
ever an agent doesn’t perform a certain action that she is, in a relevant
sense, able to perform,” 141), one of the things that the agent omits to do
in this case is to form the intention to jump in (since she didn’t form the
intention to jump in and, arguably, she was, in the relevant sense, able to
do so). But, surely, her omitting to intend to jump in isn’t intentional
(jumping in didn’t even cross her mind). By contrast, take one of Clarke’s
characters, Diana. Diana is fully aware of the presence of the child but still,
after deliberating about it for a while, decides not to save him and to con-
tinue to eat her ice cream on the shore. Again, according to the extant
conception of omissions, Diana omits to intend to jump in. Is this omis-
sion intentional? Surely, it is: one of the differences between the two cases
is that, although both agents omit to form the intention to jump in, Diana
does so intentionally, unlike the agent in the first case.2
But, if Diana’s omitting to form the intention to jump in is intentional,
and if her omitting to form the intention to jump in causes her omitting
to jump in, then why is it that some intention of hers must also play a
causal role in the situation, in order for her omitting to jump in to be
intentional?

2 Clarke’s Argument for (ii)

Clarke argues that, when Diana decides not to save the drowning child
and to continue to eat her ice cream on the shore, her intention not to
jump in causes her subsequent action of continuing to eat ice cream on
the shore. This is an interesting proposal that would (if true) preserve the
causal efficacy of negative intentions, and would also bring actions and
omissions a bit closer together, as Clarke points out. I will argue, however,
that some plausible assumptions (including assumptions that Clarke
explicitly accepts) suggest that Diana’s negative intention doesn’t actually
play such a causal role.
For Clarke, a main reason not to identify omissions with actions (at least
generally) is that actions and omissions tend to have different causal
powers. As an example of this, he gives the following case (137). Ann
Comments on Clarke’s “Intentional Omissions” 159

promised to pick up Bob at the airport late at night. Feeling lazy, and
thinking that he can take a cab, she decides to stay home playing piano
instead. This keeps her neighbor awake. Clarke believes that, whereas Ann’s
playing piano causes her neighbor to stay awake, her omitting to pick up
Bob doesn’t. Again, this is a reason not to identify Ann’s omission to pick
up Bob with her action of playing piano. I agree.
Now consider the intentions that Ann forms in this case. Presumably,
Clarke would say that we should distinguish a “positive” intention—the
intention to play piano—from a “negative” intention—the intention not
to pick up Bob at the airport. Which of these intentions causes her to play
piano? Surely, her intention to play piano causes her to play piano. But,
does her intention not to pick up Bob also cause her to play piano? Given
what Clarke wants to say about the Diana case, it seems that he would
have to say that it does: given that Ann’s omitting to pick up Bob was
intentional, her intention not to pick up Bob must have caused her sub-
sequent thought and action, including her playing piano. But this is an
odd result, given Clarke’s initial assumption about the causal powers of
actions and omissions. Why think that bodily positive and negative acts
have different causal powers but their mental counterparts (positive and
negative intentions) don’t? If there is good reason to think that Ann’s
omitting to pick up Bob doesn’t cause what her playing piano causes, isn’t
there also good reason to think that her intending to omit to pick up Bob
doesn’t cause what her intending to play piano causes?
I believe that there is, in fact, good reason to think this. Presumably,
the reason to think that Ann’s omitting to pick up Bob doesn’t cause her
neighbor to stay awake is that her neighbor stays awake, intuitively, not
because of what she doesn’t do that night, but because of what she does.
Similarly, it also seems that Ann plays piano that night, not because
of what she intends not to do that night, but because of what she intends
to do.
One could try to object to this by saying: there is a sense in which she
plays piano because of what she intends not to do. Namely: in the circum-
stances, were it not for the fact that she intended not to pick up Bob, she
couldn’t have intended to stay at home playing piano, and then she
wouldn’t have played piano. But the same is true of her omitting to pick
up Bob and the outcome of her neighbor’s staying awake: were it not for
the fact that she omitted to pick up Bob, she couldn’t have stayed at home
playing piano, and then her neighbor wouldn’t have stayed awake. So it
seems that whatever reasons we have for thinking that Ann’s omitting to
pick up Bob doesn’t cause her neighbor to stay awake are also reasons for
160 C. Sartorio

thinking that her intending not to pick up Bob doesn’t cause her playing
piano.
In support of his claim that the agent’s negative intention must cause
the subsequent positive behavior in order for his omission to be inten-
tional, Clarke imagines a case where the agent (Diana) forms a negative
intention but then the intention gets to play no causal role at all (it is
preempted by a chip implanted by neuroscientists, which causes Diana to
continue eating her ice cream on the shore). In this case, Clarke argues
that Diana’s omitting to jump in isn’t intentional. But, even if this is true,
this doesn’t show that, for Diana’s omission to be intentional, her intend-
ing not to jump in must cause her continuing to eat ice cream. Perhaps
she must intentionally omit to intend to jump in and such omission must
cause her omitting to jump in (as I propose). Or perhaps I am wrong and
her intending not to jump in must cause her omitting to jump in. Or,
finally, perhaps three things have to happen: (i) she must intend not to
jump in; (ii) she must intend to perform a different act incompatible with
jumping in (e.g., eating ice cream on the shore at the time); and (iii) such
an intention must cause that action. There could be other possibilities.
Presumably, none of these sets of conditions is met in the neuroscientist
scenario. The chip, and only the chip, accounts for Diana’s subsequent
behavior in that case. Hence the neuroscientist case doesn’t support
Clarke’s claim that, for an agent’s omission to be intentional, the agent’s
negative intention must cause her subsequent thought and action.

Acknowledgment

I am grateful to Andrei Buckareff and Randolph Clarke for helpful

discussion.

Notes

1. All page numbers refer to Clarke’s chapter 9 of this volume.

2. A natural thing to say, at least in this case, is that what makes Diana’s omission
to intend to jump in intentional is the fact that she considered possible reasons to
jump in but she took those reasons to be outweighed by reasons to continue eating
her ice cream on the shore.
11 Reply to Sartorio

Randolph Clarke

After drinking with his buddies one evening, Tom was tired. While they
vowed to carry on all night—and did—he went home and slept. Tom
intentionally omitted to join them in toasting the sunrise. Was he engaged
in some kind of behavior at dawn? Something dormitive, perhaps, but
nothing of the kind that action theory aims to characterize.
We can perfectly well use the term “behavior” in a broader sense, to
cover all things done intentionally, including Tom’s omitting to drink till
dawn. But if we do, we should see that “behavior” is a disparate category,
including actions and things that aren’t actions. And we should then not
expect a highly uniform theory of this broader phenomenon.
If, then, the right account of intentional omission doesn’t precisely
parallel a causal theory of action, does this fact make trouble for “causalism
as an attempt to explain what it is for an agent to behave intentionally”
(as Sartorio, this vol., chap. 8, 115, claims)?1 That depends on just how the
correct account goes and to what exactly “causalism” is committed.
Sartorio alleges that a proponent of causalism will hold that when one
intentionally omits to A, one’s omission is caused by one’s forming an
intention not to A (or an intention with some other relevant content). But
a causal theory of action doesn’t commit one to this view, since it isn’t a
theory of things that aren’t actions; and one can be a causalist about
intentional behavior, broadly construed, without holding this view.
Indeed, Sartorio’s account of intentional omission is causalist while reject-
ing the indicated view, for she holds that a paradigmatic intentional omis-
sion is caused by one’s omission to intend.
Still, as her account is spelled out, something that a standard causal
theory requires for intentional action—an important causal role for mental
states or events—is not said to be required for intentional omission. If there
is no such requirement, that is an important fact, even if it doesn’t force
the rejection altogether of a causal approach to intentional behavior.
162 R. Clarke

However, I’m not convinced that no such requirement falls on inten-

tional omission. Consider, first, what Sartorio says suffices for intentionally
omitting to A: that the omission is caused by one’s intentionally omitting
to intend to A (or to have some other pertinent intention). It strikes me
as a misstep to try to account for one omission’s being intentional by
appealing to yet another omission’s being intentional. Further, the notion
of intentionally omitting to intend is rather obscure, and I don’t see that
Sartorio provides much clarification of it.
One does not intentionally omit to intend to A, she holds, if A-ing does
not even cross one’s mind; that much seems right. One intentionally
omitted to intend to A, she says, if one “voluntarily failed to form that
intention, after deliberating about whether to do so, after considering
reasons for and against doing so, and so on” (126). As I observed in my
essay (chapter 9 of this vol.), the claim that the omission to intend was
voluntary seems to presuppose that, rather than explain how, it was inten-
tional. And the fact that one had deliberated and considered relevant
reasons and still not formed the intention doesn’t suffice to make the
omission to intend intentional, for one might have simply failed to make
up one’s mind.
It’s not just that Sartorio hasn’t specified “precisely the conditions under
which an omission to intend to A is intentional” (chap. 10, 158). There
are few philosophically interesting phenomena for which we can do this.
Rather, the notion appealed to here is left quite unclear, and the appeal
seems both unnecessary and unhelpful. We are not told, for example,
whether intentionally omitting to intend to A requires that the omission
to intend to A be caused by something in particular (some further omis-
sion, or the individual’s mental states), or whether in order for that omis-
sion to be intentional, there must be some further intentional omission to
intend. The latter requirement, of course, would threaten an endless
regress; but if the regress can be stopped at this point, why can’t it be
avoided altogether?
Although Sartorio doesn’t claim that intentionally omitting to intend
to A (or to have some other pertinent intention) is necessary for intention-
ally omitting to A, it is perhaps worth observing that it doesn’t seem to
be. One might intentionally omit to A without having deliberated about
whether to A, and without having made a decision on the matter. One
might spontaneously—and nonactively—acquire an intention not to A
without having raised the question of whether to A, just as one can so
acquire an intention to act. (Audi 1993, 64 provides a nice example of
the latter. As for the former: I see a snake crossing the path ahead, and I
Reply to Sartorio 163

immediately intend not to take another step forward. I refrain from walking
further.) If one’s omitting to intend to A (e.g., to continue walking) might
nevertheless be intentional, it remains to be explained how this can be so.
Indeed, strictly speaking, lacking an intention to A isn’t necessary for
intentionally omitting to A. One can have intentions of which one is
unaware, just as one can have beliefs and desires of which one is unaware.
Unaware that one intends not to A, one might A, and do so intentionally—
meaning then to A (and A-ing attentively, carefully, expertly). Similarly,
unaware that one intends to A, one might refrain from A-ing, meaning
then not to A.
Just how far Sartorio’s view of intentional omission diverges from what
causal theorists require for intentional action depends on what she thinks
about omitting for reasons, something that she doesn’t discuss. Things
done intentionally are typically done for reasons. If one’s omitting to
intend on some occasion is intentional, and done for reasons, we might
ask in virtue of what the latter is so. Must certain of one’s reason-states—
one’s beliefs, desires, affections, aversions, and the like—be causes of one’s
omitting to intend? If Sartorio accepts such a requirement, that will render
her account more thoroughly causalist, even if it still denies intentions any
necessary causal role. If she denies the requirement, then we might fairly
request a sketch of some alternative view.
I claimed in my essay that in order to intentionally omit to A, one’s
intention not to A (or some other pertinent intention) must cause some
of one’s subsequent thought and action. It must play a causal role with
respect to what does happen, even if it need not cause any absences. Sar-
torio asks what my argument for this claim is. I observed that in standard
cases of intentional omission, such an intention does in fact cause such
things. And when I considered a case in which this was not so, it seemed
to me that the omission was then not intentional, and that it failed to be
intentional because the intention in question didn’t play the indicated
causal role.
In the imagined case (this vol., chap. 9, 147–148), Diana decided not
to jump into a pond to save a drowning child. She then wondered what
to do instead, formed an intention to walk over to get a better view of the
impending tragedy, and did so. But her intention not to jump in caused
none of these things; they were all caused by a chip that had been implanted
in Diana’s brain earlier by a team of neuroscientists, who (unaware of
Diana’s decision) just happened to have picked this moment to test their
device. They got lucky, for the chip, I noted, had an unforeseen flaw: it
would have remained inert if, when the activation signal was sent, the
164 R. Clarke

agent in whose brain it was implanted had just decided to jump into
water.
Sartorio finds no support here for the causal requirement I proposed.
There are, she says, other possible explanations of why Diana’s not jumping
in isn’t intentional. She suggests, first, that Diana’s omission might fail to
be intentional because her omitting to intend to jump in doesn’t cause her
not jumping in. But is that so? The chip causes Diana’s thought and action,
but, as Sartorio recognizes, it’s a further question what causes Diana’s omit-
ting to jump in. As the case is imagined, if Diana had intended to jump
in, she would have; her not so intending made a difference to what she
did. If omissions to intend ever cause omissions to act, it isn’t clear why
this one doesn’t.
Second, Sartorio suggests that Diana’s omitting to jump in might fail to
be intentional because her intending not to jump in doesn’t cause that
omission. But if that is correct, it suggests that intentions not to act must
play an even more robust causal role in intentional omissions than what
I argued, and the right account of intentional omission will, after all,
closely parallel a causal theory of intentional action.
Finally, Sartorio suggests that in order for Diana to have intentionally
omitted to jump in, she must have intended to perform some act incom-
patible with her jumping in, and that intention must have caused that
action. But, as I think the paint case from my essay (this vol., 138) shows,
intentionally omitting to A doesn’t require performing any action that one
takes to be incompatible with A-ing; it doesn’t require, either, intending
to perform any such action.
There might, in fact, be more than one correct explanation of why
Diana’s omitting to jump in isn’t intentional, as there will be if several
necessary conditions are unsatisfied. Though the case seems to me sup-
portive, perhaps no single example will decisively show my proposal
correct. The proposal can, however, be undermined if there is a case in
which someone intentionally omits something and yet no relevant inten-
tion plays the indicated causal role.
I said that in standard cases of intentional omission, one’s intention
not to act (or other relevant intention) is in fact a cause of one’s subsequent
thought and action. Sartorio disputes this claim. In my case involving Ann
(135), she maintains, it is Ann’s intention to play piano, not her intention
not to pick up Bob, that causes her piano playing. But the two intentions
aren’t competitors; they are, respectively, later and earlier members of a
causal sequence leading to Ann’s playing piano. The intention not to pick
up Bob causes Ann to consider what to do instead, which causes her to
Reply to Sartorio 165

decide to play piano, which causes her piano playing. One need not hold
that causation is necessarily transitive to find plausible the claim that Ann’s
intention not to pick up Bob indirectly causes her piano playing.
Comparing Ann’s intention not to pick up Bob with her omitting to
pick him up, Sartorio asks, “Why think that bodily positive and negative
acts have different causal powers but their mental counterparts (positive
and negative intentions) don’t?” (chap. 10, 159). But Ann’s intention not
to pick up Bob is an intention, whereas her omitting to pick him up is not
an action. Were the case real, the former would be an actually existing
being, negative only in its content (just as is my belief that Santa Claus
doesn’t exist). In contrast, the latter (I’m inclined to think) would be an
absence of being—the absence of an act by Ann of picking up Bob at the
airport. A causal impotence of the latter is no reflection on the former.
Although I don’t, in my essay, dispute Sartorio’s claim that omissions
to intend cause intentional omissions, I’m in fact doubtful about this. At
bottom, I find it doubtful that an absence of being can cause something.
Sartorio writes of omissions having causal powers; but I don’t see how
nonbeings can have any such powers.
I’ve advanced considerations favoring an account of intentional omis-
sion that makes no appeal to causation by absences. The proffered view
accords a causal role to mental states, including intentions, and sees inten-
tional omissions as resulting—even if noncausally—from such states. If an
account along these lines is correct, omissions make trouble neither for
causal theories of action nor for causal approaches to intentional behavior,
broadly construed.

Note

1. Unless otherwise noted, all page numbers refer to Sartorio’s chapters in this
volume.
12 Causal and Deliberative Strength of Reasons for Action:
The Case of Con-Reasons

David-Hillel Ruben

An agent’s having of a reason for an action (hereafter, simply “a reason”)

is often said to be among the causes or causal conditions of the action for
which it is a reason (in this wide sense, “action” includes many cases of
inaction).1 Hereafter, this view is referred to as causalism, or (1). Causalism
as here understood is a thesis about causation, not about causal
explanation.
Causation and causal explanation might come apart in both directions,
as it were. There might be causal explanations that do not cite any or only
causes, under any description. In any case, it is clear that there can be true
causal statements such that the cause is not explanatory of its effect
because of the description of the cause in that statement, but there might
even be causes such that, under no description of the cause, does the cause
explain its effect. In this essay, I remain agnostic about all such
possibilities.
An agent can, of course, have a reason for an action without its being
a cause of that action, but in the case in which the agent performs the
action because of that reason, the reason is said, according to causalism,
to cause the action. If we conflate just for the moment questions about
causation and causal explanation, Donald Davidson’s well-known argu-
ment attempts to demonstrate just this: “If . . . causal explanations are
‘wholly irrelevant to the understanding we seek’ of human actions then
we are without an analysis of the ‘because’ in ‘He did it because . . .,’ where
we go on to name a reason” (Davidson 2001a, 86–87). The reasons David-
son focuses on, in this argument and elsewhere, are reasons which function
as “pro-reasons,” in two closely connected senses: (a) these are reasons that
are relevant to, or bear upon, the action the agent does in fact take; (b)
they are reasons that favor that action. Davidson’s reasons consist of a
belief and a pro-attitude toward the kind of action the agent does. Other
views might identify intentions or some other mental items as these
168 D.-H. Ruben

reasons, but precisely which mental items count, on a causalist view, as

reasons will not concern me here (von Wright 1978, 46–62). I also take no
view here about the nature of such mental items; they might or might not
be token-identical to brain or other physiological states.
The reasons that (1) is about are sometimes called “explanatory” reasons
in contradistinction to “justifying” reasons, or “motivating” reasons in
contradistinction to “normative” reasons, or “internal” reasons in contra-
distinction to “external” reasons. These are three different distinctions,
related to one another in complicated ways. The reasons required by (1)
are surely at least motivating: these are the reasons that have actual psy-
chological “purchase” on the agent, and are not just there and merely in
principle available to the agent in some wholly objective sense. A well-
known view, which is also adopted by the position I am considering,
asserts that the only reasons that could play the causal role required by
the idea of motivation, and hence explain why the agent did what he did,
are the agent’s internal psychological states. So the reasons required by (1),
as I construe it, are internal, explanatory, and motivating, although these
reasons, just insofar as they are reasons for action, are being asked to play
something of a weakly normative role as well.
The literature in action theory has tended to overlook the fact that
reasons function in another way too.2 (a) One can have reasons for an
action that one does not take; (b) one can have reasons that disfavor an
action taken (a “con-attitude,” to parallel Davidson’s “pro-attitude”).
Sometimes a person has conflicting reasons for acting, one set of which is
a set on which he does not act, and both sets of conflicting reasons can be
rationally or deliberatively relevant to, or bear upon, the same choice situ-
ation and rationally or deliberatively relevant to the same action finally
chosen. The con-reason is relevant to the action taken, in the sense that
that it is the action it disfavors. Both the pro- and the con-reason, as I shall
continue to say, are “rationally or deliberatively relevant” to, or “bear
upon,” the same eventual choice made or action taken. The pro-reason is
relevant to the action taken because it favors it; the con-reason is relevant
to the action taken because it disfavors it. Each set of reasons justifies or
supports a different proposed action on the agent’s part, and the agent is
not able to perform both actions, because it is impossible to act in both
ways at the same time.
The reasons might strongly conflict, in the way in which a reason to do
some token act of type A (or, as I sometimes elliptically say, a reason to A)
and a reason not to do any token act of type A conflict; or the reasons
might weakly conflict, in the way in which a reason to do some token act
Causal and Deliberative Strength of Reasons 169

of type A and a reason to do some act of type B (or, as I sometimes ellipti-

cally say, a reason to B) conflict on any occasion on which one cannot as
a matter of fact do both. On an occasion on which one cannot do both,
a reason to do an act of type B must also be a reason not to do an act of
type A, but only modulo the additional information that one cannot do
both acts in the circumstances. In the case of strong conflict, no additional
information is similarly required.
Suppose that X has a reason to do a token action of type A and a reason
to do a token action of type B, where the two reasons weakly conflict on
the particular occasion. Suppose further that X chooses to perform a token
action of type A. The first reason, which favored doing A, was rationally
or deliberatively weightier (in the circumstances, of course) than the
second, which favored doing B; the first counted for more, as far as X was
concerned. As we say, all things considered, X chose to perform a token
action of type A. (I write as if choice precedes every action, but my argu-
ment would be unaltered if choice were not ubiquitous in this way.) The
single final choice made or action taken is made or taken because of the
pro-reason and in spite of the con-reason. Indeed, that is what “all things
considered” must mean: because of the one set of reasons and in spite of
the other.3 As Dancy (2004, 4) says about these con-reasons: “But still I
was influenced by them [the con-reasons] and they do figure in my moti-
vational economy.”
We use the language of reasons, weights, and strength, in describing
our deliberations. Such language is metaphorical, but, metaphorical or not,
it certainly seems irreplaceable. We can order reasons for action by their
comparative strengths: one reason for action can be stronger than another,
weaker than a third. Whether or not there are nonrelational truths about
the deliberative strength of reasons, on which the relational truths about
them supervene, it is relational information about reasons that is crucial
for understanding the deliberative story.4 (Note that if there is such super-
venience, the same relational story can supervene on different nonrela-
tional stories. What supervenience rules out is the possibility of different
relational stories supervening on the same nonrelational or intrinsic story.)
If we have both pro- and con-reasons, we want to know which reason wins
the deliberative contest. It is only the relational information about the
rational strength of reasons for action that interests us in cases of choices
between actions.
The notion that reasons are causes, causalism, ties rationality and cau-
sality, in some way yet to be discerned. This would allow us to distinguish
two kinds of strengths for reasons: rational or deliberative strength (as
170 D.-H. Ruben

above) and causal strength. Deliberative or rational strength is a measure

of the extent to which a reason for action supports that action. Rational
strength is an epistemologically normative idea, sometimes called “weak
normativity.” There is an analogy between this idea and the idea of the
degree of support given by evidence to a conclusion in an inductive argu-
ment. A reason’s comparative or relational rational strength compares it
to the support competing reasons give to the alternative actions they
support.
The problem of weakness of the will arises for the causalist as an issue
because it challenges the natural tie, on the causalist program, between
rational and causal strength of reasons for action. For the causalist, weak-
ness of the will appears to force a wedge between the normative/rational
and the causal/motivational ideas of a reason. Davidson himself says that

(2) “if reasons are causes, it is natural to suppose that the strongest reasons
are the strongest causes” (Davidson 2001a, xvi).

Any causal view is going to have to address the question: how are rational
and causal strength related? It is this question that gives rise to the problem
that the causalist faces with weakness of the will. In a case of weakness of
will, the rationally stronger reason is not the reason that causes or moti-
vates the agent to act, if reasons do in fact cause actions. The agent acts
on a rationally weaker reason, but one that is causally strong enough,
where the rationally stronger reason is not causally strong enough. “Caus-
ally strong enough or not so” just means: the rationally weaker reason
causes the action it supports, and the rationally stronger reason does not
cause the action it supports.
I do not know whether the causalist can really successfully deal with
the phenomenon of weakness of the will. But I want, in this essay, to
address a different issue, unconnected to weakness of the will. Let’s start
by trying to trace out the causal chains that lead from the con-reasons, for
on the causalist view con-reasons must have some effects, whatever they
might be. The thought that there are events, con-reasons, which have no
effects at all, is not one likely to appeal to the causalist. To be part of the
causal order is surely to have both causes and effects.5
There are two importantly different cases that I want to distinguish
(from the causalist point of view). In cases of Type I, the pro-reason and
the con-reason jointly cause the same effect; in cases of Type II, they have
separate and causally independent effects.
Type I: These cases are the ones that will most naturally spring to mind
on a causalist view, but I believe that such cases are more limited than one
Causal and Deliberative Strength of Reasons 171

might unreflectively assume. Let’s start with an analogy from natural

science. In circumstances c, a ball falls toward the Earth because of gravi-
tational attraction and despite the presence of an upward wind. One could
say: the cause of the ball falling to Earth is jointly both the gravitational
attraction and the relative weakness of the wind’s counteracting force. The
two causal factors jointly contribute to the same result, the ball’s final
trajectory. The relative weakness of the wind is a causal factor in the ball’s
actual fall to Earth. If there had been no, or even less, weak counteracting
force of the wind, the ball would have fallen to Earth faster, somehow
differently, or some such. Perhaps the difference is temporal: if there had
been no or less a counteracting force of the wind, the ball would have
fallen to Earth sooner, earlier. Both factors were parts of the full cause; had
either been absent, the result would have been (or probably would have
been) different, or at least differently placed temporally.
A Type I example in the case of action would also not be hard to find,
and, as I said above, it is these cases that spring most readily to mind:
suppose Buridan’s ass is drawn to hay pile A because it is more attractive,
hay pile B is less attractive. (In a real Buridan’s ass case, the piles are equally
attractive, but not so in the case I am now imagining.) No starving ass
here. Both the attractiveness of A and the relative unattractiveness of B
result in the ass’s choosing hay pile A. We are all such asses much of the
time; many cases are of this type. Again, the same causal factors jointly
contribute to the same result, the ass’s final choice. The relative unattrac-
tiveness of B is a causal factor in the ass’s actual choice of A over B. If pile
B, although unattractive, had not even been available, or had been even
less attractive, but pile A had remained the same, the ass would have
chosen hay pile A more quickly, earlier, more determinedly, with less hesi-
tation, or some such. Since both factors, the degree of attractiveness that
pile A had and the degree of unattractiveness that pile B had, were both
parts of the full cause of the ass’s token choice of A, had either been absent,
the result would have been (or probably would have been) different, either
in character or in time. For example, absent pile B, the ass would have still
chosen pile A, but the token choosing would have been different in some
significant way from the actual choosing, as a result of the difference in
its cause.
In examples of cases of Type I, it is assumed that both causal factors
jointly influence the same outcome, so the outcome would have been dif-
ferent, or probably would have been different, if either of the causal factors
had been different. Both causal factors matter to the same outcome.
Although something like this might be true, and indeed no doubt is true
172 D.-H. Ruben

in many cases, it need not be. That is, on the causalist program, there could
be other cases (Type II) in which the presence of the con-reason has some
effect, but has no effect of any sort on the actual action taken, but instead
has some effect on something else. I do not know how to prove that there
must be such cases for the causalist, but it seems to me intuitively clear
that there could be.
What would the causalist have to hold, in order to deny this claim that
I have just made? Since in cases of joint causation of a single effect by
multiple part causes, it follows that the effect would have been different,
or probably would have been different, had any one of the part causes
been different (or altogether absent), the causalist who wishes to deny the
possibility of cases of Type II would have to say that:

(3) In every case of action for which the agent has both pro- and con-
reasons that figure into his deliberations, had the agent not had such a
con-reason, the action he took would have been different or altered, or
probably would have been different or altered in some way, or occurred
at a different time.6

I just don’t think that (3) could be true. I can envisage many cases in which,
were the con-reason absent, the action taken could be qualitatively the
same (in all nontrivial respects) as the action that was actually taken. We
can say of such cases: “the agent had, and acknowledged that he had, a
less weighty reason not to do something, which figured into his delibera-
tive activity, but that less weighty reason did not at all causally influence
his eventual choice to do what he did, in any way.” Perhaps such a case
might be one in which the agent has, and acknowledges that he has, a
weak moral reason that he does consider in his deliberations, but the weak
moral reason has in the end no actual effect on his eventual choice or
behavior. Or a case in which the ass considers both hay piles in its delib-
eration but is so determined to get to hay pile A that he would make exactly
the same choice regardless of what he acknowledges to be the lesser but
not negligible attractiveness of hay pile B, and so the ass would make the
identical choice—a choice qualitatively identical in character, timing, and
so on—had hay pile B not been available at all. That is, had the agent not
have had the con-reason, his actual choice would have been (or probably
would have been) qualitatively the same in all relevant respects. Cases of
Type II already presume that reasons and causation even on the causalist
program can part company to this extent: a con-reason must cause some-
thing, but the con-reason might not be a part-cause of the same effect that
the pro-reason causes or part-causes.
Causal and Deliberative Strength of Reasons 173

The case of temporal location might merit special consideration. Perhaps

the action chosen in the face of both pro- and con-reasons could be intrin-
sically qualitatively the same as the action the agent would have chosen
had he only had the pro-reason. But at least won’t the time of the actions
be different? If the agent had had no con-reason, he could not have delib-
erated and weighed up pro- and con-reason, and however short a time that
deliberation might have taken, had he had no con-reason, the action
would have occurred just that much sooner.
It might be so, but then again, it might not be so. Consider cases in
which the con-reason, although it remains a con-reason, is so obviously
(to the agent) weaker than the pro-reason that the agent has no need to
go through some actual deliberative process. The con-reason is not strong
enough to act as the countervailing wind did in my earlier example, as a
force dragging and delaying the decision to act. So the action taken and
the action the agent would have taken had he had no con-reason would
display no difference in time of occurrence.
I don’t think these cases are at all far-fetched. If we focus on cases in
which decisions are difficult, in which pro- and con-reasons are finely bal-
anced, it will seem that the con-reason, if causalism is to be believed, must
exert some sort of causal influence on the action taken, even if only a dif-
ference in temporal placement. But we are not normally so conflicted in
our choice of action. Think instead of cases, and I suggest that these will
be the vast majority of cases, in which, although the agent has a con-
reason, or con-reasons, that in some sense “weigh” with him, the relative
weighting of the pro- and the con- is clear and obvious. Deliberation is not
necessary. In those sorts of cases, I can see no reason to believe that the
con-reason, if it has causal influence, must display that causal influence by
affecting the character or time of the action actually taken.
So, let it just be stipulated that we are considering a case of Type II, in
which the agent does something in the circumstances in which he does
have conflicting pro- and con-reasons, but that he would have also done
that action in an intrinsically qualitatively identical manner, and at the
same time, had he only had the one set of pro-reasons. His “opposing”
con-reason does not make him hesitate, or dither, in doing whatever it is
that he does, in any way. There must, therefore, be cases in which the
con-reason is not causally necessary for the actual action taken, if it is, as
the causalist insists it is, a cause at all.
But if causalism is true, the con-reason must be a cause or part-cause or
causal factor of something else other than the choice or action actually
taken. In such cases, if causalism is true, the pro-reason will initiate a causal
174 D.-H. Ruben

chain leading to the action taken, and the con-reason must initiate a dif-
ferent, independent causal chain that leads to something else. One thing
that the con-reason can certainly not cause is the action it favors, since
that action never happened and therefore nothing could cause it. To be
sure, that something does not occur can have a cause, but what does not
occur can have no cause since it does not exist.
If we assume that the con-reason does not also contribute to the causa-
tion of the action it disfavors, but rather would have to cause something
else, there is any number of possible candidates for the effects of such
con-reasons available to the causalist. Perhaps a person’s con-reason directly
causes regret (Williams 1981, 27ff.), or causes some other change in his
mental landscape (his dispositions to act, for example), or causes some
psychological illness in him. He does the action favored by the pro-reason,
but since he had reasons against it, his con-reason ends in him regretting
what he did, or some such. Or perhaps the effect of the con-reason is not
even at the personal level at all. Might its effect not be some physiological
or brain event of which the actor is perhaps ignorant or unaware?7 (Or,
“some further physiological or brain event,” if the having of a con-reason
is such a physical event too.)
The important feature of all these candidate effects for cases of Type II
is that they require a second causal chain, in addition to the one that goes
from the stronger pro-reason to the action taken. If so, there would be one
causal chain leading from his having a pro-reason to his subsequent action.
There would be another quite independent causal chain leading from his
con-reason to his subsequent regret, or illness, or to some (further) physi-
ological or similar event. The causal chains would not converge causally
on the final choice or action, as they would if both pro- and con-reasons
causally contributed to the same action taken, as we sketched above in
cases of Type I.
On this rather simple picture, the pro-reason initiates a causal chain
leading to the action; the con-reason initiates a wholly independent,
second causal chain, leading to the regret or brain state or whatever. One
thing to note about this view is that it might not permit us to capture
causally the idea that both pro- and con-reason are rationally or delibera-
tively relevant to the same token final choice or action. The con-reason
might not be a reason against acting in a certain way in virtue of whatever
causal role it plays. A con-reason could not be the con-reason it is (a reason
not to do what was done) in virtue of its causing something else other
than that action. At the level of reasons for choice and action, the two
reasons bear differently on (one favors and the other disfavors) the same
Causal and Deliberative Strength of Reasons 175

choice or action, but the causal story might not mirror this in any way.
There would be just two distinct causal chains, each of which leads to a
different result; one leads to an action, the other to some psychological or
neurophysiological or dispositional state. But perhaps a causal model of
how pro- and con-reasons work in choice situations need not capture
within the causal model this fact about the rational significance that both
types of reasons have to the same action or choice, one in favor of it and
one against it, so I don’t take this as a decisive objection to the suggestion
under discussion.
My argument now focuses only on cases of Type II, such that pro- and
con-reason initiate independent causal chains. I do not deny that there
can be many cases of Type I, but these are not the ones I wish to consider.
In the cases on which I now focus, the pro-reason initiates one causal chain
and the con-reason initiates another, whatever it might be and to wherever
it leads.
In Type II examples, consider the actual situation, c. In c, the pro-reason
to A is rationally weightier for the agent than the con-reason to B. Caus-
ally, assuming (2), if there is no weakness of will, it is the pro-reason that
causes the agent to A, rather than the con-reason causing the agent to B
(so the con-reason causes something else, whatever that might be). But
now consider a counterfactual situation, c*.
c* is just like c, save in one feature, and whatever is a causal consequence
of that one feature: in c*, although the pro-reason retains the same delib-
erative weight that it has in c, the con-reason becomes much weightier.
This sort of scenario is very common. At a later time, an agent can assess
a reason as having more “gravitas” than he earlier imagined it had. It might
weigh more with him than it did before. So in c*, the con-reason counts
more for the agent. The agent does not judge that the reason to A has
become less strong than it was; it is just that the reason to B has become
deliberatively stronger, and so stronger than the reason to A.
The reason to B now rationally outweighs the reason to A in the agent’s
deliberations, so the agent now Bs rather than As. In c*, the reason to B
has become the pro-reason and the reason to A has become the con-reason.
Something about the reason to B has changed, and consequently the
ordinal information about relative strength of reasons has changed. But
nothing about the reason to A need have changed, other than certain
relational, ordinal truths about its deliberative strength.
At the level of decision, choice, and reason, this is all straightforward.
But how should we represent the allegedly underlying causal facts of the
matter in c* (in order to obtain a coherent causalist story)? In c*, the reason
176 D.-H. Ruben

to B is rationally weightier than the reason to A, so assuming that (2) is

true and that there is no weakness of will, the reason to B will cause the
agent to B instead of causing whatever it did cause in c. In c*, there will
now be a causal chain leading from the reason to B all the way to the
agent’s B-ing. But what does the reason to A now cause in c*?
Before we try to answer that question, let’s return for a moment to the
question about the relational and intrinsic strengths of reasons for action
that we briefly mentioned earlier on. Aside from the requirements of cau-
salism, I do not know whether there are intrinsic truths about the delibera-
tive strengths of reasons for action, on which their relational deliberative
ordering supervenes. I would prefer to remain agnostic on that issue too.
But it seems clear that causalism is committed to there being an intrinsic
reality to the causal strength of reasons. On causalism, reasons are also
causes (to put it succinctly), and there certainly must be an intrinsic reality
to causal strength. The whole truth about causes and their relative strengths
cannot be exhausted by only relational information. If c1 is causally stron-
ger than c2, there must be something intrinsic about c1 and c2 that makes
this relational fact true. One of the consequences of causalism’s tying
together rational and causal strength is that reasons must have strength
both in a rational/normative and a causal/motivational sense, as I claimed
earlier, and, as a result of that, reasons must also have causal strength in
an intrinsic sense.
Now revert to our two circumstances, c (the actual situation in which
the reason to A outweighs the reason to B) and c* (the counterfactual situ-
ation in which the reason to B outweighs the reason to A). In c, the reason
to A, the deliberatively strongest reason, caused the agent to A. Remember
that we are supposing that the only difference between c and c* is the fact
that the reason to B in c* rationally outweighs the reason to A and hence
causes the agent to B. In c*, the reason to A also undergoes a deliberative
relational change, since in c* it is outweighed by the reason to B but was
not so outweighed before in c. But there is no reason to think that there
must be some intrinsic causal change to the reason to A, simply in virtue
of the deliberative and causal changes to the reason to B, remembering
that in cases of Type II the causal chains initiated by the pro- and con-
reasons are independent. Whatever changes the relational deliberative
change to the reason to do A supervenes on, in the case described, they
may not include any intrinsic causal change in the reason to A. The delib-
erative relational change to the reason to A (for it is now outweighed in
c*) may supervene only on changes, deliberative and causal, to the reason
to B.
Causal and Deliberative Strength of Reasons 177

If so, then the reason to A should have the same causal strength in c*
as it had in c (even though it is now rationally outweighed by the reason
to B), and since the reason to A caused the agent to A in c, then the reason
to A should cause the agent to A in c* as well (with one possible exception,
described below). If the reason to A in c was strong enough to cause the
agent to A, then it should still have the same causal strength in c*, and
therefore should be strong enough to cause in c* whatever it caused in c,
given that there are no causal nonrelational differences between c and c*
as far as the reason to A is concerned. If the reason to A has the same causal
strength or power in both, then its effects should be the same in both
circumstances. What it is strong enough to cause in one, it should be
strong enough to cause in the other. The relational difference that in c*
the agent’s reason to A is outweighed rationally by his reason to B can’t
make a difference to what the former is causally strong enough to do, since
its causal strength is intrinsic.
So why doesn’t the agent do A in c*, just as he did in c? If the reason
to A is able in c to cause the agent to A, and if it has the same causal
properties in c* that it had in c, then it should still cause the agent to A in
c*. True, the reason to B gains in deliberative strength in c* (and so the
relational facts about the relative strength of both reasons will change from
c to c*). Given the causalist’s (2), what the reason to B causes, what its
causal strength is, must have changed from c to c*, a causal change on
which its new deliberative strength supervenes. So the reason to B should
also cause the agent to B in c*. There should be, in c*, as far as we can tell,
a standoff: the agent should be caused both to do A and to do B.
To be sure, the agent can’t do both A and B; by assumption, the agent
is not able to do both on a single occasion. But in the counterfactual situ-
ation c*, causally speaking, there should be no grounds for thinking that
the con-reason will now win out “over” the pro-reason. The con-reason is
now rationally and hence causally strong enough to cause the agent to B,
but the pro-reason remains at the same intrinsic causal strength and hence,
on the causalist view, should still be strong enough to cause the agent to
A. So why should we expect the agent to do one or the other? Why doesn’t
the agent do A rather than B, even in the counterfactual situation, since
his reason to do A remained in principle strong enough to cause him to
do A, or why doesn’t he do nothing at all, as in a true Buridan’s ass
example, since the two causes might cancel themselves out?
I mentioned one possible exception, above, to the claim that “since the
reason to A was strong enough to cause A in c, then the reason to A should
be strong enough to cause the agent to A in c* as well.” We need to take
178 D.-H. Ruben

note of this qualification. Perhaps the causal chains initiated by the reason
to A and the reason to B are independent in the actual circumstances, c
(they are not joint causes of a single effect or joint effects of a single cause),
but they might not remain independent in the counterfactual situation c*.
If the reason to B gains deliberative strength in c*, this relational change
might supervene on some changed intrinsic causal fact about it. Suppose
that in c* the reason to B, in addition to causing the agent to B, is now
able to interrupt the causal chain that would otherwise lead from the
reason to A to the agent’s A-ing, and that explains why the agent does not,
after all, do A in c*. Let’s consider this possible rejoinder to the difficulty
we have detected.
There would be some flexibility in deciding just where, in c*, the req-
uisite inhibitor blocked or stopped the chain commencing with the reason
to A from leading to its “natural” conclusion, A, as long as the chain did
not get all the way to that action. For the sake of argumentative simplicity,
let us suppose that the reason to B inhibited the very next link on the
chain. On such a chain, let m be the node that would have followed imme-
diately after the reason to A. So let us say that, in the counterfactual situ-
ation, what happened is that the reason to B inhibited or prevented m
from occurring, prevented or inhibited the reason to A from causing m,
and hence prevented the action A. That is why the agent does B instead
of A in the counterfactual situation, and why his reason to A does not lead
to his A-ing in c*, an explanation entirely consistent with the causalist
position.8 (In fact, the same result would be achieved if something else
other than his reason to B was the blocker or inhibitor, but the reason to
B is going to prove the most likely candidate for that role.)
The problem with this solution is simply that it is not true to the phe-
nomenological facts of the case. What this purported solution does is to
try and construe an agent’s not acting on a causally strong enough reason
that he has as a case of having that reason blocked or impeded by a con-
flicting reason that he also has. The identification doesn’t succeed.
Even apart from cases of weakness of the will, there is an indefinitely
large number of ways in which an agent’s wishes, wants, desires, and so
on can be thwarted. Bad luck affects us all. A typical sign of this happening
is agent frustration. In weak-willed cases, according to causalism, the
agent’s rationally strongest reason does not commence a causal chain
leading to an action because it is not causally strong enough; a rationally
weaker but causally strong enough reason does.
On the other hand, in the rather different case we are now considering,
the causalist rejoinder has it that the agent’s otherwise-causally-strong-
Causal and Deliberative Strength of Reasons 179

enough reason commences a causal chain that simply gets blocked. If an

agent’s causally strong-enough reason does not lead to action only because
the causal chain leading from it to action is blocked in some way, the agent
will be and feel thwarted. Something that he is causally driven to do, as
much as he is causally driven to do what he does do, gets closed off to
him.
Recall that on this view, since the reason to A is meant to retain in c*
whatever causal properties it had in c, it should therefore cause the agent
to act in c* as well as in c. If the agent failed to A in c*, only because the
causal efficacy of his reason to A had been blocked, even if by another
reason, the agent would feel this as some sort of failure. In one kind of
frustration or failure, an agent might feel frustrated because he did not do
what he thought was rationally the best thing to do. This kind of frustra-
tion is more properly, perhaps, thought of as a kind of defeat. It is the kind
of frustration that an agent experiences in cases of weakness of the will.
The agent knew that he had a stronger reason to do A, but did B instead.
Alas, his reason to B produced more psychic turbulence than did his reason
to A, in spite of the fact that rational or deliberative strength should have
inclined him otherwise. This is a case in which the agent acts on the ratio-
nally less weighty, but the causally more effective, reason.
On the other hand, the case we would need to envisage if the solution
being proposed worked is different but would also give rise to a kind of
frustration. It is a case in which the agent would fail to act on the reason
(albeit the less weighty reason, rationally speaking) that otherwise is
(equally) causally strong enough to drive him to act. In the case we are
considering, the agent has a rationally stronger reason to do B, and a
rationally weaker reason to do A. But (according to my argument) both his
reason to B and his reason to A should cause him to act, since each is
causally strong enough to lead to the action it supports.
But of course he does B in c*, not A, and we asked why this should be
so. We asked: in c*, why doesn’t his reason to A lead him to act too? The
suggested reply is that it does not lead him to act only because something
blocks the path leading from this otherwise causally strong enough reason
to the action for which he has that reason, thereby preventing him from
so acting. Here too, the agent would experience a kind of frustration. It is
not defeat or rational frustration, as in the case of weakness of the will. It
is the inability to act on a reason that would otherwise be causally strong
enough for an action, whatever its relational rational weight might be. It
is more like being unable to scratch an itch severe enough to drive you to
scratch (but, to be sure, when you have a rationally overriding reason not
180 D.-H. Ruben

to do so), because the frustration arises from the causal failure, not from
a rational failure. At a rational level, the agent would be happy that the
reason to A did not lead to his A-ing, since he had more reason to B than
to A. But at a causal level, he was primed to do A just as much as to do B.
But if he does not do what he is causally primed to do, he would feel
frustrated. It would indeed be like not being able to scratch an itch, when
the desire to scratch was as causally strong as any reason not to do so was.
How much weight should we put on these sorts of phenomenological
facts in deciding metaphysical matters? It is, I think, too easy to be a skeptic
about this. What we are deciding are not just metaphysical matters gener-
ally, but specifically issues in action theory. If some view in action theory
attributes to agents various kinds of mental states, or has the consequence
that they have those states, what better check is there than introspection?
I think that many views in action theory can be judged in this way, for
example, ones that attribute various second-order mental states to agents,
or ones that require of the agent almost a limitless stock of beliefs (Ruben
2003, chap. 4). One might dispute my argument that the causalist view
does imply that the agent would experience the kind of frustration that I
claim. But if the argument about this implication is sound, then introspec-
tion is the only way I know to test the claim.
What the causalist was trying to do was to give a causal model for the
case in which the agent Bs, because his reason to B has become delibera-
tively weightier than his reason to A, even though his reason to A has
retained its original nonrelational, causal strength. In this case, surely the
truth of the matter is that nothing needs to be thwarted and the agent
need feel no frustration. He gladly “surrenders” his reason to A, at least in
the circumstances, to his now-superior-because-weightier reason to B. It is
not true that his reason to B prevents or blocks him from acting on his
otherwise causally strong enough reason to A. In the case at hand, he
chooses not to do A, because he takes his reason to do A as relatively of
less importance or weight than his reason to do B, and in the case as we
have constructed it, I do not see how this fact can be modeled causally.
There is a perfectly clear deliberative story about what goes on in this case,
but it is a story for which the causalist can provide no convincing causal
counterpart.
There is, I submit, no fully convincing way causally to model decision
making that includes con-reasons, at least for cases of Type II. It is the
element of relational deliberative weight, comparative strength, which
cannot be captured causally, at least in those cases in which the con-reason
Causal and Deliberative Strength of Reasons 181

does not contribute causally to the action taken. What matters in delibera-
tion is the comparative or relative strength of reasons. If reasons were
causes, there would be nonrelational truths about the causal strength of
reasons. Because of these nonrelational causal truths, the two scenarios,
the causal/motivational and the rational/normative, won’t mesh. As long
as one thinks only about pro-reasons for action causing the actions they
favor, the point is not salient. But once con-reasons are introduced, it
becomes clearer that there is no plausible causal modeling for all the ways
in which con-reasons work in our deliberation scheme.

