Professional Documents
Culture Documents
(Oxford Handbooks in Philosophy) Brian P. McLaughlin - The Oxford Handbook of Philosophy of Mind-Clarendon Press (2009)
(Oxford Handbooks in Philosophy) Brian P. McLaughlin - The Oxford Handbook of Philosophy of Mind-Clarendon Press (2009)
[UNTITLED]
The Oxford Handbook of Philosophy of Mind
Edited by Ansgar Beckermann, Brian P. McLaughlin, and Sven Walter
Print Publication Date: Jan 2009 Subject: Philosophy Online Publication Date: Sep 2009
(p. iv)
With offices in
Page 1 of 2
[UNTITLED]
You must not circulate this book in any other binding or cover
and you must impose this same condition on any acquirer
Data available
The Oxford handbook of philosophy of mind / edited by Brain McLaughlin with Ansgar
Beckermann and Sven Walter.
p. cm.—(Oxford handbooks in philosophy)
Includes index.
ISBN 978–0–19–926261–8
1. Philosophy of mind. I. McLaughlin, Brain P. II. Beckermann, Ansgar, 1945–III.
Walter, Sven.
BD418.3.094 2008
128′.2—dc22 2008039119
ISBN 978–0–19–926261–8
1 3 5 7 9 10 8 6 4 2
Page 2 of 2
Dedication
Dedication
The Oxford Handbook of Philosophy of Mind
Edited by Ansgar Beckermann, Brian P. McLaughlin, and Sven Walter
Print Publication Date: Jan 2009 Subject: Philosophy Online Publication Date: Sep 2009
Dedication
(p. v) For Bryant, Daniel, Kathryn, Will, Grace, and Megan
Page 1 of 1
Acknowledgements
Acknowledgements
The Oxford Handbook of Philosophy of Mind
Edited by Ansgar Beckermann, Brian P. McLaughlin, and Sven Walter
Print Publication Date: Jan 2009 Subject: Philosophy Online Publication Date: Sep 2009
Page 1 of 1
List of Contributors
List of Contributors
The Oxford Handbook of Philosophy of Mind
Edited by Ansgar Beckermann, Brian P. McLaughlin, and Sven Walter
Print Publication Date: Jan 2009 Subject: Philosophy Online Publication Date: Sep 2009
Email: lantony@philos.umass.edu
Anita Avramides: Fellow and Tutor, St. Hilda's College, University of Oxford, Cow
ley Place, Oxford, OX4 1DY. United Kingdom.
Email: anita.avramides@st‐hildas.ox.ac.uk
Email: kbach@sfsu.edu
Page 1 of 10
List of Contributors
Email: lrbaker@philos.umass.edu
Email: katalin.balog@yale.edu
Email: ansgar.beckermann@uni‐bielefeld.de
Email: bermudez@wustl.edu
Email: abrook@ccs.carleton.ca
Jessica Brown: Professor in Arché Philosophical Research Center for Logic, Lan
guage, Metaphysics and Epistemology, Philosophy Departments, The University of St
Andrews, Edgecliffe, The Scores, St Andrews, KY16 9AL. UNITED KINGDOM.
Email: jab30@st‐andrews.ac.uk
Page 2 of 10
List of Contributors
Email: abyrne@mit.edu
Email: jjcampbell@berkeley.edu
Email: chalmers@anu.edu.au
Email: tim.crane@ucl.ac.uk
Daniel Dennett: Co‐director of the Center for Cognitive Studies, Austin B. Fletcher
Professor of Philosophy, Department of Philosophy, Tufts University, Medford, MA
02155–7059. USA.
Email: daniel.dennett@tufts.edu
Page 3 of 10
List of Contributors
Email: Fred.dretske@mindspring.com
Email: fegan@rci.rutgers.edu
Email: tamar.gendler@yale.edu
George Graham: A.C. Reid Professor of Philosophy, Wake Forest University, Depart
ment of Philosophy, 1834 Wake Forest Road, Winston‐Salem, NC 27106. USA.
Email: grahamg@wfu.edu
John Heil: Professor of Philosophy, School of Philosophy and Bioethics, Building 11,
Monash University VIC 3800. AUSTRALIA. And: Professor of Philosophy, Washington
University in St Louis, Department of Philosophy, Wilson Hall, One Brookings Dr.,
Campus Box 1073, St. Louis, MO 63130–4899. USA.
Page 4 of 10
List of Contributors
Email: thorgan@email.arizona.edu
Email: Jaegwon_Kim@brown.edu
Email: klawlor@stanford.edu
Email: jle@philos.umass.edu
Email: e.j.lowe@durham.ac.uk
Page 5 of 10
List of Contributors
Email: c.macdonald@qub.ac.uk
Email: cmcginn@mail.as.miami.edu
Email: rjm@rci.rutgers.edu
Email: brianmc@rci.rutgers.edu
Email: almele@fsu.edu
Page 6 of 10
List of Contributors
Email: ruth.millikan@uconn.edu
Email: michellemontague@mac.com
Email: bmontero@gc.cuny.edu
Email: adam.morton@ualberta.ca
Email: david.papineau@kcl.ac.uk
Page 7 of 10
List of Contributors
Email: cp2161@columbia.edu
Email: john@csli.stanford.edu
Email: jesse@subcortex.com
Paul Raymont: Assistant Professor, Ryerson University, 350 Victoria Street, Toronto,
Ontario M5B 2K3. CANADA.
Email: praymont@ryerson.ca
Email: robinson@ceu.hu
Email: davidrosenthal@nyu.edu
Page 8 of 10
List of Contributors
Email: seager@utsc.utoronto.ca
Email: Gabriel.segal@kcl.ac.uk
Email: g.strawson@reading.ac.uk
Email: jtienson@memphis.edu
Email: mtye@mail.utexas.edu
Page 9 of 10
List of Contributors
Email: rnvangul@syr.edu
Email: s.walter@philosophy‐online.de
Ralph Wedgwood: Professor of Philosophy, Merton College, Oxford OX1 4JD. UNIT
ED KINGDOM.
Email: ralph.wedgwood@merton.ox.ac.uk
Email: yoo@oxy.edu
(p. xvi)
Page 10 of 10
Introduction
Introduction
Brian P. McLaughlin
The Oxford Handbook of Philosophy of Mind
Edited by Ansgar Beckermann, Brian P. McLaughlin, and Sven Walter
The philosophy of mind is a core area of philosophy. It has roots tracing back to ancient
Greece, where the idea that the soul is distinct from the body originated. Philosophy of
mind flourished during much of the period from the publication in the seventeenth centu
ry of René Descartes's magnificent works on the mind to the first quarter of the twentieth
century, which closed with the publication of C. D. Broad's The Mind and Its Place in Na
ture. With the rise of behaviourism in the second quarter of that century, however, philo
sophical interest in the mind waned. Still, some mid-twentieth-century work stands out.
With the fall of behaviourism in the late 1950s and early 1960s and the rise of cognitive
science, interest in the mind and in its place in nature was renewed — and with enormous
vigour. Since then research in the philosophy of mind has been booming.
Keywords: philosophy of mind, body and soul, behaviourism, cognitive science, René Descartes, ancient Greece
THE philosophy of mind is a core area of philosophy. It has roots tracing back to ancient
Greece, where the idea that the soul is distinct from the body originated. Philosophy of
mind flourished during much of the period from the publication in the seventeenth centu
ry of René Descartes's magnificent works on the mind to the first quarter of the twentieth
century, which closed with the publication of C. D. Broad's The Mind and Its Place in Na
ture (1925). With the rise of behaviourism in the second quarter of that century, however,
philosophical interest in the mind waned. Still, some mid‐twentieth‐century work stands
out; we think here especially of Gilbert Ryle's The Concept of Mind (1949), Herbert Feigl's
‘The “Mental” and the “Physical” ’ (1958), J. J. C. Smart's ‘Sensations and Brain Process
es’ (1959), and the literature on Ludwig Wittgenstein's private‐language argument that
appeared during this time period. With the fall of behaviourism in the late 1950s and ear
ly 1960s and the rise of cognitive science, interest in the mind and in its place in nature
was renewed—and with enormous vigour. Since then research in the philosophy of mind
has been booming. It is our hope that this volume will serve not only as a sourcebook on
philosophy of mind and a passageway into the vast contemporary literature, but also as a
stimulus to research in the twenty‐first century.
Page 1 of 25
Introduction
Even within the focus of the present volume we have relied on a division of labour. The
Oxford Handbook of Free Will (Kane 2002) rendered it unnecessary to include an essay on
free will; The Oxford Handbook on Wittgenstein (McGinn, forthcoming), rendered it un
necessary to include an essay on the private‐language argument; The Oxford Handbook of
Rationality (Mele and Rawlings 2004) rendered it unnecessary to include an essay on ra
tionality; The Oxford Handbook of Ethical Theory (Kopp 2007) rendered it unnecessary to
include an essay on moral psychology; and certain topics relevant to intentionality (e.g.
rule following and interpretation) are covered in The Oxford Handbook of Philosophy of
Language (Lepore and Smith 2006).
Still, even with all of this help, difficult editorial choices remained. Contributions to our
understanding of the mind trace back to the work of Plato and Aristotle, and forward
from them to the work of such giants as Augustine, Aquinas, Descartes, Spinoza, Leibniz,
Hobbes, Malebranche, Hume, Locke, Reid, and Kant. A number of the essays in the vol
ume include historical discussion. But the volume contains no essays exclusively devoted
to historical figures or to historical periods. Sweeping survey chapters treating the philos
ophy of mind during various periods of history would have been out of line with the level
of depth of the other chapters in the volume and, moreover, can be found elsewhere in
the literature; and there was no space for a proper treatment of historical figures. Rele
vant historical topics will be covered in future Oxford handbooks; for example, The Ox
ford Handbook of Plato (Fine, forthcoming) will contain an essay entitled ‘Plato on the
Soul’.
Page 2 of 25
Introduction
Had modern humans evolved with the physical prowess of Neanderthals, Oxford would
have allowed us to include more material. We are certain of that. As it is, Oxford allowed
us a volume that a normal human being can hold in two hands, not constraining the heft
of the volume to what can be held in a single hand alone. We are thus deeply grateful to
the press for the enormously generous way in which it resolved our initial dilemma. We
have a volume covering a cornucopia of core issues in contemporary philosophy of mind.
The volume addresses the cluster of issues that fall under the heading ‘the mind–body
problem’, and much more. If your interest lies in the place of mind in nature, the nature
of consciousness, the nature of intentionality, the relationship between consciousness and
intentionality, subjectivity, mental causation, mental content, propositional attitudes,
whether thought requires language, our epistemic access to our own and other minds,
folk psychology, personal identity, the unity of consciousness, or the nature of the self,
then you have come to the right Oxford handbook. Dualism, idealism, physicalism, func
tionalism, and panpsychism are examined. So are the leading psychosemantic theories—
the leading naturalistic theories of how the contents of mental states are determined.
There are, moreover, essays focusing on mental abilities such as the ability to think, to re
member, to imagine, to feel emotions, to focus our attention (p. 3) on perceived objects
and properties, and to engage in intentional action. The topics covered comprise the
heart of contemporary philosophy of mind.
All of the essays in the volume are previously unpublished. The contributors were chosen
both because of their expertise in the subject of their essay and because of the passion
they bring to the subject. They include many of the leading figures in the philosophy of
mind, as well as a number of rising stars. We are honoured to edit a volume with such dis
tinguished contributors.
The essays are long and substantive, engaging issues in depth. They aim not only to chart
the terrain as concerns a philosophical issue, but also often to argue for particular posi
tions. We encouraged the authors to indicate where they think the truth lies. In some cas
es authors were asked to present and defend philosophical positions that originated with
them. The authors, one and all, do philosophy, rather than simply reporting what contem
porary philosophical work has been done on a particular topic. (There are currently many
excellent encyclopedia and philosophical‐dictionary entries that provide such reports; in
deed a wealth of such material is available for free on the web.) The essays are one and
all contributions to the philosophy of mind.
As editors are wont to do, we have sorted the essays into sections based on interrelated
themes. But even given our broad section titles, it seemed in some cases an arbitrary de
cision to place an essay in one section rather than in another. We hope, in any case, that
the sections provide a useful guide to exploring the volume. We should note that the or
der of essays in each section was decided entirely independently of considerations con
cerning either the relative centrality of the topic covered to the philosophy of mind or any
assessment of the closeness of the content of the essay to the truth. Order was decided
solely by pedagogical considerations. But the essays can be profitably read in any order.
Each essay stands entirely on its own. A reader can begin anywhere. And a reader can
Page 3 of 25
Introduction
skip around from one section to another following only the direction of his or her inter
ests. To supplement the Table of Contents, we provide below a glimpse of the material in
each of the essays. These strictly matter‐of‐fact summaries are self‐contained, and so they
too can be read independently. Given that they are self‐contained, if one reads them in
one sitting one will encounter some redundancy. But it is also the case that if one reads
them in one sitting the relationships between various essays will become apparent.
Virtually all of the essays note that one or another issue in the philosophy of mind re
mains unresolved. Indeed, fundamental issues in the philosophy of mind remain unre
solved. The heart of the philosophy of mind continues to beat to the drum of dialectical
debate. This volume is testament to that. Positions defended in one essay are sometimes
attacked in another; a presupposition of one essay is sometimes challenged in another; a
problem presented as insuperable in one essay is sometimes claimed in another to be re
solvable; in some cases there is an important question whether a position taken in one es
say is in conflict with a position taken in another. We won't highlight these matters; high
lighting is unnecessary, since they will leap from the page even in the brief matter‐of‐fact
summaries below. Nor will we try to adjudicate any of the disputes or to indicate where
we believe the truth lies. We (p. 4) refrain from doing so not only because it would be in
appropriate in an introduction such as this one, but also because on a number of issues
there is no consensus even among the editors. We leave the adjudication of these philo
sophical disputes to the reader. One learns philosophy by doing philosophy. Engaging
with the essays in this volume will take one to the cutting edge of debates that animate
the heart of contemporary philosophy of mind.
As Kim points out, the problem of mental causation—the problem of how mental states
and events can have causal effects—is essentially coeval with the mind–body problem.
Descartes famously argued that mental substances are distinct from physical substances
in that mental substances are not located in space. Given this substance dualism, the
question arises as to how the mental and the physical are related. Descartes's answer was
that they are causally related; specifically that minds and brains causally interact. Certain
states of and changes in the brain cause states of and changes in a non‐spatial mental
substance; and states of and changes in the mental substance cause certain states of and
changes in the brain. In response, Princess Elisabeth wrote to Descartes: ‘I admit that it
would be easier for me to concede matter and extension to the mind than it would be for
me to concede the capacity to move a body and be moved by one to an immaterial
thing’ (Garber 2001: 172). As Kim notes, this may well be the first statement of a causal
case for physicalism. The debate over whether dualism can capture mental causation had
begun.
Page 4 of 25
Introduction
Papineau presents the considerations in favour of the principle of the causal closure of
the physical—roughly, the principle that every caused physical event has a sufficient
physical cause—from a historical perspective, examines ways in which the thesis can be
made more precise, and explores the connections between the principle and the issue of
whether two varieties of naturalism are true. He points out that the causal‐closure princi
ple is invoked in the causal argument for physicalism, the argument that anything with
physical effects must in some sense be physical: it must either be identical with some
thing physical or at least be metaphysically necessitated by something physical. Pap
ineau, moreover, discusses the role of the principle of the causal closure of the physical in
the defence of the philosophical doctrine of methodological naturalism—the doctrine that
‘philosophy uses the same methods of investigation as the natural sciences’—and in de
fence of the doctrine of ontological naturalism, ‘which says that the subject matter of phi
losophy coincides with that of the natural sciences’.
3. E. J. Lowe's ‘Dualism’.
As Lowe points out, mental–physical dualism comes in two varieties: substance dualism
and property dualism. He discusses both. He distinguishes Cartesian substance dualism—
the view that mental substances are immaterial entities that bear only mental properties
—from non‐Cartesian substance dualism, which claims that mental substances are the
Page 5 of 25
Introduction
bearers of mental properties but allows that they bear physical properties as well. Sub
stance dualists claim that a person is not her body (or any part of it). Two traditional ar
guments for this view are the conceivability argument and the indivisibility argument.
The conceivability argument runs (roughly) as follows: I can conceive that I could exist
without having a body; so I can exist without having a body; thus, I am not my body. The
indivisibility argument runs: My body is divisible into parts; I am not so divisible; so I am
not identical to my body. Lowe maintains that both arguments are wanting. But in his
view there are two more promising arguments for substance dualism; namely, the re
placement argument and the unity argument. The gist of the replacement argument is
this: If all parts of a person's body were gradually replaced by artificial substitutes, the
person would survive the procedure; neither her biological body nor any part of it would
survive it; thus a person is (p. 6) not her body or any part of it. The central premises of
the unity argument, which Lowe takes to be the more compelling of the two arguments,
are that (1) I am the subject of all and only my own mental states, and that (2) neither my
body as a whole nor any part of it could be the subject of all and only my own mental
states. In his discussion of property dualism Lowe focuses mainly on the question of how,
if property dualism is true, mental causation can be reconciled with the causal closure of
the physical—the thesis that every caused physical event has a sufficient physical cause.
The alleged tension between mental causation and the causal closure of the physical can
be overcome, he argues, if we recognize that there is a distinction between fact causation
and event causation. He maintains that a mental fact (a fact to the effect that someone is
in a certain mental state) can cause a physical fact without a mental event causing a
physical event.
One response to the problem of mental causation is to deny a presupposition of the ques
tion of how mental states and events have causal effects. The question of how they have
effects presupposes that they have effects. Epiphenomenalism is the thesis that mental
states and events have no causal effects. If epiphenomenalism is true, then although it ap
pears that mental states and events causally contribute to the production of our behav
iour, they are in fact devoid of behavioural effects, since they are devoid of any effects.
Behaviour and mental states are, rather, dual effects of common physical causes in the
brain. Walter discusses two prominent objections to epiphenomenalism. The first is the
argument from evolution, according to which since epiphenomena cannot be selected for,
and since the mind was selected for in the course of evolution, epiphenomenalism must
be false. The second is the argument from other minds, according to which epiphenome
nalism undermines any justification for our confidence that others enjoy a mental life.
Walter also considers an issue that has so far been largely neglected in discussions of
epiphenomenalism; namely, the issue of which account of causation would allow for a co
herent formulation of the epiphenomenalist's position that mental events are caused by
brain events but have no causal effects. He argues that on all the leading approaches to
causation either brain events fail to cause mental events or mental events have causal ef
fects; and so the epiphenomenalist's position may turn out to be incoherent.
Page 6 of 25
Introduction
The doctrine of anomalous monism was proposed and defended by Donald Davison. It is
the doctrine that every particular mental event is identical to some particular physical
event or other (token physicalism), but mental properties are not identical to physical
properties (property dualism); and, moreover, there can be no strict psychophysical laws
(psychophysical anomalism). Yoo presents Davidson's argument for token physicalism,
and two of his main lines of argument for psychophysical (p. 7) anomalism. She also ad
dresses two of the main concerns about the doctrine of anomalous monism. Given token
physicalism, Davidson can maintain that mental events are causes since every mental
event is identical with some physical event that has causal effects. One of the central
premises of Davidson's argument for token physicalism, however, is that causation re
quires subsumption under a strict physical law. This raises the concern that Davidson is
committed to a kind of type or property epiphenomenalism, according to which events
have causal effects only in virtue of falling under types (or instantiating properties) cited
in strict physical laws, and so never in virtue of falling under mental types (or instantiat
ing mental properties). Another concern about anomalous monism is that token physical
ism is too weak to be a substantive form of physicalism since it is silent about whether
mental properties are dependent on physical properties. Advocates of anomalous monism,
including Davidson, however, hold that mental properties are dependent on physical prop
erties in that they supervene on them: two events cannot differ in their mental properties
without differing in their physical properties. But the supervenience thesis in question
can be interpreted in a number of non‐equivalent ways, and it remains an issue whether
there is a psychophysical‐supervenience thesis strong enough for anomalous monism to
count as a substantive kind of physicalism yet weak enough to be consistent with psy
chophysical anomalism.
According to Baker, non‐reductive materialism is the view that although every concrete
particular is either microphysical or made up entirely of microphysical phenomena (so
that there are no immaterial souls, entelechies, or the like), it is nevertheless the case
that (1) there are mental properties that are distinct from any physical properties; that
(2) mental properties depend on physical properties; and that (3) mental properties make
a causal contribution to what happens. She points out that proponents of the doctrine
face the issue of how to reconcile (3) with (1) and (2). She examines in detail the leading
arguments that given the causal closure of the physical, (1) and (2) can be reconciled
with (3) only by assuming a kind of untenable overdetermination of physical events by
mental events. And she discusses her own version of non‐reductive materialism, the prop
erty‐constitution view, according to which mental properties are constituted by physical
properties. This version of non‐reductive materialism, she maintains, can accommodate
the causal efficacy of mental properties and intention‐dependent properties in a way that
is consistent with the causal closure of the physical and involves no untenable overdeter
mination.
Page 7 of 25
Introduction
Van Gulick discusses the historical roots of the functionalist approach to the mind and
distinguishes a variety of versions of functionalism. These versions all have in common
that types of mental states (events, processes) are to be understood in terms (p. 8) of the
functional roles that their instances play in suitably organized systems. The roles in ques
tion consist of relations to inputs to the system, outputs of the system, and to internal
states of the system. On one version of functionalism the relevant functional roles are
those specified by folk psychology; on another version they are the functional roles speci
fied by scientific psychology. The notion of functional role, moreover, has been specified
in a variety of ways—as computational role, as causal role, and as teleological role, or as
some combination of these. Van Gulick also distinguishes ontological functionalism and
analytical functionalism. Ontological functionalism concerns the nature of types of mental
states (events, etc.). On one version, occupant functionalism, a type of mental state is tak
en to be the occupant of a certain functional role; on another version, role functionalism,
it is taken to be the higher‐order state of being in some state or other that occupies the
functional role. Analytical functionalism is a claim about mental concepts; namely, that
such concepts can be analysed in functional terms. Van Gulick also discusses arguments
for and against the functionalist approach to the mind. He notes that the overall plausibil
ity of the functionalist approach depends on how it comes to grips with ‘the three big
C's’—content, causation, and consciousness. He maintains that with regard to content,
functionalism comes off well since virtually all current major theories of mental content
are in one way or another functionalist. As concerns causation, matters are more difficult,
given that it has been argued that the price of functionalism is the causal impotence of
the mental; but, van Gulick claims, this issue remains unresolved. Consciousness seems
to be especially difficult for functionalism because of the apparent possibility of inverted
and absent‐qualia scenarios. But even here, van Gulick argues, matters are not settled.
He maintains that for opponents of functionalism to simply insist that any functional roles
that may be specified by functionalists in the future will allow for inverted or absent
qualia would beg the question against the functionalist programme. Van Gulick concludes
that the objections that have been raised against functionalism are indecisive.
property physicalists must maintain that mental properties are reductively explainable in
terms of physical properties.
As Robinson points out, idealism is not a theory about the nature of mind, but rather a
theory about the nature of the physical world: idealists claim that the physical world is
dependent on mind; indeed, they hold that mind is the ground of the physical world.
There are, he notes, two main versions of this idea: (1) the physical world exists only as a
complex feature of experience; and (2) the physical world itself is a mind, or consists of
minds. He presents the two main lines of argument for idealism, one of which appeals to
sensory qualities (such as colour, taste, etc.), and the other of which appeals to features
of our conception of the world. The first line of argument is that there will be a vicious
regress of dispositional powers unless such powers are ultimately grounded in intrinsic
sensory qualities. The second line of argument is that our conception of the world, as
manifested in our language about it, possesses certain features that cannot be under
stood as representations of aspects of a concrete, mind‐independent reality, and so that
our conception of the world is not a conception of a mind‐independent reality; indeed,
even if anti‐idealists posit abstract entities such as universals, they cannot form any co
herent conception of how a world might be in itself, independently of categories that are
modes of understanding or interpretation. Robinson examines both of these lines of argu
ments in great detail, offering responses to the leading attempts to rebut them. He con
cludes that anti‐idealists have yet to show why these lines of argument fail.
Page 9 of 25
Introduction
Perry provides a detailed characterization of our subjective life within the framework of
understanding the mind as an informational system. He maintains that all of our experi
ences have both ‘feels’—‘what‐it‐is‐like for us as subjects’ aspects—and contents that can
be satisfied or fail to be satisfied. He thus holds that bodily sensations such as aches and
pains have contents as well as feels, and that even experiences of thinking something
have feels as well as contents. He examines a number of basic metaphysical and episte
mological issues concerning subjectivity. And he appeals to the different roles concepts
can play in our information‐processing systems to argue that Frank Jackson's knowledge
argument—an argument to the effect that one could in principle know all physical truths
and yet not know what it is like to have certain experiences—fails to refute physicalism.
Rosenthal points out that higher‐order theories of consciousness are concerned with the
problem of state consciousness; namely, the problem of what it is for a state to be a con
scious state. The basic idea of higher‐order theories is that a state is conscious if one is
conscious of the state in some suitable way. There is no circularity here since the mental
state by which one is conscious of the state need not itself be a conscious state. There are
two broad kinds of higher‐order theory. According to inner‐sense theories, the relevant
way in which the subject must be conscious of the state is by sensing the state; according
to higher‐order‐thought theories, the way in which the subject must be conscious of the
state is by having in a certain way a thought that she is in the state. Rosenthal discusses
both the considerations in favour of each of the two different kinds of higher‐order‐theo
ries and the difficulties that each faces. He distinguishes two kinds of higher‐order‐
thought theories: dispositional (p. 11) higher‐order‐thought theories and intrinsic higher‐
order‐content theories. He concludes with a discussion of the central issue of how being
conscious of a state with mental qualities can make the difference between there being
something it is like for one to be in that state and there being nothing that it is like.
Page 10 of 25
Introduction
Byrne claims that sensible qualities are qualities that we perceive, such as redness,
roundness, sweetness, and motion. When we veridically perceive a red apple, we perceive
not only the apple but also the redness of the apple, which seems to be a quality of the
apple itself. Sensory qualities are supposed to be qualities of experiences, rather than of
things that are experienced; other terms for sensory qualities, he says, are ‘qualia’ and
‘the phenomenal (or qualitative) character of an experience’. Sensational qualities, Byrne
claims, are supposed to be qualities of mental entities that are the immediate objects of
experience—sense data. Byrne claims that many historical figures, and some contempo
rary ones, try to identify sensible qualities with sensory qualities or with dispositions to
produce experiences with sensory qualities. He maintains, however, that although there
are sensible qualities such as redness, sweetness, and the like, it should be controversial
whether there are any sensory qualities. He further claims that if there are none, then a
lot of the motivation for secondary‐quality theories of sensible qualities—theories of sensi
ble qualities as dispositions to produce certain kinds of experiences or as the grounds of
such dispositions—evaporates. Moreover, he says that the hard problem of consciousness
—roughly, the problem of how to reductively explain the phenomenal character of con
scious states—arguably presupposes that there are sensory qualities; and so if there are
none, the problem dissolves. He recommends that we eschew positing sensory qualities
and see how quickly—or indeed whether—we run into sand.
The explanatory gap thesis is the claim that there is an unbridgeable explanatory gap be
tween the qualitative aspects of experiences (their ‘what‐it‐is‐like for the subject’ aspects)
and physical and functional properties. It thus entails that a reductive explanation of the
qualitative aspects of experiences in terms of such properties is impossible. Levine expli
cates the explanatory gap thesis, and discusses how it relates to certain arguments for
dualism. He notes that materialists respond to the thesis either by maintaining that the
explanatory gap is eliminable in that it can be bridged or else by arguing that although
there is an unbridgeable explanatory gap, the gap can be adequately explained without
assuming property dualism, and indeed in a way compatible with materialism. The cur
rent leading strategy for reconciling materialism with an ineliminable explanatory gap is
Page 11 of 25
Introduction
to try to explain the explanatory gap as a conceptual gap, rather than an ontological gap,
by appeal to the distinctive role of phenomenal concepts in our cognitive architecture.
Some implementations of the strategy maintain that phenomenal concepts afford us an
acquaintance with the qualitative characters of our experiences, and appeal to this ac
quaintance relation to offer an epistemic explanation of the explanatory gap. Levine
notes, however, that this faces the problem of how the relation of acquaintance itself can
be accounted for in physicalist or functional terms. Moreover, he maintains that all of the
extant implementations of the phenomenal‐concept strategy face formidable challenges
that their proponents have yet to meet.
Phenomenal concepts are supposed to be concepts that we can apply directly to our oc
current experiences without any mediating mode of presentation of the properties in
virtue of which they apply. The notion of a phenomenal concept figures prominently in re
cent debates between physicalists and property dualists concerning consciousness. Dual
ists tend to argue for their position by appeal to epistemological gaps between physical
(and functional) truths and truths about the qualitative aspects of experience. Proponents
of the phenomenal‐concept strategy argue that the existence of these epistemological
gaps can be fully explained by appeal to some (physicalistically explicable) features of
phenomenal concepts, and so without having to invoke property dualism. Balog examines
in great detail the phenomenal‐concept strategy and its uses in defence of physicalism.
She discusses as well both the various extant theories of phenomenal concepts and the
challenges that remain for proponents of the strategy.
A number of the leading arguments for dualism concerning consciousness start from a
premise about an epistemic gap between physical (or functional) truths and truths (p. 13)
about consciousness, and then infer that there is an ontological gap between physical
(and functional) phenomena and consciousness. These arguments include the knowledge
argument and various conceivability arguments. One conceivability argument runs as fol
lows: (1) it is conceivable that a physical duplicate of an individual who enjoys conscious
ness could be devoid of consciousness (and so be ‘a zombie’); (2) given that this is con
ceivable, it is metaphysically possible; (3) hence, it is metaphysically possible; (4) there
fore, materialism (or physicalism) is false for consciousness. One leading materialist strat
egy for responding to such arguments is to concede the epistemic gap but to deny that it
entails ontological dualism. Thus, a materialist might accept (1) and yet reject (2). The
framework of two‐dimensional semantics has been invoked to justify certain principles
linking certain epistemic facts to modal facts concerning what is possible. According to
two‐dimensionalism, terms (and on some versions concepts) have two intensions, a prima
ry intension and a secondary intension; and the primary intension, but not the secondary
intension, is relevant to matters of cognitive significance. It is claimed that this frame
work can accommodate Kripkean cases of a posteriori necessity (such as that it is neces
Page 12 of 25
Introduction
sary yet a posteriori that water is identical with H20), and that it can do so without avail
ing materialists of an avenue of response to the dualist arguments in question. Chalmers
spells out in detail how the two‐dimensional framework can be invoked to try to justify
certain principles linking the epistemic and the modal that can be appealed to in order to
establish dualism concerning consciousness on the basis of certain epistemic facts.
The idea that mental content is wide content is the idea that it is individuated at least in
part by environmental factors or social factors. If mental content is wide, (p. 14) then two
individuals can be exactly alike in every intrinsic physical respect, and yet have mental
states with different contents. Egan discusses in detail the Twin Earth thought‐experi
ments of Hilary Putnam and of Tyler Burge offered in support of different externalist the
ories of mental content, which entail that mental content is wide. She also discusses an
argument of Gareth Evans that certain thoughts have wide content in that they contain
individuals as constituents. She examines various challenges to these externalist theses.
She critically examines the idea that there are psychological contents that are narrow—
that are determined by intrinsic properties of appropriate individuals. Moreover, she ad
dresses a number of issues concerning wide content, including the issue of whether wide
content is causally relevant to behaviour. She defends the view that wide contents are
causally explanatory.
The idea that the content of a mental state is narrow is the idea that it is determined sole
ly by intrinsic conditions of the individual in the state. If the content of a mental state is
narrow, then two individuals exactly alike in every intrinsic respect would be such that
the one is in a mental state with that content if and only if the other is. Twin Earth
thought‐experiments developed by Hilary Putnam and Tyler Burge have been used by
Page 13 of 25
Introduction
them and others to argue that mental content is not narrow content, but rather wide con
tent in that it is partly individuated by environmental or social factors. Segal discusses
some theories according to which there is a kind of mental content that is narrow, and so
a kind of mental content that an individual on Earth and the individual's doppelgänger on
Twin Earth share. But he also presents a case that the various Twin Earth thought‐experi
ments simply fail to show that the contents we ordinarily attribute to mental states using
that‐clauses are wide, rather than narrow.
Page 14 of 25
Introduction
ing that only a normative kind of dispositionalism can give an adequate account of the na
ture of intentional mental states.
As Peacocke points out, the nature of concepts is a controversial issue that interacts with
the theory of thought, and with fundamental metaphysical and epistemological issues. He
discusses the view that concepts are abstract objects, the view that they are mental rep
resentations with syntactic and semantic properties, and the pleonastic view that they are
grammatical fictions. He discusses Frege's principle of individuation of concepts, which
he states as the principle that concept A is distinct from concept B if and only if a thinker
can rationally judge some complete (truth‐evaluable) content containing the concept A
without judging the corresponding content containing B. He discusses as well the prob
lem of what distinguishes conceptual content from non‐conceptual content. Moreover, he
explores a number of questions concerning concepts that remain open. Types of con
cepts, he holds, typically have rationality profiles, and one open question is: Can all as
pects of rationality specific to a given type of concept be explained in terms of the nature
of that type of concept? Another open question is: How should we conceive of the relation
between concepts and knowledge? Yet another is: What is the role of reference and truth
in the individuation of concepts? A fourth open question is: Is (p. 17) there some unified
explanation of the range of apparent characteristics of conceptual content? Do, for exam
ple, all these characteristics flow from the idea of concepts individuated in part by rea
sons for being in states involving them? Some concepts, he maintains, are individuated by
external factors. And he cites yet another open question: What is the correct account of
the relation between external individuation of mental states, epistemic norms for judging
given conceptual content, and the identity of the concepts in that content? The task re
mains, he notes, to integrate our metaphysics, epistemology, and theory of thought for the
domain of concepts themselves.
Bermúdez points out that notions of non‐conceptual content have been invoked in discus
sions of perceptual experience, of subpersonal computational states, and of the mental
states of non‐linguistic animals and pre‐linguistic humans. Thus, it has been argued, for
instance, that something can look F (e.g. right‐angled) to a visual perceiver who lacks the
concept of F, and thus that a perceiver's visual experience can represent something as F
without that involving any conceptualization of F by the perceiver. Also, it has been
claimed that computational cognitive states have contents, but that the subpersonal‐level
processes in which they participate do not consist of the exercise of concepts. And it has
been claimed that higher non‐linguistic animals and pre‐linguistic humans have proposi
tional attitudes despite lacking concepts. Bermúdez provides a detailed survey of these
and a number of other theoretical discussions. Moreover, he examines a number of argu
ments for and against the existence of non‐conceptual content. Maintaining that there is
non‐conceptual content, he engages the fundamental issues of what non‐conceptual con
Page 16 of 25
Introduction
tent is, and how non‐conceptual content is different from conceptual content. Further, he
sketches how Peacocke's notion of a scenario content, a kind of non‐conceptual visual
content, could be developed with the help of his (Bermúdez's) notion of canonical object‐
properties to yield a more adequate account of non‐conceptual visual content.
Crane embraces Brentano's thesis that all mental phenomena are intentional. All mental
states, not only states such as beliefs, desires, and intentions, but also emotions, percep
tual experiences, and sensations are intentional in that they are about or directed upon or
concern something. He explicates Brentano's thesis and presents various ways in which
the thesis has been developed in the literature. Moreover, he compares and contrasts rep
resentationalism with intentionalism. Representationalism is the position that the phe
nomenal or qualitative character of a conscious mental state (its ‘what‐it‐is‐like for the
subject’ aspect) is determined by its intentional (p. 18) content. Intentionalism is the view
that the phenomenal character of a conscious mental state is determined by its entire in
tentional nature. Crane argues that intentionalism is the superior of the two. He points
out that there is a distinction between intentional content and intentional mode. The be
lief that p and the hope that p have the same intentional content, but the states have dif
ferent intentional modes: one is a belief, the other a hope. Perceptual experiences, too,
can have different modes: the visual mode, the auditory mode, etc.; indeed, Crane main
tains that all conscious states have intentional modes. According to intentionalism, the
phenomenal character of a conscious mental state is determined by its content and its
mode, rather than by its content alone. Thus, phenomenal character supervenes on the
combination of intentional content and intentional mode: conscious states cannot differ in
their phenomenal character without differing either in their intentional content or their
intentional mode.
Montague addresses the question: What is the content of a perceptual experience? She
maintains that the answer is: Whatever is given to one in having the experience. She ar
gues that the content of an experience includes more than the worldly state of affairs that
it presents, the state of affairs that we express using a that‐clause when attributing the
perceptual experience. She argues that the content includes as well the phenomenal
character of the experience—the ‘what‐it‐is‐like for the subject’ aspect of the experience.
She argues further that it also includes an awareness of the experience itself, even
though this awareness is normally in the background, rather than in the foreground of at
tention, since attention is normally focused on the environmental state of affairs that we
are experiencing. Montague discusses this view of the content of perceptual experience
in the context of a wide range of central debates: the debate over whether experience is
phenomenologically transparent, the externalism versus internalism debate about inten
tional content, the direct versus indirect‐realism debate, and the debate over whether
Page 17 of 25
Introduction
As Graham, Horgan, and Tienson note, we are both conscious beings and intentional be
ings. We are conscious in that there is something it is like to be us; and we are intentional
beings in that we occupy representational mental states. They note that the currently
dominant view is that there are two independent features that a mental state can have: a
mental state can have representational (intentional) content and (p. 19) it can have a phe
nomenological character (a ‘what it‐is‐like for the subject’ aspect). They call the thesis
that phenomenology and intentionality are mutually independent ‘separatism’. More
specifically, separatism is the thesis that a mental state can have intentionality without
having a phenomenal character and a mental state can have a phenomenal character
without having intentionality. Separatists will typically claim, for example, that a belief
state has intentionality but no phenomenal character, and that a feeling of pain has a phe
nomenal character but no intentional content. Inseparatism is the thesis that phenome
nology and intentionality are inseparable. Graham, Horgan, and Tienson advocate a the
sis they call ‘moderate inseparatism’. They formulate the thesis as follows: mental states
that uncontroversially have intentional content have phenomenal character, and mental
states that uncontroversially possess phenomenal character have intentional content.
They spell out the separatism/inseparatism debate, present a case for moderate insepa
ratism, and address objections to it.
Strawson claims that when used by a human being the word ‘I’ sometimes refers to the
human being as a whole and that it sometimes refers instead to the self of the human be
ing at the time in question. In Strawson's view, a self is a subject of experience, and it is
not identical with a whole human being. The reason that selves are not identical to whole
human beings is that a self exists only as long as it is consciously experiencing. Human
beings, in contrast, can of course continue to exist when they have no experiences at all;
for example, in periods of dreamless sleep. There is not a single self that is the subject of
all of the experiences that a human being has over his or her lifetime. Rather, there are
many, many selves, each of which lasts exactly as long as the conscious experience of
which it is the subject. Strawson maintains, however, that selves, like whole human be
ings, are physical objects or substances, and so rejects mental–physical substance dual
ism.
Page 18 of 25
Introduction
Raymont and Brook discuss five varieties of unity of consciousness: (1) unified conscious
ness of an individual object, (2) unified consciousness of the content of an experience, (3)
unified consciousness of acts of experiencing, (4) unified consciousness of one's self, and
(5) unity of focal attention. They point out that the first four ideas are found in Kant and
the fifth in Wilhelm Wundt. They discuss how mental (p. 20) disorders such as dissocia
tive‐identity disorder, severe schizophrenia, dysexecutive syndrome, and simultagnosia
reveal how these ways in which consciousness is normally unified can break down. They
discuss as well the disunity of the consciousness in individuals who have undergone com
missurotomies.
Issues concerning personal identity include: (1) Given a person X at a time t, with which
past and future entities (if any) is X (numerically) identical? (2) Which facts determine the
answer to (1)? (3) On what bases do we ordinarily make judgements about (1) and on
what bases could we do so in principle? Gendler examines in detail John Locke's views on
such issues, and poses problems for any Lockean approach. She also discusses in detail
three contemporary approaches to issues of personal identity.
Imagination, McGinn claims, is a basic faculty of the mind intimately connected to various
mental capacities, including the creative capacities expressed in the arts and in science.
He distinguishes two types of imagination: sensory imagination, the forming of a mental
image, and conceptual imagination, entertaining a possibility in the conceptual style. He
contrasts sensory imagination with sensory perception, arguing both against the Humean
view that imagination is a kind of perception and against the Kantian view that percep
tion is a kind of imagination. As concerns cognitive imagination, McGinn argues that
imagining‐that is not a kind of believing‐that; unlike believing‐that, imagining‐that is sub
ject to the will, requires attention, and does not purport to be knowledge of the world. He
claims that imagining‐that gives us a (fallible) means to consider the merely possible, and
therefore bears on linguistic meaning. As he puts it, understanding the meaning of a sen
tence is the imaginative apprehension of the possibility expressed by it.
Page 19 of 25
Introduction
ture‐sensitive processes defined over the compositional structures of mental symbols pos
sessing contents.
Campbell maintains that our ability to know the reference of our singular and general
terms is based, ultimately, on our ability to focus our conscious attention on objects and
properties. Consciousness of objects and of properties, he tells us, provides us, respec
tively, with knowledge of the reference of our demonstrative singular terms and of the
reference of certain of our general terms. In spelling out this thesis he relies on Russell's
distinction between knowledge of truths and knowledge by acquaintance, and on
Russell's idea that the latter is more basic than the former. Campbell discusses both con
sciousness of objects and consciousness of properties within the context of G. E. Moore's
idea of the transparency of conscious awareness. Campbell argues, among other things,
that Russell's idea of conscious acquaintance with properties has been misunderstood in
the literature. As concerns perceptual conscious awareness of objects, he argues, among
other things, that it consists of a three‐place relation between a person, a standpoint, and
the object: the person is conscious of an object from a standpoint. The standpoint will dif
fer depending on the kind of sense modality in question. Campbell maintains that it is
consciousness of the object demonstrated that provides us with knowledge of the refer
ence of our demonstrative terms.
Lawlor provides a wide‐ranging survey of recent work in the psychology and in the philos
ophy of memory. She shows, moreover, the centrality of issues concerning memory in a
wide range of core philosophical subjects. These include autonomous agency, self‐con
sciousness, reference, mental content, and reasoning.
Page 20 of 25
Introduction
Prinz reviews the leading theories of emotions, and points out that those theories can be
divided into two categories: cognitivist and non‐cognitivist. Cognitivists claim that emo
tions are or require thoughts, while non‐cognivists deny this. He presents evidence in
support of a non‐cognitivist approach that builds on the James/Lange theory, according to
which emotions are feelings of changes in the body. He argues that emotions are an im
portant source of information, even though they are bodily feelings. Emotions, he con
tends, make a distinctive contribution to decision making since, unlike thoughts or judge
ments, they present information in a way that motivates action.
We have the ability to plan courses of action and to intentionally carry them out. What is
the relationship between intentions and intentional actions? What is it to act intentionali
ty? Intentions, Mele argues, differ from desires by their settledness. While one can desire
to do something without being at all settled on doing it, to intend to do something is, in
part, to be settled on doing it. Intentions are formed by decisions, and should cohere with
one's beliefs. Also, Mele examines two analyses of what makes an action intentional. The
first is that an agent S intentionally A‐ed if and only if S A‐ed the way S intended to A. The
second is that an agent S intentionally A‐ed if and only if S A‐ed for a reason. He points
out that both analyses face problems. How one responds to these problems depends in
part, he maintains, on what aim one pursues as a philosopher of action: on whether one
aims to analyse the folk concept of intentional action or to build an adequate theory of in
tentional action. Mele addresses a third topic as well: the causalism versus anti‐causalism
debate. According to Mele, standard causalism can be characterized by two theses: (1) An
event's being an action depends on how it was caused; and (2) proper explanations of ac
tions are causal explanations. Mele examines two objections to causalism—the problem of
causal deviance and the problem of vanishing agents—and argues that neither is decisive.
VI Epistemic Issues
41. Adam Morton's ‘Folk Psychology’.
Social life is based on our ability to ascribe mental states to other people and to predict
and explain their behaviour in terms of such states. This ability is called ‘mind reading’ in
the current literature. It has been claimed that this ability is based on knowledge of folk
psychology. But what, exactly, are the resources we rely on in making such ascriptions
and predictions? Two answers are much debated. Subtleties (p. 23) aside, the theory‐theo
ry tells us that we rely on an explicit theory that tells us what mental states a person
would be in in various circumstances, and what a person would do if she were in those
states. The simulation theory holds that we rely on the ability to simulate what mental
states we would be in various circumstances and what we would do. Morton maintains
that in fact both simulation and theory play a role, and so finds Stephen Stich and Shaun
Page 21 of 25
Introduction
Avramides critically examines two would‐be solutions to the problem of other minds and
defends another. The first is the argument from analogy. From my knowledge that I have
a mind, a mind that causes certain behaviour, and from the observation that there are
others who are physically similar to me and who behave similarly to me, I conclude that
these other individuals have a mind. Several objections have been raised against this ar
gument, including that the inference rests on an induction from a single case, and that it
is impossible to check the conclusion. The second argument is the argument from best ex
planation. The ascription of mental states to others provides the best explanation of their
behaviour under certain circumstances. Avramides claims that both of these solutions are
inadequate because they leave room for scepticism about other minds. According to
Avramides, our belief that we are not alone in the universe, that there are other people
who, for example, feel pain, is not a hypothesis; it is, rather, as Thomas Reid argued, a
first principle that leaves no room for coherent doubt. It is deeply rooted in our concep
tion of mind. Avramides claims that if we abandon the mistaken Cartesian conception of
mind, we can come to see that the relation between mind and behaviour is not a contin
gent one, but rather a necessary one.
Macdonald argues for her own version of an introspectionist account of authoritative self‐
knowledge by comparing and contrasting it with alternative accounts. As she points out,
introspectionist accounts of self‐knowledge fall within the broader domain of theories of
self‐knowledge, understood as views about the nature of and basis for one's knowledge of
one's own mental states and events, including one's beliefs, desires, conscious episodes of
thinking, and sensations. Theories of self‐knowledge are motivated by the apparent need
to account for a number of striking features of at least some such knowledge, which ordi
nary empirical knowledge, including knowledge of the mental states of others, is typically
Page 22 of 25
Introduction
thought to lack. Knowledge of certain of one's mental states is said to be epistemically di
rect or immediate in some sense (for example, in being non‐inferential and/or non‐evi
dence‐based), and privileged and/or authoritative, perhaps in being incorrigible, or infalli
ble, or transparent to oneself (or all three of these). Introspectionist theories attempt to
account for some, or all, of these features by reference to a special method by which this
knowledge is obtained. Macdonald argues that the main alternatives to her view account
for some of these features at the expense of others, or suffer from additional problems
that her account avoids.