Acknowledgments

An earlier version of this essay appeared as “Con-Reasons as Causes” in

New Essays on the Explanation of Action, ed. C. Sandis (Basingstoke: Palgrave
Macmillan, 2009), 62–74. Richard Bradley, LSE, has helped in improving
this essay from what it otherwise would have been.

Notes

1. (1) is understood here to speak of causation, not necessarily only of deterministic

causation. The causation in question might be probabilistic or stochastic. I fre-
quently add “or probably” to cover cases of probabilistic or stochastic causation. My
discussion should apply equally whether the causation in question is deterministic
causation or stochastic causation.
What are causes? No essay can do everything and, with that, I intend to beg off
any responsibility for explicating the idea of causation. I am presupposing a fairly
standard account of causation, on which causes are token events or token states,
and that a causal chain is a series of such.

2. Although Jonathan Dancy (2000, 4) notes their existence: “but still I will nor-
mally speak as if all the reasons that do motivate all pull in the same direction.”

3. A con-reason is also a pro-reason in its own right for the action not taken, and
is a con-reason only in the sense that it counts against the action that was taken.
Similarly, a pro-reason is only a pro-reason for the action taken and is itself also a
con-reason for the action not taken. In what follows, to simplify terminology, I will
only use the idea of a pro-reason to be the reason that counts for the action one
takes, and the con-reason to be the reason that counts for the action one does not
take, the reason that gets outweighed. In the light of this, it would be wrong to
think of pro-reasons and con-reasons as two different sorts of reasons. I was careful
above only to say that reasons can function in these two different ways, depending
on context.
182 D.-H. Ruben

4. Instead of “relational” versus “intrinsic” (or, “nonrelational”), I would have

spoken of “ordinal” versus “cardinal,” but there seem to be presuppositions built
into the latter contrast that do not necessarily exist in the former contrast. Even
apart from the demands of causation that I discuss below, I doubt whether the whole
truth about reasons is exhausted by merely relational information, although nothing
in this essay requires that to be true. In a theoretical syllogism, a set of premises can
confer a nonrelational probability on a conclusion, making the conclusion rational
to believe to some degree; similarly, reasons for action can confer nonrelational
support on an action, making it a rational action to perform to some degree. Reasons
have an intrinsic strength as well as a comparative strength relative to other reasons
one has, if indeed one has any others.

5. I have often wondered why the principle “Every event has an effect” does not
have quite the same intuitive appeal as “Every event has a cause.” It might seem
obvious that they should stand or fall together.

6. Of course, the choice to A that he would have made or the A-ing he would have
performed had he not had a reason to B must differ from the choice to A that he
did actually make or the A-ing he actually did do in at least one way, simply in
virtue of the fact that it would have been a choice made in the absence of having
a conflicting reason to B. The qualification “in some intrinsic way” is meant to
exclude such trivial differences.

7. I do not think that one should underestimate the importance of the shift from
the personal to the subpersonal level, in order to maintain (1) and (2), broadened
to include con-reasons. It is a major concession on the part of the causalist. I do
not intend to develop the point here, but certainly the hope that lay behind the
causalist program for reasons for action was that reasons could be construed as
causes, yet doing so was compatible with understanding reasons and actions in their
own terms, sometimes called “the space of reasons.” This program was not neces-
sarily committed to construing reasons and actions as “really” about brain states
and gross behavior (even if they turn out to be identical to brain states and gross
behavior). The language of psychology and action was meant to have an internal
coherence and integrity all its own. To that extent, this option can easily take the
causalist program somewhere it had not intended to go.

8. Note that this example is not one of preemption, as some have suggested to me.
If it were a case of preemption, one would have two reasons both favoring the same
line of action, the first of which causes the action and the other of which did not
cause the action but would have caused the same action, had one not had the first
reason. In causal preemption, the inhibition or prevention is by the preempting
cause of some node on the chain that would have led from the preempted cause to
the effect. This is certainly not the case we are considering. But, arguably, all cases
of preemption involve some sort of causal inhibition or prevention, as does the case
we are considering.
13 Teleological Explanations of Actions: Anticausalism
versus Causalism

Alfred R. Mele

Teleological explanations of human actions are explanations in terms of

aims, goals, or purposes of human agents. According to one familiar causal
approach to analyzing human action and to explaining instances thereof,
human actions are, essentially, events that have appropriate mental items
(or neural realizations of those items) among their causes.1 Many causalists
appeal, in part, to such goal-representing states as desires and intentions
(or their neural realizers) in their explanations of human actions, and they
take acceptable teleological explanations of human actions to be causal
explanations. Some proponents of the view that human actions are
explained teleologically regard all causal accounts of action explanation as
rivals.2 I dubbed this position “anticausalist teleologism” (AT for short;
Mele 2003, 38) and argued against it (ibid., ch. 2).
I revisit AT in this essay. In section 1, after providing some background,
I rehearse an objection raised in Mele 2003 to a proposal George Wilson
(1989) makes in developing his version of AT. In section 2, I assess Scott
Sehon’s (2005, 167–171) recent reply to that objection. In section 3, I
articulate a version of Donald Davidson’s (1963/1980) challenge to anti-
causalists about action explanation and assess Sehon’s (2005, 156–160)
reply to Davidson’s challenge.

1 Wilson’s Proposal and My Objection

Are there informative, conceptually sufficient conditions for such things

as a human being’s acting in pursuit of a particular goal that do not invoke
causation? In chapter 2 of Mele 2003, I argued that attempts to identify
such conditions by leading anticausalists about action explanation—
Carl Ginet, Scott Sehon, R. Jay Wallace, and George Wilson—are unsuc-
cessful. In the present section, I review Wilson’s proposal and my objection
to it.
184 A. R. Mele

Regarding a particular case, Wilson claims that facts of the following

kind “about the context of the action, about the agent’s perception of that
context, and about the agent’s sentient relations to his own movements
. . . make it true that he intended of those movements that they promote
his getting back his hat [read: make it true that those movements were
sentiently directed by him at promoting his getting back his hat]”:3

[1] The man, wondering where his hat is, sees it on the roof, fetches the ladder, and
immediately begins his climb. [2] Moreover, the man is aware of performing these
movements up the ladder and knows, at least roughly, at each stage what he is about
to do next. [3] Also, in performing these movements, he is prepared to adjust or
modulate his behavior were it to appear to him that the location of his hat has
changed. [4] Again, at each stage of his activity, were the question to arise, the man
would judge that he was performing those movements as a means of retrieving his
hat. (Wilson 1989, 290)

Here Wilson apparently offers what he takes to be conceptually sufficient

conditions for its being true that the man sentiently directed certain move-
ments of his at promoting his getting back his hat. Are these conditions
in fact sufficient for this?4
Two points need to be made before answering this question. The first
concerns the second sentence of the passage just quoted. What knowledge
is attributed to the man there? Not the knowledge that he is about to
perform a sentiently directed movement of a certain kind, if sentient direc-
tion is a notion that this passage is supposed to explicate. However, this
does not block the supposition that the man knows, in some sense, that
he is about to perform a movement of his left hand onto the next rung,
for example, in Wilson’s broad sense of “perform a movement” (according
to which “A man performs a convulsive and spasmodic movement when
he clutches and cannot loose a live electric wire, and someone undergoing
an epileptic seizure may perform a series of wild and wholly uncontrollable
movements” [Wilson 1989, 49]).
Second, Wilson claims that “to try to [A] is (roughly and for the perti-
nent range of cases) to perform an action that is intended to [A]” (ibid.,
270). Brief attention to trying will prove useful. Suppressing Wilson’s
qualification for a moment, his claim amounts, for him, to the assertion
that to try to A is to perform an action that is sentiently directed by the
agent at A-ing. Regarding the following plausible proposition, there is no
need to worry about limiting the range of cases: (T1) if one is doing some-
thing that one is sentiently directing at A-ing, then one is trying to A, in
an utterly familiar, unexacting sense of “trying.” Although people may
often reserve attributions of trying for instances in which an agent makes
Teleological Explanations of Actions 185

a considerable or special effort, this is a matter of conversational implica-

ture and does not mark a conceptual truth about trying.5 A blindfolded,
anesthetized man who reports that he has raised his arm as requested quite
properly responds to the information that his arm is strapped to his side
by observing that, in any case, he tried to raise it—even though, encounter-
ing no felt resistance, he made no special effort to raise it.6 And when, just
now, I typed the word “now,” I tried to do that even though I easily typed
the word. Now, T1 entails (T2) that one who is not trying to A, even in
the unexacting sense of “trying” identified, is not sentiently directing one’s
bodily motions at A-ing. T2 is very plausible, as is the following, related
proposition: (T3) one who is not trying to do anything at all, even in
the unexacting sense of “trying,” is not sentiently directing one’s bodily
motions at anything.7
In the following case, as I will explain, although Norm is not, during a
certain time, trying to do anything, even in the unexacting sense of
“trying” that I identified, he satisfies all four of the conditions in the quo-
tation at issue during that time. The moral is that these conditions are not
sufficient for a person’s sentiently directing “movements” of his at the
time. Their insufficiency follows from the details of the case and the plati-
tude that an agent who is not trying (even in the unexacting sense) to do
anything is not sentiently directing his bodily motions at anything.
Norm has learned that, on rare occasions, after he embarks on a routine
activity (e.g., tying his shoes, climbing a ladder), Martians take control of
his body and initiate and sustain the next several movements in the chain
while making it seem to him that he is acting normally. He is unsure how
they do this, but he has excellent reason to believe that they are even more
skilled at this than he is at moving his own body, as, in fact, they are. (The
Martians have given Norm numerous demonstrations with other people.)
The Martians have made a thorough study of Norm’s patterns of peripheral
bodily motion when he engages in various routine activities. Their aim
was to make it seem to him that he is acting while preventing him from
even trying to act by selectively shutting down portions of his brain. To
move his body, they zap him in the belly with M-rays that control the
relevant muscles and joints. When they intervene, they wait for Norm to
begin a routine activity, read his mind to make sure that he plans to do
what they think he is doing (e.g., tie his shoes or climb to the top of a
ladder), and then zap him for a while—unless the mind-reading team sees
him abandon or modify his plan. When the team notices something of
this sort, the Martians stop interfering and control immediately reverts to
Norm.
186 A. R. Mele

A while ago, Norm started climbing a ladder to fetch his hat. After he
climbed a few rungs, the Martians took over. Although they controlled
Norm’s next several movements while preventing him from trying to do
anything, they would have relinquished control to him if his plan had
changed (e.g., in light of a belief that the location of his hat had changed).
Return to facts 1 through 4. Fact 1 obtains in this case. What about fact
2? It is no less true that Norm performs his next several movements than
that the man who clutches the live electric wire performs convulsive move-
ments. And the awareness of performing movements mentioned in fact 2
is no problem. The wire clutcher can be aware of bodily “performances”
of his that are caused by the electrical current, and Norm can be aware of
bodily “performances” of his that are caused by M-rays. Norm also satisfies
a “knowledge” condition of the sort I identified. If Wilson is right in think-
ing that an ordinary ladder climber knows, in some sense, that he is about
to perform a movement of his left hand onto the next rung, Norm can
know this too. What he does not know is whether he will perform the
movement on his own or in the alternative way. But that gives him no
weaker grounds for knowledge than the ordinary agent has, given that the
subject matter is the performance of movements in Wilson’s broad sense
and given what Norm knows about the Martians’ expertise. Fact 3 also
obtains. Norm is prepared to adjust or modulate his behavior, and one
may even suppose that he is able to do so. Although the Martians in fact
initiated and controlled Norm’s next several movements up the ladder
while preventing him from trying to do anything, they would not have
done so if his plans had changed. Fact 4 obtains too. In Wilson’s sense of
“perform a movement,” Norm believes that he is performing his move-
ments “as a means of retrieving his hat.” (He does not believe that the
Martians are controlling his behavior; after all, he realizes that they very
rarely do so.)
Even though these facts obtain, Norm does not sentiently direct his next
several movements up the ladder at getting his hat because he is not sen-
tiently directing these movements at all. Wilson maintains that sentiently
directing a bodily movement that one performs entails exercising one’s
“mechanisms of . . . bodily control” in performing that movement (Wilson
1989, 146). However, Norm did not exercise these mechanisms in his
performance of the movements at issue. Indeed, he did not make even a
minimal effort to perform these movements; owing to the Martian inter-
vention, he made no effort at all—that is, did not try—to do anything at
the time. And it is a platitude that one who did not try to do anything at
all during a time t did not sentiently direct his bodily motions during t.
Teleological Explanations of Actions 187

It might be suggested that although Norm did not directly move his
body during the time at issue, he sentiently directed his bodily motions
in something like the way his sister Norma sentiently directed motions of
her body when she vocally guided blindfolded colleagues who were carry-
ing her across an obstacle-filled room as part of a race staged by her law
firm to promote teamwork. If Norma succeeded, she may be said to have
brought it about that she got across the room, and her bringing this about
is an action.8 Notice, however, that there is something that she was trying
to do at the time. For example, she was trying to guide her teammates. By
hypothesis, there is nothing that Norm was trying to do at the relevant
time, for the Martians blocked brain activity required for trying. And this
is a crucial difference between the two cases. The claim that Norma sen-
tiently directed motions of her body at some goal at the time is consistent
with T3; the comparable claim about Norm is not.9
Wilson proposed sufficient conditions for its being true that a person’s
movements were sentiently directed by him at promoting his getting back
his hat. Norm satisfies those conditions even though it is false that the
“movements” at issue were sentiently directed by him. So those conditions
are not in fact sufficient.
Can Wilson’s proposal be rescued simply by augmenting it with an
anti-intervention condition? No. If the addition of such a condition does
contribute to conceptually sufficient conditions for a person’s sentiently
directing his movements at a goal, it may do so because the excluded
kinds of intervention prevent, for example, the obtaining of normal causal
connections between mental items or their neural realizers and bodily
motions. An anticausalist who augments Wilson’s proposal with an anti-
intervention condition also needs to produce an argument that the condi-
tion does not do its work in this way.

2 Sehon on Norm

In a book defending AT, Scott Sehon (2005, 167–171) replies to my

objection to Wilson’s proposal. The present section is an assessment of his
reply.
Sehon contends that it “is not obvious that Norm’s behavior fails to be
an action” (ibid., 168). In support of this claim, he sketches a version of
my story in which Norm is about to shoot someone when “the Martians
take over his body and make it carry out the dirty deed.” Sehon asks, “Does
this completely absolve Norm of responsibility for shooting his professor?”
He reports that it is not obvious to him how to answer this question.
188 A. R. Mele

“Accordingly,” he writes, “even in the routine case of going up the ladder,

I take it not to be obvious that Norm failed to act.”
Things are not nearly as murky as Sehon thinks. I start with an obvious
point. Norm is responsible for shooting his professor only if Norm shoots
his professor. This is an instance of the truism that a person P is responsible
for doing A only if P did A. But notice that P may have some responsibility
for the shooting of X even if, because P did not shoot X, P has no respon-
sibility for shooting X. For example, if P hires—or forces—someone to shoot
X, then, other things being equal, P has some responsibility for the shoot-
ing of X. In Sehon’s story, the Martians “make it seem to Norm as if he is
acting, and if Norm had changed his mind and decided to put the gun
down [they] would have immediately relinquished control” to Norm
(ibid.). (Here Sehon is following my lead, of course.) So readers may be
strongly inclined to see Norm as responsible for more than, in Sehon’s
words, merely “having a plan to commit murder.” In light of the distinc-
tion I have drawn between P’s having some responsibility for shooting X
and P’s having some responsibility for the shooting of X, those who, like
me, are convinced that Norm did not shoot the professor can maintain
that, even so, he has some responsibility for the shooting of the
professor.
Imagine that Sehon’s story remains basically the same, except that a
Martian makes himself invisible and then, with his slim but powerful
tentacle, pulls Norm’s paralyzed finger down on the trigger. (The Martian
also makes it seem to Norm as though Norm is pulling the trigger, and “if
Norm had changed his mind and decided to put the gun down, the
[Martian] would have immediately relinquished control” to Norm.) Obvi-
ously, Norm did not pull the trigger. So Norm did not shoot the professor.
That entails that Norm is not responsible for shooting the professor. Even
so, he may have some responsibility for the shooting of the professor. After
all, the Martian would not have pulled the trigger with Norm’s finger if
Norm had not been bent upon shooting the professor, and the professor’s
life would have been spared if Norm had changed his mind. Sehon’s uncer-
tainty about whether Norm shot the professor seems to derive from his
failure to distinguish having some responsibility for shooting the professor
from having some responsibility for the shooting of the professor.
Some readers may wonder why my story about Norm in Mele 2003
portrays the Martians as using M-rays rather than moving Norm with their
tentacles. Recall that Wilson’s conditions include the man’s “performing
movements”; and although he is happy to say that a wire clutcher whose
body is being jerked about by the wire’s electrical discharge is performing
Teleological Explanations of Actions 189

movements in his thin sense, I doubt that he would count a man whose
limb motions are caused by Martians pulling and pushing on his limbs as
performing movements with those limbs. Naturally, I thought the M-rays
were just fine for my purposes, but instead I could have portrayed the
Martians as moving Norm’s paralyzed body with just the right sorts of
electrical jolts to muscles and joints. Call this E-manipulation.
Sehon wonders why my Martians interfere with Norm. “What’s in it for
the Martians?” he asks (Sehon 2005, 168). In Mele 2003, I neglected to
mention that the Martians had read page 290 of Wilson’s (1989) book and
wanted to provide a living counterexample to his proposal. In any case,
Sehon’s reflection on his question leads him to the following claim:

Mele stipulates that the Martians are going to make Norm’s body do exactly what
Norm planned to do anyway. If this were an ironclad promise from the Martians,
or better yet, something that followed necessarily from their good nature, then
. . . I have little problem saying that Norm is still acting, despite the fact that the
causal chain involved is an unusual one. If he commits a murder under these cir-
cumstances, we will definitely not let him off. (Sehon 2005, 169)

Yes, if Norm commits a murder, he should be blamed for that. But if

he is not acting, he commits no murder. Is Norm acting in Sehon’s sce-
nario? Presumably, for the purposes of his thought experiment, Sehon
means to retain as much as he can from my story about Norm while
turning the Martians into beings whose good nature entails that they
always make “Norm’s body do exactly what Norm” plans to do. Evidently,
Sehon is not impressed by the following details of my story: the Martians
prevent Norm “from even trying to act by selectively shutting down por-
tions of his brain,” and they move his body by zapping “him in the belly
with M-rays that control the relevant muscles and joints” (Mele 2003, 49).
I am not sure why. Possibly, he rejects T3 (the thesis that one who is not
trying to do anything at all, even in the unexacting sense of “trying,” is
not sentiently directing one’s bodily motions at anything). And possibly,
he accepts T3 and believes that Norm counts as trying to do things when
his Martians replace mine.
Consider a scenario in which, instead of using M-rays, Sehon’s good
Martians paralyze Norm’s body and then move it by E-manipulation while
making it seem to Norm that he is acting normally. When, for example,
Norm intends to climb a ladder to get his hat, the Martians paralyze him
and E-manipulate his body up the ladder. (They do all this while making
it seem to Norm that he is acting normally, and if Norm were to change
his mind his paralysis would immediately cease and control would revert
190 A. R. Mele

to him.) I do not see how the claim that, in this scenario, Norm is climbing
the ladder can be regarded as anything but utterly preposterous. Yet, unless
Sehon can identify a crucial difference between the use of M-rays and this
alternative mode of Martian body manipulation, he is committed to having
“little problem saying” that Norm is climbing it.
Sehon is willing to grant that when my Martians are at work rather than
his, Norm is not acting (Sehon 2005, 168). He contends that, in my story,
“since Norm fails . . . to satisfy” the following condition, “his behavior
does not count as goal directed” on his “account of the epistemology of
teleology” (ibid., 169): (R1) “Agents act in ways that are appropriate for
achieving their goals, given the agent’s circumstances, epistemic situation,
and intentional states” (ibid., 155). If I am right, Norm is not acting at all,
in which case invoking R1 is overkill. And if Norm is not acting, as Sehon
is willing to grant, then Wilson’s proposal about sufficient conditions for
its being true that a person’s movements were sentiently directed by him
at promoting his getting back his hat is false, which is what I set out to
show with the Martian example in Mele 2003.
Some readers may feel that they have lost the plot. The following obser-
vation will help. One thing that Sehon would like to show is that a pro-
ponent of AT can “accommodate our intuition that Norm is not acting”
in my story (Sehon 2005, 170). He argues that “Norm’s motion is not that
of an agent, because in a range of nearby counterfactual situations his
behavior is not appropriate to his goals. Specifically, in all those situations
in which the Martians simply change their mind about what they want to
have Norm’s body do, Norm’s body will do something quite different”
(ibid.).
Sehon’s explanation of why Norm is not acting is seriously problematic.
Imagine a case in which the Martians consider interfering with Norm but
decide against doing that. Norm walks to the kitchen for a beer without
any interference from the Martians. There are indefinitely many variants
of this case in which the Martians change their minds about not interfering
and make Norm’s body do something else entirely. So “in a range of nearby
counterfactual situations his behavior is not appropriate to his goals”
(ibid.). But this certainly does not warrant the judgment that Norm is not
acting in the actual scenario. Obviously, he is acting in that scenario: he
is walking to the kitchen for a beer. If Sehon is thinking that his counter-
factual test for whether an agent is acting is to be applied in scenarios in
which the Martians interfere with Norm but not in scenarios in which they
do not interfere with him, he does not say why this should be so.
Teleological Explanations of Actions 191

The problems with Sehon’s explanation of why it is that Norm is not

acting in my case do not end here. He considers a woman, Sally, who “has
an odd neurological disorder” (ibid.). When she tries to move her finger
in way W, her finger often becomes paralyzed and “her body goes through
any number of other random motions.” In a particular case, Sally success-
fully tries to move her finger in way W when pulling a trigger and murder-
ing a professor. Sehon contends that because Sally’s “behavior is generally
very sensitive to her goals”—after all, it is “subject to these flukes only
when it involves a finger pulling”—she, “unlike Norm, satisfies the condi-
tion imposed by (R1) well enough to make her an agent at the time in
question” (ibid., 171).
This will not do. Imagine a variant of Norm’s story in which a rogue
Martian interferes with Norm only on one occasion. (The Martian is
imprisoned for life by the Martian authorities immediately afterward and
no one else ever interferes with Norm.) He moves Norm’s paralyzed body
up the ladder by E-manipulation while making it seem to Norm that he is
acting normally. Sally’s behavior is “generally very sensitive to her goals,”
and I stipulate that Norm’s behavior is generally even more sensitive to
his goals. Even so, he is not acting as his body moves up the ladder. That
Sally is acting whereas Norm is not is not explained by a difference in the
general sensitivity of their behavior to their goals.10
Sehon concludes his discussion of my objection to Wilson’s proposal
with the following report:

One could alter the example by making Sally’s neurological disorder much more
general, such that she rarely does what she intends; but with that revision, my own
intuitions about the case grow flimsy. I’m not sure what to say about her agency in
such a case, and I’m not too troubled by the conclusion that she is not exhibiting
genuine goal-directed behavior at any particular moment. (Ibid.)

Imagine that you read in a medical journal a study of a sixty-year-old

patient, Pat, who for the past ten years has suffered from a horrible illness
that has the consequence that in about 99 percent of the cases in which
he tries to do something, his “body goes through any number of other
random motions” and his attempt fails. About 1 percent of the time, he
succeeds in doing what he tries to do. Sometimes, for example, he makes
a successful attempt to signal his nurse by pressing a button. Apparently,
Sehon would not be “too troubled by the conclusion” that Pat never exhib-
its “genuine goal-directed behavior.” But if unwarranted conclusions are
troublesome, Sehon should be troubled by this. Pat’s story clearly is con-
ceptually possible. In it, some of his behavior is genuinely goal directed.
192 A. R. Mele

Two conclusions may now be drawn. First, Sehon has not shown that
my objection to Wilson’s proposal is unsuccessful. In fact, insofar as he
concedes that Norm is not acting in my story, he apparently concedes that
the objection is successful. (He nowhere claims that Norm does not satisfy
Wilson’s proposed conditions.) It is perhaps worth mentioning in this
connection that Sehon might have misled not only his readers but also
himself by treating my objection to Wilson’s proposal as though it were
an argument for the claim that no version of AT can “accommodate our
intuition that Norm is not acting” in my story (ibid., 170). Other anticau-
salists about action explanation—for example, Carl Ginet (1990) and R.
Jay Wallace (1999)—have proposed other sufficient conditions for a human
being’s performing an action or acting in pursuit of a particular goal, and
my objections to Ginet’s and Wallace’s proposals in Mele 2003 were very
different from my objection to Wilson’s proposal.11 Naturally, the objec-
tions I offered were designed to apply to the details of the specific propos-
als. Second, Sehon’s attempt to produce a version of AT that distinguishes
cases of action from cases like Norm’s is unsuccessful.

3 Causalism versus AT: Davidson’s Challenge Revisited

What can causalists about action-explanation and proponents of AT rea-

sonably challenge one another to do? Distinguishing the project of produc-
ing a conceptually sufficient condition for a human being’s acting in pursuit
of a particular goal from the project of producing an analysis of this is
useful in answering this question. The latter project obviously is more
challenging than the former: producing conditions that are individually
conceptually necessary and jointly conceptually sufficient for X is more
challenging than producing a conceptually sufficient condition for X. And
if it were demonstrated, for example, that having a mental state or event
of a certain kind among its causes is a necessary condition for an event’s
being an action performed in pursuit of a goal G and that no attempted
explanation of an action that did not appeal either explicitly or implicitly
to such a cause is an adequate explanation, that demonstration alone
would show that AT is false. A causalist does not need to generate an
acceptable analysis of an action’s being performed in pursuit of a goal G
in order to show that AT is false.12
Similarly, a proponent of AT does not need to generate an acceptable
analysis of an action’s being performed in pursuit of G in order to show
that causalism about acting in pursuit of a goal is false. I have never chal-
lenged proponents of AT to produce an analysis of this. (To ask for an
Teleological Explanations of Actions 193

analysis of any contested philosophical concept that will be widely accepted

is to ask for a great deal.) My challenge to proponents of AT concerns a
much more modest task: producing an informative conceptually sufficient
noncausal condition for an action’s being performed in pursuit of a goal
G. Teleologists contend that “teleological explanations explain by specify-
ing an action’s goal or purpose; for example, when we say that Jackie went
to the kitchen in order to get a glass of wine, we thereby specify the state
of affairs at which her action was directed” (Sehon 1997, 195). But in virtue
of what is it true that a person acted in pursuit of a particular goal? Wilson
and Sehon have failed to answer this question successfully, as I have argued
here. (So have Ginet and Wallace, as I argued in Mele 2003, ch. 2.)
Can causalists do any better? Elsewhere, I have argued that they can. In
chapter 2 of Mele 2003, I offered, from a particular causalist perspective,
what I argued to be an informative, conceptually sufficient condition for
a being’s acting in pursuit of a particular goal. And I observed that if the
condition I offered is indeed conceptually sufficient for this, causalists are
in much better shape in this connection than proponents of AT, none of
whom have succeeded in offering noncausal conceptually sufficient condi-
tions for this (ibid., 59). Interested readers are encouraged to consult
chapter 2 of Mele 2003. The present section is concerned with some related
business.
If I had ever thought I knew of a convincing direct argument for the
thesis that an agent’s acting in pursuit of a particular goal requires the
satisfaction of a specific causal condition, I would have gone public with
it. My way of defending causalism has been indirect. I have challenged
proponents of AT to produce informative, conceptually sufficient non-
causal conditions for acting in pursuit of a particular goal, argued against
their proposals, and developed a causalist view of what actions are, how
they are explained, and how they are produced (Mele 1992, 2003). If
readers who compare my causalist view with anticausalist alternatives in
light of the arguments I offered for the former and against the latter are
justified in judging that my view is considerably closer to the truth, I am
satisfied.
It is possible that some opponents of some causal views of action or
action explanation oppose these views partly because they misunderstand
them. For example, significant differences between how some causalists
and some of their opponents understand causation may help to explain
some of the opposition. In the literature defending or attacking causal
theories of action or action-explanation, not much is said about what
causation is. This is understandable: both sides may be thinking that,
194 A. R. Mele

whatever the best account of causation is, their view about what actions
are or how they are to be explained is the correct one to take, and they
may want to avoid hitching their wagon to a specific theory of causation.
The same goes for causal explanation.
David Lewis defends the thesis that “to explain an event is to provide
some information about its causal history” (Lewis 1986, 217). Presumably,
any anticausalist about action explanation who takes even some actions
to be events would reject this thesis. But suppose it were agreed on all sides
that a sufficient condition for an explanation of an event being a causal
explanation is that the explanation explains the event at least partly by
providing some information about its causal history. With this agreement
in place, one can ask, for example, whether every acceptable teleological
explanation of an action is a causal explanation of the action. Are there
acceptable teleological explanations of actions that do not explain the
actions even partly by providing some information about their causal
history? (Teleological explanations of actions, again, are explanations in
terms of aims, goals, or purposes of the agents of those actions.)
Sehon asserts that “Teleological explanations simply do not purport to
be identifying the cause of a behavior” (Sehon 2005, 218). But, as Lewis
observes, speaking in terms of “the cause of something” can easily generate
confusion (Lewis 1986, 215). Lewis adds: “If someone says that the bald
tire was the cause of the crash, another says that the driver’s drunkenness
was the cause, and still another says that the cause was the bad upbringing
which made him so reckless, I do not think any of them disagree with me
when I say that the causal history includes all three” (ibid.). In any case,
causalists like me do not purport to be identifying the cause of an action
when we offer causal explanations of actions in terms of agents’ aims,
goals, or purposes. The basic idea—oversimplifying a bit—is that a putative
teleological explanation of an action in terms of a goal, aim, or purpose G
does not explain the action unless the agent’s wanting or intending to (try
to) achieve G has a relevant effect on what he does. Obviously, the notion
of having an effect is a causal notion; and the assertion, for example, that
an agent’s intending to achieve G had an effect on what he did places his
intending to do that in the causal history of what he did.
At this point, some causalists part company with others. Some causalists
posit token mental states such as intentions, desires, and beliefs and attri-
bute causal roles to these states or to their neural realizers in the production
of actions.13 Other causalists are wary of postulating such token states of
mind.14 I take no stand on this issue here.
Teleological Explanations of Actions 195

Over four decades ago, Donald Davidson posed a much-discussed chal-

lenge to philosophers who oppose causalism about action-explanation
(Davidson 1963/1980, 9, 11). I set the stage for a version of his challenge
with a story. Al has a pair of reasons for mowing his lawn this morning.
One reason has to do with schedule-related convenience and the other
with vengeance. Al wants to mow his lawn this week and he believes that
this morning is a convenient time, given his schedule for the week. But
he also wants to repay his neighbor for the rude awakening he suffered
recently when she turned on her mower at the crack of dawn, and he
believes that mowing his lawn this morning would constitute suitable
repayment. As it happens, Al mows his lawn this morning only for one of
the two reasons mentioned.
Suppose that Al’s wanting to mow his lawn this week and his believing
what he does about convenience have no effect at all on what Al does with
his lawnmower this morning. Might it be the case, even so, that he mows
his lawn for the reason having to do with convenience rather than for the
reason having to do with vengeance? A philosopher who answers in the
affirmative faces the challenge of producing an informative, conceptually
sufficient condition for “Al mows his lawn this morning for the reason
having to do with convenience” that is consistent with this answer and
the details of the story.15
Sehon seemingly intends to tackle Davidson’s challenge head on. He
writes: “Briefly put, teleological explanations support certain counterfac-
tual conditionals, and this will allow us to distinguish the reason for which
an agent acted from other nonmotivating reasons” (Sehon 2005, 157). He
contends that “when we are determining which of” two alternative teleo-
logical explanations “is true, we can look at a variety of counterfactual
situations” (ibid., 158). “Basically speaking, we look at the agent’s behavior
in the counterfactual situations and determine the goal or goals for which
her behavior would have been appropriate. . . . The general point,” Sehon
reports, “is that we are looking at counterfactual situations to see what
account of the agent’s behavior makes the most rational sense. Thus, the
sort of case that Davidson proposes is not enough to undermine the teleo-
logical alternative to causalism” (ibid.).
A fatal flaw in Sehon’s reply to the challenge is easily identified. Suppose
you know Al pretty well and you know that he mowed his lawn this
morning. Al’s friend Ann tells you that he had the two reasons for doing
this that I mentioned, and she voices her confidence that he did it for only
one of these reasons. She promises to give you ten dollars if you figure out
196 A. R. Mele

for which of the two reasons he mowed his lawn this morning and tell her
how you figured it out. You decide to follow Sehon’s lead and to consider
various counterfactual scenarios. You know that Al dislikes mowing his
lawn in even a light rain, and you start by asking yourself what he would
have done this morning if there had been a light rain. You think that if
he would have mowed his lawn anyway, “that is good evidence that in
the actual circumstances [he] was directing [his] behavior” (Sehon 2005,
158) at getting revenge, because the rain, for Al, would outweigh schedule-
related convenience. “Would he have mowed it anyway?” you ask yourself.
And you find that you are stumped. You realize that if you had substantial
grounds for believing that Al mowed his lawn to get revenge, you could
use those grounds to support the claim that he would have mowed it even
in a light rain; and you realize that if you had substantial grounds for
believing that Al mowed his lawn only for reasons of convenience, you
could use them to support the claim that he would not have mowed it if
it had been raining. It dawns on you that the strategy of trying to identify
the reason for which Al actually acted by trying to figure out what he
would have done in the counterfactual scenario I mentioned and other
such scenarios puts the cart before the horse. Asking your counterfactual
question about the rain scenario is nothing more than a heuristic device—
and not a very useful one. The truth about what Al would have done in a
light rain is grounded partly in the truth about the reason for which he
actually acted.
As I pointed out in Mele 2003, 51, in response to an earlier proposal by
Sehon that featured counterfactuals, the truth of true counterfactuals is
grounded in facts about the actual world; and if, for example, relevant
counterfactuals about Al are true for the reasons one expects them to be,
their truth is grounded partly in Al’s acting for the reason for which he
acted. As far as Davidson’s challenge is concerned, we are back to square
one. Certain counterfactuals about Al are true partly because he acted for
a certain reason. But in virtue of what is it true that he acted for that
reason? Sehon’s proposal about counterfactuals leaves this question
unanswered.
For a critical examination of several leading anticausalist treatments of
action-explanation, see chapter 2 of Mele 2003. In the present essay I have
focused on some elements of Sehon’s recent defense of AT. If I am right,
Sehon has not undermined my objection to Wilson’s proposal, has not
produced a version of AT that distinguishes cases of action from cases like
Norm’s, and has not offered an adequate reply to Davidson’s challenge.
This is bad news for at least one version of AT.
Teleological Explanations of Actions 197

Acknowledgments

This essay was completed during my tenure of a 2007–2008 NEH Fellow-

ship. (Any views, findings, conclusions, or recommendations expressed in
this essay do not necessarily reflect those of the National Endowment for
the Humanities.) I am grateful to Andrei Buckareff and Randy Clarke for
comments on a draft of this essay.

Notes

1. See Bishop 1989; Brand 1984; Davidson 1980; Goldman 1970; Mele 1992, 2003;
Thalberg 1977; and Thomson 1977.

2. See, e.g., Sehon 1994, 1997, 2005; Taylor 1966; and Wilson 1989, 1997.

3. For a justification of the instruction in brackets, see Mele 2003, 47.

4. This paragraph and the next eight are borrowed, with some minor modifications,
from Mele 2003, 48–50.

5. See Adams and Mele 1992, 325; Armstrong 1980, 71; McCann 1975, 425–427;
and McGinn 1982, 86–87.

6. See James 1981, 1101–1103. For discussion of a case of this kind, see Adams and
Mele 1992, 324–331.

7. How readers interpret my action-variable “A” should depend on their preferred

theory of action individuation. Davidson writes: “I flip the switch, turn on the light,
and illuminate the room. Unbeknownst to me I also alert a prowler to the fact that
I am home” (1963/1980, 4). How many actions does the agent, Don, perform?
Davidson’s coarse-grained answer is one action “of which four descriptions have been
given” (ibid.). A fine-grained alternative treats A and B as different actions if, in
performing them, the agent exemplifies different act-properties (Goldman 1970).
On this view, Don performs at least four actions, because the act-properties at issue
are distinct. An agent may exemplify any of these act-properties without exemplify-
ing any of the others. (One may even turn on a light in a room without illuminating
the room: the light may be painted black.) Componential views represent Don’s
illuminating the room as an action having various components, including his
moving his arm (an action), his flipping the switch (an action), and the light’s going
on (Ginet 1990). Where proponents of the coarse-grained and fine-grained theories
find, respectively, a single action under different descriptions and a collection of
intimately related actions, advocates of the various componential views locate a
“larger” action having “smaller” actions among its parts. Readers should understand
the variable “A” as a variable for actions themselves (construed componentially or
otherwise) or actions under descriptions, depending on their preferred theory of
198 A. R. Mele

action-individuation. The same goes for the expressions that take the place of “A”
in concrete examples.

8. Readers who regard the claim that Norma brought it about that she got across
the room as an exaggeration may be happy to grant that she helped to bring that
about. Her helping to do that is an action.

9. On a case that may seem to be problematic for T3, see Mele 2003, 64–65, n. 22.

10. In my original story, the Martians interfere with Norm only “on rare occasions”
(Mele 2003, 49). So, seemingly, they may interfere with him much less often than
Sally has her finger problem. The variant of Norm’s case just sketched renders specu-
lation about this comparative issue otiose.

11. For my discussion of Ginet 1990 and Wallace 1999, see Mele 2003, 39–45.

12. Incidentally, I have never offered an analysis of action nor of acting in pursuit
of a particular goal, although I have defended causalism in both connections (Mele
1992, 2003). Paul Moser and I (Mele and Moser 1994) have offered an analysis of
what it is for an action to be an intentional action.

13. See, e.g., Brand 1984; Davidson 1980, ch. 1; and Mele 1992, 2003.

14. See, e.g., Child 1994; Hornsby 1993.

15. There is considerable disagreement about what reasons for action are, and I have
spun my story in a way that is neutral on this issue. For example, the story does
not identify the reasons mentioned in it with belief-desire pairs, as Davidson (1980,
ch. 1) does; nor does it deny that this identity holds.
14 Teleology and Causal Understanding in Children’s
Theory of Mind

Josef Perner and Johannes Roessler

In “The Emergence of Thought,” Donald Davidson argues that while we

have no difficulty in describing, on the one hand, mindless nature and,
on the other, mature adult psychology, “what we lack is a way of describ-
ing what is in between.” He claims that there is a deep and “perhaps
insuperable problem in giving a full description of the emergence of
thought”; and he expresses relief at not working “in the field of develop-
mental psychology” (Davidson 2001, 128). In this chapter we argue that
Davidson was right about the depth and difficulty of the problems involved
in describing the emergence of thought. But we think Davidson was unduly
pessimistic about the prospect of making progress, empirical and philo-
sophical, with these problems. Indeed we hope to show that describing the
emergence of thought may help to shed light on the nature of thought.
We should make it clear immediately that our concern here will not be
with Davidson’s completely general problem, the “conceptual difficulty”
he sees with the very idea of attributing beliefs and other propositional
attitudes to immature thinkers. For the purposes of this chapter, we simply
assume that such attributions can be literally correct. We will be concerned
with a much more specific issue: the question of how young children
understand intentional action. The problem is akin to Davidson’s, though,
in that it involves a “conceptual difficulty.” According to current philo-
sophical orthodoxy, to understand what it is to act intentionally one has
to be able to find actions intelligible in terms of the agent’s reason for
acting, the agent’s purpose in doing what he or she does. Call this the
Reason claim, or R. Current orthodoxy also holds that reasons for action
are provided, or constituted, by suitable pairs of beliefs and desires. Call
this the Belief-Desire claim, or BD. (The classical source for both R and BD
is Davidson’s [1963] paper “Actions, Reasons, and Causes.”) Put together,
R and BD suggest that to see people as acting intentionally you have to
think of them as acting on the basis of what they believe: understanding
200 J. Perner and J. Roessler

intentional action requires a grasp of the explanatory role of beliefs. There

is a large body of evidence to suggest that 2- and 3-year-olds do not satisfy
this condition. On the other hand, there is also convincing evidence that
2- and 3-year-olds do have some grasp of what it is to act intentionally.
Hence the “conceptual difficulty”: the hypothesis that children have some
understanding of intentional action enjoys empirical support, yet it is hard
to see how it can be sustained, given that children apparently lack a basic
conceptual prerequisite for such understanding.
In what follows we develop and defend the following diagnosis. What
is to blame for our “conceptual difficulty” is a dogma of contemporary
philosophy of mind, that reason-giving explanations of actions are expla-
nations in terms of beliefs and desires. The suggestion we will pursue is
that young children find actions intelligible in terms of reasons that are
not conceived as mental states at all—but that can nevertheless intelligibly
be taken to provide causal explanations. We’ll call this the teleological
account of children’s conception of intentional action. We argue that the
teleological account is of more than developmental interest. Its genealogy
of commonsense psychology has an important bearing on the nature of
the adult conception of intentional action. For, as Wittgenstein remarked:
“the thing about progress is, it tends to look more momentous than it
really is.”1

1 The Puzzle

What evidence is there to suggest that young children have some under-
standing of intentional action? Some psychologists maintain that even
toward the end of the first year, as infants begin to engage in joint atten-
tion interactions with others, they perceive and understand others’ actions
as goal-directed (Tomasello 1999). Others have argued that such under-
standing manifests itself in 18-month-olds’ more sophisticated capacities
for imitation of intended actions (Meltzoff 1995). Here we will focus on
evidence provided by (slightly older) children’s performance on classical
false-belief tasks. The question put to children in such tasks is what the
protagonist in some story will do next. For example: Suppose Maxi’s mother
transferred the chocolate Maxi put into the green kitchen cupboard to the
blue cupboard while he is out playing. Maxi, feeling peckish, returns to
the house. Where will he look for the chocolate? Three-year-olds’ performance
on this task is poor, but far from random. They reliably predict that Maxi
will look in the blue cupboard. They don’t suggest that he will look under
the kitchen table or in the playground or in the loft. What explains this
Teleology in Children’s Theory of Mind 201

general pattern in 3-year-olds’ performance? The natural, and standard,

interpretation is that they tend to predict what Maxi will do on the basis
of what it makes sense for Maxi to do, given the actual location of the choco-
late. If you want to find the chocolate, looking in the blue cupboard is
precisely what you should be doing—what you have a good reason to do.
Thus, on the natural interpretation, children do think of Maxi as an inten-
tional agent. They assume that Maxi acts in a purposive manner; that he
will do what, given his purpose, he has good reason to do.
The puzzle, then, is this. Young children’s performance on false-belief
tasks simultaneously provides evidence that they think of others as inten-
tional agents and that they have a systematic deficit in understanding the
explanatory role of beliefs.2 Yet, if recent philosophical orthodoxy is right,
failure to grasp the explanatory role of beliefs should make it impossible
to understand what it is to act intentionally. For (to repeat) recent ortho-
doxy holds that

R: Understanding what it is to act intentionally requires finding actions

intelligible in terms of agents’ reasons for acting

and

BD: Reasons for action are provided, or constituted, by suitable pairs of

beliefs and desires.

2 Desire Psychology

There is, in the developmental literature, an influential view that may seem
to offer a simple solution to our puzzle. The basic idea is that while think-
ing of someone as acting intentionally certainly requires understanding
something about the mental states causally responsible for intentional
actions, such understanding may be more or less comprehensive. A fully
developed conception of intentional action will require a large and complex
set of psychological notions, including, of course, the notion of belief. But,
the claim is, children may have a rudimentary grasp of intentional action
in virtue of understanding something about the explanatory role of desires,
without yet appreciating how desires tend to interact with other states,
such as beliefs (Bartsch and Wellman 1995).
Straight off, though, it is not obvious how this suggestion speaks to our
puzzle. To understand Maxi’s behavior as intentional, you have to put two
things together: the purpose of the action and the means by which Maxi
seeks to accomplish his purpose. A simple “desire psychology” may enable
you to identify Maxi’s purpose (his purpose is to get hold of his chocolate),
202 J. Perner and J. Roessler

but it is silent on what he might be expected to do to accomplish his

purpose. So, on the face of it, the proposal has nothing to say on how
children understand the intentional, purposive nature of the instrumental
action they expect Maxi to perform (his opening the blue cupboard). Of
course, you might say that the desire to locate his chocolate is at least part
of what gives Maxi a reason to act. But understanding part of what gives
someone a reason for acting intentionally is not the same as a partial
understanding of intentional action.
Perhaps one reason the problem tends to go unnoticed is this. It is
common practice in developmental psychology to equate goals with
desires (Gopnik and Meltzoff 1997;3 Tomasello et al. 2005). On this way
of thinking, the ability to explain bodily movements in terms of desires
may be redescribed as the ability to explain actions in terms of goals. The
latter ability, in turn, is naturally equated with an understanding of goal-
directed action. This can make it look as if a grasp of the explanatory role
of desire just is a grasp of goal-directed action. But that appearance is
deceptive. To ascribe to children a conception of goal-directed action is to
say that they have mastered a distinctive way of explaining actions. They
understand what it means for someone to do one thing in order to achieve
another. This mode of explanation essentially involves putting together
someone’s purpose in acting with a particular way of pursuing the purpose.
Put differently, it involves an understanding of instrumental reasons
for action. In contrast, explaining a bodily movement in terms of some
mental cause, even if that cause is a desire, does not necessarily involve
any such “putting together” of end and means.4 A “desire psychologist”
might correctly explain someone’s salivating in terms of a desire for choco-
late. This does not mean that she thinks of the activity as an intentional
action. The ability to explain bodily movements as the effects of desires is
not sufficient for even a rudimentary understanding of what it is to act
intentionally.

3 Objective Reasons

The conception of practical reasons articulated by BD (practical reasons are

constituted or provided by suitable pairs of beliefs and desires) is arguably
not the only conception familiar to commonsense psychology. We often
think of practical reasons not as mental states but as worldly facts. When
deliberating about what to do—say, where to look for some prized object—
we tend to be more interested in the facts than in our beliefs about the
facts. We think of the fact that looking in a certain location will enable us
Teleology in Children’s Theory of Mind 203

to retrieve the object as giving us a reason to look there. From the perspec-
tive of deliberation, only true propositions—facts—can provide genuine
reasons. This point is not inconsistent with BD. The claim is not that BD
is false. It is only that BD does not exhaust the commonsense psychology
of practical reasons. In this observation, we suggest, lies the solution to
our “conceptual difficulty.” Young children find intentional actions intel-
ligible in terms of “objective” practical reasons.
Let’s clarify the basic idea with the help of a relatively uncontroversial
example from Bernard Williams. Suppose you believe of the content of a
certain bottle that “this stuff is gin,” when in fact the bottle contains
petrol. You feel like a gin and tonic. Should we say that you have a reason
to mix the stuff with tonic and drink it? Williams suggests that you do not
have such a reason, although you think you do, and although it would
certainly be rational for you to drink the stuff (Williams 1981b, 102). One
might object that to say that it would be rational for you to drink the stuff
just is to say that you have a reason to drink it—a reason that might be
appealed to, for example, in offering a “reason-giving” explanation of your
action. But Williams is surely right in drawing our attention to the fact
that you are mistaken in thinking you have a reason to drink the stuff. Your
putative reason can be set out as follows: “I need a gin and tonic. This stuff
is gin. So I should mix it with tonic and drink it.” Certainly the inference
reveals your action to be rational from your point of view. Correlatively,
appeal to the inference may figure in a reason-giving explanation of your
action. But the fact remains that the inference is unsound. Given that the
second premise is incorrect, the inference fails to establish the truth of its
conclusion: you are mistaken in taking the premises to establish that you
should drink the stuff. In this sense, you are mistaken in thinking you
have a reason for drinking it. This is perfectly consistent with acknowledg-
ing that there is a sense in which you do have such a reason. The point is
sometimes put by saying that you lack a justifying or guiding reason, but
have an explanatory reason to drink the stuff (Raz 1978). But this can be
misleading, given that an explanatory reason too may be said to justify, at
least in the “anaemic” sense (Davidson 1963/1980) of revealing the action
to be justified or rational from your perspective. Marking the distinction as
one between justifying and explanatory reasons would, in the current
context, be awkward in another way. For the suggestion we want to pursue
would now (confusingly) have to be put by saying that young children
explain intentional actions in terms of justifying (rather than explanatory)
reasons. So we’ll simply distinguish between subjective and objective reasons:
you have a subjective reason to drink the stuff, but you lack an objective
204 J. Perner and J. Roessler

reason to do so. Note that by itself the distinction does not imply that we
are talking about different sorts of things here. Subjective reasons need not
be taken to be mental states. Instead they might be taken to be propositions
forming the contents of mental states. The distinction turns on whether
or not a reason statement is to be understood as relativized to the agent’s
current perspective. To say that you have a subjective reason to drink the
stuff is to say that, from your perspective, it looks as if you have an (objec-
tive) reason to do so.
It might be said that even objective reasons have to be relativized in
one respect: they must be relativized to the agent’s set of desires or objec-
tives or projects. But on reflection it is not clear that this is so. As is often
pointed out by critics of the Humean theory of motivation5 (and as is
acknowledged by some of its defenders6), practical deliberation does not
always or even typically start from reflection about one’s current desires.
Practical inferences are often premised on evaluative propositions, to the
effect that some action or some state of affairs is important or desirable;
or, more specifically, on propositions involving “thick” evaluative con-
cepts (e.g., promise, treachery, brutality, courage), apparently embodying
“a union of fact and value” (Williams 1985, 129); or again, as in the
example above, on propositions to the effect that someone needs, or needs
to do, a certain thing, where claims of need are also best understood as a
species of evaluative propositions, not to be confused with, or reduced to,
ascriptions of desire (Wiggins 1987).7 This suggests that there may after all
be such a thing as a fully objective reason, relativized neither to the sub-
ject’s instrumental beliefs nor to her set of desires and projects. A possible
illustration might be the suggestion that the subject in Williams’s example
not only lacks an (objective) reason to drink the stuff, but actually has an
(objective) reason not to drink it—a reason provided by the fact that drink-
ing petrol is bad for your health.8

4 Teleology and the Hybrid Account

We are now in a position to formulate two possible solutions to our puzzle.

One proposal might be that young children find actions intelligible in
terms of partially objective reasons: reasons that are relativized to the
agent’s desires but not to her instrumental beliefs. For example, children
might take it that given Maxi’s desire to eat some chocolate, he has a good
reason to look in the blue cupboard (for doing so will as a matter of fact
enable him to satisfy his desire). Put differently, they conceive of Maxi’s
reason as a combination of a desire and an objective instrumental fact.
Teleology in Children’s Theory of Mind 205

Children predict and explain intentional actions by “putting together” a

mental state, defining the objective of the action, and an objective fact,
determining the means to achieve the objective. We will call this the hybrid
account of children’s conception of intentional action. (The hybrid account
is one way of developing the idea that young children are “desire psycholo-
gists.”9) The second solution agrees with the hybrid account that under-
standing the way young children think about intentional actions requires
going beyond BD. But it gives a more radical twist to that strategy. It
maintains that children find actions intelligible in terms of fully objective
reasons, relativized neither to the agent’s instrumental beliefs nor to her
pro-attitudes.10 They conceive of Maxi’s reason in terms such as these:
“Maxi needs his chocolate. (Or: It is important, or desirable, that Maxi
obtain his chocolate.) The way to get it is to look in the blue cupboard.
So he should look in the blue cupboard.” We call this the teleological
account.11
How might we tell which of the two explanatory schemes young chil-
dren go in for? We can spell out the key difference between them as
follows. Finding actions intelligible in terms of objective reasons implies a
certain lack of detachment. It involves commitments we do not normally
associate with reason-giving explanations (accustomed as we are to taking
such explanations to be a matter of citing subjective reasons). The two
explanatory schemes differ in the kinds of commitments they entail. An
explanation following the hybrid scheme carries with it a commitment
concerning the effectiveness of some means for achieving the agent’s
purpose. The teleological scheme in addition commits the interpreter to
some evaluative proposition. Both schemes share a crucial limitation: they
break down when it comes to actions informed by false instrumental
beliefs. But the teleological scheme is more limited still. It also breaks down
in relation to actions informed by values the interpreter does not share.
(Thus the teleological account might be said to involve an unusually strong
version of the principle of charity.) There are two obvious sorts of cases
where this extra limitation makes itself felt. One is the case of interpreting
actions informed by objectionable values (e.g., hurting someone because
it’s fun to do so). The other is the case of interpreting the activities of two
agents pursuing conflicting goals. Consider a competitive game. If you
understand A’s activities in terms of the fact that it’s desirable that A win
the game you will be hard pressed simultaneously to understand B’s reasons
for his contributions to the game, on pain of finding yourself committed
to inconsistent evaluative propositions: It’s desirable that A should win (by
beating B), and it’s desirable that B should win (by beating A).12 Thus the
206 J. Perner and J. Roessler

teleological account would predict that young children have difficulties in

understanding competitive games; more precisely, that they have difficul-
ties in understanding the reasons informing the activities of participants
in such games. By contrast, if young children subscribe to the hybrid
scheme, they should find it easy to understand each of the competing
activities, viz. in terms of the desires of each of the protagonists. (That A
wants to win is plainly consistent with B’s wanting to win also.) In general,
for a teleologist, cooperation is utterly natural; practical reasons are subject
to a kind of preestablished harmony. As conceived by the teleologist,
reasons are essentially intersubjective. If we explain Maxi’s action in terms
of the fact that he needs his chocolate, reference to that very fact may also
help to understand the actions of other agents. For example, the fact may
explain why someone else is doing something that will assist Maxi in
retrieving his chocolate.
How should we understand the issue between the hybrid view and the
teleological account? And what, if anything, is there of philosophical sig-
nificance in this debate? Well, it certainly looks initially as if the issue were
a straightforwardly empirical one. And there is some suggestive initial evi-
dence in favor of the teleological account: young children do seem to have
difficulties in understanding both wicked behavior and competitive games.
On the other hand, some will be inclined to doubt whether the teleological
scheme is even intelligible. If this were right, it would be implausible to
attribute the scheme to any rational thinker (even to very young thinkers).
Thus the teleological account might be rejected on a priori grounds. One
philosophical issue, then, is whether an objection of this sort can be sus-
tained. As will become clear later, there is also much of philosophical
interest in interpreting the empirical evidence. But there is a further way
in which the debate bears on the philosophy of action. Suppose the teleo-
logical account is intelligible and enjoys a measure of empirical support—
as we will argue it is and does. Suppose, in other words, that at least a
rudimentary conception of intentional action is available prior to, and
independently of, grasping concepts of propositional attitudes such as
belief or various kinds of pro-attitudes. This would raise the question of
how to understand the relation between the teleological conception and
the mature adult conception of intentional action. We return to this ques-
tion in section 7.

5 Teleology and Causal Understanding

We begin with the issue of intelligibility. A preliminary point to make

about the idea of objective practical reasons is that it amounts to nothing
Teleology in Children’s Theory of Mind 207

very grand or eccentric, metaphysically speaking. It is not the idea that

values belong to the fabric of the world as it is in itself, or that we have,
or need, a special faculty for detecting values. It is merely the idea that
there are true evaluative propositions. While this commits one to the view
that evaluative statements can be true or false, such a commitment need
not be seen as a particularly contentious matter. For it may be argued that,
as Williams put it, “truth in itself isn’t much” (Williams 1996, 19). The
central point here is that evaluative statements show the sorts of “surface
phenomena” (e.g., possible embedding in conditionals) that make it plau-
sible to think of them as bearers of truth-values. There is no commitment
to any substantive account of what it is in virtue of which such statements
are true (nor even to the possibility of any such account).
If there is a serious worry about the intelligibility of teleology, it is surely
over the claim that objective reasons can intelligibly be regarded as explan-
atory. One version of the challenge may be summarized as follows: “The
teleological scheme purports to explain actions in terms of facts that are
taken to constitute justifying reasons for it. Now, as Davidson has taught
us, it is one thing for someone to have a reason to perform a certain action;
it is another for that reason to explain her action—to be the reason for
which the agent acts. (For one thing, the agent may simply fail to do what
she has a reason to do. For another, even if she does it, and does it inten-
tionally, she may do it for some other reason.) There are well-known and
powerful considerations that suggest the notion of causation holds the key
to a proper understanding of the difference: reason-giving explanations
have to be conceived as a species of causal explanations (Davidson
1963/1980). But of course, causal explanations have to appeal to causes.
Yet causes are conspicuous by their absence in the teleological scheme.”
One might wonder whether the hybrid view fares any better than the
teleological account vis-à-vis the Davidsonian challenge. After all, the
hybrid account, too, envisages action explanations in terms of external
facts rather than just in terms of mental states. But it might be said, quite
plausibly, that there is a key difference. Suppose the upshot of the David-
sonian challenge is that, as one recent commentator put it, intentional
actions are to be explained “in terms of mental items—such as beliefs,
desires, intentions, and associated events . . . or their physical realizers”
(Mele 2003, 51, our emphasis). The idea is that for reason explanation to
be intelligible it has to make reference to particulars—events or, perhaps,
states—that stand in a causal relation to someone’s doing something inten-
tionally. The teleological account does not provide for that requirement.
Values are not events. It would be fanciful to think of Maxi’s need for
chocolate, or the importance of his getting his chocolate, as an “item” that
208 J. Perner and J. Roessler

stands in a causal relation to the event of his opening the green cupboard.
In contrast, the hybrid view may seem to provide for causes, conceived as
particulars. True, even on the hybrid view, the explanatory force of early
reason explanations cannot be exhaustively explained in terms of causal
relations between “mental items” and actions. External facts also play a
vital explanatory role. But defenders of the hybrid view might argue that
such facts should be seen as “standing conditions,” whose explanatory role
essentially depends on that of causally efficacious “items.” Their role may
be not unlike that of the dryness of the ground in an explanation of a
forest fire caused by someone’s dropping a cigarette. In brief, the Davidso-
nian challenge might seem to provide materials for an a priori argument
in favor of the hybrid view.
We want to suggest that this argument rests on an implausible premise.
The Davidsonian challenge consists of two central claims. One is that
action-explanation must be a species of causal explanation. This is usually
motivated by arguing, convincingly, that action-explanations are explana-
tions of the occurrence of events, and that it’s hard to see how the occur-
rence of an event can be made intelligible other than in causal terms. The
second claim is that causal explanation has to appeal to causes, conceived
as particulars. The correct teleological response to the a priori argument
for the hybrid view, we suggest, is to accept the first but reject the second
claim. The key question here is: What does it mean for an explanation to
be causal? Adapting Bernard Williams’s remark about truth, perhaps the
right thing to say here is that causal explanation in itself isn’t much. It’s
not explanation in terms of laws of nature; it’s not explanation in terms
of event causation; it’s not explanation in terms of causal processes or
causal mechanisms. The basic idea of a more minimalist account is that
causal explanations advert to facts that “make a difference,” where this is
to be spelled out in terms of patterns of counterfactual dependence.13 Our
suggestion, then, is that teleological explanation is a species of causal
explanation. So we agree with the idea underpinning the Davidsonian
challenge to teleology, that the development of the commonsense concep-
tion of intentional action is inextricably entwined with the development
of causal understanding. But we deny that causal understanding in this
area takes the form of understanding causal relations between “mental
items” and actions.
Specifically, the version of the difference-making approach we propose
to draw on is the so-called interventionist approach to causation and causal
explanation (Woodward 2003). The central idea of interventionism is this.
To say that there is a causal relation between two variables X and Y is to
Teleology in Children’s Theory of Mind 209

say that if there were to be an intervention changing the value of X, there

would also be a change in the value of Y. To say that taking aspirin causes
relief from headache is to say that if there were an intervention on the
amount of aspirin ingested by someone suffering from headache, there
would also be a change in the intensity of her headache. The notion of an
intervention is itself a causal notion. Crudely, an intervention on X is a
causal relation between some variable I and X such that I causes a change
in the value of X in an “exogenous” way—I takes complete control of X
without affecting Y in any other way than through X.14 This sort of account
aims to elucidate the meaning of causal claims in terms of interventionist
counterfactuals—in terms of how the value of Y would change under
interventions on X—without, of course, undertaking to offer a reductive
analysis.
Now on an interventionist approach there is, on the face of it, nothing
a priori objectionable about the idea that objective reasons may causally
explain intentional actions. The idea comes to this: were there to be an
intervention on the evaluative or instrumental facts that give someone a
practical reason, there would be a corresponding change in her action. The
suggestion that young children think about intentional actions in this way
is not at all implausible, especially in light of the sorts of social interaction
characteristic of early childhood. Consider games of mutual imitation. The
simple rule defining such games is that one participant has to imitate the
actions of the other. In the context of this game, I can reliably manipulate
your actions by intervening on your reasons. For example, my banging my
toy gives you a reason to bang yours. (Only by banging yours can you
achieve the defining purpose of the game, viz., to imitate my action.) So
the fact that I am banging my toy offers a good causal-teleological explana-
tion of your action, a causal explanation in terms of reason-giving facts.15
To grasp such explanations it is not enough to understand that you have
a reason to perform the action. Children also have to understand the
reason as explanatory of the action. This they do, in virtue of their devel-
oping understanding of interventionist counterfactuals.16 On this view, the
causal understanding that goes into children’s conception of intentional
action does not require, or consist in, a conception of unobservable enti-
ties—“mental items”—conceived as causes of bodily movements. Rather,
it is a matter of understanding what will, or would, happen under interven-
tions on someone’s practical reasons.
Note that the teleological account is concerned exclusively with the core
element of children’s understanding of intentional action, their capacity
for reason-giving explanation. The account does not imply that the
210 J. Perner and J. Roessler

capacity operates in isolation. Nor does it imply that psychological proper-

ties play no role in such explanations. One way in which teleology may
be enriched is by adding the idea, which may be in place quite early in
development, that people’s responsiveness to their practical reasons is
subject to certain enabling (and disabling) conditions. For example, the
agent has to be awake, not too tired or distracted, she may have to be able
to perceive some critical object, and so on. Of particular interest here is
the notion of attention: even young children may have some understand-
ing of someone’s attention or “engagement,”17 as something that has to
be suitably focused if the person is to respond to the reasons afforded by
a situation. Children’s understanding of the causal relevance of this condi-
tion may manifest itself in their practical interventions on others’ focus of
attention; for example, their attempts at attracting someone’s attention by
pointing or shouting. The point is important since it helps to allay the
worry that the resources of teleology are simply too limited to offer a cred-
ible picture of early social understanding. Pure teleology admittedly leaves
many things unexplained. A teleologist who thinks it’s time for a game of
mutual imitation should be baffled if his chosen partner shows herself to
be unresponsive. Of course, being unable to make rational sense of an
action need not prevent a teleologist from registering that the action (or
inaction) is taking place. (Much of the time, mature commonsense psy-
chologists are in the same position.) Still, the limitation may seem crip-
pling. But there are two ways in which teleologists may overcome this
limitation. An initially inscrutable action may become intelligible once the
teleologist comes to understand, and share, the value informing it.18 Alter-
natively, someone’s (in-)action may remain rationally unintelligible, but
the agent’s failure to respond to her reasons may itself be unsurprising
given the satisfaction of some disabling condition. (She may be asleep.)
Teleologists need not be at sea in the social world.
What we suggest would be a non sequitur is to move from the involve-
ment of psychological material in young children’s understanding of
intentional agency to the conclusion that they must be finding actions
intelligible in terms of subjective reasons—reasons relativized to the agent’s
cognitive/evaluative perspective. There may be important roles for the
psychological material to play other than to delineate what, relative to the
agent’s subjective perspective, it is rational for her to do. (The same is true
in the case of mature commonsense psychology.) It is no trivial task to
decide whether a particular example of early mentalism manifests an
understanding of subjective reasons—as becomes clear when we turn to
the experimental evidence.
Teleology in Children’s Theory of Mind 211

6 The Evidence

It is sometimes said that children appreciate the subjectivity of desires

many years before they grasp the subjectivity of belief (e.g., Repacholi and
Gopnik 1997; Rakoczy, Warneken, and Tomasello 2007). There is more
than one way to understand this claim. On one reading, it amounts to an
endorsement of the hybrid view: children are able to understand the rela-
tivity of practical reasons to the agent’s desires before they are able to rela-
tivize practical reasons to the agent’s beliefs. We want to suggest that, on
this reading, the claim lacks empirical support. There is a rough distinction
we can make between two sorts of evidence that are relevant here. On the
one hand, there is evidence that might be appealed to in support of the
hybrid view. We discuss two such findings: concerning infants’ under-
standing of subjective preferences, and concerning verbal references to
desires in action-explanation. On the other hand, there are findings that
appear to support the teleological account. As will become apparent, our
empirical defense of the teleological account is somewhat qualified. While
there is some suggestive evidence in its favor, and no persuasive evidence
against it, we believe that more tightly controlled experimental work is
needed to make further empirical progress with these matters.

6.1 Subjective Preferences

Even 18-month-olds seem to be aware of the subjectivity of desire in the
following sense: they are sensitive to individual preferences. Evidence for
this comes from a study where infants were presented with two kinds of
food (yummy crackers and yucky broccoli). In a control condition, when
another person asked them for something to eat, they almost always gave
that person a cracker rather than a piece of broccoli. However, in the dis-
crepant desire condition the other person first demonstrated deviant pref-
erences. Taking the crackers she exclaimed “Yuck, that’s awful!,” underlined
by the appropriate facial expression. When she tasted broccoli she looked
pleased and also said so: “Mmm, that’s good!” Now when this person later
asked the infant for something to eat, most children younger than 18
months still handed her the crackers, but most children over 18 months
handed her the broccoli (Repacholi and Gopnik 1997). A similar finding
is that 2-year-old children freely verbalize subjective preferences: “Daddy
likes shaving cream. I hate shaving cream” (Bartsch and Wellman 1995).
One sort of question raised by these findings is how we should charac-
terize children’s conception of subjective preferences. Questions under this
heading include the following: Do children think of “liking” as a mental
212 J. Perner and J. Roessler

state, something like a desire or an emotion—associated perhaps with a

distinctive experience—or rather as something like a character trait (on the
lines of “he likes to be in charge”)? Another question raised by the findings
is: what sort of explanatory role do subjective preferences occupy in chil-
dren’s theory of mind? Do children think of subjective preferences as
desires in the sense of BD—as part of what makes it rational for the subject
of the desire (but for no one else, unless he or she has a desire to cooperate)
to use suitable means to satisfy the desire? Or do they think of subjective
preferences in some other way, giving them a different sort of explanatory
role?
Note that these are quite different issues. Suppose children do think of
subjective preferences as involving conscious mental states, say, desires for
objects. This should not automatically lead us to assume that they conceive
of such desires as states that provide agents with subjective reasons. That
would be to ignore the possibility that children think of “likings” as being
essentially linked to objective purposes. Thus children’s conception of sub-
jective preferences may be part of their conception of objective needs.
Subjective preferences may be taken to determine what is good for different
sorts of people or creatures. Broccoli is good for A, but not for B. Eating
grass is good for cattle but not for humans. This is a relatively sophisticated
(perhaps in one sense, nonegocentric) conception of needs. But it does not
require any understanding of subjective reasons. On this conception, the
import of A’s and B’s differential preferences is that it is objectively desir-
able that A, but not B, should obtain some broccoli (Perner, Zauner, and
Sprung 2005). Correlatively, there is no privileged explanatory relation
between A’s preference and A’s action: A’s preference makes it as reason-
able, and as intelligible, for B to pass A a piece of broccoli as for A to set
off for the greengrocer. The findings concerning subjective preferences are
consistent with this interpretation, and hence with the teleological
account.

6.2 Talk about Desires

We mentioned two potential instances of early psychological understand-
ing: children’s grasp of attention (and perhaps perceptual experience), and
their understanding of subjective preferences. Neither of them has any
tendency to suggest that children appeal to mental states as providers of
subjective reasons; both are naturally incorporated into the teleological
scheme of explanation. But there is another sort of connection between
teleology and psychology. There is a sense in which teleology provides for
certain psychological notions. If the teleological account is right, finding
Teleology in Children’s Theory of Mind 213

someone’s actions (causally) intelligible in terms of reason-giving worldly

facts provides for a primitive notion of intentional action—arguably, a
psychological phenomenon. Now setting out a teleological explanation in
full would require reconstructing the practical inference that articulates
the reason for which the agent performed the action, with an evaluative
and instrumental premise and a conclusion to the effect that, say, the agent
should open a particular cupboard. A more minimalist way of explaining
the action would be merely to identify its goal. Rather than saying Maxi
opened the cupboard because he needs his chocolate etc., a teleologist may
simply tell us that it was in order to get his chocolate that Maxi opened
the cupboard. Or: Maxi opened the cupboard with the intention of taking
out his chocolate. In this way, the teleological scheme of explanation can
be seen to provide for a simple notion of intention (the intention with
which someone acts). This is arguably a psychological notion in good
standing, but it is not one that helps to explain actions on the model of
BD. The intention with which someone acts, as it is understood by a tele-
ologist, may be classified as a (kind of) desire, but a teleologist does not
conceive of it as a state that provides the agent with a subjective reason—a
desire that, together with a belief, helps to rationalize intentional actions.
Rather, that an agent has this intention follows from the fact that his action
is open to a reason-giving, teleological explanation.
The point is significant given that it may help to interpret some inter-
esting data concerning early verbalizations. Children talk about desires
substantially earlier than about beliefs (see Bartsch and Wellman 1995).
Much of this talk is a matter of expressing their own requests and wishes.
Some of it is a matter of reflecting on sometimes contrasting subjective
preferences (see above). Least frequently, but perhaps most interestingly,
some examples have been recorded of early appeals to desires in action-
explanations. A striking illustration is the following (Bartsch and Wellman
1995, 118):

Ross (2;10): Look, there’s a car up in the air. [on a hoist]

Adult: Oh. Why is it up there?
Ross: Man put it up there.
Adult: Why did he put it up there?
Ross: He want to fix it.

On one reading, “want to fix it” is to be construed as an ascription of a

mental state distinct from, and causally responsible for, the intentional
action in question—a desire that both gives the man a subjective reason
to put the car up in the air and causally explains his doing so. But there
214 J. Perner and J. Roessler

is an alternative reading. The child may make the man’s action intelligible
by stating the intention with which it was performed: his intention in
putting the car up was to fix it. The example reveals the child’s understand-
ing that people act on the basis of good reasons, and provides a partial
reconstruction of the reason operative in this case, identifying the purpose
informing the action. Note that on this reading, it’s possible that Ross takes
the reason to be provided by the objective desirability of the man’s fixing
the car, rather than by the man’s desire to fix it. Thus, the example pro-
vides no evidence against the teleological account.

6.3 Competitive Games

When introducing teleology we mentioned the importance of two sorts of
cases: actions informed by (to the interpreter) unacceptable goals (e.g.,
throwing a ball at someone because one finds it desirable to hurt her); and
competitive actions. If children make sense of intentional actions via the
teleological scheme, they should find these sorts of cases unintelligible. If
they use the hybrid scheme, these cases shouldn’t pose any special chal-
lenge. As we indicated, there is some evidence to suggest that young
children do find these cases challenging. Now the evidence relating
to unacceptable goals (Yuill 1984) relies on an experimental paradigm that,
at least in the current context, is to be treated with some caution (as we
argue in appendix 2). So we will focus here on the second kind of case.
A good test case between teleology and the hybrid view arises when it
comes to understanding competitive behavior as intentional actions, done
for a reason. In a competitive (zero-sum) game (as are most simple strategic
or probabilistic games) one player’s win means the other player’s defeat.
So their respective contributions to the game cannot be understood as
attempts to attain some objectively desirable end, valid for both. The teleo-
logical account therefore predicts that competition involving reasoned
actions can only be appreciated by children who understand differences
of perspective, including the case of false beliefs. In contrast, on the hybrid
theory, understanding competitive behavior should pose no particular
challenge: children should be able to appreciate that relative to their
respective objectives (to win the game), both A’s and B’s behavior makes
rational sense. The question is where to find relevant developmental evi-
dence. The only directly relevant evidence that we are aware of is based
on the hand-guessing (commonly also known as “penny hiding”) game
used in a study by Gratch (1964).
The penny-hiding game is a two-person game in which the subject is
actively involved either as a guesser or as a hider. Both try to win. The
Teleology in Children’s Theory of Mind 215

hider hides the penny in one hand or the other, and then invites a guess.
This is repeated over a row of trials, after which the participants change
roles. In the role of hider “the child was judged to be competitive if he
expressed displeasure on any of the trials in which E found the marble, or
if any of the following events occurred: 1. when E selected the marble-
holding hand, S refused to show the marble or made an attempt to transfer
it to his other hand; 2. S extended an empty hand for E to guess from. A
child was judged to be non-competitive if none of the above events occurred.
The non-competitive children frequently told E where the marble would
be hidden, after a trial in which E had failed to guess correctly; and many
of them extended the marble in an open hand on all trials” (Gratch 1964,
53–54).
The proportion of children who displayed competitive spirit increased
steadily from 5 percent below the age of 3 years to 58 percent at 4.5 years
to practically 95 percent at around 6 years. This fits very well the typical
developmental trend on false-belief tests (see Perner, Zauner, and Sprung
2005, figure 4, for a graphic display). Gratch’s study is particularly helpful
insofar as he analyzed indicators of competitive spirit separately from
indicators of the ability to deceive, fool, or conceal information from the
opponent. Since deception requires an understanding of false belief, a
developmental link between understanding the deceptive aspects of the
game and understanding false belief would not be terribly interesting and
no support for the teleological theory developed here. The shortcoming of
Gratch’s data for present purposes is, of course, that there is no direct
comparison of how many of the children in his sample would have passed
the false-belief test. Ideally one would also see the use of a control task
with similar cognitive demands except for the competitiveness and in
which children of all ages can succeed.
The good news is that more recently several studies included the penny-
guessing game and the false-belief tasks (Baron-Cohen 1992; Chasiotis et
al. 2006; Hughes and Dunn 1997, 1998). The bad news is that, without
exception, these studies only analyzed the hand-guessing behavior for
indicators of deceptive abilities (or a mix of combative spirit and decep-
tion). Hence the reported correlations with false belief understanding
provide no convincing evidence against the hybrid theory and, therefore,
also no support for teleology.
We should also mention two other pieces of evidence that go well with
teleology. Although children seem to have problems understanding the
point of competition, they are quite concerned about obeying the rules of
a game (Rakoczy, Warneken, and Tomasello 2008) even at the age of 2
216 J. Perner and J. Roessler

years when they just start to use “desire” terms for other people (Bartsch
and Wellman 1995; Imbens-Bailey, Prost, and Fabricius 1997). At 3 years
(36 months) this concern becomes almost obsessive. Clearly they expect
people to act a certain way because it is the right, conventional way, and
they seem to have little understanding for idiosyncratic deviation.
Following the pioneering work by Shultz and Shamash (1981), several
studies reported that children have difficulty distinguishing intentions
from desires (see reviews by Astington 1999, 2001). The latest study by
Schult (2002) included children as young as 3 years. They had to toss bean
bags into three different colored buckets, some of which contained a ticket
for a prize. For each toss they had to indicate which bucket they intended
to hit. On some trials they hit the intended bucket, on others they missed
it; on some they won a prize, on others they didn’t, resulting in four dif-
ferent combinations. The 4- and 5-year-olds were remarkably accurate in
answering all types of questions. The 3-year-olds, on the other hand, had
serious problems with questions about their intentions, in particular when
satisfaction of their intention contrasted with satisfaction of their desire,
as shown in table 14.1.
This pattern of results follows from our assumption that children remain
basic teleologists until about 4 years, when they can understand differences
of perspective. They have no problem knowing what they want, i.e. the
desirable goal of the action (winning the prize), and whether they got it
or not. They also understand intentions to hit a particular bucket, though
only insofar as there are objective reasons for such intentions. Consider now
a case of fortuitous success, where children accidentally get the prize after
hitting a bucket they didn’t intend to hit. To understand that they didn’t
intentionally hit the bucket, children have to understand that they had

Table 14.1
Data from Schult (2002): number of children giving correct answers to the satisfac-
tion questions, “Did you do what you were trying to do?” (intention), and “Did you
get what you wanted?” (desire).

Condition (correct answer) 3-Year-Olds 4-Year-Olds 5-Year-Olds

Miss target/get prize: fortuitous success

Intention (“no”) 4/15 8/8 8/8
Desire (“yes”) 12/15 8/8 8/8

Hit target/no prize: bad luck

Intention (“yes”) 8/15 8/8 8/8
Desire (“no”) 13/15 8/8 8/8
Teleology in Children’s Theory of Mind 217

no reason for hitting that particular bucket, despite the fact that doing so
turned out to be conducive to reaching their goal. Or consider a case of
bad luck, where they hit a certain bucket without getting a prize. To under-
stand that they hit the bucket intentionally, children have to understand
that they did have a reason for hitting that bucket, despite the fact that
doing so turned out not to be conducive to reaching their goal. Under the
teleological interpretation, it is unsurprising that young children have
problems under these kinds of circumstances. Correct judgment of these
cases only becomes possible when one understands that one acted on the
assumption that the prize might be in the bucket one was aiming for. Since
in the critical cases this assumption turns out to be false, the intentionality
of the intended action can only be understood if one can understand it in
terms of the perspective of that assumption.
In sum, there is some suggestive evidence against the hybrid theory and
some support for the teleological account in Gratch’s finding that children
show little competitive spirit before the age at which they are able to
understand false belief as a motivating reason. More importantly (for the
developmental psychologists), our account also provides us with a clearer
analysis of what young children should find difficult about competition
and incompatibility of goals. In appendix 2, we discuss further evidence
that is prima facie relevant to the debate between teleology and the hybrid
view (evidence concerning children’s understanding of emotional reac-
tions to the satisfaction or frustration of desires). In appendix 3, we present
an outline of a new experimental paradigm to test for children’s ability to
understand competitive actions.

7 Teleology in Perspective

We conclude with a brief suggestion as to how the teleological account

bears on our understanding of adult commonsense psychology. Suppose
young children find intentional actions intelligible in terms of reason-
providing facts, rather than mental states. We can distinguish two different
ways in which this initial understanding subsequently gets refined and
enriched, corresponding to two distinct elements of the mature adult con-
ception of intentional action. Both involve giving explanatory relevance
to an agent’s perspective, though in quite different ways.
First, a central element of a more sophisticated conception of practical
reasons is an appreciation of the role of knowledge and ignorance. We may
think of this as a more subtle application of the possibly earlier developing
idea, that agents’ responsiveness to reasons is subject to enabling and
218 J. Perner and J. Roessler

disabling conditions. In its subtler (adult) form, the idea is this: if someone
doesn’t know that action x causes event y, it’s unsurprising that she won’t
perform x, despite having reason to cause y—for she lacks a subjective
reason to perform x. Correlatively, her performing x can be intelligible in
terms of her knowledge that x causes y. Note that explanations in terms
of knowledge are simultaneously teleological (they accord an explanatory
role to reason-giving facts, knowledge being a factive state) and psychologi-
cal (knowledge being a psychological state).
Second, mature interpreters are also able to explain actions in terms of
considerations the agent takes to provide her with a reason, without
endorsing the consideration or the claim that they constitute reasons. It
is this dimension of the “adult theory” that enables us to figure out why
someone is adding petrol to her tonic or to understand competition. And
it is here that we find the rationale for BD: adopting a detached, relatively
noncommittal stance toward others’ intentional activities requires finding
them intelligible in terms of reasons provided by their propositional
attitudes.
The teleological genealogy of the adult theory has important implica-
tions for the sorts of psychological properties invoked in the mature con-
ception of intentional action. It should lead us to question the relentless
focus on just two kinds of mental states, beliefs and desires, encouraged
by the “belief-desire model” of action-explanation. The teleological geneal-
ogy highlights the explanatory role given in the adult theory to knowl-
edge.19 It also suggests that the notion of desire in BD should be interpreted
not in the narrow Humean sense but along the lines of Davidson’s “pro-
attitudes,”20 as subsuming the immense variety of attitudes that constitute
agents’ perspectives on the purposes for which they act. Reflection on the
teleological origin of commonsense psychology also helps to shed light on
the sort of explanation required for understanding intentional action. It is
not enough to think of certain mental states as the causes of bodily move-
ments. What matters is the ability to see how some of the agent’s psycho-
logical properties provide her with considerations that from her point of
view can be seen to amount to a practical reason. Understanding the sub-
jective reason informing someone’s intentional action requires delineating
what, from her perspective, presents itself as an objective reason.

Appendix 1: Direct and Indirect Tests

The recent investigations of children’s understanding of rational action and

belief have made it necessary to distinguish two sources of evidence. The
Teleology in Children’s Theory of Mind 219

“traditional” evidence is based on what is called in the consciousness litera-

ture (Reingold and Merikle 1993; Richardson-Klavehn and Bjork 1988) a
direct test of knowledge: Children are asked a question (or set a task) that refers
(directly) to an actor’s action in question. For instance, in the false-belief
task children are told a story or made to observe a scene in which a protago-
nist puts an object in one of two locations, which is unexpectedly trans-
ferred to the other location in the protagonist’s absence. When he returns
children are asked: “Where does the protagonist think the object is?” or
“Where will the protagonist go to get the object (he desires)?” On these
measures a large amount of data shows that children start to give correct
answers between the ages of 3 to 5 years. (See the meta-analysis by Wellman,
Cross, and Watson 2001.) The younger children’s difficulty stands in stark
contrast to their ability to predict with conviction where someone will go
for the object when the protagonist has witnessed the transfer and no false
belief is involved (e.g., Clements and Perner 1994; Ruffman et al. 2001).
Indeed, children not only fail to take belief into account in their predictions
of the protagonist’s actions but also fail to use it in explanations of the
erroneous actions. For instance, when shown that the protagonist in the
false-belief story goes to the empty location and they are asked why he went
there, many of them answer (not incorrectly but uninformatively) “because
he wants to get the object.” When asked to specify why he then went to the
empty location, they admit their ignorance: “Don’t know” (Wimmer and
Mayringer 1998; Perner, Lang, and Kloo 2002).
The fact that children feel comfortable talking about desires (want)
before they address beliefs is also apparent in Bartsch and Wellman’s
(1995) study of naturally occurring speech. The authors tried, as best as
they could, to identify genuine references to beliefs and desires from the
limited context of the transcribed utterances (their figure 5.1). Reference
to desire started with the first words as early as 18 months, peaking just
before 3 years (33–36 months) to level off at about 3 percent of all utter-
ances. In contrast, references to beliefs started at 3 years, reaching a con-
stant level of about 2 percent of all utterances at 4 years. A similar picture
also emerged from induced utterances in a study by Imbens-Bailey, Prost,
and Fabricius (1997), conveniently displayed in our figure 14.1. Also, by
using a method for eliciting extensive explanations, Bartsch, Campbell,
and Troseth (2007) were able to get half the 3- to 4-year-olds to give belief-
explanations of erroneous actions, but emotion- and desire-explanations
were still much more frequent. However, when relying on nonverbal indi-
cators of understanding belief and desire, recent research has complicated
this seemingly clear picture.
220 J. Perner and J. Roessler

0.9
Cumulative Proportion of Children

0.8

0.7

0.6

0.5

0.4

0.3

0.2

0.1

0
18 21 24 27 30 33 36 39 42
Age in Months

Self-perception Self-desire Self-belief

Other-perception Other-desire Other-belief

Figure 14.1
Proportion of children who use different categories of mental verbs.

Starting with Clements and Perner (1994), research has recently focused
on using indirect tests of knowledge: Children are not asked any question
about the protagonist’s belief or action, but one of three other measures
is used. (1) Looking time: The duration of looking at an erroneous action
is compared to looking at a successful action. Longer looking at the suc-
cessful action than at the erroneous action in cases of false belief is inter-
preted as children being surprised about a successful action when the actor
has a false belief. These data indicate sensitivity to the protagonist’s belief
as early as 14 or 15 months (e.g., Onishi and Baillargeon 2005; Surian,
Caldi, and Sperber 2007). One problem with these data is that the looking-
time differences are multiply interpretable and not a clear indicator of
expectation (Perner and Ruffman 2005; Sirois and Jackson 2007). (2)
Looking in expectation: A better measure in this respect is children’s direc-
tion of eye gaze as an indication of where they expect the protagonist to
reappear (Clements and Perner 1994). This method has recently indicated
Teleology in Children’s Theory of Mind 221

that children understand belief-based action as early as 2 years (Southgate,

Senju, and Csibra 2007) or even earlier (Southgate 2008). (3) Interpretation
of referential gestures: Southgate (2008) used a paradigm developed by
Carpenter, Call, and Tomasello (2002) and Happé and Loth (2002) on
children as young as 17 months. These infants observed a confederate
putting object A into container X and object B into container Y. In her
absence the infants watched the experimenter switch the objects (A now
in Y and B in X). When the confederate returned she pointed to, say, box
X and said, “I want this one.” Then the covers of the boxes on the child’s
side were opened so that the child, but not the confederate, could see
inside and the confederate asked the child to give the object to her. Chil-
dren above chance handed her the object in box Y, which contained the
object that the confederate had put inside box X but was no longer in
there when the confederate expressed her desire for the object in box X.
We concentrate in our discussion on the evidence from direct tests.
Answers to questions more clearly express a reasoned commitment, that
is, an understanding that the protagonist will act the indicated way because
he has good reason to do so. Indirect measures contain no commitment.
I can look to a location because of a vague, unreasoned hunch even when
to my proper reasoning he will appear somewhere else. Indeed, that is what
the looking-in-expectation data seem to indicate: Children look (in expec-
tation) to where the protagonist thinks the object is, but steadfastly and
with conviction (Ruffman et al. 2001) they claim that he will appear at the
location where the object actually is. Perner, Rendl, and Garnham (2007)
proposed that the false-belief data from infancy (see also, e.g., Moll and
Tomasello 2007; Tomasello and Haberl 2003; Southgate, Senju, and Csibra
2007) show an unexpectedly strong preoccupation with what other people
can and cannot see, and keep an “experiential record” for the other person,
using this record to compute likely actions and referential gestures but
without clear understanding that these records constitute beliefs that direct
action in the real world.
Investigations of children’s understanding of knowledge presents a
similar picture. In line with the evidence of early sensitivity to beliefs,
children as young as 14 or 18 months keep track of what another person
has or has not yet encountered (Moll and Tomasello 2007; Tomasello and
Haberl 2003; Southgate, Senju, and Csibra 2007) and make their com-
municative behavior dependent on it by preferring to show the person
objects that the person has not yet encountered over encountered objects
(regardless of the child’s own familiarity with the object). In all these cases,
the use of this experiential record (Perner, Rendl, and Garnham 2007;
222 J. Perner and J. Roessler

“registration of encounters,” Apperly and Butterfill 2008) is limited to a

single purpose, namely, to bring the other person into informational
contact with yet unencountered objects. In the only other experimental
paradigm used, children are asked to decide whom to ask for information,
the one who has seen the hiding of the object or the one who was not
able to see it. Here children seem oblivious to the other person’s informa-
tional access until 4 years (Povinelli and deBlois 1992), or, with improved
methodology, not before 3 years (Sodian, Thoermer, and Dietrich 2006).