According to semantic externalism, the contents of a subject's thoughts are at least partly
individuated by factors in her environment. One of the main objections to semantic exter
nalism is that it is incompatible with the fact that we normally have privileged access to
our thoughts; privileged in the sense that we can normally know what we are thinking a
priori; a priori in the sense that the knowledge is not based, directly or indirectly, upon in
vestigation of the environment. Brown notes that there are two main problems that pro
ponents of the view that semantic externalism and such privileged access are compatible
must address: the achievement problem and the consequence problem. The achievement
problem is that given semantic externalism it is difficult to see how a subject could
achieve privileged access; for it seems that if a subject's thought contents are partly indi
viduated by her environment, then the subject's knowledge of what she is thinking would
have to be based, at least indirectly, on empirical information about her environment. The
consequence problem is that the conjunction of semantic externalism and privileged ac
cess seems to have the consequence that we can have a priori knowledge of facts about
our environment that it patently seems could be known only a posteriori. Brown points
out several promising avenues for compatibilists to explore to resolve the achievement
problem. As concerns the consequence problem, she argues that it in fact does not follow
from semantic externalism that there are a priori knowable entailments from the fact that
one is having certain thoughts to contingent facts (p. 25) about the environment. She con
cludes that semantic externalism may well be compatible with privileged access.
Bach examines the so‐called paradoxes of self‐deception as well as a wide variety of other
issues concerning self‐deception. He notes that the view that being self‐deceived requires
having intentionally deceived oneself seems to have paradoxical consequences. It seems
to have the consequence that the self‐deceiver has succeeded in carrying out an intention
to believe something that the self‐deceiver believes is false, while retaining the belief that
it is false and knowledge of the stratagem. But how could one successfully carry out the
stratagem in such a circumstance? And how could one believe that p and believe that not
p at the same time? Some philosophers attempt to answer the first question by positing
unconscious intentions; some claim instead that in self‐deception one part of a person in
tentionally deceives another part of the person, so that the part that successfully carries
out the stratagem is not the part that is taken in by it. Some philosophers attempt to an
Page 23 of 25
Introduction
swer the second question by distinguishing two kinds of beliefs, for example, central be
liefs and avowed beliefs. Bach critically examines these various would‐be answers to the
questions. He argues in favour of rejecting the idea that self‐deception requires intention
ally deceiving oneself, and maintains instead that in self‐deception one unintentionally
misleads oneself in certain ways. And he argues that rather than self‐deception requiring
contradictory beliefs, in self‐deception one has a complex disposition to resist consciously
thinking something that one believes (or at least takes there to be strong evidence for
and normally would believe), and to affirmatively think something contrary to that propo
sition. Moreover, he explores in detail the role of attention and inattention in the various
activities by which a self‐deceiver unintentionally acquires and sustains the disposition in
question, activities that include evasion, rationalization, and what he calls jamming. The
last involves using one's imagination to conjure fanciful alternatives to the unwanted
truth, and using the mere possibilities in question to support the case that the matter is
unsettled. Bach discusses many other issues concerning self‐deception as well, including
how self‐deception can be twisted. Typical cases of self‐deception involve wishful think
ing. But in cases of twisted self‐deception a person is self‐deceived that p, yet wishes that
p were not the case. Whereas the straight self‐deceiver cannot face up to some unpleas
ant truth, and so evades it, the twisted self‐deceiver cannot get his mind off some unde
sirable falsehood, and ends up disposed to affirmatively think it.
References
Broad, C. D. (1925), The Mind and Its Place in Nature (London: Routledge & Kegan Paul).
Dunbar, R., and Barrett, L. (2007) (eds.), The Oxford Handbook of Evolutionary Psycholo
gy (Oxford: Oxford University Press).
Feigl, H. (1958), ‘The “Mental” and the “Physical”’, Minnesota Studies in the Phi
(p. 26)
Fine, G. (forthcoming) (ed.), The Oxford Handbook of Plato (Oxford: Oxford University
Press).
Garber, D. (2001), ‘Understanding Interaction: What Descartes Should Have Told Elisa
beth’, in D. Garber (ed.), Descartes Embodied: Reading Cartesian Philosophy through
Cartesian Science (Cambridge: Cambridge University Press), 168–88.
Kane, R. (2002) (ed.), The Oxford Handbook of Free Will (Oxford: Oxford University
Press).
Kopp, D. (2007) (ed.), The Oxford Handbook of Ethical Theory (Oxford: Oxford University
Press).
Lepore, E., and Smith, B. (2006) (eds.), The Oxford Handbook of Philosophy of Language
(Oxford: Oxford University Press).
Page 24 of 25
Introduction
Mele, A. R., and Rawlings, P. (2004) (eds.), The Oxford Handbook of Rationality (Oxford:
Oxford University Press).
Ryle, G. (1949), The Concept of Mind (Chicago, Ill.: University of Chicago Press).
Smart, J. J. C. (1959), ‘Sensations and Brain Processes’, Philosophical Review, 68: 141–56.
Stich, S., Samuels, R., and Margolis, M. (forthcoming) (eds.), The Oxford Handbook of
Philosophy of Cognitive Sciences (Oxford: Oxford University Press).
Brian P. McLaughlin
Page 25 of 25
Mental Causation
The problem of mental causation is essentially coeval with the mind–body problem.
Descartes arguably invented the latter when, in Meditation 2, he asked ‘But what then am
I?’ to which he replied ‘A thing which thinks’, and then went on to argue, in Meditation 6,
that ‘it is certain that this I is entirely and absolutely distinct from my body, and can exist
without it’. As every student of western philosophy knows, Descartes's view was that
minds and bodies constitute two disjoint categories of substance: minds are immaterial
substances whose essential nature is thinking, while bodies are material substances
located in physical space whose essence consists in being extended in space. Presumably,
substance dualism of this form was not startling news to anyone at the time. However,
Descartes, alone among the great rationalists of his day, urged a further view: minds and
bodies are in causal interaction with each other, minds influencing bodies in voluntary
actions and bodies influencing minds in perception and sensation.
Keywords: mental causation, mind–body problem, René Descartes, substance dualism, causal interaction,
voluntary actions
Page 1 of 28
Mental Causation
THE problem of mental causation is essentially coeval with the mind–body problem.
Descartes arguably invented the latter when, in Meditation 2, he asked ‘But what then am
I?’ to which he replied ‘A thing which thinks’, and then went on to argue, in Meditation 6,
that ‘it is certain that this I is entirely and absolutely distinct from my body, and can exist
without it’.1 As every student of western philosophy knows, Descartes's view was that
minds and bodies constitute two disjoint categories of substance: minds are immaterial
substances whose essential nature is thinking, while bodies are material substances
located in physical space whose essence consists in being extended in space. Presumably,
substance dualism of this form was not startling news to anyone at the time. However,
Descartes, alone among the great rationalists of his day, urged a further view: minds and
bodies are in causal interaction with each other, minds influencing bodies in voluntary
actions and bodies influencing minds in perception and sensation. Leibniz famously
denied causal relations between all monads; occasionalists like Malebranche allowed no
causal relations anywhere in the created world, with God as the sole causal agent; and
there probably was no room for genuine mind‐body causation in Spinoza's doctrine of a
single substance with mind and body as its two parallel attributes. It is not surprising
then that Descartes's contemporaries lost no time in contesting his interactionist thesis,
which, to many of us today, is probably the most commonsensical and plausible of the
doctrines that make up Descartes's theory of the mind. Among the prominent critics of
the interactionist thesis were Pierre Gassendi, Antoine Arnauld, and (p. 30) Princess
Elisabeth of Bohemia. The challenge posed to him was in essence this question: How can
substances with such diverse natures, one with thinking as its essence and not even
located in space and the other with extension and location in space, causally influence
each other? Thus was the problem of mental causation born.
Some commentators have attributed the fall of Cartesian dualism to Descartes's inability
to come up with a satisfactory response to this challenge (Watson 1987). What is
remarkable is the fact that the problem of mental causation didn't go away with the
downfall of substance dualism. Although substance dualism has not totally vanished from
the contemporary scene and there have recently been some attempts to resuscitate it,
philosophy of mind since the mid‐twentieth century has been dominated by a strong
physicalist outlook, and the agenda of the field has consisted largely in various
physicalist/naturalist programmes—the projects of ‘naturalizing’ centrally important
mental phenomena, such as intentionality, content, and consciousness. The re‐emergence
of mental causation as a major problematic in the 1960s was occasioned by the
reluctance on the part of most philosophers of mind to go all the way with physicalism.
These philosophers reject the dualism of mental and material substances, recognizing
only material things as the inhabitants of the world, and yet they are deeply committed to
the preservation of mental properties and kinds as genuine but physically irreducible
properties and kinds. And part of what makes mental properties ‘genuine’ properties is
their possession of causal efficacy. That is to say, these property dualists hold the view
that a mental property when possessed by a material thing endows it with distinctive
causal powers, powers that are irreducible to those of physical properties. The question
that has been raised concerning this combination of doctrines parallels the challenge
Page 2 of 28
Mental Causation
posed to Cartesian interactionist dualism: Given the irreducible diversity of the two
domains of properties, mental and physical, how is it possible for an object's mental
properties to have causal powers to affect, and be affected by, its physical properties?
Just as the essential diversity in their natures turned out, for substance dualists, to be an
insuperable obstacle to explaining causal relations between mental and material
substances, the purported autonomy of mental properties vis‐à‐vis physical properties has
proved, for physicalist property dualists, to be a seemingly intractable difficulty in
accommodating mind‐body causation.
Before we turn to the contemporary debate, it will be instructive to see wherein the real
difficulties lie for substance dualism in accounting for mental causation.
This suggests that physical causation requires contact—the object that causally influences
another must come into contact with it. In view of the transitivity of causation, the cause need
not be in direct contact with the object in which a change is caused, but it is clear that, on
Descartes's view, where there are causally related objects there must be objects in contact. And,
evidently, the very idea of contact requires spatiality. Given this, Princess Elisabeth's request to
Descartes, in her letter of May 1643, seems eminently natural and reasonable; she asked
Descartes to explain
how the mind of a human being can determine the bodily spirits in producing
voluntary actions, being only a thinking substance. For it appears that all
determination of movement is produced by the pushing of the thing being moved,
by the manner in which it is pushed by that which moves it, or else by the
qualification and figure of the surface of the latter. Contact is required for the first
two conditions, and extension for the third. [But] you entirely exclude the latter
from the notion you have of the soul, and the former seems incompatible with an
immaterial thing.2
Descartes's reply was that mind‐body causation must be understood through the idea of mind‐
body union, and that this is a primitive notion that is intelligible per se. Causation among
material things is causation by contact; however, the issue of contact does not arise with respect
to mind‐body union. I will comment on this approach later, but for now let us return to Princess
Elisabeth. Apparently unsatisfied by Descartes's response, she writes back:
Page 3 of 28
Mental Causation
And I admit that it would be easier for me to concede matter and extension to the
mind than it would be for me to concede the capacity to move a body and be
moved by one to an immaterial thing.3
I find this response remarkable—in fact, fascinating; it is, to my knowledge, the first causal
argument for physicalism that I know. She would rather physicalize the mind (‘easier for me to
concede matter and extension to the mind’) than accept causal relations between immaterial
minds and material bodies.
Princess Elisabeth's difficulty with mental causation within the Cartesian setting
depended on a particular conception of physical causation; namely, that physical
causation requires contact between cause and effect, whether direct or mediated. But
perhaps causation at a distance, without mediating chains of contiguous cause‐effect
pairs, cannot be ruled out a priori. I believe, though, that it can be shown that the
difficulties that beset causal relations involving immaterial, non‐spatial entities go deeper
and do not depend on the assumption that physical causation requires contact or
contiguity. They would arise even for causation, or action, at a distance (if such (p. 32)
should exist). A plausible argument can be developed, at a more abstract and general
level, that shows that causation is inseparable from spatiality, and that causal relations
between immaterial souls—that is, mind‐mind causation—are just as problematic as mind‐
body causation and for the same reason.
Here is the argument.4 Let us first consider an example of physical causation. A gun, call
it A, is fired and this causes the death of a person, X. Another gun, B, is fired at the same
time (say, in A’s vicinity—but the time and place are unimportant), and this results in the
death of another person, Y. What makes it the case that the firing of A caused X’s death
and the firing of B caused Y’s death, and not the other way around—that is, A’s firing
causing Y’s death and B’s firing causing X’s death? That cannot be an unexplainable brute
fact. There must be a relation R that grounds and explains the ‘cause‐effect pairings’, a
relation that holds between A’s firing and X’s death and also between B’s firing and Y’s
death, but not between A’s firing and Y’s death or between B’s firing and X’s death. What
is this R, the ‘pairing relation’ as we might call it? We are not supposing at this point that
there is a single such R for all cases of physical causation, only that some relation must
ground the fact that a given cause is a cause of the particular effect that is caused by it.
Two ideas immediately come to mind. The first is the idea of a causal chain: there is a
continuous causal chain from A’s firing to X’s death, as there is from B’s firing to Y’s
death, whereas no such chains link A’s firing with Y’s death or B’s firing with X’s death.
The second idea is that each gun when it was fired was at an appropriate distance and in
appropriate orientation relative to the person it killed, but not to the person it did not kill.
That is, spatial relations do the job of pairing causes with their effects.
A moment of reflection shows that the causal chain idea will not work as an independent
solution to the problem. A causal chain, after all, is a chain of causally connected events,
and interpolating more cause‐effect pairs will obviously not solve the pairing problem. It
is plausible to think that spatial relations, and more broadly spatio‐temporal relations,
must play a role in generating pairing relations. Intuitively, space seems to have nice
Page 4 of 28
Mental Causation
causal properties; for example, as distance increases, causal influence diminishes, and it
is often possible to set up barriers, at intermediate points, to block or impede the
propagation of causal influence. In any case, I can state my fundamental assumption in
general terms, and it is this: It is metaphysically possible for there to be two distinct
physical objects, a and b, with the same intrinsic properties and hence the same causal
potential or powers; further, one of these, say a, causes a third object, c, to change in a
certain way but object b has no causal effect on c. Now, the fact that a, but not b, causes c
to change must be grounded in some fact about a, b, and c. Since a and b have the same
intrinsic properties, it must be their relational properties with respect to c that provide an
explanation of their different (p. 33) causal roles vis‐à‐vis c. What relational properties, or
relations, can do this job? The only plausible answer seems to be that it is the spatial
relation between a and c, and that between b and c, that are responsible for the causal
difference between a and b vis‐à‐vis c (a was in the right spatial relation to c; b was ‘too far
away’ from c to exert any influence). At least, there is no other possible explanation that
comes to mind. Later I will give an explanation of what it is about spatial relations that
enables them to serve as causal‐pairing relations.
Now consider the possibility of immaterial souls, outside physical space, causally
interacting with material objects in space. The following again should be a metaphysically
possible situation: Two souls that have the same intrinsic properties5 act in a certain way
at the same time, and as a result a certain material object undergoes a change. Moreover,
it is the action of one of the souls, not that of the other, that is the cause of the physical
change. What makes it the case that this is so? What pairing relation pairs the first soul,
but not the second, with the physical object? Since souls, as immaterial substances, are
outside physical space, it is not possible to invoke spatial relations to do the pairing. What
possible relations could provide causal pairings across the two domains, one of spatially
located material things and the other of immaterial minds outside space?
Consider a variation on the foregoing example: There are two physical objects, P 1 and P 2,
with the same intrinsic properties, and an action of an immaterial soul causally affects
one of them, say P 1, but not P 2. How can we explain this? Since P 1 and P 2 have identical
intrinsic properties, they must have the same causal capacity (‘passive’ as well as ‘active’
causal powers), and it would seem that the only way to make them discernible in a causal
context is their spatial relations to other things. Doesn't that mean that any pairing
relation that could do the job in this case must be a spatial relation? If so, the pairing
problem for this case is unsolvable, since the soul is not in space and cannot bear spatial
relations to anything. The soul cannot be any ‘nearer’ to, or ‘more properly oriented’
toward, one physical object than another. Nor can we say that there is a causal barrier
‘between’ the soul and one of the physical objects but not the other; for what can
‘between’ mean as applied to something in space and something outside it? It is a total
mystery what non‐spatial relations there can be that might help distinguish, from the
point of view of an immaterial soul, between two intrinsically indiscernible physical
objects.
Page 5 of 28
Mental Causation
According to Descartes, the pineal gland is where the soul and the body meet for mutual
causal action. The soul of course cannot literally be in the pineal gland, though it was
claimed to move the gland to and fro and thereby initiate the motion of animal spirits.
One implicit assumption here is that my mind acts on my pineal gland, and your mind acts
on yours. More generally, a person is a ‘union’, Descartes said, of an immaterial mind and
a material body, and, as noted, he claimed the idea of mind‐body union to be a primitive,
per se intelligible notion. It is possible to construe this (p. 34) as Descartes's answer to
the pairing problem. That is, in our foregoing example, what distinguishes the two souls
in their causal relationship to a material thing, when this is a human body, is that the
first, not the second, forms a ‘union’ with it. Moreover, through this union relationship
with a material body your soul can causally reach other material things, since spatial
relations are now available to do the causal pairings between your body and material
things around it. Something like this may well seem to the committed dualist like a
reasonable way out, but we can hardly accept it as a solution to the problem at hand. The
reason is that mind‐body causation is implicitly, and ineliminably, involved in the notion of
a union of a mind and a body. A mind is united with that body with which it is in ‘direct’
causal relation, that is, without another body or mind serving as a causal intermediary.
This pineal gland, not that one, counts as mine precisely because my mind is in direct
causal contact with it and only with it. It is difficult to see what other explanation is
possible. Unless we understand mind‐body causation, therefore, we do not understand
mind‐body union. And we do not understand mind–body causation unless we have a
solution to the pairing problem for minds and bodies. Simply declaring mind‐body union a
primitive and yet intelligible notion, as Descartes apparently did, does not help.
But could there be causal interactions among immaterial minds? Ruling out mind‐body
causation does not in itself rule out the possibility of an autonomous domain of immaterial
minds in which minds are in causal commerce with other minds. Perhaps that is the
picture of a purely spiritual afterlife offered by some religions and theologies. Is that a
possibility? The pairing problem makes such an idea a dubious proposition. Again, any
substance dualist who wants causation in the immaterial realm must allow the possibility
of there being three mental substances, M 1, M 2, and M 3, such that M 1 and M 2 have the
same intrinsic properties, and hence the same causal powers, and yet an action by M 1,
but not the same action by M 2 at the same time, is causally responsible for a change in M
3. In such a situation, what pairing relation could connect M 1, but not M 2, with M 3? If
causation is to be possible within the mental domain, there must be an intelligible and
motivated answer to this question. But what mental relations could serve this purpose? It
is difficult to think of any; I don't think we even know where to begin.
Consider what space does for physical causation. In the kind of picture envisaged, where
a physical thing causally acts on only one of the two objects with identical intrinsic
properties, what distinguishes these two objects is their spatial locations. Space provides
a principle of individuation of material objects. Pure qualities and causal powers do not.
And what enables space to serve this role is the fact that physical objects occupying
exactly the same location in space at the same time are one and the same object.6 This is
in effect the venerable principle of ‘impenetrability of matter’, which can be understood
Page 6 of 28
Mental Causation
as an exclusion principle for space: Material things (p. 35) compete for, and exclude one
another from, spatial regions. From this it follows that if physical objects a and b bear
exactly the same spatial relations to some object c, a and b are one and the same object.
This principle is what enables space to individuate material things with identical intrinsic
properties. The same goes for causation in the mental domain. What is needed to solve
the pairing problem for immaterial minds is a kind of mental coordinate system, a ‘mental
space’, in which minds are each given a unique ‘location’ at a time. Further, a principle of
‘impenetrability of minds’ must hold in this mental coordinate system; that is, minds that
occupy the same ‘location’ in this space must be one and the same. I don't think we have
any idea how a mental space of this kind could be constructed.7 Moreover, even if we
could develop such a space for immaterial minds, that still would fall short of a complete
solution to the pairing problem; to solve it for causal relations across the mental and the
physical domains, we would need to somehow coordinate the two spaces to yield unitary
pairing relations across the domains. I doubt that we have any idea how something like
this could be done. No wonder Descartes claimed the mind‐body union to be an
unexplainable primitive.
consciousness, although it arises out of neural processes, can somehow loop back and
causally influence the course of lower‐level neural occurrences from which it has
emerged (Sperry 1969). Note that both Huxley and Sperry were primarily scientists; it
Page 7 of 28
Mental Causation
appears that for over three hundred years no philosopher of great prominence took up
the issue of minds’ causal powers as a major philosophical problem. But everything
changed in 1970 with the publication of Donald Davidson's ‘Mental Events’.
What is a ‘strict’ law? Strict laws are exceptionless and not hedged by ceteris paribus clauses.
And laws, unlike mere empirical generalizations, are able to ground subjunctives and
counterfactuals and confirmable by observation of positive instances (that is, they are
inductively projectible). There is also the suggestion that strict laws are part of a
‘comprehensive’ and ‘complete’ theory over some domain (Davidson 1970: 219); but a precise
construal of ‘strict’ in ‘strict laws’ will not concern us here. Note that if we take the mental/
physical dichotomy to be exhaustive (as Davidson seems to do), the anomalism of the mental
implies that if there are any strict laws they will all be physical laws.9
The position Davidson calls anomalous monism is a monistic physicalism claiming that all
individual mental events are physical events. An event is mental (or physical) just in case
a mental (or physical) description, or predicate, is true of it—or, as we might say, in case
it falls under a mental (or physical) kind. Let c be a mental event; in accordance with
Premise 2 above, let us suppose that c is a cause of some physical event e. By the
nomological character of causality (Premise 1), it follows that some strict law L must
subsume c and e. But the anomalism of the mental (Premise 3) says that L cannot be a
psychophysical law or a purely psychological law. So L must be a purely physical law.
Since c and e are subsumed by a physical law, each must fall under a physical kind (or
there must be a true physical description of each)—whence it follows that c, the mental
event, is a physical event. More generally, it follows that all mental events causally
connected with physical events are themselves physical (p. 37) events; and, since
causation is transitive, those mental events that are not directly causally related to
physical events are likely to be so related through other mental events. That, Davidson
thinks, should pretty much cover almost all mental events. In fact, Davidson's argument
can be made more general, yielding a stronger conclusion: it is plain that Premises 1 and
3, the nomological character of causality and the anomalism of the mental, together entail
the proposition that any event (of any kind) that enters into a causal relation with another
Page 8 of 28
Mental Causation
event (of any kind) is a physical event. This leaves out only events that are both causeless
and effectless, and it seems safe enough to disregard them.10 The conclusion, in any case,
is a form of physical monism: all mental events are physical events.
This is a causal argument for physicalism—or, more precisely, token physicalism. As such
it is reminiscent of Princess Elisabeth's remark to Descartes quoted above, that she
would rather ‘concede matter and extension’ to the mind—that is, take the mind as
something material—than accept the thesis that immaterial minds and material bodies
can causally influence each other. The core idea shared by Davidson and Elisabeth is that
the possibility of mind‐body causation requires minds to be physical. For Davidson, even
mental‐to‐mental causation requires both cause and effect to be physical events.
Davidson and his critics have had exchanges on this charge of epiphenomenalism;12
however, discussions of mental causation have largely moved away from Davidson's
anomalous monism. The problem of mental causation confronting Davidson is in essence
to answer this question: If the mental is anomalous (that is, there are no laws about the
mental), then, given that causation requires laws, how can mental events (p. 38) be
causes or effects? The ensuing discussion helped to bring the general issue of
epiphenomenalism to prominence, and philosophers began exploring the causal status of
mental properties in a broader context. As a result, the focus of discussion has shifted to
such questions as how mental causation is possible given that the physical world is
causally closed (the physical‐causal‐closure problem), whether purported mental causes
of physical events are excluded by physical causes (the exclusion problem), whether
events can be systematically overdetermined by having both a mental and a physical
cause (the overdetermination problem), and so on. We turn to these issues in the sections
to follow.
Page 9 of 28
Mental Causation
Let us now see how the spectre of epiphenomenalism has returned to haunt property
dualists. To begin, most ontological physicalists will accept the following closure
principle:
Causal closure of the physical domain. If a physical event has a cause at t, it has a
sufficient physical cause at t.
To deny this would imply that there might be non‐physical causal agents, outside the physical
domain, injecting causal influences into it, and that there might be physical events whose
explanations must invoke these non‐physical causal agents and forces. If this were the case,
theoretical physics would be in principle incompletable. Note that physical causal closure
understood this way does not exclude non‐physical causes of physical effects; the closure
principle only says that if there are to be such causes, there must also be sufficient physical
causes. No non‐physical cause of a physical event is essential, and there is no need to go outside
the physical domain to give a causal explanation of any physical event. Nor does the closure
principle imply (p. 39) that every physical event has a cause; there could be uncaused physical
events. In this sense, the closure thesis is consistent with the failure of physical determinism.
Further, as far as closure goes, there may be all sorts of non‐physical entities and events,
perhaps including mental ones, and there may be all kinds of interesting and significant causal
relations among them. The only thing that physical causal closure protects is the causal and
explanatory self‐sufficiency of the physical domain.
The dualist component of property dualism finds expression in the following proposition:
Anti‐reductionism. Mental properties are distinct from, and irreducible to, physical
properties.
Page 10 of 28
Mental Causation
Causal exclusion. No event can have more than one sufficient cause occurring at any
given time—unless it is a genuine case of causal overdetermination.
Given these propositions as premises, we can quickly show that property dualism runs into
trouble with mental causation.
Suppose some mental event, say an instance of mental property M, is a cause of a
physical event, an instance of physical property P. (For simplicity and uniformity, we take
events as instances, or instantiations, of properties; a mental event is an instance of a
mental property at a time, and similarly for physical events.) By physical causal closure,
there is a physical event, say an instance of P*, that is a sufficient cause of the instance of
P. Anti‐reductionism entails that M ≠ P*, whence it follows that the mental cause of the P‐
instance must be distinct from its physical cause. And we may stipulate that this is not a
genuine case of overdetermination; to contest this would result in the seemingly absurd
claim that all cases of mental‐to‐physical causation are cases of systematic
overdetermination involving a sufficient physical cause as well as the mental cause. At
this point the causal‐exclusion principle applies, and we must exclude either the mental
cause or the physical cause. Excluding the physical cause will not do any good, since if we
try to retain the mental cause, physical causal closure will kick in again and bring back
the excluded physical cause. The mental cause, therefore, must be let go.
That is the argument. It applies to all forms of property dualism that accept physical
causal closure. It is often referred to as an ‘exclusion argument’ because its point is that
any putative mental cause of a physical event is always ‘excluded’, or ‘screened off’, by a
physical cause. Physical causal closure implies that any putative non‐physical cause of a
physical event will face competition from a physical event for causal status. The exclusion
principle comes into play at this point, yielding the conclusion that in such competition
the physical cause always prevails. The overall implication of the argument, therefore, is
that mental events—and, more generally, all events in the special sciences—are
epiphenomenal with respect to physical events.
Property dualists come in two groups: those who accept mind‐body supervenience and
those who do not. At this time most property dualists belong to the former, and among
them are the so‐called non‐reductive physicalists. Ontological physicalism—or the
rejection of non‐physical stuff, like Cartesian mental substance—is the first step (p. 40)
toward physicalism. It is usually thought, however, that this is not nearly enough for
physicalism, and that any robust physicalist must accept a strong dependence of mental
properties on physical properties. It is customary to explain this dependence relation in
terms of supervenience: Mental properties supervene on physical properties. For our
present purposes, we may state this thesis in the following way:
Page 11 of 28
Mental Causation
According to supervenience in this sense, there are no mental events or states that float free
from physical processes; every mental phenomenon must be grounded in, or anchored to, some
underlying physical base (presumably, a neural state). This means that mental states can occur
only in systems that can have physical properties; namely, physical systems. This precludes
Cartesian souls as subjects of mentality. Moreover, supervenience entails that any two things
that are physically identical must be identical in all mental respects—to put it another way,
physical twins are necessarily psychological twins as well. This captures the idea that the
physical nature of a being wholly determines its mental nature. In these ways, mind‐body
supervenience, as a physicalist thesis, goes well beyond ontological physicalism.
As noted, the exclusion argument presented above applies to all forms of property
dualism, including those that do not accept supervenience. For property dualists who
accept supervenience, in particular non‐reductive physicalists, we can develop a second
argument that reveals a further epiphenomenalist implication of their position. The
exclusion argument has shown that property dualism has the consequence that mental‐to‐
physical causation is not possible. The argument to follow will show that property dualism
plus mind‐body supervenience is committed to epiphenomenalism tout court; that is,
mental events are not capable of causing other mental events any more than causing
physical events.
Here is the argument: Suppose that an instance of mental property M causes an instance
of mental property M*. From mind‐body supervenience it follows that M* has a physical
property, P*, as its supervenience base. This means that the instantiation of P* at t is
necessarily sufficient for M* to be instantiated at t, no matter what happened before t—
and, in particular, as long as P* is there, M* will be there even if the M*‐instance's
putative cause, the M‐instance, had not been there at all. This puts the causal status of
the M‐instance vis‐à‐vis the M*‐instance in jeopardy. The only way to salvage the M‐
instance's claim to be a cause of the M*‐instance appears to be this: we say that the M‐
instance caused M* to instantiate by causing its supervenience‐base property P* to
instantiate. This involves mental‐to‐physical causation: For any M‐instance to cause an
M*‐instance, it must first cause a P*‐instance, where P* is a physical supervenience base
of M*. More generally, the argument shows that, under supervenience, same‐level
causation implies ‘downward’ causation. We may call this the ‘supervenience’
argument.14
(p. 41)
To conclude, then, the exclusion argument has shown that mental‐to‐physical causation is
not possible. The supervenience argument shows that mental‐to‐mental causation is
possible only if mental‐to‐physical causation is possible. The two arguments together,
therefore, show that neither mental‐to‐physical nor mental‐to‐mental causation is possible
—that is, the mental is epiphenomenal tout court.
Page 12 of 28
Mental Causation
If you are persuaded by the arguments, you will have to reject at least one of the four
propositions, and your choice will determine, to a significant degree, the kind of position you will
occupy as regards the mind–body problem.
Who might want to reject (i), the physical causal closure? Emergentism is a form of
property dualism that does exactly that; for the emergentists typically believe in
‘downward’ causation—they claim that emergent phenomena, like consciousness, once
they have emerged from their ‘basal’ physical/biological conditions, can causally loop
downward on to their own physical emergence bases and influence the course of events
at the lower level (Sperry 1969; for discussion see Kim 1999). This is in clear violation of
the causal closure of the physical domain, since emergentism considers emergent
phenomena to be irreducible to and distinct from the (physical) phenomena from which
they emerge. Whether or not emergentism so conceived is a coherent position is a
question that has received some attention (see e.g. Kim 1999, 2006; O'Connor and Wong
2005); however, this goes beyond the scope of the present chapter. There are perhaps
other ontological physicalists who would reject (i); there seems nothing in ontological
physicalism per se that commits the position to (i).
(p. 42)
As noted, non‐reductive physicalists accept mind–body supervenience. For us, that is true
by definition; the point, though, is that there are a large number of physicalists who fit
the definition. Moreover, these philosophers will not regard giving up physical causal
closure as an option; their physicalist commitment is too deep for that. What this means
is that if they want to remain non‐reductivists, they must accept epiphenomenalism. This
Page 13 of 28
Mental Causation
One possible response that should not be neglected is a partial rejection of (iv) combined
with a partial rejection of (iii). For we should not think that the mental as a totality is
either reducible or irreducible; it may well be that part of the mental is reducible, though
not all of it is. It may well be—I believe this to be the case—that intentional/cognitive
states are physically reducible while phenomenal aspects of conscious experience, or
‘qualia’, are not so reducible. A position like this has been defended by David Chalmers
(1996; also Kim 2005). A physicalist who embraces a position like this must accept
epiphenomenalism for the physically irreducible mental residue.
(i) The generalization objection: the exclusion/supervenience arguments show too much,
and therefore they must be incorrect. If these arguments are sound, they generalize and
will show not only that there is no mental causation, but also that there is no biological
causation, no chemical causation, and no geological causation, leaving physical causation
as the only causation that exists. In fact, not all physical causation we normally accept is
safe because causal relations between macro‐level objects and events, such as a baseball
Page 14 of 28
Mental Causation
breaking the window and an overnight cold spell freezing the lake, will give way to
microphysical causation at the level of fundamental physics. This is absurd; the exclusion/
supervenience arguments cannot be correct (Van Gulick 1992: 325; Block 2003).
This objection is called the generalization argument (Kim 1998). There are some points to
be considered in reply. First, and the most important, is the fact that the objection, if
correct, could only add to the success of the exclusion/supervenience arguments. The
reason is that the arguments can be seen in two ways: (1) as an attempt to demonstrate
epiphenomenalism, and (2) as a reductio against one of the premises of the arguments.
We can agree with the proponents of the generalization objection that epiphenomenalism
is false—not only in regard to mental properties but also biological, chemical, geological,
and other special science properties. But then one of the premises must be rejected.
Which one? A committed physicalist has only one choice: the anti‐reductionist premise.
Reductionism wins. So if you are a non‐reductive physicalist, you are ill‐advised to attack
the exclusion/supervenience arguments by deriving further unpalatable consequences.
Second, the generalization argument moves too quickly, failing to consider the question
whether or not special‐science properties in biology, chemistry, geology, and so on are
reducible to physical properties. There is a strong initial intuition, though derided by
some, that mental properties, especially the phenomenal properties of conscious
experience (qualia) and the ‘aboutness’ of thoughts, are physically irreducible. I don't
believe we have such dualist intuitions about biological or chemical properties. We should
examine carefully the assumption, often all too readily made, that the properties
investigated in the special sciences, such as biology and geology, are irreducible as well.
We should be cautious—in fact, suspicious—about quick and loose arguments invoking
the so‐called multiple realizability of special‐science properties.16
(p. 44)
(ii) The causal‐drainage argument: current physics does not rule out the possibility that
there is no fundamental, ‘bottom’ level in microphysics; if there is no bottom level, causal
powers will drain away, and there will be no causation anywhere (Block 2003).
In a way, this is a continuation of the preceding objection. It contends that what the
exclusion/supervenience arguments show is that causation at one level gives way to
causation at the next, lower level. This means that if there is a fundamental microphysical
level, with absolutely basic particles (true metaphysical atoms), all causation will seep
Page 15 of 28
Mental Causation
down and get deposited at this level. That would mean that there is causation only at the
most fundamental micro‐level. But if there is no fundamental level, then the seepage of
causal powers cannot be stopped, with the result that there is no causation anywhere.
Full discussion of this interesting argument cannot be given here (see Block 2003; Kim
2003, 2005). It should be noted, though, that this objection, like the previous one, can
only play into the reductionist's hand. For what the objection does is to continue the
exclusion/supervenience arguments and derive further unwelcome consequences,
reinforcing the need for a reductio. The situation remains the same for physicalists: they
must choose between reductionism and epiphenomenalism.
This line of argument has been forcefully developed by Barry Loewer (2002). It is
appropriate to raise questions about the concept of causation operative in the exclusion/
supervenience arguments, and to consider, in particular, the question of what sorts of
causal concepts are required to underwrite causal exclusion and physical causal closure.
I believe Loewer is correct in saying that a robust notion of generation and production is
involved in the concept of causation driving the arguments.
This raises a host of substantive questions. To begin, is it really true that the notion of
causation as generation/production is absent from theoretical physics? We must
distinguish physics from the mathematical equations and formalisms taken to express
fundamental physical laws. It could be argued that it is our understanding and
interpretation of these formalisms, especially in the context of explanation and practical
control, that is crucial to the question of what sorts of causal concepts are operative in
physics. Moreover, the mere fact that a generative concept of causation is not present at
the fundamental microphysical level does not mean that it cannot appear at higher levels
(perhaps like the directionality of time). Further, we care about mental causation, it
seems to me, chiefly because we care about human agency, and evidently agency involves
a productive/generative notion of causation. An agent is someone who brings about a
state of affairs for reasons. If there indeed are no productive causal relations in the
world, that would effectively take away agency—and our worries about mental causation
along with it.
(iv) We can show that mental causation exists on the basis of the Humean regularity
account, the nomological account, or the counterfactual account of causation.
No one would take it seriously if anyone were to suggest that there is no problem for
mental causation since causation, as Hume is sometimes thought to have shown, is only a
matter of regularities (or ‘constant conjunctions’) and there surely are mental‐physical
regularities. Things don't improve much if this suggestion is strengthened by adding that
these regularities must be nomological, with a modal force appropriate for laws of nature,
and be able to support counterfactuals, be projectible, and so on. The reason is that non‐
causal nomological regularities can be generated by underlying causal processes, in the
way regularities governing medical symptoms are generated by the progress of an
underlying pathology. The crucial question, therefore, is whether these mind‐body
regularities represent not just laws but causal laws. There surely are interesting and
useful laws and regularities in the special sciences. According to the exclusion/
supervenience arguments, these regularities are not causal regularities, although they
may arise from genuine underlying causal regularities; mental‐mental regularities are no
more causal than the regular connections seen in a series of shadows cast by a moving
car.
Some appeal to the counterfactual approach to causation to show that mind‐body causal
relations are unproblematic (see Baker 1993; Loewer 2002). In a simple case the
argument would run like this: We can see that Sally's desire to get to the airport caused
her to hail a cab because we see that the counterfactual ‘If Sally had not wanted to go to
the airport, she wouldn't have hailed a cab’ is true. Here again we need to ask what
grounds the truth of counterfactuals of this kind, and whether their truth always provides
sufficient grounds for the attribution of a causal connection (in an appropriate direction).
The counterfactual approach to causation has been controversial, though influential. One
point to keep in mind is that there appears to be a general consensus that whatever the
virtues of the counterfactual theory of causation may be, it cannot give us a productive/
generative relation of causation. (Of course, some advocates of the counterfactual
approach, like Loewer, will not consider that a liability.)
(p. 46)
(v) Overdetermination is a feasible option. It has been suggested by some that we should
accept mental‐physical overdetermination; that is, the position that in every case in which
a mental event causes a physical event (or any kind of event) there is also a physical
cause of it. This proposal is a bit bizarre, and it is unclear whether, even if it can be
defended, it should count as a sufficient vindication of mental causal efficacy. However,
Page 17 of 28
Mental Causation
the approach has been seriously argued by a number of advocates (see Mellor 1995: 103–
5; Mills 1996; Garrett 1998). (Papineau (2001: 26–8) criticizes this approach; see also
Bennett 2003.)
x has property F at t = def. there exists some property P such that x has P at t, and
input of kind J applied to x causes P to instantiate in x and P's instantiation in turn
causes x to emit an output of kind K.
Both input and output can be complex—that is, input can be a combination of various stimulus
conditions and other mental states, and output can include further mental states as well as
behaviour and physical changes. J and K are the two parameters that fix the kind of property F
is. As an example, something like the following could qualify as a functional definition of pain (or
being in pain):
x is in pain at t = def. there exists some property P such that x has P at t, and tissue
damage or trauma to x causes P to instantiate in x and P's instantiation causes x to
wince and groan and to enter into a state of distress.
Page 18 of 28
Mental Causation
A human is in pain in virtue of having his C‐fibres stimulated, and for him C‐fibre stimulation is
the property that fills the causal role specified in the functional definition of pain. So we say that
C‐fibre stimulation is a realizer of the functional property of being in pain—it is a realizer for
humans. But, as Hilary Putnam famously argued (p. 47) (1967), pain's realizer in octopuses is
probably entirely different; it may well be that octopuses don't even have C‐fibres. Let's say then
that O‐fiber stimulation is the realizer of pain in octopuses. If there were sentient Martians, their
pain realizer might be X‐fibre stimulation. There is no upper bound to the number of possible
realizers of a functional property. According to functionalism, this phenomenon of multiple
realizability is the hallmark of mental properties, and it is what makes reductionist physicalism
false. This is so because mental properties, say pain, cannot be reductively identified with any
single physical realizer, say C‐fibre stimulation. Rather, mental properties as functional
properties are ‘second‐order’ properties, at a higher level of abstraction; pain as functionally
conceived abstracts away the physical/physiological details of its diverse multiple realizers, and
focuses instead on the role it plays in the psychological economy of organisms. What all
instances of pain have in common is not some single physical property or mechanism; it is the
causal role they play in the psychologies of the organisms and systems endowed with ‘tissue
damage detectors’.
Given this conception of a mental property, what can we say about its causal efficacy? Let
us first note the following two evident points:18
(1) A given instance of pain has the same causal powers as the instance of its
physical realizer. In general, the causal powers of an instance of a functional
property F are identical with (or perhaps a subset of) the causal powers of its
realizer. To put it another way, instances of a functional property ‘inherit’ their causal
powers from their realizers. (This is sometimes called the causal‐inheritance
principle.)
(2) P1 and P2 count as diverse realizers because they are causally (and
nomologically) diverse. If P1 and P2 have identical, or largely identical, causal
powers, they should count as one property and one realizer (the principle of causal
individuation of properties and kinds).
Page 19 of 28
Mental Causation
From these two points it follows that pain, as a functional property, is highly heterogeneous as a
causal type—as heterogeneous as its diverse physical realizers. Here we must count all
nomologically possible realizers of pain, not just its actual realizers. Given this, it is difficult to
believe that instances of pain—that is, all possible as well as actual instances—can have
anything causally in common beyond the causal role definitive of pain; namely, their being
caused by certain input conditions and their causing certain outputs. That is to say, pain as a
causal category is in danger of losing its scientific interest, for it is unlikely to enter into any
significant contingent lawlike regularities. The only general causal truth about all instances of
pain may well be the analytic one that directly follows from its functional definition. To be sure,
there may be interesting and significant regularities about human pains, or mammalian pains, or
octopus pains; but that would be because humans, mammals, or octopuses each share the same,
or similar, pain realizers, and there are likely to be significant empirical laws about each such
realizer. This means that there is virtually no likelihood that there can be a scientific theory of
pain as such, a theory that applies to all actual and (p. 48) nomologically possible pains. The
same goes for the possibility of a theory of perception as such, of learning and memory as such,
and so forth. What there can be—and what seems sufficient—is a theory of human pains, of
octopus pains, and so on; and a theory of human learning and memory, of octopus learning and
memory, of Martian learning and memory, and so on. If there are commonalities in the theory of
human learning and the theory of octopus learning, that must be due to the shared features of
the physical/biological realizers of learning in the two species (which in turn could be the results
of parallel evolution in shared environments).
Another consequence of (1) and (2) worthy of note is this: psychological properties
conceived as functional properties do not bring with them any new causal powers. All the
causal powers they represent are those of their realizers. We do not have to say that this
makes them epiphenomenal; it's only that no new and novel causal powers magically
materialize at the psychological level, as the emergentists envisaged. Some have touted
functionalism about the mental as a form—in fact, the principal form—of non‐reductive
physicalism, a position that protects the autonomy of psychology—and other special
sciences. But our results show that the autonomy purchased with functionalism does not
amount to much: there is a good chance that psychology will be sundered into many
psychologies along the lines dividing its many physical realizers, and these ‘local’ or
‘realizer‐specific’ psychologies are arguably reducible to the physical/biological theories
of the underlying realizer states and processes.19
Page 20 of 28
Mental Causation
(p. 49)
Essentially the same question arises when we think about how our propositional attitudes
—states with contents (usually specified by declarative sentences)—can serve as causes.
My belief that my turn is coming up for a dive from the high platform causes my muscles
to go tense. My belief that I've just pulled off an excellent dive makes me sigh with relief.
The two beliefs cause what they cause because of their contents. For the physicalist, what
ultimately cause these events, the sighing and the onset of muscle tension, are neural
events and states. As physicalists, we think of states of the brain as carrying contents—
that is, as representing certain states of affairs—and it is now widely accepted that what
particular content a given belief—that is, a given brain state—carries depends crucially
on the causal/cognitive/historical relations that the subject bears to his external
environment. This is content externalism. But, just like computational processes, brain
processes—successions of causally connected mental states—seem entirely insensitive to
what these states represent; they depend solely on the physical/biological configurations
of these states. How one brain state causes another depends, it seems, only on the local
physical/biological properties of these states, not at all on what these states might or
might not represent. In particular, if content externalism is true, what the states
represent is determined, at least in part but crucially, by the historical and environmental
circumstances in which the subject has been embedded. But neural causation strikes us
as being entirely local; that is, it depends only on what is here and now physically and
biologically.
Page 21 of 28
Mental Causation
So the problem is how to explain the causal efficacy of content. This is a complex
problem; although there have been important and valuable contributions in this area, it is
still essentially an open problem. I must conclude by listing some of the more significant
works in this area: Allen (1995); Baker (1995); Crane (1988); Drestke (1988, 1998); Fodor
(1987, 1989); Jacob (1997); Segal and Sober (1991); Yablo (1997).
References
Alexander, S. (1920), Space, Time, and Deity, II (London: Macmillan).
Allen, C. (1995), ‘It Isn't What You Think: A New Idea about Intentional Causation’, Noûs,
29: 115–26.
Baker, L. (1993), ‘Metaphysics and Mental Causation’, in J. Heil and A. Mele (eds.),
Mental Causation (Oxford: Oxford University Press), 75–95.
Bennett, K. (2003), ‘Why the Exclusion Problem Seems Intractable and How, Just Maybe,
to Tract It’, Noûs, 37: 471–97.
Block, N. (1990), ‘Can the Mind Change the World?’, in G. Boolos (ed.), Meaning and
Method. Cambridge: Cambridge University Press), 137–70.
Broad, C. D. (1925), The Mind and Its Place in Nature (London: Routledge & Kegan Paul).
Page 22 of 28
Mental Causation
repr. in Davidson, D. Essays on Actions and Events (Oxford: Oxford University Press,
1980), 207–25.
—— (1993), ‘Thinking Causes’, in J. Heil and A. Mele (eds.), Mental Causation (Oxford:
Oxford University Press), 3–17.
—— (1998), ‘Minds, Machines, and Money: What Really Explains Behaviour’, in J. Bransen
and S. E. Cuypers (eds.), Human Action, Deliberation and Causation (Dordrecht: Kluwer),
157–73.
Fair, D. (1979), ‘Causation and the Flow of Energy’, Erkenntnis, 14: 219–50.
—— (1989), ‘A Defence of Dualism’, in J. R. Smythies and J. Beloff (eds.), The Case for
Dualism (Charlottesville, Va: University Press of Virginia).