Appendix 2: Action versus Emotion

As we mentioned in the text, there is more than one way to interpret the
claim that children understand the “subjectivity of desires.” One reading
is: they understand subjective preferences. A second reading is: they under-
stand that a desire provides an agent with a subjective reason, a reason
that makes it rational for the agent (but for no one else—unless they wish
to cooperate) to act in a certain way. We argued that evidence regarding
subjective preference provides no support for the claim that children
understand subjective reasons. Some of the experimental work in this area,
however, is concerned with a third claim. This is the claim that young
children are able to understand and predict someone’s emotional reaction
to the satisfaction or frustration of some desire even when they don’t share
the desire. For example, Yuill (1984) addressed the question of whether
children can predict that an agent will take pleasure in the satisfaction of
a wicked desire (such as a desire to hit someone). Similarly, Perner, Zauner,
and Sprung (2005) investigated children’s ability to attribute emotional
reaction to the satisfaction/frustration of desires in the case of two pro-
tagonists with mutually incompatible desires.
Does evidence from these paradigms help to settle the issue between
teleology and the hybrid view? There are two reasons for skepticism. One
is simply that the experimental state of play regarding the third claim is
currently inconclusive. Crudely: Yuill’s (1984) and Perner, Zauner, and
Sprung’s (2005) findings suggest that until they pass the false-belief task,
children have great difficulty attributing emotional reactions to the satis-
faction of goals they don’t take to be objectively desirable. But more
recently, Rakoczy, Warneken, and Tomasello (2007) and Rakoczy,
Warneken, and Tomasello (2008) reported some evidence that children
can attribute appropriate emotions in competitive situations before they
pass the false-belief test.
Teleology in Children’s Theory of Mind 223

A second, weightier reason for skepticism is this. It is not clear that

explaining emotional reactions to the satisfaction/frustration of desires is
a case of reason-giving explanation. It is not clear, for example, that to
understand and predict that A will feel sad about the frustration of his
desire, children have to appreciate that A has a reason for feeling sad. Emo-
tional reactions happen to us: they are caused, for example, by perceived
events, not formed for reasons. Correlatively, understanding such reactions
is not a matter of reconstructing the agent’s reason for undergoing it. To
understand A’s sadness you merely have to know that the frustration of a
desire tends to lead to sadness, not that it makes it rational to feel sad.
Admittedly, this crude contrast doesn’t do justice to the subtleties of adult
commonsense psychology in this area. (See Goldie 2000, ch. 2, for some
helpful distinctions.) But it is enough to suggest that it would be a non
sequitur to move from evidence (if there were such evidence) that children
are able to understand emotional reactions to desires they don’t share
to the conclusion that they grasp subjective reasons—that they think of
A’s desires as making certain actions (or emotions) rational from A’s
perspective.

Appendix 3: Sabotage

The empirically pressing question now is how to derive testable predictions

from the assumption that young children use teleology before they are
able to understand perspective differences. We need a task for which
simple teleology is insufficient, i.e. which requires understanding the pro-
tagonists’ intentional actions in terms of their perspectives on what’s
desirable. We need another, control task, identical to the experimental task
except that teleology can provide the correct answer without need for
understanding perspective differences. The predictions then are that the
experimental task should be mastered after the control task and not before
(within a margin of error) other typical perspective tasks (involving per-
spective differences, e.g., false-belief task) are mastered. Here is a suggestion
involving sabotage.
Children can engage in sabotage before understanding false belief
(Sodian 1991)—in seeming contradiction to our claim. In Sodian’s tasks
the children were asked to think of a way to prevent a robber from getting
to the treasure. They correctly locked the treasure box when the robber
approached but not when the good prince came to look for treasure. Criti-
cally for us, correct responding may not require children to interpret the
 224 J. Perner and J. Roessler

The Coconut Story

Conflict:
far
Abe wants the coconut. ?
Bea, too, wants the
coconut.

Base manipulation:
Abe
Collaborative: Tree either on far
Abe wants the coconut. or near island.
Bea wants to help him
bring it home.

?
Neutral: No Bea. near

Bea

Figure 14.2
Test question: Which board will Abe use to get to the coconut?

robber’s behavior—to find it intelligible in terms of his reasons. The child’s

task is only to find a means to prevent something bad. A new paradigm
must ensure that prediction of the saboteur’s act must be based on under-
standing what the sabotee will do for (his) good reasons.
In the experimental conflict story of the coconut scenario in figure 14.2
the girl, Bea, wants to get the one coconut left on the tree on the island.
She crosses over the black board to the center island. Then she remembers
that she needs her cart. She goes back to get the cart. Meanwhile, the boy,
Abe, has arrived also with the intention to get the last coconut. To get to
the tree he needs another board. He has the choice between the black and
the brown board. Test question: Which board will he take to cross the
moat, the brown or the black one? Correct answer: the black board.
Teleology in Children’s Theory of Mind 225

In a collaborative condition, Bea indicates that she wants to help Abe

pick the coconut. The correct answer to the test question in this condition
is: the brown board. In a neutral condition, without Bea, we can get an
estimate of children’s preference for the “near” board (brown) over the
“far” board (black). Presumably most children will make him use the nearer
one. By counterbalancing the competitive and collaborative conditions
with whether the palm tree is on the near (to Bea’s approach) or the far
island, we can check whether the understanding of the collaborative and
the competitive aspects, respectively, can overrule the natural tendency to
predict use of the most convenient nearer board.
We expect young children who fail false-belief tasks (as a measure of
their ability to understand conflicting perspectives) to be able to predict
correctly what Abe will do in the cooperative condition (use the brown
board), but to fail to see that he will act differently in the competitive
condition (he will use the black board). Children who pass the false-belief
test, we expect, will make correct predictions in both conditions.

Notes

1. Or rather Johann Nestroy, as quoted in the motto of Wittgenstein (1958).

2. It is tempting to think the solution to our puzzle may be provided by some recent
experimental findings, widely taken to show that under certain conditions even very
young children have some grasp of the causal role of false beliefs. In Appendix 1
we explain why we think the temptation should be resisted.

3. Compare Gopnik and Meltzoff’s claim that 18-month-olds who participated in

a study of imitation of failed attempts were able to treat different kinds of move-
ments “as the causal consequence of the same underlying mental state, the same
goal” (1997, 151).

4. See Schueler 2009 for illuminating discussion of the importance, and the theoreti-
cal implications, of the “putting-together” point.

5. See e.g., Parfit 1997; Scanlon 1998; Schueler 2003, 2009; Hornsby 2008.

6. “[T]ypically, in deliberation what I do pay attention to are the relevant features

of the external world [rather than one’s own desires or impulses]: the cost of the
alternatives, the quality of the food, the durability of the cloth, the fact that I made
a promise” (Blackburn 1998, 253).

7. Crudely, claims of need have implications to do with matters such as harm and
flourishing, but they have no immediate implications regarding the agent’s desires.
(That you need some vitally important medicine does not imply that you want to
take it.)
226 J. Perner and J. Roessler

8. Are fully objective reasons external reasons in Williams’s (1981a) sense—reasons

that cannot be derived by practical deliberation from the agent’s “subjective moti-
vational set”? The question raises complex interpretative and substantive issues we
cannot pursue here. Two brief comments will have to suffice.
One comment is that objective reasons in our sense are not necessarily external
in Williams’s. At least in some of his later writings on external reasons, Williams
(1995b) allows that the application of “thick ethical concepts” can provide us with
practical reasons. Of course, Williams insists that whether a certain application of
a “thick” concept provides an agent A with a reason essentially depends on whether
A uses the concept (“chastity is an example that focuses the mind”; ibid., 37). The
idea here is that the reason-giving role of “thick” facts turns on the presence in A’s
subjective motivational set of the sorts of dispositions tied up with mastery of the
relevant thick concept. Still, it is significant that this version of internalism is not
committed to the view that only certain kinds of mental states (passions or desires)
can constitute the noninstrumental element of practical reasons. Suppose someone
is tempted to make a remark that, unbeknownst to him, would be a very tactless
remark to make in the circumstances. Williams’s “liberal” internalism allows us to
say that—provided only that tactless is one of the person’s concepts—he has a reason
not to make the remark, even though from his current perspective the remark seems
unobjectionable. In our terminology, this would be an objective reason; but the
reason would still count as internal by Williams’s lights.
Our second comment is that nevertheless, the developmental suggestion we will
be pursuing may be put by saying that we all started life as external reasons theorists.
For the suggestion is that young children are familiar with objective reasons before
even understanding subjective reasons—before they grasp that there are two sorts of
perspectives from which to consider what someone has reason to do. It is not clear,
though, that there is anything in Williams’s account that would be inconsistent
with this suggestion. (The “liberal” internalism Williams espouses in this later writ-
ings does not imply that external reasons statements are meaningless; it merely says
they are false.) In fact our developmental account may shed some light on what is
going on when people make what appear to be external reasons statements. It is
significant that Williams’s most convincing example of an external reasons state-
ment (his gloss on James’s story of Owen Wingrave) turns on a conflict among
members of a family. (Despite Owen’s loathing for the military, “[h]is family might
have expressed themselves by saying that there was a reason for Owen to join the army”;
Williams 1981, 106.) Our developmental suggestion would make it unsurprising if
the family were a context in which the idea of external reasons could sometimes be
seen to linger.

9. Although this is not the way Bartsch and Wellman put things, some of their
suggestions fit well with the hybrid account; they stress the importance of desire
psychologists’ drawing on their own knowledge of the world in predicting how
someone will go about satisfying his or her desires (see Bartsch and Wellman 1995,
155).
Teleology in Children’s Theory of Mind 227

10. We use the term “pro-attitude” in Davidson’s sense, as subsuming “desires,

wantings, urges, promptings, and a great variety of moral views, aesthetic principles,
economic prejudices, social conventions, and public and private goals and values”
(Davidson 1963/1980, 4).

11. Compare Aristotle’s conception of teleological explanation: “the fourth cause is

the goal: i.e., the good” (quoted in Charles 1984, 198). More recently, Csibra and
Gergely (1998) have used the notion of a “teleological stance” in their theory of
infants’ expectations concerning rational actions. Their account differs from ours
(and Aristotle’s) in that it does not appeal to evaluative facts as the source of goals,
but takes the notion of a goal as primitive. Another difference is that their account
is concerned with children’s expectations as to the movement of certain kinds of
agents under certain circumstances, not with children’s ability to explain intentional
actions in terms of agents’ reasons.

12. Of course, there is a (to adults, natural) way of taking these latter propositions
that would make them conflicting without being inconsistent, along the lines of
“there is something to be said for A’s winning, and there is something to be said
for B’s winning.” But construed in this way, the propositions could not do the work
a teleologist expects them to do. To provide teleological explanations, the relevant
evaluative propositions have to license conclusions as to what the agent should do
or has most reason to do; prima facie reasons, on their own, do not license such
conclusions. That there is something to be said for A’s winning, and that A can win
by doing x, only gives A a reason to do x if there are no other more important con-
siderations in favor, e.g., of letting B win. In brief, teleology, as we conceive it here,
has to appeal to “all-out” practical reasons.

13. See Woodward forthcoming for a suggestive discussion of the contrast between
“difference-making” and “causal process” theories of causation. See also Steward
1997 and Hornsby 1993.

14. See Woodward 2003; see also Campbell 2007 for helpful discussion.

15. The case of imitation illustrates that teleology has the resources to conceive of
reasons as “agent-specific.” Although teleology appeals to worldly facts rather than
the individual agent’s mental states, this does not mean that two agents cannot
have reason to perform different actions. One way in which different agents may
be seen to have reasons to do different things is in virtue of their distinctive roles.
Having just banged my toy a couple of times, I, currently performing the part of
the model, have no reason to keep banging; but you, performing the part of the
imitator, have every reason to bang your toy. Again, the very same objective purpose
or value can be seen to yield different sorts of practical reasons, depending on the
agent’s skills and circumstances.

16. Of course, it is not obvious whether imitation is, or when it begins to be, tied
up with explanation. Certainly to begin with, infant imitation may well be a more
228 J. Perner and J. Roessler

primitive phenomenon than the ability to make rational sense of others’ behavior.
All we are saying here is that imitation provides compelling materials for teleological
explanations. Acquiring a conception of intentional action may partly be a matter
of learning to exploit such materials.

17. See Doherty 2009 for helpful discussion of the notion of engagement.

18. For a “grown-up” version of this point, compare John McDowell: “Finding an
action or propositional attitude intelligible, after initial difficulty, may not only
involve managing to articulate for oneself some hitherto merely implicit aspect of
one’s conception of rationality, but actually involve becoming convinced that one’s
conception of rationality needed correcting, so as to make room for this novel way
of being intelligible” (McDowell 1998, 332).

19. See Williamson 2000, and especially Hornsby 2008, on the explanatory role of
knowledge.

20. Davidson’s gloss on “pro-attitude” bears repeating: “desires, wantings, urges,

promptings, and a great variety of moral views, aesthetic principles, economic
prejudices, social conventions, and public and private goals and values” (1963/
1980, 4).
15 Action Theory Meets Embodied Cognition

Fred Adams

1 Introduction

In cognitive science, embodied cognition is sweeping the planet.1 Interest

in this perspective on cognition is becoming wildly popular. Some of the
tenets of embodied approaches to cognition are compatible with the
received theories of action, but some aren’t. In this essay, I will look at the
problem caused for the received view by the claim that much of embodied
cognition is situated and time-pressured in such a way that the received
view of how intentions work in the production of actions in these cases
cannot be correct. I shall defend a compatibility approach, arguing for the
view that even if action is time-pressured and situated, the requirements
of the received view can still be met.

2 Action Theory (AT)

There are theories of action that are not causal theories, but I will not
discuss them here because few (maybe no) sparks fly when combining
them with claims of embodied cognition. Potentially, sparks fly at the
intersection of causal theories of action and embodied theories of cogni-
tion, so in the spirit of generating light, we shall explore there.
Causal theories of bodily action2 maintain that what makes a mere
bodily movement an action is its causal history. This is so regardless of
whether one considers the action the bodily movement (a product of the
right kinds of mental causes) that is brought about in the right way or
considers the action the very causing of the movement by those right kinds
of mental causes. On the product view (Mele 1992), my clenched fist opens
and is an action because I intended to open it and my intention produces
my open fist in the right way. On the causing view (Dretske 1988), my
action is my opening of my fist, where the opening is a causing of the
230 F. Adams

movement by my intention (in the right way). On the product view a

relevant intention is not a constituent of the action, but on the causing
view it is actually a component of the action.
On both views the important causal antecedents of actions are mental
states and their representational (i.e., intentional) content. Not only must
the proper mental states bring about the action in the proper way, they
must represent the act as desirable, not impossible, and as one’s intended
goal for the action to unfold. So there is a constellation of desires (I want
my fist to open), beliefs (I can open it), and intentions (I shall open it
soon/now) in the background and production of the movement that con-
stitutes the action. The representational details of the contents of these
mental states are every bit as important as the details of their causing. The
contents are essential to the determination of what is done. Two qualita-
tively identical bodily movements can be completely different actions if
brought about by different intentions (with qualitatively the same bodily
movements, Tom is scratching his itchy nose, Guido is signaling a mafia
hit).
Among the shortcomings sometimes attributed to some versions of
causal theories of action is that they are incomplete. Elisabeth Pacherie
(2006, 2) puts it this way: “They account at best for how an action is initi-
ated, but not for how it is guided, controlled and monitored until comple-
tion.” Pacherie is surely right. This is why Al Mele and I (Adams and Mele
1989) modeled our accounts of action and intentional action on control
models of goal-directed behavior generally. On our view a piece of behavior
is goal-directed behavior only if it is behavior of a goal-directed system. A
piece of behavior B is directed toward a specific goal G due to an informa-
tion-processing network of control within the system S. S must be causing
B and comparing information about whether B obtains (is about to obtain)
with a representation of goal state G, and be prepared to monitor progress
or make error-corrections as it detects that B is not going to match G. Thus
a system S is a goal-directed system in its behavior B toward a goal-state G
if and only if:

(1) S has an internal representational state R capable of fixing G as S’s

goal-state and S is capable of detecting G’s presence or absence;

(2) Information about S’s ongoing behavior B is fed back into the system
as input and is compared with R;

(3) S’s modification of output behavior B in response to comparison

between S’s present state and S’s goal-state causally depend on the correc-
tion processes of (1) and (2).
Action Theory Meets Embodied Cognition 231

All goal-directed behavior fits this model.3 I (Adams 1986a, 1997) think4
all intentional action fits a similar model that has come to be known as
the “Simple View.” So on my account, S does A intentionally only if: (1)
S intends to A; (2) S does A; and (3) S’s A-ing causally depends on S’s inten-
tion to A (in the right way)—the intention guides and sustains S’s A-ing
via goal-directed feedback control systems. Here is how Mele and I (1989)
described the role of intentions in their functional roles as motivators of
action, initiators of action and sustainers of action.

Intentions are functionally specifiable mental states. The control model identifies
the essence of the functional role of intentions. They must: (i) set the goal or plan
of the action; (ii) be involved in causally initiating and informationally updating
progress toward the action; (iii) provide a standard for determining error and cor-
rection or damage control when the plan goes awry; (iv) provide the criterion for
goal-success (help determine when the intended action has been completed, disen-
gage the plan’s implementation, and so on; (v) play a crucial role in the counter-
factual dependency of output behavior (bodily movement) upon intention and
information input (the perception of present state as compared with one’s intended
goal-state). (Adams and Mele 1989, 514)

Mele and I (Adams and Mele 1992) later extended our discussion to include
trying. We said that tryings were simply intentions at work. They do not
involve special effort. They are not willings. Tryings exist in every case
where an intention to A issues in an intentional A-ing. Trying is the agent’s
contribution to the action. Tryings are not mediators between proximal
intentions (an intention to A here and now) and actions, in the sense that
intentions cause tryings and then tryings cause actions. Instead, trying is
a continuous unfolding of the intention’s doing its work. Intentions are
not ballistic. We put it this way:

For us, tryings are effects of the normal functioning of appropriate intentions.
Roughly, trying to A is an event or process that has A-ing as a goal and is initiated
and (normally) sustained by a pertinent intention. Successful tryings to A, rather
than causing A-ings, are A-ings. . . . On the view to be defended, tryings begin in
the brain. Their initiation is the immediate effect of the formation or acquisition of
a proximal intention. Action begins where trying begins—in the brain. (Adams and
Mele 1992, 326)

3 Embodied Cognition (EC)

There is a great deal of excitement in cognitive science these days over

what has come to be known as embodied cognition. Before we get to specific
claims, the general framework of embodied cognition is motivated by the
232 F. Adams

idea that the mind and cognition are for action and consequently cognitive
processing has its roots and grounding (Pecher and Zwaan 2005) in sensory
(Barsalou 1999) and motor processing (Jeannerod 2006). The sensory and
motor systems are not just input–output systems for cognition, contin-
gently causally connected to cognition, but are constitutive of cognition
(just how varies among the proponents). Hence, all concepts (their content
and how they drive thought and action) can be understood properly only
in relation to their sensory and motor origins. That’s the positive side of
the program. On the negative side is an urging away from theories of
cognition that see concepts as arbitrary abstract symbols understandable
(and functioning) independently of their contingent connections to per-
ception or action (Turing, Fodor, Chomsky).
Much of the excitement is due to new empirical findings that link cog-
nitive activity and behavior to sensory and motor priming, such that one
unexpectedly finds faster cognitive reaction times when subjects are cog-
nitively tasked with experimental paradigms that involve sensory or motor
priming. In most cases, the conclusion drawn (Glenberg and Kaschak 2002)
is that the best explanation for the empirical results is that cognition is
not only grounded but happens in the sensory and motor systems. Some
of the excitement is also due to the claims made that a paradigm shift to
embodied cognition will even help us solve the symbol-grounding problem
(Searle 1980; Harnad 1990).
Among the specific claims being made by proponents of embodied
cognition (Wilson 2002), I will look mainly at those that have the most
direct relevance for models of human action. Take the claim that cognition
is situated. Here is how Wilson unpacks “situated”:

Simply put, situated cognition is cognition that takes place in the context of task-
relevant inputs and outputs. That is, while a cognitive process is being carried out,
perceptual information continues to come in that affects processing, and motor
activity is executed that affects the environment in task-relevant ways. Driving,
holding a conversation, and moving around a room while trying to imagine where
the furniture should go, are all cognitive activities that are situated in this sense.
(Wilson 2002, 4)

Wilson points out that off-line cognition (planning, remembering, day-

dreaming) is not situated, on this understanding, and that this fact has
been pointed out before (Clark and Grush 1999; Grush 1997). Such cogni-
tive activity is also said to be “decoupled” from immediate interaction with
the environment. She also points out that an argument could be made
that:
Action Theory Meets Embodied Cognition 233

situated cognition is nevertheless the bedrock of human cognition due to our evo-
lutionary history . . . invoking a picture of our ancestors relying almost entirely on
situated skills . . . obtaining food, avoiding predators. Thus, situated cognition may
represent our fundamental cognitive architecture, even if this is not always reflected
in the artificial activities of our modern world. (Wilson 2002, 5)

To her credit, Wilson finds such appeals strained. As she notes, it is likely
that even our ancestors engaged in counterfactual reasoning about food
and predators, while hunting and gathering or, while parenting, warning
children about possible dangers to avoid. However, for our purposes, what
matters is not whether all cognition is situated. The clear implication of
the embodied cognition movement is that cognitive abilities arise as situ-
ated and never lose their connection to their sensory and motor origins.
Furthermore, we will be interested mainly in action where the cognitive
states involved in action may be seen as situated. The best case for this is
in ongoing activity—where current cognitive states are causally involved
in guiding and sustaining action. How much more situated can it get? We
will be interested in what embodied cognition tells us about cognitive
states driving situated action.
Closely associated with cognition’s being situated is its being time-
pressured. As Wilson puts it, “agents must deal with the constraints of ‘real
time’ or ‘runtime’” (Wilson 2002, 6). She continues:

These phrases are used to highlight a weakness of traditional artificial intelligence

models, which are generally allowed to build up and manipulate internal representa-
tions of a situation at their leisure. A real creature in a real environment, it is pointed
out, has no such leisure. It must cope with predators, prey, stationary objects, and
terrain, as fast as the situation dishes them out. (Ibid.)

As with whether all cognition is situated, surely not all is time-pressured.

Wilson points out that making sandwiches, paying bills, and working
crossword puzzles don’t fit the bill. Yet, as before, we can take the point
embodied cognitive theorists make as that cognition has this property in
its etiology. In addition, there will be cases where this feature will be
needed to explain cognition’s role in guiding and sustaining human activ-
ity. As Wilson notes, this will include cases from walking, to making skilled
hand movements, driving in traffic, steering a plane in a dogfight, playing
a sport, or just roughhousing with the kids or family pet.
Cognition is for action. This claim hits particularly close to home for
action theory when embodied cognition considers how the visual system
drives purposive behavior. Wilson nicely points out that the traditional
view of perception has been that its cognitive role is to build up
234 F. Adams

consciously accessible perceptual representations for the planning and

guidance of action. However, recent empirical findings of Goodale and
Milner (1992), Jeannerod (1997, 2006), and others threaten to unravel at
least some of the traditional view because certain kinds of visual inputs
prime for motor activity (Borghi 2005) types of grip (precision/power grip
size, distance between index finger and thumb). And the activation of some
perceptual and cognitive states fire motor neurons directly (even in the
absence of task demands for the perceiver) (Gallese et al. 1996, Rizzolati
et al. 1996). So the suggestion is that some connections between percep-
tions and motor activity are quite direct (literally “for” one another), going
in both directions.5 What is more, this can happen independently of con-
scious perceptual awareness by the subject. Wilson points out here (as
elsewhere) that the claims are far from holding across the board, however.
There is still, as she calls it, perception for perception’s sake. Our visual
processing, at least in the ventral stream, seems more flexible than this
overall. Yet there is one more aspect of embodied cognition that is espe-
cially relevant to our attention here. Though normally ventral stream
visual processing (sometimes called the “what” system generating a con-
scious visual field) and dorsal stream processing (now seen as a “how”
processing system that guides actions such as reaching and grasping and
that can guide behavior without a consciously accessible visual field) work
together, they can come apart.6 Indeed, as we will soon see, some of the
most interesting cases for the meeting of action theory and embodied
cognition are cases where purposive bodily movement is driven by dorsal
stream visual processing and the motor system at several stages of remove
(or dissociated) from ventral stream processing. How are these cases of
behavior to be described, explained, and evaluated as actions?
It is mainly these aspects of embodied cognition that will occupy us
here. Elsewhere (Adams and Aizawa 2008) I’ve dealt with some of the
others including that cognitive systems off-load cognitive work to the
environment, that the environment actually becomes a part of the cogni-
tive system (the thesis of extended mind), and the claim that off-line
cognition is body based (Adams manuscript).

4 Sparks, or Where AT Meets EC

At the intersection of action theory and embodied cognition is an impor-

tant worry over the nature and role that representations (particularly
intentions) are supposed to play in turning a bodily movement into an
action. As described on the Adams–Mele control model of the role of inten-
Action Theory Meets Embodied Cognition 235

tions in actions (and tryings), intentions have a complex set of cognitive

functions and they are causal antecedents of actions. The worry is that
some of the constraints of embodiment placed on causal antecedents of
bodily movement will turn out to disqualify those causal antecedents as
intentions, at least as intentions traditionally conceived of as typical kinds
of representations.7
Let’s consider two features of embodied cognition that we have dis-
cussed, that cognitive states are situated and time pressured. Let’s set aside
whether all cognitive states are like this (presumably they are not). We are
interested primarily in intentions. Surely not all intentions have these
features. Pacherie (2006, 2008) nicely distinguishes three kinds of inten-
tions with three kinds of intentional contents: future intentions (I will
retire some day), proximal intentions (I will now type the letter L), and
motor-intentions (and here is where it gets tricky . . . what is the content
of a motor-intention?). Motor-intentions are sent in the here and now
from various motor regions telling the body to move (specific muscles to
contract or movement to be made in specific directions). What is the
content of such a motor-intention? It doesn’t represent a future state or
movement, because, being an activated motor state, it is happening now.
It is not a proximal intention for the very reasons imposed by the con-
straints of being situated and time pressured and updated in the flow of
information being fed back during the process of moving. I’ll explain more
why this is so shortly. Does its content even involve concepts? What is its
representational mood (indicative or imperative)? How can it have a mood,
if it is not conceptual? Imperatives are like commands—do this . . . do
that—where there are concepts of doing the “this”s and “that”s. How can
it have a representational content if the thing it represents has not even
yet happened, but is not so far ahead in the future that there is even room
or distance between cause and effect? Normally effects (footprints in the
snow) represent causes (feet making the prints). But feet don’t normally
represent the footprints they are about to make or are making. Motor-
intentions, if they are representations, represent states of affairs they are
involved in the act of bringing about at that very instant. So how can
motor-intentions simultaneously fill the cognitive role they are supposed
to fill on the control theory of action and satisfy the constraints of embod-
ied cognitive states that are situated and time constrained? That’s the
question.8
Why do we need to appeal to these motor-intentions? There are several
kinds of reasons, such as the result of experimental research on differences
between dorsal stream visual processing and ventral stream processing. But
236 F. Adams

others are more familiar from everyday experience with situated and time-
constrained action. In the shower, I accidentally dislodge the soap from
the soap tray. In a blink of an eye I bump it up against the wall of the
shower, visually and tactileley track its trajectory, and catch it before it
hits the shower floor. Catching it was lucky, but not accidental or unin-
tentional. If it was purposive behavior, then action theory (AT, along the
lines of Adams and Mele 1992) is committed to there being intentional
and representational mental states active in the motivation, initiation,
production, guiding, sustaining, and terminating of this bodily behavior.
Mark Rowlands (2006, 102–104) uses the term “deed” to designate activ-
ity that is both situated and time compressed. Two of his examples of deeds
include catching a cricket ball, where “typically you will have less than
half a second before the ball, which may be traveling in excess of 100 mph,
reaches you.” Another of his examples is playing Chopin’s Fantasie
Impromptu in C# Minor, where “your fingers have to traverse the keys in
the sort of bewildering display necessary to successfully negotiate this
notoriously difficult piece.” Rowlands coins the term “deed” to cover these
because he maintains that “they do not fit the strict conception of action.”
Specifically, he maintains that the direct antecedents of these deeds are
not intentional or representational states, and he maintains that a general
antecedent intention (to catch the ball or to play Chopin) is not sufficient
for the relevant doings to be individuated properly as actions. What makes
an action the action that it is cannot, thus, be the intentional content of
a prior intention, because any number of deeds (online corrections) can
be involved in satisfying an antecedent intention. He appeals to both
phenomenology (we are not consciously aware of all the moves we make
while carrying out these deeds) and science (research on dorsal vs. ventral
stream processing supports the first point) to support his claim that the
direct antecedents of deeds are not representational or intentional in the
normal sense.
One might think, as I do, that motor-intentions can come to the rescue
in just such cases. Rowlands does not, and he develops a view of deeds
that gives these movements themselves the representational properties
that the received view of action reserves for the intentional states that
cause the bodily movements. Here is what Rowlands says about his rejec-
tion of motor-intentions:

We might explain the status of a deed in terms of its connection to a general ante-
cedent intention, but then individuate the deed in terms of the motor representa-
tion that causally produces it. However, if this line is to be convincing, two questions
Action Theory Meets Embodied Cognition 237

need to be addressed. First, do motor representations possess content of a sort suf-

ficient to individuate deeds? Second, if so, what is the source or basis of this content?
In answering the second question, we will be required to tell a certain story about
how motor representations come to satisfy the essential conditions of, or constraints
on, representation: how they carry information about the environment, how they
have the function of tracking environmental items, or enabling the organism to
achieve some task in virtue of tracking such features, how they are capable of mis-
representation, and so on. (Rowlands 2006, 109)

Rowlands believes that the case for motor-intentions is no more compel-

ling than the theory he develops on which it is the deeds themselves that
satisfy these requirements for being representations.9 In what follows, I
want to try to explain how to answer the questions Rowlands asks. To do
so, I will begin by reviewing a taxonomy of intentions offered by Elisabeth
Pacherie.
Here are some of the ways Pacherie (2006) characterizes the role of
motor-intentions10 (M-intentions) versus present or proximal intentions
(P-intentions): “P-intentions and M-intentions are both responsible for the
online control of action, but whereas the time scale at which the former
operate is the time scale of the consciously experienced present, the latter’s
time scale is that of the neurologically micro-present, which only partially
overlaps the conscious present” (Pacherie 2006, 7).
Pacherie (ibid., 8) isolates three important characteristics of motor-
intentions. These are normally associated with dorsal stream processing
(vision-for-action system), which extracts visual information about proper-
ties of objects (situations) relevant to action and uses this information to
build or elicit motor representations used in effecting rapid visuomotor
transformations. First, objects and their properties are represented in a
format that is relevant for immediate selection of motor routines. When I
caught the falling bar of soap, my visual system represented the soap’s
spatial position in terms of the movements I needed to make to catch it
(my reach, direction, grip size, and shape). Second, the representations of
the necessary movements reflect implicit or tacit knowledge of the biome-
chanical constraints and the kinematics and dynamics of movement (none
of it consciously accessed). Third, the motor representation normally codes
for cascades of movements, where the goal of the action determines the
global organization of the motor sequence. So, for example, the goal of a
movement will partially determine the grip size and type of motor repre-
sentation. One might catch a falling pen with a precision grip, if one were
planning to use it to write . . . or with a power grip, if one were planning
to use it as a weapon.
238 F. Adams

Echoing the worries of Rowlands, in a recent paper Shaun Gallagher

(2008) forcefully challenges AT from the point of EC on whether there
could be representations that fit the bill to both be representations with
the appropriate intentional content and to be explanatorily useful. Being
a representation-skeptic for what he calls “minimal representations,” Gal-
lagher thinks the AT model breaks down when it meets these situated,
time-pressured types of cases. Gallagher is not totally a representation
skeptic. When it comes to future (F-intentions) and nonsituated reasoning
and planning, he is prepared to embrace them. But he would prefer a
dynamic systems model free of representations entirely for explaining the
types of actions minimal intentions are supposed to explain.

One important problem is that a majority of cognitive scientists continue to use the
R-word and do so in ways that are often not clear. In the case of action it is nothing
more than a handy, but often confused and misleading term, a bad piece of heuris-
tics, an awkward place-holder for an explanation that needs to be cast in dynamical
terms of an embodied, environmentally embedded, and enactive model. . . . It may
take more energy to define and distinguish any legitimate sense of representation
. . . than it would to explain the phenomenon in non-representationalist terms. And
if one can explain the phenomenon in non-representationalist terms, then the
concept of representation is at best redundant. (Gallagher 2008, 365–366)

I think the dynamic systems view cannot do without something that plays
the role of representations, and I will explain why. I will also do my best
to give an account of how motor-intentions get their representational
content and what that content might be. For now, let’s see why Gallagher
thinks that no account of representation is likely to fit the bill for the kinds
of situated, time-pressure actions we are considering.
To be sure we are talking about the same thing, paraphrasing Michael
Wheeler (2005, 197) on “action-oriented representations” (AORs), here is
what Gallagher says about these “minimal representations” (what we are
calling motor-intentions). They are “temporary egocentric motor maps of
the environment . . . fully determined by the situation-specific action
required” (Gallagher 2008, 353). They don’t represent a preexisting world
via a type of internal image or neural pattern (a visual field from ventral
stream processing, say). Instead, “how the world is is itself encoded in
terms of possibilities for action” (ibid.). Sound familiar? This is very similar
to the remarks of Rowlands, and involves the situated, time-pressured type
of situation of EC applied to action-oriented representations. Gallagher
adds that according to Wheeler, what is represented (in an AOR) is not
knowledge that the environment is x, but knowledge of how to negotiate
the environment. “AORs are action-specific, egocentric relative to the
Action Theory Meets Embodied Cognition 239

agent, and context-dependent” (ibid., 354). So, we are indeed talking about
the same thing as Pacherie’s M-intentions.
Now let’s consider Gallagher’s reasons for thinking that AORs or
M-intentions cannot be representations. The first is decoupleability.11 The
argument is that representations are necessarily decoupleable and AORs
aren’t, so AORs can’t be representations:

According to . . . the standard definition . . . of representation, representation is

decoupleable from x (I can represent x even if x is absent from the immediate envi-
ronment). But once we do decouple . . . I suggest we are no longer talking about
action in the same sense. Indeed, it is difficult to see how . . . acts can be decoupled
from x . . . or the context without becoming something entirely different from an
element of the action at stake, or an AOR. Off-line cognition, imaging, remember-
ing, or even re-enacting an action decoupled from its original context and absent x
may (or may not) require representation—but this says nothing about representa-
tion in action. (Gallagher 2008, 357)

Generally, it is true that if X represents Y, X and Y can come apart (decou-

ple). I can take a picture of you and later look at it when you are not
around. It still represents you. I can think about you, when you are not
around. I can long for you when you are not around . . . and so on. Clear
enough. First, is this an essential property of all representations? And
second, is it true that AORs (M-intentions) cannot decouple?
Consider sensory states. If I touch you, you’ve got to be there. If I see
you, you’ve got to be there. If I hear you, you’ve got to be there (somewhere
causing my hearing). These token sensory states, when veridical, are
coupled. Now one can be in qualitatively identical states and be uncou-
pled, as when one is hallucinating. When hallucinating, I only seem to be
touching you (you are not really there but it feels exactly as if you were).
I only seem to see you. I only seem to hear you. So, in the veridical cases,
the coupling must be there. In the nonveridical cases, things may become
decoupled. No one denies that such decoupled experiential states are rep-
resentations. I don’t think even Gallagher does. So if things are similar
with motor-intentions (AORs), then they should count as representations
as much as sensory states.
Searle (1983) reports cases performed by Penfield where the motor
system is decoupled from the muscles. The subject does what he normally
does when he makes a fist (but is unable to see if his hand closes and his
arm is anesthetized . . . so he gets no proprioceptive feedback about his
hand). Yet to some subjects it feels for all the world as though they did
indeed close their hand (make a fist). It feels this way because they get
efferent copy feedback. The motor system tells the brain that it sent the signal
240 F. Adams

that normally closes the hand. Signals received generate an actish experi-
ential state in the subject. We might call it a hallucination of acting (not
unlike a hallucination of you that I might have as a sensory state, when
you are not actually there). If they don’t actually cause a bodily movement,
does this mean that these are not genuine representations? I don’t see why
that would be true of motor-intentions any more than it would be true of
a hallucination as of you, if you were not there. Hallucinations are repre-
sentations, if anything is.
True, one would still have to explain where these AORs (M-intentions)
get their representational contents and what those contents are, if they are
representations. But the fact that they can decouple doesn’t seem to mean
they can’t be representations any more than the fact that I can suffer hal-
lucinations of you means that when I actually see you, I am not in a state
that represents you. Gallagher12 is right that off-line representations are
different from online ones. One type is veridical (seeing you) or a genuine
trying (I’m trying to catch the soap) precisely because it is online. Yet, this
does not mean that the same types of representations cannot in principle
decouple. Indeed, they can even still be tryings. When Penfield told his
subject to try to close his fist (while his efferent path to the muscles was
blocked) it would be false to say the subject didn’t try because the subject
didn’t succeed. He did try. His brain sent the signal. Penfield just blocked
it. The subject’s motor-intention was decoupled, but it still was sent “to
the hand” and was telling the hand “to make a fist.”
Today one can perform the same type of experiment with subjects using
transcranial magnetic stimulation (TMS) to stop the motor signals from
reaching the muscles. Subjects may be tapping with their fingers (to the
beat of a sound they are shadowing). When TMS interrupts their tapping,
the subjects report that they were trying to continue tapping but were
unable to produce the movement. This is a type of decoupling of motor
command from implementing a movement in the muscles. It is not exactly
“off-line.” The subjects are not just “imagining” making the movement.
They were making them prior to the TMS activation and they continue
making them after the TMS activation is turned off, and the subjects didn’t
do anything different in between (from their perspective).13
Okay, what’s next? Gallagher mentions two other features14 of represen-
tation that he thinks motor-intentions (or AORs) won’t have. Discussing
why he thinks a dynamic system approach is preferable, he says: “Nothing
in this dynamically dissipating process amounts to a representation, if we
take representation to involve: an internal image or symbol or sign, a
Action Theory Meets Embodied Cognition 241

discrete duration, and decoupleability” (Gallagher 2008, 362). We have

discussed decoupleability, so it is on to the other two. Take discrete dura-
tion first. Gallagher doesn’t really argue for this. At most he adds to the
claim that they are not discrete that “it is more like a temporal, dynamic,
and distributed process” (ibid., 364). He also does not explicitly argue that
these are not internal images or symbols or signs. At most he adds that
they extend “to include embodied environmental aspects . . . only ‘weakly’
neuronal” (ibid.). These are both argued for by Wheeler (2005) and
Rowlands (2006) and accepted (but not argued for) by Gallagher.
So what are these two objections to motor-intentions being representa-
tions all about? They are about the rejection of a language of thought (LOT)
model of cognition that EC is campaigning against. On the LOT model (or
any naturalized15 model of semantics), representations are thoughts, beliefs,
desires, or future intentions. They are symbolic and language-like and
propositional. They have a compositional syntax and semantics. They have
discrete symbols that serve as the vehicles for their representational
content. The vehicles might be different though their contents are the
same (Frege puzzles).
For motor-intentions, Gallagher is suggesting that the LOT model
cannot be right because the representations are situated and time-pressured
and they are in the motor system. LOT does not portray the basic building
blocks of thoughts, namely, concepts, as being in the motor system. So
Gallagher is thinking of motor-intentions as pulses of electricity not unlike
those the coil (or now computer) sends to the spark plugs to fire. No LOT
there. No language-like representations there. No discrete symbols there.
If the motor system works like this, then the LOT model does not apply
to motor-intentions.
Gallagher agrees with Berthoz and Petit (2006) that the brain is an
“organ for action rather than an organ for representation. They want to
move away from a philosophy that puts language in first place and that
models action on language-like representation, where action is equivalent
to movement plus representation” (Gallagher 2008, 353). He also quotes
Berthoz and Petit (2006, 23), saying “By applying this representational
filter, everything in the external as well as the internal world appears
frozen, fixed and stabilized by the projection of the propositional
form, which implicitly structures representation.” Gallagher agrees with
Wheeler (2005) that “representations on the propositional model are
discrete static structures, states of affairs that lack the kind of dynamics
found in action. . . . what takes the place of representations in non-
242 F. Adams

representationalist accounts of action is a form of perceptually based

online intelligence which generates action through complex causal inter-
actions in an extended-body-environment system” (Wheeler 2005, 193).
In reply to all of this, first notice that these are attacks on thoughts,
beliefs, desires, and other such cognitive representations that on many
views are likely to have symbolic elements with language-like structure.
To lump these together with motor-intentions is to assume that motor-
intentions have the same kinds of representational vehicles and represen-
tational contents as these other cognitive states. There is good reason to
think that they don’t. They are coming out of the motor system—not a
central cognitive system. When in the dorsal stream, they are not even
consciously accessible. And to some extent, motor-intentions are informa-
tionally encapsulated and not subject to the usual cognitive penetrability
of rationality constraints (Bratman 1987).
There is little reason to think that they generate intensional contexts
(Frege puzzles). If my motor-intention is trying to catch the bar of soap I
dislodged from the shower tray and the bar of soap is the oldest bar of
soap in the house, then my motor-intention is in the act of catching the
oldest bar of soap in the house.16 Very likely there are no differences in
motor vehicles such that two different vehicles could share the same motor
content and thereby generate intensional contexts (Frege puzzles).
They may even be nonconceptual. Jeannerod (2006) puts matters this
way:

The visual percept . . . has a rich informational content about the object, but it has
no conceptual content: it remains non-conscious and is ignored by the perceiver.
If visual processing were to stop at this stage, as may occur in pathological condi-
tions (Jacob and Jeannerod, 2003), the object could not be categorized, recognized
or named. It is only at the later stage of the processing that conceptualization occurs.
The representation of a goal-directed action operates the other way around. The
conceptual content, when it exists (i.e. when an explicit desire to perform an action
is formed), is present first. Then, at the time of execution, a different mechanism
comes into play where the representation loses its explicit character and runs auto-
matically to reach the desired goal. (Jeannerod 2006, 4–5)

So, if it is true that motor-intentions do not have conceptual content, it is

probably true that motor-intentions are neither language-like, nor image-
like, nor syntactic. This does not disqualify them from being propositional.
Information can have propositional content and be coded in analog form
(Dretske 1981). Whether the representations are distributed is relevant to
their implementation level in the neurons, but it may not be relevant to
their content one level up in description. My thought that today is Sunday
Action Theory Meets Embodied Cognition 243

has a discrete and propositional content. How it is realized or implemented

in the brain is another story.
With respect to the motor system, as populations of neurons fire and
sum together, a vector of electrical potential forms. Some of these vector
summations become tuned to control specific muscles; some to control for
specific directions (directional tuning) and configurations of hands or grip.
An example of the latter is that you can write your name (however badly)
with the pen in normal position, in your fist, in your opposite hand . . .
maybe even between your toes or with the pen in your mouth. So for some
vector summation there is commonality of direction and force, but no
unique signal dedicated to specific muscles. Hence, some motor cells are
representing types of movement and direction independently of which
muscles are used to achieve this. This level of abstraction is an element of
representation. The activation can also be contextual (fire in some sequences
of activity, but not in others). This shows that there is something represen-
tational going on (not just a pulse being sent to a muscle as from the coil
to the spark plugs). Therefore, that fact that there is no single discrete
symbolic structure that implements a motor-intention does not disqualify
motor-intentions from being representations.
Jeannerod (1985) has long noted that an act as simple as taking a book
from a shelf is extremely complex and involves complex sensorimotor
involvement at what he calls the “operational” level:

This consideration thus joins those concerning the distributed nature of the opera-
tional level. One can easily imagine that this level consists of different sensorimotor
channels, each independent, each characterized by the type of sensory information
it is capable of processing and by the type of movement it is capable of producing.
The sensorimotor channel which processes “book thickness” information produces
a movement that is adapted to this information, namely, contraction of the muscles
which form a pincer grasp; the channel which processes the “distance with respect
to the body” information produces movements which extend the arm, etc. In order
to produce the correct movement to grasp a book, the different channels involved
must be activated simultaneously. Simple observation reveals, in fact, the pincer
grasp is forming at the same time as the arm is being extended, and that the fingers
do not wait until they are close to the book before adopting the correct posture.
The channels therefore function in a parallel fashion . . . because of their respective
independence—an advantageous arrangement when a large quantity of information
must be processed in a minimum amount of time, as is the case here. (Jeannerod
1985, 125–126)

Hence, a complex motor-intention (active in picking a book from a

shelve) may be compositional in content, analog, distributed across several
244 F. Adams

sensorimotor channels, and still satisfy the general nature of being a rep-
resentation. I will now say more about why these motor-intentions are
representations.
Earlier I said that the dynamic systems account that Gallagher prefers
probably cannot succeed without representations. When I’m trying to
catch the soap, something is still directing my reach for the soap, and this
cannot simply be due to the causal influence of the soap itself and its
structural properties impacting my sensory systems. Why not? Because
what I do and how I shape my hand is dependent on my larger background
goals (don’t waste time in the shower, don’t knock things from their proper
places without trying to catch them) and so on. And as in reaching for a
book, my goals are guiding the direction of my reach and the nature of
my grip. The fact that these are not random but are being purposively
coordinated by my motor system tells us that the elements for repre-
sentation are there—nomic coordination and direction of a purposive
movement.
There are two more matters to address, if motor-intentions are indeed
representations. The first (earlier raised by Rowlands) is the matter of from
where they acquire their representational content. The second is the format
of their representational content. Let’s consider the first matter. Motor-
intentions function largely in service of proximal (P-intentions) or more
distal future intentions (F-intentions). As noted earlier, my future and
proximal intentions are normally conscious and constrained by consider-
ations of rationality. As I plan to interact in the world and leave my mark,
I consider all the possibilities of interacting with various goal objects and
the array of goal states that I may desire to bring about. With a range of
possibilities identified, I must go from the future intention (or what some
call a distal intention, D-intention) to the more proximal intention, and
then this must all be translated into the motor-intentions. Pacherie (2008)
describes beautifully the cascade of content and dependency between these
various intentions:

One can . . . distinguish two moments in the dynamics of M-intentions. The

upstream dynamics lead to the selection of one among the typically several pre-
potentialized motor programs. When an M-intention is governed by a P-intention
and inherits its goal from it, the presence of the goal tends to increase the salience
of one of these possible pragmatic organizations of the situation and thus allow for
the selection of the corresponding motor program. Forming a P-intention to act on
an object, say reach for a pen, typically involves focusing one’s attention on the
object that is to be the target of the action. As Campbell (2002) points out, we need
to explain how the motor system manages to connect to the very object identified
Action Theory Meets Embodied Cognition 245

on the basis of perceptual experience. Campbell proposes that “conscious attention

to the object will include some awareness of the location of the object, and that the
target for processing by the visuo-motor system can be identified as ‘the object at
that location’” (2002, p. 55). In other words, common location is the binding prin-
ciple for perception and action, for objects represented at the level of P-intentions
and objects represented at the level of M-intentions. The role of conscious attention
is to define a target for the visuo-motor system. Once the target is defined, it is the
job of the visuo-motor system to set the parameters for action on the object. . . .
The guidance and monitoring functions of M-intentions are exercised as part of the
downstream dynamics. They are responsible for setting the precise parameters of
motor commands and for fine motor adjustments and rapid corrections during
execution. (Pacherie 2008, 186–187)

In this Pacherie agrees with Jeannerod (1997, 2006) and MacKay (1981)
that there is a hierarchy of motor representations such that the goals and
parameters of acts coded for at the higher levels act as constraints on the
lower-level representations.17 The lower-level representations that drive the
movement represent both the body in motion (a generator of forces) and
a goal of the action encoded in a “pragmatic” mode, distinct from the
“semantic” mode. For one, she cites his and other studies that suggest the
“amount” of force needed for the movement is encoded. (On the subjective
side there is a “sensation of effort” that accompanies this representation.)
The representation of a goal represents both an object and a final state.
Jeannerod suggests that these representations “fall between” a sensory
function and a motor function. Pragmatic representations activate prede-
termined motor functions related to the specific objects and their affor-
dances (Gibson 1979). Motor representations are relational, representing
neither states of the body nor states of the environment, but states of the
relations between the two: motor patterns that objects afford the agent.
Pacherie adds that motor-intentions are at least partially modularly encap-
sulated and only moderately cognitively penetrable. However, some cogni-
tive penetration is possible; how an object is grasped is not just a function
of its size, but what we intend to do with it. So the systems and levels of
intention have some degree of cross-talk. And the environment “affords”
much more than the motor system responds to. So the response is deter-
mined by our other cognitive states (and thus there are limits to the
encapsulation of M-intentions).
Subjects are not even always aware of what bodily movements their
motor-intentions are producing. The role of F-intentions (or distal, D-inten-
tions) is not lost, once an action begins. There must still be a causal depen-
dency on the goal state represented in that D-intention. Pacherie puts it
246 F. Adams

this way: “the relation between the three levels is not one of mere co-
existence. They form an intentional cascade,18 with D-intentions causally
generating P-intentions and P-intentions causally generating in turn
M-intentions” (Pacherie 2008, 188).
M-intentions inherit their representational content from distal and
proximal intentions. As the agent forms goals (whether long-term or short-
term) and plans how to achieve these goals, the other levels of intentions
chain together and form an intentional cascade and a dependency of
content working its way up and down the cascade. This helps explain why
motor-intentions are context sensitive (e.g., changing grip posture to a
precision grip or a power grip in time-pressured reaching depends on the
goal of the use of the object being gripped). Hence there is a mechanism
of dependency and inheritance of content that runs through the three
types of intentions described by Pacherie.19
Now let’s consider further the format of the representations that are
motor-intentions. Consider the following types of acts. One might pick
up: a needle, a pencil, a baby shoe, a bronzed baby shoe, a book, a copy
of the OED, a bar of gold. Additionally, one might catch: a feather, a ping
pong ball, a golf ball, a baseball, a bowling ball, an egg, a water balloon.
fMRI studies show two things. If one is imagining doing these acts, the
motor cortex is activated. What is more, just reading about them activates
the sensorimotor areas that would be activated if one were to perform the
acts. There is a first-person phenomenology of “what it’s like” to do these
acts that one can almost feel in just reading about or imagining them. And
this neural imagining is just the sort of activity in the brain that one can
use in brain–machine interfaces to learn to manipulate external robot arms
and computer devices just by imagining acting.
Earlier I talked about “directional tuning” and other sorts of fine-tuning
in the cortex that allow the brain to form correlations between firings of
populations of neurons and bodily movements (whether muscle-specific
or direction-specific types of movements). These types of movements are
not innate. Yes, there are innate20 types of motor routines, such as infant
tongue protrusion and the ability to imitate shape of mouth (Meltzoff and
Moore 1977), but these types of action listed above aren’t some of them.
These require a history of fine-tuning. That is, the agent has to make these
movements, witness their outcomes, and then select for doing them again
in order to set up a repertoire of abilities to make them. The movement
types are learned, skilled.
What, then, is the format of the content of a motor signal that is sent?
It is in the form of an imperative (Mandik 2005). It is in the form of a type
Action Theory Meets Embodied Cognition 247

of movement or type of direction of movement or sequence of types

chained together (for complex movements).21 The signals represent move-
ment types: precision grip, power grip, move left, now up, now right, and
so on.
Isn’t this just the sort of thing that worried Gallagher? How can these
be representations that are situated, time pressured, and coupled? These
require a kind of representational particularity that falls short when a
representation is run off-line or decoupled. The answer is that when it
comes to particularity of representational content, this particularity is
secured by causal connections. Consider perception. What makes my per-
ception a perception of you (not your identical twin) is that I’m looking at
you (not your twin). If you two look indistinguishably alike to me, it is not
the qualitative content that secures particularity of my visual percept; it is
causal origin that secures the particularity of content (Stampe 1977).
The same is true in the motor system. A signal that is sent to my index
finger to contract might just as easily make my ring finger contract, if the
signal were diverted to that finger. But what makes it a signal specifically
directed to my index finger is that this is how the motor is wired to my
body, to my hand. What makes it a signal to move my body to the right
or left, open or close a hand or move a finger is that the signals are sent
from an egocentric reference frame of my body and its connection to my
brain. One’s brain exploits the connections with which it comes packaged.
The background conditions (channel conditions of the information flow)
secure some of the conditions of the content that does not need to be sent
in the signal itself (Searle 1983; Dretske 1981). So Gallagher’s worry about
particularity, about how this motor-intention can be about this token
action, is solved. It is about this particular action (token) in virtue of being
causally connected to this body—the body to which it has become finely
tuned and to which it is causally connected.
There is, as Searle (1983) puts it, a world-to-mind direction of fit in the
representations of motor-intentions. They are representations of types of
movements and have an imperative force (“make x happen!”). They can
be compared with sensory information to monitor the degree of goal-
success or failure. Fed back through the sensory system will be representa-
tions of whether the particular movements are moving toward achieving
one’s goal (or have succeeded or have missed the mark and require a con-
nection). And all of this will be a bit faster than the speed of thought in
many cases.
Equally important, in order to fine-tune the motor system, one’s brain
must already be able to make the various movements. Then information
248 F. Adams

is available via the sensory system about the type of movement the
motor signal produced. So there are two systems of representation with
different directions of fit (mind–world, world–mind) harnessed and
working closely together—and both serving the purposes of the goal-
directed system. Hence, motor-intentions (and their motor signals) become
recruited by the brain for the types of movements they are able to
make, based on past experience and “fine-tuning.” As Mandik (2005) says,
instead of being selected for their ability to indicate the way the world is
(Dretske 1988) on the sensory side, these representations are being selected
for the ability to perform various types of manipulations on the body and
world.

5 Conclusion

As I set out to show, there is no incompatibility between the control-

theoretic model of action and the fact that many actions are situated and
time pressured—the fact that cognition is for action. The dependency
between different types of intentions and their contents on the higher–
level, consciously held goals of the agent and what Pacherie calls the
“intentional cascade” explains the relation between future and proximal
intentions and motor-intentions. Motor-intentions themselves are types of
representation, and I have tried to map out the format of their content.
Thus, I have provided a response to skeptics such as Gallagher and Row-
lands who believe that motor-intentions are not representations and that
representations are not needed in explaining the intentionality of “deeds”
or situated and time-pressured purposive actions. Hence, on these matters
there is compatibility between the claims of AT and EC. On other matters,
I think there does turn out to be some incompatibility, but those matters
will have to await another occasion for their airing.

Acknowledgments

Thanks to those who read and commented or corresponded helpfully on

this essay. This is especially true of Ken Aizawa, Santiago Amaya, Steve
Beighley, David Chan, Shaun Gallagher, Al Mele, Pete Mandik, Elisabeth
Pacherie, Mark Rowlands, and Rob Wilson. I would also like to thank
Richard Ivry for podcasting his Cognitive Science C127 on i-Tunes, where
I learned some of the facts mentioned here about fine-tuning and the
motor system.
Action Theory Meets Embodied Cognition 249

Notes

1. Ken Aizawa says I’m understating the current influence of the movement.

2. Mental actions are interesting in their own right, but I won’t discuss them here.

3. Mele and I (Adams and Mele 1989) realized at the time that there were many
types of feedback that would be involved in these processes (Adams 1986b). Today
Grush’s (2004) work on emulators has done an excellent job of bringing things up
to date. Our view does not require that things do not go “ballistic” at some point.
Indeed, probably all bodily actions become ballistic at some point, but to be goal-
directed they must be products of systems that have these goal-directed mechanisms
and actions must originate in such systems.

4. Mele (1992) and I part company here, though his view, the “single phenomenon”
view, is mainly a departure due to Bratman’s (1987) attack on the Simple View. I
have elsewhere explained why I don’t find Bratman’s attack persuasive, and Mele
and I have skirmished in various journals over related points from time to time, but
our views remain very close and we both accept the control model overall. For these
differences with Mele, see Adams 1997, 1994a,b.

5. Aren’t cases of “automaticity” of movement a problem for my view? I don’t think

so. There are actions and there is mere bodily movement. Cases of mere bodily
movement can be triggered in the brain stem of hydraencephalic infants; another
example is mere involuntary eye-blinks in normal human beings. These are mere
behavior—things my body does, but not things I do. As in the view outlined for
AT, actions will have intentional states with representational contents in their etiol-
ogy. In fact, it is the worry that without this even involuntary movements could be
cast as actions on a view like that of Rowlands (2006) (which puts the representa-
tional content in the movement itself) that keeps me from being persuaded by his
view. More about this below.

6. Of course, though they can come apart, Goodale (2004, 1170) also says that “both
systems are required for purposive behavior—one system to select the goal object
from the visual array, the other to carry out the required computations for goal-
directed action.” It is in the latter instance that the motor system is involved and
in which I will argue below that representations are involved.

7. David Chan (1995) introduces a new category of “nonintentional” actions for

such cases. He does so because he thinks intentions have to meet rationality con-
straints, like those introduced by Bratman (1987). This means that for him manner-
isms or habitual actions don’t qualify as intentional actions, and we need a new
category. I (Adams 1997, 2007) don’t think all intentions do need to meet these
“rationality” constraints, though most do. But “one-off” intentions, irrational inten-
tions, or motor-intentions do not. As I’ll explain below, motor-intentions can satisfy
250 F. Adams

the properties that Chan introduces to differentiate nonintentional actions from

mere movements, viz., foreknowledge and an actish feel. So I don’t think we need
the new category. These are all tryings of various kinds, and tryings are
intentional.

8. I’ll say more below about how future intentions, proximal intentions, and motor-
intentions are related on Pacherie’s views and will try to answer this question about
the content of motor-intentions. I’ll also say why Searle’s “intentions in action,”
though relevant, cannot give us the content of motor-intentions.

9. I won’t here attempt to counter the theory Rowlands develops. Let me just say
that if I can answer the questions he asks about motor representations, then I will
have undercut some of the motivation for replacing the received view with this new
view of “deeds.”

10. The feature of being beneath the level of consciousness means that motor-
intentions are not equivalent to Searle’s (1983) “intentions in action.” For Searle
says a person correctly expresses the content of his intentions in action to raise his
arm “quite precisely when he says ‘I am raising my arm.’” (Searle 1983, 106ff). He
adds that “if one wants to carve off the intentional content from its satisfaction he
can say, ‘I am trying to raise my arm’” (ibid., 107). All of this clearly takes place at
the conscious level and is a type of processing that one may find in the ventral
visual stream (not dorsal).

11. Gallagher (2008, 358) notes that Wheeler (2005, 219) gives up that AORs must
be decoupleable, but Gallagher continues to press the point.

12. Gallagher considers but rejects that emulators (Clark and Grush 1999) might be
the way to go. He rejects this because he thinks that once the representation and
what it represents come apart, the game is over. The representation cannot then be
guiding the action: “it ceases to be part of a forward motor control mechanism”
(Gallagher 2008, 358). Actually this is not so. That is the beauty of emulators. They
are part of the forward control. They anticipate what is happening before it does,
so the system does not have to delay for feedback from the muscles and world. But
even if Gallagher were right, the same is true for sensory states. Once you aren’t
there I’m not seeing you. The question is whether these representation types can
decouple (in principle). It is not whether they can decouple while doing their cogni-
tive work guiding an action or directly perceiving the world. No one thinks that for
veridical perceptual states. So why hold motor-intentions to those standards? Curi-
ously, Gallagher seems to agree that decoupled, emulator states and others may be
representations when not guiding action. But he questions why they must be rep-
resentations when coupled and guiding action. However, I think it would be far
more curious if a representational state lost its representational status just because
it became coupled to and began guiding and sustaining a purposive bodily
movement.
Action Theory Meets Embodied Cognition 251

13. There is a two-second window where what is to be sent is already “in the pipe-
line.” So for two seconds after the TMS is activated, subjects continue to tap nor-
mally. Then, after the two-second delay, their tapping is interrupted. What is more,
research on brain–machine interfaces (neural prostheses) show that subjects can
learn to send the proper motor signals to a machine to make a robotic arm move a
device and serve a purpose. This shows that one can learn to transform a decoupled
motor-intention (before one learns to make the robot obey) into a coupled motor-
intention (when one learns to successfully drive a robotic device) from the motor
cortex.

14. Actually, Gallagher gives six things that he thinks disqualify motor-intentions
from being representations, but they are not all about EC—(1) they are not internal
but extended into the environment, (2) are not discrete, enduring, (3) are not
passive, but enactive, (4) are not decoupleable, (5) are not strongly instructional,
and (6) are not homuncular, involving interpretation. I have dealt with extended
mind issues elsewhere (Adams and Aizawa 2008). I will be addressing their being
enactive in addressing their content. The whole point of a motor-intention is to be
enactive. I won’t here discuss the claim that they are not “strongly instructional.”
Gallagher doesn’t either. He merely says that only EC can solve the frame problem
and says representations would have to contain something like Searle’s (1983)
“background” to do so. I am working indirectly on this in thinking about closure
and knowledge. If knowledge is not closed, it is precisely because there are channel
conditions necessary for knowledge representations that contain information not
contained in things they enable the subject to know (Adams, Barker, and Figurelli
manuscript). And I won’t discuss the homuncular objection because no one in the
naturalized semantics camp thinks there is anything to “interpretation” of a repre-
sentation that can’t be reduced to purely natural causal networking. Gallagher
doesn’t argue for these so much as refer to others who have. Since I’ve addressed
these other items elsewhere, I won’t address them here.

15. See Adams 2003a,b.

16. Though I think there is nothing like a Frege puzzle in the motor system, Pete
Mandik reminds me that there may be something similar to failure of existential
generalization. For instance, when one TMS’s the motor system (causing decoupling)
and a signal is sent to the muscles “to move,” the muscles may not move. Of course,
this is not exactly a failure of existential generalization because these are impera-
tives, not indicatives. But there will be a failure of compliance conditions being met.
If this counts as an intensional phenomenon, then there is intentionality even in
the context of the content of a motor-intention. However, as soon as one backtracks
up the intentional cascade to the proximal or future intentions, then one finds
intensionality with a vengeance.

17. Al Mele points out to me that MacKay (1981, 630) also believes that motor
schemas involved in handwriting extend to lower components of movements
252 F. Adams

representing neuromuscular activity to achieve the required results in the script

being produced. For MacKay, too, the proximal intention forms an intentional
cascade down to the level of the motor-intention.

18. Of course, Pacherie admits that some actions are performed “on the fly” and
involve only the P-intentions and M-intentions (Pacherie 2008, 189).

19. Some may think that it is not necessary for the proximal intentions to cascade
down to the level of the motor signals. For instance, Mele (1992, 222) suggests that
the proximal intention to button one’s shirt in the normal way may be sufficient
to trigger the motor routines of buttoning, saying, “In normal cases, the plan is, by
default, the agent’s normal plan for that activity.” But there is no normal plan for
catching the bar of soap in the shower or deciding to catch the falling pen with a
precision or power grip. Some of this is done on the fly and in compensation for
changes in environmental contingencies. This tells me that there must be a cascade
of information and representation downward through the layers of the control
systems, as described by Pacherie, Jeannerod, and MacKay.

20. Gallagher (2005) thinks these might even by conscious, but does not argue for
it. Both Pacherie (personal communication) and I think they are purposive and
under control of feedback-controlled mechanisms. They are likely not consciously
intentional. It turns out that this ability shows up in neonates, then diminishes,
and then later returns. This is likely because the basal ganglia are inhibitory. In the
interim between when this imitative activity starts and stops, the basal ganglia
develop. When developed they only allow motor impulses that achieve a certain
threshold to get through and cause these activities once again. At this later point,
it is likely that the infants are indeed protruding their tongues voluntarily and
perhaps fully intentionally.

21. Pacherie says these sequenced signals have a “grammar.”

16 Intentions as Complex Dynamical Attractors

Alicia Juarrero

1 Background

What is the difference between a wink and a blink? Intuitively, we would

say that a wink is intentional and a blink is not. From a philosophical
point of view, answering the question, what is an intention and how does
an agent’s intention cause his or her behavior, has a long history. This
“problem of action” dates back to ancient times. In the Crito, Socrates
wonders how philosophers such as Anaxagoras or Anaximenes would
explain his refusal to escape from prison. How could merely physical phe-
nomena explain how his reasons for remaining keep him in jail? In the last
half of the twentieth century, papers and books on action theory prolifer-
ated. Can reasons be causes? Doesn’t the purposiveness or goal-directedness
of action imply the occurrence of backward causation when an anticipated
outcome plays a role in bringing about present behavior? If intentions are
just brain events, how can they carry meaning such that it is the meaning
or content of the intention that results in the appropriate kind of
behavior?
From these few questions alone it becomes clear that the philosophical
problem of action is fundamentally tied to the problem of causation: how
does meaningful behavior come about? In this essay I will argue that a
uniquely modern combination of views about causality—that only effi-
cient causality is true causality, and that recursive or circular causality is
impossible—is responsible for the impasse in which philosophical action
theory finds itself. Elsewhere (Juarrero 1999; Juarrero-Roqué 1983a,b, 1985,
1987–88, 1988), I chronicle in detail those philosophical equivalents of
epicyclic contortions designed to shoehorn the analysis of action into the
modern notion of efficient causality, where intentional causes are under-
stood as instantaneous forces that push the body into motion in billiard-
ball-like fashion. For that reason alone none of these standard causalist
254 A. Juarrero

theories of action succeeds. But then neither do the attempts that offer
behaviorist or identity-theory reductions of the purposiveness of action.
The impediment, I argued in those earlier works, is their inability to frame
a scientifically acceptable way of understanding mereological (especially
top-down) causality. I claim that concepts borrowed from complex dynam-
ical systems theory can provide just such a theory-constitutive metaphor
that makes intentional causation tractable.
During classical times it is left to Aristotle to provide a thorough analysis
of the difference between voluntary, involuntary, and nonvoluntary
behavior. As is well known, Aristotle accounts for all phenomena, not just
intentional behavior, in terms of four causes. Suppose I intend to write a
book. While my arm and hand movements serve as the efficient cause of
the actual writing, the goal of producing a book functions as its final or
purposive cause. Its material cause, the stuff from which it is physically
constructed, includes the ink, paper, and so on. What makes the behavior
a case of writing a book—instead of something else—is the formal defining
or essential cause that sustains the behavior along that essential path and
guides it to completion.
Less well known, however, is the role that another of Aristotle’s prin-
ciples plays in the history of action theory: the principle that there exists
no circular or recursive causality; no self-cause. The concepts of potentiality
and actuality lead to the conclusion that whatever happens is caused to
occur because of something other than itself. Since nothing can be both
potential and actual at the same time with respect to the same phenom-
enon, mover (actual) and moved (potential) cannot be identical. Even in
the case of the (apparent) self-motion of organisms, one aspect of the
organism qua active principle changes a second aspect from passive to
active; this aspect, in turn, now qua active principle can move a third . . .
and so on, until the animal moves.
By the time the intellectual giants of the seventeenth century and their
followers got through changing the way philosophy and science are done,
Aristotle’s formal and final causes had been discarded as superstitious
nonsense. With material cause left to one side, only efficient cause—the
instantaneous, billiard-ball-collision-type causality of mechanistic sci-
ence—qualified as cause. While discarding three of the four causes, however,
modern science retained the Aristotelian thesis that nothing causes itself.
When combined with the reductionist belief that wholes are no different
from aggregates, the claim that anything that is caused must be caused by
something other than itself—and only as a consequent of an efficient
Intentions as Complex Dynamical Attractors 255

cause—produces an account of action that has bedeviled philosophy of

mind and science for hundreds of years.
The inevitable philosophical upshot of this worldview is either a Carte-
sian substance dualism, or a thoroughgoing materialism. Either there exists
a substance, the mind—along with its will—that because of its different
(nonphysical) nature causes actions—winks, not blinks; or there must exist
a sui generis physical process in the brain that triggers such behavior. The
problem with the former disjunct is that it violates the laws of conservation
of matter and energy: how can a nonphysical phenomenon such as an
intention insert itself in (and alter) the physical space-time continuum to
produce bodily effects without violating conservation laws? The central
problem with the latter alternative, as Socrates questioned, is its inability
to account for how a physical (say, neural) process can embody meaning
and intentionality such that the intention’s meaningful content causes,
guides, and sustains behavior that terminates in action (Searle 1981). For
nearly four hundred years, philosophical reflections relating to action
theory have consisted in epicyclic contortions designed to circumvent
these two main objections. Elsewhere (Juarrero 1999, chs. 2, 4, and 5) I
chronicle these various attempts to fit the causes of action into a modern
understanding of efficient cause. None succeeds.
More generally, the atomism and reductionism that pervaded modern
science reduced integral wholes to mere epiphenomenal aggregates, and
temporal processes to instantaneous snapshots. With the belief that only
primary qualities such as mass are real, secondary relational properties too
were stripped of their ontological status and labeled subjective. According
to this framework, therefore, there is no way to conceive of how particles
might interact to produce an emergent whole with qualitatively novel
properties that can loop back down and affect the component parts. Even
Immanuel Kant, who attributed his philosophical awakening to David
Hume’s influence, recognized that modern science’s understanding of cau-
sality cannot account for living organisms. In the Critique of Teleological
Judgment, Kant (1980) explicitly discusses this failure of mechanistic causal-
ity. As an example he mentions trees, which, because they produce leaves
but are at the same time produced by the leaves, exhibit “a kind of causal-
ity unknown to us” (Kant 1980, 23, §65, AkV, 375). What was missing,
Kant saw, was a scientific framework that explains mereological (part–
whole, whole–part) relationships: (1) how numerous particles can interact
to create wholes with emergent properties, and (2) how the novel pro-
perties of these organized wholes then can exert causal power so as to
256 A. Juarrero

determine a different direction for the whole (Kant 1980; Juarrero-Roqué

1985). Neither the theory of evolution nor the classical thermodynamics
of the nineteenth century provided solutions to these puzzles. Despite the
fact that cosmological and biological evolution present us with increasing
complexity and order over time, the second law of classical thermodynam-
ics’ continually reminded thinkers of the inevitability of dissolution and
decay. What was missing, we can now say, was a conceptual framework
that accounts for mereological interactions and provides a satisfactory
explanation of causal relationships among levels of organization without
reducing the higher to the lower—and is compatible with contemporary
science.

2 Nonlinear Dynamical Systems Theory

Enter complexity theory. With the assistance of powerful desktop comput-

ers that could simulate nonlinear dynamic phenomena that were analyti-
cally intractable, by the last third of the twentieth century the phenomenon
of turbulent flow, physical processes such as Bénard cells, and chemical
phenomena such as the Belousov–Zhabotinsky (B-Z) reaction, showed that
positive feedback loops among particles can result in self-organizing pro-
cesses that, despite their superficially chaotic appearance, in fact embody
very sophisticated degrees of order. These emergent higher-level phenom-
ena, moreover, exert top-down influence on the very components that
make them up (Belousov 1959; Zhabotinsky 1964).
Bénard cells appear when water is heated uniformly from below. At first,
water temperature is close to thermal equilibrium, but as the water in the
bottom of the pan becomes hotter a temperature gradient develops. At
a certain temperature, convection begins, but the random fluctuations—
the bubbles—are initially suppressed. Increasing the heat, however, causes
the system to become unstable, and beyond a critical temperature any
fluctuation can become reinforced such that it drives the system over the
instability threshold and the dynamics suddenly switch: ordered macro-
scopic molecular flows abruptly form into visible rolling hexagonal col-
umns—the Bénard cells. Suddenly, millions of previously independent
water molecules synchronize into a complex, streamlined pattern that dis-
sipates the temperature gradient and lowers internal entropy production.
What’s important for our purposes is that once the rolling hexagonal
convection columns appear, the behavior of each individual water mole-
cule is no longer independent of the others; it is determined top-down by
the cell in which it is caught up. Once each water molecule is entrained
Intentions as Complex Dynamical Attractors 257

into the dynamics of a Bénard cell, that is, its behavior is constrained by
its role in the overall rolling process. If one is willing to broaden one’s
understanding of causality and also to take causal potency so understood
as the mark of the real, even in such elementary phenomena as these, and
despite the claims of mechanistic science, dynamical processes provide
empirical evidence that wholes can be more than just epiphenomenal
aggregates reducible to the sum of their component parts. The newly orga-
nized arrangement shows emergent macroscopic characteristics that cannot
be derived from laws and theories pertaining to the microphysical level;
they also represent eddies of local order that exert active power on their
constituents, top-down. Moreover, the integrity, identity, and characteris-
tics of the overall pattern that constitutes a Bénard cell are decoupled from
its material basis in the sense that different types of viscous fluids can
present the same form of organization, which is no longer identified by
the microarrangement of its constituents. Theories of individual identity
based on token-identifications lose all relevance in these cases.
As intriguing as Bénard cells are, it is nevertheless undeniable that the
conditions within which such dissipative structures appear are set from
without: the rolling hexagonal cells form only because we’ve placed the
fluid in a container of a certain size; because we’ve cranked up the heat
that created the gradient that took the system far from equilibrium and
precipitated the discontinuous transformation from conduction to convec-
tion. Other dissipative structures such as hurricanes and dust devils also
form only because of the external metereological and atmospheric condi-
tions. Clearly, in each of these cases, if the boundary conditions are
removed the structure will disintegrate. Because the constraints within
which the self-organization takes place are set externally, the type of emer-
gence on display in these examples is only a weak form of emergence.
Although such structures can therefore only be said to self-maintain—and
not self-create—they nonetheless offer a “theory-constitutive metaphor” for
rethinking autonomy and top-down causality in a scientifically respectable
fashion.
Things get even more interesting at the chemical level, as Peirce, Mill,
and other precursors of complexity theory recognized (see Juarrero and
Rubino 2008). In the case of the B-Z reaction, the fourth step of the process,
a positive feedback loop in which the product of the process is necessary
for the process itself, represents precisely the type of circular causality
forbidden since the time of Aristotle. As the fourth autocatalytic step iter-
ates, the reinforced, accelerating hypercycle drives the system further and
further from equilibrium until a threshold of instability is reached, at
258 A. Juarrero

which point a random internal fluctuation or external perturbation can,

once again and in a burst of overall entropy production, drive the system
over the threshold of instability and cause it to reorganize at a higher level.
The new regime, as always in nonlinear dynamical systems, is characterized
by lower internal entropy production. What’s novel in chemistry is that
the dramatic colorful waveforms of the B-Z reaction are caused to occur by
the dynamics of the process itself, in the sense that the constraints within
which the new organization comes about are produced by the process’s
endogenous dynamics—the self-organization is precipitated by the far-
from-equilibrium gradient created by the reinforcing loop. Because the
constraints within which the reorganization takes place are in this sense
self-produced, chemical processes truly self-create, not only self-maintain
the way physical dissipative structures do. The term autopoiesis—self-
organization—now applies, as does the term strong emergence.

3 Constraints

Borrowing from information and communication theory, and the work of

Lila Gatlin (Gatlin 1972), I attempted in an earlier essay to analyze this
kind of interlevel—mereological—causality in terms of context-sensitive
constraints (Juarrero 1999). Doing so brings back a role for formal and final
cause, but reworked in scientifically acceptable fashion.
Just as a gas-filled container at thermodynamic equilibrium can do no
work, random taps on an old-fashioned telegraph device, or random pixels
on a computer screen, can transmit no messages. Constraints that take the
system away from equiprobability and equilibrium are necessary to perform
work and communicate, respectively. Be it a piston in the first case or
communication rules in the second, constraints represent or embody
changes in probability distribution. A piston inserted into the container
and moved to one side compresses more molecules to one side than the
other. Pistons thus embody constraints that take the system away from
randomness or equiprobability. In any language, some letters of the alpha-
bet are more likely than others. Following Gatlin I called this kind of
constraint context-free because it embodies the system’s prior probability.
The fact that different kinds of neurons fire at different intervals is
an example of an innate context-free constraint, as are all innate
propensities.
Context-free constraints, however, quickly reach a bottleneck that stifles
communication because reliability of the information thereby produced is
inversely related to message variety. At the limit, only one message would
Intentions as Complex Dynamical Attractors 259

be transmitted with probability 1, all other alternatives being reduced to

probability 0. By flashing at regular (nonrandom or nonequiprobabilistic)
intervals, lighthouses signal “message, message, message.” No pattern
conveys no information; but the same pattern repeated time and again
conveys no new information. Similarly, in nature, mitotic reproduction
faithfully transmits “another, another, another” to a new generation. But
the choke point of mitosis is that it allows no message variety except in
error; and when errors do occur, the earlier biological information is lost.
As Shannon (1948) discovered, to ensure message variety as well as
information transmission a different type of constraint is required, one
that promotes message variety at the same time as it ensures reliable infor-
mation content. Enabling constraints become necessary, in other words.
Shannon’s contribution to information theory was to demonstrate that
reducing error was not incompatible with reliable message transmission if
the message is encoded correctly using what can be called context-sensitive
constraints. The key is making particles and processes interdependent by
correlating and synchronizing them. Doing so establishes conditional prob-
ability distributions among the components.
Consider the following heuristic example, also from linguistics. In the
English language, for example, in addition to the context-free constraints
that make as and es more likely than xs and qs, the probability distribution
of alphabet letters is also related to their interdependence: in English, given
that the sequence of letters …tio has already occurred, the probability that
the next letter will be n is almost 1. The probability that the next letter
will be k is almost nil. The likelihood of a particular letter’s occurrence
therefore depends not only on its own prior probability in that language;
it also depends on the letter or sequence of letters that preceded it. When
the appearance of the letter q (in English) increases the probability that u
will appear next, the two become systematically related: they form an
i-tuplet, a higher-level organization with q and u as components. As rela-
tions among i-tuplets themselves become contextually constrained, words
become possible; further context-sensitive syntactical constraints on top of
those then make sentences possible. And so on. Systems and systems of
systems can emerge. This is true of any language: for instance, since Man-
darin Chinese restricts words to only one or two syllables, differences in
inflection are necessary to ensure message variety.
Words can say more than phonemes, sentences more than words. Since
levels are screened off from each other, new levels of dynamical organiza-
tion involve the appearance of novel capabilities at the uppermost level
(Salthe 1985, 2002). Context-sensitive constraints are therefore generative;
260 A. Juarrero

they create information. The differentiation into a hierarchical system that

is nevertheless dynamically integrated creates entirely new realms of
meaning and possibilities. The overall system AB represents an enlarged
phase space with more degrees of freedom than constituents A and B had
separately: amino acids folded into a three-dimensional protein structure
possess properties and capabilities that the linear sequences on their own
do not, much less the individual codons; neural patterns can carry meaning
that individual neuron firings do not, and so forth. In contrast to context-
free constraints, which limit message variety, the closure of context-
sensitive constraints thus opens up possibilities and increases Shannon
entropy by freeing up previously unavailable message variety.1 Context-
sensitive constraints thus make complexity possible. By correlating and
coordinating previously independent particles into a more complex, dif-
ferentiated whole, context-sensitive constraints enlarge the variety of states
the system as a whole can access.
It is important not to reify these dynamic “structures of process” (Earley
1981). I suggested earlier that context-sensitive constraints are conditional
probability distributions that capture the interdependent relationships
between the components. As just mentioned, they exercise top-down
causal power by resetting the lower-level phenomena’s context-free con-
straints—their prior probability, that is. Such “resetting” should not be
understood as an exercise of efficient causality. The better analogy is with
a game of cards: Whereas at the outset all four players have an equal likeli-
hood of getting an ace, once one of the players has two aces the probability
of any player being dealt an ace ipso facto alters. Their changed probability
distribution just is—just embodies—the distribution of cards in the deck,
combined with the dynamical interdependencies among the four players
over the course of the game. These relationships—a higher-level phenom-
enon—constrain the particulate level, top-down; despite the claims of
seventeenth-century thinkers, relationships can thus be more than just
epiphenomenal.
Even in an artificial neural network this integration can precipitate the
appearance of semantics.2 A dramatic example of what positive feedback
loops can bring about in nonlinear dynamical systems is found in word-
reading neural networks (Hinton and Shallice 1991). Even though it is an
artificial network, when the neural network is trained without feedback
loops and subsequently lesioned, its output errors resemble those of surface
dyslexia: If the word presented was bed, the erroneous output might be
bad, or bid. But if the same network is trained with feedback loops and
then lesioned below the feedback loops, the erroneous output mimics
Intentions as Complex Dynamical Attractors 261

those of patients suffering from deep dyslexia: If the word presented was
bed, the erroneous output might be cot; if the word presented was orchestra,
the erroneous output might be band. The authors conclude that these
remarkable results can be explained only by postulating that as a result of
the circular loops, the network self-organizes a semantic attractor, that is,
a high-dimensional dynamic pattern whose emergent properties embody
semantic relationships. Circular causality is real, and it is responsible for a
creative evolutionary spiral. Instead of representing meaning in a symbol
structure, however, a dynamical neurological organization embodies
meaning in the topographical configurations—the set of self-organized
context-dependent constraints—of its phase space.
The trajectories of complex dynamical processes are characterized by
so-called strange or complex attractors, patterns of behavior so intricate that
it is difficult to identify an overarching order amid the variations they
allow. Strange attractors describe ordered global patterns with such a high
degree of local fluctuation that individual trajectories never quite repeat
exactly, the way a regular pendulum does. Complex attractors are therefore
said to be “thick” because they allow individual trajectories to diverge so
widely that even though they are located within the attractor’s basin, they
are uniquely individuated. The width and convoluted trajectories described
by strange attractors imply that the overall pathways they describe are
multiply realizable. The butterfly-shaped Lorenz attractor is by now a well-
known complex attractor.
Context-sensitive constraints are important not only for information
transmission or in connectionist networks; they are also responsible for
the creation of natural complexity. Feedback loops and chemical catalysts,
as we saw in the B-Z reaction, are natural first-order context-sensitive
constraints. As natural embodiments of first-order context-dependent
constraints, catalysts are one example of how natural dynamics create
complexity by interrelating and correlating what were heretofore indepen-
dent particles. Reentrant loops in the nervous system are another. Once
the catalytic loop achieves closure and a new level of complexity emerges,
the lower-level constituents are henceforth characterized by conditional
probability distributions different from that embodied in their prior prob-
ability. As distributed process wholes, complex structures such as Bénard
cells and B-Z chemical waves thus impose what one might call second-order
contextual constraints on their components by restricting their degrees of
freedom. Particles behave differently once caught up in the higher-level
organization than they would have as independent isolated particles.
Whereas first-order contextual constraints are enabling and generative
262 A. Juarrero

constraints that create novel global systems with increased degrees of

freedom, second-order constraints are restrictive and limiting constraints
that curtail the components’ state space. The evolutionary payoff of restric-
tions at the particulate level, as mentioned earlier, is increased message
variety at the global level, which can now do more and different things
than the parts could on their own.

4 Top-down Constraints as Semiotic

Once top-down second-order contextual constraints are in place, energy

and matter exchanges across an autocatalytic cycle’s boundaries are regu-
lated by the overall cycle, in the service of the whole. The organization itself
determines the stimuli to which it will respond; doing so preserves and
enhances its cohesion and integrity, its organization and identity. A sort
of incipient autonomy therefore appears as control and direction is progres-
sively turned over to the global level’s needs by second-order contextual
constraints that decouple and buffer the overall system from the merely
energetic exchanges. It is important to emphasize that this selection process
operates on the basis of criteria set at the higher level (Ellis 2007), on the
basis of how well the higher-level operation is thereby executed. By altering
the go of things at the lower level in the service of the whole, chemical
processes thus display a self-direction and autonomy absent in those found
at the physical level. Toward the end of this essay I will discuss the impli-
cations of these ideas for the thorny philosophical problem of free will.
Autocatalytic cycles degrade gracefully in large measure as a result of
their robustness in the face of the addition and deletion of component
particles. Autocatalytic cycles, that is, typically add, delete, and replace
components such that the overall cycle’s integrity, persistence, and effi-
ciency are preserved and enhanced. Top-down causality described in this
fashion is thus a process of selection, a kind of semiosis that interprets—
from the point of view and for the benefit of the higher-level whole—which
among all the possible component addition, deletion, or replacement
alternatives best satisfies the requirements of the higher level. Cycle abc,
over time, can thus be transformed into a cycle composed of molecules
def, all the while remaining itself, that is, while continuing to function as
the same type of process—while continuing to embody the same function
(Ulanowicz 1997, 2005). A sort of incipient selfhood can therefore be said
to emerge with complex dynamical systems. Indeed, insofar as a complex
system acts from what can be called its own point of view, in its own
benefit, an adaptive system’s conatus toward continued organizational
Intentions as Complex Dynamical Attractors 263

integrity and robustness is complexity science’s answer to Aristotle’s final

and formal cause. And from chemistry onward, it is a product of a process
of self-creation.
Complex systems also highlight the importance of interlevel contextual
relationships and the relative unimportance of primary qualities—and
thereby also introduce an element of normativity: it now becomes possible
to evaluate the various microarrangements in light of the higher-level func-
tion they carry out. Some will do so better, or more elegantly, and so on,
than others. In light of these remarkable processes, thorny philosophical
topics such as the problem of identity suddenly need revisiting: Once a
system’s feedback loop extends into the environment and back in this
manner, the environment is thereby brought into the phenomenon’s very
structure—into its very identity, in fact. Inside–outside distinctions become
more difficult to specify. By making a system’s current state and behavior
depend systematically on its earlier trajectory, feedback loops also incor-
porate the system’s history into its current state and behavior. Feedback,
in other words, threads a system through both time and space, thereby
allowing a system’s temporal and contextual relationships to become a part
of its very overall structure and identity, making a self-organizing network
of components and its environment “in fact one system” (Rocha 2001, 97).
More generally, once the chemical stage emerges from the merely physical,
true chemical autopoiesis represents the advent of the type–token distinc-
tion, where multiple tokens can embody the same kind of phenomenon,
a phenomenon that must be type-identified as, say, a particular chemical
function.
Our current understanding of nonlinear dynamical phenomena thus
allows us to appreciate that ontological categories can no longer be defined
as in Newton’s time, based solely on internal primary properties. No longer
can one do science solely by isolating a phenomenon in the laboratory
and analyzing its components. The concepts of essence and substance,
identity and individuation, as just mentioned, must be rethought, but it
is in any case undeniable that complexity theory effectively reembeds phe-
nomena in context and history, thereby putting the lie to much of the
atomistic flavor of modernity. When even a snowflake carries the trajectory
it has traversed embedded in its very structure; when the trajectories of
two pendulums—or the same pendulums swung at two different times—do
not tend toward classic equilibrium but instead diverge dramatically
despite virtually identical origins; when the context in which a genome is
embedded alters its expression, causal relationships do not resolve simply
into linear chains of efficient causes. And when self-organized systems
264 A. Juarrero

create the very constraints that control matter-energy flows—when, in

other words, the constraints that give rise to self-organization have been
imported into the system, the capacity for “recursive production of struc-
tures becomes possible.” This happens with the appearance of chemistry
(Moreno 2008). It is thanks to the consequences of autopoiesis that a veri-
table explosion of structural variety becomes possible.

5 Mind as Complex Attractor

In earlier work I hypothesized that as a result of neural feedback loops

(Edelman’s reentrant processes), globally coherent wave patterns similarly
entrain and self-organize into high-dimensional dynamic structures in the
brain possessing emergent mental properties such as consciousness, inten-
tionality (sensu Brentano), qualia, and so on.3 Recent empirical research
provides confirmatory evidence of a global workspace model of conscious-
ness, dynamically understood in this way. Just a few months prior to this
writing, Raphael Gaillard’s team in France compared the way epileptic
patients undergoing intracranial electroencephalograms consciously and
unconsciously processed briefly flashed words using a visual masking pro-
cedure. Although neural activity in multiple cortical areas showed that the
patients were processing masked words nonconsciously, these signals quickly
degraded and the neural activity showed only local synchrony. In contrast,
neural activity indicating conscious processing of unmasked words showed
a sustained voltage increase, particularly in the prefrontal cortex. Even
more distinctly, conscious processing showed long-range phase synchrony
across different portions of the brain. Moreover, these widely distributed
synchronized signals activate others, as evidenced by increases in long-
range Granger causality. The authors conclude that the evidence supports
Bernard Baars’s theory of a global workspace model of consciousness:
bottom-up propagation combined with top-down attentional amplifica-
tion results in a whole-brain neural assembly and “ignites into a self-
sustained reverberant state of coherent activity that involves many
neurons distributed throughout the brain” (Gaillard et al. 2009.). “Con-
sciousness is more a question of dynamics than of local activity,” says
Gaillard (2009, 7).
A self-organized neural state is representational and symbolic if its
central features are given not by the configuration’s intrinsic physical
properties but by the information it carries (Wheeler and Clark 1999). Since
feedback loops of nonlinear dynamical processes extend outward into the
environment and back in time, it is also reasonable to maintain that
Intentions as Complex Dynamical Attractors 265

mental phenomena are embodied processes with tentacles that extend

outward to the environment and history. Although I will use language that
refers primarily to neural or neurological processes, one must keep in mind
that from a complex systems’ perspective a mind’s content is not just in
the brain (Noë 2004).
If I am correct, moreover, mental phenomena should be describable
mathematically as neural attractors. The more abstract and general a
semantic category, the broader the neural attractor it describes; conversely,
the more specific a semantic category, the narrower the attractor. Discover-
ing high-dimensional patterns in the brain that lower local entropy pro-
duction would provide confirmatory data for this hypothesis. Earlier I
noted that in connectionist networks without feedback units, after a few
training runs the connection weights of the intermediate neurons for bid
and bed might be close, representing lexicographical relationships. In net-
works with feedback loops, on the other hand, the dynamic reconfigura-
tion means that bed and cot are now nearer in semantic attractor space
than are bed and bid. According to this view, then, semantic closeness in the
mind too would be tantamount to relative location in dynamic space, a
framework that, incidentally, would also account for the source of Freudian
slips. I would not be surprised if the neural pruning that takes place in
early childhood turned out to be a feature of this streamlining neural
self-organization.