Page 23 of 28
Mental Causation
Gillett, C., and Rives, B. (2001), Does the Argument from Realization Generalize?
Responses to Kim’, Southern Journal of Philosophy, 39: 79–98.
Hall, N. (2004), ‘Two concepts of causation’, in J. Collins, N. Hall, and L. A. Paul (eds.),
Causation and Counterfactuals (Cambridge, Mass.: MIT Press), 225–76.
Huxley, T. H. (1874), ‘On the Hypothesis that Animals are Automata, and Its History’,
Fortnightly Review, 16: 555–80;
(p. 51) Jacob, P. (1997), What Minds Can Do (Cambridge: Cambridge University Press).
Kim, J. (1973), ‘Causation, Nomic Subsumption, and the Concept of Event’, Journal of
Philosophy, 70: 217–36.
—— (1993), ‘Can Supervenience and Non‐strict Laws Save Anomalous Monism?’, in J. Heil
and A. Mele (eds.), Mental Causation (Oxford: Oxford University Press), 19–26.
—— (1998), Mind in a Physical World: An Essay on the Mind–Body Problem and Mental
Causation (Cambridge, Mass.: MIT Press).
—— (2001), ‘Lonely Souls: Causality and Substance Dualism’, in K. Corcoran (ed.), Soul,
Body, and Survival: Essays on the Metaphysics of Human Persons (Ithaca, NY: Cornell
University Press), 30–43.
—— (2003), ‘Blocking Causal Drainage and Other Maintenance Chores with Mental
Causation’, Philosophy and Phenomenological Research, 67: 128–53.
Lepore, E., and Loewer, B. (1987), ‘Mind Matters’, Journal of Philosophy, 84: 630–42.
Page 24 of 28
Mental Causation
Loewer, B. (2002), ‘Comments on Jaegwon Kim's Mind and the Physical World’,
Philosophy and Phenomenological Research, 65: 655–62.
McLaughlin, B. (1989), ‘Type Epiphenomenalism, Type Dualism, and the Causal Priority
of the Physical’, Philosophical Perspectives, 3: 109–35.
Menzies, P. (2003), ‘The Causal Efficacy of Mental States’, in S. Walter and H.‐D.
Heckmann (eds.), Physicalism and Mental Causation (Exeter: Imprint Academic), 195–
223.
O'Connor, T., and Wong, H. J. (2005), ‘The Metaphysics of Emergence’, Noûs, 39: 658–78.
Searle, J. R. (1980), ‘Minds, Brains, and Programs’, Behavioral and Brain Sciences, 3:
417–24.
Segal, G., and Sober, E. (1991), ‘The Causal Efficacy of Content’, Philosophical Studies,
63: 1–30.
Page 25 of 28
Mental Causation
Sosa, E. (1993), ‘Davidson's Thinking Causes’, in J. Heil and A. Mele (eds.), Mental
Causation (Oxford: Oxford University Press), 3–17.
Stich, S. (1983), From Folk Psychology to Cognitive Science: The Case Against Belief
(Cambridge, Mass.: MIT Press).
Stoutland, F. (1980), ‘Oblique Causation and Reasons for Action’, Synthese, 43: 351–67.
Notes:
(1) Descartes (1641/1984). The quotations are taken from The Philosophical Works of
Descartes, trans. Elizabeth S. Haldane and G. R. T. Ross (Cambridge: Cambridge
University Press, 1931).
(2) This and other quotations from the correspondence between Princess Elisabeth and
Descartes are taken from Garber (2001). The present quotation is from p. 172 (except for
a correction of an apparent typo: in the last sentence ‘body’ has been replaced with
‘soul’).
(4) I will give only a brief sketch of the argument; for a fuller discussion see Kim (2001)
or Kim (2005: ch. 3). I first discussed the pairing problem in Kim (1973). My reflections
were prompted by Foster (1968). Foster was the first to discuss the pairing problem, but
he believes that a substance dualist can live with it: see Foster (1989). Haskell Fain
discussed a similar problem in Fain (1963).
(5) If you are inclined to invoke the identity of intrinsic indiscernibles for souls to
dissipate the issue, the next situation we will consider involves only one soul and the
supposed remedy does not apply. Moreover, the pairing problem can be generated
Page 26 of 28
Mental Causation
without assuming that there can be distinct intrinsic indiscernibles. This assumption,
however, helps present the problem in a simple and compelling way.
(6) There is the familiar problem of coincident objects—e.g. the statue and the bronze.
Some claim that although these occupy the same region of space and coincide in many of
their properties, they are distinct objects. I must set this problem aside, but it should be
noted, though some will dispute this, that coincident objects share the same causal
powers, except perhaps those associated with coming into being and going out of
existence.
(7) It will not help to try to bring souls into physical space and give them physical
locations. For details see Kim (2001) or Kim (2005: ch. 3).
(8) The term ‘downward causation’ is due to the biologist Donald Campbell, who
introduced it in Campbell (1974).
(9) In a later paper (1993) Davidson says that strict laws are found only in ‘developed
physics’.
(10) This amounts to the suggestion that Davidson's second premise be replaced by the
statement that every mental event has either a cause or an effect.
(12) For a defence of Davidson see Lepore and Loewer (1987); for Davidson's own reply
see Davidson (1993); for rejoinders and further discussion see Kim (1993), McLaughlin
(1993), and Sosa (1993).
(13) Thus, ‘property pluralism’ is a better name for this position than ‘property dualism’.
(14) In many places I have combined the two arguments as one and called it the
‘supervenience’ argument. This combined argument is also called the ‘exclusion’
argument in the literature. See Kim (1998) and Kim (2005). I believe that the usage I am
suggesting here is an improvement—more apt and clarifying. See also Yablo (1992).
(16) On multiple realization and reduction see Fodor (1974); Kim (1992); Bickle (1998);
Polger (2004); Shapiro (2004).
(19) For further discussion of functionalism and mental causation see Block (1990).
Page 27 of 28
Mental Causation
Jaegwon Kim
Page 28 of 28
The Causal Closure of the Physical and Naturalism
Over the latter half of the last century English-speaking philosophy became increasingly
committed to naturalistic doctrines. Much of this naturalistic turn can be attributed to the
widespread acceptance of the thesis that the physical realm is causally closed. This
article contains four sections. The first section offers an initial formulation of the thesis
that physics is causally closed and explains its immediate implications. The second
section then discusses the evidence for the thesis from a historical perspective. The third
section considers ways of making the thesis properly precise. Finally, the fourth section
explores the connections between the thesis and the more general issue of naturalism.
Keywords: naturalistic doctrines, physics, causal relation, naturalism, causal closure, causal–closure thesis
Introduction
OVER the latter half of the last century English‐speaking philosophy became increasingly
committed to naturalistic doctrines. Much of this naturalistic turn can be attributed to the
widespread acceptance of the thesis that the physical realm is causally closed.
This chapter will contain four sections. Section 2.1 offers an initial formulation of the
thesis that physics is causally closed and explains its immediate implications. Section 2.2
then discusses the evidence for the thesis from a historical perspective. Section 2.3
considers ways of making the thesis properly precise. Finally, Section 2.4 explores the
connections between the thesis and the more general issue of naturalism.
Page 1 of 14
The Causal Closure of the Physical and Naturalism
Physics, by contrast, does seem to be causally closed. If you consider any physical effect,
then there will arguably always be some prior sufficient physical cause: we expect to be
able to account for physical effects without leaving the physical realm itself. In particular,
this seems to hold even for physical effects which take place within the bodies of
conscious beings. When the muscle fibres in my arm contract, this is presumably due to
electrochemical activity in my nerves, which is due to prior physical activity in my motor
cortex, and so on. In principle, it would seem possible to account for this entire sequence
solely in terms of the resources offered by physics itself, and without making any
essential appeal to any other subject matter. (Isn't the causal closure of physics
undermined by quantum‐mechanical indeterminism? Well, quantum mechanics does call
for a more careful formulation of the causal‐closure thesis, but this doesn't matter for the
philosophical implications. Let us shelve this complication until Section 2.3.)
The causal closure of physics is solely a claim about how things go within physics itself. It
does not assert that everything is physical, but only that those things that are physical
have a physical cause. So it does not rule out realms of reality that are quite distinct from
the physical realm. It is entirely consistent with the causal closure of physics that there
should be self‐sufficient realms that operate quite independently of physical goings‐on (a
realm of ghostly spirits, say). The causal closure of the physical says only that when we
are within the physical realm we will find that every physical effect has a physical cause.
Even so, the causal closure of the physical does give rise to a powerful argument for
reducing many prima facie non‐physical realms to physics: for it indicates that anything
that has a causal impact on the physical realm must itself be physical. This is because the
causal closure of the physical seems to leave no room for anything non‐physical to make a
causal difference to the physical realm, since it specifies that every physical effect
already has a physical cause.
To see the significance of this point, note that many prima facie non‐physical realms—
such as the biological, meteorological, and mental realms—certainly do seem to have
physical effects. Not only are these realms causally affected by physical events, as noted
Page 2 of 14
The Causal Closure of the Physical and Naturalism
above, but they in turn exert a causal influence on the physical realm. An infection
(biological) can cause a physical rise in temperature; a hurricane (meteorological) can
destroy physical houses; my current thoughts (mental) can give rise to physical
movements of my fingers on a keyboard. However, the causal closure of the physical says
that these physical effects must already have physical causes: the rise in temperature will
be caused by cellular‐level chemical processes; the houses' destruction by fast‐moving
gases; the movements of my fingers by neuronal activity (p. 55) in my brain. How then
can the infection, the hurricane, and my thoughts also make a causal difference? The
causal closure of the physical would seem to leave no room for these prima facie non‐
physical causes to matter to the physical effects. Unless, that is, we identify the prima
facie non‐physical causes with the physical causes. If we equate the infection, hurricane,
and thoughts with the cell‐level activity, fast‐moving gases, and neuronal activity
respectively, then their causal efficacy will be respected. The biological, meteorological,
and mental causes won't be eclipsed by the physical causes, simply because they will be
one and the same as the physical causes—they will be ‘non‐physical’ only in the sense
that they are normally referred to using specialist (biological, meteorological, mental)
terminology, and not because they are ontologically different.
The thesis of the causal closure of the physical thus argues that many prima facie non‐
physical occurrences—all those that exert an influence on the physical realm—must
themselves in fact be physical. For otherwise it is hard to see how they could have any
physical effects.
At first sight it might seem as if causal closure follows from the presence of conservation
laws in physics: if there are laws specifying that important physical quantities stay
constant over time, won't this show that the later values of physical quantities must be
determined by earlier values? However, it depends on what conservation laws you have.
Not any set of physical‐conservation laws rules out non‐physical causes for physical
effects.
Page 3 of 14
The Causal Closure of the Physical and Naturalism
Descartes's physics might allow an independent mind to affect the brain, but Descartes's
physics is wrong. Later in the seventeenth century Leibniz replaced Descartes's law of
the conservation of ‘motion’ with the two modern laws of conservation of (vectorial)
momentum and of (scalar) kinetic energy, and thereby arrived at what we now regard as
the correct laws governing impacts. Now, Leibniz's physics, unlike Descartes's, did indeed
imply that the later values of all physical quantities are determined by their earlier values,
and therewith the causal closure of the physical. Leibniz thus faced the argument
outlined in my first section above: How can a non‐physical mind have physical effects,
given that the causal closure of the physical seems to leave no room for anything non‐
physical to make a causal difference to the physical realm (see Woolhouse 1985)?
However, Leibniz did not draw the modern physicalist conclusion that the mind must
therefore be identical to the brain. Rather, he denied that the mind in fact has any
physical effects. Since it seemed incontrovertible to him that mind and brain must be
ontologically separate, he instead inferred from the causal closure of the physical that the
mind is impotent to affect the physical world. (It only appears to do so because of the
‘pre‐established harmony’ with which God has arranged both the mental and physical
worlds.)
While Leibniz's physics implies the causal closure of the physical, this is not true of the
Newtonian system of physics that replaced it at the end of the seventeenth century. The
crucial difference is that, where Leibniz upheld the central principle of the ‘mechanical
philosophy’ and maintained that all changes of velocity are due to impacts between
material particles, Newton allowed that accelerations can also be caused by disembodied
forces, such as the force of gravity. Moreover, Newton's system was quite open‐ended
about the range of different forces that exist. In addition to gravity, Newton and his
followers came to recognize magnetic forces, chemical forces, forces of adhesion—and
indeed vital and mental forces, which arose specifically in living bodies and sentient
brains respectively. If we count vital and mental forces as ‘non‐physical’ (we shall return
to this point in the next section), then the admission of such forces undermines the causal
closure of the physical. For it means that physical effects, in the form of accelerations of
particles of matter, will sometimes be due to the operation of non‐physical vital or mental
Page 4 of 14
The Causal Closure of the Physical and Naturalism
causes. Given this, Newtonian science once more allowed that a non‐physical mind can
influence the physical world.
Doesn't Newtonian physics coincide with Leibniz's at least to the extent of upholding the
conservation of momentum and energy? So why don't these principles imply the causal
closure of the physical, as they did for Leibniz? The answer is that Newton's physics
differs from Leibniz's in the way the conservation of energy must be understood. Leibniz
upheld the conservation of kinetic energy. However, the existence of Newtonian force
fields means that this quantity is not always conserved: for example, two bodies receding
from each other will slow down due to their mutual gravitational (p. 57) attraction, and so
lose kinetic energy. Newtonian energy conservation specifies rather that the sum of
kinetic and potential energy is constant. Potential energy is the latent energy stored when
bodies are ‘in tension’ in force fields, as when two receding gravitating bodies cease to
move apart and are about to accelerate together again. The notion of potential energy
was not prominent in early Newtonian physics, but by the middle of the nineteenth
century physicists concluded that all forces operated so as to conserve the sum of
potential and kinetic energy—any loss of kinetic energy would mean a rise in potential
energy, and vice versa.
The emergence of this modern version of the ‘conservation of energy’ placed strong
restrictions on what kinds of forces can exist, but it by no means rules out vital and
mental forces. Provided that the fields of these forces store in latent form any losses of
kinetic energy they occasion (cf. the notion of ‘nervous energy’), the presence of non‐
physical forces will be perfectly consistent with the conservation of kinetic plus potential
energy. True, the conservation of kinetic plus potential energy does apparently imply that
all forces must be governed by deterministic force laws (otherwise what would ensure
that they always ‘paid back’ any kinetic energy they ‘borrowed’?) and this greatly
exercised many Victorian thinkers, especially given that nothing in early Newtonian
physics had ruled out spontaneously arising mental forces. But, even so, the Newtonian
conservation of energy does not stop deterministic vital and mental forces affecting the
physical realm.
Nevertheless, during the late nineteenth and the twentieth centuries an increasing
number of scientists have come to doubt the existence of vital and mental forces. The
most significant evidence seems to have come directly from physiology and molecular
biology, rather than from physics. Over the last hundred and fifty years a great deal has
come to be known about the workings of biological systems (including brains), and there
has been no indication that anything other than basic physical forces is needed to account
for their operation. In particular, the twentieth century has seen an explosion of
knowledge about processes occurring within cells, and here too there is no evidence of
anything other than familiar physical chemistry. The result has been that the
overwhelming majority of scientists now reject vital and mental forces, and accept the
causal closure of the physical realm (see Papineau 2002: app.; see also McLaughlin 1992).
Page 5 of 14
The Causal Closure of the Physical and Naturalism
At first sight it might seem that ‘physical’ can be defined by reference to the fundamental
categories of physical theory. However, this strategy generates an awkward dilemma for
advocates of the causal‐closure thesis (Hempel 1980). On the one hand, they can equate
‘physical’ with the category of phenomena recognized by current (p. 58) physical theory.
But then it seems implausible that ‘physics’ in this sense is closed: past form suggests
very strongly that physics will in time come to posit various new fundamental causal
categories. Alternatively, advocates of causal closure might wish to equate ‘physical’ with
the ontology of some ideal future physical theory. But then it is hard to see how the causal
closure of the ‘physical’ could have any current philosophical significance, given that we
are as yet quite ignorant of exactly what this ideal future physical theory will include.
However, this dilemma is by no means inescapable. True, neither current physics nor
ideal future physics gives us a suitable notion of ‘physics’ for framing the causal‐closure
thesis. But this does not mean there are no other notions of ‘physics’ that will serve this
purpose. Indeed there are arguably a number of different ways of understanding ‘physics’
that will yield a well‐evidenced and contentful causal‐closure thesis (Jackson 1998: 7–8).
For a start, we could simply define ‘physical’ as not sui generis mental or biological. This
understanding of ‘physical’ was in effect assumed at the end of the last section, in the
argument that the non‐existence of vital or mental forces establishes the causal closure of
physics. Note that nothing in that argument assumed a definitive list of fundamental
physical categories; rather the thought was simply that this list would not include any sui
generis mental or vital entities. This is a relatively inclusive understanding of ‘physical’: it
counts as a ‘physical’ cause anything that isn't mental or vital, and to this extent renders
the causal closure of the physical a relatively weak thesis. But even so it remains a thesis
of much philosophical interest, since it still argues that any mental or vital causes of
physical effects must be identical to causes that are not themselves sui generis mental or
vital.
A rather stronger reading of ‘physical’ would take it to cover any categories of the same
general kind as are recognized by current physical theory. The idea here is to understand
‘physical’ as referring to some mathematically characterizable set of primary qualities,
recognizably descended from the categories of modern physics, even if not identical to
them. On this reading the list of fundamental ‘physical’ categories will be taken to include
not just anything non‐vital‐or‐mental, but more specifically items that display the same
kind of simplicity as the categories of contemporary physics. Again, there seems good
Page 6 of 14
The Causal Closure of the Physical and Naturalism
reason to suppose that ‘physics’ in this sense is casually closed, and therefore that
anything which in this sense has ‘physical’ effects must itself be ‘physical’.
Finally, and even more specifically, there is the option of equating ‘physical’ with
microscopic. Modern physical theory characteristically deals with fundamental entities
that are very small. Given this, we might define ‘physical’ as referring to anything that is
composed of entities below a certain size (Pettit 1993). The causal‐closure thesis will then
hold that every effect that is microscopically constituted must have a cause that is
similarly microscopically constituted. Again, this thesis seems plausible, and argues that
everything that has microscopically constituted effects must itself be microscopically
constituted.
(p. 59)
Apart from worries about the meaning of ‘physical’, there are various other complications
with the causal‐closure thesis. So far this thesis has been presented fairly informally. A
more careful formulation would be this:
Every physical effect has an immediate sufficient physical cause, in so far as it has
a sufficient physical cause at all.
Consider first the requirement that the physical cause be ‘immediate’. This is needed to rule out
the possibility of physical causes which produce their physical effects only via non‐physical
intermediaries. Imagine, for example, that every bodily movement is indeed fully determined by
some prior brain state, but that these brain states only produce the bodily movements indirectly,
by first producing some sui generis mental effect, which then affects the nervous system, and so
on (see Lowe 2000). This set‐up would ensure that every bodily movement had some (indirect)
sufficient physical cause—namely, the prior brain state—but intuitively it wouldn't ensure the
causal closure of physics, since it wouldn't require that the direct causes of physical effects must
be physical, and therefore wouldn't ensure that every physical effect had a purely physical prior
history. The requirement that every physical effect have an immediate physical cause will rule
out any mixed physical/mental/physical histories of this kind, since it ensures that any physical
effect has an immediately prior physical cause, which in turn has an immediately prior physical
cause, and so on.
Now consider the requirement that the physical cause be ‘sufficient’. This is needed to
ensure that it causes the physical effect by itself, and not solely in virtue of its conjunction
with some sui generis non‐physical cause. Imagine, for example, that some neuronal
activity is caused by the conjunction of some chemical state and some sui generis mental
cause. Then that neuronal activity would have a physical cause (the chemical state), but
this cause would not have sufficed on its own, in the absence of the sui generis mental
factor. This is clearly less than we need for a philosophically significant closure thesis. To
make sure we have the right kind of closure thesis, we thus need to require that every
physical effect have a physical cause that suffices on its own.
However, this last requirement now highlights the need for the final qualification entered
above—that every physical effect has a sufficient immediate physical cause ‘in so far as it
Page 7 of 14
The Causal Closure of the Physical and Naturalism
has a sufficient immediate cause at all’. The reason for this latter qualification is to
accommodate the indeterminism of modern quantum mechanics, which tells us that
certain physical effects are random, without any sufficient determining cause.
At first sight it might seem as if this qualification will empty the causal‐closure thesis of
any interest, and in particular prevent it from implying that everything with physical
effects must itself be physical. For it might seem that quantum indeterminism creates
room for sui generis non‐physical causes (operations of the will, perhaps) to exert a
‘downwards’ influence on the physical realm, by influencing whether or not random
quantum events occur. However, this appearance is illusory. Quantum mechanics still
specifies that random physical effects have their probabilities fixed by sufficient
immediate physical causes. If we understand the causal closure of the physical as
covering this kind of physical determination of physical probabilities, (p. 60) then it will
once more rule out any sui generis non‐physical cause for a physical effect. For any such
sui generis cause would have to make a difference to the probability of the relevant
physical effect, and this would once more run counter to the causal‐closure thesis,
understood now as implying that this probability is already fixed by some sufficient
immediate physical cause.
The most immediate way in which causal closure bears on methodological naturalism is
by itself providing an important illustration of the methodological thesis. Recall how
Section 2.2 showed the causal‐closure thesis to be a highly empirical claim, whose
acceptance derives from detailed empirical evidence about the causes of physical effects.
This clearly does not stop the causal‐closure thesis from having substantial philosophical
consequences for the nature of mind and other philosophically interesting categories.
Methodological naturalists will say that this is a prime example of the way that
philosophical results, just like scientific results, characteristically depend on empirical
findings.
The causal closure of physics also promises to bring philosophy into close contact with
science at a more detailed level. The causal‐closure thesis promises to show that the
general run of prima facie non‐physical entities are in fact physical. If this is right, then
Page 8 of 14
The Causal Closure of the Physical and Naturalism
Exactly how far scientific investigation will be able illuminate such categories, however,
will depend on whether the causal‐closure thesis really does show them to be physically
constituted, and if so in what precise sense. So far we have not looked very closely at the
argument from causal closure to physical constitution. There is plenty of room for debate
about the range of items to which this argument applies, and about the precise sense in
which it shows these items to be physical. This now brings us to the bearing of the causal‐
closure thesis on ontological naturalism. In the rest of this final section I shall examine
exactly what the causal‐closure thesis implies about the physical constitution of reality.
(p. 61)
It will be convenient to proceed in two stages. First I shall consider those items to which
the causal‐closure thesis applies directly; namely, those entities that have physical effects.
After that I shall consider whether the causal‐closure thesis has any implications for
entities that are not themselves physically efficacious.
Take some prima facie non‐physical cause that has a physical effect. According to the
causal‐closure thesis, this physical effect already has a sufficient physical cause. So, on
pain of deeming this effect to have two independent causes, we need somehow to collapse
the non‐physical cause into the physical cause. But what exactly does this require? There
are different views about what is needed to render the two causes indistinct.
A minimum requirement is that the two causes be token identical, in the sense that they
both involve the same spatio‐temporal particulars. If some thought and some brain state
are both to be counted as the cause of my fingers moving over the keyboard, say, then it
seems clear that these causes cannot involve spatio‐temporally divergent particulars. If
they did, then we would have a substantial dualism of causes, a pair of causes involving
different particular substances.
Still, it is generally agreed that token identity on its own is not enough to rule out a
divergence of causes. It may rule out ‘substance dualism’, but it does not yet exclude
‘property dualism’. To see why, note that token identity is a very weak doctrine, since it
does not imply any relationship at all between the properties involved in the physical and
non‐physical causes—it is enough that the same particular entity should possess both
these properties. In this sense, my height is token identical with my weight, since these
properties are both possessed by the same particular; namely, me. Yet it seems wrong to
conclude on this account that my height causes what my weight causes. Similarly, it
seems unwarranted to conclude that my thoughts cause what my brain states cause,
simply on the grounds that these causes must involve the same particulars. Causes are in
Page 9 of 14
The Causal Closure of the Physical and Naturalism
some sense property‐involving, and any claim that two causes coincide will require that
the properties they involve are in some sense also coincident.
Type identity is the most obvious way to ensure that non‐physical and physical causes do
not comprise two distinct causes: if exactly the same particulars and properties comprise
a non‐physical and a physical cause, the two causes will certainly themselves be fully
identical. Still, type identity is a very strong doctrine. Type identity about thoughts, for
example, would imply that the property of thinking about the square root of two is
identical with some physical property. And this seems highly implausible. Even if all
human beings with this thought must be distinguished by some common physical
property of their brains—which itself seems highly unlikely—there remains the argument
that other life forms, or intelligent androids, (p. 62) will also be able to think about the
square root of two, even though their brains may share no significant physical properties
with ours.
This ‘variable realization’ argument has led many philosophers to seek an alternative way
of reconciling the efficacy of non‐physical causes with the causal‐closure thesis, one
which does not require the strict identity of non‐physical and physical properties. They
suggest that it is enough if the non‐physical properties metaphysically supervene on the
physical ones. According to this doctrine, a being's physical properties will fix its non‐
physical properties, in the sense that any two beings who share physical properties will
necessarily share the same non‐physical properties, even though the physical properties
which so ‘realize’ the non‐physical ones can be different in different beings. This arguably
ensures that nothing more is required for any specific instantiation of a non‐physical
property than its physical realization—even God could not have created my brain states
without thereby creating my thoughts—yet avoids any reductive identification of non‐
physical properties with physical ones. (Those who hold that non‐physical properties
generally supervene on physical ones are therefore called ‘non‐reductive physicalists’.
For discussion of this position, see the essays in Kim 1993.)
Page 10 of 14
The Causal Closure of the Physical and Naturalism
independent causes, each of which would have sufficed on its own—as when someone
dies because they are shot and struck by lightning simultaneously. But it seems quite
wrong to hold that all the physical effects of non‐physical causes are similarly
overdetermined by two independent causes.
The issue here hinges on the acceptability of different kinds of overdetermination. All can
agree that ‘strong overdetermination’ is absurd—the physical effects of non‐physical
causes surely don't always have two quite independent causes—and (p. 63) thus that
causal closure implies at least the metaphysical supervenience of the non‐physical on the
physical cause. But this leaves open the possibility of ‘weak overdetermination’, with the
relevant physical effects always having both a physical cause and a supervenient but
distinguishable non‐physical cause. Opinions differ on the acceptability of weak
overdetermination. Advocates of type identity argue that this is just as absurd as strong
overdetermination. (What difference can the non‐physical cause make, once the physical
cause has done its work?) Friends of supervenience counter that there is nothing wrong
with weak overdetermination. (In particular, it doesn't seem to carry the unhappy
implication that the result would have occurred anyway, even if one of the causes had
been absent—for the supervenience relation seems tight enough to ensure that both
causes would have been absent if either were; see Bennett 2003.)
Given the plausibility of the arguments for variable realization, the majority of
contemporary philosophers probably favour supervenience over type identity. Note how
this considerably dilutes the ontological implications of the causal‐closure thesis. It is one
thing to hold that any causally efficacious non‐physical property must be type identical
with some physical property. It is far weaker to claim that any such non‐physical property
must supervene on physical properties. A wide range of different kinds of non‐physical
properties arguably satisfy the latter requirement. In particular, to return briefly to the
methodological aspect of naturalism, it is by no means obvious that any property that
supervenes on physical properties must be the kind of property that is open to scientific
investigation. True, the most familiar kind of supervenient properties are functional
properties, properties which are constituted by the role they play in some system of
scientific laws, and these by their nature are appropriate objects of scientific
investigation. But functional properties are by no means the only kind of properties that
Page 11 of 14
The Causal Closure of the Physical and Naturalism
can be held to supervene on physical properties. For example, various kinds of properties
that are constituted by normative facts are widely held to supervene on the physical, and
it is doubtful that these are appropriate objects of scientific investigation.
Let me now turn to properties that are not themselves physically efficacious, and so
escape the immediate impact of the causal‐closure thesis. As observed in section 4.1, the
causal closure of physics itself leaves it quite open that there should be non‐physical
realms which are causally isolated from the physical—a realm of ghostly spirits, perhaps.
The argument from causal closure bites only on those items that do have physical effects,
implying that they can only have those effects if they are themselves physical.
This now opens the possibility that there are features of reality that do not even
supervene on the physical realm, notwithstanding the causal closure of physics. There are
two philosophically significant options to consider under this heading: first,
epiphenomenal mental kinds; second, commonsensically non‐causal kinds like
mathematics, modality, and morality.
The first option relates specifically to conscious mental states. Pretheoretically, these
might be supposed to have physical effects (most obviously of a behavioural kind) and
therefore to supervene on the physical realm. But many philosophers feel that (p. 64)
there are compelling reasons to deny that conscious properties are metaphysically
supervenient on the physical. (Aren't zombies metaphysically possible—beings who share
all your physical states yet have no conscious life at all?) Given causal closure, this non‐
supervenience threatens to imply strong overdetermination of all the physical effects of
conscious causes—unless, that is, we conclude that conscious causes don't have physical
effects after all. This epiphenomenalism option has been explored by various
contemporary philosophers. Persuaded that conscious properties can't possibly be
physical, they conclude that conscious states must be causal ‘danglers’, causal upshots of
brain processes, but without any effects themselves (see Jackson 1982; Chalmers 1996).
(Other philosophers, perhaps the majority, run the argument the other way, sticking to
the claim that conscious properties have behavioural effects, and concluding that they
must therefore supervene on the physical, despite initial philosophical impressions to the
contrary.)
Now consider mathematical, modal, and moral facts. Here there is plenty of room from
the start to doubt that they have physical effects. Mathematical entities are abstract,
outside space and time; modal facts concern realms of possibility other than the actual;
moral facts seem to transcend any natural facts. If it follows from such considerations
that these facts have no physical effects, then they escape the argument for physicalism.
Causal closure argues that anything with a physical effect must itself be physical. But this
carries no immediate implications for mathematical, modal, or moral facts, if they lack
physical effects. For all that has been argued so far, they need have no particular relation
to the physical world at all.
Page 12 of 14
The Causal Closure of the Physical and Naturalism
The idea of facts without physical effects does face other difficulties, however. In
particular, there are obvious epistemological queries about such facts. If mathematical,
modal, and moral facts—or, for that matter, epiphenomenal conscious facts—lack any
physical effects, then how can we possibly know about them? The standard route to
knowledge of the actual world is via perception, or at least via some kind of causal
interaction between our cognitive systems and the facts known about. And since our
cognitive systems are physically realized, this would seem to require that the facts known
about have physical effects. This makes it hard to see how we can know about realms of
reality which lack physical effects.
This argument is not conclusive. There remains the option that we can know about these
realms a priori, or in some other way that does not require causal interaction. A number
of contemporary philosophers defend theories along these lines. Still, it is not always
obvious that such non‐causal epistemologies can be made to work. An alternative
response to the epistemological difficulties is to conclude that the putative non‐causal
entities do not really exist, and that our discourse about them is no more than a useful
fiction. In addition, there remains the option of insisting that the problematic area is
metaphysically supervenient on the physical after all, and so capable of interacting
physically with our cognitive systems.
Let me conclude by summing up this final section. The causal‐closure thesis argues that
anything with physical effects must in some sense be physical. There may be room in
reality for non‐physical things that lack physical effects, but it is not obvious that we can
have knowledge of such things. However, even if our knowledge (p. 65) is restricted to
physical things, the requirement of physicality is not necessarily as demanding as it may
seem. Some take it to require type identity between physical and prima facie non‐physical
properties. But as many hold that it only requires the metaphysical supervenience of the
prima facie non‐physical on the physical, and this is a far weaker requirement.
References
Bennett, K. (2003), ‘Why the Exclusion Problem Seems Intractable, and How, Just Maybe,
to Tract It’, Noûs, 37: 471–97.
Crane, T., and Mellor, D. (1990), ‘There Is No Question of Physicalism’, Mind, 99: 185–
206.
Page 13 of 14
The Causal Closure of the Physical and Naturalism
Lowe, E. (2000), ‘Causal Closure Principles and Emergentism’, Philosophy, 75: 571–85.
David Papineau
Page 14 of 14
Dualism
Keywords: dualism, contemporary philosophy, substance dualism, property dualism, mental substances, individual
object
Page 1 of 21
Dualism
Substance dualism, I have just said, maintains that mental and physical substances are
distinct. But what are we to understand by a mental or physical substance in this context?
I have already explained that by a substance, here, we are to understand (p. 67) an
individual object or bearer of properties. By a mental substance, then, we are to
understand a bearer of mental or psychological properties, and by a physical substance
we are to understand a bearer of physical properties. So, by this account at least, we can
see that, whether we are talking about substance dualism or property dualism in the
philosophy of mind, the fundamental distinction in play is the distinction between mental
and physical properties. The property dualist holds that mental and physical properties
are distinct, while the substance dualist additionally holds that the bearers of those
properties are distinct. This being so, what is crucially needed at this point is a precise
and defensible account of the notion of a mental property, on the one hand, and that of a
physical property, on the other. These are, it seems clear, distinct notions, although
whether the properties of which they are notions are themselves distinct is, of course, one
of the main points at issue. However, it is one thing to say that these notions are distinct
—which would, I think, be almost universally accepted—and quite another to provide an
account of that notional distinction that would be to everyone's satisfaction. In fact, it has
proved remarkably difficult to produce an uncontentious characterization of either notion.
Some, but by no means all, philosophers have been happy to adopt Brentano's contention
that intentionality is the mark of the mental, for example (see Brentano 1874/1995 and,
for a modern defence, Crane 2001). Similarly, the proposal that physical properties are
Page 2 of 21
Dualism
those that form the subject matter of the science of physics (whether contemporary
physics, or an ideal or ‘completed’ physics)—or which, at least, depend ontologically upon
the properties studied by physics—is controversial, even if widely espoused. (For doubts
see Crane and Mellor 1990.) Fortunately, it is much easier to provide paradigm examples
of mental and physical properties which almost all parties to the debate will be happy to
accept as such. For instance, pain and desire are universally recognized as being mental
properties, while mass and velocity are universally recognized as being physical
properties. In what follows I shall take it for granted that the notional distinction now at
issue is a genuine one and that for practical purposes it can be captured by appeal to
such paradigm examples.
This accepted, let us return to the topic of substance dualism. The adherents of substance
dualism, as I have so far characterized it, contend that the bearers of mental properties,
such as pain and desire, are distinct from—that is, are not to be identified with—the
bearers of physical properties, such as mass and velocity. What are these ‘bearers’,
though? Well, the bearers of mental properties might be called, quite generally, subjects
of experience—understanding ‘experience’ here in a broad sense, to include not just
sensory and perceptual experience, but also introspective and cognitive states (‘inner’
awareness and thought) (see further Lowe 1996: ch. 1). Human persons—we ourselves—
provide prime examples of subjects of experience, but no doubt we should also include
examples drawn from the ‘higher’ reaches of the non‐human animal domain. I am not
insisting, then, that a subject of experience be capable of reason or self‐reflection, even
though possession of these capacities is very arguably a necessary condition of
personhood. Nonetheless, in what follows I shall normally be assuming that we are
talking about subjects of experience like ourselves; that is, persons. As for the bearers of
physical properties, for the purposes of (p. 68) the present discussion I shall mostly be
referring to bodies, or parts of bodies—on the understanding that what we are talking
about here are not mere lumps or masses of matter, but organized bodies and their parts,
the paradigm examples being the human body and its organic parts, such as the brain
and the neurons and other kinds of cell making up the brain and central nervous system.
In these terms, then, the substance dualist may be construed as holding that a person is
not to be identified with his or her body, nor with any part of it, such as the brain. On this
view, a person—not the person's body or brain—feels pain and has desires, even if it is
true to say that a person feels pain or has desires only because his or her body or brain is
in a certain physical state. The physical state in question—a certain pattern of excitation
in nerve cells, say—is not to be identified with the pain or desire consequently
experienced by the person, according to the substance dualist. A property dualist could
and no doubt would agree with this latter claim, however, while not accepting the further
claim that the person experiencing the pain or desire is distinct from that person's body
or any part of it. For the time being I shall focus on this further claim that is distinctive of
substance dualism, and return later to the common ground between substance and
property dualism. For, clearly, no argument to the effect that mental events or states
Page 3 of 21
Dualism
cannot be identified with physical events or states can suffice to establish the truth of
substance dualism, as opposed to the weaker doctrine of property dualism.
At this point I need to draw a distinction, for which I have not so far catered, between two
types of substance dualism. An apparent implication of what I have said so far concerning
substance dualism may be that, according to it, a bearer of mental properties—a subject
of experience—only bears mental properties, just as a bearer of physical properties, such
as a human body or brain, only bears physical properties. This was indeed the view of the
most famous substance dualist of all, René Descartes, for whom the human self or ego is
an immaterial substance (see Descartes 1641/1984 and, for prominent modern
sympathizers, Swinburne 1986 and Foster 1991). However, even if I am distinct from—not
to be identified with—my body or any part of it, as Descartes held, it does not
automatically follow that I can have only mental, not physical, properties. And, indeed,
there is a modern form of substance dualism—which may be called, aptly enough, non‐
Cartesian substance dualism—which differs from Cartesian substance dualism precisely
over this point. According to non‐Cartesian substance dualism, it is I, and not my body nor
any part of it, who am the bearer of mental properties, just as Descartes maintained.
However, unlike Descartes, the non‐Cartesian substance dualist does not make the
further claim that I am not the bearer of any physical properties whatsoever. This sort of
substance dualist may maintain that I possess certain physical properties in virtue of
possessing a body that possesses those properties: that, for instance, I have a certain
shape and size for this reason, and that for this reason I have a certain velocity when my
body moves (see Lowe 1996: ch. 2 and also Baker 2000). It doesn't follow that such a
substance dualist should allow that every physical property possessed by my body is also
possessed by me, however, for some of these properties may entail that the thing
possessing them is a body—and the non‐Cartesian substance dualist wants to deny, of
course, (p. 69) that I am a body. One such property, for instance, would appear to be the
property of being wholly composed of bodily parts, which is possessed by my body but
presumably not by me.
Prescinding, for the time being, from the distinction between Cartesian and non‐
Cartesian substance dualism, what sorts of arguments can be advanced in favour of such
dualism, and how good are they? Some of the best‐known arguments have been inherited
from Descartes himself and hence their contemporary versions may be described as ‘neo‐
Cartesian’. Two neo‐Cartesian arguments in particular are worthy of consideration: an
argument from the conceivability of disembodiment and an argument from the
indivisibility of the self. For brevity's sake, I shall call them the conceivability argument
and the indivisibility argument respectively.
The conceivability argument has both a strong and weak version, the difference in
strength being a difference in the strength of their premises—that is to say, the premises
of the strong version of the argument entail those of the weak version, but not vice versa.
That being so, one might suppose that the weak version is to be preferred, because it
assumes less. The weak version may be constructed as follows:
Page 4 of 21
Dualism
(1) It is clearly and distinctly conceivable that I should exist without possessing a
body.
(2) What is clearly and distinctly conceivable is possible.
Hence (3) It is possible that I should exist without possessing a body.
(4) If it is possible that I should exist without possessing a body, then I must be
distinct from my body.
Therefore (5) I am distinct from my body.
which clearly entails (1) (see Meixner 2004: ch. 3). The historical source of (1*) is, of course,
Cartesian doubt about the very existence of the physical world in its entirety—a doubt which at
least appears to be coherent and therefore to describe a possible state of affairs. As I say, one
might suppose the weak version of the argument to be preferable to the strong version because
it assumes less. However, it could be contended that (1) is only plausible, or at least is most
plausible, in the context of (1*), on the grounds that it is difficult to conceive of oneself as
existing in a disembodied state save under the hypothesis that the existence of the entire
physical world is an illusion.
Whether we consider the strong or the weak version of the conceivability argument, it
presents certain difficulties. First, in order to be valid it needs to be supplemented by a
further premise; namely,
Why? Because, clearly, if my body is not necessarily a body, then I cannot assume that a possible
world in which I exist without possessing a body is one in which I am distinct from my body: for
in that world my body might still exist and be identical (p. 70) with me, without being a body
(see Merricks 1994). However, (4+) should not be regarded as particularly controversial, I think.
This is because it is very plausible to maintain that it is an essential property of any entity that it
belongs to a certain ontological category, such as the category of bodies. To suppose that my
body might not have been a body is rather like supposing that the number 2 might not have been
a number, and seemingly absurd for the same reason.
Much more controversial is premise (2), that what is clearly and distinctly conceivable is
possible. (For well‐informed discussion see Gendler and Hawthorne 2002.) Let us grant
the truth of premise (1*), that it is clearly and distinctly conceivable that I should exist
without any body whatever existing, basing this claim on the coherence of Cartesian
doubt about the existence of the physical world. The content of such doubt is something
like this: Perhaps, for all that I know, the entire physical world as it seems to be presented
to me in perception is non‐existent and that perception is wholly illusory. This is a doubt
about the nature of the actual world, amounting to a surmise that the actual world
contains no physical objects although it does contain me and my mental states. I am
inclined to think that the surmise is at least a coherent, or logically consistent, one. But
the question is whether this is enough to establish that there is a possible world in which
I and my mental states exist but no physical objects exist. Of course, if the surmise is
correct, then the actual world is just such a world. But we are not given that the surmise
Page 5 of 21
Dualism
is correct, only that it is coherent. To this it may be replied that it suffices that the
surmise could be correct—it doesn't have actually to be correct. But the trouble, I think, is
that we simply don’t know whether or not it could be correct, because there may, for all
we know, be some reason why it couldn't be correct—a reason that we haven't yet thought
of. For instance, it might be that there simply couldn't be a world containing no physical
objects, whether or not it also contained me and my mental states.
We might sum up this response to the conceivability argument by saying that the trouble
with premise (2) is that it illicitly conflates ‘real’ or metaphysical possibility with mere
epistemic possibility. That is to say, (2) together with either (1) or (1*) does not serve to
ground the truth of (3), that it is possible that I should exist without possessing a body, in
the requisite sense of ‘possible’. The most that can be established by these means is that I
might actually exist without possessing a body, in an epistemic sense of ‘might’. This is
the sense of ‘might’ in which we can say, for instance, that there might be an even
number greater than 2 which is not the sum of two prime numbers, because we don't
know whether or not Goldbach's conjecture is true. But in the metaphysical sense of
‘necessary’, it is either necessarily the case or else necessarily not the case that every
even number greater than 2 is the sum of two prime numbers, so the matter is not in this
sense a contingent one. Likewise, then, we cannot assume that it is a contingent matter
whether or not I possess a body just because it is true that, in the epistemic sense, I
might or might not possess a body.
Let me pass on now to the indivisibility argument. This may be constructed as follows. (p. 71)
In this case both premises are open to question, although (7) less obviously so than (6). With
regard to (7), although it is indeed plausible to say that any extended region of space is divisible
into smaller subregions, it is not self‐evident that any object exactly occupying that larger region
must as a consequence be composed of parts occupying the smaller subregions. Of course, if the
object in question were wholly composed of homogeneous and infinitely divisible matter, this
would indeed be a consequence. But modern physics does not support the supposition that
matter of this kind exists. Indeed, it would appear that the so‐called fundamental particles
postulated by modern physics, such as the electron, are neither point‐like (since that would
imply that they possessed an infinite energy‐density) nor composed of anything smaller. So it
would seem that they provide actual counter‐examples to the assumption underlying premise
(7). Even so, whether or not that questionable assumption is correct, it seems evident enough
that my body is in fact divisible into parts of which it is composed—and so for the purposes of
the indivisibility argument we can replace the contentious (7) by the uncontentious
(7*) My body is composed of parts.
Page 6 of 21
Dualism
As for premise (6), however, it may now appear to be straightforwardly question‐begging, since
it simply denies that I possess a property that (7*) uncontentiously attributes to my body—
namely, the property of being a composite entity—and hence, it may be said, already presumes
the truth of the conclusion, (5), that I and my body are distinct. Certainly, if the indivisibility
argument is to acquire any persuasive force, an independent reason needs to be advanced in
support of premise (6).
My own view, I should say, is that premise (6) is indeed true, but that the most plausible
argument for its truth requires (5) as a premise, so that (6) cannot without circularity be
appealed to in an argument for the truth of (5). If (5) is to be successfully argued for,
then, we need to look elsewhere than to the indivisibility argument. I shall suggest an
alternative shortly. As for my chief reason for thinking (6) to be true—that is, for holding
the self to be a simple or non‐composite entity—it is that I consider the following
argument (and note that its first premise includes (5) as a conjunct) to be not just valid
but sound. (For a fuller account see Lowe 2001.)
(8) I am not identical with my body nor with any part of it.
(9) If I am composed of parts, all of those parts must be parts of my
(10) Anything that is wholly composed of parts of my body must either itself be a
part of my body or else be identical with my body as a whole.
Hence (11) I am a simple entity, not composed of any parts,
which is another way of expressing (6). The crucial premise here is, of course, (8), to which I
shall return in a moment. As for premise (9), this should be uncontentious (p. 72) in the context
of a debate between substance dualism and its physicalist opponents, since they will naturally
agree with (9), holding as they do that I am identical with my body or some part of it, such as my
brain. Premise (10) seems equally uncontentious—but more of that in a moment. I should
acknowledge, however, that not all substance dualists will be happy to assert premise (9). Some,
for instance, adopt the following view of the self: they hold that I am distinct from—not identical
with—my body, but am composed of it and another, immaterial, entity: my soul. On this view I am
a body–soul composite. (For discussion and criticism see Olson 2001 and Kim 2001.) Such a
composite is still a ‘substance’—that is, an individual object or property‐bearer—but one which,
in violation of premise (9), contains both parts of my body and something else, my soul, as parts.
Indeed, Descartes himself sometimes writes as if he endorses this view. I can only say that I find
it implausible and unattractive myself.
Another kind of substance dualist will reject premise (10), holding that I am wholly
composed of parts of my body and yet am not identical with any part of it nor with my
body as a whole. This kind of substance dualist sees the relation between me and my body
as being analogous to that between a bronze statue and the lump of bronze of which it is
made. On this view I am constituted by, but not identical with, my body (see especially
Baker 2000). And, indeed, the example of the bronze statue may be seen as posing a
threat to a generalized version of (10). For doesn't it show that it simply isn’t true that
anything that is wholly composed of parts of an object x must either itself be a part of x or
else be identical with x as a whole? For the bronze statue, it may be said, is wholly
composed of parts of the lump of bronze and yet is neither itself a part of the lump of
bronze nor identical with the lump of bronze as a whole. However, here a great deal turns
on the question of how, precisely, we are to understand the assertion that the bronze
Page 7 of 21
Dualism
statue is ‘wholly composed of parts of the lump of bronze’. If the assertion is taken to
mean that we can decompose the statue into parts all of which, without remainder, were
parts of the lump of bronze, then it is certainly true. For we can decompose the statue
into bronze particles, all of which were parts of the lump. But if, instead, the assertion is
taken to mean that all of the parts of the statue are also parts of the lump of bronze—
which is, mutatis mutandis, the interpretation that I was assuming in proposing premise
(10)—then it is far from evident that it is true. For example: the head of the statue—
assuming it to be a statue of a man—is a part of the statue and yet is not, plausibly, a part
of the lump of bronze (see further Lowe 2001).