6 Intentional Action

Philosophical debates about top-down causality by supervenient mental

states have invariably raised the following concern: If mental states have
causal potency in virtue of being physical states, and one assumes causal
closure of the physical realm, how can overdetermination be avoided? The
type–token decoupling characteristic of complex systems points to a solu-
tion. A self-organized neural state can also be said to be causal qua repre-
sentational and symbolic—qua mental, that is—if what matters is that the
system’s output is dependent on the neurological pattern, not because of
the configuration’s intrinsic physical properties but because of the mental
content it carries (Wheeler and Clark 1999). The word-reading neural
network and the fact that autocatalytic cycles select molecules for import
on the basis of criteria determined at the global level are just two examples
of how complex dynamical systems satisfy this requirement. The answer
to the threat of causal overdetermination can thus be found in the reverse
formulation: Brain states can have causal efficacy in virtue of being (embodying,
266 A. Juarrero

being entrained into) complex neural states with emergent mental properties.
That is, brain states can have causal efficacy in virtue of the mental content
they carry. The same brain states, token-identified, may have different (or
no) causal effects depending on whether or not they are entrained into a
mental attractor at all, or depending on whether or not they are entrained
in the same mental attractor, type-identified. On this view the micro-
physical configuration of the nervous system’s dynamical network exer-
cises its causal power and produces a particular output in virtue of
embodying the top-down context-sensitive constraints of emergent mental
properties.
Does this way of looking at the mind–brain problem resolve concerns
over causal overdetermination and conservation laws? Paul Humphreys
maintains that unlike aggregates, whose individual components retain
their identities, emergence happens only when microstates fuse (Hum-
phreys 1997). When they fuse, particles comprising unified wholes “no
longer exist as separate entities and therefore do not have all their indi-
vidual causal powers available for use at the global level.” Because compo-
nents “go out of existence” when they fuse, Humphreys maintains, worries
concerning causal overdetermination and the causal closure of the physical
are avoided and top-down causality is possible. Humphreys warns, however,
that fusion and multiple realizability are incompatible: any claim that
mental properties can be variously instantiated in components that do not
“go out of existence” reintroduces the threat of overdetermination, he
insists.
Complex systems have taught us that dynamic self-organization pro-
duces high-level emergents capable of top-down causation without thereby
violating conservation laws. Phase transitions, symmetry breaking, and
other forms of dynamic transformations entrain components into higher-
level wholes without thereby fusing the particles. Instead, the global pat-
terns that emerge as a result of these qualitative changes are embodied as
the conditional probability distributions of the components. The operation
of fusion, a static notion that implies that once fused, there is no going
back, is unlike the operation of integration, despite the thermodynamically
irreversible nature of the latter.4 Fusion is like the operation of context-free
constraints, which, as we saw, close off possibilities in a bottleneck that
prevents open-ended evolution. In contrast, bottom-up context-sensitive
constraints represent interactions that are Goldilocks-like—not too tight,
not too loose—and allow the same microarrangement token-identified to
take part in different global dynamics, type-identified, both synchronically
and diachronically. If the disruptive perturbation or fluctuation is strong
Intentions as Complex Dynamical Attractors 267

enough the global structure disintegrates, but while the constraints hold,
the complex dynamics remain coherent over time (Ulanowicz 2005). The
emergence of dynamical integration, I proposed earlier, is nothing but the
effects of second-order context-sensitive constraints, embodied as a set of
conditional probabilities that are invariant over time and that modulate
and direct the behavior of particular—but now no longer independent—
microphysical constituents in such a way that the mental content carried
by the overall dynamics cascades into the action performed. In graph-
theoretic terms, this unique balance between integration and differentia-
tion can be measured in terms of a network’s causal density (the fraction
of interactions among nodes in a network that are causally significant).
Analysis shows that high causal density is consistent with a high dynami-
cal balance between differentiation and integration, and therefore with
high complexity (Seth 2005, 2006; Seth and Edelman 2007). The robust-
ness characteristic of complex systems is thus due to dynamics that are
globally coordinated while component details remain distinct; compo-
nents do not fuse, and yet the overall system displays a remarkable resil-
ience and metastability in its functional powers despite radical differences
in the arrangements of its component parts.
We are at last in a position to understand how top-down causality of
the sort described above makes intention-caused actions possible. Accord-
ing to this framework, intentions are similarly high-dimensional, neurologi-
cally embodied long-range attractors with emergent properties. No doubt
the billion-plus-neuron human brain possesses an indefinite number of
imbricated dynamics. Intentions, by definition, involve motor attractors,
as well as others embodying high-level properties of meaning (the semantic
content of the intention), emotional valence, and so on. Aesthetic, ethical,
or moral value judgments, I suppose, are embodied in even higher-level
attractors, which self-organize later in both individual development and
neurological evolution.
When intentions strongly entrain motor neurons such that they pre-
cipitate a certain type of behavior with probability near 1, they constitute
proximate intentions; when they merely prime the likelihood of future
motor activity they embody prior intentions (Bratman 1987). Prior inten-
tions restructure a multidimensional neural state into a new organization
characterized by a new set of coordinates and a new dynamics; the context-
sensitive constraints that partition a prior intention’s contrast space carry
emergent properties of meaning, emotional valence, and so on. On this
account logical and syntactical relationships are also embodied in the
higher-level relationships between the various attractors.
268 A. Juarrero

Dynamically, this means that once I form the prior intention to do A,

not every logical or physically possible alternative remains open down-
stream, and those that do are contoured differently, probabilistically speak-
ing. By imposing novel conditional propensities on available pathways,
the dynamics of a newly self-organized semantic space constrain move-
ments within it, top-down. Through the assignation of conditional prob-
abilities, some act-types are eliminated, others are made more likely. Once
I settle on the prior intention of going on a diet, say, the likelihood of
certain future behaviors—such as gorging on cookies—recedes. (Other con-
siderations such as the robustness of the agent’s character, which can now
be understood as the combined strength, stability, and resilience of the
agent’s mental attractor set, must also be taken into account.) As a result
of further interactions among the prior intention, other internal dynamics,
and the environment, one option from those weighted alternatives estab-
lished by the new organization can become a proximate attractor. By
strongly entraining motor dynamics to its own, this pathway in turn
guides and constrains behavior in the immediate future. Formulating a
proximate intention (do A now) can thus be understood as positioning a
particular attractor just ahead of the system’s immediate position in phase
space and entraining the necessary motor attractors to execute the action.
In this way the context-sensitive constraints that embody the meaningful
content of intentions cascade into behavior top-down and carry out inten-
tional behavior. And so the person acts. In recent years, nonlinear meth-
odologies suitable to complex systems are providing increasing empirical
confirmation that indeed intentions can be naturalized as soft-assembled
context sensitive constraints operating as control parameters (Riley and
Turvey 2001; Van Orden, Kloos, and Wallot 2009).
If I am right, one needs to be careful about conclusions drawn from the
curious research on “choice blindness” (Hall and Johannson 2009). In a
series of experiments, subjects were asked to select between, for example,
two varieties of jam. Immediately after the volunteers made their choice,
a different jar was substituted for the one selected. When asked why they
chose the (substituted) jar, not only did the participants not notice the
mismatch, they even offered “reasons” for their “choice.” On my account,
prior and proximate intentions can be more or less vague and ambiguous,
or more or less specific and concrete; dynamically, the attractors will be
more or less shallow and broad, or more or less narrow and deep. My
proximate intention can be framed as either “stop at the next convenience
store to buy something—anything—to eat for dinner,” or “stop at the next
7-11 to buy a Healthy Choice brand frozen chicken dinner.” The latter’s
Intentions as Complex Dynamical Attractors 269

attractor is represented by a much narrower and deeper well; the former’s

attractor would be embodied in a much shallower and broader basin. If
my proximate intention had been formulated as the former, I would most
likely not even notice that the item I selected from the freezer had been
switched, no matter how conscious the decision. It seems to me therefore
that Hall and Johannson’s conclusion that “choice blindness drives a
wedge between intentions and actions in the mind” follows necessarily
only if the relationship between intention and action is conceptualized as
billiard-ball causality between crisply defined objects (Hall and Johannson
2009, 27). One has to remain alert to how uncritical presuppositions about
mechanistic causality can creep into the design of experiments and the
conclusions we draw from them.
A similar caution applies to research that extrapolates from the timing
of the appearance of the readiness potential to the reality of free will (Libet
2004). The Libet experiments went roughly as follows: participants were
asked to perform activities such as pressing a button, or flexing a finger or
wrist, within a certain time frame. Participants were also asked to note the
position of the dot on the oscilloscope timer when “he/she was first aware
of the wish or urge to act.” The instant the button was actually pressed,
or the finger flexed, was also recorded on the oscillator, this time electroni-
cally. By comparing the moment the button was pushed and time of the
subject’s conscious decision to act, researchers were able to calculate the
time between the subject’s initial volition and the ensuing action.
The unexpected results showed that brain activity initiating behavior (the
readiness potential) preceded (by up to 300 milliseconds) any conscious
decision to act; even more remarkable, in their reporting the subjects often
antedated the volition such that the intention and decision (was felt to)
precede the action.
Since regulatory records operate in a time-independent mode and exert
top-down causal influence on lower-level metabolic processes operating in
a dynamic and time-dependent mode, arguments such as these purporting
to refute the possibility of free will based on linear temporal relations
between neural events and experience call for closer scrutiny. Libet’s exper-
iments clearly disallow mechanistic interpretations of the concept of free
will that conceive of voluntary intentions as conscious representations
distinctly separate from and preceding the actions they forcefully bring
about as efficient causes. On such an account the subject’s antedating is
of course illusory, as is, a fortiori, any claim that the volition efficiently
caused the behavior. But if one envisions top-down causality as the opera-
tions of second-order context-sensitive constraints, and one keeps in mind
270 A. Juarrero

the fact that feedback loops embodied in those constraints extend outward
into the environment and back in time, then the following alternative
interpretation opens up: If the subject’s intention is formulated in vague
and general terms as suggested earlier—“I will press the button every so
often”—the agent can then let the environment carry out the detailed
movements without thereby obviating the higher level’s control on the
lower. Rodney Brooks calls the process “letting the world serve as its own
model” (Brooks 1991). If I decide merely to “drive home” (as opposed to
the more precise and specific “drive home along route X”), I can let the
lay of the land—the traffic pattern or road conditions—determine for me
whether to turn right on Oak Street or left on Poplar Street. A well-known
example is the following: after driving home from work along our daily
commuter route all of us have experienced the peculiar sensation of real-
izing that we can’t even remember if the traffic light at Elm and Maple
was red or green that day. Following Libet’s reasoning, it is obvious that
in such situations the actions of depressing the accelerator or brake bypass
conscious decision making. But are they not therefore not constrained by
awareness? I don’t think so. Had a rabbit or a deer suddenly jumped in
our path the automaticity would have been quickly replaced by con-
sciously aware decision making.
The problem with the Libet-like experiments, on my view, is their reli-
ance on a traditional understanding of causality as series of discrete events
acting as efficient causes. In contrast, the thesis put forth in this essay has
been that functional, informational, symbolic, and representational pro-
cesses operate in the brain as top-down, second-order context-sensitive
constraints. Unlike efficient causes, which have traditionally been under-
stood to be instantaneous, atomistic events, the dynamics of higher
informational regulatory systems—the genetic and neural systems, for
example—often operate at a slower and longer time scale than that of their
constituents; that these slower and longer dynamics can nonetheless con-
strain the lower-level dynamics that constitute them is an illustration of
what Haken calls “slaving.” Phase separations between levels of organiza-
tion are often speed and time differences between levels of organization.
Phases are clearly demarcated wherever crisp differences in time scales exist
between the higher-level emergent dynamics and processes at the lower
level; the Gaillard research mentioned earlier is just one of the most recent
to show that dynamic phase separations are significant. Salthe (1985, 2002)
notes that higher-level laws and principles can apply to dynamics that are
slower and longer than those of the lower level; but the higher level is not
always slower or faster than the lower: The regulatory genetic system oper-
Intentions as Complex Dynamical Attractors 271

ates faster than the metabolic system; the regulatory neural system works
faster than the metabolic system. My point here is merely to point out that
“there’s more to heaven and earth than is dreamt of in your [mechanistic]
philosophy, Libet,” and so empirical research designed from a mechanistic
framework may be chasing red herrings. Research on the role of timing in
brain processing is still in its early stages (Carey 2008), but I am confident
that future research into dynamic phase separation (such as that between
alpha and beta patterns in the brain) will shed light on causal propagation
and the process of top-down modulation control and regulation. In light
of the discussion Libet’s work has occasioned, further work on this feature
of brain dynamics is warranted, especially phase differences between levels
of neural organizations such as the regulatory informational system and
the lower-level motor networks under the former’s control.
Elsewhere I argued that the information-theoretic concepts of equivoca-
tion and ambiguity might help track the effectiveness of intentional con-
straints over the course of the behavior’s trajectory. Newly developed
techniques designed to capture the interdependencies between multiple
time series should prove even more useful in determining the flow of
information from one part of the brain to another. Granger causality,
mentioned earlier in connection with the research on the distributed char-
acter of conscious processes in the brain, is one such technique that com-
pares streams of data known as time series, such as fluctuations in neural
firing patterns. Granger causality helps determine whether correlations are
mere coincidences or reflect one process influencing another process. In
one recent study,

Researchers gave volunteers a cue that a visual stimulus would be appearing soon
in a portion of a computer display screen, and asked them to report when the
stimulus appeared and what they saw. Corbetta’s group previously revealed that this
task activated two brain areas: the frontoparietal cortex, which is involved in the
direction of the attention, and the visual cortex, which became more active in the
area where volunteers were cued to expect the stimulus to appear.
Scientists believed the frontoparietal cortex was influencing the visual cortex,
but the brain scanning approach they were using, functional magnetic resonance
imaging (fMRI), can only complete scans about once every two seconds, which was
much too slow to catch that influence in action. When researchers applied Granger
causality, though, they were able to show conclusively that as volunteers waited for
the stimulus to appear, the frontoparietal cortex was influencing the visual cortex,
not the reverse. (Washington University School of Medicine 2008)

Because Granger causality only applies to pairs of variables, misleading

results may occur when the true relationship involves three or more
272 A. Juarrero

variables: If both variables tested are in fact being “caused” by a third,

Granger tests may yield positive results even though the two variables have
no causal relations with each other. In cases such as these a related test
called vector autoregression (VAR) can be applied to better track information
flow in the brain. Neuroscientists are currently using both techniques in
their studies, whose purpose is to capture the conditional probabilities
carried by context-sensitive constraints.

7 Autonomy

I would like to close with some thoughts concerning the implications that
complex adaptive systems have with respect to the philosophically thorny
subject of free will. I propose that these dynamical processes allow one to
rethink autonomy and self-direction in a scientifically respectable way and
with enough of a payoff to warrant the appellation a kind of free will worth
wanting.
We saw earlier that the change from physical Bénard cells to chemical
autocatalytic cycles occurs when the boundary constraints within which
the organization takes place are created by the process itself—when, in
other words, the production of the order parameter is brought inside the
system itself, dynamically speaking. When the endogenous dynamics
themselves control the overall process, the slack or decoupling between type
and token that turns regulatory control and direction over to type-identified
criteria provides a measure of self-direction and autonomy to chemical
processes that is absent in merely physical ones.
Increased structural complexity through chemical catalysis production,
however, in the end becomes brittle and reaches a bottleneck; chemistry
soon reaches its limit, that is, at a choke point that prevents further evolu-
tion. As we saw earlier, because it relies solely on context-free constraints,
mitotic biological reproduction quickly does too. Lila Gatlin, whom I cited
earlier, argues that, evolutionarily speaking, the evolutionary bottleneck
was breached and truly open-ended evolution made possible only after
vertebrates discovered a way to retain context-free constraints stable while
simultaneously allowing context-sensitive constraints to expand. The
research team of Alvaro Moreno, Juli Peretó, and Kepa Ruiz-Mirazo at the
University of the Basque Country (Spain) published a series of papers (Ruiz-
Mirazo and Moreno 1998, 2000, 2006; Ruiz-Mirazo, Peretó, and Moreno
2004; Moreno 2008) individually and severally in which they expand on
Gatlin’s insight: open-ended evolution necessitated a way of preserving
earlier evolutionary advances while at the same time continuing the
Intentions as Complex Dynamical Attractors 273

process of catalytic creation. It required, in short, the creation of hereditary

lineages. The appearance of what the Basque team calls dynamic decoupling
guarantees both the open-endedness of evolution and the emergence of
strong autonomy.
Robust and ongoing biological evolution becomes possible, that is, only
when a second type of biological function appears that preserves prior evo-
lutionary advances without interfering with the (ongoing) catalytic pro-
duction from which those advances originated in the first place. This new
type of functional component—of which DNA is a prime example—serves
as what Moreno and his colleagues call a register (or record). This decoupling
also results in the concomitant emergence of strong autonomy, because once
again registers construct the constraints within which the ongoing cata-
lytic function operates. The important point I want to make here is that
when dynamic decoupling appears, the transformation that took place
with the emergence of chemistry is taken one step further: with the appear-
ance of registers the control and regulatory function is brought even
further inside the system dynamics themselves. Constraints within which
ontogenetic development is canalized not only embody phylogenetic
information—they are now on board. Adding an additional set of internal
top-down contextual constraints further decouples organisms from buffet-
ing by energetic forces, thereby making it even more self-directing. As
keeper of biological lineages, DNA faithfully preserves a relatively invariant
record of phylogenetically earlier traits without thereby becoming involved
in or interfering with those ongoing metabolic functions that allow varia-
tions to continue to appear over epigenetic time. Neuronal self-organiza-
tion and the use of tools are other examples of this type of dynamic
decoupling, claims the Basque team. As was the case in the emergence of
chemical from physical processes, when register function decouples from
catalyst function, the control and regulatory criteria that will direct the
future course of events have been handed over even more so to the
organism.
Because the two functions (registers and catalysis creation) cannot be
directly linked, a novel, information-type of semiosis is required to translate
between the two: “code-type information semiosis,” of which DNA’s genetic
code is the prime example, is charged with this task (Moreno 2008). In my
language, dynamic decoupling with indirect code-type information semio-
sis is a version of top-down causation in which semantic higher-level code
regulates and modulates—constrains—lower-level energetic exchanges
according to criteria set at the higher level. The law-like relations between
the two phase-separated dynamics carry the code-type information
transmitted.
274 A. Juarrero

I propose that the appearance of consciousness and related mental phe-

nomena such as intentions constitutes the emergence of yet another regis-
ter-type function, a function that once again serves as record or register—in
this case of earlier ontogenetic experience. Having access to a conscious
mental workspace while sensory mechanisms continue their ongoing work
permits an explosive expansion of state space. It permits, that is, a new
dimension of message variety, with trajectories regulated top-down accord-
ing to mental content. The agent can now do a lot more and different
things than before. Whereas with the appearance of DNA heredity became
encapsulated in a new anatomical structure, the new conscious, mental
workspace functions as a dynamical register, encoding mental content on a
different dynamical scale (see Salthe 1985, 2002, on scalar and specification
hierarchies; see Gaillard 2009). It is tempting to speculate whether emo-
tions serve as the indirect information-type code that translates between
the two. Freedman’s research on the way olfactory stimulus simultaneously
encodes meaning and emotion when remembered could be interpreted in
this way (Freeman 1991).
What seems certain is that human symbolic language functions as the
context-sensitive translation code between self-consciousness or long-term
declarative memory on the one hand and ongoing sensory input on the
other. The transformation of short-term memory into long-term declara-
tive memory might turn out to be a component of this process. And when
human language becomes entrained into culture and then recorded in
written form, the added sedimentation of linguistic meaning encoded in
dictionaries is an additional example of yet another high-level register
(sensu Moreno) that enables cultural evolution to continue, and explo-
sively so. The task of each new register is the same as DNA’s: to preserve
established information while allowing human beings to create new
meaning as they experience new things. It should come as no surprise
therefore that the meaning of concepts and even of “forms of life,” as it
becomes sedimented in dictionaries and moral codes, can bias the condi-
tional probability distributions of patterns in individual nervous systems,
both developmentally and phylogenetically.5
The process by which actions can be value-informed is therefore less
mysterious once this framework is adopted. When from among the various
alternatives presented to a potential agent she chooses one on the basis of
very high-level value-infused criteria, behavior turns into ethical and moral
action (moral as opposed to amoral, not immoral) in virtue of the axiologi-
cal content the intention constrains, not the physical microarrangement
of the neurological pattern that carries that content). Here again we have
Intentions as Complex Dynamical Attractors 275

another example of a regulatory control function being “brought inside”—

not structurally like DNA but dynamically—and therefore even more
strongly decoupled from external, energetic control and direction.
The central thesis of this essay has been that the emergence, operation,
and evolution of complex dynamical systems show how global systemic
levels grow more and more self-directed as their capacity to constrain,
modulate, and regulate their constituent levels is increasingly modularized
and dynamically decoupled from merely energetic exchanges—in recent
evolution by being split off as independent registers or records. Complex
systems with this feature, which begins with biological heredity, are strongly
autonomous. And they become even more so over the course of evolution
with the emergence of consciousness, self-consciousness, values, and sym-
bolic language. If mind and language are the most evolutionarily advanced
dynamically decoupled records or registers; if decoupling from external buf-
feting is precisely the sort of trait one needs to determine autonomy; and
if top-down context-sensitive constraints operating in complex adaptive
systems can explain how the course of things can be regulated and modu-
lated by higher-level criteria such as meaning and values; then it seems to
me that nonlinear dynamical systems theory makes room for a kind of free
will worth wanting.

Notes

1. It does not just activate previously existing potential; complexification creates

heretofore nonexistent possibilities—and therefore new directions for the system.

2. See Wheeler and Clark 1999 for additional examples of new properties created
by the causal spread of context-sensitive constraints.

3. These constraints are on top of those preexisting constraints, both context-free

and context-sensitive, that present as innate predispositions or propensities—such
as those embodied in mirror neurons, for example.

4. The irreversibility is provided not by unalterably fusing the micro-level compo-

nents into a global structure, but by their historicity and context-dependence. These
are embodied in the system’s internal dynamics, which “carry on their back” the
conditions under which the systems were created and the trajectory they have
undergone. Even snowflakes, for example, carry in their very structure information
about both the atmospheric conditions that caused them and those they traversed
before reaching the ground.

5. Like DNA, however, and despite the claims of Foucault and other postmodernists,
the top-down constraints of language are not completely rigid; instead they allow
for the creation of new meaning accomplished through play and exploration.
17 The Causal Theory of Action and the Still Puzzling
Knobe Effect

Thomas Nadelhoffer

One can test attempted philosophical analyses of intentional action partly by ascer-
taining whether what these analyses entail about particular actions is in line with
what the majority of non-specialists would say about these actions. . . . [I]f there is
a widely shared concept of intentional action, such judgments provide evidence
about what the concept is, and a philosophical analysis of intentional action that
is wholly unconstrained by that concept runs the risk of having nothing more than
a philosophical fiction as its subject matter.
—Alfred Mele (2001, 27)

1 Introduction

It is a common assumption among philosophers that whether or not

something counts as an intentional action depends on what was going on
“in the head” of the agent at the time it was performed. On this view,
intentional actions are events that are caused by a combination of anteced-
ent conative and cognitive mental states such as wishes, desires, beliefs,
reasons, decisions, and intentions.1 And although the philosophers who
adopt this kind of causal theory of action may admittedly disagree when
it comes to the precise nature of the relationship between various mental
states and human agency—for example, whether intentions can simply be
reduced to pairs of beliefs and desires or whether intentions to x are neces-
sary for intentionally x-ing—the orthodox view is that the proper etiologi-
cal explanation of intentional action will be couched exclusively in
mentalistic terms.
As intuitive as this view may initially appear, however, there is gathering
evidence that folk ascriptions of intentional action are sometimes driven
not only by judgments concerning the mental states of the agents in ques-
tion but also by the moral valence of the outcome of the action. These
findings—which are collectively referred to as the “Knobe effect” (Nichols
278 T. Nadelhoffer

and Ulatowski 2007, 348)—have led some philosophers to conclude that

the concept of intentional action is inherently evaluative (Knobe 2003a,b).
To the extent that this is the case, it potentially poses problems for the
causal theory of action—at least if this theory is supposed to adequately
capture the folk concept of intentionality.2 After all, if an action can be
intentional so long as its consequences are sufficiently bad even if the
agent who brought it about neither desired, wanted, intended, nor tried
to do so, the project of explaining intentionality solely in terms of the
mental states of the agent is on shaky ground.
In this chapter, I am first going to briefly argue for the relevance of data
on folk intuitions to the philosophical project of providing an account of
intentional action. Then, I will examine some of the empirical research
that has been done on this front—paying particular attention to Steve
Guglielmo and Bertram Malle’s recent attempts to resolve some of the
puzzles that have arisen in light of Joshua Knobe’s findings (Guglielmo
and Malle n.d.a, n.d.b). More specifically, I will try to show that Guglielmo
and Malle have yet to establish that the so-called Knobe effect is merely
an artifact of an impoverished experimental design. In making my case, I
will present the results of two new studies that suggest that even though
Guglielmo and Malle have done much to advance our understanding of
the folk concept of intentional action, they have yet to fully explain away
Knobe’s puzzling data.
At the end of the day, I will argue not only that philosophers of action
need to pay close attention to the research on folk intuitions, but that this
research potentially puts pressure on the common assumption that whether
an action is intentional depends entirely on the mental states of the agent.
That being said, it is admittedly too early in the day to determine the full
extent of the threat. Plumbing these depths would require more data than
we currently have at hand. As such, my primary goal in this chapter is
merely to survey a small piece of the empirical terrain, present the results
of two new studies, and consider some of the potential implications for
the philosophy of action. Resolving all of the thorny underlying issues is
an arduous task for another day that will require philosophers and psy-
chologists to continue to work together on the important project of under-
standing the fundamental nature of human agency.

2 Folk Concepts and the Philosophy of Action

Oliver Wendell Holmes once famously remarked that “even a dog distin-
guishes between being stumbled over and being kicked” (Holmes 1963,
The Causal Theory of Action and the Knobe Effect 279

7)—a difference that permeates our ordinary judgments about responsibil-

ity and blame. In the event that someone accidentally or unintentionally
harms you, this usually mitigates how much blame you think the person
deserves for doing so. If, on the other hand, someone purposely and inten-
tionally harms you under normal circumstances, this is grounds for judging
the person to be fully morally (and possibly legally) culpable for the harm.
In short, judgments about intentions and intentional actions lie at the
heart of the sphere of our ascriptions of responsibility. Moreover, a stan-
dard direction of fit is commonly assumed to exist between the two—
namely, judging whether an agent is fully responsible for x-ing requires
one to first decide both whether she intended to x and whether she x-ed
intentionally.
For present purposes, I am going to call this the standard model of inten-
tionality and blame. In short, it is a unidirectional view whereby judgments
about intentionality serve as inputs to judgments concerning responsibil-
ity, but not the other way around. Understandably, this seemingly com-
monsensical model has been adopted by philosophers and legal theorists
alike. However, as intuitive as it may appear at first blush, this view has
been called into question by recent data on folk ascriptions of intentional
action. But before I examine the gathering evidence for the competing
bidirectional model—the view that judgments concerning the badness of
an action (or the blameworthiness of an agent) sometimes influence judg-
ments of intentionality and not just the other way around3—I first want
to briefly discuss the relevance of folk intuitions to the philosophy of
action more generally.
The first thing worth pointing out on this front is that there is a storied
tradition in philosophy whereby philosophers are in the business of ana-
lyzing folk beliefs, judgments, and concepts. Frank Jackson has arguably
put forward the most thorough and persuasive recent defense of this posi-
tion. On his view, “our subject is really the elucidation of the situations
covered by the words we use to ask our questions—concerning free action,
knowledge . . . or whatever” (Jackson 1998, 33). And given that ordinary
language is often the proper subject matter of philosophy, “consulting
intuitions about possible cases” is the proper method (ibid.). Although I
entirely agree with Jackson’s contention that philosophers should often
focus on our ordinary beliefs and judgments,4 his view is nevertheless
based on the questionable assumption that the intuitions of analytic phi-
losophers are representative of those of laypersons.
This is a problem that Jackson acknowledges when he says that he is
“sometimes asked—in a tone that suggests that the question is a major
280 T. Nadelhoffer

objection—why, if conceptual analysis is concerned to elucidate what

governs our classificatory practice, don’t I advocate doing serious opinion
polls on people’s responses to various cases? My answer is that I do—when
it is necessary” (ibid., 36–37). I, for one, agree with Jackson that when it
comes to some issues such as free will, moral responsibility, and intentional
action, people’s ordinary intuitions are the appropriate starting point for
philosophical investigation. However, I nevertheless believe that Jackson’s
method of getting at them—namely, armchair reflection coupled with
informal polls of students and friends—is methodologically problematic.
On my view, to the extent that philosophers are going to continue making
claims concerning the contours of folk intuitions, the task of testing these
claims should be attended to with the same carefulness and rigor one finds
in psychology and cognate fields. In light of this concern, researchers
working under the rubric of “experimental philosophy” have recently
begun using techniques borrowed from the social sciences to get at folk
judgments concerning a wide variety of philosophical issues.5
One topic that has received a lot of attention among experimental
philosophers is the concept of intentional action.6 After all, as Alfred Mele
has correctly pointed out, any adequate philosophical analysis of inten-
tional action should be “anchored by common-sense judgments” about
particular cases (Mele 2001, 27)—even if it admittedly need not capture or
reflect all of these judgments. On this view, one way of testing an analysis
of intentional action would be to see whether it agrees with our pretheo-
retical intuitions. And the only method of determining what nonspecialists
think about particular cases is to actually ask them. Having done so, if we
find that an analysis of intentional action is entirely inconsistent with folk
intuitions, we will be in a good position to suggest that it “runs the risk
of having nothing more than a philosophical fiction as its subject matter”
(ibid.).
Data about the folk concept of intentional action become all the more
important for philosophers such as Hugh McCann who explicitly claim to
be interested in ordinary concepts and not their philosophical counterparts
(McCann 1998, 210). So, whereas those philosophers who are merely
interested in technical concepts and other so-called “terms of art” may
dismiss empirical data concerning folk intuitions tout court, philosophers
such as Jackson and McCann, who are explicitly interested in analyzing
our everyday concepts, are beholden to the results of the kind of research
being done by experimental philosophers.7 As such, even if experimental
philosophy is admittedly not relevant to all philosophical projects, it is
inescapably relevant to some of them.
The Causal Theory of Action and the Knobe Effect 281

Unfortunately, until fairly recently, there was a dearth of empirical

research on how people actually think and talk about intentionality. Malle
and Knobe tried to fill in this lacuna in their groundbreaking paper entitled
“The Folk Concept of Intentional Action” (1997). In order to sketch the
boundaries of ordinary judgments, Malle and Knobe ran a series of surveys.
In light of their findings—the details of which need not concern us here—
they developed the following five-component model of intentionality:

Belief Desire

ෑෑ

Intention

Skill Awareness

ෑෑෑ

Intentionality

According to this model, performing an action intentionally “requires the

presence of five components: a desire for an outcome; beliefs about an
action that leads to that outcome; an intention to perform the action; skill
to perform the action; and awareness of fulfilling the intention while per-
forming the action” (Malle and Knobe 1997, 12). Moreover, Malle and
Knobe suggest that these five components are “hierarchically arranged,
such that belief and desire are necessary conditions for attributions of
intention and, given an intention, skill and awareness are necessary condi-
tions for attributions of intentionality” (ibid., 15).
Malle and Knobe’s five-component model has several strengths. First, it
is based on controlled and systematic studies rather than armchair specula-
tion or informal polls of students, friends, and colleagues. Second, by
purportedly identifying all of the necessary components of the folk concept
of intentional action, their model not only integrates a number of the past
analyses of intentionality, but it also reveals where these previous analyses
went wrong. However, despite the obvious advantages provided by Malle
and Knobe’s model, there is gathering evidence that when participants are
given morally valenced vignettes, they often ascribe intentionality even
though some of the conditions of the five-component model have not
been satisfied. In light of this research, it appears that the aforementioned
bidirectional view may better comport with the data on folk intuitions
than its unidirectional counterpart. If true, this would not only put
282 T. Nadelhoffer

pressure on the standard model of intentionality, but it would also cast

doubt on the causal theory of action more generally. But we are getting
ahead of ourselves. Before we start worrying about the possible implica-
tions of these recent findings, we should first take a closer look at the
findings themselves.

3 The Knobe Effect

In a series of recent studies, Knobe set out to determine whether folk intu-
itions about the intentionality of foreseeable yet undesired side effects are
influenced by moral considerations (Knobe 2003a,b). Each of the 78 par-
ticipants in the first of these side-effect experiments was presented with a
vignette involving either a “harm condition” or a “help condition.” Those
who received the harm condition read the following vignette:

The vice-president of a company went to the chairman of the board and said, “‘We
are thinking of starting a new program. It will help us increase profits, but it will
also harm the environment.’” The chairman of the board answered, “‘I don’t care
at all about harming the environment. I just want to make as much profit as I can.
Let’s start the new program’.” They started the new program. Sure enough, the
environment was harmed. (Knobe 2003a, 191)

They were then asked to judge how much blame the chairman deserved
for harming the environment (on a scale from 0 to 6) and to say whether
they thought the chairman harmed the environment intentionally. Of the
participants, 82 percent claimed that the chairman harmed the environ-
ment intentionally. Participants in the help condition, on the other hand,
read the same scenario except that the word “harm” was replaced by the
word “help.” They were then asked to judge how much praise the chair-
man deserved for helping the environment (on a scale from 0 to 6) and to
say whether they thought the chairman helped the environment inten-
tionally. Only 23 percent of the participants claimed that the chairman
intentionally helped the environment (ibid., 192).
In another side-effect experiment, Knobe got similar results. This time
each of the 42 participants received one of the following two vignettes:

Harm Condition
A lieutenant was talking with a sergeant. The lieutenant gave the order:
“‘Send your squad to the top of Thompson Hill.’” The sergeant said: “‘But
if I send my squad to the top of Thompson Hill, we’ll be moving the men
into the enemy’s line of fire. Some of them will surely be killed!’”
The lieutenant answered: “‘Look, I know that they’ll be in the line of fire,
The Causal Theory of Action and the Knobe Effect 283

and I know that some of them will be killed. But I don’t care at all about
what happens to our soldiers. All I care about is taking control of
Thompson Hill’.” The squad was sent to the top of Thompson Hill. As
expected, the soldiers were moved into the enemy’s line of fire, and some
of them were killed (ibid.).

Help Condition
A lieutenant was talking with a sergeant. The lieutenant gave the order:
“‘Send your squad to the top of Thompson Hill.’” The sergeant said: “‘But
if I send my squad to the top of Thompson Hill, we’ll be taking them out
of the enemy’s line of fire. They’ll be rescued!’” The lieutenant answered:
“‘Look, I know that we’ll be taking them out of the line of fire, and I know
that some of them would have been killed otherwise. But I don’t care at
all about what happens to our soldiers. All I care about is taking control
of Thompson Hill’.” The squad was sent to the top of Thompson Hill. As
expected, the soldiers were moved out of the enemy’s line of fire, and some
of them were saved (ibid.).

Once again, the two conditions yielded drastically different responses:

77 percent of the participants who read the harm condition said the agent
intentionally brought about the negative side effect, whereas only 30
percent of those who read the help condition said the agent brought about
the positive side effect intentionally (ibid.). When Knobe combined the
praise and blame ratings from the two experiments, he got the following
results: Whereas the participants who were given the harm condition said
the agent deserved a lot of blame (M=4.8), those who were given the help
condition said that the agent deserved virtually no praise (M=1.4). More-
over, these results were correlated with their judgments about whether or
not the side effect was brought about intentionally (ibid., 193).
Needless to say, philosophers and psychologists alike found these find-
ings to be both surprising and puzzling. If nothing else, the results of
Knobe’s studies provided support for the bidirectional model of intention-
ality while at the same time putting pressure on the more “valence-neutral”
models that have traditionally populated the action theory literature.8
However, although the asymmetry between participants’ responses in the
harm and help conditions is admittedly noteworthy, trying to fully explain
it would take us beyond the scope of the present essay. Multiple explana-
tions of the “Knobe effect” have been put forward in the literature and
a number of follow-up studies have been run. Now is neither the time
nor the place to examine either all of the existing data or the plethora of
competing views that have been put forward to explain them.9 For present
284 T. Nadelhoffer

purposes, the most important take-home lesson from Knobe’s research is

that in the harm conditions, participants judged that the side effects were
brought about intentionally, even though neither the CEO nor the lieuten-
ant cared at all about bringing them about. After all, these findings offer
prima facie evidence against the view that Michael Bratman has dubbed
the “simple view” (Bratman 1984)— the view whereby intending to x is
necessary for intentionally x-ing.10
To see how the results of Knobe’s studies put pressure on the simple
view, we must first make a few initial assumptions. On the one hand, if
an agent does not care at all about doing x (or bringing x about), then she
lacks any desire to x. On the other hand, if an agent does not want or
desire to bring x about, then she does not intend to do so. If these assump-
tions are correct, then Knobe’s participants were willing to judge that the
side effects were brought about intentionally even though the agents did
not want—and hence did not intend—to do so. Thus, at first blush, Knobe
has provided empirical evidence that calls both the simple view and the
aforementioned five-component model of intentionality into question.
More importantly, for present purposes, to the extent that folk ascriptions
of intentional action are driven not only by judgments concerning the
mental states of the agents but also by the moral valence of the outcome,
Knobe’s findings also create problems for the causal theory of action more
generally.
Keep in mind that at least according to the aforementioned philosophi-
cal approach to action theory that is espoused by Jackson, McCann, and
others, an adequate theory of intentionality will be one that comports with
salient folk intuitions and judgments. As such, if it turns out that whether
someone is judged to have done something intentionally depends not only
on her mental states at the time but also on the moral valence of the
outcome, then the causal theory of action falls short as a complete analysis
of the folk concept of intentional action. Unsurprisingly, defenders of the
more traditional valence-neutral model have not taken Knobe’s findings
lying down. Indeed, as we are about to see, recent studies suggest that
perhaps the conclusions drawn by Knobe and others have been
premature.

4 Resurrecting the Standard Model

In a pair of recent papers, Guglielmo and Malle set out to put pressure on
Knobe’s two central findings—namely, (a) the “side-effect findings,” and
(b) the “skill findings.” The former—which potentially undermine the
The Causal Theory of Action and the Knobe Effect 285

“intention” condition of Malle and Knobe’s five-component model—pur-

portedly show that people sometimes judge that an agent can bring about
a side effect intentionally even if she neither wants, tries, nor intends to
bring it about. The latter findings—which potentially undermine the skill
condition of the five-component model—purportedly show that people
sometimes judge than an agent can perform an action intentionally even
if she does not exercise any salient skill in bringing it about. By my lights,
Guglielmo and Malle’s attacks on both findings are illuminating and
important. However, for present purposes, I am going to limit my attention
to their attempts to undermine the side-effect findings since these are the
ones that potentially cast the most doubt on both the simple view and the
causal theory of action. At the end of the day, if the latter two views are
to be viable as accounts of the folk concept of intentional action, Knobe’s
side-effect findings would need to be explained away—which is precisely
what Guglielmo and Malle try to accomplish. But since discussing all of
the data they marshal forward is beyond the scope of the present essay, I
will simply focus on the two studies that I take to pose the greatest poten-
tial threat to the Knobe effect.
In the first study, Guglielmo and Malle presented participants with ver-
sions of Knobe’s earlier CEO harm vignettes. However, unlike Knobe’s
earlier studies, which forced participants to make simple dichotomous
intentionality judgments, Guglielmo and Malle provided participants with
more answer choices in the hopes that this would yield greater insight into
how they actually viewed the CEO’s primary actions. Upon reading a
version of Knobe’s CEO harm vignette, participants were first once again
forced to make a yes/no intentionality judgment. Having done so, they
were then provided with multiple action descriptions to choose from and
asked to select “the most accurate description of what the CEO did.” The
choices were as follows:

1. “The CEO willingly harmed the environment.”

2. “The CEO knowingly harmed the environment.”
3. “The CEO intentionally harmed the environment.”
4. “The CEO purposely harmed the environment.”

Unsurprisingly, 73 percent of the participants answered the dichotomous

intentionality question in the affirmative—which is roughly what one
would expect based on Knobe’s earlier findings. Surprisingly, however,
across all four conditions of the study, only 1 percent of the participants
deemed that “intentionally harmed the environment” was the most accu-
rate description of what the CEO did. By contrast, 82 percent selected
286 T. Nadelhoffer

“knowingly harmed the environment,” and 14 percent selected “willingly

harmed the environment.” On the surface, these results put serious pres-
sure on the Knobe effect. After all, if the bidirectional model were correct,
one would expect that the participants in these studies would find it most
correct to say that the CEO “intentionally” or “purposely” harmed the
environment. But contrary to this expectation, these were the two least
popular options.
As Guglielmo and Malle point out, “this finding is particularly note-
worthy because the same participants who first provided the forced-
choice intentionality response afterwards picked a different label as the
most accurate behavior description” (n.d.a). On their view, these results
suggest that when participants “are freed from the forced dichotomous
choice and have a chance to select a more subtle interpretation of the
situation at hand,” their answers no longer provide evidence for the Knobe
effect (ibid.). In light of these findings, Guglielmo and Malle believe
they have both undermined the bidirectional model while at the same
time vindicating the valence-neutral five-component model. By their
lights, people do not normally conceptualize the CEO as having intention-
ally harmed the environment. Instead, when people are given the choice,
they view his behavior as “the intentional action of adopting a profit-
raising program while fully knowing (and not preventing) that it would
harm the environment.” But insofar as this is the case, the Knobe effect
starts to look like little more than an artifact of an impoverished experi-
mental design. Indeed, the results of a follow-up study appear to further
undermine the purported evidence for the bidirectional model of
intentionality.
In this second study, participants were once again presented with
Knobe’s CEO harm vignette. They were then told that “this situation has
certain ambiguities and leaves some questions open. How can we best
describe what the CEO did? Please pick the most accurate description . . .
then pick the second-most accurate description.” The action descriptions
available to participants this time around were as follows:

1. “The CEO intentionally harmed the environment.”