However, is it not open to the constitution theorist to agree, now, with premise (10),
interpreted in the manner I intend and instead reject premise (9), although not for the
same reason that this was rejected by the proponent of the body–soul‐composite theory?
Cannot the constitution theorist say that, just as the statue has parts, such as its head,
that are not parts of the lump of bronze, so I have parts that are not parts of my body—
but not because I have any immaterial part or parts, any more than the statue has. In
principle, I agree, the constitution theorist could say this. However, I simply don't see
what these ‘additional’ parts could at all plausibly be. The reason why the statue has
parts that are not parts of the lump of bronze is that it (p. 73) has parts—such as its head
—that are, like the statue, constituted by, but not identical with, a portion of bronze. If,
analogously, I were to have parts that are not parts of my body, they would have to be
parts that are constituted by, but not identical with, parts of my body—just as, according
to the constitution theorist, I am constituted by my body as a whole. But there are, surely,
no such parts of me—no parts of me that are related to parts of my body in the way that I
am related to my body as a whole. As a self or subject of experience I do not, for example,
have lesser or subordinate selves or subjects as parts of me, each of them associated with
different parts of my body—as though I were a kind of collective or corporate self, on the
model of a company or club. (For more on the latter notion see Scruton 1989.) At least, it
certainly doesn't seem that way to me!
Page 8 of 21
Dualism
Entities possessing different identity‐conditions cannot be identified with one another, on pain of
contradiction (see further Lowe 1989: ch. 4). But what are ‘identity‐conditions’? Speaking quite
generally, the identity‐conditions of entities of a kind K are the conditions whose satisfaction is
necessary and sufficient for an entity x of kind K and an entity y of kind K to be identical; that is,
for them to be one and the same K. Thus, for example, the identity‐conditions of sets are these: a
set x and a set y are one and the same set if and only if x and y have exactly the same members.
In the case of things that persist through time, their identity‐conditions will also provide their
persistence‐conditions, since a thing persists through time just in case that same thing exists at
every succeeding moment during some interval of time. Now, there are, of course, notorious
difficulties attaching to the question of personal identity, and particularly to the question of what
conditions are necessary and sufficient for the identity of the self over time. However, even
without being able to settle this question, we may well be in a position to determine that the
identity‐conditions of the self, whatever they may be, are different from those of the body or any
part of it, such as the brain.
Here is one sort of consideration that seems quite compelling in this respect. We know
already that parts of the human body can be replaced by artificial substitutes which serve
the same function more or less equally well, as far as the person possessing that body is
concerned. For example, a ‘bionic’ arm can replace a natural arm and serve the person
who owns it pretty much as well as the original did. And, indeed, it seems perfectly
possible in principle that every part of a person's biological body should, bit by bit, be
replaced in this fashion, with nerve cells gradually being replaced by, say, electronic
circuits mimicking their natural function (see Lowe 1989: 120; Baker 2000: 122–3). If
such a procedure were carried out completely, as it seems it (p. 74) could be, the person
whose biological body had been replaced by an entirely artificial one would, very
plausibly, survive the procedure and so still exist at the end of it. And yet, clearly, neither
that biological body nor any part of it would have survived and still exist. If correct, this
shows that the persistence‐conditions of human persons are different from those of their
biological bodies and their various parts, such as their biological brains, and hence that
such persons—we ourselves—are not identical with those bodies nor with any of their
parts. In short, it establishes the truth of (8), the main claim of substance dualism. (I
should emphasize, however, that the cogency of claim (12) and the consequent support
that it provides for (8) do not stand or fall with the success of the preceding argument,
since there may be many other ways to show that my identity‐conditions differ from those
of my body or any part of it.)
Notice, however, that the foregoing argument for substance dualism—the replacement
argument, as we may call it—while it serves the purposes of non‐Cartesian substance
dualism well enough, is not sufficient to establish the truth of Cartesian substance
dualism, since the latter maintains that the self possesses only mental properties, not any
physical ones. For the replacement argument doesn't show that the self could survive in a
completely disembodied state and hence doesn't show that the self might exist even in
circumstances in which no physical properties whatever, such as shape or mass, could
possibly be attributed to it. The conceivability argument does purport to show this, of
course, but has been found wanting in persuasive force. As for the indivisibility argument,
Page 9 of 21
Dualism
it, like the replacement argument, cannot be used specifically in support of Cartesian
dualism, even setting aside the other difficulties that attach to it—for its conclusion is
only that I am distinct from my body, not that I lack, or could lack, any physical properties
whatever.
(14) Neither my body as a whole nor any part of it could be the subject of all and
only my own mental states.
Of course, (14) is the crucial premise, so let us see how it might be defended. First, then,
observe that my body as a whole does not need to exist in order for me to have each and every
one of the mental states that I do in fact have. If, for instance, I were to lack the tip of one of my
little fingers, I might as a consequence lack some of (p. 75) the mental states that I do in fact
have, but surely not all of them. I might perhaps lack a certain mildly painful sensation in the
fingertip—a sensation that I do in fact have—but many of my other mental states could surely be
exactly the same as they actually are, such as the thoughts that I am in fact having in composing
this essay. Indeed, I could still even have that sensation ‘in my fingertip’, because the
phenomenon of ‘phantom’ pain is a well‐attested one. However, I venture to affirm that no entity
can qualify as the subject of certain mental states if those mental states could exist in the
absence of that entity. After all, I certainly qualify as the subject of my mental states, as (13)
asserts, but for that very reason those mental states could not exist in my absence. Mental states
must always have a subject—some being whose mental states they are—and the mental states
that in fact belong to one subject could not have belonged to another, let alone to no subject at
all (see further Strawson 1959: ch. 3; Lowe 1996: ch. 2). But, as we have just seen, many and
quite possibly all of my own mental states could exist even if my body as a whole were not to
exist—that is to say, even if certain parts that my body actually possesses were not to exist. This,
I suggest, indicates that my body as a whole cannot qualify as the subject of all and only my own
mental states and so cannot be identified with me.
Now, many physicalists may agree with my reasoning so far but draw the conclusion that,
rather than being identical with my body as a whole, I am identical with some part of it,
the most obvious candidate being my brain. However, it is easy to see that the foregoing
reasoning can now just be repeated, replacing ‘my body as a whole’ by ‘my brain as a
whole’ throughout. For it seems clear that, although I may well need to have a brain in
Page 10 of 21
Dualism
order to have mental states, neither my brain as a whole nor any distinguished part of it
is such that it needs to exist in order for me to have each and every one of the mental
states that I do in fact have. Indeed, even if each and every one of my mental states
depends upon some part of my brain, it by no means follows, of course, that there is some
part of my brain upon which each and every one of my mental states depends. (To
suppose that this does follow would be to commit a so‐called quantifier‐shift fallacy.) And
yet I, being the subject of all and only my own mental states, am such that each and every
one of those mental states does depend upon me. Hence, we may conclude, neither my
brain as a whole nor any part of it can qualify as the subject of all and only my mental
states and so be identical with me. Putting together the two stages of this train of
reasoning, we may thus infer that (14) is true and from that and (13) infer the truth of (8),
the main claim of substance dualism.
However, here it may be objected that the foregoing defence of premise (14) depends
upon an illicit assumption; namely, that if my body as a whole were to lack a certain part,
such as the tip of one of my little fingers, then it—my body as a whole—would not exist.
This assumption is unwarranted because it presupposes, questionably, that every part of
my body is an essential part of it, without which it could not exist. As it stands, this may
be a fair objection—although it should be acknowledged there are some philosophers who
do hold that every part of a composite object is essential to it (see e.g. Chisholm 1976: ch.
3). However, I think that the reasoning in favour of premise (14) can in fact be formulated
slightly (p. 76) differently, so as to make it independent of the truth of this assumption.
The initial insight still seems to be perfectly correct—that, as I put it, my body as a whole
does not need to exist in order for me to have each and every one of the mental states
that I do in fact have. Thus, to repeat, the thoughts that I am having in composing this
essay plausibly do not depend upon my body including as a part the tip of one of my little
fingers. Call these thoughts T. Consider, then, that object which consists of my body as a
whole minus that fingertip. Call this object O and call my body as a whole B. (It should be
conceded here that there are some philosophers who would deny that any such object as
O exists—see e.g. van Inwagen 1981—but that is, to say the least, a controversial claim.)
Suppose, now, that it is proposed that I am identical with B, and hence that B is the
subject of the thoughts T. Then we can ask: On what grounds can B be regarded as the
subject of T in preference to O, given that T do not depend upon B's including the part—
the fingertip—that O does not include? Isn't the material difference between B and O
simply irrelevant to the case that can be made in favour of either of them qualifying as the
subject of T? But in that case we must either say that both B and O are subjects of T, or
else that neither of them is. We cannot say the former, however, because B and O are
numerically distinct objects, whereas the thoughts T have just one subject—myself. We
may conclude, hence, that neither B nor O is a subject of T and thus that I, who am the
subject of T, am identical with neither of them. This sort of reasoning can then be
repeated, as before, with respect to any specific part of B, such as my brain.
Page 11 of 21
Dualism
However we exactly formulate the defence of premise (14), the basic point of the unity
argument, as I call it, is that my mental states do not all depend on my body as a whole or
on any part of it in the unified way in which they all depend upon me as their subject.
This point, it seems to me, is a good one. Indeed, between them, the unity argument and
the replacement argument provide, I think, very compelling grounds for belief in the
truth of non‐Cartesian substance dualism. Moreover, if this form of dualism is true then
so, a fortiori, is property dualism, in its guise as a dualism of mental and physical events
or states—on the widely accepted assumption, at least, that events or states consist in the
possession of properties by substances at times. For if a mental event or state, such as a
thought or a sensation, always and only consists in the possession of a mental property by
a subject at a time, and a subject cannot be identified with its physical body or any part of
it, then it follows that such an event or state cannot be identified with any physical event
or state which consists in the possession of a physical property by that body or any part of
it. For, on this account of the identity‐conditions of events or states, events or states are
individuated in part by the objects in whose possession of properties at times they consist
(see further Lowe 2002: ch. 12). It is interesting—and perhaps even surprising—that we
should be able to arrive at this conclusion without entering at all into the question of how,
exactly, mental and physical properties are themselves related. But I think that we can
indeed do so.
However, compelling though the unity and replacement arguments seem to be to me, I
recognize that there are many other philosophers whom they will fail to convince. So at
this point I want to turn to the doctrine of property dualism in its own (p. 77) right, rather
than as an adjunct of non‐Cartesian substance dualism. For there are independent
considerations in favour of property dualism which have little or nothing to do with any
arguments in favour of, or indeed against, substance dualism. In fact, as I have already
indicated, property dualism is typically advanced not as a consequence of substance
dualism, but as a supposedly more defensible alternative to it.
Perhaps the most intuitively persuasive case that can be made for property dualism in its
own right is by appeal to the apparently irreducible character of the properties of
phenomenal consciousness, or ‘qualia’, as revealed to us by introspection. It is hard,
perhaps, to see how mental qualities such as the perceived redness of a rose, tasted
bitterness of a lemon, or the felt sharpness of a pain could simply be physical properties
of the brain or nervous system, manifested in complex patterns of neuronal activity. There
seems to be an unbridgeable ‘explanatory gap’ between such sensory qualia and the
neurophysiological properties that appear to be correlated with them, which prevents us
from either simply identifying these properties or even somehow ‘reducing’ the former to
the latter (see further Levine 2001; Chalmers 1996: ch. 4). However, physicalists may
urge that this is just because our access to our own mental properties is radically
different from our access to the physical properties of other things, being via
introspection rather than via our sensory organs, so that these properties appear to be
different to us, when in some cases they could very easily be the same (see Papineau
2002). After all, it may be said, if sensory qualia were neurophysiological properties of the
brain, one would hardly expect the appearance of such properties to be the same for the
Page 12 of 21
Dualism
subject possessing them as for another person looking at the brain in question. Even if A's
pain, for example, just is a neurophysiological property of A's brain, the appearance of A's
pain to another person, B, who is looking at the relevant part of A's brain, will be
determined by the character of B's own visual qualia, which will presumably be quite
different neurophysiological properties of B's brain.
However, the idea that we may distinguish in such a way between ‘appearance’ and
‘reality’, where mental properties such as pain are concerned, is open to question, for the
following reason (see Kripke 1980: 144–55). Suppose, for the sake of argument, that the
mental property of pain just is a certain neurophysiological property, such as the firing of
C‐fibres, by analogy with the empirically confirmed fact that the property of heat in a gas
just is the kinetic activity of gas molecules—where the ‘is’ is understood as being the ‘is’
of identity. Saul Kripke has famously argued that any such identity, if it obtains, must
obtain necessarily, not merely contingently. So, for example, given that heat in a gas just
is the kinetic activity of gas molecules, it could not have been anything else. It must be
conceded, however, that the identity seems contingent—and this is something that needs
to be explained. Here we may appeal to the fact that we can, of course, imagine that heat
in a gas might have been something else, explaining this in turn by the fact that there
could have been a different kind of physical activity which had the same appearance as
heat; that is, which gave rise to the same phenomenal experience as the kinetic activity of
gas molecules gives rise to in us. This, then, is why the identity, despite being necessary,
seems contingent. But the putative identity of pain with the firing of C‐fibres—or any
other (p. 78) neurophysiological property—seems equally contingent, a fact which ought
to be similarly explicable, one might think, if the identity really does obtain. This,
however, is precisely where the intended analogy with the identity between heat and the
kinetic activity of gas molecules appears to break down. For if pain just is the firing of C‐
fibres, say, then the analogy requires us to explain the seeming contingency of this
identity by saying that there could be a kind of physical activity that has the appearance
of pain without being pain, just as there could be a kind of physical activity which has the
appearance of heat without being heat. And yet it seems evident that there is no such
possibility in the case of pain, because whatever seems like pain is pain, even if not
everything that seems like heat is heat.
But I think that this line of argument is vulnerable to the following riposte. If pain just is
the firing of C‐fibres, then, indeed, the seeming contingency of this identity needs to be
explained, and it can't be explained in the foregoing fashion, because it is very plausibly
true that whatever seems like pain is pain. Notice, however, that the reverse is not at all
obviously true—that whatever is pain seems like pain. I have already indicated why this
may not be the case. When B looks at the firing of C‐fibres in A's brain, how that
neurophysiological property seems to B is determined by the character of B's own visual
qualia, which are very different from the pain qualia that B would experience if his own C‐
fibres were firing. It is, very arguably, this difference between B's visual qualia and his
pain qualia that makes it seem to B that the firing of C‐fibres is different from pain and
hence that their identity, if it obtains, is merely a contingent one. In this case pain in A
does not seem like pain to B, but this is quite consistent with the claim that whatever
Page 13 of 21
Dualism
seems to B like pain in B is pain in B. For this reason, I am not convinced that qualia‐
based arguments for property dualism are nearly so compelling as their advocates
suppose. This is not to say that we should simply accede at once to the physicalists’ claim
that phenomenal properties are indeed identical with, or reducible to, neurophysiological
properties. The failure of an argument against a certain claim does not constitute any sort
of argument in favour of that claim. Indeed, the burden of proof clearly lies with the
physicalist to provide a positive argument in favour of the present claim, since it must
still be conceded that phenomenal properties do not seem to be identical with
neurophysiological properties, even if we have seen that this fact is not incompatible with
their identity. A positive argument is always required in support of a claim that reality is,
in some respect, not as it seems to be.
This is the point at which causal considerations come to the fore in the debate between
dualism and its opponents. Perhaps the most powerful argument against interactive
dualism is the so‐called causal‐closure argument. (For further background see Lowe
2000a: ch. 2.) By interactive dualism I mean the doctrine that mental events or states are
not only distinct from physical events or states, but are also included amongst the causes
and effects of physical events or states. Of course, the causal‐closure argument can have
no force against either epiphenomenalist or non‐interactive‐parallelist dualism, but since
even the first and more credible of these positions has relatively few modern advocates
(see e.g. Robinson 2004) I shall not consider them here. In any case, even those who do
support them would presumably (p. 79) concede that they would prefer to endorse
interactive dualism if they thought that it could meet the physicalists’ objections, so let us
concentrate on seeing how those objections can indeed be met, focusing on the causal‐
closure argument.
The key premise of the causal‐closure argument against interactive dualism is the
principle of the causal closure of the physical domain. This principle has received a
number of different formulations—some of which are really too weak for the physicalist's
purposes (see further Lowe 2000b)—but the relatively strong version of the principle that
I shall chiefly consider here is this:
(15) No chain of event causation can lead backwards from a purely physical effect to
antecedent causes some of which are non‐physical in character. (See Kim 1993.)
It may be objected on behalf of interactive dualism that (15) is simply question‐begging, because
it rules out by fiat the possibility of there being non‐physical mental causes of some physical
effects. However, as we shall see, (15) does not in fact rule out this possibility. Dialectically, it is
in the dualist's interests to concede to the physicalist a version of the causal‐closure principle
that is as strong as possible—provided that it still falls short of entailing the falsehood of
interactive dualism—because if the causal‐closure argument in its strongest non‐question‐
begging form can be convincingly defeated, the physicalist will be left with no effective reply.
Weaker versions of the causal‐closure principle can, of course, be countered by interactive
dualists relatively easily, but tend to be countered by them in implausible ways which leave the
physicalist with a telling response.
Page 14 of 21
Dualism
To illustrate the latter point, consider the following widely advocated version of the
causal‐closure principle:
(16) Every physical effect of a mental cause has a sufficient physical cause.
An interactive dualist may accommodate (16) by, for example, espousing the doctrine of
interactive parallelism, which maintains that there is a one‐to‐one correlation between the
mental and physical causes of any physical event that has a mental cause, such that both the
mental and the physical causes of any such event are sufficient causes of it. (For an exposition
and defence see Meixner 2004: ch. 8.) (By a sufficient cause of a given event I mean an event or
conjunction of events that causally necessitates the event in question.) The physicalist may
object that this doctrine has the highly implausible implication that every physical effect of a
mental cause is causally overdetermined by that mental cause and the physical cause that is,
supposedly, one‐to‐one correlated with it. To this the interactive parallelist may reply that such
causal overdetermination is no mere accident but, rather, the upshot of psychophysical laws, so
that the fact that it occurs is a matter of nomic or natural necessity. However, it may nonetheless
appear surprising to the impartial bystander that psychophysical laws of this character should
be thought to govern the causal interactions of mind and body, when so many other possibilities
are compatible with interactive dualism. The non‐interactive parallelist has, it seems, much
better reason to suppose that there is a one‐to‐one correlation between the apparent mental
causes of physical events and their actual physical causes, because (traditionally, at least) he
sees this as (p. 80) being the upshot of a divinely instituted pre‐established harmony between
the mental and physical domains. Equally, the physicalist has a perfectly good reason to suppose
that there is a one‐to‐one correlation between the mental and physical causes of physical events,
because he identifies those causes, and identity is a one‐to‐one relation par excellence. But the
interactive parallelist, it seems, must simply regard it as a brute fact that psychophysical laws
sustain such a one‐to‐one correlation—a fact that is all the more remarkable because so many
other arrangements are consistent with the truth of interactive dualism. Neutral parties to the
debate could be forgiven for suspecting that the interactive parallelist postulates the one‐to‐one
correlation of mental and physical causes simply in mimicry of the physicalist's position, with a
view to denying the physicalist recourse to any empirical evidence of a causal character that
could discriminate between the two positions. For wherever the physicalist claims to find
evidence of one and the same cause of a certain physical effect—a single cause that is both
mental and physical—the interactive parallelist will be able to reply that we in fact have two
distinct but correlated causes, one of them mental and the other physical.
Let us now consider a version of the causal‐closure argument against interactive dualism
that appeals to the very strong formulation of the causal‐closure principle embodied in
premise (15)—that no chain of event causation can lead backwards from a purely physical
effect to antecedent causes some of which are non‐physical in character. Two additional
premises are needed. First,
Page 15 of 21
Dualism
(19) All of the mental causes of purely physical effects are themselves physical in
character.
which contradicts the defining thesis of interactive dualism. My defence of interactive dualism
will rest upon a challenge to premise (18). Moreover, it will endorse a version of interactive
dualism which combines it with the sort of non‐Cartesian substance dualism defended earlier.
What seems plausible is that if we were to trace the purely bodily causes of any bodily
event, such as the movement of my arm on a given occasion, backwards indefinitely far,
we would find that those causes ramify, like the branches of a tree, into a complex maze
of antecedent events in my brain and nervous system—these neural events being widely
distributed across large areas of those parts of my body and having no single focus
anywhere, the causal chains to which they belong possessing, moreover, no distinct
beginnings (see further Lowe 1996: ch. 3 see Fig. 3.1 below). And yet, my mental act of
decision or choice to move my arm seems, from an introspective point of view, to be a
singular and unitary occurrence which somehow initiated my action of raising my arm.
How, if at all, can we reconcile these two apparent (p. 81) facts? It seems impossible to
identify my act of choice with any neural event, or even with any combination of neural
events, because it and they seem to have such different causal features or profiles. The
act of choice seems to be unitary and to have, all by itself, an ‘initiating’ role, whereas the
neural events seem to be thoroughly disunified and merely to contribute in different ways
to a host of different ongoing causal chains, many of which lead independently of one
another to the eventual arm movement. (For a response to doubts about the ‘initiating’
role of acts of choice, founded on the well‐known experiments of Benjamin Libet 1985,
see Lowe (2008: ch. 4).)
Page 16 of 21
Dualism
can say is that if this or that neural event, or set of neural events, had not occurred, the
arm movement might have proceeded in a somewhat different manner—more jerkily,
perhaps, or more quickly—not that my arm would have remained at rest, or would instead
have moved in a quite different kind of way. (In other words, in the ‘closest’ possible
worlds in which the relevant neural events do not occur, it's not the case that an arm
movement of the given kind does not occur at
It may be asked: But what about the causes of my act of decision or choice? Are these
bodily, or mental, or both? My own opinion is that an act of decision or choice is free, in
the ‘libertarian’ sense—that is to say, it is uncaused (see further Lowe 2003a). This is not
to say that decisions are simply inexplicable, only that they demand explanations of a non‐
causal sort. Decisions are explicable in terms of reasons, not causes. That is to say, if we
want to know why an agent decided to act as he did, we need to enquire into the reasons
in the light of which he chose so to act. Since decisions are, according to NCSD,
Page 17 of 21
Dualism
attributable to the self and not to the body or any part of it, there is no implication here
that any bodily event is uncaused.
It may now be wondered: How is it really possible for mental acts of decision to explain
anything in the physical domain, if that domain is causally closed, in the sense defined
earlier? Let us recall how, precisely, we defined the causal closure of the physical domain.
According to the principle of physical causal closure, I said, no chain of event causation
can lead backwards from a purely physical effect to antecedent causes some of which are
non‐physical in character. This was premise (15) above. But intentional causation on the
NCSD model, as I have described it, does not violate the principle of physical causal
closure, since it does not postulate that mental acts of decision or choice are events
mediating between bodily events in chains of event causation leading to purely physical
effects: it does not postulate that there are ‘gaps’ in chains of physical causation that are
‘filled’ by mental events (see Lowe 1996: 82). (p. 83) Thus, it is consistent with premise
(15) of the causal‐closure argument and avoids the conclusion of that argument by
repudiating, instead, premise (18).
On the NCSD model, a decision can explain the fact that a bodily movement of a certain
kind occurred on a given occasion, but not the particular movement that occurred. Even
so, it may be protested, if physical causation is deterministic, then there is really no scope
for intentional causation on the NCSD model to explain anything physical, because the
relevant counterfactuals will all be false. It will be false, for instance, to say that if I had
not decided to raise my arm, a rising of my arm would not have occurred; for precisely
the same bodily movement would still have occurred, caused by the same physical events
that actually caused it. Maybe so. But we know that physical causation is not in fact
deterministic, so the objection is an idle one.
The NCSD model of intentional causation may nonetheless seem puzzling to many
philosophers. I suggest that that is because they are still in the grip of an unduly simple
conception of causation—one which admits only of the causation of one event by one or
more antecedent events belonging to one or more chains of causation which stretch back
indefinitely far in time. Since this is the only sort of causation recognized by the physical
sciences, intentional causation on the NCSD model is bound to be invisible from the
perspective of such a science (see Lowe 2003b). To a physicalist, this invisibility will seem
like a reason to dismiss the notion of intentional causation as spurious, because ‘non‐
scientific’. To more broad‐minded philosophers, I hope, it will seem like a reason to
discern no conflict between explanation in the physical sciences and another, more
humanistic, way of explaining our intentional actions.
References
Baker, L. R. (2000), Persons and Bodies: A Constitution View (Cambridge: Cambridge
University Press).
Page 18 of 21
Dualism
Chisholm, R. M. (1976), Person and Object: A Metaphysical Study (London: George Allen
& Unwin).
Crane, T., and Mellor, D. H. (1990), ‘There Is No Question of Physicalism’, Mind, 99: 185–
206.
Foster, J. (1991), The Immaterial Self: A Defence of the Cartesian Dualist Conception of
the Mind (London: Routledge).
Gendler, T. S., and Hawthorne, J. (2002) (eds.), Conceivability and Possibility (Oxford:
Clarendon).
and A. Mele (eds.), Mental Causation (Oxford: Oxford University Press), 189–210.
—— (2001), ‘Lonely Souls: Causality and Substance Dualism’, in K. Corcoran (ed.), Soul,
Body, and Survival: Essays on the Metaphysics of Human Persons (Ithaca, NY: Cornell
University Press), 30–43.
Kripke, S. A. (1980), Naming and Necessity (Cambridge, Mass: Harvard University Press).
Levine, J. (2001), Purple Haze: The Puzzle of Consciousness (Oxford: Oxford University
Press).
Libet, B. (1985), ‘Unconscious Cerebral Initiative and the Role of Conscious Will in
Voluntary Action’, Behavioral and Brain Sciences, 8: 529–66.
Lowe, E. J. (1989), Kinds of Being: A Study of Individuation, Identity and the Logic of
Sortal Terms (Oxford: Blackwell).
Page 19 of 21
Dualism
—— (2001), ‘Identity, Composition, and the Simplicity of the Self’, in K. Corcoran (ed.),
Soul, Body, and Survival: Essays on the Metaphysics of Human Persons (Ithaca, NY:
Cornell University Press), 139–58.
Olson, E. T. (2001), ‘A Compound of Two Substances’, in K. Corcoran (ed.), Soul, Body, and
Survival: Essays on the Metaphysics of Human Persons (Ithaca, NY: Cornell University
Press), 73–88.
Page 20 of 21
Dualism
E. J. Lowe
Page 21 of 21
Epiphenomenalism
THE idea that our mental states can causally affect our bodily behaviour, and thus
indirectly affect our environment, is an integral part of what one could, following Wilfrid
Sellars (1962), call our ‘manifest image of the world’. We are, it seems, agents who act
because of our beliefs, desires, sensations, intentions, perceptions, etc.: a headache
makes us frown, the intention to make a bid at an auction causes us to raise our hand,
and the desire to hear a loved one's voice leads us to make a phone call. That our mental
life is causally effective is hard to deny. How it can be so, however, is far from clear. Four
centuries have passed since Descartes's pioneering discussion of the problem of mental
causation, but we still lack a satisfying account of how our mental life fits into the causal
structure of the world. Our failure to understand the how of mental causation may be a
Page 1 of 12
Epiphenomenalism
reason to reconsider Thomas Huxley's (1874) suggestion that our mental life is a mere
epiphenomenon—a by‐product of the neurophysiological causes of behaviour that does
not itself contribute anything to ‘the go’ of the world.
However, epiphenomenalism is not only at odds with our intuitive conception of ourselves
as autonomous agents. It also faces a whole series of theoretical difficulties. Among other
things, it appears to undermine freedom of the will, the possibility of an evolutionary
account of the mind, our conviction that others enjoy a mental life similar to ours, the
application of epistemic norms like justification, warrant, or reasonableness to processes
of belief formation, the distinction between reasons for an action and the reasons for
which it was performed, and our ability to refer to, have knowledge of, and have
memories about mental states. In this short chapter there is no room for a comprehensive
discussion of all these objections to the view that mental phenomena are epiphenomena
(see Walter 2007a, 2008a). I will only briefly sketch two of the most important of them
below, saying why the epiphenomenalist's standard response to them appears
unsatisfactory. Then I consider an issue that has so far been mostly neglected; namely, the
question of which account of causation (p. 86) would allow for a coherent formulation of
the epiphenomenalist's position. Finally, I will suggest an alternative version of
epiphenomenalism, which, though not free from problems, may be superior to the
traditional version.
4.1 Epiphenomenalism
Epiphenomenalists typically maintain that every mental event has a complete physical, in
particular neurophysiological, cause, while no mental event is ever a complete or even a
partial cause of any other event. It seems as if our behaviour is caused, or causally
influenced, by our beliefs, desires, sensations, etc., but in fact the regular succession of
mental and physical events—our headache being followed by our frowning, our desire to
call a loved one being followed by our making a phone call, etc.—is the result of causal
processes at the underlying neurophysiological level. Epiphenomenalism evolved in the
late eighteenth and the nineteenth century as a consequence of two apparently
discordant beliefs about the world: (1) the growing scientific confidence that our world is
a world of purely physical causes governed by physical laws and driven by physical forces
only, and (2) a decidedly dualistic trust in the distinctness of mind and brain.
Epiphenomenalism managed to reconcile these two conflicting convictions. The view that
modern science allows only for physical causation is compatible with dualism if the
mental has no causal impact upon the physical.
Page 2 of 12
Epiphenomenalism
Thesis (CD) plays a pivotal role in the epiphenomenalist's typical response to standard
objections. (CD) ensures that whenever the epiphenomenalist's opponent appeals to a mental
cause m in order to explain some phenomenon, there is a physical event p, namely, m's physical
cause, which, the epiphenomenalist can claim, actually plays the causal role her opponent
attributes to m (see sections 4.2 and 4.3 for examples of this strategy). Nor, obviously, can the
epiphenomenalist give up (CI) and remain an epiphenomenalist. Finally, something like (IR) is
accepted by most philosophers of mind, and there would be little or no reason to endorse
epiphenomenalism in the first place, if mental events were identical with or reducible to physical
events. Arguably then, (IR), (CD), and (CI) form the core of traditional (p. 87)
epiphenomenalism. Yet we will see in Section 4.4 how they may cause trouble for the
epiphenomenalist. First, however, let us consider two of the most prominent objections against
the view that the mental is a mere epiphenomenon.
Page 3 of 12
Epiphenomenalism
Usually that is all the epiphenomenalist has to say in response to the objection from
evolution (Two laudable exceptions are Robinson (2007) and Corabi 2007.) However, that
response seems wanting for several reasons.
First, suppose a university's admissions office proclaimed that an applicant's religion has
no influence upon her chances of being accepted, but it turned out that all accepted
students have the same religion. That would cry out for an explanation, and the same
holds if it is claimed that our mind is an epiphenomenon but that as a matter of (brute)
fact evolution has seen to it that any (healthy) brain of sufficient complexity is
accompanied by a conscious mind.
Second, the usual analogies show that evolution can propagate a trait that is neutral or
detrimental for an organism's fitness. They do not show, however, that evolution can
propagate a trait that does not make any causal difference whatsoever. (After all, the
polar bear's heavy coat affects its behaviour: it slows it down.) The mind, it seems, would
be the only co‐evolved trait which has no effect at all.
Third, we understand perfectly well why a selection pressure for a warm coat has led to a
heavy coat—the insulatory properties of a polar bear's coat depend upon its thickness,
and the thicker the coat is, the heavier it is. In contrast, we do not have (p. 88) the
slightest idea why a selection pressure for certain neurophysiological structures should
necessarily have led to a conscious mind (due, of course, in part to what Joe Levine calls
the ‘explanatory gap’; see Ch. 16 below).
This does not show that the epiphenomenalist's standard response is irreparably flawed,
but it shows that the typical analogies alone are an insufficient response to the objection
from evolution.
Although this is not the place to go into the details, let me suggest the outline of a
different rejoinder. Our mind is no doubt the result of a process of evolution (we have one,
but we have ancestors that did not have one). That does not entail, however, that our
mind is the result of a process of natural selection. Natural selection is one of the
mechanisms behind evolution (the change of allele frequencies within populations), but
not the only one. That point was made by Gould and Lewontin (1979), and has recently
been backed up by researchers in evolutionary development (so‐called ‘evo‐devo’), who
argue that factors in the embryonic development of an organism can affect evolution
without there being any specific selection pressures, for instance because there is not
enough variation in the gene pool for natural selection to be effective (e.g. Carroll 2005).
This of course falls short of establishing that our mind is not a result of natural selection,
but it shows that a crucial premise of the objection from evolution—that natural selection
is the only mechanism that could explain the evolution of our mind—needs more support
than it is usually given.
Page 4 of 12
Epiphenomenalism
First, the epiphenomenalist's version is phenomenologically much less adequate than her
opponent's story: it is obvious that the attribution of mental states to others depends
upon their behaviour, but it is not at all obvious that it depends upon the assumption of a
similar neurophysiological make‐up (see e.g. Dennett 1987: 17–25).
(p. 89)
Third, since the epiphenomenalist's story contains one more inference than her
opponent's, namely, the one from a similar neurophysiological make‐up to a similar
mental life, and since this inference is not at all obvious, her alternative line of
justification is epistemically less secure. Everyone must hold that the inference from
behaviour to the underlying causes is justified. But the epiphenomenalist must also hold
Page 5 of 12
Epiphenomenalism
that the inference to similar mental states as further effects is justified, and that seems
risky, especially since we have little idea of why neurophysiological states should be
accompanied by mental states at all. Hence, the epiphenomenalist's appeal to apparently
uncontroversial cases where we infer one effect from another via a joint cause does not
show that her alternative solution to the other‐minds problem is satisfactory.
This and the preceding section have shown how important (CD) is for the
epiphenomenalist. (CD) ensures that for any mental event there is a physical event of
which the epiphenomenalist can claim that it actually plays the causal role usually
attributed to the mental event in question (enhancing our fitness, causing our behaviour,
etc.). And, as stated in Section 4.1, (CI) and (IR) are essential for the epiphenomenalist.
Yet, Section 4.4 will show that the combination of (CD), (CI), and (IR) can lead to some
unexpected problems concerning the epiphenomenalist's account of causation.
(PC) Every physical event which has a cause has a (set of) physical event(s) as its
complete cause. [physical closure]2
(p. 90)If there is at least one caused physical event, (PC) entails physical‐to‐physical causation.
Moreover, if there is at least one caused mental event, (CD) entails physical‐to‐mental causation.
Since the epiphenomenalist is making these causal claims, she thus ought to have some theory
of causation, even though she maintains that mental‐to‐physical and mental‐to‐mental causation
are impossible. What has gone largely unnoticed so far in discussions of epiphenomenalism (but
see Lachs 1963) is that it is far from clear what conception of causation could account for the
causal transactions posited in (PC) and (CD), while at the same time also respecting (CI). Given
the limited space, it is impossible to consider each and every account of causation in detail, but
the following should at least illustrate the nature and force of the problem for the
epiphenomenalist.
Obviously, adopting a simple regularity account of causation which dispenses with any
kind of causal necessity and analyses causes as sufficient conditions of their effects will
not do for the epiphenomenalist (a point already noted by Broad 1925). Roughly, the idea
would be that if c and e are distinct events (and c is earlier than e), c causes e iff there are
event types C and E such that c is of type C, e is of type E, and events of type C are
regularly followed by events of type E. Yet according to the epiphenomenalist, a mental
event m will regularly be followed by a behavioural event b; since m and b have a common
neurophysiological cause p, p will be regularly followed by both m and b, so that m will be
regularly followed by b, too. Hence, if the epiphenomenalist invokes a regularity
conception of causation to account for physical‐to‐physical and physical‐to‐mental
causation, she is thereby violating (CI).
Page 6 of 12
Epiphenomenalism
The same holds for Mackie's INUS account (1974), according to which causes are insufficient
but non‐redundant parts of an unnecessary but sufficient condition. My belief that I have
left my laptop behind at the check‐in counter may not by itself be sufficient for my
returning to the check‐in counter (I may be convinced that catching my flight is more
important), but only in combination with other conditions, where in addition I would not
have returned to the check‐in counter under these conditions had I not believed that I
had left my laptop there. My belief would then be an INUS condition of my return to the
check‐in counter. Hence, if the epiphenomenalist invokes an INUS conception of
causation to account for physical‐to‐physical and physical‐to‐mental causation, she is once
again violating (CI).
On the other hand, however, if it is unintelligible how the mental could transfer energy or
momentum to the physical, then for the very same reason it is also unintelligible how the
physical could transfer energy or momentum to the mental, so that the kind of physical‐
to‐mental causation required by (CD) would be impossible, too. Hence, if the
epiphenomenalist invokes a conserved‐quantity conception of causation to account for the
impossibility of mental‐to‐physical and mental‐to‐mental causation, she is violating (CD).
Apart from that, (CI) and (CD) together entail the existence of ‘causal dead ends’—
unidirectional causal chains leading from the physical to the mental, but not back again
to the physical. Hence, if causation consisted in the transfer of some physical magnitude,
Page 7 of 12
Epiphenomenalism
then (CI) and (CD) would apparently lead to a ‘drainage’ of energy, momentum, etc. that
would violate physical conservation laws.
Although these brief remarks are far from exhaustive or conclusive, they show that it is
unclear how a coherent overall account of the various causal claims to which the
epiphenomenalist is committed would look. This objection goes deeper than the usual
ones which purport to demonstrate that epiphenomenalism is incompatible with some
apparently undeniable feature of our manifest image of the world. If correct, it shows that
epiphenomenalism cannot even be coherently formulated—not, at least, by adopting one
of the currently leading accounts of causation. If epiphenomenalism is to have any chance
of being credible, then—regardless of how it fares (p. 92) with regard to other objections
—much more attention has to be paid than hitherto to explaining what exactly according
to the epiphenomenalist causation is supposed to be.
(NCD) Every mental event is realized by a (set of) physical event(s). [non‐causal
dependence]5
Page 8 of 12
Epiphenomenalism
The position spanned by (IR), (CI), (NCD), and (PC) avoids any problematic ‘causal dead ends’.
Although there still is a unidirectional dependency of the mental upon the physical, that
dependency is non‐causal, so that (NCD) and (CI) do not violate physical conservation laws if a
conserved‐quantity conception of causation is adopted. Moreover, adopting such a conception
still accounts for the truth of (CI), but once (CD) is replaced by (NCD) it no longer conflicts with
alleged cases of physical‐to‐mental causation, for there aren't any. Finally, if a conserved‐
quantity conception of causation works at all, it arguably works best for the kind of physical‐to‐
physical causation posited by (PC). Hence, if the epiphenomenalist could defend a conserved‐
quantity account of causation (which is admittedly no easy task, but not unprecedented either;
see e.g. Kistler 1998, Hall 2004), replacing (CD) by (NCD) would make (p. 93) her position
consistent. Moreover, it seems that there would not be any disadvantage. The standard strategy
for dealing with objections remains unaffected, as seen above (although it is still not altogether
convincing; see, for example, Sections 4.2 and 4.3), and, since realization is standardly taken to
be compatible with a broadly physicalistic attitude, a general dismissal of dualism does not do as
an objection against this alternative version of epiphenomenalism.
However, why would one adopt such a position? If one is prepared to accept that the
mental is realized by the physical, and thus in some broad (non‐reductive) sense part of
the physical, then why would one still hold that it is an epiphenomenon? Generally
speaking, the motivation for adopting the alternative version of epiphenomenalism is the
same as the motivation for adopting traditional epiphenomenalism. On the one hand,
there is the intuition that the mental and the physical are in some robust sense distinct;
on the alternative version of epiphenomenalism that intuition is accounted for by (NCD)
and (IR).6 On the other hand, there is the intuition that as far as causation is concerned,
the primacy is with the physical; on the alternative version of epiphenomenalism that
intuition is accounted for by the conserved‐quantity account of causation, which arguably
restricts causation to the (fundamental) physical level. As a consequence, it seems that
the mental cannot contribute anything to ‘the go’ of the universe; that is, mental events
cause neither physical nor mental events. One will be led to such a position if one thinks
that traditional epiphenomenalism is untenable (for instance for the reason stated in
Section 4.4) but also finds the various accounts of mental causation that have been
offered by so‐called non‐reductive physicalists unconvincing. It is a position for all those
who (like me) are reluctant to endorse traditional epiphenomenalism, and instead replace
its unabashedly dualistic world‐view with a more naturalistic view according to which the
mental is realized by the physical, but at the same time find mental causation problematic
because they just cannot see how an adequate account of mental causation would go.7 , 8
References
Benecke, E. (1901), ‘On the Aspect Theory of the Relation of Mind to Body’, Proceedings
of the Aristotelian Society, 1: 18–44.
Broad, C. D. (1925), The Mind and Its Place in Nature (London: Routledge).
Carroll, S. (2005), Endless Forms Most Beautiful: The New Science of Evo Devo and the
Making of the Animal Kingdom (New York: Norton).
Page 9 of 12
Epiphenomenalism
Gould, S., and Lewontin, R. (1979), ‘The Spandrels of San Marco and the Panglossian
Paradigm: A Critique of the Adaptationist Programme’, Proceedings of the Royal Society
London, B205: 581–98.
Hall, N. (2004), ‘Two Concepts of Causation’, in J. Collins, N. Hall, and L. A. Paul (eds.),
Causation and Counterfactuals (Cambridge, Mass.: MIT Press), 225–76.
Huxley, T. (1874), ‘On the Hypothesis that Animals are Automata, and Its History’,
Fortnightly Review, 22: 555–80.
Lachs, J. (1963), ‘Epiphenomenalism and the Notion of Cause’, Journal of Philosophy, 60:
141–5.
Popper, K. (1978), ‘Natural Selection and the Emergence of Mind’, Dialectica, 32: 339–55.
Sellars, W. (1962), ‘Philosophy and the Scientific Image of Man’, in R. Colodny (ed.),
Frontiers of Science and Philosophy (Pittsburgh, Pa: University of Pittsburgh Press), 33–
78.
Walter, S. (2005), ‘Program Explanations and Mental Causation’, Acta Analytica, 20: 32–
47.
Page 10 of 12
Epiphenomenalism
—— (2008a), ‘Ist der Epiphänomenalismus absurd? Ein neuer Blick auf eine tot geglaubte
Position’, Zeitschrift für philosophische Forschung, 62: 415–32.
Notes:
(1) (CD) allows for the possibility of uncaused mental events. Epiphenomenalism is thus
compatible with mental events that are not causally grounded in physical events, as long
as they have no effects.
(2) Given (CI), no mental event is a complete or even a partial cause of a physical event;
hence, if a physical event has a cause at all, it must have a physical cause.
(3) ‘Roughly’ because since counterfactuals are not in general transitive, causation
should not be defined as counterfactual dependence per se, but rather as a chain of
counterfactual dependencies (see Lewis 1973).
(4) There is a lot more to say on this issue though. Lewis has argued that if m and b are
effects of a common cause p, ‘¬m ¬b’ is false (see Lewis 1979 and the postscripts to his
1973 and 1979 in Lewis 1986). According to Lewis, p is not sufficient for m in the relevant
possible worlds. When assessing similarity across worlds we must weigh up the degree to
which there are violations in the fundamental laws against the degree of departure from
particular matters of fact. Lewis argues that a possible world in which m does not occur
but (due to a minor miracle) p occurs and causes b is closer to the actual world than any
possible world in which m and p are both absent and we have to change the entire causal
history of the world in question. These are contested issues, however. Suffice it to say
that if the epiphenomenalist adopts a counterfactual account of causation, she at least
owes us a detailed explanation for why counterfactuals like ‘¬m ¬b’ are supposed to
be false.
(5) Of course, being a realizer here cannot mean being a filler of a causal role, as a
functionalist would usually hold, for the mental has no causal role to play according to the
epiphenomenalist. An epiphenomenalist account of realization would have to comprise, at
a minimum, the claim that realization is a relation between properties such that a
property G (a set of properties G 1,…,G n) realizes a property F only if the instantiation of
G in an object o (of G 1,…,G n in objects o 1,…,o n) is sufficient for the instantiation of F in
an object o′, but not vice versa. (See, for instance, Shoemaker's subset model of
Page 11 of 12
Epiphenomenalism
(7) See Walter (2005) for an argument against Jackson and Pettit's attempt to ground
mental causation in the mental's ability to figure in so‐called program explanations,
Walter (2007b) for an argument against Lynne Baker's attempt to ground mental
causation in the mental's special explanatory status, and Walter (2007c) for an argument
against Stephen Yablo's appeal to the determinable/determinate distinction in the attempt
to prevent a pre‐emption of the mental by the physical.
(8) I'm indebted to Brian McLaughlin for helpful comments on earlier drafts of this paper
and to Lena Kästner and Nicole Troxler.
Sven Walter
Page 12 of 12
Anomalous Monism
Anomalous monism is a view about the relationship between the mind and the body,
which attempts to strike a delicate balance between the thesis of materialism, on the one
hand, and the irreducibility of the mental, on the other. Its current formulation is found in
Donald Davidson's landmark paper, ‘Mental Events’, and concerns only intentional states
— contentful mental states, such as the belief that p, the desire that q, and other
propositional attitudes. Anomalous monism consists of two theses, one concerning
monism, the other concerning anomalism. The ‘monism’ part of anomalous monism is the
claim that all events, including the mental ones, essentially fall under one class; namely,
the class of physical events.
Keywords: anomalous monism, mind–body relation, thesis of materialism, intentional states, propositional
attitudes, physical events
ANOMALOUS monism is a view about the relationship between the mind and the body, which
attempts to strike a delicate balance between the thesis of materialism, on the one hand,
and the irreducibility of the mental, on the other. Its current formulation is found in
Donald Davidson's landmark paper, ‘Mental Events’ (1970), and concerns only intentional
states—contentful mental states, such as the belief that p, the desire that q, and other
propositional attitudes.