2. “The CEO intentionally adopted an environment harming program.”
3. “The CEO intentionally adopted an environment-harming program and
a profit-raising program.”
4. “The CEO intentionally adopted a profit-raising program that he knew
would harm the environment.”
5. “The CEO intentionally adopted a profit-raising program.”
The Causal Theory of Action and the Knobe Effect 287

As Guglielmo and Malle once again correctly point out, if the bidirectional
model of intentionality were correct, one would expect participants to
select the first and second action descriptions as the most accurate.
However, as was the case in the earlier study, only a minority of partici-
pants found these to be the “most accurate descriptions.” In fact, the first
two descriptions were selected as either the most or second most accurate
by only 16 percent of the participants. Conversely, 83 percent of the par-
ticipants deemed the fourth action description to be either the most or the
second most accurate, while 20 percent selected the third description. In
short, the results of this study provide evidence that “when people were
asked to indicate which action the CEO performed intentionally, they
showed striking agreement: He intentionally adopted a profit-raising
program that he knew would harm the environment; he did not intention-
ally harm the environment” (ibid.). Guglielmo and Malle believe these
findings lay to rest the bidirectional model of intentionality favored by
Knobe and others. On their view, participants in earlier studies only judged
that the CEO harmed the environment intentionally because they were
forced to give a dichotomous intentionality judgment. However, once
participants are given more choices, their intuitions no longer seem to
provide any evidence for the bidirectional model. On the surface, these
new findings admittedly pose a serious challenge to the Knobe effect.
However, I don’t think things are as straightforward as Guglielmo and
Malle have assumed. By my lights, more work needs to be done before they
“disconfirm” the claim that moral considerations sometimes influence folk
ascriptions of intentional action.
Perhaps the biggest shortcoming of Guglielmo and Malle’s new studies
is that they don’t actually directly test the Knobe effect. To see why, keep
in mind that Knobe’s key finding was that participants’ intuitions were
different in the harm condition than they were in the help condition—that
is, participants judged bad side effects to be more intentional than good
side effects. In order to fully explain away these findings, Guglielmo and
Malle would have needed to include both harm and help conditions—
something they failed to do. After all, even if their participants were admit-
tedly less inclined than Knobe’s participants to judge that the CEO harmed
the environment intentionally, it is still possible that had they been pre-
sented with a help condition, their ratings of intentionality would have
been even lower still. In short, Knobe’s findings were relational. It wasn’t
just that participants in the harm condition overwhelmingly judged
that the CEO intentionally harmed the environment. It was also that those
in the help condition overwhelmingly judged that the CEO did not
288 T. Nadelhoffer

intentionally help the environment. To fully undermine this finding, Gug-

lielmo and Malle would have needed to run both conditions.
Another worry arises with respect to the “multiple action description”
method Guglielmo and Malle adopted. In short, just because participants
did not select “intentionally harmed the environment” as the most accu-
rate way of describing the CEO’s behavior, it doesn’t follow that they
judged that the CEO did not harm the environment intentionally. All that
follows is that participants thought it sounded more correct to state that
the CEO knowingly harmed the environment. But insofar as knowing
that doing x will bring about y is a necessary condition for intentionally
bringing about y, it is unsurprising that participants may have deemed that
it is more accurate to say that the CEO knowingly harmed the environment
than it is to say that he intentionally harmed the environment. Consider,
for instance, the following example: “Paige went to the beach to play vol-
leyball. She ended up playing volleyball all afternoon.” Now, imagine I
gave you the following list of action descriptions and asked you to specify
which one sounded the “most accurate”:

1. “Paige willingly played volleyball.”

2. “Paige knowingly played volleyball.”
3. “Paige intentionally played volleyball.”
4. “Paige purposely played volleyball.”

If forced to select one of these as the most accurate, I suspect some people
might judge that it sounds most accurate to say that she knowingly or
willingly played volleyball. After all, both are presumably necessary condi-
tions of intentionally or purposely playing volleyball. But this would not
mean that they do not also think it is accurate to say that Paige played
volleyball intentionally or purposely. Conversely, even if the majority of
people selected “intentionally” (or “purposely”) as the most accurate
description of Paige’s behavior, it wouldn’t follow that they think it is in
inaccurate to say that she played volleyball willingly and knowingly. By
my lights, if one truly wanted to get at the salient intuitions in this case,
one would need to provide participants with conjunctive choices as well
such as:

5. “Paige willingly, knowingly, intentionally, and purposely played

volleyball.”

Had Guglielmo and Malle provided participants with both individuated

and conjunctive action descriptions, they may very well have gotten results
that were more in line with Knobe’s earlier findings. But to the extent to
The Causal Theory of Action and the Knobe Effect 289

which they failed to do so, they cannot rule out the possibility that par-
ticipants judged that the CEO harmed the environment intentionally.
Of course, this criticism admittedly only applies to the first aforemen-
tioned study from Guglielmo and Malle. Their second study, on the other
hand, is immune to this worry since it provides participants with precisely
the kind of more fine-grained action descriptions I was lobbying for above.
However, a closer look at the actual descriptions they used raises additional
worries. To see why, consider once again the choices that were available
to participants:

1. “The CEO intentionally harmed the environment.”

2. “The CEO intentionally adopted an environment-harming program.”
3. “The CEO intentionally adopted an environment-harming program and
a profit-raising program.”
4. “The CEO intentionally adopted a profit-raising program that he knew
would harm the environment.”
5. “The CEO intentionally adopted a profit-raising program.”

By my lights, it is unsurprising that the majority of participants selected

the third and fourth action descriptions as the most accurate since these
contain more detailed and nuanced information than the other choices.
As such, it doesn’t follow from the fact that these were the two most
popular choices that participants would not have judged that the CEO
harmed the environment intentionally had they been provided with a
more detailed description. In order to fully undermine the Knobe effect on
this front, Guglielmo and Malle would have needed to provide participants
with the option of selecting an “intentionally harmed the environment”
action description that was just as detailed as the other lengthier choices.
Consider, for instance, the following:

6. “The CEO intentionally adopted a profit-raising program that he knew

would harm the environment. Hence, the CEO intentionally harmed the
environment.”

By providing participants with fine-grained action descriptions for the

“knowingly but not intentionally” view without at the same time provid-
ing them with similarly fine-grained descriptions of the “knowingly and
intentionally” view, Guglielmo and Malle may have unwittingly stacked
the deck against the Knobe effect. As such, I believe that it is still too early
in the day to proclaim a victory for the valence-neutral model of inten-
tionality. Of course, determining whether the worries I have just raised
about Guglielmo and Malle’s studies are justified is an empirical affair.
290 T. Nadelhoffer

Accordingly, I ran two new studies that were designed to put these worries
to the test.

5 The Knobe Effect Revisited

Before I present my latest findings, I should first briefly summarize what

I took to be the three main shortcomings of Guglielmo and Malle’s
aforementioned studies. First, they only used a harm condition rather
than using both harm and help conditions. As such, they left open the
possibility that participants could nevertheless have displayed the kind of
asymmetrical intuitions identified earlier by Knobe. Second, the partici-
pants in Guglielmo and Malle’s first study were given individuated action
descriptions without also being given conjunctive action descriptions. As
a result, it is quite possible that participants judged that the CEO harmed
the environment intentionally even if they did not select this as the most
accurate action description. Finally, the participants in Guglielmo and
Malle’s second study were given fine-grained action descriptions for the
“knowingly but not intentionally” view but they were not provided with
similarly fine-grained descriptions of the “knowingly and intentionally”
view.
In light of these worries, I ran two new studies. In the first one, partici-
pants were 162 people who completed an online questionnaire.11 Each
participant was randomly assigned to one of two conditions—a harm
condition (83 participants) or a help condition (79 participants). The two
conditions were as follows:

The vice-president of a company went to the chairman of the board and said, “We
are thinking of starting a new program. It will help us increase profits, but it will
also harm [help] the environment.” The chairman of the board answered, “I don’t
care at all about harming [helping] the environment. I just want to make as much
profit as I can. Let’s start the new program.” They started the new program. Sure
enough, the environment was harmed [helped].
This situation has certain ambiguities and leaves some questions open. How can
we best describe what the CEO did? Please circle the number of the most accurate
description:

1. The CEO intentionally adopted a profit-raising program that he knew would

harm [help] the environment. Since he knew the program would harm [help] the
environment and he intentionally adopted the program, the CEO intentionally
harmed [helped] the environment.
2. The CEO intentionally adopted a profit-raising program that he knew would
harm [help] the environment. Even though he knew the program would harm [help]
The Causal Theory of Action and the Knobe Effect 291

the environment and he intentionally adopted the program, the CEO did not inten-
tionally harm [help] the environment.

In the harm condition, 77 percent of the participants judged that the CEO
intentionally harmed the environment whereas only 23 percent judged
that he did not intentionally harm the environment. In the help condi-
tion, on the other hand, only 19 percent of the participants judged that
the CEO intentionally helped the environment whereas 81 percent judged
that he did not intentionally help the environment. These results—which
are in line with Knobe’s earlier results—are statistically significant ( p < 0.01,
FET).
On the surface, these findings put pressure on the explanation of
Knobe’s studies put forward by Guglielmo and Malle. After all, on their
view, people find it most natural to say that the CEO harmed the environ-
ment knowingly but not intentionally. However, at least with respect to
this first study, participants were far more likely to judge that the CEO
knowingly and intentionally harmed the environment than they were to
say that he knowingly but did not intentionally harm it. Conversely, par-
ticipants were far more likely to say that the CEO knowingly but did not
intentionally help the environment than they were to say that he inten-
tionally helped it. These results are precisely what one would expect based
on the earlier side-effect studies run by Knobe and others.
Of course, one might worry that I may have unwittingly stacked the
cards in favor of the Knobe effect given the specific wording of the two
choices. For instance, in the harm condition, the first answer actually
explains why one might deem the side effect intentional—namely, the CEO
intentionally adopted a program that he knew would harm the environ-
ment. The second answer, on the other hand, did not provide participants
with the reason why one might judge that the CEO did not intentionally
harm the environment—namely, that he did not intend to do so. Instead,
the second option once again provided participants with the same reason
for deeming the side effect intentional that the first option provided. By my
lights, this is a legitimate concern. Consequently, I ran another study that
was designed to correct for it.
In the second study, participants were 130 people who completed an
online questionnaire.12 Each was randomly assigned to one of two condi-
tions—a harm condition (59 participants) or a help condition (71 partici-
pants). Once again, the wording of the two respective vignettes—which
were based on the original Knobe CEO studies—was the same but for the
moral valence of the outcome. The two conditions were as follows:
292 T. Nadelhoffer

The vice-president of a company went to the chairman of the board and said, “We
are thinking of starting a new program. It will help us increase profits, but it will
also harm [help] the environment.” The chairman of the board answered, “I don’t
care at all about harming [helping] the environment. I just want to make as much
profit as I can. Let’s start the new program.” They started the new program. Sure
enough, the environment was harmed [helped].
This situation has certain ambiguities and leaves some questions open. How can
we best describe what the CEO did? Please circle the number of the most accurate
description:

1. The CEO intentionally adopted a profit-raising program that he knew would

harm [help] the environment. Since he knew the program would harm [help] the
environment and he intentionally adopted the program, the CEO intentionally
harmed [helped] the environment.
2. The CEO intentionally adopted a profit-raising program that he knew would
harm [help] the environment. But since he did not intend to harm [help] the envi-
ronment, the CEO did not intentionally harm [help] the environment.

In the harm condition, 88 percent of the participants judged that the CEO
intentionally harmed the environment whereas only 12 percent of the
participants judged that the CEO did not intentionally harm the environ-
ment. In the help condition, on the other hand, only 25 percent judged
that the CEO intentionally helped the environment whereas 75 percent
judged that the CEO did not intentionally help the environment. Once
again these results—which are statistically significant (p < 0.01, FET)—are
perfectly in line with Knobe’s earlier findings. Moreover, they put further
pressure on the explanation of the Knobe effect that was put forward by
Guglielmo and Malle.
Keep in mind that on their view, when participants are provided with
the option of saying that the CEO knowingly harmed the environment
but neither intended nor intentionally harmed it, they will not judge that
the CEO harmed the environment intentionally. However, in my second
study, participants were provided with both options but they nevertheless
overwhelmingly preferred to say that the CEO both knowingly and inten-
tionally harmed the environment. As such, it is unclear that Guglielmo
and Malle have accomplished their goal of shielding the standard unidi-
rectional view from Knobe’s side-effect findings. Indeed, in both of my two
latest studies, participants’ responses provide new support for the bidirec-
tional view while further challenging both the simple view and the causal
theory of action. In light of these latest findings, I minimally believe that
I have shown that more work still needs to be done before either side in
this ongoing debate can claim a decisive victory. As it stands, I believe that
The Causal Theory of Action and the Knobe Effect 293

the puzzling nature of the Knobe effect remains something that needs to
be either further explained or further explained away.
That being said, it is worth mentioning that there are several lingering
shortcomings with my two latest studies. First, as Guglielmo and Malle
point out, nearly all of the studies on the folk concept of intentionality
that have been run thus far share the limitation of using a vignette design.
Mine are no different in this respect. I find their suggestion of using mock
jury designs and visual stimuli very intriguing. The results of these kinds
of studies would obviously shed important new light on the nature of folk
intentionality judgments. Second, I did not collect data on the cognitive
timing of my participants’ responses to the vignettes. Guglielmo and Malle
are apparently already running some reaction time studies, and I very
much look forward to seeing the fruits of their labors. Finally, and most
importantly, my studies admittedly did not address the “pro-attitude”
hypothesis put forward by Guglielmo and Malle to partly explain the
harm–help asymmetry. On their view, not caring about harming the envi-
ronment is not on par with not caring about helping it. I entirely agree.
Indeed, I voiced the same worry about Knobe’s original vignettes in Nadel-
hoffer 2004b. However, for present purposes, my main goal was to test
Guglielmo and Malle’s “knowingly but not intentionally” explanation of
the Knobe effect. Further testing their “pro-attitude” hypothesis is a task
for another day. Hopefully, psychologists and philosophers will continue
to work together on these issues in an effort to better understand the
nature of our intuitions and beliefs concerning intentional action.

6 Conclusion

In this chapter, I set out to provide an overview of one of the more hotly
contested recent debates in the philosophy of action. More specifically, I
wanted to show that Guglielmo and Malle’s recent attempt to lay the
Knobe effect to rest falls short even if they have admittedly taught us many
important lessons along the way. By my lights, it is clear that they have
thrown down the gauntlet in defense of the standard model of intentional-
ity (and the causal theory of action more generally). Their studies are both
more sophisticated and powerful than the previous research that has been
done on this front. As such, any subsequent work on the folk concept of
intentional action must carefully take Guglielmo and Malle’s insightful
findings into account. Unfortunately, I was admittedly only able to scratch
the surface of their research in this chapter. In the future, I hope to give
their work the further attention it deserves. For now, I have merely tried
294 T. Nadelhoffer

to take a few small steps toward better understanding the still puzzling
nature of the folk concept of intentionality. Whether people think I pushed
the debate forward or backward remains to be seen.

Notes

1. One of the paradigmatic papers defending this view is Davidson 1963/1980.

2. In this chapter, when I use the term “intentionality” I am using it as shorthand

for “intentional action.” My usage is not to be confused with Brentano’s notion of
intentionality as the mind’s “direction towards an object” (Brentano 1874) even if
the two are admittedly related in some respects.

3. Though it need not ultimately concern us here, it is nevertheless worth pointing

out that Guglielmo and Malle misconstrue Knobe’s position throughout their work.
In short, Knobe claims that the badness of the action—but not the blameworthiness
of the agent—sometimes influences people’s intentionality judgments. Moreover,
he does not think that this influence ought to be viewed as a bias. Rather, Knobe
claims that the folk concept of intentionality has an inherent moral component
that gets implicated when people use the concept competently. My view, on the
other hand, is that (a) blameworthiness as well as moral badness can influence folk
ascriptions of intentional action, and (b) this influence represents an affectively
driven bias or performance error. As such, Guglielmo and Malle (n.d.a) mistakenly
lump our views together when they claim that we both believe that “intentionality
judgments are biased by evaluative considerations, in particular by feelings or judg-
ments of blame.” In light of this mistake, many of their objections to Knobe’s view
are more directly objections to my own. As such, some of the evidence that they
provide against what they call the “blame→intentionality” view fails to undermine
Knobe’s view even if this evidence admittedly puts pressure on mine. By my lights,
the only general way to refer to both Knobe’s view and my own without obscuring
important differences between the two is to say that we both believe that moral
considerations sometimes influence folk judgments of intentionality. Which moral
considerations have this effect and whether this influence represents a bias or
mistake are issues that are hotly disputed.

4. Two points of clarification are in order at this point. First, just because I think
that data concerning folk intuitions are relevant to some philosophical debates—e.g.,
free will—it does not follow that I believe that these data are relevant to all philo-
sophical debates—e.g., mereology. Second, just because I think that folk intuitions
are relevant to some philosophical problems, it does not follow that I believe they
solve these problems. To my knowledge, no experimental philosopher tries to move
from “the folk think that x is the case” to “x is the case.” Instead, to the extent that
the folk intuitions are philosophically relevant, they serve as starting points and
constraints to philosophical investigation and not final arbiters of philosophical
truth.
The Causal Theory of Action and the Knobe Effect 295

5. For an overview of recent work in experimental philosophy, see Nadelhoffer and

Nahmias 2007. For now, it is worth pointing out that the boundary between experi-
mental philosophy and social psychology is admittedly blurry. This is partly because
some scientists are interested in the same kind of intuitions that interest philoso-
phers, and they, too, discuss the philosophical implications of their research—see,
e.g., Carlsmith and Darley 2008; Damasio 1994; Darley, Carlsmith, and Robinson
2000; Greene 2003, 2007; Haidt 2001, 2003; Hauser 2006; Malle 2001, 2006; Nisbett
2003; Nisbett and Ross 1980; Wegner 2002. Because “experimental philosophy” is
perhaps best viewed as a family resemblance term, trying to explicate the movement
in terms of necessary and jointly sufficient conditions would be wrongheaded even
if well intentioned.

6. See, e.g., Adams and Steadman 2004a,b; Cushman and Mele 2008; Feltz and
Cokely 2007; Hindriks 2008; Knobe 2003a,b, 2004a,b), 2005a,b; Knobe and Burra
2006a,b; Knobe and Mendlow 2004; Leslie, Knobe, and Cohen 2006; Machery 2008;
Malle 2001, 2006; Malle and Knobe 1997; Mallon 2008; McCann 2006; Meeks 2004;
Nadelhoffer 2004a,b,c, 2005, 2006a,b,c; Nado forthcoming; Nanay forthcoming;
Nichols and Ulatowski 2007; Phelan and Sarkissian 2008, 2009; Turner 2004; Wiland
2007; Wright and Bengson 2009; Young et al. 2006.

7. Indeed, McCann has actually run some studies of his own on the folk concept
of intentional action (McCann 2006). For my response to his interpretation of the
data he collected, see Nadelhoffer 2006c.

8. See, e.g., Mele 1992; Mele and Moser 1994; Mele and Sverdlik 1996; and Malle
and Knobe 1997.

9. For two recent overviews of the action theory literature in experimental philoso-
phy, see Feltz 2008 and Nado forthcoming.

10. Proponents of this view defend it on a number of grounds. First and foremost,
the simple view purportedly captures our pretheoretical intuitions and coheres with
our ordinary usage of the concepts of intending and intentional action (McCann
1998, 210). After all, in ordinary contexts it would admittedly sound strange for me
to say that I dialed my friend’s phone number intentionally even though I did not
intend to do so. Second, given that the SV is the seemingly uncontroversial claim
that intending to x is necessary for intentionally x-ing, the view has the virtue of being,
well, simple or “uncluttered” (Adams 1986a, 284). Third, it “gives us reason to
believe that our intentions causally guide our actions in virtue of their content”
(ibid.)—thereby supporting our ordinary view of ourselves whereby the contents of
our intentions to x play an important role in our intentionally x-ing.

11. Participants were recruited via this Web site: http://www.philosophical-

personality.com/.

12. Participants were recruited via this Web site: http://www.philosophical-

personality.com/.
References

Ackrill, J. 1978. Aristotle on action. Mind 87:595–601.

Adams, F. Manuscript. Embodied cognition.

Adams, F. 1986a. Intention and intentional action: The simple view. Mind and Lan-
guage 1:281–301.

Adams, F. 1986b. Feedback about feedback: Reply to Ehring. Southern Journal of

Philosophy 24:123–131.

Adams, F. 1994a. Of epicycles and elegance. Canadian Journal of Philosophy

24:627–636.

Adams, F. 1994b. Trying, desire, and desiring to try. Canadian Journal of Philosophy
24:613–626.

Adams, F. 1997. Cognitive trying. In Contemporary Action Theory, vol. 1, ed. G.

Holmstrom-Hintikka and R. Tuomela. Dodrecht: Kluwer.

Adams, F. 2003a. Thoughts and their contents: Naturalized semantics. In The Black-
well Guide to the Philosophy of Mind, ed. T. Warfied and S. Stich. Oxford: Blackwell.

Adams, F. 2003b. The informational turn in philosophy. Minds and Machines

13:471–501.

Adams, F. 2007. Trying with the hope. In Rationality and the Good, ed. M. Timmons,
J. Greco, and A. Mele. Oxford: Oxford University Press.

Adams, F., and K. Aizawa. 2008. The Bounds of Cognition. Oxford: Blackwell.

Adams, F., Barker, J., and J. Figurelli. Manuscript. Towards closure on closure.

Adams, F., and A. Mele. 1989. The role of intention in intentional action. Canadian
Journal of Philosophy 19:511–532.

Adams, F., and A. Mele. 1992. The intention/volition debate. Canadian Journal of
Philosophy 22:323–338.
298 References

Adams, F., and A. Steadman. 2004a. Intentional action in ordinary language: Core
concept or pragmatic understanding. Analysis 74:173–181.

Adams, F., and A. Steadman. 2004b. Intentional actions and moral considerations:
Still pragmatic. Analysis 74:264–267.

Aguilar, J. 2007. Interpersonal interactions and the bounds of agency. Dialectica

61:219–232.

Aguilar, J., and A. Buckareff. 2009. Agency, consciousness, and executive control.
Philosophia 37:21–30.

Alvarez, M. 2005. Agents, actions, and reasons. Philosophical Books 46:45–58.

Alvarez, M., and J. Hyman. 1998. Agents and their actions. Philosophy 73:219–245.

Annas, J. 1978. How basic are basic actions? Proceedings of the Aristotelian Society
78:195–213.

Anscombe, G. E. M. 1963. Intention. Ithaca, N.Y.: Cornell University Press.

Anscombe, G. E. M. 1966. A note on Mr. Bennett. Analysis 26:208.

Anscombe, G. E. M. 1989. Von Wright on practical inference. In The Philosophy of

Georg Henrik Von Wright, P. A. Schilpp and L. E. Hahn, eds., La Salle, Ill.: Open Court.
Reprinted in Human Life, Action, and Ethics, ed. M. Geach and L. Gormally (Exeter:
Imprint Academic, 2005).

Apperly, I. A., and S. A. Butterfill. 2008. Do humans have two systems to track beliefs
and belief-like states? Unpublished manuscript, Department of Psychology, Univer-
sity of Birmingham, UK.

Aquinas, T. 1999. On Human Nature. Ed. T. Hibbs. Indianapolis: Hackett.

Aristotle. 1934. The Nicomachean Ethics with an English Translation by H. Rackham.

Cambridge, Mass.: Harvard University Press.

Aristotle. 1978. De motu animalium. Trans. M. Nussbaum. Princeton: Princeton Uni-

versity Press.

Aristotle. 1983. Physics, Books III and IV. Trans. E. Hussey. Oxford: Oxford University
Press.

Aristotle. 1999. Nicomachean Ethics. Trans. T. Irwin. Indianapolis: Hackett.

Armstrong, D. 1973. Acting and trying. Philosophical Papers 2:1–15. Reprinted in

Armstrong 1980.

Armstrong, D. 1975. Beliefs and desires as causes of actions: A reply to Donald

Davidson. Philosophical Papers 4:1–8.

Armstrong, D. M. 1978. A Theory of Universals: Universals and Scientific Realism, vol.

2. Cambridge: Cambridge University Press.
References 299

Armstrong, D. 1980. The Nature of Mind. Ithaca, N.Y.: Cornell University Press.

Armstrong, D. 1983. What Is a Law of Nature? Cambridge: Cambridge University

Press.

Armstrong, D. 1997. A World of States of Affairs. New York: Cambridge University

Press.

Armstrong, D. 1999. The open door: Counterfactual versus singularist theories of

causation. In Causation and Laws of Nature, ed. H. Stanken. London: Kluwer
Academic.

Astington, J. W. 1999. The language of intention: Three ways of doing it. In Develop-
ing Theories of Intention: Social Understanding and Self-Control, ed. P. D. Zelazo, J. W.
Astington, and D. R. Olson. Mahwah, N.J.: Erlbaum.

Astington, J. W. 2001. The paradox of intention: Assessing children’s metarepresen-

tational understanding. In Intentions and Intentionality: Foundations of Social Cogni-
tion, ed. B. F. Malle, L. J. Moses, and D. A. Baldwin. Cambridge, Mass.: MIT Press.

Audi, R. 1993. Intending. In his Action, Intention, and Reason. Ithaca, N.Y.: Cornell
University Press.

Baron-Cohen, S. 1992. Out of sight or out of mind? Another look at deception in

autism. Journal of Child Psychology and Psychiatry, and Allied Disciplines
33:1141–1155.

Barsalou, L. 1999. Perceptual symbols systems. Behavioral and Brain Sciences

22:577–660.

Bartsch, K., M. D. Campbell, and G. L. Troseth. 2007. Why else does Jenny run?
Young children’s extended psychological explanations. Journal of Cognition and
Development 8:33–61.

Bartsch, K., and H. M. Wellman. 1995. Children Talk about the Mind. Oxford: Oxford
University Press.

Beebee, H. 2004. Causing and nothingness. In Causation and Counterfactuals, ed. J.

Collins, N. Hall, and L. A. Paul. Cambridge, Mass.: MIT Press.

Bedau, M., and P. Humphreys, eds. 2008. Emergence: Contemporary Readings in the
Philosophy of Science. Cambridge, Mass.: MIT Press.

Belousov, B. P. 1959. A periodic reaction and its mechanism. Compilation of Abstracts

on Radiation Medicine 147:145.

Bennett, D. 1965. Action, reason, and purpose. Journal of Philosophy 62:85–96.

Bennett, J. 1988. Events and Their Names. Indianapolis: Hackett.

Bennett, J. 1995. The Act Itself. Oxford: Clarendon Press.

300 References

Bennett, J. 2008. Accountability (II). In Free Will and Reactive Attitudes: Perspectives
on P. F. Strawson’s “Freedom and Resentment,” ed. M. McKenna and P. Russell.
Farnham: Ashgate Press.

Bentham, J. 1996. An Introduction to the Principles of Morals and Legislation. Ed. J. H.

Burns and H. L. A. Hart. Oxford: Clarendon Press.

Berthoz, A., and J. L. Petit. 2006. Phenomenologie et physiologie de l’action. Paris: Odile
Jacob.

Bishop, J. 1989. Natural Agency: An Essay on the Causal Theory of Action. Cambridge:
Cambridge University Press.

Bishop, J. 2003. Prospects for a naturalist libertarianism: O’Connor’s Persons and

Causes. Philosophy and Phenomenological Research 66:228–243.

Bishop, J. 2007. Believing by Faith: An Essay in the Epistemology and Ethics of Religious
Belief. Oxford: Clarendon Press.

Bittner, R. 2001. Doing Things for Reasons. New York: Oxford University Press.

Blackburn, S. 1998. Ruling Passions. Oxford: Oxford University Press.

Borghi, A. 2005. Object concepts and action. In Grounding Cognition, ed. D. Pecher
and R. Zwaan. Cambridge: Cambridge University Press.

Brand, M. 1971. The language of not doing. American Philosophical Quarterly

8:45–53.

Brand, M. 1984. Intending and Acting: Toward a Naturalized Action Theory. Cambridge,
Mass.: MIT Press.

Bratman, M. 1984. Two faces of intention. Philosophical Review 93:375–405.

Bratman, M. 1987. Intention, Plans, and Practical Reason. Cambridge, Mass.: Harvard
University Press.

Bratman, M. 2000. Reflection, planning, and temporally extended agency. Philo-

sophical Review 109:35–61.

Bratman, M. 2001. Two problems about human agency. Proceedings of the Aristotelian
Society 101:309–332.

Brentano, F. 1874. Psychologie vom empirischen Standpunkt. Leipzig: Duncker &

Humblot.

Brentano, F. 1995. Psychology from an Empirical Standpoint. London: Routledge.

Brooks, R. 1991. Intelligence without representation. Artificial Intelligence

47:139–159.

Buccino, G., L. Riggio, G. Melli, F. Binkofski, V. Gallese, and G. Rizzolati. 2005.

Listening to action-related sentences modulates the activity of the motor system: A
References 301

combined TMS and behavioral study. Brain Research. Cognitive Brain Research
24:355–363.

Buckareff, A. 2007. Mental overpopulation and mental action: Protecting intentions

from mental birth control. Canadian Journal of Philosophy 37:49–66.

Buckareff, A., and J. Zhu. 2009. The primacy of the mental in the explanation of
human action. Disputatio 3(26):73–88.

Byrd, J. 2007. Moral responsibility and omissions. Philosophical Quarterly 57:56–67.

Campbell, J. 2002. Reference and Consciousness. Oxford: Oxford University Press.

Campbell, J. 2007. An interventionist approach to causation in psychology. In

Causal Learning, ed. A. Gopnik and L. Schulz. Oxford: Oxford University Press.

Care, N. S., and C. Landesman, eds. 1968. Readings in the Theory of Action. Bloom-
ington, Ind.: Indiana University Press.

Carey, B. 2008. Anticipating the future to “see” the present. New York Times. June
10.

Carlsmith, K., and J. Darley. 2008. Psychological aspects of retributive justice.

Advances in Experimental Social Psychology 40:193–236.

Carpenter, M., J. Call, and M. Tomasello. 2002. A new false belief test for 36-month-
olds. British Journal of Developmental Psychology 20:393–420.

Chan, D. 1995. Non-intentional actions. American Philosophical Quarterly

32:139–151.

Charles, D. 1984. Aristotle’s Philosophy of Action. London: Duckworth.

Chasiotis, A., F. Kiessling, J. Hofer, and D. Campos. 2006. Theory of mind and
inhibitory control in three cultures: Conflict inhibition predicts false belief under-
standing in Germany, Costa Rica, and Cameroon. International Journal of Behavioral
Development 30:249–260.

Child, W. 1994. Causality, Interpretation, and the Mind. Oxford: Clarendon Press.

Chisholm, R. 1964. The descriptive element in the concept of action. Journal of

Philosophy 61:613–625.

Chisholm, R. 1966. Freedom and action. In Freedom and Determinism, ed. K. Lehrer.
New York: Random House.

Clark, A. 1997. Being There: Putting Brain, Body, and World Together Again. Cambridge,
Mass.: MIT Press.

Clark, A., and R. Grush. 1999. Towards a cognitive robotics. Adaptive Behavior
7:5–16.
302 References

Clarke, R. 1994. Ability and responsibility for omissions. Philosophical Studies

73:195–208.

Clarke, R. 2010. Intentional omissions. Noûs 44:158–177.

Clements, W. A., and J. Perner. 1994. Implicit understanding of belief. Cognitive

Development 9:377–397.

Collins, J., N. Hall, and L. A. Paul, eds. 2004. Causation and Counterfactuals. Cam-
bridge, Mass.: MIT Press.

Coope, U. 2004. Aristotle’s account of agency in Physics III.3. Boston Area Colloquium
in Ancient Philosophy 20:201–221.

Coope, U. 2007. Aristotle on action. Proceedings of the Aristotelian Society (suppl. vol.)
81:109–138.

Csibra, G., and G. Gergely. 1998. The teleological origins of mentalistic action
explanations: A developmental hypothesis. Developmental Science 1:255–259.

Cushman, F., and A. Mele. 2008. Intentional action: Two-and-a-half folk concepts?
In Experimental Philosophy, ed. J. Knobe and S. Nichols. New York: Oxford University
Press.

Damasio, A. 1994. Descartes’ Error: Emotion, Reason, and the Human Brain. London:
Penguin.

Dancy, J. 2000. Practical Reality. Oxford: Oxford University Press.

Danto, A. C. 1963. What we can do. Journal of Philosophy 60:434–445.

Darley, J., K. Carlsmith, and P. Robinson. 2000. Incapacitation and just deserts as
motives for punishment. Law and Human Behavior 24:659–683.

Davidson, D. 1963. Actions, reasons, and causes. Journal of Philosophy 60:685–700.

Reprinted in Davidson 1980.

Davidson, D. 1967. Causal relations. Journal of Philosophy 64. Reprinted in Davidson

1980.

Davidson, D. 1971. Agency. In Agent, Action, and Reason, ed. R. Binkley et al. Toronto:
University of Toronto Press. Reprinted in Davidson 1980.

Davidson, D. 1973. Freedom to act. In Essays on Freedom of Action, ed. Ted Hond-
erich. London: Routledge & Kegan Paul. Reprinted in Davidson 1980.

Davidson, D. 1976. Hempel on explaining action. Erkenntnis 10(3). Reprinted in

Davidson 1980.

Davidson, D. 1978. Intending. In Philosophy of History and Action, ed. Y. Yovel. Dor-
drecht: D. Reidel. Reprinted in Davidson 1980.
References 303

Davidson, D. 1980. Essays on Actions and Events. Oxford: Oxford University Press.

Davidson, D. 1985. Reply to Bruce Vermazen. In Essays on Davidson, ed. B. Vermazen

and M. Hintikka. New York: Oxford University Press.

Davidson, D. 1987. Problems in the explanation of action. In Metaphysics and Moral-

ity: Essays in Honour of J. J. C. Smart, ed. P. Pettit, R. Sylvan, and J. Norman. Oxford:
Blackwell. Reprinted in Davidson 2004.

Davidson, D. 2001a. Essays on Actions and Events, 2nd ed. Oxford: Oxford University
Press.

Davidson, D. 2001b. The emergence of thought. In his Subjective, Intersubjective,

Objective. Oxford: Oxford University Press.

Davidson, D. 2004. Problems of Rationality. Oxford: Clarendon Press.

Dennett, D. C. 1979. Brainstorms: Philosophical Essays on Mind and Psychology.

Hassocks, Sussex: Harvester Press.

Dennett, D. C., and K. Lambert. 1978. The Philosophical Lexicon. Privately printed.

Doherty, M. J. 2009. Theory of Mind: How Children Understand Others’ Thoughts and
Feelings. Hove: Psychology Press.

Donagan, A. 1987. Choice: The Essential Element in Human Action. London:

Routledge.

Dowe, P. 2000. Physical Causation. Cambridge: Cambridge University Press.

Dretske, F. 1981. Knowledge and the Flow of Information. Cambridge, Mass.: MIT Press.

Dretske, F. 1988. Explaining Behavior: Reasons in a World of Causes. Cambridge, Mass.:

MIT Press.

Duff, A. 2004. Action, the act requirement, and criminal liability. In Agency and
Action, ed. J. Hyman and H. Steward. Cambridge: Cambridge University Press.

Earley, J. 1981. Self-organization and agency: In chemistry and process philosophy.

Process Studies 11:242–258.

Ellis, G. F. R. 2007. On the nature of causation in complex systems. Royal Society of

South Africa. http://www.sabinet.co.za/abstracts/royalsa/royalsa_v63_n1_a6.xm.

Enç, B. 2003. How We Act: Causes, Reasons, and Intentions. Oxford: Oxford University
Press.

Enç, B., and F. Adams. 1992. Functions and goal-directedness. Philosophy of Science
59:635–654.

Farahany, N. 2009. The interface between freedom and agency. Stanford Technology
Review. http://www.stlr.stanford.edu.
304 References

Feinberg, J. 1984. Harm to Others: The Moral Limits of the Criminal Law, vol. 1. New
York: Oxford University Press.

Feltz, A. 2008. The Knobe effect: A brief overview. Journal of Mind and Behavior
28:265–278.

Feltz, A., and E. Cokely. 2007. An anomaly in intentional action ascription: More
evidence of volk diversity. In Proceedings of the 29th Annual Meeting of Cognitive
Science Society, ed. D. S. McNamara and G. Trafton. Mahwah, N.J.: Lawrence Erlbaum.

Fine, K. 1982. First order modal theories III—Facts. Synthese 53:43–122.

Fischer, J. M. 1985–1986. Responsibility and failure. Proceedings of the Aristotelian

Society 86:251–270.

Fischer, J. M., and M. Ravizza. 1998. Responsibility and Control: A Theory of Moral
Responsibility. Cambridge: Cambridge University Press.

Fletcher, G. 1994. On the moral irrelevance of bodily movements. University of

Pennsylvania Law Review 142:1443–1453.

Frankfurt, H. 1969. Alternate possibilities and moral responsibility. Journal of

Philosophy 66: 829–839. Reprinted in Frankfurt 1988.

Frankfurt, H. 1978. The problem of action. American Philosophical Quarterly 15:

157–162. Reprinted in Frankfurt 1988. Page numbers refer to the latter edition.

Frankfurt, H. 1988. The Importance of What We Care About. New York: Cambridge
University Press.

Frankfurt, H. 1998. Necessity, Volition, and Love. New York: Cambridge University
Press.

Freeland, C. 1985. Aristotelian actions. Noûs 19:397–414.

Freeman, W. 1991. The physiology of perception. Scientific American

(February):78–85.

Gaillard, R. 2009. “Consciousness signature” discovered spanning the brain. New

Scientist 21:7.

Gaillard, R., S. Dehaene, C. Adam, S. Clémenceau, D. Hasboun, et al. 2009. Converg-

ing intracranial markers of conscious access. PLoS Biology 7:472–492.

Gallagher, S. 2005. How the Body Shapes the Mind. Oxford: Oxford University Press.

Gallagher, S. 2008. Are minimal representations still representations? International

Journal of Philosophical Studies 16:351–369.

Gallese, V., L. Fadiga, L. Fogassi, and G. Rizzolati. 1996. Action recognition in pre-
motor cortex. Brain 119:593–609.
References 305

Gatlin, L. 1972. Information and the Living System. New York: Columbia University
Press.

Gibson, J. 1979. The Ecological Approach to Visual Perception. Boston:

Houghton-Mifflin.

Ginet, C. 1990. On Action. Cambridge: Cambridge University Press.

Ginet, C. 2002. Reasons explanations of action: Causalist versus noncausalist

accounts. In The Oxford Handbook of Free Will, ed. R. Kane. New York: Oxford Uni-
versity Press.

Ginet, C. 2004. Intentionally doing and intentionally not doing. Philosophical Topics
32:95–110.

Glenberg, A., and M. Kaschak. 2002. Grounding language in action. Psychonomic

Bulletin and Review 9:558–565.

Goetz, S. 1988. A noncausal theory of agency. Philosophy and Phenomenological

Research 49:303–316.

Goldman, A. 1970. Theory of Human Action. Englewood Cliffs, N.J.: Prentice-Hall.

Goldie, P. 2000. The Emotions. Oxford: Oxford University Press.

Goodale, M. 2004. Perceiving the world and grasping it: Dissociations between
conscious and unconscious visual processing. In The Cognitive Neurosciences III, ed.
M. Gazzaniga. Cambridge, Mass.: MIT Press.

Gopnik, A., and A. N. Meltzoff. 1997. Word, Thoughts, and Theories. Cambridge,
Mass: MIT Press.

Goodale, M. A., and A. D. Milner. 1992. Separate visual pathways for perception
and action. Trends in Neurosciences 15:20–25.

Gratch, G. 1964. Response alternation in children: A developmental study of orien-

tations to uncertainty. Vita Humana 7:49–60.

Greene, J. 2003. From neural “is” to moral “ought”: What are the moral implications
of neuroscientific moral psychology? Neuroscience 4:847–850.

Greene, J. 2007. The secret joke of Kant’s soul. In Moral Psychology, vol. 3: The Neu-
roscience of Morality: Emotion, Disease, and Development, ed. W. Sinnott-Armstrong.
Cambridge, Mass.: MIT Press.

Grush, R. 1997. Yet another design for a brain? Review of Port and van Gelder (eds.),
Mind as Motion. Philosophical Psychology 10:233–242.

Grush, R. 2004. The emulation theory of representation: Motor control, imagery,

and perception. Behavioral and Brain Sciences 27:377–442.
306 References

Guglielmo, S. and B. F. Malle (n.d.a). Can unintended side effects be intentional?

Solving a puzzle in people’s judgments of intentionality and morality.

Guglielmo, S. and B. F. Malle (n.d.b). Enough skill to kill: Intentional control and
the judgment of immoral actions. University of Oregon.

Haddock, A. 2005. At one with our actions, but at two with our bodies: Hornsby’s
account of action. Philosophical Explorations 8:157–172.

Haggard, P., and M. Elmer. 1999. On the relations between brain potentials and the
awareness of voluntary movements. Experimental Brain Research 126:128–133.

Haidt, J. 2001. The emotional dog and its rational tail: A social intuitionist approach
to moral judgment. Psychological Review 108:814–834.

Haidt, J. 2003. The emotional dog does learn new tricks: A reply to Pizarro and
Bloom. Psychological Review 110:197–198.

Hall, L., and P. Johansson. 2009. Choice blindness: You don’t know what you want.
New Scientist 18:26–27.

Hall, N. 2004. Two concepts of causation. In Causation and Counterfactuals, ed. J.

Collins, N. Hall, and L. A. Paul. Cambridge, Mass.: MIT Press.

Happé, F., and E. Loth. 2002. “Theory of mind” and tracking speakers’ intentions.
Mind and Language 17:24–36.

Harman, G. 1976. Practical reasoning. Review of Metaphysics 79:431–463. Reprinted

in Mele 1997.

Harnad, S. 1990. The symbol grounding problem. Physica D. Nonlinear Phenomena

42:335–346.

Hart, H. L. A., and A. Honoré. 1959. Causation and the Law. Oxford: Oxford Univer-
sity Press.

Hauk, O., I. Johnsrude, and F. Pulvermüeller. 2004. Somatatopic representation of

action words in human motor and premotor cortex. Neuron 41:301–307.

Hauser, M. D. 2006. Moral Minds. New York: HarperCollins.

Haynes, J., K. Sakai, G. Rees, S. Gilbert, C. Frith, and R. E. Passingham. 2007. Reading
hidden intentions in the human brain. Current Biology 17:323–328.

Hempel, C. G. 1961. Rational action. In Proceedings and Addresses of the American

Philosophical Association, vol. 35. Reprinted in N. S. Care and C. Landesman 1968,
285–286.

Higginbotham, J. 2000. On events in linguistic semantics. In Speaking of Events, ed.

J. Higginbotham, F. Pianesi, and A. Varzi. Oxford: Oxford University Press.

Hindriks, F. 2008. Intentional action and the praise–blame asymmetry. Philosophical

Quarterly 58:630–641.
References 307

Hinton, G., and T. Shallice. 1991. Lesioning an attractor network: Investigations of

acquired dyslexia. Psychological Review 98:74–95.

Hobbes, T. [1651] 1994. Leviathan. Indianapolis: Hackett.

Hobbes, T. [1654] 1999. Hobbes’s Treatise Of Liberty and Necessity. In Hobbes and
Bramhall on Liberty and Necessity, ed. V. Chappell. New York: Cambridge University
Press.

Hobbes, T. [1656] 1999. Selections from Hobbes, The Questions Concerning Liberty,
Necessity, and Chance. In Hobbes and Bramhall on Liberty and Necessity, ed. V. Chap-
pell. New York: Cambridge University Press.

Holmes, O. W., Jr. 1963. The Common Law. Boston: Little, Brown.

Hornsby, J. 1980. Actions. London: Routledge & Kegan Paul.

Hornsby, J. 1993. Agency and causal explanation. In Mental Causation, ed. J. Heil
and A. Mele. Oxford: Clarendon Press.

Hornsby, J. 1997. Postscript: A disjunctive conception of bodily movements. In J.

Hornsby, Simple Mindedness: Essays in Defence of Naïve Naturalism in the Philosophy
of Mind, 102–110. Cambridge, Mass.: Harvard University Press.

Hornsby, J. 2004. Agency and actions. In Agency and Action, ed. J. Hyman and H.
Steward, 1–23. Cambridge: Cambridge University Press.

Hornsby, J. 2008. A disjunctive conception of acting for reasons. In Disjunctivism,

ed. A. Haddock and F. Macpherson. Oxford: Oxford University Press.

Hughes, C., and J. Dunn. 1997. Pretend you didn’t know: Preschoolers’ talk about
mental states in pretend play. Cognitive Development 12:381–403.

Hughes, C., and J. Dunn. 1998. Understanding mind and emotion: Longitudinal
associations with mental-state talk between young friends. Developmental Psychology
34:1026–1037.

Hume, D. [1748] 2007. An Enquiry Concerning Human Understanding. Oxford: Oxford

University Press.

Hume, D. [1777] 1975. Enquiries Concerning Human Understanding and Concerning the
Principles of Morals. Oxford: Clarendon Press.

Humphreys, P. 1997. How properties emerge. Philosophy of Science 64:1–17. Reprinted

in Bedau and Humphreys 2008.

Hursthouse, R. 1991. Arational actions. Journal of Philosophy 88:57–68.

Imbens-Bailey, A. L., J. H. Prost, and W. V. Fabricius. 1997. Perception, desire, and

belief in me and you: Young children’s reference to mental states in self and others.
Paper presented at the Biennial Meeting of the Society for Research in Child Devel-
opment, Washington, D.C., April 3–6, 1997.
308 References

Jackson, F. 1998. From Metaphysics to Ethics: A Defense of Conceptual Analysis. New

York: Oxford University Press.

Jacob, P., and M. Jeannerod. 2003. Ways of Seeing: The Scope and Limits of Visual
Cognition. Oxford: Oxford University Press.