Anomalous monism consists of two theses, one concerning monism, the other concerning
anomalism. The ‘monism’ part of anomalous monism is the claim that all events, including
the mental ones, essentially fall under one class; namely, the class of physical events. To
say that an event is physical is to say that there is a physical description that correctly
picks out the event; alternatively, one can say that an event is physical if it exemplifies a
physical property (mutatis mutandis) for mental events). While Davidson prefers to
formulate his views in terms of the application of predicates rather than the instantiation
Page 1 of 17
Anomalous Monism
Anomalous monism, then, asserts that every mental event aside from having its mental
property also has a physical property. For this reason the monistic part of anomalous
monism is also known as token physicalism: the idea that every mental event is
numerically identical with a physical event, though not the converse, as there are events,
such as hurricanes, that are physical but not mental. The following example illustrates
the idea of token identity for events. It is uncontroversial that physical events can
exemplify more than one property: the sinking of the Titanic, say, can be the same event
as the greatest maritime disaster. Here being a sinking (p. 96) and being a maritime
disaster are the distinct properties or types that belong to one and the same event.
Likewise, believing that it is raining (a mental property) and being a certain neural
occurrence (a physical property) can belong to one and the same event. The idea of token
physicalism, then, is the idea that all mental events also exemplify a physical property,
which is to say that mental events, qua events, are reductively identifiable with physical
events.
The thesis that expresses the ‘anomalous’ part of anomalous monism is that there are no
strict lawful regularities that govern how someone will think or act, exceptionless
regularities that constitute a closed and comprehensive theory, which is essentially a
fundamental theory of all natural phenomena that physics is thought to provide. It
encompasses psychological anomalism—the denial of strict regularities that would
connect only mental events with other mental events—and psychophysical anomalism—
the denial of strict regularities that would connect mental events with physical events.
The thesis of psychological anomalism is directed against the possibility of regimenting
folk psychology into a closed and comprehensive theory in its own right, and the thesis of
psychophysical anomalism is directed against the requisite inter‐theoretic bridge
principles that could facilitate a reduction of folk psychology to physics.
Page 2 of 17
Anomalous Monism
Davidson's argument for token physicalism relies upon the claim that a mental event is
causally efficacious in virtue of its token identity with an efficacious physical event. The
argument consists of three principles. The first is that mental causation really happens;
one's desire to stay dry in the rain, for instance, can cause one to use an umbrella. The
second is that strict laws are necessary for causation. And the (p. 97) third is that there
are no strict psychological laws or psychophysical laws. Here are Davidson's formulations
of the principles (1970: 208):
Principle of Causal Interaction: [A]t least some mental events interact causally
with physical events.
Anomalism of the Mental: [T]here are no strict deterministic laws on the basis of
which mental events can be predicted and explained.
The argument for token physicalism proceeds thus. According to the anomalism of the mental,
no mental event m can cause a physical event p (or any other event, for that matter) on account
of its governance under a psychological law or psychophysical law, as no such laws are properly
strict (if there are any such laws at all). By the principle of the nomological character of
causality, all causal relations must be covered by a strict law. Such strict laws are found only in
the domain of fundamental physics, meaning that only physical events—events exemplifying
physical properties—can enter into causal transactions. Thus, if m causes p, as the principle of
causal interaction affirms, it is because m has a physical property, from which it follows that m is
a physical event.
The main objections to the argument are these. The most prominent is that it ends up
precluding the possibility of mental causation, not grounding its possibility (this is
discussed in Section 5.5). A different criticism is that mental events cannot be
individuated into discrete tokens, on account of their holistic interrelations with other
mental events, so that it is impossible to identify mental events with discretely
individuated physical events (Lurie 1978; Antony 1989; Child 1993). A third criticism is
that there is no unique physical token with which a mental token can be identified at a
time, but rather an array of different possible candidates, undermining the idea that there
is a unique token identity that can be established (Hornsby 1980; Horgan and Tye 1985).
Page 3 of 17
Anomalous Monism
This is to say that a theory of this kind describes a closed and comprehensive system—
closed in that there are no occurrences whose causal origin goes beyond the laws of the
theory, and comprehensive in that its laws fully suffice to encompass the course of all
natural occurrences. Let us say that T is a theory that consists of a body of laws. An event
is a T‐event just in case a law of T governs its occurrence. T, then, is closed if and only if a
T‐event is caused by only other T‐events; that is, if and only if there are no foreign non‐T‐
events causing a T‐event. T is comprehensive if and only if the laws of T fully suffice to
govern the occurrence of all T‐events; that is, if and only if there are no T‐events whose
occurrence escapes the jurisdiction of the laws of T. Although Davidson adds that strict
laws covering causal relations must be deterministic, it could turn out that the laws of T
are not deterministic. But being probabilistic does not mean having exceptions; a
probabilistic law can still be strict in that it has no need of being qualified by ceteris
paribus conditions. (In fact, our best current physics points in this direction.)
can be rendered strict and exceptionless if we stick only to its proprietary home
vocabulary. In such a case, according to Davidson, a non‐strict law, at best, provides
evidence that some strict law or other is at work in securing the causal relation between
the events.
(p. 99)
To appreciate the thesis of psychological anomalism, let us consider some well‐worn folk‐
psychological generalizations:
A. If an agent dearly wants X, but discovers that not‐X, then the agent will be
disappointed that not‐X.
B. If an agent believes that Y and that Y entails Z, then the agent believes that Z.
C. If an agent fears X, then the agent wants it to be the case that not‐X.
These familiar generalizations go a long way in supporting many of our explanations and
predictions about our thoughts and actions, but we could never say of any of them that they are
strict. Take, for instance, (B); there are all sorts of defeaters that may obstruct an agent from
making the inference—distraction, inattention, fatigue, and so on. This is true for all
generalizations of folk psychology. While good at serving as rules of thumb, none of them is
immune to contravention.
Is there a way of refining the generalizations of folk psychology so that they may be
turned into strict laws? For there to be strict psychological laws, psychology needs to be
either closed and comprehensive, or reducible to, or transmutable into, such a theory. As
Davidson argues, this is not possible. It isn't closed, because non‐physical events affect
the mental (consider distal physical causes of perceptions), nor is it comprehensive, as
mental events do not constitute the full scope of natural phenomena (such as short
circuits and hurricanes). And it isn't reducible to fundamental physics, the theory
designated as closed and comprehensive, because there are no bridge principles
connecting mental predicates (properties) with physical predicates (properties) to effect
the reduction—principles that connect terms or properties of the reduced theory with
those of the reducing theory so that the reducing theory can incorporate the phenomena
explained by the reduced theory. (This is the thesis of psychophysical anomalism, covered
more fully in the next section.)
Page 5 of 17
Anomalous Monism
D. An agent has a mental property M if and only if the agent has a physical property
P.
The psychophysical‐anomalism thesis is that there are no reductive bridge principles conforming
to schema (D), but it is also directed at the weaker one‐way conditionals along the lines of (E):
E. If an agent has a physical property P, then the agent has a mental property M.
Since (D) entails (E), arguments directed against (E) will automatically count as arguments
directed against (D). For the rest of this discussion, we will only concern ourselves with (E).
Davidson's claim is not merely that psychophysical correlations have not yet been discovered;
his is an a priori pronouncement against their possibility.
Although the argument for psychophysical anomalism is still thought to be sketchy in
places, there are two prominent reconstructions of it, each of which relies upon the
following two principles:
Page 6 of 17
Anomalous Monism
The holism assumption is that a propositional attitude derives the content that it has by playing
its unique inferential role(s) within the network of the agent's total propositional attitudes. And
the rationality assumption is that we must interpret the content of a belief, say, by placing it
within a larger pattern of beliefs (desires and actions) whose contents rationally cohere with the
content of the belief we ascribe. None of this is to deny that mentally endowed agents falter in
their rationality. Instances of local irrationality have to be acknowledged by any adequate theory
of mind. The point is that an agent's beliefs, desires, and other propositional attitudes cannot be
systematically and globally irrational, for then the attributions no longer count as mental
attributions. Let us now turn to the reconstructions.
Page 7 of 17
Anomalous Monism
5.4.1 Reconstruction 1
This reconstruction appeals to the idea that there are different rules—what Davidson
calls ‘constitutive principles’—essential to the attribution of intentional states and
physical states (Davidson 1970, 1974). The constitutive principles that govern the
attribution of physical states are distinctly non‐rational, which is to say that physical
properties are not attributed in a way that makes them rationally coherent with other
physical attributions. This is obvious: physical things, such as travelling (p. 101) along an
elliptical orbit, weighing 50 lb, or having negative charge, are not, as a categorical
matter, the kinds of things that have rational bearing. Things are different when it comes
to the attribution of propositional attitudes. The constitutive principles governing the
attribution of intentional states are rationality‐preserving. These ‘constitutive principles
of rationality’ enjoin us to attribute attitudes that, for the most part, rationally cohere
with the ones ascribed. So while physical things and their relations to other physical
things have no bearing, our thoughts about physical (and other) things do have rational
bearing.
The fact of this disparity constitutes one of the premises of Kim's reconstruction. The
other crucial premise is a novel spin on the nature of inter‐theoretic reductive bridge
laws. Kim suggests that we think of bridge laws as having the capacity to ‘transmit’
constitutive principles. More specifically, the constitutive principles governing the
attribution of the reducing property get transmitted to the one reduced. Taking ‘(x) (Px →
Mx)’ to represent the form of psychophysical laws, Kim's idea is that the non‐rational
constitutive principles governing P will carry over to M, so that the attribution of M will
then be governed by distinctly non‐rational constitutive principles. Suppose, then, the
usual method of attributing mental properties according to principles of rationality leads
to the attribution of M for an agent on a certain occasion. As the attribution of M would
also be governed by non‐rational constitutive principles of the physical, it is possible that
the psychophysical law covering M enjoins us to make a different attribution. If this is
possible, then psychophysical laws invite predication conflict. Worse, they run the risk of
compromising the nature of mind, which is that it is rational. As Davidson has often
stressed, something is a mind only if it consists of a network of propositional attitudes
that rationally cohere with each other. If all the attitudes were to be attributed by
appealing to psychophysical laws, so that only non‐rational constitutive principles come
to govern the attribution of mental properties, we could end up with attitude attributions
that are rationally incoherent, undermining the very nature of mind itself.
This reconstruction, while provocative, has been resisted. Tiffany (2001), along with
McLaughlin (1985) and Latham (1999), has argued that attributing intentional states in a
way that has no regard for their rational interconnections does not ineluctably lead to
Page 8 of 17
Anomalous Monism
5.4.2 Reconstruction 2
This reconstruction, offered by William Child (1993), is based upon the claim that the
constitutive principles of rationality are ‘uncodifiable’. This means that the principles of
rationality—such as ‘If you believe that p and that p entails q, then you (p. 102) ought to
believe that q’, ‘If you desire p, and believe that doing q is the best way of getting p, then
you ought to do q’, and so on—cannot be agglomerated into a complete system of rules
that uniquely determines what an agent ought to think or do. Take the first principle:
suppose the agent discovers, on independent evidence, that q is not true. The agent then
has two options: either drop your belief that p or drop your belief that p entails q. The
original principle does not tell you which of these two options to take, and therefore
underdetermines which attitude you are to attribute. Such indeterminacies are true of all
applications of principles of rationality. As a consequence, there can be no psychophysical
laws, since mental properties cannot be uniquely fixed by physical properties.
Page 9 of 17
Anomalous Monism
One of the desiderata of anomalous monism was to establish a framework that can
demonstrate how mental causation is possible. Because anomalous monism embraces the
token‐identity thesis, a mental event is causally relevant on account of its token identity
with a causally relevant physical event. If one accepts the token‐identity thesis,
establishing the causal efficacy of a mental event is easy: because physical events
unproblematically enter into causal relations and mental events simply are physical
events, mental events are causally efficacious. But mere token identity is still insufficient
for establishing the possibility of mental causation, because we need further assurance
that the mental event is efficacious in virtue of its mental property. Many argue that the
causal relevance of mental properties is what really matters for an account of mental
causation. One of the most damning objections to Davidson's argument for anomalous
monism is that while it may successfully establish the causal efficacy of mental events, it
precludes the causal relevance of mental properties.
To see the difference between the causal efficacy of an event, on the one hand, and the
causal relevance of a property of an event, on the other, consider an explosion that causes
a fire. Suppose the explosion was also the loudest sound that happened in Smith's cellar.
If they are token‐identical events, then it is also true that the loudest sound in Smith's
cellar caused the fire. However, surely the cause was not causally (p. 103) efficacious in
virtue of it's having the property of being the loudest sound in Smith's cellar. Rather, it
was the fact that the cause was an explosion. According to the critics, mental properties,
under the rubric of Davidson's anomalous monism, are as causally irrelevant to behaviour
as loudness is to the fire. Contrary to what we want in a demonstration of the possibility
of mental causation, anomalous monism entails that mental events cause other events in
virtue of their physical properties, not in virtue of their mental properties. To use a
different locution among certain critics, mental events are efficacious qua physical event,
not qua mental event.
Page 10 of 17
Anomalous Monism
Such a view is what McLaughlin (1989) has called type epiphenomenalism, the idea that
the type or property satisfied by the event is causally irrelevant. The reasoning goes as
follows. Recall Davidson's requirement that causal relations must be backed by strict
laws. According to this requirement, events can enter into a causal relation only if they
are subsumed by a strict law. Only physical laws have what it takes to be strict, so this
means that events can enter into causal relations only in virtue of their physical
properties. It follows that what secures the causal relation between a mental event and a
physical event is the mental event's instantiation of the relevant physical property, not the
instantiation of its intentional property. Therefore, the critics conclude, the fact that the
mental event has the mental property (propositional content) it has, makes no causal
difference to whether or not the behavioural reaction occurs. By the very lights of the
premises of the argument for anomalous monism, the critics argue, it is not the
instantiation of a mental property, but rather the instantiation of a physical property, that
brings about the effect. If this is right, then there is no support for mental
counterfactuals: if the content of a thought were altered or deleted altogether, this would
not affect a mental event's causal status as long as its physical properties remained
intact, because in the end it is the physical properties that fix all causal relations.
Davidson's response challenges the very idea that there are properties in virtue of which
events bring about their effects (1993). Davidson, contra his critics, insists that (p. 104)
the idea that events are causes in virtue of some fact about the events is a feature of
causal explanations, an intensional relation between predicates or descriptions. It is not a
feature of causation, an extensional relation between events simpliciter, and so holds no
matter how the causes and effects are described. Consequently, the problem of
epiphenomenalism is based upon conflating causation with causal explanation.
Davidson's displacement of the ‘in virtue of’ locution from the domain of causation to the
domain of explanation is motivated by his particular views about the different logical
properties of causal statements as opposed to those of explanatory statements (1967).
Page 11 of 17
Anomalous Monism
Explanations of the form, ‘—— because ——’ create an intensional context where the
‘because’ typically relates statements or facts. Causal statements of the form, ‘—— causes
——’, on the other hand, are extensional, where the blanks are filled in with nominalized
descriptions of events. With respect to causal claims, as long as the events are described
by some appropriate event name, such as perfect nominals or definite descriptions,
redescriptions along these lines are intersubstitutible. Davidson's contention is that when
event names are modified by the ‘in virtue of’ idiom, or any of its cognates, the name gets
transformed into a fact description or a statement so that intersubstitution fails, turning
the extensional context into an intensional one. This leaves us with one of two options
with respect to the favoured phrase of the critics, ‘c qua F causes e qua G’. Either we
regard it as an explanatory statement disguised as a causal claim, or we drop the qua
modifiers along with the properties they highlight if we want the phrase to convey its
intended causal meaning. Whatever the choice, Davidson argues, we're left with the more
simple ‘c causes e’.
(a) the fact that c has F is what causes the fact that e has G; or
(b) c causes e in virtue of some fact (namely, that c has F and e has G, and F and G
are suitably related).
Davidson has argued against the reading offered by (a), the idea that facts could count as proper
causal relata, maintaining that facts belong only within an intensional context, and thus could
only serve as explicanda at best. But, as McLaughlin points out, the reading offered by (b) does
not appeal to facts as the relata of the causal relation. The idea behind (b) is that the holding of
a causal relation can be explained by some fact. The important thing to appreciate is that
specifying that a relation holds in virtue of some fact does not mean that the relata of the
relation are themselves facts; so anyone who wished to embrace (b) as the proper interpretation
need not be committed to (a). On McLaughlin's diagnosis of the dispute between Davidson and
his critics, Davidson conflates (a) with (b), and while (a) may be questionable, (b) certainly is
not. As McLaughlin explains, to violate extensionality, the ‘in virtue of’ qualification in (b) would
have to make the cause of an event essentially description‐dependent, but this qualifier doesn't
play that kind of role. This is (p. 105) because what makes it the case that a collection of
particulars are related in a certain way does not involve the need to cite what descriptions are
true of the particulars.
Page 12 of 17
Anomalous Monism
(WS) F* weakly supervenes upon G* if and only if necessarily for any world w and
individuals x and y, if x and y are indiscernible with respect to G* in w, then x and y
are indiscernible with respect to F* in w.
(SS) F* strongly supervenes upon G* if and only if necessarily for any worlds w 1 and
w 2 and any individuals x and y, if x in w 1 and y in w 2 are indiscernible with respect to
G*, then x in w 1 and y in w 2 are indiscernible with respect to F*.
(GS) F* globally supervenes upon G* if and only if worlds w 1 and w 2 indiscernible
with respect to G* are indiscernible with respect to F*.
Page 13 of 17
Anomalous Monism
(p. 106)Briefly, the differences between the three supervenience theses comes down to this
(see McLaughlin 1995). Weak supervenience is a claim about the distribution of properties
pertaining to individuals within a given world. When applied to the family of mental properties
and physical properties, saying that mental properties weakly supervene upon physical
properties amounts to the claim that if two individuals are physically alike, then they are also
mentally alike. The thesis is weak in that the consequence follows with no modal force, as it
allows for physical twins each lodged in different worlds to be mentally different. Strong
supervenience, on the other hand, is a modally stronger claim, because it is about the
distribution of properties pertaining to individuals across worlds. To say that mental properties
strongly supervene upon physical properties is to say that two individuals who are physically
alike must be mentally alike. This rules out cross‐world physical twins having different mental
properties. Finally, global supervenience is about worlds in their entirety, not about individuals
within or across worlds. According to global supervenience, worlds that have the same
distribution of physical properties must have the same distribution of mental properties. While
consistent with the failure of weak supervenience, if unqualified, it still has modal significance,
as the idea is that physically identical worlds must also be mentally the same.
As stated before, Davidson appeals to weak supervenience to supplement his theory of
anomalous monism. Several philosophers, however, have worried that it is too weak to
express an interesting form of dependence of the mental upon the physical (McLaughlin
1985). Consider the physical twins Abe and Babe. While they think the same thoughts
within a given world, placing each in diverse worlds allows for them to entertain different
thoughts. One might think that this is a virtue of weak supervenience, as it is consistent
with the externalist theory of content individuation, where facts about one's physical
environment and social linguistic conventions partly determine the content of an attitude
(Putnam 1975; Burge 1986). While this is certainly a feature of weak supervenience one
may exploit, the worry about the weakness of weak supervenience is that it allows for
physically identical twins lodged in physically identical environments to have differing
thoughts, something more radical than the psychophysical relations entailed by content
externalism.
Strong supervenience, on the other hand, imposes exact similarity in what Abe and Babe
are thinking as each occupies a different world. But here there is the worry that it may be
too strong to be consistent with Davidson's considerations for psychophysical anomalism
(Kim 1993). Strong supervenience implies that for any member of the supervening family
there exists some member or members of the base family that necessitates the
supervening member, where the nature of the necessitation is either nomological or
metaphysical. This follows if we assume that physical properties are closed under
complementation—that not having a certain physical property is itself a physical property
—so that for any given mental state there will exist some physical state that strictly
necessitates the mental state. Such a possibility looks like it compromises Davidson's
psychophysical‐anomalism thesis.
But this issue is complicated. Depending upon what, exactly, we require of a law, it is
possible to read Davidson's psychophysical‐anomalism thesis as being (p. 107) consistent
with strong supervenience (Child 1993). On this reading, psychophysical‐necessitation
relations generated by strong supervenience could be so highly specific that they are
Page 14 of 17
Anomalous Monism
instantiated only once in a world. Laws of nature, on the other hand, have to have greater
generality, which is what makes it possible for them to be confirmable by their positive
instances. If there is no positive instance of a psychophysical‐supervenience relation
other than the one time it occurs, then strong supervenience may not threaten the
psychophysical‐anomalism thesis after all. But then it would carry very little significance.
References
Anscombe, G. E. M. (1957), Intention (Oxford: Blackwell).
—— (1970), ‘Mental Events’, in Davidson, Essays on Action and Events (Oxford: Oxford
University Press, 1980), 207–25.
—— (1993), ‘Thinking Causes’, in J. Heil and A. Mele (eds.), Mental Causation (Oxford:
Oxford University Press), 3–17.
Dray, W. (1957), Laws and Explanation in History (Oxford: Oxford University Press).
Page 15 of 17
Anomalous Monism
Fodor, J. (1989), ‘Making Mind Matter More’, Philosophical Topics, 17: 59–80.
Heal, J. (1995), ‘Simulation, Theory, and Content’, in P. Carruthers and P. K. Smith (eds.),
Theories of Theory of Mind (Cambridge: Cambridge University Press), 75–89.
Horgan, T., and Tye, M. (1985), ‘Against the Token Identity Theory’, in E. Lepore
(p. 108)
and B. McLaughlin (eds.), Actions and Events: Perspectives on the Philosophy of Donald
Davidson (Oxford: Blackwell), 427–43.
Hornsby, J. (1980), ‘Which Physical Events are Mental Events?’, Proceedings of the
Aristotelian Society, 55: 73–92.
Kim, J. (1985), ‘Psychophysical Laws’, in E. Lepore and B. McLaughlin (eds.), Actions and
Events: Perspectives on the Philosophy of Donald Davidson (Oxford: Blackwell), 369–86.
Latham, N. (1999), ‘Davidson and Kim on Psychophysical Laws’, Synthese, 118: 121–44.
Lepore, E., and Loewer, B. (1987), ‘Mind Matters’, Journal of Philosophy, 84: 630–42.
—— (1989), ‘Type Epiphenomenalism, Type Dualism, and the Causal Priority of the
Physical’, Philosophical Perspectives, 3: 109–35.
Putnam, H. (1975), ‘The Meaning of “meaning”’, in Putnam, Mind, Language and Reality:
Philosophical Papers, ii (Cambridge: Cambridge University Press), 215–71.
Tiffany, E. C. (2001), ‘The Rational Character of Belief and the Argument for Mental
Anomalism’, Philosophical Studies, 103: 285–314.
Yalowitz, S. (1997), ‘Rationality and the Argument for Anomalous Monism’, Philosophical
Studies, 87: 235–58.
Page 16 of 17
Anomalous Monism
Julie Yoo
Page 17 of 17
Non‐Reductive Materialism
The expression ‘non-reductive materialism’ refers to a variety of positions whose roots lie
in attempts to solve the mind–body problem. Proponents of non-reductive materialism
hold that the mental is ontologically part of the material world; yet mental properties are
causally efficacious without being reducible to physical properties. After setting out a
minimal schema for non-reductive materialism (NRM) as an ontological position, this
article canvasses some classical arguments in favour of NRM. Then it discusses the major
challenge facing any construal of NRM: the problem of mental causation, pressed by
Jaegwon Kim. Finally, it offers a new solution to the problem of mental causation.
Keywords: non-reductive materialism, material world, mental properties, ontology, mental causation, Jaegwon Kim
Page 1 of 23
Non‐Reductive Materialism
6.1 Introduction
THE expression ‘non‐reductive materialism’ refers to a variety of positions whose roots lie
in attempts to solve the mind–body problem. Proponents of non‐reductive materialism
hold that the mental is ontologically part of the material world; yet mental properties are
causally efficacious without being reducible to physical properties. After setting out a
minimal schema for non‐reductive materialism (NRM) as an ontological position, I'll
canvass some classical arguments in favour of NRM.1 Then I'll discuss the major
challenge facing any construal of NRM: the problem of mental causation, pressed by
Jaegwon Kim. Finally, I'll offer a new solution to the problem of mental causation.
Materialism
For present purposes I consider non‐reductive and reductive materialism to be
ontological positions.2 All forms of materialism, reductive and non‐reductive, disavow
immaterial souls, vital spirits, entelechies, and the like. According to any materialist,
every concrete particular is made up entirely of microphysical items. Non‐reductionists
part company with reductionists, however, with respect to identity or non‐identity of
various kinds or types.
The non‐reductionist distinguishes mental kinds from physical kinds, where the mental
includes sensation and thought, and the physical is roughly the domain of the physical
sciences, including neurophysiology. Even after three and a half centuries we still cast
discussion of the mind/body problem in blatantly Cartesian terms, albeit now with a
materialistic twist: sensation and thought turn out to be part of the material world. It is
just assumed by most parties to the dispute that there is an antecedent unproblematic
pretheoretical distinction between the physical and the mental. Although I believe that
this initial ‘mental/physical’ distinction itself needs scrutiny, I'll follow the mainstream
and take for granted the (unexamined) distinction between the mental and the physical.
Reductionists in the philosophy of mind (e.g. Kim 1998; Lewis 1999; recently Heil 2003)
hold that there are no properties that are distinctively mental properties. There are
mental predicates, of course. There are distinct levels of explanation. We have different
Page 2 of 23
Non‐Reductive Materialism
(1) There are mental properties that are distinct from any physical properties.
However, (1) is compatible not only with materialism, but also with substance dualism and
neutral monism; hence, (1) does not suffice for non‐reductive materialism. So we need to add
another thesis to yield non‐reductive materialism. Typically, non‐reductive materialists hold that
the mental depends on the physical. The kind of dependence at issue is usually a relation of
determination—some kind of supervenience relation. Since the kind of dependence differs in
different versions of non‐reductive materialism, I'll formulate the thesis in the most general (and
imprecise) way:
(2) Mental properties depend on physical properties.
Non‐reductionists may elucidate thesis (2) in more or less strict ways, but, as
(p. 111)
materialists, they agree that mental properties do not depend on anything other than physical
properties. Finally, non‐reductive materialists eschew epiphenomenalism about mental
properties: mental properties make a causal difference (whether or not there are any properties
that are epiphenomenal). Different non‐reductive views construe the causal difference that
mental properties make in different ways, but for all of them
(3) Mental properties make a causal contribution to what happens.
I take the conjunction of theses (1)–(3) to be a minimal schema for any variety of NRM. However,
the conjunction of theses (1)–(3) remains incomplete in several ways.
The conjunction of (1)–(3) yields only a schema because different versions of NRM result
from different elucidations of ‘depend on’ in thesis (2) and ‘make a causal contribution to’
in thesis (3). For example, ‘depend on’ in thesis (2) may be understood variously as:
‘weakly supervene on’ (e.g. Davidson 1980), ‘strongly supervene on’ (e.g. Kim (1998), or
‘globally supervene on’ (e.g. Van Gulick 1993). Any of these kinds of dependence may be
invoked by proponents of NRM who hold that mental properties are realized by physical
properties. And ‘make a causal contribution’ in thesis (3) may be understood as
nomological sufficiency (e.g. Kim 1998) or counterfactual dependence (e.g. Lewis 1973),
or something else.
The schema defined by the conjunction of theses (1)–(3) is only minimal, because different
versions result from supplementing (1)–(3) by various further theses. For example, some
non‐reductive materialists (e.g. Horgan 1993; Antony 1999) hold that mental properties
must be fully explainable and predictable in principle in the vocabulary proper to the
physical sciences. But others (e.g. Davidson 1980; Baker 1993; Burge 1993) hold the
negation: mental properties cannot be both fully explained and predicted, even in
principle, in the vocabulary proper to the physical sciences.3 And many materialists (e.g.
Page 3 of 23
Non‐Reductive Materialism
Lepore and Loewer 1989; Kim 1998) supplement (1)–(3) with a thesis of the causal
closure of the physical: Every physical event has a complete physical cause.
(p. 112)
Page 4 of 23
Non‐Reductive Materialism
The special sciences (e.g. psychology) appeal to mental properties that are neither
identical to physical properties nor reducible to physical properties. A science like
psychology is reducible to a science like neurophysiology only if the laws of psychology
are reducible to laws of neurophysiology by means of biconditional bridge laws
containing predicates of both psychology and neurophysiology (Fodor 1974).5 Bridge laws
would connect the kinds to which psychology appeals to the kinds to which
neurophysiology appeals. But because of multiple realizability, a psychological kind is not
correlated with a single neurophysiological kind, but with a vast range of disjoint
neurophysiological kinds. Moreover, the same issues surrounding the reduction of mental
to neurophysiological properties also arise for the reduction of neurophysiological
properties to cellular properties: neurophysiological properties are multiply realized at
the cellular level.6 (So (ii) is a special case of (i) that also supports thesis (1).) Further,
events recognized by the special sciences supervene on events recognized by
microphysics. (This point supports thesis (2).) Nevertheless, lawlike, counterfactual‐
supporting generalizations of the special sciences are causal: If someone is thirsty, and
believes that the bottle has water in it, then ceteris paribus, he will drink. (This point
supports thesis (3).)
According to Davidson, there are no strict psychophysical laws at all. Mental and physical
predicates are not made for each other. Correct application of mental language is
constrained by holism and normativity that have no place in correct (p. 113) application of
physical language. (This point supports a Davidsonian analogue of thesis (1). Davidson
appeals to predicates or descriptions, instead of to properties.) Furthermore, Davidson
holds that the mental weakly supervenes on the physical in that there is no mental
change without a physical change. (This point supports a version of thesis (2).) Davidson
endorses a token‐identity theory. All events are physical events, but some physical events
are described in mental terms (namely, in the vocabulary of propositional attitudes). The
difference between mental and physical events is merely a difference in how they are
described. Since causation is a relation between events no matter how they are
described, mental events (i.e. physical events described in mental language) can cause
physical events. (This point supports a version of thesis (3).) Since explananda are events
only as described, mental events cannot be given purely physical explanations. Davidson
argues that mental events weakly supervene on physical events. Modifying theses (1)–(3)
for Davidson's appeal to descriptions instead of properties, anomalous monism supports
NMR.
Page 5 of 23
Non‐Reductive Materialism
In the past fifteen or so years Jaegwon Kim has mounted a sustained attack on various
versions of NRM in numerous articles. Unless mental properties are reducible to physical
properties, he argues, they are causally inert or else there is massive (and implausible)
overdetermination. In Kim (1998: ch. 2) he pressed his objections from several directions.
I shall focus on two arguments against NRM: the overdetermination argument, which I'll
sketch briefly, and the ‘downward‐causation’ argument, which I'll set out in detail. Both
arguments need recourse to the idea of higher‐level properties, which Kim roughly takes
to be this: P2 is a higher‐level property than P1 iff the entities where P2 makes its ‘first
appearance’ have ‘an exhaustive decomposition, without remainder, into entities
belonging to the lower levels’ (1998: 15). Each of Kim's arguments against NRM relies on
one or more of the following metaphysical assumptions:
Page 6 of 23
Non‐Reductive Materialism
Assume that mental events are realized by physical events (in the sense of the physical‐
realization thesis), and hence that mental events supervene on physical events. If one
mental event, M, caused another, M*, then there would be a physical event P* that
realized M*, and M* would supervene on P*. On the assumption that the physical is
causally closed, P* has a complete physical cause. Since M* supervenes on P*, the
complete physical cause of P* is also a cause of M*. In that case, M* is overdetermined—
by M and by the complete physical cause of P*. So if mental properties are not identical
with physical properties, and mental events have physical effects, then these physical
effects are overdetermined: All mentally caused events have complete physical causes as
well as mental causes. But it is implausible, claims Kim, that every event with a mental
cause is causally overdetermined.
To bolster his case Kim bids us consider an example of overdetermination. Suppose that
there are two assassins acting independently who shoot a politician at the same time. As
Kim says, it is not plausible that all events with mental causes are overdetermined in that
way. However, as Barry Loewer points out (Loewer 2001, 2002), in contrast to the case of
the two assassins, a mental event and a physical realizer of it are not independent; they
are metaphysically connected. So the analogy misfires.
There is a single argument that can be reconstructed from Kim's writings that, I believe,
is his most forceful and sweeping assault on NRM. After stating the overall argument (as
I–IV below), I'll give it in greater detail. Each of the premises in the overall argument is
defended by one of the sub‐arguments (1–3, 4–7, 8–13, respectively).
Page 7 of 23
Non‐Reductive Materialism
I. If higher‐level properties are both irreducible and causally efficacious, then there
is downward causation by irreducible higher‐level properties.
II. If there is downward causation by irreducible higher‐level properties, then there
are two distinct nomologically sufficient conditions of a single event.
III. There are not two distinct nomologically sufficient conditions of a single event.
∴ IV. Higher‐level properties are not both irreducible and causally efficacious.
Now turn to the arguments for Premises I–III. If mental states are causally efficacious, as NRM
holds, then one irreducible and causally efficacious mental state may cause another mental
state. Suppose that M and M* are mental states realized by physical states P and P* respectively,
and that M ≠ P and M* ≠ P*.
Argument for Premise I:
(p. 116)Hence, the supposition that M causes M* leads to a contradiction (1 and 13). The only
causally efficacious properties are microphysical (or micro‐based macrophysical properties that
are mereological aggregates of subatomic properties—see next section). Therefore, it appears
that if NRM is correct, mental states are causally inert, and epiphenomenalism carries the day. I
shall respond to the downward‐causation argument by proposing another model of NRM that
satisfies the schema for NRM given in Section 6.2. If my model is correct, then the downward‐
causation argument is unsound. (In particular, lines (2) and (8) are false.) Before proposing my
own model, however, I want to revisit an old controversy about the scope of Kim's conclusion.
Page 8 of 23
Non‐Reductive Materialism
(i) Kim's first prong: Micro‐based (or microstructural) macroproperties are properties of
macro‐objects that can be characterized in terms of microstructure: ‘P is a micro‐based
property just in case P is the property of being completely decomposable into
nonoverlapping property parts, a1, a2, …, an, such that P1(a1), P2(a2), …, Pn(an), and
R(a1, …, an)’, (1998: 84). For example, being a water molecule is a micro‐based property:
it is the property of having two hydrogen atoms and one oxygen atom in a certain
bonding relationship. Micro‐based properties, Kim argues, are both macroproperties and
causally efficacious.8 For example, my table's having a mass of 10 kg is a micro‐based
property: it is the property of being completely decomposable into 10 non‐overlapping
parts each weighing 1 kg. Having a mass of 10 kg is a property of the table that is
causally efficacious (it makes the pointer on the scale read ‘10 kg’) and is not a property
of the table's proper parts. Hence, says Kim, we were mistaken to suppose that all
macrophysical properties fall to the argument against non‐reductive mental causation.
Micro‐based macroproperties are causally efficacious.
(p. 117)
(ii) Kim's second prong: Mental properties and their realizers are on the same level. All
properties of a single bearer are at a single level. So my property of intending to lock the
door is at the same level as my property of having microparts with such‐and‐such
microproperties and related in a certain way. So the competition between mental and
physical properties is intralevel. Belief properties and the neural properties that realize
them are at the same level; I have both. So a belief is a second‐order functional property:
the property of having a first‐order property that plays a certain causal role. The
distinction between first‐ and second‐order properties should be distinguished from the
micro‐macro hierarchy of levels: ‘the realization relation does not track the micro‐macro
relation’ (Kim 1998: 82). Neural properties are the first‐order properties—the realizers—
that play the causal role. Mental and neural properties are at the same level, and neural
properties have the causal powers.
Page 9 of 23
Non‐Reductive Materialism
On Kim's view, a property can have a realizer only if it can be ‘functionalized’–that is, only
if it can be construed ‘as a property defined by its causal/nomic relations to other
properties, specifically properties in the reduction base’ (1999: 10).9 Kim ties realization
to supervenience: If P realizes M, then M supervenes on P (1993a: esp. 196–7). The
functional property and its realizers—the supervening property and its base—are on the
same level. In short, says Kim, the problem of mental causation does not generalize to
cross‐level causation, because mental and neural properties are at the same level, and
micro‐based properties are macroproperties that are not susceptible to an analogue of
the problem of mental causation.
Let me respond to Kim's argument. Even if we accept everything that Kim says, there
remains a huge class of important properties that Kim's view will render epiphenomenal
(and hence, by Alexander's Dictum, non‐existent). These are properties mentioned in
causal explanations of psychology, economics, and political science, as well as in everyday
life. They are properties without which we cannot begin to make sense of the world that
we encounter.
I shall coin the term ‘intention‐dependent’, or ID for short, for such properties. They are
properties that cannot be instantiated in a world without beings with propositional
attitudes; for example, being in debt, being a driver's licence, being a delegate. Nobody
can be in debt and nothing can be a driver's license in a world without beings with
propositional attitudes. Call any property either that is a propositional‐attitude property
(like believing, desiring, or intending) or whose instances presuppose that there are
beings with beliefs, desires, and/or intentions an ID property.10 These are properties, non‐
mental as well as mental, whose instances depend on there being creatures with
intentionality. ID properties that we are familiar with include being a wedding, being a
carrot scraper, being a treaty, and so on. Other communities may be familiar with other
kinds of ID properties; but all communities (p. 118) recognize many kinds of ID properties
—as well as other ID objects like pianos and pay‐cheques, and ID phenomena like
conventions and obligations.11 All artefacts and works of art and most human activities
(getting a job, going out to dinner, etc.) are ID phenomena: They could not exist or occur
in a world without beliefs, desires, and intentions. ID properties are not plausibly
construed as micro‐based properties.
However, ID properties are causal properties. By ‘causal properties’ I mean roughly the
properties in virtue of which an object can have some effect: Property P is a causal
property of x iff it's possible that there is some event E such that x causes E in virtue of
having P. Without ID properties we could explain almost nothing that happens in the
world—a president's ordering an invasion; a dean's cutting the departmental budget; a
person's arrest on charges of fraud.
It is highly unlikely that on Kim's account ID properties would turn out to be causally
efficacious. In order to take ID properties to be causally efficacious Kim would have to
construe them as functional properties, whose causal powers are just the causal powers
of their non‐intentional realizers. Consider the property of being a payment of a debt.
Page 10 of 23
Non‐Reductive Materialism
Your payment of a debt confers on you the causal power of clearing your name, and
putting an end to harassing phone calls from your creditor. So payment of a debt is
causally efficacious. Can Kim's view count payment of a debt as causally efficacious? The
answer is affirmative only if three conditions are met.
The first condition, on Kim's view, is that the property must be ‘functionalized’. It is not at
all clear to me that the property of being payment of a debt has a single causal role, or
even how to determine whether it does or not. The second condition is that the causal
powers of an instance of paying a debt reside in its non‐intentional realizer. It is difficult
even to find a candidate to be a non‐intentional realizer of a payment of a debt. Here's
why.
The third condition is to find a theory at the base level that explains how the realizers of
the higher‐level property can instantiate the functional specification. Since we have no
idea of the identity of any non‐intentional realizers, we are in no position to find such a
theory. Thus, it is highly unlikely the property of paying a debt can be functionalized. So it
is unlikely that Kim's reductive approach can rescue ID properties as causally efficacious.
An objector may be tempted to brush aside my argument that Kim's conditions cannot be
met by ID properties, on the grounds that it is merely epistemological. The fact that we
do not know how to carry out the reduction, as we are reminded often, does not imply
that there is no reduction to be carried out (see Antony and Levine 1997). To such an
objector I reply that if one advocates a particular strategy to meet a challenge, one
should give reason to think that the strategy can succeed. If we have no idea of what a
reduction would look like, we are in no position to claim that it can be carried out in
principle. Without the ‘merely epistemological’, one has little reason to believe the loftily
metaphysical.
Page 11 of 23
Non‐Reductive Materialism
Finally, even if Kim's conditions for functionalization were met, ID properties like being
the payment of a debt would violate Kim's causal‐realization principle. The causal‐
realization principle, you recall, is this: If an instance of S occurs by being realized by an
instance of Q, then any cause of this instance of S must be a cause of this instance of Q
(and of course any cause of this instance of Q is a cause of this instance of S). So any
cause of ending the harassing phone calls must bring about the non‐intentional realizer of
the instance of the property of putting an end to harassing phone calls. Hence, if Jones's
payment of his debt is to have the effect of putting an end to harassing phone calls, it
must bring about the non‐intentional realizer. But the non‐intentional realizer, as we have
just seen, includes all the properties on which this instance of putting an end to
harrassing phone calls supervenes, and these include properties far beyond any
individual's control today.
So if Kim is correct, there may well be no intentional causation whatever. Not only is
mental causation at stake, but all causation by objects' having properties whose instances
depend on there being things with propositional attitudes; for example, being written in
Dutch, being in debt, being a delegate. If we are realists about causal explanation (as Kim
and I both are), then without ID properties we would have no causal explanations of, say,
the president's vetoing a spending bill—or of any other historical, political, economic,
social, and legal phenomena. So the problem of mental causation may not generalize to
all macroproperties, but it does seem to generalize to a great swath of macroproperties
that we cannot do without.
Page 12 of 23
Non‐Reductive Materialism
case, as causal properties. There are three elements of the PC view to be explained: the
ideas of properties at different ontological levels, of property constitution, and of
independent causal efficacy.
There are distinct ontological levels: atoms (and aggregations of atoms) are on a lower
level than are credit cards or passports.14 Every object is of some primary kind or other
(see Baker 2000). An object's primary‐kind property determines its level and confers on it
causal powers that cannot be manifested at lower levels. But an object also has other
causal powers at lower levels, as well as at the level of its primary‐kind property. For
example, a bronze statue has some causal powers in virtue of being a statue and some
causal powers at a lower level in virtue of being made of bronze. (So I reject Kim's
conception of levels according to which all properties with a single (p. 121) bearer are on
the same level.) An ordinary woman has causal powers at a personal level (she can make
her friends feel good), as well as at a sub‐personal level (she can rearrange air molecules
when she dives into the pool).
Although there is much more to be said about levels, I have said enough to state the first
thesis of my view: Mental properties are distinct from physical properties, because they
confer causal powers at different levels. The PC view thus satisfies thesis (1) of NRM:
There are mental properties that are distinct from physical properties.
Second, consider the notion of property constitution: Property instances are constituted
by other property instances at a lower level.15 A property's constituter on a given
occasion may be a proper part of a supervenience base for the property, but the
constituting instance (e.g. being an extension of an arm out of a car window) does not
suffice for the constituted instance (e.g. being a left‐turn signal). Property constitution is
much weaker than supervenience: Whether or not one property instance constitutes
another depends on circumstances. For example, raising one's hand in certain
circumstances constitutes voting; one's hand's going up in certain circumstances
constitutes raising one's hand; certain muscles' contracting in certain circumstances
constitutes one's hand's going up; certain molecular motions in certain circumstances
constitutes the muscles' contracting; and so on.16 Even the relation of atomic properties
to molecular properties requires that the atomic‐property instantiations be in certain
circumstances: Instances of being a sodium atom and being a chlorine atom do not
constitute anything unless they are in circumstances of bonding.17
Page 13 of 23
Non‐Reductive Materialism
Although I can only be brief here, let me informally introduce another term needed for a
schema for property constitution:18 ‘G‐favourable circumstances’. G‐favourable
circumstances are the milieu in which something can have the property of being a G. The
G‐favourable circumstances are conditions such that the addition of an appropriate F‐
instance makes it the case that there is a G‐instance, but not so comprehensive that just
anything in G‐favourable circumstances guarantees that there is a G‐instance. Then,
when a suitable F‐instance is in G‐favourable circumstances, it comes to constitute a G‐
instance. (To be suitable, an F‐instance cannot cease to be an F‐instance when put in G‐
favourable circumstances.) G‐favourable circumstances may be characterized by open
sentences which are satisfied by an appropriate F‐instance. G‐favourable circumstances
are conditions that are necessary but not (p. 122) sufficient for a G‐instance. If an F‐
instance is in G‐favourable circumstances, then ipso facto there is a G‐instance. For
example, if a hand raising is in vote‐favourable circumstances, then ipso facto there is a
vote. Here is a schema for property constitution:
The potential constituters of an instance of G may have nothing in common, other than their
suitability to constitute an instance of G in various circumstances.20 For example, a single
instance of the property of voting may be constituted by an electronic signal, a mark on paper, a
hole in paper, or something else.21 There is no general answer to the question of how much
latitude there is among potential lower‐level property instances that may constitute a single
higher‐level property instance. My only point is that there is some latitude: a constituted
property instance may have any of a variety of different kinds of non‐intentional constituters,
and there may be no physical similarities among the potential constituters.
The definition (PC) is too broad. It allows that, say, an instance of having mass constitutes
an instance of being a passport. To remedy that, we may define ‘direct property
constitution’ as follows:
Wherein, then, lies the dependence of the mental on the physical, or, more generally, of
the constituted property instances on their constituters? Although constitution is not
itself a supervenience relation, where there is constitution, there is a supervenience
Page 14 of 23
Non‐Reductive Materialism
Third, consider the notion of independent causal efficacy. Constituted property instances
confer causal powers that are ‘over and above’ the causal powers of their constituters.
Some non‐reductionists hold that a property instance has independent causal efficacy if
and only if it would have had its effect even if its constituter had been different.22 I would
add that the causal powers of the constituted properties are not determined by those of
the constituter alone. The G‐favourable circumstances are required for the constituted
property to be instantiated. So
(IC) A property instance has independent causal efficacy if and only if (i) it would
have had its effect even if its constituting property instance had been different, and
(ii) it confers causal powers that could not have been conferred by its constituting
property instance alone.
Any property whose instances have independent causal efficacy is a genuine causal property. My
thesis, then, is this: ID properties generally (with mental properties as a special case) are causal
properties because their instances have independent causal efficacy. Consider an example.
Let V be Jones's voting against Smith at t.
P be Jones's hand going up at t.
V* be Smith's getting angry at Jones at t′.
P* be Smith's neural state at t′.
C be circumstances that obtain at t in which a vote is taken by raising hands (‘vote‐
favourable circumstances').
The causal powers conferred by the constituted property instance (Jones's voting against Smith)
are independent of the causal powers conferred by the constituter (Jones's hand’ going up). The
causal powers conferred by Jones's hand going up include the power to move air molecules. The
causal powers conferred by Jones's voting against Smith include the power to anger Smith—no
matter how the vote was cast. In short, the causal efficacy of constituted property instances—of
mental (p. 124) property instances and of instances of intention‐dependent properties generally
—is independent of the causal efficacy of their constitutors. The PC view thus satisfies thesis (3)
of NRM: Mental properties make a causal contribution to what happens.
Thus, the PC view satisfies the schema for NRM given in Section 6.2. The PC view, I
believe, vindicates non‐reductive mental causation.
Page 15 of 23
Non‐Reductive Materialism
Finally, note that the PC view does not violate the causal‐closure principle. The causal‐
closure principle says, roughly, that any physical event that has a cause at t has a
complete physical cause at t (Kim 1993b: 280).23 On my view, all property instances are
physical in this respect: any property instance is either identical to or ultimately
constituted by microphysical property instances. ID properties thus are physical
properties. So their causal efficacy does not violate the causal‐closure principle.
Someone may object that ID properties as I have construed them are not really physical
properties: the only physical properties are microphysical, or micro‐based (p. 125)
properties that are just aggregates of microphysical properties (see Kim 1998: 114). Even
so, the PC view would not violate the causal‐closure principle. Consider a case of basic
action: Suppose that Jane is going through the security gate at a US airport, and she is
instructed by a federal agent to raise her arms, so that the agent can ‘wand’ her. Jane
wills24 to raise her arms (M) and she raises them (M*). Suppose that her willing to raise
her arms causes her to raise them. Let MP be the microphysical constituter of Jane's
willing to raise her arms and let MP* be the microphysical constituter of Jane's raising
Page 16 of 23
Non‐Reductive Materialism
her arms. (Note that the relation between MP and M, on the one hand, and MP* and M*,
on the other hand, is not Kim's realization relation but my constitution relation.)
On the PC view, MP is not a complete cause of MP*. The causal‐closure principle requires
only that MP* have a complete microphysical cause, not that MP be that complete cause
of MP*. MP is only a proper part of a larger aggregate of microproperties that is
nomologically sufficient for MP*.25 There is no difficulty for the property‐constitution view
in saying: (i) Jane's willing to raise her arms causes her to raise her arms; (ii) Jane's
willing to raise her arms is constituted by MP; (iii) Jane's raising her arms is constituted
by MP*; but (iv) MP does not cause MP*. If the microphysical state of one sizable spatio‐
temporal region that ends at the time of Jane's willing caused the microphysical state of
that slightly later sizable region that begins at the time of Jane's raising her arms, then
the causal‐closure principle is honoured.26 So although the PC view does not require MP
to be causally sufficient for MP*, it nevertheless does not violate the causal‐closure
principle.27
6.8 Conclusion
Non‐reductive materialism is the most promising metaphysical view for understanding
the world as we encounter it—the world filled with ordinary things like people (p. 126)
and artefacts and works of art. Non‐reductive materialism alone offers a metaphysics that
takes ordinary things and their interactions at face value and makes them intelligible; and
it alone respects a common‐sense conception of reality in the context of a broadly
scientific outlook.28
References
Antony, L. (1999), ‘Making Room for the Mental: Comments on Kim's ‘Making Sense of
Emergence’, Philosophical Studies, 95: 37–44.
Antony, L., and Levine, J. (1997), ‘Reduction With Autonomy’, Philosophical Perspectives,
11: 83–106.
Page 17 of 23
Non‐Reductive Materialism
Baker, L. (1993), ‘Metaphysics and Mental Causation’, in J. Heil and A. Mele (eds.),
Mental Causation (Oxford: Oxford University Press), 75–96.
Boyd, R. (1999), ‘Kinds, complexity and multiple realization’, Philosophical Studies, 95:
67–98.
Burge, T. (1993), ‘Mind‐body Causation and Explanatory Practice’, in J. Heil and A. Mele
(eds.), Mental Causation (Oxford: Oxford University Press), 97–120.
Crisp, T., and Warfield, T. (2001), ‘Kim's Master Argument’, Noûs, 35: 304–16.
Heil, J., and Mele, A. (1993) (eds.), Mental Causation (Oxford: Oxford University Press).
repr. in Kim, Supervenience and Mind (Cambridge: Cambridge University Press, 1993),
237–64.
repr. in Kim, Supervenience and Mind (Cambridge: Cambridge University Press, 1993),
309–35.
—— (1993a) ‘The nonreductivist's troubles with mental causation,’ in J. Heil and A. Mele
(eds.), Mental Causation (Oxford: Oxford University Press), 189–210;
Page 18 of 23
Non‐Reductive Materialism
repr. in Kim, Supervenience and Mind (Cambridge: Cambridge University Press, 1993),
336–57.
repr. in Kim, Supervenience and Mind (Cambridge University Press, 1993), 265–84.
—— (1998), Mind in a Physical World: An Essay on the Mind‐Body Problem and Mental
Causation (Cambridge, Mass.: MIT Press).
(p. 127) —— (1999), ‘Making Sense of Emergence,’ Philosophical Studies, 95: 3–36.
Lepore, E., and Loewer, B. (1989), ‘More on Making Mind Matter’, Philosophical Topics,
17: 175–91.
repr in Lewis, Philosophical Papers, ii. (Oxford: Oxford University Press, 1986), 159–72.
Loewer, B. (2001), review of J. Kim, Mind in a Physical World, Journal of Philosophy, 98:
315–24.
Nagel, E. (1961), The Structure of Science (New York: Harcourt, Brace & World).
Page 19 of 23
Non‐Reductive Materialism
Putnam, H. (1975), ‘The Nature of Mental States’, in Putnam, Mind, Language, and
Reality: Philosophical Papers, ii (Cambridge: Cambridge University Press), 429–40.
Searle, J. (1992), The Rediscovery of the Mind (Cambridge, Mass.: MIT Press).
Segal, G., and Sober, E. (1991), ‘The Causal Efficacy of Content’, Philosophical Studies,
63: 1–30.
Shoemaker, S. (2003), ‘Causality and Properties’, in P. van Inwagen (ed.), Time and Cause
(Dordrecht Reidel, 1980), 109–35;
repr. in Shoemaker, Identity, Cause and Mind, 2nd edn. (Oxford: Clarendon 2003), 206–
34.
Van Gulick, R. (1993), ‘Who's In Charge Here?’, in J. Heil and A. Mele (eds.) Mental
Causation (Oxford: Oxford University Press), 233–58.
Notes:
(2) Historically reduction has been taken to be a relation between theories. According to
Ernest Nagel's theory reduction, theory T2 is reduced to theory T1 just in case there are
‘bridge laws’ connecting the predicates of T2 and T1, and T2 is deducible from T1
together with the bridge laws (see Nagel 1961).
(3) There is also room to hold that mental properties are predictable, but not explainable,
in the vocabulary of the physical sciences.
(4) There is an enormous literature on multiple realization (see e.g. Block 1997).
(5) For a different non‐reductive conception of the special sciences see Boyd (1999).
(6) Since the special sciences include all the sciences above the level of microphysics, if
all the special sciences were reduced to microphysics there would be no macro‐causation
at all. I'll return to this matter in Section. 6.5.
(7) See Baker (1993), Burge (1993) and Van Gulick (1993) for arguments that Kim's
claims against mental causation generalize to all macroscopic properties.
Page 20 of 23
Non‐Reductive Materialism
(9) Kim argues that functional properties are reducible to the bases that realize them. He
gives a procedure for property reduction. To reduce E to a reduction base B, first, give a
functional definition of E in terms of its causal relations to other properties (properties in
reduction base B); second, find the physical realizers of E in B; third, find a theory at the
level of B that explains how realizers of E fulfil the causal role specified in the definition.
(11) In other places I've used the expression ‘intentional property’ to refer to ID
properties, and, less fortunately, ‘intentional object’ to refer to ID objects. Although I
characterized what I meant by ‘intentional object’ carefully, I am now resorting to the
technical term ‘ID object’ in order to avoid confusion with uses of ‘intentional object’
associated with Brentano and Meinong.
(12) Pereboom and Kornblith (1991) Clapp (2001), and Pereboom (2002) all use the term
‘constitution’, but my view differs significantly from each of theirs.
(14) Kim defines levels mereologically: objects having properties at one level become
parts of objects having properties at higher levels.
(15) I say ‘property constitution’ for convenience. What is constituted are property
instances, not properties themselves. Property constitution is analogous to the idea that I
developed in Baker (2000) for understanding material objects in terms of what I simply
called ‘constitution’.
(16) Pace Kim (1998), I take the voter to be the bearer of all the properties at the
different levels.
(17) I individuate property instances in such a way that the same property instance could
have occurred in different circumstances. Any G‐instance must be in G‐favourable
circumstances, but if a G‐instance is constituted by an F‐instance, the F‐instance (which
in fact is in G‐favourable circumstances) could have occurred in non‐G‐favourable
circumstances.
(18) The schema for constitution of properties differs from the one for constitution of
particulars given in Baker (2000) and elsewhere. In the schema for constitution of
particulars, F and G are x's and y's primary‐kind properties, respectively, and x and y are
Page 21 of 23
Non‐Reductive Materialism
(19) x has F but lacks G at t if the F‐instance is not in G‐favourable circumstances. X's
having the property of being a salt molecule is constituted by x's having the compound
property of being a sodium atom and being a chlorine atom—but only in salt‐favourable
circumstances. If the properties of being a sodium atom and being a chlorine atom were
instantiated in circumstances that prevented bonding, there would be no salt molecule.
But necessarily when the chlorine and sodium atoms are in salt‐favourable
circumstances, there is a salt molecule. The necessity here is metaphysical necessity.
(20) This feature distinguishes my idea of property constitution from ideas of constitution
found in Pereboom (2002), and from Clapp (2001).
(21) For a defence of this claim see Pereboom and Kornblith (1991); see also Pereboom
(2002). My view differs from Pereboom's in the latter article in several important ways.
Most significantly: (i) Pereboom sets aside ‘any fundamentally relational causal powers’.
(ii) Pereboom takes the relation between levels to be realization, where a realizer is
nomologically sufficient for the realized property. (iii) Pereboom takes the causal powers
of the realized property to be determined by (‘constituted by’) those of the realizer. I
differ on all scores: (i) Assuming that causal powers derive from properties in virtue of
which something has an effect, I take almost all intentional causal powers to be
relational. (ii) I take the relation between levels to be constitution, where a constituter is
not nomologically sufficient for the constituted property. (iii) I take the causal efficacy of
ID properties not to be determined by their constituters.
(22) Pereboom (2002) and Pereboom and Kornblith (1991) explain this point fully and
persuasively.
(23) This principle is important, says Kim, because to deny it ‘is to accept the Cartesian
idea that some physical events have only nonphysical causes’ (1993b: 280).
(24) I am using ‘will’ as an all‐purpose term that covers choosing, deciding, forming an
intention for the immediate future. ‘Will’ carries no metaphysical weight here.
(25) See Noordhof (1999) and Segal and Sober (1991). I discovered these articles after I
had written the paragraph to which this note is appended.
(26) There is much more to be said about the causal‐closure principle. Kim holds that
physicalism ‘need not be, and should not be, identified with micro‐physicalism’ (Kim
1998: 117). In that case, if we disentangle the causal‐closure principle from the thesis of
mereological supervenience, my own non‐reductive view satisfies the causal‐closure
principle (see Kim 1998: 117).
Page 22 of 23
Non‐Reductive Materialism
(27) We may still have a harmless kind of overdetermination. But note that the
overdetermination is generated by the whole supervenience base, not by the constituter.
(28) I am grateful to Gareth B. Matthews, Hilary Kornblith, and Jonathan Schaffer for
reading drafts of this paper and making helpful suggestions.
Page 23 of 23
Functionalism
Functionalism at its core is the thesis that minds and mental kinds are to be understood
in terms of the roles or functions that specific states and processes play within suitably
organized systems. From a functionalist perspective, minds differ from non-minds not in
any distinctive substance or fundamental substrate, but in their systemic organization
and the roles played by their parts and sub-parts within it. A minded system is simply one
that is organized in the right sort of way, though just which ways those are is a difficult
and disputed matter. Functionalists classify states or processes largely, if not solely, in
terms of the relevant roles or functions that they play in some such system.
Keywords: substrate, functionalism, systemic organization, behaviourism, mental states, brain processes
FUNCTIONALISM at its core is the thesis that minds and mental kinds are to be understood in
terms of the roles or functions that specific states and processes play within suitably
organized systems. From a functionalist perspective, minds differ from non‐minds not in
any distinctive substance or fundamental substrate, but in their systemic organization
and the roles played by their parts and sub‐parts within it. A minded system is simply one
that is organized in the right sort of way, though just which ways those are is a difficult
and disputed matter.
Functionalists classify states or processes largely, if not solely, in terms of the relevant
roles or functions that they play in some such system. A process is treated as one of
perceiving, remembering, or desiring in virtue of its systemic role, and so too are specific
states such as believing that the climate has changed, intending to buy a tuna sandwich,
or hearing the wind in the pines. Mental types and kinds are to be understood in term of
their associated roles. The underlying substrate of mind matters only in so far as it
supports and affects its system's organization and operation and the roles thus realized
within it. From a functionalist perspective, mental differences are primarily differences in
Page 1 of 28
Functionalism
function or role, and substrate variations that do not alter role thus do not matter
mentally.
Functionalism per se is not a specific thesis but more a general perspective on the
mental. Particular functionalist theories and models share that basic view but differ
greatly in how they develop it and what they take its import to be. Thus, in discussing
functionalism, one must be clear about just what claim is at issue.
7.1 History
Though functionalism no doubt has some historical antecedents, it is basically a
contemporary view that originated in the 1960s and early 1970s. According to one
(p. 129) standard and oft‐repeated history (Block 1980a), functionalism arose in response
to the problems confronting the two main versions of physicalism then extant: logical
behaviourism and the psychophysical type‐type identity theory.
Behaviourism had many supporters but confronted many objections, of which some
concerning sufficiency and holism may have played a role in the origin of functionalism.
Various examples were raised to challenge the sufficiency of the behaviourist analyses.
Imagined examples appealing to perfect actors or super‐Spartans were alleged to show
that all the relevant behavioural facts and dispositions might be true of a person who
nonetheless lacked the relevant mental state. A perfect actor who was not in any actual
pain might nonetheless be disposed to behave in every pain‐relevant way in all the
relevant situations, and a super‐Spartan (Putnam 1965), stoically trained to never show
pain, might conversely have no dispositions to behave in any pain‐related ways despite
really being in pain. The behavioural and dispositional facts in such cases would not
suffice to fix the mental facts.
Others alleged that the holism of the mental made behaviourism viciously circular, at
least in so far as it was intended as a reductive proposal for explaining the mental in
terms of the non‐mental. The difficulty was with the antecedents of the conditional
dispositions that figured in the behaviourist analyses. Wanting some coffee will lead one
to drink a cup one is offered, but only if one believes that the cup contains coffee and has
Page 2 of 28
Functionalism
Like behaviourism, the type‐type identity theory had both supporters and critics. Among
the issues raised against it, the multiple‐realization objection seems to have played a role
in the development of functionalism (Putnam 1967a). The objection turns on the empirical
likelihood that the neural bases of mental states will vary in different organisms, across
or even within species. Human pain, lizard pain, and octopus pain are quite likely
different at the neural level, and thus pain might be said to be multiply realized. If the
neural basis of pain can vary so widely across cases, then it would seem the property of
being in pain is not strictly identical with any of those particular neural properties. Alien
creatures might be even more unlike us physiologically, yet it seems plausible to suppose
they too could have beliefs, desires, or pains. Thus, according to the critics, the link
between mental states and physical states should not be regarded as one of type‐identity
but as a matter of realization, with many physical bases able to realize a mental state of a
given type. Being a pain or a belief would be more like being an eye or a stomach than
being water or heat.
According to the standard history (Block 1980a), behaviourism failed in part because it
was ‘too liberal’ in counting things as having minds or particular mental states when in
fact they lacked them, while type‐type identity erred in the opposite direction by being
‘too conservative’ in refusing to count creatures as having minds or pains simply because
they failed to share our specific human neurophysiology. Functionalism, on that story,
developed to avoid both pitfalls. Such functionalists defined types of mental states in
terms of the roles they played within an internal network of states mediating an
organism's or system's engagement with its environment. A state's role thus included not
only links to situational inputs and behavioural outputs but also its interactions with other
internal states. The appeal to internal organization thus added conditions beyond the
purely behavioural ones that had proved insufficient for the behaviourist. The
functionalists could rule out purported behavioural counter‐examples on the ground that
Page 3 of 28
Functionalism
the behaviour was not produced by the right sort of internal organization. Moreover,
since the relevant organization might be realized by many different neural or even non‐
neural structures, the functionalist could avoid the chauvinism charge. Any system, no
matter how unhuman in the details of its physiology, might nonetheless count as having a
mind, as long its specific structure realized the relevant sort of organization.
The standard history is no doubt part of the story, but only part of it. Other factors also
played a role in the origin of functionalism; and functionalism in some of its forms has
stayed closer to behaviourism or to the identity theory than that history suggests.
Philosophic thinking was surely influenced by other developments of the time. The rise of
cognitive psychology and of attempts to build data‐driven models of mental processes put
the emphasis on the systemic organizational aspect of mind. General advances in
computer technology and the widespread acceptance of the software‐hardware metaphor
also suggested an analogous realization view of the mind‐body relation. Software is to
hardware as function is to structure, and as mind is (p. 131) to brain. Brain processes
provide the substrate for the mental processes they realize. The notion of the mind as an
information‐processing system, which arose in the 1960s, reflected the influence of both
computational and cognitivist trends (Lindsay and Norman 1977).
Conversely, not all versions of functionalism broke as cleanly with their behaviourist and
identity‐theorist predecessors. Some functionalists, most notably Daniel Dennett (1978,
1987) but others as well (Bennett 1992; Baker 1995), interpret the view in a basically
behaviourist way. They accept the importance of holistic interdependence in analysing
mental‐state concepts but nonetheless regard the truth makers for mental attributions as
solely facts about actual and counterfactual behaviour. They acknowledge that the role
played by a given desire, for example, cannot be specified without taking into account
other desires, beliefs, and intentions, but they unpack those dependencies solely in terms
of complex conditionals about patterns of behaviour. Such functionalists reject any inner‐
route constraint that requires that the relevant behaviour be produced by a specified type
of internal process. All that is needed is that there be some internal mechanism able to
support the relevant behavioural facts.
Other functionalists have appealed to facts about the roles associated with mental kinds
to support a version of type‐type identity theory (Armstrong 1968; Lewis 1972). Rather
than grounding identity claims on correlational evidence, they argue for a two‐step
process in which mental concepts are analysed in terms of their roles and then empirical
science discovers the actual structures or properties, probably neural in nature, that in
fact play those roles. For example, we might analyse our concept of pain in terms of
relations to noxious stimuli, avoidance behaviour, and other mental states such as desires,
memories, and emotions. Neuroscientists might then discover the specific brain
structures or properties that play those roles in humans; for example, x‐firings in the
cingulate cortex. The conclusion drawn by such functionalists would be that pain, at least
human pain, is type identical with x‐firing in the cingulate. Functionalism construed in
Page 4 of 28
Functionalism
this way is not really an alternative to the type‐type identity theory but rather a way to
articulate and support the identity theory. (See occupant functionalism in the next
section.)
(p. 132)
Some functionalists intend their claim to apply to all mental states, but others restrict it
to a particular subset. Some take it to be a claim about our ordinary common‐sense or so
called folk mental kinds, such as pain, anger, thought, and desire. Others restrict the
range further to just those folk states that figure in our rational‐agent model such as
believing, wanting, and intending, or to those that involve propositional attitudes, such as
believing that p, hoping that q, or seeing that s. Of particular importance is whether one
includes conscious and experiential states within the scope of the claim, since many
objections to functionalism appeal to its supposed limits in dealing with the phenomenal
or experiential aspect of mind. The scope of the claim also depends on whether one
interprets it as solely a claim about actual human mental states, or as extending to animal
mental states, or even more broadly to the possible mental states of robots, aliens, or
radically non‐human systems.
Other functionalists interpret the claim as referring to the mental kinds that are used (or
should be used) in some present or future psychological theories, especially empirically
based scientific theories. Those kinds might or might not match up very well with our folk
kinds, but if the theories are successful in describing and explaining human or non‐
human mental processes, there may be good reason to accept them as real mental kinds.
Functionalism about such kinds is thus a sort of description or prescription of how we do
or should scientifically theorize about the mind. Claims of this sort are sometimes
labelled ‘psycho‐functionalist’ (Block 1980a).
The notion of functional role gets interpreted in an equally diverse range of ways. Specific
versions of functionalism differ in how they unpack that notion and in the theoretical
resources they require to specify such roles. Though each admits of many subtypes, the
Page 5 of 28
Functionalism
Some interpret functional role in terms of the mathematical functions, and thus as
covering pretty much any mapping from inputs to outputs. To specify such a function, one
needs to specify the items (or types of items) in the input set, as well as in the output set,
and the rule or mapping between them. Thinking of roles in this way is especially apt for
computational or machine‐state versions of functionalism, which explicate mental states
and processes in terms of abstract computational relations such as those that define
Turing machines and other sorts of formal automata. Treating functional roles in this way
also naturally links up with work in artificial intelligence and computational psychology
(Putnam 1967b).
Others interpret the relevant roles in terms of causal relations, in particular in terms of
fairly simple relations such as ‘causing’, ‘jointly causing’, ‘inhibiting’, and ‘causing unless
… ’, which may seem to be the sorts of causal links invoked by folk psychology (Lewis
1972). A desire for coffee together with a belief that there is a café round the corner will
jointly cause me to walk there unless I have some contrary attitude such as an intention
to get to my office as quickly as possible. Mental states are treated as nodes in a linked
network which is explicitly interpreted only in terms of its inputs and its outputs. The
inputs might be specified as stimuli or environmental situations, and the outputs as
responses or behaviours. The internal states are characterized implicitly and solely in
terms of their location within the network of links.
(p. 133)
A key issue is the sorts of conceptual resources one may use to specify such causal
networks. On an austere model one may appeal only to simple and general causal notions
like those listed just above in specifying the links, and only to basic physical descriptions
in characterizing inputs and outputs, such as ‘looking at a red cup’ and ‘raising one's
right hand’. Specifying causal networks in this restricted way would be apt if one views
functionalism as a reductive account of the mental; that is, one that aims to reduce the
mental to the non‐mental.
However, it is not clear that any such restricted networks could adequately capture the
relevant mental roles. In particular, they do not seem in accord with the sorts of causal
generalizations one finds in folk psychology, which typically invoke more mentalistic
notions in characterizing the relevant links as well as in specifying inputs and outputs.
Folk theory, for example, holds that if one is in an embarrassing situation one is likely to
apologize or attempt to explain away one's gaffe. How could one specify in purely
physical terms what counts as an embarrassing situation or an attempt to explain away
one's error? Folk theory includes many similar mentalistic generalizations; for example, if
one comes to believe that p, one will typically also come to believe the obvious and
plausible consequences of p; one's set of goals and preferences tends toward rational
coherence; and if one perceives a dangerous threat one will react in fearful and evasive
ways. Such platitudes are among those that figure in the roles associated with our
Page 6 of 28
Functionalism
ordinary mental kinds, but they characterize the relevant sorts of inputs, outputs, and
causal links in ways that involve mental and intentional concepts, and thus could not be
used by the reductive functionalist. By contrast, other functionalists of a less reductive
bent might well use such concepts in specifying causal roles (Van Gulick 1982).
Causal‐role versions also differ in how widely or narrowly they specify the relevant causal
networks. Narrow versions restrict the networks to the internal causal structure of the
organism or system (Stich 1983), and thus individuate roles or mental states solely in
terms of factors internal to the individual. Wide versions allow the relevant networks to
extend out into the world beyond the organism, and thus they may appeal to
environmental, social, or historical factors in specifying and individuating roles and states
(Putnam 1975; Burge 1986). The wide/narrow distinction can also be applied to other
sorts of functionalism, but it is especially relevant with regard to causal‐role versions.
Some may object that construing ‘function’ in a teleological sense would inject a
problematic notion of purpose or ‘final causation’ into functionalism, a worry that may be
of special concern to those who view functionalism as a reductive thesis. If mental kinds
are to be understood in terms of teleological functions, then the prospects for a reductive
account of mind will depend on the degree to which teleology can itself be reductively
understood or naturalized (Wright 1973; Nagel 1977; Millikan 1984).
Page 7 of 28
Functionalism
Many accounts of teleology have been offered in the recent literature, and most aim to
explain it in a naturalistic way. The largest division is between those that treat historical
facts about a structure's origin as a necessary condition on its function (Wright 1973;
Millikan 1984) and those who do not (Cummins 1975; Nagel 1977). In particular, some
hold that a structure S can have the teleological function of doing X only if S has been
selected for doing X. Though accounts that include such a historical requirement are in
keeping with evolutionary theory, they are open to so‐called swampman objections that
appeal to a hypothetical case in which a molecular duplicate of a normal human arises by
accident (Davidson 1987). Since such a swampman would have no selectional history, its
inner states would lack any functions according to historical accounts of teleology. Yet in
so far as it would be physically and behaviourally indistinguishable from ordinary
humans, it seems wrong to deny that it would have mental states of the normal sort. Non‐
historical accounts of teleology can avoid such counter‐examples, but they may face other
problems of their own. The teleological notion of function thus remains in dispute.
Nonetheless, the various teleological accounts—together with the computational and
causal readings—provide the three main ways to interpret the functionalist notion of
‘function’.
The third parameter along which versions of functionalism vary is the nature of the link
they assert between the mental and the functional. What does understanding the mental
in terms of the functional involve? Again there are three main families of views, which
respectively treat functionalism as an ontological claim about property identity, an
analytic claim about mental concepts, or merely as a pragmatic thesis about the utility of
viewing minds and mental kinds from a particular theoretical perspective.
(p. 135)
Ontological versions of functionalism typically concern the identity and true nature of
mental kinds or properties, where properties are considered as objective constituents of
reality distinct from whatever concepts or predicates we may use to describe or refer to
them. If feeling a pain and believing that lilies smell sweet are real properties
instantiated in the world, then what sort of properties are they? What is essential to their
instantiation? And how are they related to other properties, whether physical, functional,
or of some other sort? Different functionalists answer in different ways.
As noted in the history section, some early functionalists were motivated to reject the
psychoneural identity thesis on the basis of the multiple‐realization objection. If many
different neural (or non‐neural) substrates might realize pain in different organisms, then
the property of being in pain could not be identical with any of those particular neural
properties. According to such functionalists, the relation between mental and neural
properties is better considered as a matter of realization rather of identity.
What then is the relation between the functional and the mental? Like mental properties,
the relevant functional properties are realized by the underlying neural substrate, but
how many distinct acts of realization are involved? If the functional properties are
identical with the mental properties, then only one act of realization is required. If they
Page 8 of 28
Functionalism
are not identical, there are at least two options. Either they are independently realized by
the physical, or they get realized in a connected way perhaps by a two‐level process
through which the physical realizes the functional, which then in turn realizes the mental.
Among those who interpret functionalism as a claim of property identity there is a major
division between so‐called role functionalists and occupant functionalists. Role
functionalists identify a given mental property M1, for example being an instance of pain,
with the property of playing the relevant role R1. The properties of being pain and of
playing the pain role are regarded as one and the same property. If one thinks instead in
terms of properties had by whole persons, such as that of ‘being in pain’, the role
functionalist would identify such mental properties with the property of ‘being in some
state which plays the relevant role’; for example, the pain role (Putnam 1967a; Lycan
1987).
By contrast, occupant functionalists (Armstong 1968; Lewis 1972) identify the mental
property with the structural or neural property that in fact occupies or plays that role in a
given reference population, such as normal humans. If neuroscience discovers that
patterns of chaotic firing in the cingulate cortex play the pain role in humans, then the
occupant functionalist would conclude that the property of human pain is identical with
the property of having chaotic cingulate firings. (The cingulate does in fact appear to play
part of the pain role, but the neural substrate of human pain seems to be widely
distributed across multiple brain regions.) From the perspective of occupant
functionalism, the role associated with a given mental kind serves as a means to secure
reference to the relevant property, but it does not typically give the identity or essence of
that property, which must instead be discovered through empirical investigation. The
roles serve more or less as a list of symptoms might do in the discovery of a disease. One
might refer to some unknown agent as ‘the cause (p. 136) of river blindness’ and only
later discover that the relevant property is that of ‘being infected with the parasitic
nematode Onchocerca volvulus’.
Thus, both role and occupant functionalists interpret the view as an identity claim about
mental properties, but the two groups identify them with very different sets of non‐mental
properties: either with role properties or with the specific structural properties that
occupy or realize such roles in a given population. The identities proposed by the
occupant functionalist may in fact be very much the same as those asserted by the
psychoneural identity theorists, at least if the proposed identification is for a population‐
or species‐specific property such as ‘human pain’. The fact that very different structural
properties might realize the pain role in different populations need pose no conflict for a
suitably restricted version of occupant functionalism. If the structural properties that
realize the pain role in octopuses or Martians are structurally very different from those
that do so in humans, the occupant functionalist will conclude that human pain, octopus
pain, and Martian pain are three distinct properties which merely play parallel roles in
their respective contexts (Lewis 1980). The role functionalist might more probably regard
the three groups as sharing a common mental property in virtue of instantiating the same
Page 9 of 28
Functionalism
role property; namely, the property of having some state or other that plays the pain role
for that population.
Given the ability to draw the distinction and to individuate mental properties in both
ways, one may wonder if there is a genuine dispute between the occupant and role
functionalist, or merely a verbal disagreement. Is there really an independent fact of the
matter about whether humans and Martians are in identical mental states or only in
functionally similar mental states? Are there principled and compelling reasons to
individuate mental properties in the one way rather than the other? If there are any such
reasons, the burden would seem to be on the disputants to produce them, and it is not
clear they have yet done so.
Such conceptual claims may vary in strength. A strong version of analytical functionalism
may claim the ability to give strict necessary and sufficient conditions for mental
concepts in solely functional terms. A more modest version might claim to give conditions
that are necessary though not sufficient, or the converse, but ones that nonetheless
elucidate critical aspects of our mental concepts. Claims of that latter sort might not
qualify as offering a conceptual reduction, but may still explicate a critical functional
aspect of our mental concepts. A non‐reductive analytical functionalist, for example,
might concede that it is impossible to give strictly sufficient (p. 137) conditions for the
mental in solely functional terms, but still hold that functional factors are essential for
distinguishing among mental kinds; for example, our concept of a pain may be that of a
state that is mental and that plays a given role.
Conceptual claims of the more modest sort overlap at the borders with the third family of
views about the relation between the mental and the functional, which might be labelled
‘pragmatic functionalism’ in contrast to the ontological and conceptual versions of the
view. According to the pragmatic functionalist, attending to the functional roles played by
mental states and processes can be of great use in theorizing about the mental, in helping
us both to discover the nature of mental reality and to clarify our mental concepts. She
regards functional models as a powerful tool for enhancing our understanding of the
mental and its relation to the non‐mental. However, the pragmatic functionalist is not
Page 10 of 28
Functionalism
committed to any ontological or conceptual reduction, and does not claim that functional
roles and relations exhaust the nature of mental properties or mental concepts.
There are thus many different versions of functionalism. To interpret the basic claim that
mental states/kinds should be understood in terms of functional roles and relations one
needs to specify which mental states are covered, which notion of function is involved,
and what understanding the former in terms of the latter involves. The diverse answers
one might give to those three questions generate a wide variety of functionalist theses as
summarized in Table 7.1.
Various cross‐combinations from the three columns are possible. The functionalist might
ontologically identify the mental properties referred to by some empirical psychological
theory with computational properties. Alternatively, she might claim our propositional‐
attitude concepts can be fully analysed in terms of teleological roles. Other (p. 138)
functionalists might assert that human mental states can be usefully and practically
modelled in terms of the reductively specified causal roles that they play. Indeed, a given
functionalist might consistently support several such combinations. For example, some
have analysed our folk concepts in terms of causal roles, and then argued that as an
ontological matter the properties satisfying those concepts in humans should be identified
with the neural properties occupying those roles.
Occupant
Page 11 of 28
Functionalism
Page 12 of 28
Functionalism
Coherence
The very coherence of functionalism has been challenged from early in its history. The
basic general claim that mental concepts or properties are functional in nature does not
amount to much unless one can give a clear idea of what counts as functional and what
distinguishes the functional from the non‐functional. Without any such distinction, the
basic functionalist claim threatens to collapse from a lack of content. Some early versions
of functionalism were explicitly intended as an alternative to type‐type psychoneural
identity theory, and thus as offering a fundamentally different solution to the mind–body
problem. If mental properties were functional, and neural properties were structural,
then their relation should be understood as a matter of realization rather than identity,
especially given the possibility of multiple realization.
Moreover, the notion of realization provided a means of vindicating the physicalist claim
that everything real is physical without needing to embrace outright type‐type identity.
The physicalness of reality consists in the fact that everything real is physically realized.
Realization can thus support a more robust physicalism than other non‐identity relations,
such as the supervenience of the mental on the physical, which turned out to be
compatible with non‐physicalist views such as property dualism.
However, as critics noted, the structural/functional distinction extends well into the realm
of the physical and the neural, and is relative rather than absolute. What counts as
functional depends largely on how one chooses to individuate inputs and outputs and on
the scheme of abstraction one selects (Kalke 1969). Are neurons, (p. 139) synapses, and
dendrites structural kinds and realizers, or are they functional kinds open to multiple
realizations? The answer will depend on our particular explanatory project. Should we
count a twenty‐amp fuse and a twenty‐amp electromagnetic circuit breaker as different
structures that realize the same functional kind? They both allow currents not exceeding
twenty amps to flow through the circuit, and block currents above twenty amps. Or are
they of different functional kinds because one burns through when overloaded and the
other is pulled back by magnetic attraction and can be reset? Such questions have no
absolute answer.
Page 13 of 28
Functionalism
Thus, the general functionalist point about realization remains. Indeed, realization is
more crucial than ever to understanding the mind‐body relation; it is the basic link that
ties together the many levels one needs to explain. As to more specific versions of
functionalism, their coherence will depend on the particular ways in which they spell out
their respective notion of function and on how successfully they draw the needed
distinctions at each relevant level. The multilevel aspect of functionalism also provides
the means to reply to a common charge of circularity lodged against functionalism.
In so far as the functionalist aims to explain key mental features such as intentionality,
rationality, or consciousness in terms of functional relations, she confronts a version of
the theoretician's dilemma. If she restricts herself to relations that do not presuppose
those very features, her model is likely to prove inadequate, but if they do involve such
features then circularity threatens. The functionalist confronts the general problem of
trying to explain or construct consciousness or intentionality from components that are
themselves devoid of those features, and the task may seem insuperable. How could
items devoid of content or phenomenal character satisfy the sorts of relations that seem
needed to realize intentional or conscious states. The gap between non‐mental realizers
and realized mental features seems too large.
Here the multilevel nature of functionalism is again crucial. The gap between non‐mental
realizers and sophisticated mental features such as consciousness and propositional
intentionality need not be bridged by a single level of realization, but might (p. 140)
instead be produced by a series of nested levels of realization, many of which involve
cruder, less fully realized versions of the critical mental aspects. The approach relies on
the assumption that the relevant features of mind, such as intentionality and content, can
occur in many forms, some less sophisticated than others, and that items with those
simpler forms might be combined in suitable systems to realize more sophisticated forms.
One need not jump from non‐contentful neurons that operate without any understanding
to full‐blown intelligent rational agents in a single step. Rather, one might realize simple
subsystems with only a crude and limited form of content or understanding and then
combine them into successively more complex levels with an increase in the type and
degree of intelligence and content at each ascending level (Dennett 1978).
Page 14 of 28
Functionalism
person whom they realize. Each of those sub‐personal modules or ‘agents’ would then in
turn be decomposed into systems of interacting sub‐subcomponents that are even less
capable, with the basic process being reiterated as many times as needed until all the
intelligence and understanding have been discharged at a level that can be fully realized
by purely non‐contentful components that require no understanding to perform their
roles—what Dennett (1978) has called the ‘army of idiots’. Whether the strategy can in
fact succeed for all the relevant mental features remains controversial, with special
reservations about those features associated with phenomenal consciousness. Even if
intelligence, understanding, and content admit of degrees in the way required by
homuncular functionalism, it is less clear that phenomenality does so. Nonetheless, it may
do so, and the multilevel approach in general gives the functionalist important theoretical
tools for bridging the gap between the mental and the non‐mental.
The overall status of functionalism as a theory of mind depends largely on its success in
dealing with the major features of mind. In so far as some supposed mental aspect is
indeed a real feature of mind, the functionalist should be able to explain it within her
general framework. Of course, not everyone agrees about which features of mind are real
and need to be accommodated. Some, for example, might deny the reality of free will or
the self. However, almost everyone agrees that functionalism needs to address the three
big Cs: content, causation, and consciousness. Indeed, many of the standard anti‐
functionalist objections appeal to its supposed inadequacy with regard to one of the
three.
There is, however, disagreement about what an adequate account would require. More
reductionist versions of functionalism may need to provide strict necessary and sufficient
functional conditions for the relevant concepts or to specify the essential nature of the
relevant properties. However, other more pragmatic versions of the (p. 141) view might
be content merely to show that functional considerations can provide important and
useful insight into describing and explaining those features of mind.
Content. Intentionality is a pervasive feature of the mental. Many, if not all, mental states
are about things and have content. It is difficult to imagine how anything could be a
perception, a memory, a belief or desire, or even an emotion or a sensation if it lacked all
content and was not about anything. An adequate theory of mind needs to provide both a
general account of how intentionality comes to exist and also some account of why
specific mental items have the particular intentional contents that they do. Moreover, it
needs to address not only what given states are about but how they conceptualize or
represent their objects. It thus must provide an account of the sense and modes of
presentation associated with mental states as well as of their objects and conditions of
satisfaction.
Giving such accounts would be a project far beyond the scope of this chapter. What
matters for present purposes is that virtually all the current major theories of
intentionality and content are in one way or another functionalist. There are many
disagreements in the intentionality literature—for example between individualist models
Page 15 of 28
Functionalism
and externalist ones or between causal‐role theories and teleological ones— but most of
them are ‘in house’ disputes; that is, disagreements about how to proceed within a
general functionalist framework.
Among the current models of intentional content are those that appeal to conceptual role,
causal history, informational semantics, nomic covariance, teleological function, and role
in adaptive behavioural guidance. There is no consensus as to which is best, and indeed
there is good reason to want a plurality of different ways of thinking about content for
various explanatory purposes. Nonetheless, they are all basically functionalist theories
that explain content in terms of the roles that mental items play within complex relational
networks.
Functionalism provides the basic structure for thinking about intentionality, and most of
the major controversies in the literature concern how to set various parameters within
that framework. Some limit the relevant networks to relations purely within the mental
agent, for example narrow conceptual‐role theories, while others extend the network out
into the physical, social, and historical world, as in various externalist or wide‐conceptual‐
role theories. Some accounts specify the roles in a purely causal or computational way,
while others appeal to teleological notions. Some models are holistic and appeal to a
large array of relations in fixing content; others fix content more locally on a smaller
relational base. Some take the primary bearers of content to be representations, and
others regard mental states as primary, but in either case the respective models of
content appeal to relational facts about the relevant items. Despite their disagreement
along these many parameters, all the major theories explicate content in terms of realized
roles within relational networks, and they are thus all functionalist in their basic
approach.
One might of course deny that any of the current theories is likely to provide an adequate
account of intentionality. But in so far as one has any optimism about the state of current
theorizing on the topic, one should regard that as a factor in favour of functionalism.
The causal status of the mental is more controversial. For example, some philosophers
deny that ordinary intentional explanations are causal (Dennett 1978; Bennett 1992).
They regard folk psychology as a heuristic interpretative structure that is useful in
predicting and making sense of complex patterns of behaviour; indeed, one that is
essential for our normal social interaction. Nonetheless, they deny that such explanations
are causal or that they appeal to mental states as inner causes. However, most
philosophers do regard at least many mental states and properties as causally relevant or
potent, and they fear that any account of mind that could not account for its causality
would collapse into an unacceptable form of epiphenomenalism; that is, the view that the
mental is causally inert.
Page 16 of 28
Functionalism
The price of functionalism according to such critics is the causal impotence of the mental.
To get real mental causality, they argue, one must go back to some form of identity
theory. Since the underlying physical properties are doing all the causal work, the only
way that mental properties could themselves be causal is if they are indeed the very same
properties. Thus, if the physicalist wishes to accommodate our strong intuition that
mental properties are causal, she must reject the functionalist's view of the mind–body
relation as a matter of realization between distinct properties, and go back to some form
of at least localized or species‐specific type‐type identity theory.
The problem of mental causation raises far more issues than can be addressed in this
chapter, but the functionalist has some reasonable basic responses that can be briefly
stated. Most importantly, it is far from obvious that real causality resides only at the
realizer level. This might be so for those versions of functionalism that regard functional
properties as abstract logical constructions, but many versions do not and instead treat
them as concrete systemic features. It is true that their reality depends upon their
underlying realizers, but the causal powers of the realized feature derive not only from
those of its realizers but also from the systematic way in which they are combined and
organized in order to realize that feature.
The amplifier in a stereo system really can boost the power of a signal while maintaining
its frequency profile, and one's immune system really can produce resistance to many
pathogens one encounters. Similarly, one's memory can keep track of where one left one's
keys and direct where one looks when one needs them. The abilities and powers of
immune systems and memory systems admittedly depend upon the more limited powers
of their many realizers. But that does not imply that such systems do not really have
those powers. They do indeed have them. And they do so because of the way in which
their specific realizers have been recruited and incorporated into the (p. 143) requisite
organized general pattern, which is often highly stable and self‐maintaining despite a
continuous interchange of realizers (Van Gulick 1993a).
Moreover, given that such features are open to a wide range of multiple realizations, it is
useful to classify them as of a common kind in virtue of their shared systemic causal
powers. Distinguishing among them on the basis of their differences in realization would
lead to a loss of descriptive and explanatory power; real causal regularities would be
hidden. Given the important patterns that persist at the higher level and that support
causal and counterfactual reasoning, there is real value in recognizing causally relevant
kinds at the system level (Fodor 1974).
Page 17 of 28
Functionalism
Perhaps this would be enough to accommodate our intuitions about the causal status of
minds and mental kinds. There might be a stricter more demanding sort of causal status
that is found only in ultimate realizers which do not derive their causal powers from any
more basic realizers. If so, mental features may lack that status, but that need not be an
embarrassment to the functionalist. There is little reason to believe that we have any
reliable intuitions about whether mental features are causal in that very demanding
sense, and thus the functionalist has no need to accommodate them. Moreover, mental
features turn out to be no worse in this regard than physiological, geological, or economic
features. None of them is causal in that ultimate sense either, and if mental features are
as causally robust as chemical and neural ones, then the functionalist need not be much
concerned with the threat of epiphenomenalism.
Consciousness
The largest number of anti‐functionalist objections concern its alleged inability to deal
with consciousness. Of course, what doing so requires will depend on which properties
consciousness really has. The functionalist may deny some of its traditional properties,
but at a minimum she must give an account of the basic difference between creatures or
mental states that are conscious and those that are not. In so far as conscious perceptions
or memories differ importantly from non‐conscious ones, functionalism must address that
difference and the general nature of state consciousness. Similarly, it needs an account of
creature consciousness and the distinctive mental lives of conscious beings.
Both distinctions—state and creature—get drawn in diverse ways, some of which are
easier for the functionalist to accommodate. States can be said to be conscious in the
access sense if their contents are reportable and broadly available for a rich range of
inference and application. They can alternatively be said to be conscious in the
phenomenal sense if there is an experiential feel associated with being in such a state,
some ‘way it is like from the inside to be in such a state’. The two senses may not always
coincide and the former access sense fits more easily into the functionalist framework,
since it is primarily a matter of relations and interactions among various contentful items.
By contrast, the phenomenal aspect of consciousness may seem more difficult and
resistant to the functionalist's approach.
(p. 144)
Page 18 of 28
Functionalism
a state one is aware of being in. That self‐awareness consists in having a higher‐order
thought (HOT) or higher‐order perception (HOP) directed at the relevant lower‐order
state. What makes a desire a conscious desire is the accompanying higher‐order state
directed at it. Thus, higher‐order theories also explicate consciousness in terms of
intentional relations. In so far as intentionality is itself to be explained functionally, such
theories ultimately explain consciousness in functional terms.
Yet other theories explain the difference between conscious and non‐conscious states as a
matter of degree of influence and interaction within the mind. On such models, a mental
state's being conscious is a matter of its cerebral celebrity (so called ‘fame in the brain’)
or of its content being ‘broadcast in a global workspace’ which makes it available for use
to a wide number of mental modules (Dennett 1991).
Thus, despite their differences, these various theories all take a basically functionalist
approach to consciousness. Nonetheless, the anti‐functionalist may deny that any of them
is really adequate. Even if most currently available theories are functionalist, that would
not count for much if none of them really does justice to the reality of consciousness.
Indeed, some philosophical arguments aim to show that no functionalist theory could ever
be adequate, because there are essential features of consciousness that inevitably fall
outside their scope. In particular, they aim to show by a priori means that functional
criteria alone are insufficient to capture the nature of consciousness in the phenomenal
sense and of the experiential qualia associated with it (Chalmers 1996).
(p. 145)
Page 19 of 28
Functionalism
Inverted‐qualia cases involve the possibility of two individuals, call them Norm and Flip,
who are totally alike in their functional organization but experience different qualia when
they perceive a given sort of object. When they look at lime, they both call it ‘green’, and
the inner states produced play the same functional role in each of them, but the quale
associated with that state in Flip is the one associated with the states produced in Norm
by ripe tomatoes. However, if their respective qualia have always been as they are, the
difference between them would be undetectable and would make no functional
difference. Thus, the critic concludes that functionalism fails to capture a real mental
difference between Norm and Flip and in general fails to capture the nature of specific
qualia.
Absent‐qualia arguments make the even more radical claim that any set of functional
conditions can be realized by systems which possess no qualia at all; that is, by systems
for which there is nothing that it is like to be them. Such a realization, call it Zip, would
be in effect a functional zombie, that is, a system that was like conscious beings in every
functional respect but which nonetheless was not itself conscious in the phenomenal
sense. If Zip were indeed possible, functionalism might be faulted for failing to capture
not merely the nature of specific qualia but even the general notion of qualia, and thus
failing completely as an account of phenomenal consciousness.
The supposed possibility of both inverted‐ and absent‐qualia cases is supported largely by
appeal to intuitions about what can be imagined or conceived. The anti‐functionalist
asserts that given any functional specification, it is always possible to imagine its being
realized in ways that invert or lack qualia. These intuitions about what is imaginable are
taken to show that they are conceivable in a sense that implies that they are at least
logically possible. However, functionalists have challenged both the possibility of such
imagined cases as well as the conclusions drawn from them.