James, W. [1890] 1981. The Principles of Psychology, vol. 2. Cambridge, Mass.: Harvard
University Press.

James, W. 1956. The will to believe. In The Will to Believe and Other Essays in Popular
Philosophy, and Human Immortality. New York: Dover.

Jeannerod, M. 1985. The Brain Machine. Cambridge, Mass.: Harvard University Press.

Jeannerod, M. 1997. The Cognitive Neuroscience of Action. Malden, Mass.: Blackwell.

Jeannerod, M. 2006. Motor Cognition. Oxford: Oxford University Press.

Johansson, P., L. Hall, S. Sikstrom, and A. Olsson. 2005. Failure to detect mismatches
between intention and outcome in a simple decision task. Science 310:116–119.

Juarrero, A. 1999. Dynamics in Action: Intentional Behavior as a Complex System. Cam-

bridge, Mass.: MIT Press.

Juarrero, A., and C. Rubino, eds. 2008. Emergence, Complexity, and Self-Organization:
Precursors and Prototypes. Mansfield, Mass.: ISCE Publishing.

Juarrero-Roqué, A. 1983a. Dispositions, teleology, and reductionism. Philosophical

Topics 12:153–165.

Juarrero-Roqué, A. 1983b. Does level-generation always generate act-tokens? Philoso-

phy Research Archives 9:177–192.

Juarrero-Roqué, A. 1985. Kant’s concept of teleology and modern chemistry. Review

of Metaphysics 39:107–135.

Juarrero-Roqué, A. 1987–88. Does action theory rest on a mistake? Philosophy

Research Archives 13:587–612.

Juarrero-Roqué, A. 1988. Non-linear phenomena, explanation, and action. Interna-

tional Philosophical Quarterly 28:247–255.

Kamm, F. 1994. Action, omission, and the stringency of duties. University of Penn-
sylvania Law Review 142:1492–1512.

Kane, R. 1996. The Significance of Free Will. Oxford: Oxford University Press.

Kant, I. 1980. The Critique of Teleological Judgement. Trans. J. C. Meredith. Oxford:

Oxford University Press.

Kim, J. 1989. Mechanism, purpose, and explanatory exclusion. Philosophical Perspec-

tives 3:77–108.
References 309

Kim, J. 1993. The non-reductivist’s troubles with mental causation. In Mental Causa-
tion, ed. J. Heil and A. Mele. New York: Oxford University Press.

Kim, J. 1998. Mind in a Physical World. Cambridge, Mass.: MIT Press.

Kleinig, J. 1976. Good Samaritanism. Philosophy and Public Affairs 5:382–407.

Knobe, J. 2003a. Intentional action and side-effects in ordinary language. Analysis

63:190–193.

Knobe, J. 2003b. Intentional action in folk psychology: An experimental investiga-

tion. Philosophical Psychology 16:309–323.

Knobe, J. 2004a. Folk psychology and folk morality: Response to critics. Journal of
Theoretical and Philosophical Psychology 24:270–279.

Knobe, J. 2004b. Intention, intentional action, and moral considerations. Analysis

64:181–187.

Knobe, J. 2005a. Theory of mind and moral cognition: Exploring the connections.
Trends in Cognitive Sciences 9:357–359.

Knobe, J. 2005b. Cognitive processes shaped by the impulse to blame. Brooklyn Law
Review 71:929–937.

Knobe, J., and A. Burra. 2006a. The folk concept of intention and intentional action:
A cross-cultural study. Journal of Cognition and Culture 6:113–132.

Knobe, J., and A. Burra. 2006b. Experimental philosophy and folk concepts: Meth-
odological considerations. Journal of Cognition and Culture 6:331–342.

Knobe, J., and G. Mendlow. 2004. The good, the bad, and the blameworthy: Under-
standing the role of evaluative reasoning in folk psychology. Journal of Theoretical
and Philosophical Psychology 24:252–258.

Kukso, B. 2006. The reality of absences. Australasian Journal of Philosophy

84:21–37.

Ladd, J. 1965. The ethical dimension of the concept of action. Journal of Philosophy
62:633–645.

Landesman, C. 1965. The new dualism in the philosophy of mind. Review of Meta-
physics 19:329–345.

Leslie, A., J. Knobe, and A. Cohen. 2006. Acting intentionally and the side-effect
effect: “Theory of mind” and moral judgment. Psychological Science 17:421–427.

Lewis, D. 1973. Causation. Journal of Philosophy 70:556–567.

Lewis, D. 1980. Veridical hallucination and prosthetic vision. Australasian Journal of

Philosophy 58:239–249.
310 References

Lewis, D. 1986. Causal explanation. In Philosophical Papers, vol. 2. New York: Oxford
University Press.

Lewis, D. 2004. Void and object. In Causation and Counterfactuals, ed. J. Collins, N.
Hall, and L. A. Paul. Cambridge, Mass.: MIT Press.

Libet, B. 1985. Unconscious cerebral initiative and the role of conscious will in
voluntary action. Behavioral and Brain Sciences 8:529–539.

Libet, B. 2004. Mind Time: The Temporal Factor in Consciousness. Cambridge, Mass.:
Harvard University Press.

Libet, B., C. A. Gleason, E. W. Wright, and D. K. Pearl. 1983. Time of conscious

intention to act in relation to onset of cerebral activities (readiness potential): The
unconscious initiation of a freely voluntary act. Brain 106:623–642.

Louch, A. R. 1966. Explanation and Human Action. Berkeley: University of California

Press.

Lowe, E. J. 2008. Personal Agency: The Metaphysics of Mind and Action. New York:
Oxford University Press.

Machery, E. 2008. The folk concept of intentional action: Philosophical and experi-
mental issues. Mind and Language 23:165–189.

MacKay, D. 1981. Behavioral plasticity, serial order, and the motor program. Behav-
ioral and Brain Sciences 4:630–631.

Mackie, J. L. 1974. The Cement of the Universe: A Study of Causation. Oxford: Claren-
don Press.

Malle, B. 2001. Folk explanations and intentional action. In Intentions and Intention-
ality: Foundations of Social Cognition, ed. L. Moses, B. Malle, and D. Baldwin. Cam-
bridge, Mass.: MIT Press.

Malle, B. 2004. How the Mind Explains Behavior: Folk Explanations, Meaning, and Social
Interaction. Cambridge, Mass.: MIT Press.

Malle, B. 2006. Intentionality, morality, and their relationship in human judgment.

Journal of Cognition and Culture 6:87–112.

Malle, B., and J. Knobe. 1997. The folk concept of intentional action. Journal of
Experimental Social Psychology 33:101–121.

Mallon, R. 2008. Knobe vs. Machery: Testing the trade-off hypothesis. Mind and
Language 23:247–255.

Mandik, P. 2005. Action-oriented representation. In Cognition and the Brain, ed. A.

Brook and K. Akins. Cambridge: Cambridge University Press.

Martinich, A. P. 2005. Hobbes. New York: Routledge.

References 311

Mathis, S. 2003. A plea for omissions. Criminal Justice Ethics (spring–fall):15–31.

McCann, H. 1975. Trying, paralysis, and volition. Review of Metaphysics

28:423–442.

McCann, H. 1998. The Works of Agency: On Human Action, Will and Freedom. Ithaca:
Cornell University Press.

McCann, H. 2006. Intentional action and intending: Recent empirical studies. Philo-
sophical Psychology 18:737–748.

McDowell, J. 1982. Criteria, defeasibility, and knowledge. Proceedings of the British

Academy 62:455–479.

McDowell, J. 1998. Functionalism and anomalous monism. In his Mind, Value, and
Reality. Oxford: Oxford University Press.

McGinn, C. 1982. The Character of Mind. Oxford: Oxford University Press.

McGrath, S. 2005. Causation by omission: A dilemma. Philosophical Studies

123:125–148.

McIntyre, A. G. 1985. Omissions and other acts. Ph.D. Dissertation, Princeton

University.

McIntyre, A. G. 1994. Compatibilists could have done otherwise. Philosophical Review

103:458–488.

Meeks, R. 2004. Unintentionally biasing the data: Reply to Knobe. Journal of Theoreti-
cal and Philosophical Psychology 24:220–223.

Melden, A. I. 1961. Free Action. London: Routledge.

Mele, A. 1981. The practical syllogism and deliberation in Aristotle’s causal theory
of action. New Scholasticism 55:281–316.

Mele, A. 1992. Springs of Action: Understanding Intentional Behavior. New York: Oxford
University Press.

Mele, A. 1997a. Agency and mental action. Philosophical Perspectives 11:231–249.

Mele, A., ed. 1997b. The Philosophy of Action. Oxford: Oxford University Press.

Mele, A. 2001. Acting intentionally: Probing folk notions. In Intentions and Inten-
tionality: Foundations of Social Cognition, ed. B. F. Malle, L. J. Moses, and D. A.
Baldwin. Cambridge, Mass.: MIT Press.

Mele, A. 2003. Motivation and Agency. New York: Oxford University Press.

Mele, A. 2006. Free Will and Luck. Oxford: Oxford University Press.

Mele, A. 2007. Reasonology and false beliefs. Philosophical Papers 36:91–118.

312 References

Mele, A. 2009. Effective Intentions: The Power of Conscious Will. Oxford: Oxford Uni-
versity Press.

Mele, A., and P. Moser. 1994. Intentional action. Noûs 28:39–68.

Mele, A., and S. Sverdlik. 1996. Intention, intentional action, and moral responsibil-
ity. Philosophical Studies 82:265–287.

Mellor, D. H. 1995. The Facts of Causation. London: Routledge.

Meltzoff, A. 1995. Understanding the intentions of others: Re-enactment of intended

acts by 18-month-old children. Developmental Psychology 31:838–850.

Meltzoff, A., and M. Moore. 1977. Imitation of facial and manual gestures by human
neonates. Science 198:75–78.

Menzies, P. 1989. A unified account of causal relata. Australasian Journal of Philosophy

67:68–69.

Moll, H., and M. Tomasello. 2007. How 14- and 18-month-olds know what others
have experienced. Developmental Psychology 43:309–317.

Molnar, G. 2003. Powers: A Study in Metaphysics. Oxford: Oxford University Press.

Moore, M. 1988. Mind, brain, and the unconscious. In Mind, Science and Psycho-
analysis, ed. P. Clark and C. Wright. Oxford: Blackwell.

Moore, M. 1993. Act and Crime: The Philosophy of Action and Its Implications for
Criminal Law. Oxford: Clarendon Press.

Moore, M. 1997. Intentions and mens rea. In M. Moore, Placing Blame: A General
Theory of the Criminal Law. Oxford: Oxford University Press.

Moore, M. 2006. Intention, responsibility, and the challenges of recent neuro-

science. In Does Consciousness Cause Behavior? ed. S. Pockett, W. P. Banks, and S.
Gallagher. Cambridge, Mass.: MIT Press.

Moore, M. 2009a. Causation and Responsibility: An Essay in Law, Morals, and Metaphys-
ics. Oxford: Oxford University Press.

Moore, M. 2009b. Intentions, responsibility, and the challenges of recent neurosci-

ence. Stanford Technology Review. http://www.stlr.stanford.edu.

Moore, M. 2010. Intentions as a marker of moral culpability and legal punishment.

In The Philosophical Foundations of the Criminal Law, ed. R. Duff and S. Green. Oxford:
Oxford University Press.

Moore, M. Forthcoming. Consciousness, brain states, and the causal efficacy of

intentions. In Conscious Will and Responsibility: A Tribute to Benjamin Libet, ed. L.
Nadel and W. Sinnott-Armstrong. Oxford: Oxford University Press.
References 313

Moreno, A. 2008. Organization and evolution: The principle of dynamic decoupling.

Paper delivered at Complejidad 2008: 4th Biennial Conference on the Philosophical,
Epistemological, and Methodological Implications of Complexity Theory. Havana,
Cuba.

Moya, C. 1990. The Philosophy of Action. Oxford: Polity.

Mumford, S. 2009. Passing powers around. Monist 92:94–111.

Nadelhoffer, T. 2004a. The butler problem revisited. Analysis 64:277–284.

Nadelhoffer, T. 2004b. Praise, side effects, and intentional action. Journal of Theoreti-
cal and Philosophical Psychology 24:196–213.

Nadelhoffer, T. 2004c. Blame, badness, and intentional action: A reply to Knobe and
Mendlow. Journal of Theoretical and Philosophical Psychology 24:259–269.

Nadelhoffer, T. 2005. Skill, luck, control, and intentional action. Philosophical Psy-
chology 18:343–354.

Nadelhoffer, T. 2006a. Bad acts, blameworthy agents, and intentional actions: Some
problems for jury impartiality. Philosophical Explorations 9:203–220.

Nadelhoffer, T. 2006b. Foresight, moral considerations, and intentional actions.

Journal of Cognition and Culture 6:133–158.

Nadelhoffer, T. 2006c. On trying to save the simple view. Mind and Language
21:565–586.

Nadelhoffer, T., and E. Nahmias. 2007. The past and future of experimental philoso-
phy. Philosophical Explorations 10:123–149.

Nado, J. Forthcoming. Effects of moral cognition on judgments of intentionality.

British Journal for the Philosophy of Science.

Nanay, B. Forthcoming. Morality or modality: What does the attribution of inten-

tionality depend on? Canadian Journal of Philosophy.

Nichols, S., and J. Ulatowski. 2007. Intuitions and individual differences: The Knobe
effect revisited. Mind and Language 22:346–365.

Nisbett, R. E. 2003. The Geography of Thought: How Asians and Westerners Think Dif-
ferently . . . and Why. New York: Free Press.

Nisbett, R. E., and L. Ross. 1980. Human Inference: Strategies and Shortcomings of Social
Judgment. Englewood Cliffs, N.J.: Prentice-Hall.

Noë, A. 2004. Action in Perception. Cambridge, Mass.: MIT Press.

O’Connor, T. 2002. Persons and Causes: The Metaphysics of Free Will. Oxford: Oxford
University Press.
314 References

Onishi, K. H., and R. Baillargeon. 2005. Do 15-month-old infants understand false

beliefs? Science 308:255–258.

Pacherie, E. 2006. Towards a dynamic theory of intentions. In Does Consciousness

Cause Behavior? An Investigation into the Nature of Volition, ed. S. Pockett, W. P. Banks,
and S. Gallagher. Cambridge, Mass.: MIT Press.

Pacherie, E. 2008. The phenomenology of action: A conceptual framework. Cognition

107:179–217.

Parfit, D. 1997. Reasons and motivation. Proceedings of the Aristotelian Society (suppl.
vol.) 71:99–129.

Peacocke, C. 1979a. Deviant causal chains. Midwest Studies in Philosophy

4:123–155.

Peacocke, C. 1979b. Holistic Explanation: Action, Space, Interpretation. New York:

Oxford University Press.

Pecher, D., and R. Zwaan, eds. 2005. Grounding Cognition. Cambridge: Cambridge
University Press.

Perner, J. 2004. Wann verstehen Kinder Handlungen als rational? In Der Mensch—ein
“animal rationale”? Vernunft—Kognition—Intelligenz, ed. H. Schmidinger and C.
Sedmak. Darmstadt: Wissenschaftliche Buchgemeinschaft.

Perner, J., B. Lang, and D. Kloo. 2002. Theory of mind and self-control: More than
a common problem of inhibition. Child Development 73:752–767.

Perner, J., B. Rendl, and A. Garnham. 2007. Objects of desire, thought, and reality:
Problems of anchoring discourse referents in development. Mind & Language
22:475–517.

Perner, J., and T. Ruffman. 2005. Infants’ insight into the mind: How deep? Science
308:214–216.

Perner, J., P. Zauner, and M. Sprung. 2005. What does “that” have to do with point
of view? The case of conflicting desires and “want” in German. In Why Language
Matters for Theory of Mind, ed. J. W. Astington and J. Baird. New York: Oxford Uni-
versity Press.

Peters, R. 1958. The Concept of Motivation. London: Routledge.

Peters, R. 1967. Hobbes. Harmondsworth: Penguin.

Phelan, M., and H. Sarkissian. 2008. The folk strike back: Or, Why you didn’t do it
intentionally, though it was bad and you knew it. Philosophical Studies
138:291–298.

Phelan, M., and H. Sarkissian. 2009. Is the “trade-off hypothesis” worth trading for?
Mind and Language 24:164–180.
References 315

Pietroski, P. 2002. Causing actions. Oxford: Oxford University Press.

Pink, T. 1997. Reason and agency. Proceedings of the Aristotelian Society

97:263–280.

Pink, T. 2004. Suarez, Hobbes, and the Scholastic tradition in action theory. In The
Will and Human Action: From Antiquity to the Present Day, ed. T. Pink and M. Stone.
London: Routledge.

Povinelli, D. J., and S. deBlois. 1992. Young children’s (Homo sapiens) understanding
of knowledge formation in themselves and others. Journal of Comparative Psychology
106:228–238.

Price, A. 2004. Aristotle, the Stoics, and the will. In The Will and Human Action: From
Antiquity to the Present Day, ed. T. Pink and M. Stone. London: Routledge.

Prichard, H. A. 1949. Moral Obligations. Oxford: Clarendon Press.

Rakoczy, H., F. Warneken, and M. Tomasello. 2007. “This way!,” “No! that way!”—
3-year olds know that two people can have mutually incompatible desires. Cognitive
Development 22:47–68.

Rakoczy, H., F. Warneken, and M. Tomasello. 2008. The sources of normativity:

Young children’s awareness of the normative structure of games. Developmental
Psychology 44:875–881.

Raz, J. 1978. Introduction. In Practical Reasoning, ed. J. Raz. Oxford: Oxford Univer-
sity Press.

Reid, Thomas. [1788] 1983. Essays on the Active Powers of Man. In The Works of
Thomas Reid, D.D., ed. W. Hamilton. Hildesheim: G. Olms Verlagsbuchhandlung.

Reingold, E. M., and P. M. Merikle. 1993. Theory and measurement in the study of
unconscious processes. In Consciousness, ed. M. Davies and G. W. Humphreys.
Oxford: Blackwell.

Repacholi, B. M., and A. Gopnik. 1997. Early reasoning about desires: Evidence from
14- and 18-month-olds. Developmental Psychology 33:12–21.

Richardson-Klavehn, A., and R. A. Bjork. 1988. Measures of memory. Annual Review

of Psychology 39:475–543.

Riley, M. A., and M. T. Turvey. 2001. The self-organizing dynamics of intentions

and actions. American Journal of Psychology 114:160–169.

Rizzolati, G., L. Fadiga, V. Gallese, and L. Fogassi. 1996. Premotor cortex and the
recognition of motor actions. Brain Research: Cognitive Brain Research 3:131–141.

Rocha, L. M. 2001. Evolution with material symbol systems. BioSystem 60:95–121.

Rowlands, M. 2006. Body Language. Cambridge, Mass.: MIT Press.

316 References

Ruben, D. 1985. The Metaphysics of the Social World. London: Routledge & Kegan
Paul.

Ruben, D. 2003. Action and Its Explanation. Oxford: Oxford University Press.

Ruben, D. 2008. Disjunctive theories of perception and action. In

Disjunctivism:Perception, Action, Knowledge, ed. A. Haddock and F. MacPherson.
Oxford: Oxford University Press.

Ruffman, T., W. Garnham, A. Import, and D. Connolly. 2001. Does eye gaze indicate
implicit knowledge of false belief? Charting transitions in knowledge. Journal of
Experimental Child Psychology 80:201–224.

Ruiz-Mirazo, K., and A. Moreno. 1998. Autonomy and emergence: How systems
become agents through the generation of functional constraints. In Emergence,
Complexity, Hierarchy, Organization, ed. G. L. Farre and T. Oksala. Acta Polytechnica
Scandinavica. Espoo-Helsinki: The Finnish Academy of Technology.

Ruiz-Mirazo, K., and A. Moreno. 2000. Searching for the roots of autonomy: The
natural and artificial paradigms revisited. Communication and Cognition—Artificial
Intelligence (CC-AI): The Journal for the Integrated Study of Artificial Intelligence, Cognitive
Science, and Applied Epistemology 17:209–228.

Ruiz-Mirazo, K., and A. Moreno. 2006. The maintenance and open-ended growth
of complexity in nature: Information as a decoupling mechanism in the origins of
life. In Rethinking Complexity: Perspectives from North and South, ed. F. Capra, P. Soto-
longo, A. Juarrero, and J. van Uden. Mansfield, Mass.: ISCE Publishing.

Ruiz-Mirazo, K., J. Peretó, and A. Moreno. 2004. A universal definition of life:

Autonomy and open-ended evolution. Origins of Life and Evolution of the Biosphere
34:323–346.

Russell, B. 1903. The Principles of Mathematics. Cambridge: Cambridge University

Press.

Salthe, S. N. 1985. Evolving Hierarchical Systems: Their Structure and Representation.

New York: Columbia University Press.

Salthe, S. N. 2002. Summary of the principles of hierarchy theory. General Systems

Bulletin 31:13–17.

Sandis, C., ed. 2009. New Essays on the Explanation of Action. Basingstoke: Palgrave
Macmillan.

Sartorio, C. 2005. A new asymmetry between actions and omissions. Noûs

39:460–482.

Sartorio, C. 2009. Omissions and causalism. Noûs 43:513–530.

Scanlon, T. 1998. What We Owe to Each Other. Cambridge, Mass.: Harvard University
Press.
References 317

Schaffer, J. 2005. Contrastive causation. Philosophical Review 114:327–358.

Schroeter, F. 2004. Endorsement and autonomous agency. Philosophy and Phenom-

enological Research 69:633–659.

Schueler, G. F. 2003. Reasons and Purposes: Human Rationality and the Teleological
Explanation of Action. Oxford: Clarendon Press.

Schueler, G. F. 2009. The Humean theory of motivation rejected. Philosophy and

Phenomenological Research 78:103–122.

Schult, C. A. 2002. Children’s understanding of the distinction between intentions

and desires. Child Development 73:1727–1747.

Searle, J. 1980. Minds, brains, and programs. Behavioral and Brain Sciences
3:417–457.

Searle, J. 1981. The intentionality of intention and action. Manuscrito 4:77–101.

Searle, J. 1983. Intentionality. Cambridge: Cambridge University Press.

Sehon, S. 1994. Teleology and the nature of mental states. American Philosophical
Quarterly 31:63–72.

Sehon, S. R. 1997. Deviant causal chains and the irreducibility of teleological expla-
nation. Pacific Philosophical Quarterly 78:195–213.

Sehon, S. 2005. Teleological Realism: Mind, Agency, and Explanation. Cambridge, Mass:
MIT Press.

Sellars, W. 1963. Philosophy and the scientific image of man. In Science, Perception,
and Reality. New York: Routledge & Kegan Paul.

Seth, A. K. 2005. Causal connectivity of evolved neural networks during behavior.

Network 16:35–54.

Seth, A. K. 2006. Causal networks in neural systems: From water mazes to conscious-
ness. In Proceedings of the 2006 Meeting on Brain Inspired Cognitive Systems, ed. I.
Aleksander et al. Millet, Alberta: ICSC Interdisciplinary Research.

Seth, A. K., and G. M. Edelman. 2007. Distinguishing causal interactions in neural

populations. Neural Computation 19:910–933.

Shannon, C. 1948. The mathematical theory of communication. Bell System Techni-

cal Journal (July and October).

Shultz, T. R., and F. Shamash. 1981. The child’s conception of intending act and
consequence. Canadian Journal of Behavioural Science 13:368–372.

Simons, D. J., and C. F. Chabris. 1999. Gorillas in our midst: Sustained inattentional
blindness for dynamic events. Perception 28:1059–1074.
318 References

Sirois, S., and I. Jackson. 2007. Social cognition in infancy: A critical review of
research on higher order abilities. European Journal of Developmental Psychology
4:46–64.

Smart, J. J. C. 1959. Sensations and brain processes. Philosophical Review

68:141–156.

Smith, A. 2005. Responsibility for attitudes: Activity and passivity in mental life.
Ethics 115:236–271.

Smith, M. 1998. The possibility of philosophy of action. In Human Action, Delibera-

tion, and Causation, ed. J. Bransen and S. Cuypers, 17–41. Dordrecht: Kluwer Aca-
demic. Reprinted in Smith 2004.

Smith, M. 2003a. Rational capacities. In Weakness of Will and Varieties of Practical

Irrationality, S. Stroud and C. Tappolet, 17–38. Oxford: Oxford University Press.
Reprinted in Smith 2004.

Smith, M. 2003b. Humeanism, psychologism, and the normative story. Philosophy

and Phenomenological Research 67:460–467.

Smith, M. 2004. Ethics and the A Priori: Selected Essays on Moral Psychology and Meta-
Ethics. New York: Cambridge University Press.

Smith, M. Forthcoming. The explanatory role of being rational. In Practical Reason,

ed. D. Sobel and S. Wall. New York: Cambridge University Press.

Smith, P. 1990. Contemplating failure: The importance of unconscious omission.

Philosophical Studies 59:159–176.

Sodian, B. 1991. The development of deception in young children. British Journal of

Developmental Psychology 9:173–188.

Sodian, B., C. Thoermer, and N. Dietrich. 2006. Two- to four-year-old children’s

differentiation of knowing and guessing in a non-verbal task. European Journal of
Developmental Psychology 3:222–237.

Soon, C. S., M. Brass, H.-J. Heinze, and J. Haynes. 2008. Unconscious determinants
of free decisions in the human brain. Nature Neuroscience 11:543–545.

Sorabji, R. 2004. The concept of the will from Plato to Maximus the Confessor. In
The Will and Human Action: From Antiquity to the Present Day, ed. T. Pink and M.
Stone. London: Routledge.

Sorabji, R. 1979. Body and soul in Aristotle. In Articles on Aristotle, vol. 4: Psychology
and Aesthetics, ed. J. Barnes, M. Schofield, R. Sorabji. London: Duckworth.

Sorell, T. 1986. Hobbes. London: Routledge.

Southgate, V. 2008. Attributions of false belief in infancy. Invited Presentation at

the EPS Research Workshop on Theory of Mind: A Workshop in Celebration of the
References 319

30th Anniversary of Premack and Woodruff’s Seminal Paper, “Does the Chimpanzee
Have a Theory of Mind?” (BBS 1978). Organized by A. Hamilton, I. Apperly, and D.
Samson, University of Nottingham, September 11–12, 2008.

Southgate, V., A. Senju, and G. Csibra. 2007. Action anticipation through attribution
of false belief by 2-year-olds. Psychological Science 18:586–592.

Stampe, D. 1977. Towards a causal theory of linguistic representation. In Midwest

Studies in Philosophy, vol. 2. ed. P. French et al. Minneapolis: University of Minnesota
Press.

Steward, H. 1997. The Ontology of Mind. Oxford: Clarendon Press.

Surian, L., S. Caldi, and D. Sperber. 2007. Attribution of beliefs by 13-month-old

infants. Psychological Science 18:580–586.

Symposium on B. Libet. 1985. Behavioral and Brain Sciences 8.

Symposium on M. Moore’s Act and Crime book. 1994. University of Pennsylvania Law
Review 142: 1443–1748.

Tanney, J. 2005. Reason-explanation and the contents of the mind. Ratio

18:338–351.

Taylor, R. 1966. Action and Purpose. Englewood Cliffs, N.J.: Prentice-Hall.

Thalberg, I. 1977. Perception, Emotion, and Action. New Haven: Yale University Press.

Thomson, J. 1977. Acts and Other Events. Ithaca: Cornell University Press.

Thomson, J. 1996. Critical study on Jonathan Bennett’s The Act Itself. Noûs
30:545–557.

Thomson, J. 2003. Causation: Omissions. Philosophy and Phenomenological Research

66:81–103.

Tomasello, M. 1999. The Cultural Origins of Human Cognition. Cambridge, Mass.:

Harvard University Press.

Tomasello, M., M. Carpenter, J. Call, T. Behne, and H. Moll. 2005. Understanding

and sharing intentions: The origins of cultural cognition. Behavioral and Brain Sci-
ences 28:675–735.

Tomasello, M., and K. Haberl. 2003. Understanding attention: 12- and 18-month-
olds know what is new for other persons. Developmental Psychology 39:906–912.

Tooley, M. 1987. Causation: A Realist Approach. Oxford: Oxford University Press.

Tuck, R. 1989. Hobbes. New York: Oxford University Press.

Turner, J. 2004. Folk intuitions, asymmetry, and intentional side effects. Journal of
Theoretical and Philosophical Psychology 24:214–219.
320 References

Ulanowicz, R. 1997. Ecology: The Ascendent Perspective. New York: Columbia Univer-
sity Press.

Ulanowicz, R. 2005. A revolution in the Middle Kingdom. In Micro, Meso, Macro:

Addressing Complex Systems Couplings, ed. H. Liljenström and U. Svedin. Hackensack,
N.J.: World Scientific.

Ulanowicz, R. 2008. The Third Window. West Conshohocken, Pennsylvania: Temple-

ton Press.

Van Mill, D. 2001. Liberty, Rationality, and Agency in Hobbes’s “Leviathan.” Albany:
SUNY Press.

Van Orden, G. C., H. Kloos, and S. Wallot. 2009. Living in the pink: Intentionality,
wellbeing, and complexity. In Handbook of the Philosophy of Science, vol. 10: Philoso-
phy of Complex Systems, ed. C. Hooker. General editors D. M. Gabbay, P. Thagard,
and J. Woods. Amsterdam: Elsevier BV.

Velleman, J. D. 1992. What happens when someone acts? Mind 101:461–481.

Reprinted in Velleman 2000.

Velleman, J. D. 2000. The Possibility of Practical Reason. New York: Oxford University
Press.

Vendler, Z. 1962. Effects, results, and consequences. In Analytical Philosophy, 1–15,

ed. R. J. Butler. Oxford: Blackwell.

Vermazen, B. 1985. Negative acts. In Essays on Davidson, ed. B. Vermazen and M.

Hintikka. New York: Oxford University Press.

Vihvelin, K., and T. Tomkow. 2005. The dif. Journal of Philosophy 102:183–205.

von Wright, G. H. 1978. On so-called practical inference. In Practical Reasoning, ed.

J. Raz. Oxford: Oxford University Press.

Wallace, R. J. 1999. Three conceptions of rational agency. Ethical Theory and Moral
Practice 2:217–242.

Walton, D. 1980. Omitting, refraining, and letting happen. American Philosophical

Quarterly 17:319–326.

Washington University School of Medicine. 2008. Scientists adapt economics theory

to trace brain’s information flow. ScienceDaily (October 10). http://www.science-
daily.com/releases/2008/10/081009185035.htm

Watson, G., ed. 1982. Free Will. Oxford: Oxford University Press.

Wegner, D. 2002. The Illusion of Conscious Will. Cambridge, Mass.: MIT Press.

Weinryb, E. 1980. Omissions and responsibility. Philosophical Quarterly 30:1–18.

Wellman, H. M., D. Cross, and J. Watson. 2001. Meta-analysis of theory of mind

development: The truth about false belief. Child Development 72:655–684.
References 321

Wheeler, M. 2005. Reconstructing the Cognitive World: The Next Step. Cambridge,
Mass.: MIT Press.

Wheeler, M., and A. Clark. 1999. Genic representation: Reconciling content and
causal complexity. British Journal for the Philosophy of Science 50:103–135.

Wiggins, D. 1987. Claims of need. In his Needs, Values, Truth. Oxford: Blackwell.

Wiland, E. 2007. Intentional action and “in order to.” Journal of Theoretical and
Philosophical Psychology 27:113–118.

Williams, B. 1973. Morality and the emotions. In his Problems of the Self. Cambridge:
Cambridge University Press.

Williams, B. 1981a. Internal and external reasons. In his Moral Luck. Cambridge:
Cambridge University Press.

Williams, B. 1981b. Moral Luck. Cambridge: Cambridge University Press.

Williams, B. 1985. Ethics and the Limits of Philosophy. London: Fontana.

Williams, G. 1987. Oblique intention. Cambridge Law Journal 46:417–438.

Williams, B. 1995a. Acts and omissions, doing and not doing. In Virtues and Reasons:
Philippa Foot and Moral Theory, ed. R. Hursthouse, G. Lawrence, and W. Quinn.
Oxford: Clarendon Press.

Williams, B. 1995b. Internal reasons and the obscurity of blame. In his Making Sense
of Humanity. Cambridge: Cambridge University Press.

Williams, B. 1996. Truth in ethics. In Truth in Ethics, ed. B. Hooker. Oxford:

Blackwell.

Williamson, T. 2000. Knowledge and Its Limits. Oxford: Oxford University Press.

Wilson, G. 1989. The Intentionality of Human Action, 2nd ed. Stanford: Stanford
University Press.

Wilson, G. 1997. Reasons as causes for action. In Contemporary Action Theory. vol.
1. ed. G. Holmström-Hintikka and R. Tuomela. Dordrecht: Kluwer.

Wilson, G. 2009. Debates about causalism in the theory of action. In Philosophy of

Action: Five Questions, ed. J. Aguilar and A. Buckareff. London: Automatic/VIP.

Wilson, M. 2002. Six views of embodied cognition. Psychonomic Bulletin and Review
9:625–636.

Wimmer, H., and H. Mayringer. 1998. False belief understanding in young children:
Explanations do not develop before predictions. International Journal of Behavioral
Development 22:403–422.

Winch, P. 1958. The Idea of a Social Science. London: Routledge.

322 References

Wittgenstein, L. 1958. Philosophische Untersuchungen. Frankfurt: Suhrkamp.

Wittgenstein, L. 1972. Philosophical Investigations. Trans. G. E. M. Anscombe. Oxford:

Blackwell.

Woodward, J. 2003. Making Things Happen: A Theory of Causal Explanation. Oxford:

Oxford University Press.

Woodward, J. Forthcoming. Causal perception and causal cognition.

Wright, J., and J. Bengson. 2009. Asymmetries in folk judgments of responsibility

and intentional action. Mind and Language 24:24–50.

Yablo, S. 1992. Mental causation. Philosophical Review 101:245–280.

Yaffe, G. 2009. Commentary on Moore’s “Intention, responsibility, and the chal-

lenges of recent neuroscience.” Stanford Technology Law Review. http://stlr.stanford
.edu/2009/02/intention_responsibility_and_t.html.

Young, L., F. Cushman, R. Adolphs, D. Tranel, and M. Hauser. 2006. Does emotion
mediate the relationship between an action’s moral status and its intentional status?
Neuropsychological evidence. Journal of Cognition and Culture 6:291–304.

Yuill, N. 1984. Young children’s coordination of motive and outcome in judgments

of satisfaction and morality. British Journal of Developmental Psychology 2:73–81.

Zhabotinsky, A. M. 1964. Periodic processes of malonic acid oxidation in a liquid

phase. Biofizika 9:306–311.

Zimmerman, M. 1981. Taking some of the mystery out of omissions. Southern Journal
of Philosophy 19:541–554.
Contributors

Frederick Adams Professor of Cognitive Science and Philosophy, Univer-

sity of Delaware

Jesús H. Aguilar Associate Professor of Philosophy, Rochester Institute of

Technology

John Bishop Professor of Philosophy, University of Auckland

Andrei A. Buckareff Assistant Professor of Philosophy, Marist College

Randolph Clarke Professor of Philosophy, Florida State University

Jennifer Hornsby Professor of Philosophy, Birkbeck College, University of

London

Alicia Juarrero Professor of Philosophy, Prince George’s Community

College

Alfred R. Mele William H. and Lucyle T. Werkmeister Professor of Phi-

losophy, Florida State University

Michael S. Moore Charles R. Walgreen, Jr. University Chair, Professor

of Law and Philosophy, Professor in the Center for Advanced Study, and
Co-Director of the College of the Law Program in Law and Philosophy,
University of Illinois College of Law

Thomas A. Nadelhoffer Assistant Professor of Philosophy and Contribut-

ing Faculty to Law and Policy Program, Dickinson College
Josef Perner Professor of Psychology, University of Salzburg

Johannes Roessler Senior Lecturer in Philosophy, University of

Warwick

David-Hillel Ruben Director of New York University in London and Pro-

fessor, Birkbeck College, University of London
324 Contributors

Carolina Sartorio Associate Professor of Philosophy, University of

Arizona

Michael Smith McCosh Professor of Philosophy, Princeton University

Rowland Stout Senior Lecturer in Philosophy, University College

Dublin
Index

Absent agent, 12–13, 28–30, 33–35, Bishop, John, 15–16, 87–94, 96

38–39 Bratman, Michael, 13–14, 30, 51, 242,
Adams, Fred, 21, 242 267, 284
Agency, 13, 48–55, 57–59, 61, 63, 66,
72, 85, 102–103, 106, 110, 112 Causal deviance, 10–12, 74–76, 79, 86,
agential systems, 87, 92–93 88, 93, 97
autonomous, 13, 96 Causal power, 5, 9, 14, 30, 64, 72–73,
and the CTA, 9, 12–16, 27–28, 48–49, 108, 120, 124, 128, 140, 147, 159,
101, 115–116, 135 165, 177
human, 4, 277 Chisholm, Roderick, 12, 74
moral, 2, 70 Choice, 5, 27, 168–169, 171–175,
and omissions, 135, 139, 151 268–269, 286–289, 291. See also
problem of natural, 69, 70–72, 75–78 Decision
prosthetic, 86–88, 91, 93 Cognitive science, 229, 231
Agent-causation, 9, 16, 28, 30, 38, Consciousness, 219, 264, 274–275
71–72, 74–76, 78–80 Control, 5–6, 9, 11–15, 46, 49, 64, 71,
Anscombe, Elizabeth, 103, 127 74, 77, 80, 87–94, 97, 111, 148–150,
Aquinas, St. Thomas, 5 184–186, 188–190, 209, 230–231,
Aristotle, 2–4, 6, 9, 107, 111–112, 254, 234–235, 237, 243, 248, 262, 264,
257, 263 268, 270–273, 275
Autonomy, 13, 21, 257, 262, 272–273, Counterfactuals, 29–30, 36, 54, 102,
275 117–118, 148, 175–178, 190,
195–196, 208–209, 233
Basic action, 45–47, 78, 89, 96
Belief, 6–7, 9–10, 27–28, 30, 36–38, 40, Davidson, Donald, 2, 6–10, 16–17, 30,
45–47, 51–54, 56, 62, 65, 72, 74, 87, 32–33, 45, 52–53, 61–63, 65, 72–80,
89, 116, 142, 163, 165, 167, 180, 103, 109, 116–117, 137, 167–168,
186, 194, 199–202, 204–207, 211, 170, 183, 192, 195–196, 199, 203,
213–215, 217–225, 230, 241–242, 207–208, 218
254–255, 230, 241–242, 254–255, Decision, 27, 93, 116, 138, 162, 173,
277, 279, 281, 293 175, 180, 269–270, 277
326 Index

Deliberation, 5–6, 13, 27, 119, 143, 169, Hume, David, 4, 45, 66, 105, 108–109,
172–173, 175, 181, 203–204, 226 204, 218, 255
Desire, 3–6, 8, 10, 13, 16–17, 27–28,
30, 36–38, 45–47, 49–54, 57, 62, 65, Identification, 13
87, 116, 144, 163, 178, 183, 194, Intending, 116, 124, 136, 141–148,
199–202, 204–207, 211–214, 159–160, 164, 194, 284
216–223, 230, 241–242, 244, Intention, 9–11, 17, 27–28, 30, 35–41,
277–278, 281–282, 284 65, 73–74, 80, 85–93, 96–98,
Determinism, 70, 140 116–129, 137–151, 157–165, 167,
Developmental psychology, 202. See 183, 194, 207, 213–214, 216, 224,
also Psychology 229–231, 234–248, 253, 255,
Differential explanation, 95 267–270, 274–275, 277, 279, 281
Dispositions, 30, 174–175. See also Intentional action, 1, 3, 8–9, 11, 16–17,
Causal power 73–75, 80, 87, 90, 92–95, 97,
Doing, 27–29, 34, 41, 46–50, 52–53, 115–116, 118, 129, 135–139, 141,
55, 58, 61, 63–64, 76, 101–103, 145, 147, 149–151, 161, 199–203,
111–112, 115, 120–121, 126, 138, 205, 207–209, 213–214, 217, 218,
143–144, 184–185, 188, 199, 207, 230–231, 265, 277–281, 284, 287,
235–236, 246, 284, 288 293
Donagan, Alan, 2–3 Intentional behavior, 9, 86, 129, 161,
Duns Scotus, 5 165, 254
Intentionality, 217, 248, 255, 278–279,
Embodied cognition, 229, 231–235 281–282, 284, 286–287, 293–294
Emotion, 73, 212, 219, 222–223, 267,
274 Kant, Immanuel, 255–256
Endorsement, 13 Knobe, Joshua, 277–278, 281–285,
287–293
Fischer, John Martin, 139–140 Knowledge how, 45–46, 48–49, 64
Frankfurt, Harry, 13
Free action, 8, 73, 279. See also Free will Lewis, David, 30, 80, 194
Free will, 4, 38, 69–70, 262, 269, 272, Libet, Benjamin, 43, 269–271
275, 280
Malle, Bertram, 26, 278, 281, 284–293
Gallagher, Shaun, 238–241, 244, McCann, Hugh, 280
247–248 Mele, Alfred R., 7, 17–18, 20, 69, 125,
Ginet, Carl, 7, 115–116, 141–143, 145, 230–231
183, 192–193 Mental action, 14, 19, 33–35, 77,
Goldman, Alvin, 38–39, 137 121–123, 125
Guidance, 92, 234, 245 Mental causation, 1, 3–4, 8–14, 28–29,
35–41, 69, 71, 73–77, 80, 85,
Hobbes, Thomas, 2, 4–6, 9 116–119, 123–125, 129, 145, 147,
Hornsby, Jennifer, 12, 29, 31, 34, 150, 161, 165, 183, 187, 194,
47–55, 109–111 229–230, 265–267
Index 327

Moore, Michael S., 11, 19, 27, 31, 36 Reliability, 20, 93–98, 258
Motivation, 5, 13, 168, 170, 176, 181 Responsibility, 14, 27, 33, 39–40, 69,
72, 92
Nadelhoffer, Thomas, 21 Rowlands, Mark, 236–238, 241, 244
Neuroscience, 38–40 Ruben, David-Hillel, 20
Ryle, Gilbert, 6
Omissions, 19–20, 34–35, 48–49, 55, 58,
115–129, 135–151, 157–160, 161–165 Sartorio, Carolina, 20, 143, 146, 151,
154–155, 161–165
Pacherie, Elisabeth, 230, 235, 237, Searle, John, 239, 247
244–246, 248 Sehon, Scott, 20, 187–196
Peacocke, Christopher, 10–11, 52–53, Simple View of intentional action, 231,
80, 85–86, 89–93 284–285, 292, 295. See also
Practical reason(ing), 6, 80, 154, Intentional action
202–204, 206, 209–211, 217–218, Smith, Michael, 11, 18, 19, 57–66
253. See also Reasons for action Standard story of action, 1, 6, 9, 13,
Psychology, 14, 18, 199–202 19, 27–28, 38, 45–55, 57–66,
developmental psychology, 20–21, 115–116, 149, 253
199–202, 208, 210, 214–215, 217
social psychology, 18, 295 Taylor, Richard, 12, 30
Top-down causation, 254, 256–257,
Rational(ity), 53–55, 63–66, 73, 142, 260, 26–275. See also Causation
168–170, 174–81, 203, 210, 212–214, Trying, 14, 29, 31–32, 40, 77, 136, 231,
222–223 235, 240, 244
Reasons-explanation, 1, 5–11, 13,
16–18, 20–21, 29–30, 38, 49–53, Velleman, J. David, 13
59–60, 62–65, 69–71, 115, 118–119, Volition, 27–28, 30, 37. See also Trying
150–151, 167–168, 178, 183–196, Volitionist theory of action, 14, 77
200, 202, 203, 205, 207–213,
218–219, 277 Wants, 7, 13, 17, 144. See also Desire
anti-psychologistic (objective) theory Weakness of will (akrasia), 2, 73, 170,
of, 18, 200, 205, 207–213, 218–219 175–176, 178–179
causal theory of, 16–18, 20, 38, 49–53, Will(ing), 5–6, 14, 27, 31–32, 34–35,
59, 62–65, 115, 150–151, 167–168, 77. See also Volition
178–80, 183–196 Wilson, George, 16–17, 20, 145,
teleological theory of, 16–18, 20, 183–187
183–96, 202, 207–213, 218–219, Wittgenstein, Ludwig, 2, 6, 69, 75,
227–228 77–78
Reasons for action, 1, 5–11, 16–18,
20–21, 38, 71, 116, 137, 150–51,
167–181, 195–196
con-reasons vs. pro-reasons, 20, 167,
169–170, 172–175, 180–181