The coherence of inverted‐qualia cases has been challenged both on a priori conceptual
grounds and on empirical evidence about the structure of our sensory colour space. On
the conceptual side, doubts have been raised about whether any real sense can be
attached to the idea of intersubjective differences between essentially private properties
(Wittgenstein 1953). How could be there ever be any evidence or criterion for verifying
that there was or was not such a difference, since by stipulation normal and inverted
individuals are supposed to be the same in all behavioural and functional respects? And if
no verification procedure is even possible, the inverted‐qualia hypothesis may be in
danger of becoming a meaningless pseudo‐possibility (Dennett 1991).
The conceptual coherence of inverted qualia has been defended by appeal to the
possibility of intrasubjective step‐by‐step partial inversions that would together result in
an overall inversion (Shoemaker 1982). Each partial step would be detectable. (p. 146)
However, the organization and behaviour of the person at the end of the series would be
functionally equivalent to that at its start despite the fact that the specific qualia playing
Page 20 of 28
Functionalism
the relevant roles were not the same. In so far as each of the partial shifts is coherent, so
too, it might seem, must be their sum.
Qualia inversions may thus seem possible for those who regard qualia as intrinsic
properties of perceptual states. However, those who equate qualia with representational
contents will deny the absent‐qualia possibility (Harman 1990; Dretske 1995). On such a
view, a state's qualitative character or what‐it‐is‐likeness consists in how it represents the
world as being. Most representationalists also hold that a state's content depends upon
its functional relations to items in the world; for example, on what objective properties it
reliably tracks. Thus, if all those relations were preserved, its content would remain
unchanged and so too would its qualitative feel. So given a relational‐representational
view of qualia, functionally identical qualia inversions are impossible.
Moreover, even if the functionalist concedes that inverted‐qualia cases are possible, the
consequences will depend on what her theory must explain to count as a success. If
inverted qualia are possible, then functional conditions may suffice to identify only an
equivalence class of qualia rather than a specific quale. Nonetheless, if those conditions
could be satisfied only by creatures or systems with some qualia or other, that might
suffice as a basic account of qualia and phenomenal consciousness.
Absent‐qualia cases would pose a greater problem for the functionalist. Again, many
functionalists have denied that they are possible, although their reasons for doing so have
varied, depending in part on which version of functionalism they endorse.
For the computational functionalist, the question is whether robots that shared all our
computationally specifiable functional organization might nonetheless be such that there
was nothing that it was like to be them. The causal‐role functionalist might ask whether
any system could have a state P that played all the causal roles associated with being in
pain but that nonetheless lacked any felt quale of hurtfulness. Could there be a pain
zombie?
Some functionalists have argued that nothing that was not a pain could play the full role
of pain. In particular, they have argued that non‐pains could not typically give rise to
beliefs about being in pain, at least not if one accepts a plausible (p. 147) functionalist
Page 21 of 28
Functionalism
Absent‐qualia advocates might reject such a move as circular, since it relies on beliefs
that depend on qualitative states for their identity. Whether the functionalist may do so
will depend on what resources he is allowed to invoke in specifying causal roles. There
will be a trade‐off between sufficiency and explanatory power. The more the functionalist
relies on quasi‐qualitative notions in specifying causal roles, the better his chance of
ruling out non‐qualia realizations, but also the less his prospects for explaining qualia in
non‐qualitative terms (Chalmers 1996).
As with phenomenal similarity and qualitative belief, the functionalist confronts a general
dilemma concerning which concepts she can use in specifying functional roles. Whether
those functions are thought of computationally, causally, or teleologically, there is the
further issue of the degree to which qualitative or phenomenal concepts can be implicitly
assumed in specifying the relevant relations and interactions. To the extent that they are
not relied on, absent‐qualia realizations are hard to exclude, but to the degree they are
assumed, circularity looms as a danger.
The possibility of absent qualia might also be attacked by the causal‐role functionalist on
the grounds that such cases implicitly assume that qualia are epiphenomenal; that is, that
they lack any causal efficacy. If Zip's states play the same causal roles as Norm's despite
the fact that they lack any qualia, then it would seem that Norm's qualia make no causal
difference to his states (Shoemaker 1975a; Kim 2005). But that seems implausible. Does
the hurt one feels when one bangs a shin not cause one to rub it, and could the yellow
quale associated with one's visual experience of a banana really play no part in why one
answers ‘yellow’ when asked to describe its colour? If absent‐qualia cases entail such
negative causal results, that would be a strong reason against them.
Page 22 of 28
Functionalism
The dispute again turns on the issue of how to specify roles. Do ersatz red qualia and
genuine red qualia play the same role in different ways; that is, are they alternative
realizers of one and the same role? Or do the roles they play differ in ways that matter to
the functional classification, as the critic of absent qualia will argue? How could a state
without any real qualia affect one's beliefs about one's perceptual experience in the same
intimate way that one's genuinely qualitative perceptions do (Van Gulick 1993b)? Even if
Zip says he is seeing red, the belief he expresses cannot have been produced by the same
evidential link to his perceptual states, because those states have no red qualia of the sort
his beliefs supposedly refer to. Defenders of absent qualia, on the other hand, will dismiss
any such differences as insignificant variations in how the very same role is realized in
Zip and Norm.
Thus, the possibility of absent‐qualia cases will depend on whether the functionalist can
specify the relevant roles in a way that is non‐circular but nonetheless identifies roles
that could be filled only by states with genuine qualia. In so far as the absent‐qualia
proponent aims his attack not just at present functionalist models but at the future
prospects of the basic approach, he needs to show that the functionalist cannot succeed
in specifying such roles, and that is very hard to do. To simply insist that any roles
specified in the future will always be open to absent‐qualia realizations would seem to
beg the question against the functionalist programme.
In reply, some computational functionalists have denied that such weird systems could
satisfy the relevant computational conditions, especially if real‐time constraints were
included (Dennett 1991); they would for example be far too slow. Others have simply
replied that such bizarre realizations would be phenomenally conscious, strange though
they would be (Churchland 1985). Non‐computational functionalists regard such
Page 23 of 28
Functionalism
Functionalism faces another possible challenge, from those who view qualia as intrinsic
properties of internal mental states or objects. The functionalist defines mental kinds in
terms of the roles they play within a systematic network of interrelated states and
processes, and such functional kinds seem to be paradigmatically relational. How could
any such relational kinds be identical with qualia, if the latter (p. 149) are intrinsic
properties of mental states? In so far as what is intrinsic is by definition non‐relational,
any functionalist account of intrinsic qualia would seem straightforwardly inconsistent
(Kim 2005).
Functionalists might reply in at least two ways. First, they could reject the view of qualia
as intrinsic properties of mental states. For example, some representationalists (Tye
1995) identify qualia with the contents of our perceptual states. They argue that such
contents are the only mental features to which we have introspective access, and it is
those contents that determine what is it like to be in such a state. What it is like to see a
yellow banana is just to be in a visual state whose content is that there is a yellow banana
in front of me. However, if a state's content is determined largely, if not wholly, by causal
and informational links it bears to items in the world, then that content is a relational
property of the state rather than an intrinsic one. If so, the relational nature of
functionalism poses no difficulty in dealing with qualia.
Functionalism in one version or another remains the dominant view of mind among
philosophers. The general functionalist framework provides important tools for dealing
with the mind–body problem, intentionality, mental causation, and even consciousness. It
has significant critics, but none of the objections that have been raised against it seems
decisive, and functionalists have made plausible responses in every case. Debate will no
doubt continue.
References
Page 24 of 28
Functionalism
(p. 150) Cummins, R. (1975), ‘Functional Analysis’, Journal of Philosophy, 72: 741–64.
Davidson, D. (1987), ‘Knowing One's Own Mind’, Proceedings and Addresses of the
American Philosophical Association, 60: 441–58.
Kalke, W. (1969), ‘What is Wrong with Putnam and Fodor's Functionalism’, Noûs, 3: 83–
94.
Page 25 of 28
Functionalism
—— (1980), ‘Mad Pain and Martian Pain’, in N. Block (ed.), Readings in the Philosophy of
Psychology (Cambridge, Mass.: Harvard University Press), 122–32.
Lindsay, P., and Norman, D. (1977), Human Information Processing (New York: Academic).
Lycan, W. (1981), ‘Form, Feel and Function’, Journal of Philosophy, 78: 24–50.
Place, U. T. (1956), ‘Is Consciousness a Brain Process?’, British Journal of Psychology, 47:
44–50.
Putnam, H. (1965), ‘Brains and Behaviour’, in R. J. Butler (ed.), Analytical Philosophy, 2nd
ser. (Oxford: Blackwell), 1–19.
—— (1967a), ‘The Nature of Mental States’, in W. Capitan and D. Merril (eds.), Art, Mind
and Religion (Pittsburgh, Pr.: University of Pittsburgh Press), 37–48.
Ryle, G. (1949), The Concept of Mind (New York: Barnes & Noble).
Searle, J. R. (1982), ‘The Myth of the Computer: An Exchange’, New York Review of
Books, 29/11: 56–7.
Page 26 of 28
Functionalism
(p. 151) Smart J. J. C. (1959), ‘Sensations and Brain Processes’, Philosophical Review, 68:
141–56.
Stich, S. (1983), From Folk Psychology to Cognitive Science: The Case Against Belief
(Cambridge, Mass.: MIT Press).
—— (1993a), ‘Who's in Charge Here and Who's Doing All The Work?’, in J. Heil and A.
Mele (eds.), Mental Causation (Oxford: Oxford University Press), 233–56.
Page 27 of 28
What is Property Physicalism?
Since the days of Democritus, Plato, and Aristotle the main concern of philosophers of
mind has been whether there is an independent realm of the mental beyond the realm of
the physical. This question has mainly been understood as that of whether there are
independent mental substances — souls or selves. Most of what has been written in the
philosophy of mind during the last seventy to eighty years, however, has instead been
concerned with the question of whether there are independent or irreducible mental
properties. To many it has seemed obviously true that there are no mental substances,
but with properties things are different. The history of answers to questions such as ‘Is
being in pain or thinking about Paris really something physical?’ (including the answers
associated with logical behaviourism, the identity theory, functionalism, and
supervenience theories) has been told time and again. Nonetheless, it is worth taking a
new look at it.
Keywords: mental substances, philosophy of mind, mental properties, logical behaviourism, identity theory,
functionalism
Page 1 of 26
What is Property Physicalism?
SINCE the days of Democritus, Plato, and Aristotle the main concern of philosophers of
mind has been whether there is an independent realm of the mental beyond the realm of
the physical. This question has mainly been understood as that of whether there are
independent mental substances—souls or selves. Most of what has been written in the
philosophy of mind during the last seventy to eighty years, however, has instead been
concerned with the question of whether there are independent or irreducible mental
properties. To many it has seemed obviously true that there are no mental substances,
but with properties things are different. Is being in pain or thinking about Paris really
something physical? Do we have sufficient reason to believe that all mental properties are
physical properties? However, it was not only the truth of property physicalism—the claim
that all mental properties are physical properties—that was at issue. Most of the debate
has rather been concerned with the question of what property physicalism amounts to.
What has to be the case in order to make property physicalism true? The history of
answers to this question (including the answers associated with logical behaviourism, the
identity theory, functionalism, and supervenience theories) has been told time and again.
Nonetheless, it is worth taking a new look at it.
Page 2 of 26
What is Property Physicalism?
Of course the answer to this question essentially depends on what concept of property
one endorses. Rudolf Carnap, whose articles (1932) and (1932/1933) rank among the
most important early works on property physicalism, held the view that properties are the
senses or intensions of predicates.2 For this reason, according to Carnap, two predicates
stand for the same property just in case they have the same sense; that is, just in case
they are synonymous (see also Hempel 1949). For Carnap property physicalism amounts
to the claim that for each mental predicate there is a synonymous predicate in physical
language. Or, in a nutshell: property physicalism requires synonymy.
Carnap, however, not only held this view of property physicalism. He also thought that
property physicalism in this sense is true. In his view, for example, the predicate ‘x is
excited’ is synonymous with the expression ‘x’s body (especially his nervous system) has a
physical structure that is characterized by a high pulse and rate of breathing, by
vehement and factually unsatisfactory answers to questions, by the occurrence of
agitated movements on the application of certain stimuli, etc.’ (1932/1933: 170 ff.). Why
did Carnap assume these predicates to be synonymous? Because, according to the
verificationist theory of meaning Carnap endorsed in the early thirties, predicates are
synonymous if and only if they are applied on the basis of the same observations and
because this, in Carnap's view, is true of the two predicates mentioned. Indeed, he
thought the second predicate simply comprises an enumeration of these observations.
This view, which has become known by the somewhat misleading name ‘logical
behaviourism’, however, is untenable. As soon as the verificationist account of meaning
was abandoned it became very clear that it is impossible to define the term, say, ‘wants a
beer’ without including mental terms in the definiens. If someone wants a beer she will go
to the refrigerator—but only if she believes that there is beer in the fridge, that she will
not be shot if she goes to the fridge, etc. That someone takes an umbrella may be a sign
that she believes that it is raining—but only if she doesn't take the umbrella in order to
Page 3 of 26
What is Property Physicalism?
hide from curious gazes. Generally speaking, it turned out to be impossible to define
mental predicates in physical language in a non‐circular way.
(p. 154)
This finding, however, did not mark the end but rather the beginning of what today is
known as ‘identity theory’. At the end of the 1950s Place and Smart argued that
statements like ‘The temperature of a gas is identical to the mean kinetic energy of its
molecules’, ‘Lightning is an electrical discharges’, and ‘Water is H2O’ are perfectly true
identity statements though ‘temperature’ and ‘mean kinetic energy’—or for that matter
‘lightning’ and ‘electrical discharge’ or ‘water’ and ‘H2O’—are by no means synonymous
(see Place 1956; Smart 1959). Nonetheless, ‘temperature’ and ‘mean kinetic energy’
stand for the same property—as physics has shown us.3 It is therefore at least possible
that empirical science will arrive at the result that even ‘pain’ and ‘C‐fibre firing’ stand
for the same property, though they are not synonymous. Place and Smart thus held the
view that property physicalism amounts to the claim: Each mental property is identical to
a physical property even if the corresponding predicates are not synonymous. Or, in other
words: property physicalism does not require synonymy, just identity. But if this is the
case, we are again confronted with the question: How do we find out that the mental
predicate ‘M’ stands for the same property as the physical predicate ‘P’ if it is not by
examining whether ‘M’ and ‘P’ are synonymous?
In his seminal paper (1983) Joseph Levine made an important observation that has
become the subject of much debate. Levine starts by comparing the following two
property‐identity statements:
(1) The temperature of a gas is identical with the mean kinetic energy of its
molecules.
and
(2) Pain is identical with the firing of C‐fibres.
According to Levine, statement (1) is fully explanatory while statement (2) is not. The
explanatory character of (1) rests on two facts:
(a) Our concept of temperature is exhaustively captured by the causal role of
temperature.
(b) Physics can render it intelligible that precisely this causal role is played by the
mean kinetic energy of the molecules of a given gas.
In other words, on Levine's view statement (1) is fully explanatory on the grounds that
temperature is reductively explainable in terms of the mean kinetic energy of the molecules
involved. Reductive explanation is a two‐step procedure. To show that property F is reductively
explainable in terms of property G you will have to, first, give an analysis of F and, second,
demonstrate that it follows from the basic laws of nature that all objects possessing property G
satisfy this analysis.
Page 4 of 26
What is Property Physicalism?
What is crucial is that Levine at least may be understood to claim that the truth of
property‐identity statements of the form ‘F = G’ depends on whether it is possible
(p. 155) to supply an explanatory reduction of F to G.4 If this were the case there would be
It is noteworthy that these considerations are in perfect accordance with what early
exponents of the identity theory thought about the truth of non‐analytic identity
statements. The generally accepted answer to the question why it should be supposed
that the temperature of a gas is identical to the mean kinetic energy of its molecules was:
‘Because classical thermodynamics can be reduced to statistical mechanics’. It is easy to
see that the basic idea is the same. First, on the basis of a semantics of theoretical terms
widely recognized in those days, people assumed that expressions like ‘temperature’ are
implicitly defined by the laws in the formulations of which they occur. Thus, the
expression ‘temperature’ was supposed to be implicitly defined by the laws of classical
thermodynamics; the meaning of ‘temperature’ consists in the causal role described by
those laws. Second, from the reducibility of classical thermodynamics to statistical
mechanics it follows that for all laws of classical thermodynamics it is possible to derive
image laws from statistical mechanics (see Beckermann 2001a: 107–8); that is, laws
showing that at the level of statistical mechanics there is a property (i.e. mean kinetic
energy of molecules) playing exactly the causal role characteristic of temperature. In
other words, since the meaning of the expression ‘temperature’ is implicitly defined by
the laws of thermodynamics, it follows from the reducibility of classical thermodynamics
to statistical mechanics that it is possible to give an explanatory reduction of the
temperature of a gas to the mean kinetic energy of its molecules. From the very
beginning the opinion prevailed that there is a very close connection between the identity
of properties and their reductive explainability.
Page 5 of 26
What is Property Physicalism?
When C. D. Broad developed his theory of emergence in the early 1920s his aim was to
create room for a third position mediating between these two extremes—a position he
called emergent vitalism.
Broad's first step was to point out that the problem of vitalism is only a special case of a
much more general problem—the problem of how the properties of complex systems are
related to the properties and the arrangement of their physical parts. Regarding this
question, there are in principle only two basic types of response. One can hold the view
that the properties of a complex system S can or cannot be explained by referring
exclusively to its physical parts and their arrangement. But even if such an explanation is
possible, there are two further possibilities—the properties may be either mechanistically
explainable or emergent. In sum, the difference between emergent and mechanistic
theories can be explained as follows:
Put in abstract terms the emergent theory asserts that there are certain wholes,
composed (say) of constituents A, B, and C in a relation R to each other; that all
wholes composed of constituents of the same kind as A, B, and C in relations of
the same kind as R have certain characteristic properties; that A, B, and C are
capable of occurring in other kinds of complex where the relation is not of the
same kind as R; and that the characteristic properties of the whole R(A, B, C)
cannot, even in theory, be deduced from the most complete knowledge of the
properties of A, B, and C in isolation or in other wholes which are not of the form
R(A, B, C). The mechanistic theory rejects the last clause of this assertion.
Elsewhere (Beckermann 2000) I have tried to analyse this passage in detail and to show that it
ultimately amounts to the following two definitions:
(ME) A macro‐property F of a complex system S consisting of parts C 1, …, C n
arranged in way R, i.e. of a system with micro‐structure [C 1, …, C n; R], is
mechanically explainable if, and only if, it follows from the general laws of nature
holding for the component parts C 1, …, C n, and suitable bridge laws,5 that all
objects with micro‐structure [C 1, …, C n; R] possess all features characteristic of F.
(E) A macro‐property F of a complex system S with microstructure [C 1, …, C n; R] is
emergent if, and only if, the following holds:
(a) It is a true law of nature that all systems with micro‐structure [C 1, …, C n; R]
have F; but
(p. 157)
(b) it does not follow from the general laws of nature holding for component
parts C 1, …, C n, and suitable bridge laws that all objects with microstructure [C
1, …, C n; R] possess all the features characteristic of property F.
Page 6 of 26
What is Property Physicalism?
If this reading is correct, two points become clear straightaway. First, Broad agrees that
mechanical explanation is a two‐step affair. Second, the two steps of Broad's account exactly
correspond to the steps of Levine's account. The first task is to discover the analysis of the
property F in question; that is, to find out the features characteristic of F.6 (For Broad, as a rule
these features are characteristic forms of behaviour; for instance, the characteristic forms of
behaviour of objects with the property of being magnetic.) And after that has been done, the
task consists in showing that it follows from the general laws of nature that objects with a given
microstructure possess the features characteristic of F. Thus, according to Broad, an advocate of
property physicalism has to claim that mental properties are not emergent but mechanically—or,
as we might say today, reductively—explainable.
Page 7 of 26
What is Property Physicalism?
It is, therefore, not very surprising that in recent years all attempts at formulating criteria
for the identity of properties have been subjected to a fundamental kind of criticism. In
effect, this criticism amounts to the claim that identity is a relation which does not admit
of any further analysis. Either properties are identical, or they are not. There is no
instructive answer to the question ‘Why is it that F and G are identical?’. And that is why
there are no criteria that properties would have to satisfy for being identical. The only
admissible question is how to find out whether properties are identical; that is, how to
find out whether ‘F’ and ‘G’ stand for the same property.
This position has been urged with special emphasis by David Papineau (1998).7 In his
view, every proponent of property physicalism has to endorse the identity theory. (p. 158)
Papineau says that his ‘first task is to show that physicalism is best conceived as a thesis
about property identity’ (1998: 374). But Papineau goes on to argue that the identity
theory may be true even if it is not possible to reduce mental properties to physical ones.
Identities either obtain, or they don't. It makes no sense to ask why it is that objects or
properties are identical. And for this reason the question of how P manages to give rise to
M does not even make sense for an identity theorist. Identical properties simply are
identical, they do not give rise to each other. Nothing gives rise to itself.
Objections similar to those mentioned by Papineau are raised by Block and Stalnaker in
their criticism of the assumption that physicalists are committed to the claim that mental
properties are reductively explainable. In Block and Stalnaker (1999) they argue that this
could not possibly be the case. They hold that reductive explainability presupposes that a
phenomenon F requiring explanation can be analysed in such a way that the only
concepts employed in this analysis are those which also occur in our general laws of
nature. According to Block and Stalnaker, however, this is not a general possibility,
especially if we are dealing with mental phenomena. Consequently, as a rule, attempts at
reductive explanation will come to grief. This, they insist, is no argument against
physicalism. After all, a physicalist is merely committed to an assertion of identity; and
mental properties may be identical with physical properties even if it is not possible to
give a reductive explanation. Generally speaking, talk of criteria of identity is highly
misleading. Instead of asking what conditions must be satisfied for property F to be
identical with property G, we should ask what reasons justify the claim that F = G.
Page 8 of 26
What is Property Physicalism?
In sum, the position put forward by Papineau, as well as by Hill, McLaughlin, and Block
and Stalnaker, amounts to the following: (1) Identity has nothing to do with reductive
explainability. In fact, if F and G are identical, you can't reduce one to the other; nothing
reduces to itself. (2) The only thesis physicalists are committed to is that mental
properties are identical with physical properties (see also Hill 1991; Hill and McLaughlin
1999; McLaughlin 2001).
With regard to their first claim, I think Hill, McLaughlin, Papineau, and Block and
Stalnaker are right. In early debates on the identity theory two quite different ideas were
confounded—the idea of identity and the idea of reductive explainability. It should be
noted that reductive explainability is compatible with multiple realizability; therefore it is
not a sufficient condition for identity. However, it is not even a necessary condition. Water
is a good example to elucidate this claim.
Levine (1993) argues that though the truth of ‘water = H2O’ cannot be known a priori, it
is indeed not conceivable that H2O should fail to manifest the characteristic macro‐
properties of water. For on the basis of the basic laws of nature we can deduce that H2O
boils at 212 °F at sea level, that H2O is a liquid, that H2O is transparent, etc. (Levine
1993: 128–9). We have already seen that, in his view, this marks an important epistemic
difference between the claim that water = H2O and the claim that pain = C‐fibre firing,
since it is not in the same sense inconceivable that a person's C‐fibres are firing while at
the same time the person does not feel any pain. But independently of Levine's view it is
an important question whether it is a necessary (p. 159) condition for the truth of ‘water
= H2O’ that we can deduce that H2O possesses the characteristic macro‐properties of
water.
Block and Stalnaker have argued that it is not. If we try to find out whether ‘water =
H2O’ is true, we, rather, rely on considerations like the following. We know that heating
causes water to boil. Furthermore, science tells us why increasing the mean kinetic
energy of H2O molecules leads to a certain activity M of these molecules. If we assume
that water = H2O, heat = mean molecular kinetic energy, and boiling = molecular activity
M we, therefore, ‘have an account of how heating produces boiling … Identities allow a
transfer of explanatory and causal force not allowed by mere correlations. Assuming that
heat = mke, … etc. allows us to explain facts that we could not otherwise explain. Thus,
we are justified by the principle of inference to the best explanation in inferring that
these identities are true’ (Block and Stalnaker 1999: 23–4). Assuming that water = H2O,
heat = mean molecular kinetic energy, etc. leads to a simpler and more coherent picture
of the world. This is what justifies these identity statements. It is not necessary that we
are able to deduce that H2O possesses the characteristic macro‐properties of water.
In my view, the answer to the question of how we justify the claim that water = H2O is
much simpler. We start from the background assumptions that water is a chemical
substance and that chemical substances are individuated by their molecular structure. To
find out whether water = H2O we, therefore, do nothing but collect some samples of
Page 9 of 26
What is Property Physicalism?
water, bring them into a chemical lab, and ask the chemists to analyse the molecular
structure of the samples. If the answer is that the molecular structure is the same in all
cases, namely H2O, we have a result: water = H2O.
However that may be, that reductive explainability is not a necessary condition for
identity is shown by the following consideration. Assume that chemists tell us that the
molecular structure of all submitted samples of water is H2O. And assume further that it
is not possible to deduce from the fundamental laws of nature that H2O boils at 212 °F at
sea level, that H2O is a liquid, that H2O is transparent, etc. What should we say in this
situation? In my view, the answer is obvious. Water is still H2O, but all surfaces properties
of water are emergent in Broad's sense. This may be incredible, but it is not impossible.
Thus, we have to keep apart the idea of identity and the idea of reductive explainability
(in Broad's sense).8 And we should carefully distinguish the view that physicalism
amounts to the claim that all mental properties are identical to physical properties and
the view that physicalism is true if all mental properties (p. 160) can be reductively
explained in terms of physical properties. But isn't physicalism committed to the claim
that everything is physical? Why should we think that the second view actually is a
version of physicalism?
Page 10 of 26
What is Property Physicalism?
In reaction to this argument functionalism was developed: the claim that each mental
state or property is characterized by a certain causal role. Functionalism maintains that,
say, pain is characterized by the fact that it is caused by tissue damage, that it causes
wincing and pain‐relieving behaviour, that it decreases the capacity to concentrate, etc.
According to one version of functionalism,9 ‘is in pain’ refers to just this causal role.
According to another (Lewis's) version, ‘is in pain’ works rather as a definite description
—in each case it stands for just that state that realizes the causal role characteristic of
pain in the being in question. Neither variant of functionalism, however, is a version of
property physicalism, for neither implies that the states that realize the causal roles in
question must be physical states.10 Thus, functionalism as such is neutral with regard to
ontology.11
Out of functionalism, however, there grew a real alternative to the identity theory: the
theory of supervenience. The basis of this theory is the idea that property physicalism
might be true even if mental properties are not identical with physical properties—
provided the realm of the mental is in an ontological way dependent on the realm of the
physical, provided, that is, all mental facts are ontologically (p. 161) determined by the
physical facts. The concept of supervenience was introduced to spell out this basic idea.
Page 11 of 26
What is Property Physicalism?
For purposes of clarification he adds that ‘a minimal physical duplicate of our world is a world
that (a) is exactly like our world in every physical respect (instantiated property for instantiated
property, law for law, relation for relation), and (b) contains nothing else in the sense of nothing
more by way of kinds or particulars than it must to satisfy (a)’ (1998: 13).
Taking these considerations into account, we now have an answer to the question of why
the view that property physicalism requires reductive explainability should count as a
version of physicalism. Reductive explainability implies supervenience. If all mental
properties are reductively explainable in terms of physical properties, a complete
description of the physical world analytically implies any statement about the distribution
of mental properties and relations. But identity also implies supervenience. If mental
properties are (identical with) physical properties there can, of course, be no difference
in the distribution of mental properties without a difference in the distribution of physical
properties. Thus, we seem to be confronted with two options: supervenience by reductive
explanation or supervenience by identity. At least, to my knowledge no other versions of
the supervenience theory have ever been seriously advocated. For this reason the
supervenience theory does not lead to a new answer to the question of what property
physicalism amounts to.
It is commonly agreed that philosophers who endorse the first option (‘a priori
physicalists’) are at least committed to the following claim: If we knew all physical facts,
we could infer solely from this knowledge without any further empirical research which
things have which mental properties and which things stand in which mental relations to
each other. To put it a little more precisely, a priori physicalists are committed to the
claim:
Page 12 of 26
What is Property Physicalism?
It is certainly a semantic fact that the expressions ‘Mark Twain’ and ‘Samuel Clemens’
are rigid designators; that is, they refer to the same objects in all possible worlds.15 And it
is also a semantic fact that ‘Mark Twain’ and ‘Samuel Clemens’ actually refer to the same
object—the writer Mark Twain. Thus, everyone who knows that ‘Mark Twain’ and ‘Samuel
Clemens’ are rigid designators which refer to the same object in the actual world is able
to infer solely from this knowledge that ‘Mark Twain = Samuel Clemens’ is necessarily
true. And the same holds for the proposition ‘Water is H2O’. Everyone who knows that the
expressions ‘water’ and ‘H2O’ designate the same chemical substance in the actual world
and that these expressions also are rigid terms standing for the same substances in all
possible worlds is able to infer solely on this basis that ‘water is H2O’ is necessarily true.
Thus, that (p. 163) the truth of propositions like ‘Mark Twain = Samuel Clemens’ and
‘water is H2O’ is not knowable a priori is not because these propositions are made true by
concealed modal facts that can be known to hold only a posteriori. What makes them true
is simply the fact that every object and every chemical substance is identical to itself in
each possible world, and to know this we do not have to do any empirical research.
Rather, that the truth of the propositions in question is not knowable a priori is because
even competent speakers can know some of the crucial semantic properties only on the
basis of empirical research.16
The fact that even competent speakers do not know all the semantic properties of the
expressions of their language was noticed by Frege, who insisted that true identity
statements of the form ‘a = b’ may be valuable extensions of our knowledge. According to
Frege, the reason for this is that a competent speaker has to know the sense, but not the
reference, of the expressions of her language.17 It may be the case that competent
speakers of English will have to know that the expressions ‘evening star’ and ‘morning
star’ designate heavenly bodies, but they need not know which heavenly body they
designate, and consequently they need not know that they designate the same planet.
Similarly, it may be that competent speakers of English have to know that the word
‘water’ stands for a certain chemical substance, but they need not know exactly what
substance that is, and hence they need not know that, as a matter of fact, ‘H2O’ stands for
the same substance.
Page 13 of 26
What is Property Physicalism?
With regard to the view of Hill, McLaughlin, Papineau, and Block and Stalnaker, the
upshot of these considerations is this. They claim that physicalism can be true only if all
mental properties are identical with physical properties. And they interpret property‐
identity statements on the basis of the model of object‐identity statements. When identity
theorists claim, for example, that pain is C‐fibre firing they, of course, do not mean to say
that there are two different properties pain and C‐fibre firing which, as a matter of fact,
are identical. (Different properties are never identical.) Rather, they mean that the
property expressed by ‘pain’ is identical with the property expressed by ‘C‐fibre firing’.
And since everything is identical to itself and to (p. 164) nothing else, this is the case if
and only if ‘pain’ and ‘C‐fibre firing’ stand for the same property.18 If we knew which
properties (or property) ‘pain’ and ‘C‐fibre firing’ stand for, we, therefore, could know a
priori whether ‘pain = C−fibre firing’ is true. Yet even competent speakers of English do
not know this semantic fact without empirical research. And for this reason the truth of
‘pain = C−fibre firing’ can be known only a posteriori.
Or so it seems, since things are in fact even more complicated. The initial question was
not what we could know a priori, but what we could derive a priori from the complete
knowledge of all physical facts. And, at least at first sight, on the basis of this knowledge
we could infer quite a lot of (if not all) semantic facts. Take ‘water’. Competent speakers
know that the chemical substance we call ‘water’ is the stuff that we find in rivers and
lakes, that falls from the clouds when it is raining, that comes from taps, etc. Obviously, it
is a physical fact that this stuff is H2O. From this together with the fact that ‘H2O’ is the
canonical name of H2O we can infer that ‘water’ and ‘H2O’ stand for the same substance.
Chalmers and Jackson have maintained that this result can be generalized (see especially
Chalmers 2002; Jackson 2003). Each name and each predicate, they argue, has a
descriptive content—a set of features which, as competent speakers know, the object
referred to by the name (or the substance expressed by the predicate) possesses in the
actual world. Furthermore, if the features are physical features, it is a physical question
which object or which substance actually possesses these features. On the basis of the
complete knowledge of all physical facts we must, therefore, be able to tell in this case
which object the name refers to or which substance is expressed by the predicate. A
posteriori physicalists have to claim that even if we did know all the physical facts we
would have to do more empirical research in order to find out whether ‘pain’ and ‘C‐fibre
firing’ stand for the same property. But what kind of research could this be? If we really
know all the physical facts, it after all seems to be an a priori affair to find out whether ‘F
= G’ is true; that is, whether ‘F’ and ‘G’ stand for the same property.
Page 14 of 26
What is Property Physicalism?
In the last section we saw that among identity theorists there are two factions. Members
of the first faction argue that if identity statements like ‘pain = C−fibre firing’ are true
their truth must follow a priori from a complete description of the physical world.
Members of the second faction deny this claim. In their view, these identity statements
may be true even if their truth does not follow a priori from a complete (p. 165)
description of the physical world. However that may be, with regard to the question of
what property physicalism amounts to we are finally left with the alternative: physicalism
requires identity versus physicalism requires reductive explainability in Broad's sense.
Let us assume that advocates of a posteriori identity will be able to cope with the
argument from multiple realizability.19 Which arguments can be adduced to support one
or the other of these two alternatives?
Jaegwon Kim (2005) also starts from the observation that with regard to the problem of
property physicalism there are two camps of philosophers. On the one hand, there are
philosophers like the British emergentists, who hold that property physicalism can be
true only if mental properties can be reductively explained in terms of physical
properties. On the other hand, there are identity theorists, who deny the indispensability
of reductive explanations. British emergentists were seriously worried about questions
like ‘Why does pain, not itch or tickle, arise from C‐fibre firing?’, ‘Why and how does
conscious experience arise from certain other neural states?’ (see e.g. Kim 2005: 94).
Block and Stalnaker, on the other hand, argue that these questions are misguided.
Identities cannot be explained. Properties are identical, or they are not. There is no
instructive answer to the question, ‘Why is it that F and G are identical?:
Identities [as Block and Stalnaker argue] should be seen not as helping to answer
explanatory questions like ‘Why is Jones conscious whenever pyramidal cell
activity is going on in the brain?’ but rather as neutralizing or dissipating them—
that is, as showing that there is nothing here to be explained. Why is there water
just where and when H2O is present? Why do the two correlate? The appropriate
answer here is this: Water is just H2O, and there is here no correlation to be
explained.
Kim argues that the emergentists' why‐questions have such prima facie urgency that Block and
Stalnaker have to have very good arguments in order to overcome them. And this, he adds, is
not the case. To my mind, however, there is a more direct way to show that the emergentists are
right and Block and Stalnaker wrong. The reason why Block and Stalnaker's attempt fails is that,
as I am going to argue, mental predicates belong to kinds of predicates that simply do not fit the
identity theory.
The decisive issue is the semantics of mental predicates. In former times it was almost
common ground among philosophers that the meaning of each predicate consists in a set
of necessary and sufficient conditions. A predicate applies to an object if and only if these
conditions are satisfied. With the advent of Kripke and Putnam philosophical common
sense changed radically. Nowadays many seem to believe that all predicates (including all
mental predicates) work rather like names—they, as it were, refer to properties; they
Page 15 of 26
What is Property Physicalism?
apply to an object just in case the object has the property referred to. Empirical research
may reveal the nature of this (p. 166) property. But for the predicate to be true of an
object, the object has to possess none of the known surface features of the property.
Let us briefly rehearse Kripke's arguments. His aim was to show that certain linguistic
expressions, in particular proper names and natural‐kind terms, function in ways
different from those often presupposed. Names, he argues, are rigid designators always
referring to the same object even in modal contexts. Something very similar is true with
regard to natural‐kind terms. Here Kripke starts from situations of the following kind.
Scientists invite a group of competent speakers of English to join them in their lab. The
members of this group know that the usual gold they have so far been acquainted with is
the chemical element with atomic number 79. Now the scientists indicate a substance
possessing the same surface properties as our usual gold and say: ‘This is a substance we
have discovered in a meteor that recently came down in Siberia; it is a kind of stuff the
chemical structure of which is ABC’. How will members of the invited group respond to
this? Basically, there are two possibilities. On the one hand, it is possible for them to
reply: ‘How very interesting. Thus, there is a different substance with the same surface
properties as gold’. But, on the other hand, it is also possible for the members of this
group to remark: ‘How very interesting. Evidently there are other kinds of gold besides
the chemical element with atomic number 79’.20 On the basis of linguistic intuitions
Kripke professes himself certain that competent speakers of English would react in the
former manner. And precisely this is his reason for thinking that the word ‘gold’, too, is a
rigid designator—an expression which in all possible worlds designates one and the same
chemical substance. If competent speakers of English react in the former manner this,
however, does not only show that ‘gold’ is a rigid designator, it also shows that none of
the known surface properties of gold is decisive for the applicability of the term ‘gold’.
This term is used in English to refer to a certain kind of chemical substance. And
chemical substances are individuated by their molecular structure. It is, therefore, the job
of science to tell us what substance the term ‘gold’ actually stands for. Whether ‘gold’
applies to a given object depends only on its molecular structure, not on its surface
properties.
But, of course, what I would like to call the Kripke test could have had quite a different
outcome. Take, for example, a situation in which scientists tell us: ‘Usually all substances
that are water‐soluble have the microstructure UVW. But now we have found a substance
with a very different microstructure which nonetheless dissolves when it is immersed in
water’. To my mind, there can be no doubt that competent speakers would say that this
new substance is water‐soluble. If a scientist proposed identifying water‐solubility with
having microstructure UVW we would tend to declare this to be at least a little bit
strange. We would say that the scientist just had not understood that objects are water‐
soluble if they dissolve when they are immersed in water. Unlike gold, water‐solubility
has no nature that can be discovered by science; unlike ‘gold’, the term ‘water‐soluble’
Page 16 of 26
What is Property Physicalism?
has an analysis. There are certain known characteristic surface features an object must
have for the term ‘water‐soluble’ to apply to it.
(p. 167)
Thus, there are two kinds of predicates. On the one hand, some—especially natural‐kind
terms such as ‘water’, ‘gold’, etc.—work rather like names. They apply to kinds of things
the nature of which has to be revealed by science. On the other hand, there are
predicates like ‘transparent’, ‘water‐soluble’, and ‘poison’ which admit of an analysis.
They apply on the basis of associated characteristic features.21
The fact that there are different kinds of predicates for which different kinds of questions
are appropriate22 is well reflected in scientific practice. Sometimes, indeed, scientists try
to discover the nature of the properties predicates stand for. They tell us what water is,
what cats are, what clouds are, what lightning is, etc. They tell us the nature of things
that, as it were, have a real essence. If we ask what water is, what clouds are, or what
lightning is we get answers like ‘Water is the chemical compound H2O’, ‘Clouds are large
amounts of water droplets in the sky forming white, grey, or black masses’, and
‘Lightning is a kind of electrical discharge’. And when we get these answers, there is no
point in asking a further ‘Why?’. You just cannot explain why water is H2O, why clouds are
large amounts of water droplets in the sky, or why lightning is an electrical discharges.
That's how it is, and that's that.23
On the other hand, scientists do not tell us what the nature of water‐solubility is or what
the nature of transparency is. We already know that; that is, we know the analyses of the
corresponding predicates. An object is water‐soluble if and only if it dissolves when
immersed in water. An object is transparent if and only if light passes through it. Water‐
solubility and transparency do not have a real, they just have a nominal essence.
Nonetheless, scientists can tell us a lot of things about water‐solubility and transparency.
They can tell us why salt is water‐soluble and why glass is transparent; that is, they can
give reductive explanations of these facts. With regard to water‐solubility and
transparency, scientists do not answer what‐, but why‐questions. It is just the other way
round with regard to water and clouds. Let us have a look at a couple of examples.
Common salt is a compound with the chemical structure NaCl. The metal sodium reacts
with chlorine because chlorine atoms can complete their outer electron shell with the
electrons given off by the sodium atoms. In this process sodium and chlorine ions are
created, which exert strong electrical forces on each other and which, therefore, form a
lattice structure. Because of the forces the ions exert on each other, common salt is solid.
But it is nonetheless water‐soluble, since water molecules can—due (p. 168) to their
dipolar structure—pull the sodium ions and the chlorine ions from their places within the
lattice.
The ability of soap to loosen dirt can also be traced back to the structure of its molecules.
For materials like oils or fats that are not water‐soluble can be suspended and washed
Page 17 of 26
What is Property Physicalism?
away when they are surrounded by molecules having a long lipophilic chain of
hydrocarbon on the one end and a hydrophilic group of alcylcarboxylate on the other. And
precisely this is true of soap molecules.
Finally, the discovery of the double helix was the decisive breakthrough in understanding
the chemical processes underlying the biology of reproduction and heredity.
There are a number of things to note here. First, all these examples exhibit exactly the
two‐step structure that is, as we have already seen, characteristic of reductive
explanations. The first step is to give an analysis of the predicate in question, and the
second step consists in showing that it follows from the basic laws of nature that objects
with a certain microstructure satisfy this analysis. It would not suffice to say that the
sodium and chlorine ions form a lattice structure which is responsible for the water‐
solubility of salt if it would not follow from the basic laws of nature that, due to their
dipolar structure, water molecules have the ability to pull the ions from their places
within the lattice.
Second, these examples show that reductive explanations are ubiquitous in all sciences
and especially in chemistry and biology. To give such explanations is at least as much an
aim of these sciences as to discover general laws. Actually, at the bottom of the search for
reductive explanations is what I would like to call the idea of the unity of the world. It is
commonly agreed that our world has a layered structure. Atoms consist of elementary
particles, molecules of atoms, cells of molecules, living beings of cells, social groups of
living beings (see Kim 1998: 15). Time and again, new properties occur at higher levels.
Atoms are neither solid nor fluid; only collections of molecules have these properties.
Molecules again don't breath or nourish, nor do they reproduce; only living beings have
these capacities. The idea of the unity of the world just consists in the claim that all
properties of higher‐level entities have a reductive explanation in terms of the properties
and the arrangement of their lower‐level parts. That salt is water‐soluble has an
explanation in terms of the properties of the ions that salt molecules consist of. That soap
has the capacity to loosen dirt can be explained by its molecular structure. That animals
have the capacity to digest has a chemico‐physiological explanation, etc. It is, of course,
an empirical question whether all higher‐order phenomena can be explained this way. The
unity‐of‐the‐world thesis is not a priori true. But it seems to be an aim of science to show
that it is an empirical truth.
Let us come back to the decisive point. Sometimes scientists are quite content when they
get an answer to the question ‘What is X?’. ‘Granite is a magmatic rock, composed
primarily of feldspar, quartz, and mica.’ ‘Asthma is a chronic inflammation of the
respiratory tract caused by antibodies produced by the humoral immune system in
response to inhaled allergens.’ In these cases there is no room left for asking an
additional ‘Why?’. On the other hand, there are also cases in which it is not only natural
but perfectly legitimate to ask ‘Why?’. Why is salt water‐soluble? To answer (p. 169) this
question, it will not do to just point to the fact that water has a certain molecular
structure. Nothing short of a reductive explanation will suffice. One has to show that
Page 18 of 26
What is Property Physicalism?
In accordance with scientific practice we should say that it is breaches in nature like the
one just described that would prove physicalism to be false, at least physicalism with
regard to the water‐solubility of salt. If we take scientific practice seriously we must
distinguish. With regard to properties for which it suffices to answer what‐questions and
for which why‐questions are inappropriate it is reasonable to understand property
physicalism as requiring identity. However, with regard to properties which are expressed
by predicates that admit of an analysis and for which why‐questions are perfectly
legitimate property physicalism must be understood as requiring reductive explanation.
Nothing is gained by saying water‐solubility is identical with having microstructure UVW.
Here we can be content only if it is possible to show that it follows from the basic laws of
nature that substances with this microstructure behave in just the way characteristic of
water‐solubility. That is the reason for the intuitive impact of the explanatory‐gap
argument.
Exponents of analytic functionalism claim that the predicate ‘is in pain’ has an analysis. It
is true of a person if the person is in a state which has a certain causal role—it is caused
by tissue damage, it causes wincing and pain‐relieving behaviour, it decreases the
capacity to concentrate, etc. Qualia friends hasten to add that the most important aspect
is that this state has the characteristic qualitative character, that it feels painful to be in
that state. Identity theorists oppose.25 On their view, ‘is in (p. 170) pain’ stands for a
property which empirical research will reveal to be, for example, a physical property. So
let us do the Kripke test.
Suppose empirical research has demonstrated C‐fibre firing to be the best candidate for
being identical with being in pain. In our world beings feel pain if and only if their C‐
fibres are firing. Now consider the following counterfactual situation. Scientists invite a
group of competent speakers of English to join them in their office. The members of this
group know that the neural correlate for pain is C‐fibre firing. The scientists declare:
‘Recently we examined a group of original inhabitants living on a so‐far‐unknown island.
Members of this group do not exhibit any characteristic pain behaviour even if their C‐
fibres are firing; what is more, none of them is reporting painful feelings. On the other
Page 19 of 26
What is Property Physicalism?
hand, members of the group exhibit pain behaviour and report painful feelings not if their
C‐fibres but if their D‐fibres are firing’. Maybe the members of the invited group will be
at a loss for a moment. But how will they describe the new situation? Again, there are two
possibilities. On the one hand, it is possible for them to reply: ‘How very interesting.
Thus, there are people who are in pain, but do not exhibit any characteristic pain
behaviour nor experience any painful feelings. And, what is more, there seems to be
another property which is sometimes associated with the characteristic features that
usually accompany being in pain’. On the other hand, however, it is also possible for
members of the group to remark: ‘How very interesting. Obviously, some people are not
in pain though their C‐fibres are firing whereas others are in pain if their D‐fibres are
firing’. If the members of the group described the new situation in the first manner, this
would prove that in English ‘pain’ is a rigid designator referring to a certain property the
nature of which has to be discovered by science. And if they described the situation in the
second manner, this would show that in English ‘pain’ is used as a predicate that applies
to a person just in case this person behaves in a certain way and has the right feelings. I
am quite sure that they would describe the situation in the second way. And if so, this
would show that ‘pain’ is akin to ‘transparent’ and not to ‘water’.
Incidentally, this result is also supported by the fact Kim has stressed—the fact that it
seems perfectly legitimate to ask questions like ‘Why does pain, not itch or tickle, arise
from C‐fibre firing?’ and ‘Why and how does conscious experience arise from certain
other neural states?’. For we have already seen that such why‐question are appropriate
just in case we are dealing with predicates like ‘water‐soluble’ and ‘transparent’.
If it is admitted, however, that mental predicates belong to this kind of predicates, they
no longer fit the identity theory. Identity statements make sense only if the predicates
involved work like names. It is only in this case that we can ask for the nature of the
properties these predicates stand for. And it is only in this case that we can reasonably
ask whether these properties are (identical with) physical properties. (p. 171) But if
mental predicates belong to the same group of predicates as ‘water‐soluble’ and
‘transparent’, it is reductive explanation that is at issue. Just as we can legitimately ask
for a reductive explanation of the fact that salt and sugar are water‐soluble, we can also
legitimately ask for a reductive explanation of the fact that beings whose C‐fibres are
firing feel pain. Physicalism with regard to mental properties, therefore, does not amount
to the claim that all mental properties are physical properties. It rather amounts to the
claim that all mental properties can be reductively explained in terms of physical
properties; that is, that mental properties are not emergent in the sense of C. D. Broad.26
References
Beckermann, A. (2000), ‘The Perennial Problem of the Reductive Explainability of
Phenomenal Consciousness: C. D. Broad on the Explanatory Gap’, in T. Metzinger (ed.),
Page 20 of 26
What is Property Physicalism?
—— (2001a), Analytische Einführung in die Philosophie des Geistes, 2nd, rev. edn. (Berlin,
New York: de Gruyter).
Beckermann, A. (2001b), ‘The Real Reason for the Standard View’, in A. Meijers (ed.),
Explaining Beliefs: Lynne Rudder Baker and Her Critics (Stanford, Calif.: CSLI), 51–67.
Block, N., and Stalnaker, R. (1999), ‘Conceptual Analysis, Dualism, and the Explanatory
Gap’, Philosophical Review, 108: 1–46.
Braddon‐Mitchell, D., and Jackson, F. (1996), The Philosophy of Mind and Cognition
(Oxford: Blackwell).
Broad, C. D. (1925), The Mind and Its Place in Nature (London: Routledge & Kegan Paul).
—— (1956), Meaning and Necessity, 2nd edn. (Chicago, Ill.: Chicago University Press).
Chalmers, D., and Jackson, F. (2001), ‘Conceptual Analysis and Reductive Explanation’,
Philosophical Review, 110: 315–60.
Frege, G. (1892), ‘über Sinn und Bedeutung’, Zeitschrift für Philosophie und
Philosophische Kritik, 100: 25–50;
trans. as ‘On Sense and Nominatum’, in H. Feigl and W. Sellars (eds.), Readings in
Philosophical Analysis (New York: Appleton‐Century‐Crofts, 1949), 85–102.
Page 21 of 26
What is Property Physicalism?
Hill, C., and McLaughlin, B. (1999), ‘There are Fewer Things in Reality than are
(p. 172)
—— (2005), ‘The Case for A Priori Physicalism’, in C. Nimtz and A. Beckermann (eds.),
Philosophie und/als Wissenschaft. Philosophy—Science—Scientific Philosophy (Paderborn:
Mentis), 251–65.
Kim, J. (1998), Mind in a Physical World: An Essay on the Mind–Body Problem and Mental
Causation (Cambridge, Mass.: MIT Press).
Levine, J. (1983), ‘Materialism and Qualia: The Explanatory Gap’, Pacific Philosophical
Quarterly, 64: 354–61.
—— (1993), ‘On Leaving Out What It's Like’, in M. Davies and G. W. Humphreys (eds.),
Consciousness (Oxford: Blackwell), 121–36.
McLaughlin, B. (2001), ‘In Defence of New Wave Materialism: A Response to Horgan and
Tienson’, in C. Gillett and B. Loewer (eds.), Physicalism and Its Discontents (Cambridge:
Cambridge University Press), 319–30.
Place, U. T. (1956), ‘Is Consciousness a Brain Process?’, British Journal of Psychology, 47:
44–50.
Putnam, H. (1975), ‘The Nature of Mental States’, in Putnam, Mind, Language, and
Reality: Philosophical Papers, ii (Cambridge: Cambridge University Press), 429–40.
Segal, G. (2000), A Slim Book about Narrow Content (Cambridge, Mass.: MIT Press).
Page 22 of 26
What is Property Physicalism?
Smart, J. J. C. (1959), ‘Sensations and Brain Processes’, Philosophical Review, 68: 141–56.
Notes:
(1) Gottlob Frege (1892) struggled with the question of whether sameness is a relation
between objects or between names of objects. Finally, he decided on the first option.
Nonetheless, true identity statements like ‘a = b’ never convey the information that the
object referred to by ‘a’ and ‘b’ is identical to itself. That is something we already know in
advance. And whatever new information identity statements may convey, it remains the
case that ‘a = b’ is true if and only if ‘a’ and ‘b’ refer to the same individual.
(4) This, of course, is not what Levine really claims. He only maintains that there is an
important epistemic difference between the identity statements (1) and (2) even though
this difference by itself does not imply that (2) is false.
(5) Block and Stalnaker (1999) have argued that even if the property to be explained in a
reductive explanation has an analysis, this analysis will be couched in terms that do not
belong to the vocabulary of the basic laws the explanation is based on. In order to
complete such explanations, we therefore have to add bridge laws which tell us which
higher‐level features are identical to which lower‐level phenomena. Reductive
explanations, therefore, presuppose true property‐identity statements instead of replacing
them. In my view, it is indeed true that reductive explanations presuppose appropriate
principles to bridge the gulf between different mereological levels. These principles,
however, are not a posteriori identity statements. The bridge principles actually used in
science rather seem to be unproblematic a priori conditionals. (‘If all parts of a disc
revolve around a certain point, the disc itself rotates around this point’; ‘An object is
transparent if light rays pass through it’; ‘An object dissolves in a liquid if its parts
[molecules] are untied from each other and distributed among the molecules of the
liquid’.)
(6) Actually, I think Chalmers and Jackson (2001) are right in arguing that in general
reductive explanations do not presuppose a complete analysis of F. It suffices that the
truth of the conditional ‘All systems with micro‐structure [C 1, …, C n; R] have F’ can be
known a priori on the basis of the general laws of nature holding for component parts C 1,
…, C n, suitable bridge laws, and what competent speakers know about the extension of
‘F’.
(7) Of course, Papineau was not the first advocate of this view (see e.g. Hill 1984, 1991).
Page 23 of 26
What is Property Physicalism?
(8) It is very important to distinguish Broad's account of reductive explanation from the
account that has recently been developed by Chalmers and Jackson. According to
Chalmers and Jackson, a phenomenon is reductively explainable in terms of the physical
iff it follows a priori from a complete description of the physical world plus a ‘that's all’
clause plus some indexical truths. As we will see, it may well be that identity statements
like ‘water = H2O’ are reductively explainable in Chalmers's and Jackson's sense. That
the two senses of ‘reductive explanation’ are different already follows from the fact that a
lot of physical macro‐properties that would be emergent in Broad's sense are reductively
explainable in the sense of Chalmers and Jackson (see their 2001: sect. 4). The reason for
this difference is that Broad's basis is only the fundamental laws of nature while
Chalmers and Jackson start from a complete description of the physical world which also
comprises all exceptions from these laws.
(9) For an overview of the different versions of functionalism see Braddon‐Mitchell and
Jackson (1996).
(10) This is so even if the causal role itself is entirely expressible in terms of physics.
Even then the role filler, the state by which the role is realized in a certain being, may be
a non‐physical state.
(12) Actually this formulation only applies to the concept of global supervenience. Here,
however, I do not wish to go into the difference between local and global supervenience.
(13) As already mentioned (n. 8 above), π also has to include indexical physical truths
about ourselves and a ‘that's all’ clause stating that the description really is complete—
otherwise π will not entail the mental fact that there are no immaterial ghosts having
pain. This is reflected in Jackson's demand that the relevant possible worlds have to be
minimal physical duplicates of our world (see Chalmers and Jackson 2001).
(14) Again, π also has to include indexical physical truths about ourselves and a ‘that's all’
clause stating that the description really is complete. It should also be noted that a priori
physicalism in the sense of (AP) does not maintain that physicalism is a priori true.
(15) A rigid designator refers to the same object in all possible worlds if it refers to
anything at all.
Page 24 of 26
What is Property Physicalism?
(16) To avoid misunderstandings, I would like to make clear what I mean by ‘semantic
property’ or ‘semantic fact’. Frege held the view that each expression has a sense and a
reference and that the sense and reference of complex expressions result in regular ways
from the senses and references of their components. What we may call a ‘Frege
dictionary’ thus contains for each non‐complex expression two entries, one in which its
sense and one in which its reference is given (for all possible worlds). Moreover, we may
assume that all references are given in a canonical, transparent way; that is, if two names
refer to the same object, their references are given by the same expression. For example,
since ‘Mark Twain’ and ‘Samuel Clemens’ refer to the same writer, their references are
given by the same words; say, ‘Mark Twain’. According to Frege, predicates refer to
concepts. Exponents of the identity theory, rather, seem to assume that predicates stand
for properties. Be that as it may, for each predicate a Frege dictionary contains an entry
that gives the reference of this predicate; that is, that tells us what the predicate stands
for. This again is done in a canonical, transparent way. If two predicates stand for the
same concept or property, their references are given by the same expression. A semantic
fact is what follows from all entries in a Frege dictionary. (The idea of a Frege dictionary
has already been alluded to by R. B. Marcus: 1963: 115). At least, this is what Kripke
takes her to say (1971: 142).
(17) ‘The sense of a proper name is grasped by everyone who knows the language […] A
complete knowledge of the nominatum would require that we could tell immediately in
the case of any given sense whether it belongs to the nominatum. This we shall never be
able to do’ (Frege 1892: 86; my emphasis).
(18) ‘Identity is utterly simple and unproblematic. Everything is identical to itself; nothing
is ever identical to anything else except itself. There is never any problem about what
makes something identical to itself; nothing can ever fail to be. And there is never any
problem about what makes two things identical; two things never can be identical’ (Lewis
1986: 192–3).
(19) Say, by explaining away the alleged possibility or by holding that mental properties
are identical to functional properties. Of course, the reductive‐explanation theory has no
problem whatsoever with this argument.
(20) For more realistic cases see Segal (2000: ch. 5).
(21) Of course, the distinction I want to stress could also be put in the following way. All
predicates stand for properties. But not all properties are of the same kind. Some do not
admit of an analysis, others do. I, however, prefer to speak of predicates instead of
properties. Generally, we have to be aware of the fact that the whole debate is threatened
by a confusion—the confusion between predicates and concepts on the one hand and
properties on the other. Time and again we are told that properties are to be analysed or
that we should try to deduce properties of a whole from knowledge of the properties of its
parts. Literally speaking, this makes no sense. One can only deduce sentences or
Page 25 of 26
What is Property Physicalism?
propositions, but not properties. And one can analyse concepts or the meaning of
predicates, whereas it seems at least reasonable to ask whether it is really possible to
analyse a property. It is predicates or concepts that get analysed.
(23) Of course, this does not mean that these identity claims have no justifications. See
e.g the justification mentioned on p. 159 above.
(24) This breach would be no atom smaller even if we assumed that water‐solubility is
identical to a certain molecular structure. If the water‐solubility of common salt is
emergent, this means that there are circumstances in which salt molecules behave in
ways that cannot be accounted for by the general laws holding for sodium and chlorine
ions. And this in turn implies that these laws are incomplete; that is, that there are
circumstances in which sodium and chlorine ions behave in ways that cannot be
accounted for by these laws (see Beckermann 2001b).
(25) Some philosophers, however, argue that identity theorists do not have to deny that it
is part of the analysis of ‘is in pain’ that pains feel painful, since they are not so much
concerned with the property of being in pain (a property of living beings) as with the
qualitative character of pains—the quale of feeling painful (a property of certain mental
states). This quale, these identity theorists claim, is identical to a physical or functional or
representational property. But this version of the identity theory seems to be confronted
with quite similar problems; for example, it is possible to construe a similar Kripke test
(see the next paragraph).
(26) I am very grateful to Brian McLaughlin, Christian Nimtz, Michael Schütte, Sven
Walter, and especially Wolfgang Schwarz for extremely helpful comments on earlier drafts
of this paper.
Ansgar Beckermann
Page 26 of 26
What is the Physical?
Addressing the question of what is the physical is important not only for understanding
the general thesis of physicalism, but also for assessing the strength and significance of a
number of the central arguments regarding physicalism and the physical nature of the
mind. For example, debates over the possibility of zombies pervade the literature on
consciousness. Zombies, in the relevant sense, are supposed to be creatures that are just
like us in every physical respect yet lack consciousness. Are such creatures possible?
Some have argued that while it is not conceivable, or logically possible, that there could
be, say, H2O that is not water, it is conceivable that there could be worlds that are
physical duplicates of ours yet populated by zombies rather than conscious human
beings, and thus consciousness does not logically supervene on the physical.
Keywords: physicalism, consciousness, nature of mind, physical duplicates, physicalism, physical nature
PHYSICALISM, Hilary Putnam once said, is the only metaphysical position with any real clout
(1987). Its clout comes from its alignment with physics; but, as we will see, physics is also
a cause of its troubles. Physicalists generally claim that everything, or at least some
significant subset of everything, is physical.1 Most significantly, physicalists hold that
minds and mental properties—physicalism's primary nemeses—are entirely physical. But
what is it to be physical? With a few exceptions, physicalists have had little to say about
this. Yet formulating a workable notion of the physical is of utmost importance—for what
good is clout, if you don't have substance?
Addressing the question of what is the physical is important not only for understanding
the general thesis of physicalism, but also for assessing the strength and significance of a
number of the central arguments regarding physicalism and the physical nature of the
mind. For example, debates over the possibility of zombies pervade the literature on
consciousness. Zombies, in the relevant sense, are supposed to be creatures that are just
Page 1 of 18
What is the Physical?
like us in every physical respect yet lack consciousness. Are such creatures possible?
Some have argued that while it is not conceivable, or logically possible, that there could
be, say, H2O that is not water, it is conceivable that there could be worlds that are
physical duplicates of ours yet populated by zombies rather than conscious human
beings, and thus consciousness does not logically (p. 174) supervene on the physical. But
how are we to determine whether such zombies are possible? Without some prior notion
of the physical, intuitions here are prone to run amok.
How one understands the notion of the physical also plays an important role in
determining what one thinks about what poor, dear Mary—the imprisoned, brilliant
neuroscientist—knows about the experience of seeing red. Would Mary, who knows all the
physical facts about colour vision yet has spent her entire life in a black and white
environment, be surprised upon seeing her first ripe tomato or would she just say, ‘Ho‐
hum, I knew that's what red would look like all along’? To answer this question we need
some understanding of what it is that Mary knows, which minimally involves at least
some notion of what counts as a physical fact. Are the physical facts the facts that science
tells us about? Are they the facts about the ultimate constituents of nature, whatever they
may be? Different answers to the question of what it means to be physical will ground
different intuitions about what Mary knows, and even about the coherence of the entire
thought‐experiment.2
Finally, it is often claimed, as Wayne Davis puts it, that ‘the principle argument for
physicalism is the success of physics’ (1995: 679). Or, in J. J. C. Smart's words: ‘[S]cience
is increasingly giving us a viewpoint whereby organisms are able to be seen as physico‐
chemical mechanisms … yet, that everything should be explicable in terms of physics …
except the occurrence of sensations seems to me to be frankly (p. 175)
Page 2 of 18
What is the Physical?
unbelievable’ (1959: 142).4 The reasoning here assumes that the physical world is the
world given to us by physics (or physics and chemistry)—else why would its success be
reason to think that the mental will fall under the physicalist umbrella? However, as we
will see, there are serious problems with such a view.
Although we need at least some conception of the physical to fully engage with these
issues, fortunately we need not provide necessary and sufficient conditions for being
physical before entering the debate. Understanding concepts and determining what falls
under their scope can go hand in hand, and as such the process of determining whether
the mind is physical is itself a step in clarifying our understanding of the physical.5
Nonetheless, to enter the debate we need some ground to stand on, since without any
conception of the physical at all, debates over whether the mind is physical would be as
illuminating as the discussions two five‐year‐olds might have about the nature of infinity
(and much less entertaining).
Some think that we already have an entry point into the debate; namely, common sense.
On their view, our grounds for using the term ‘physical’ in philosophical arguments are no
more shaky than, say, our grounds for using the term ‘modest’ in questions about whether
an individual is modest: although in neither case can we fully explain the concept, both
are intuitive enough. The two cases, however, are not analogous. For one thing, the
concept of being modest is a concept from ordinary‐language. To be sure, there are
ordinary‐language usages of ‘the physical’ as well, but they tend not to line up with
philosophers' needs. For example, one might say, ‘taking care of an infant is physically,
but not mentally, challenging’. Here the physical pertains to non‐mental aspects of our
bodily actions, but in a debate over whether the mind is physical this cannot be what is
meant. In ordinary language ‘physical’ is sometimes contrasted with virtual or digital, as
in ‘I'd prefer a physical copy to an email’, but no physicalist should think an email isn't
also physical. For another thing, while we can point to examples of modest people and to
examples of immodest people, it is often not clear what is supposed to serve as an
example of something non‐physical, or at least something that if it were to exist would
count as non‐physical. Yet to avoid turning the claim ‘the mind is physical’ into a trivial
truth we need to have some idea of a contrasting class, something that, if it were to exist,
would count as non‐physical. The mental clearly cannot be taken as an example. But what
can?
Page 3 of 18
What is the Physical?
were to exist, would not take up space, yet neither, it seems, should count as counter‐
examples to physicalism. It even seems unreasonable to exclude from the physical realm
things that do not exist in space‐time. Physicists have speculatively posited something
from which space‐time itself emerges, the playfully called ‘zerobrane.’ Yet since there
seems no reason to take zerobranes as nonphysical, we cannot take existing in space‐time
as a criterion for being physical.6 Richard Healey sums up the situation well when he
says, ‘[the] expanding catalogue of elementary particle states of an increasingly recondite
nature seems to have made it increasingly hard for the physicists to run across evidence
that would cast doubt on a thesis of contemporary physicalism stated in terms of
it’ (1979: 208). Or, as Bertrand Russell said years before, ‘matter has become as ghostly
as anything in a spiritualist's séance’ (1927/1992: 78).7 The moral is clear: ghosts, ghost
stuff and other pretheoretically bizarre phenomena are not going to provide viable
contrasts to the physical.
It might be claimed that we can have a useful notion of the physical even without
specifying a contrasting class. Certain philosophers have argued that we can merely start
with certain prototypical examples of physical objects and properties and extrapolate
from there. But after specifying central cases of the physical, determining how to go on
from there is no trivial matter. Let us take our prototypical examples of the physical from
everyday so‐called physical objects such as rocks and trees. In a sense, the mind, be it
physical or not, would fit much more smoothly in with rocks and trees than with the
fundamental particles of physics. For example, the notion of the mental fits easily into
ordinary language. We also take the mental to fit into the same macro‐level causal
network as everyday objects: in explaining why Sam moves his hand (as opposed to the
micro‐particles in his hand) toward the glass of water, we naturally refer to his sensation
of thirst, desire for water, and so forth. And these things would be true even if the mental
were non‐physical. But, presumably, since in formulating our concept of the physical we
are supposed to extrapolate from rocks, trees, and hands so as to cover the phenomena of
physics but not non‐physical minds, something more needs to be said.
A more sophisticated version of this idea is the view that ‘physical’ is a natural kind term.
That is, we start with our central examples of physical things and properties and then
take our basic physical phenomena to be that which constitutes or realizes such things
and properties, whatever it may be. As William Lycan states it, physicalism (p. 177) is the
view that ‘creatures with minds are made entirely of the same ultimate components as
are ordinary inanimate objects, and their properties are determined by the ways in which
those components are arranged and related to external things’ (2003: 15; see also
Snowdon 1989). As Lycan sees it, this way of putting the physicalist thesis is ‘good
enough to work with and … distinguish[es] that thesis from the views of Descartes and
Hume’ (2003: 15).
Descartes certainly did reject the view that ‘creatures with minds are made entirely of
the same ultimate components as are ordinary inanimate objects’. And Hume seems to at
least reject the view that we can know that minds and ordinary objects are made of the
same ultimate components since he holds that, in his words, ‘matter … and spirit are at
Page 4 of 18
What is the Physical?
bottom equally unknown and we cannot determine what qualities inhere in the one or in
the other’ (1783/1998: 324). But although physicalists should reject these views this isn't
all they should reject. Among other positions, they should reject panpsychism; roughly,
the view that the fundamental nature of reality involves mentality. Yet panpsychism can
be formulated so that it is consistent with Lycan's proposed criterion. Because Lycan's
criterion is too weak to exclude panpsychism, it is unclear whether such a criterion is, in
fact, ‘good enough to work with’.8
Lycan's criterion is too weak, since it does not exclude panpsychism, but in another sense
it is too strong. Physicalists are monists inasmuch as they think that all matter is
fundamentally composed of the same kind of stuff; namely, physical stuff. However, it
seems that there can be more than one kind of fundamental physical stuff. Gassendi
suggested this when he objected to Descartes's dualism, stating that while he would
grant Descartes his conclusion that mind and body are really distinct, he did not accept
that this implies that the mind is incorporeal (Haldane and Ross 1968: 237). Moreover, if,
as some physicists posit, dark matter is made entirely out of axions, a new kind of
fundamental particle, dark matter, it would seem, should still count as physical. To be
sure, the thesis that human beings contain within them fundamental stuff that does not
occur in such things as rocks and chairs is, as Lycan puts it, rather ‘loony’ (2003: 14). But
I think that even Lycan would agree that if such stuff were discovered to exist, it could
count as physical. Thus, Gassendi is correct: merely stating that the mental is not made of
the same ultimate components as ordinary inanimate objects does not suffice to make it
non‐physical. But what then does?
In respect to the mind, we want to know what is it that the mental supervenes on, is
realized by, or is identical to (fill in your favourite dependence relation) in order for it to
count as physical. To be sure, strictly speaking not all physicalists say that the mind, or
even the brain for that matter, is physical. This is because not all physicalists apply the
term ‘physical’ to higher‐level phenomena. While this can cause some confusion, the issue
here is primarily terminological, since physicalists typically (p. 178) do not take the
existence of rocks and trees to be counter‐examples to physicalism (since rocks and trees,
they will claim, depend on lower‐level physical phenomena) even if they choose to reserve
the term ‘physical’ for lower‐level phenomena. A significant issue, however, is how to
identify these lower‐level phenomena; that is, how to identify that upon which, according
to the physicalist, everything else depends. Identifying such dependence bases is, as
Jeffrey Poland puts it, ‘one of the deepest foundational issues facing physicalists’ (1994).
Most physicalists want to specify the dependence bases in terms of whatever the
physicist says exists. Wave‐particle duality, one‐dimensional strings, and other ghostly
phenomena do not count as counter‐examples to physicalism because, it is said, the
physical is whatever is given to us by physics itself. In avoiding counter‐examples in this
way, however, we run into what is called ‘Hempel's dilemma’ (see Hempel 1980).9 On the
one hand, it seems that we cannot define the physical in terms of current physics, since
current physics is most likely not correct. Despite some physicists' heady optimism that
the end of physics is just around the corner, history cautions prudence. The end of
Page 5 of 18
What is the Physical?
physics has been predicted before: toward the end of the nineteenth century, just before
the relativity revolution, Lord Kelvin remarked that all that is left for physics is the filling
in of the next decimal place; then, in the early part of the twentieth century Max Born
supposedly claimed that physics would be over in six months. And, in all likelihood,
today's claims that we've (just about) got it right are similarly unrealistic: today's physics
is probably neither entirely true, in the sense that some of our theories may look as
wrong‐headed to future generations as phlogiston theory looks to us now, nor complete,
in the sense that there are still unaccounted for phenomena.10 Yet, on the other hand,
Hempel claims that if we take physics to be some future unspecified theory, the claim that
the mind is physical is extremely vague, since we currently do not know what that theory
is. Geoffrey Hellman sums up this dilemma nicely: ‘[E]ither physicalist principles are
based on current physics, in which case there is every reason to think they are false; or
else they are not, in which case it is, at best, difficult to interpret them, since they are
based on a “physics” that does not exist’ (1985: 610).
A few philosophers take on the first horn and define the physical over current physics.
Apart from making physicalism most likely false, this approach has some other rather odd
consequences. For example, it seems to imply that neither Hobbes nor Le Mettrie count
as physicalists and, moreover, that if a new particle is discovered next week, it will not be
physical. Of course, it does provide us with a theory the truth of which is determinable
now. But this can be said as well of the theory that the earth is flat. Hellman's response is
basically to reject physicalism as a possibly true theory about the ultimate nature of the
world yet allow that physicalism as defined (p. 179) over current physics may be of
practical use, as false theories sometimes are. He thinks, for instance, that it may serve
as a criterion of adequacy for the completion of physics (Hellman 1985: 610). And
perhaps it can, but such a physicalism is far from what most physicalists take themselves
to be defending, which is a theory that tells us something substantial about the ultimate
nature of the mind.
Andrew Melnyk argues for something bolder: physicalism defined over current physics,
he argues, is almost certainly false yet nonetheless acceptable as a substantial theory
about the ultimate nature of reality (1997). It is acceptable, he argues, because the
attitude a physicalist should have towards the thesis of physicalism is the attitude
someone with realist and anti‐relativist intuitions has towards her favoured scientific
hypotheses. This attitude, he claims, amounts to taking one's favoured theory to be better
than current and historical rivals; that is, those formulated theories that are sensibly
intended to achieve a significant number of the favoured theory's goals. As such, it does
not require that one take one's hypothesis to be true, nor, even, more likely true than
false. Rather, just as, say, a string theorist asserts that everything is made out of strings
while being aware that string theory is probably far from the last word and in fact more
likely to be false than true, a physicalist asserts that everything is physical—in the sense
of either being accounted for directly by the phenomena of current physics or realized by
such phenomena—while being aware that physicalism is probably far from the last word
and in fact more likely to be false than true. According to Melnyk, then, we can take on
the first horn of Hempel's dilemma—that is, we can define the physical in terms of
Page 6 of 18
What is the Physical?
current physics—because (1) one can reasonably accept a scientific hypothesis while at
the same time accepting that it is more likely to be false than true as long as the theory is
better than its relevant rivals; (2) physicalism, as Melnyk understands it, what we can call
‘M‐physicalism’, while more likely to be false than true is better than its relevant rivals;
and (3) M‐physicalism is a scientific hypothesis. Thus, defining the physical over current
physics is supposed to give us an acceptable substantial theory.
One question to ask here is whether M‐physicalism is better than its relevant rivals. And
answering this is no easy task.11 But let us put this issue to the side, since it is not even
clear that taking physicalism to be better than its relevant rivals gives the physicalist
what she needs. It seems that in philosophy when no options are good yet one is better
than the other, it very well may not be reasonable to accept the best of this bad lot, and,
in my opinion, philosophical honesty should prevent us from doing so. Consider the case
of free will. You might think that one response to the free‐will debate is better than all
others because it is, perhaps, clearer, or perhaps it (p. 180) alone is not self‐contradictory,
yet since you think that all responses are failures, you do not accept any of them. But
Melnyk has a different take, since, as he sees it, ‘surely physicalism should be viewed as a
scientific hypothesis’ (1997: 226).
Smart proposes a different way to take on the first horn of the dilemma (1978). While
Smart defines the physical over current physics, he does not think that this probably
makes physicalism false. Rather, he argues that for the purposes of the mind–body
problem current physics is good enough, since the principles relevant to understanding
the mind are principles of ‘ordinary matter’ (for example, principles relevant to
understanding neurons, such as molecular theory) and that these principles will most
likely not be overthrown. Or, as Lycan puts it: ‘the changes in the physics underlying
biology and chemistry should not matter in any way to the mind, however much they
Page 7 of 18
What is the Physical?
matter to matter’ (2003: 15). In other words, physicalism, understood in this way, says
that once you've duplicated the phenomena at the neural level, you've duplicated the
mental, regardless of what is going on at lower levels.
But it seems that the Smart/Lycan approach to taking on the first horn of Hempel's
dilemma runs into problems as well, since if the neural itself is not physical, as would be
the case if panpsychism were true, then although it may be that once we've duplicated
the neural we've duplicated the mental, the mental would still not be physical. Moreover,
although it may be very unlikely that our understanding of the mental will not be
furthered by new developments in fundamental physics, it would seem to be within the
realm of possibility that, say, the particular visual sensation of seeing red requires a
certain type of process to be occurring at the level of fundamental physics. And it also
seems possible that neural properties do not require this particular fundamental process
(brains, as it were, are multiply realizable with respect to it). (p. 181) Such a view, it
would seem, could count as physicalistic even though it would not be true that the
principles relevant for understanding the mind are the principles relevant for
understanding ordinary matter, and even though duplication of the neural does not
suffice for duplication of the mental.14
It seems, then, that taking on the first horn of Hempel's dilemma, that is, defining the
physical‐dependence base in terms of current physics, does not succeed. But does taking
on the second horn fare any better? Many physicalists think it does. For example, David
Armstrong explicitly tells us that when he says ‘physical properties’ he is not talking
about the properties specified by current physics, but rather ‘whatever set of properties
the physicist in the end will appeal to’ (1991: 186).15 In a similar vein, Frank Jackson
holds that the physical facts encompass ‘everything in a completed physics, chemistry,
and neurophysiology, and all there is to know about the causal and relational facts
consequent upon all this’ (1986: 29). And, even if it is not always explicitly stated, it
seems that, as Barry Loewer puts it, ‘what many have on their minds when they speak of
fundamental physical properties is that they are the properties expressed by simple
predicates of the true comprehensive fundamental physical theory’ (1996: 103).16 So for
Armstrong, and others as well, it is not today's physics upon which we are to base our
notion of the physical, and not any future physics, since, presumably, physics of the future
will still be false for a long time to come, but, rather, a true and complete physics, a
physics, as it were, sub specie aeternitatis.
Can we formulate physicalism in terms of a true and complete physics? For some, one
problem with defining the physical in terms of future physics is that since we do not
currently know what future physics will be like, we cannot now determine whether
physicalism is true. And, indeed, this seems to be Melnyk's main reason to reject attempts
to define the physical over future physics. But many physicalists do not think that our
current inability to determine the truth of physicalism is problematic. Physicalism, as they
see it, can be based on a physics that does not yet exist because it is a hypothesis that
Page 8 of 18
What is the Physical?
awaits scientific confirmation (or, for that matter, refutation). Physicalists are betting that
it is correct, but do not claim to be able to now determine that it is correct.17
Physicalists, I agree, need not now determine whether physicalism is true. Unfortunately,
a more serious problem is that, far from turning physicalism into a thesis whose truth
awaits empirical support, using the notion of ‘physics in the end’ or ‘a completed physics’
to explain the physical actually seems to trivially exclude the (p. 182) possibility that the
mind is not physical, making physicalism with respect to the mind trivially true. For what
is a true and complete physics, save for one that accounts for the fundamental nature of
everything? If mentality is a real feature of the world, a completed physics will, by
definition, account for the most fundamental nature of the mental as well.18 Yet neither
physicalists nor their foes think that at this time in the debate the mind is physical simply
as a matter of definition. Physicalists think the claim needs to be argued for and, as many
hold, will ultimately depend on what scientific investigation reveals. And their foes clearly
do not think that they are denying what amounts to, more or less, an analytic truth. It
seems, then, that when physicalists who take on the second horn of Hempel's dilemma
talk about a true and complete physics, they cannot simply mean a theory of everything,
since this would make their claim that the mind is physical trivially true. Yet there is also
reason to think that they do not simply intend to refer to the temporal end of physics. For
this physics might still be inaccurate and incomplete; even worse, for all we know physics
might regress. So it seems that physicalists need another option.
Some argue that there are phenomena that physics, and perhaps scientific investigation
entirely, does not aim to cover. Rather, physicists, they argue, in their role as physicists,
are only concerned to account for a certain class of phenomena, and because of this we
can distinguish a true and complete fundamental physics from the true and complete
fundamental theory of the world; that is, the fundamental theory of the world sub specie
aeternitatis. The true and complete theory, it is claimed, would account for the
fundamental nature of everything, whereas a true and complete physics would account
for the fundamental nature of only those things that physics aims to cover.
But what sorts of phenomena are to be excluded from the scope of a true and complete
physics? According to Brian McLaughlin, physicalists must take a stand on this. For
example, the true and complete physics, McLaughlin thinks, ‘would not, for instance,
postulate a spiritual realm that is causally isolated from the realm governed by its
laws’ (2001: 11424). On McLaughlin's view, if a causally isolated spiritual realm were to
exist, the true and complete theory of the world would account for it, yet a true and
complete physics would not. And if there are things that exist outside the scope of a true
and complete physics, physicalism is not trivially true; rather, in such a situation it would
be false, since physicalists, according to McLaughlin, hold that a true and complete
physics will be a true and complete theory.
Is it reasonable to think that physics has identifiable limits? According to Quine, ‘if the
physicist suspected there was any event that did not consist in a redistribution of the
elementary states allowed for by his physical theory he would seek a way of
Page 9 of 18
What is the Physical?
supplementing his theory’ (1981: 98). And, as Bas van Fraassen sees it, ‘there are no
science stoppers’ (1996: 80). Quine and van Fraassen suggest what I take to be a good
methodological principle: scientific enquiry should not accept a priori barriers, or, in
other words, when you discover territory that does not conform to your map, (p. 183)
change the map, not the territory. Such changes might involve not only expanding our
scientific ontology, but changing our scientific method as well. For example, if standard
controlled experiments fail to reveal phenomena that we nonetheless suspect might exist
—as some have claimed could be the case with parapsychological phenomena—we should
try to find a way to change the control. If we are somehow convinced that there is a
spiritual realm that is causally isolated from our world, let us try to understand it.
Rather than placing a priori constraints on the scope of scientific enquiry, however,
physicalists may make predictions about the limits of scientific enquiry, and can take
physicalism to be refuted if such limits are surpassed. This is the point McLaughlin is
trying to establish. However, I am not sure that the example he provides of a causally
isolated spiritual realm serves this purpose. To be sure, it seems reasonable to claim that
physics will never cover that which we will never fathom, that which we will never have
any reason to think exists. And it may be that a causally isolated spiritual realm fall into
this category. This has nothing to do with spirits, whatever they may be, as it seems that
any causally isolated realm would fall into this category as well. So it is a good bet that
any causally isolated realm is outside the scope of physics. But do we really want to
count, say, a causally isolated collapsed star as non‐physical merely because physics
could never touch it? I think that the answer is ‘no’. Moreover, even if we were willing to
do this, physicalism, as defined over a true and complete physics, would still be
necessarily true of our own causally connected world, which makes the theory trivial
enough for most tastes.
The prospects for formulating a useful notion of the physical are beginning to look rather
grim. But before we lose hope, let us look at some suggestions for demarcating the
physical from the non‐physical that come from the other side. While it is not unusual for
physicalists to think of the physical so as to favour the truth of physicalism, dualists and
other anti‐physicalists, perhaps unsurprisingly, tend to do just the opposite. For example,
some anti‐physicalists define the physical over physics yet take physics to tell us only
about the purely structural features of the world. If one, then, thinks that the mental has
some sort of non‐structural, nature and also that there is a sharp divide between the non‐
structural and the structural, it is easy to be led to the view that the mental is not
physical (1996: 153, 163, and passim).19 Other anti‐physicalists think of the physical as
the objective, as that which is knowable from a third‐person point of view (see Swinburne
1994, Taliaferro 1994, and, while not a complete anti‐physicalist, Nagel 1986). When
experiential phenomena are taken, as they are by dualists, to be knowable only from the
first‐person point of view, it is also easy to see how one could be led to reject physicalism.
(p. 184)
Page 10 of 18
What is the Physical?
Just as we should reject definitions of the physical that make physicalism analytically
true, we should reject definitions of the physical, such as the ones that seem to stand
behind the dualistic views just considered, that make physicalism analytically false.
Where does this leave us? According to some, problems with formulating a working
notion of the physical dissolve the problem of physicalism and the mind–body problem
entirely. As Chomsky sees it, because we have no way to formulate the physical/non‐
physical distinction, ‘we have no coherent way to formulate issues related to the “mind–
body problem” ’ (1995: 5; see also Chomsky 1993, 1998). Chris Daly claims that our lack
of understanding of the notion of the physical shows that ‘no debate between physicalism
and dualism can even be set up’ (1998: 213), while Tim Crane and Hugh Mellor, after
finding flaws with a wide variety of proposals for defining physicalism, conclude that their
paper ‘should really be the last paper on the subject of physicalism’ (1990: 83).
I think that although the problems we have found with many of the current attempts to
demarcate the physical from the non‐physical indicate that there is at least no obvious
way to formulate a substantial question, it does not imply that there is no mind–body
problem. Rather, there seems to be an interpretation of, or perhaps a replacement for, the
physical that, while perhaps not worthy of the clout Putnam claimed was bestowed on
physicalism by physics, does give the debate some substance.20 This replacement is
hinted at by some of those who attempt to carve a path between the two horns of
Hempel's dilemma. These physicalists attempt to give content to the notion of the
physical by taking the true and complete physics at issue to be a successor to today's
physics.21 On their view, physicalists are betting that future physics will have certain
features in common with current physics; most importantly, that it will not incorporate
fundamental mental phenomena. Physicalism thus becomes both a thesis about what this
true and complete successor physics will account for (the fundamental nature of
everything) and also a thesis about what this physics will be like (it will not contain, say,
acts of fundamental consciousness). Physicalism as such, it is argued, is not trivial, since
it could be false if the true and complete physics does not account for everything or if the
true and complete physics is not a successor to current physics.
Despite appearances, the notion of the physical we have arrived at is actually not
beholden to physics. In the final analysis, the reason it provides the thesis of physicalism
with content is because it excludes the mental from the ultimate dependence base. And
the question of whether this dependence base includes fundamental mentality is distinct
from the question of whether it is exhausted by (p. 185) the phenomena posited by
current, future, or final physics. The question being posed by such compromise views,
then, is not whether mental properties are determined by the properties posited by
physics but, roughly, whether mental properties are determined by non‐mental
properties.22
Whereas these compromise views arrive indirectly at a notion of the physical as, roughly,
the fundamental non‐mental and whatever is determined by such phenomena, a number
of philosophers have suggested this understanding of the physical directly. For example,
the idea that our central concern in the mind–body problem is whether mentality is
Page 11 of 18
What is the Physical?
fundamental is captured by Jerry Fodor's claim that ‘if the semantic and intentional are
real properties of things, it must be in virtue of their identity with (or maybe
supervenience on?) properties that are themselves neither intentional nor
semantic’ (1987: 97). Joseph Levine defends this take on the mind–body problem as well,
formulating the thesis of physicalism as: ‘[O]nly non‐mental properties are instantiated in
a basic way; all mental properties are instantiated by being realized by the instantiation
of other, non‐mental properties’ (2001: 21). David Papineau suggests that we can think of
the physical as ‘the “non‐mentally identifiable”—that is, as standing for properties which
can be identified independently of this specifically mental conceptual apparatus’ (2002:
41).23 And in trying to formulate a physicalism that allows for the possibility of no
fundamental properties, I have argued elsewhere that physicalism is true if all properties
are eventually determined by non‐mental properties such that all further determinations
of these properties, if any, are non‐mental, and that physicalism is false if there are some
properties that are eventually determined by mental properties such that all further
determinations of these properties, if any, are mental Montero (2006).
I think that taking the fundamental physical as the non‐mental provides a useful starting
place for understanding the central debates in philosophy of mind that I mentioned at the
start of this paper. Zombies can be understood as creatures that are just like us in every
non‐mental respect. Mary, we should say, knows all the relevant facts about the non‐
mental aspects of colour vision. And the causal closure of the physical should be
understood as the causal closure of the non‐mental. This understanding of the physical
also gives us a starting point in formulating an interesting thesis of physicalism in
general. And the reason for this, at least as I see it, is that a central point of contention
between physicalists and anti‐physicalists is whether human beings, and perhaps also
other animals, have, in some way, a special place in the world. One way we would seem to
be special is if mental phenomena were part of the original brew that was set in motion,
as one creation story goes, in the big bang. This would seem to give us a place of
prominence, since it would hint at a world created with us in mind; that is, it would
suggest, as another creation story goes, that when God created the world, she also
created minds. Another way we would seem to be special is if mental phenomena were
added as something extra somewhere (p. 186) along the way. Thus, it seems that
physicalism should rule out fundamental mental phenomena.
But we have not arrived at the final word, since taking the non‐mental as our dependence
base may not exclude all that physicalists would like to exclude. For example, a world in
which numbers form the basic fabric of reality might not be a world in which the mental
forms the basic fabric of reality, yet some might see such a world as antithetical to
physicalism. And the same could be said of fundamental norms: a world where certain
actions are immoral and that's that might make many physicalists uncomfortable. These
examples illustrate a problem not just for formulating physicalism in general but even for
formulating physicalism with respect to the mental; for excluding the mental from the
dependence base is consistent with, say, mental phenomena being ultimately determined
only by normative phenomena.24 However, it is not too difficult to see how one can
reformulate the theory so as to exclude numbers, norms, or whatever else one is
Page 12 of 18
What is the Physical?
interested in excluding: rather than the mental/non‐mental contrast, we could employ the
numerical/non‐numerical, the normative/non‐normative contrast, or a combinations of
these instead.
Yet, one might be tempted to ask, what is it about the fundamental mental, the
fundamental numerical, the fundamental normative that makes them all antithetical to
physicalism? If we could answer this question we would be able to arrive at a unified
notion of the physical that could make sense of questions about whether anything at all,
including irreducible numbers, norms, and minds, is physical. I think there is no answer
to this question save for the rather disappointing one that they tend to make some self‐
called physicalists uncomfortable. But we shouldn't take this disappointment too much to
heart, since as long as we specify what is to be excluded from the dependence base, we
can make sense of questions about the ontological status of norms, numbers, minds, and
so forth (or at least I have not shown that we can't). As such, physicalism as a unified
general thesis dissolves, while the mind–body problem awaits a solution.
I would like to thank Jonah Goldwater, Joel David Hamkins, and Graham Parsons for their
comments on earlier versions of this paper.
References
Armstrong, D. (1991), ‘The Causal Theory of Mind’, in D. Rosenthal (ed.), The Nature of
Mind (Oxford: Oxford University Press).
Crane, T., and Mellor, H. (1990), ‘There is No Question of Physicalism’, Mind, 99: 185–
206.
Crook, S., and Gillett, C. (2001), ‘Why Physics Alone Cannot Define the “Physical”:
Materialism, Metaphysics, and the Formulation of Physicalism’, Canadian Journal of
Philosophy, 31: 333–60.
Daly, C. (1998), ‘What Are Physical Properties?’, Pacific Philosophical Quarterly, 79: 196–
217.
Page 13 of 18
What is the Physical?
Davidson, D. (1970), ‘Mental Events’, in L. Foster and J. Swanson (eds.), Experience and
Theory (Amherst, Mass.: University of Massachusetts Press), 79–101.
Haldane, E., and Ross, G. H. T. (1968) (eds.), Descartes's Collected Works (Cambridge:
Cambridge University Press).
Hellman, G. (1985), ‘Determination and Logical Truth’, Journal of Philosophy, 82: 607–16.
Hume, D. (1783/1998), ‘On the Immortality of the Soul’, in S. Copley and A. Edgar (eds.),
David Hume Selected Essays (Oxford: Oxford University Press), 324–32.
Jackson, F. (1986), ‘What Mary Didn't Know’, Journal of Philosophy, 83: 291–5.
Kant, I. (1781–7/1929), Critique of Pure Reason, trans. N. Kemp Smith (New York: St.
Martin's Press).
Levine, J. (2001), Purple Haze: The Puzzle of Consciousness (Oxford: Oxford University
Press).
Page 14 of 18
What is the Physical?
Melnyk, A. (1997), ‘How to Keep the “Physical” in Physicalism’, Journal of Philosophy, 94:
622–37.
Nagel, T. (1986), The View From Nowhere (Oxford: Oxford University Press).
and Reason: Philosophical Papers, iii (Cambridge: Cambridge University Press), 205–28.
Quine, W. V. (1981), Theories and Things (Cambridge, Mass.: Harvard University Press).
Smart, J. J. C. (1959), ‘Sensations and Brain Processes’, Philosophical Review, 68: 141–56.
Snowdon, P. (1989), ‘On Formulating Materialism and Dualism’, in J. Heil (ed.), Cause,
Mind and Reality: Essays in Honour of C. B. Martin (Dordrecht: Kluwer Press), 137–58.
Swinburne, R. (1994) ‘Body and Soul’, in R. Warner and S. Tadeusz (eds.), The Mind–Body
Problem: A Guide to the Current Debate (Oxford: Blackwell), 311–16.
Unger, P. (1999), ‘The Mystery of the Physical and the Matter of Qualities: A Paper for
Professor Shaffer’, Midwest Studies in Philosophy, 23: 75–99.
Page 15 of 18
What is the Physical?
Notes:
(1) Thoughts about the proper scope of physicalism vary; some take it to apply to
everything whatsoever, others restrict it to the empirical world, others to the causal
world, or the contingent and/or causal, and still others to the concrete realm.
(2) Dennett, for example, denies the coherence of the thought‐experiment, since he
denies that we can form any coherent thoughts about what follows from knowledge of all
the relevant physical facts. See, for example, Dennett (1991: 398–401).
(3) See Montero (2003) for further discussion of the causal closure of the physical.
(4) Although Smart represents this as a ‘confession of faith’, Melnyk (2003) develops an
extended argument for physicalism taking the success of the physical sciences as his
starting point.
(5) Moreover, when one's purpose is to argue against the view that the mental is physical,
one needs at most a necessary condition that one can show does not hold of the mind, and
when one's purpose is to argue for the view that the mental is physical one needs at most
a sufficient condition that one can show does hold of the mind.
(6) Moreover, it would seem that that which was ‘before’ the big bang would not be
spatio‐temporal, yet could be physical.
(7) For further discussion of this view and its historical precedents see Strawson (2003).
See also Unger (1999).
(8) Although most take panpsychism to be inconsistent with physicalism, not all do. See,
for example, Strawson (2006) and Dowell (2006).
(9) Melnyk (1997) dubs the problem Hempel formulates ‘Hempel's dilemma’.
(10) Going further, one might say that inasmuch as physics strives for absolute precision,
all current physics falls short of accuracy. Or, even worse, one might say that since
current physics is probably inconsistent, it is complete for the unsatisfactory reason that
from a contradiction one can prove anything.
(11) For example, is current physics better than what we can call ‘physics‐plus’; that is,
better than current physics plus one additional kind of quark? If, as it seems, it is likely
that new kinds of quarks will be discovered, and, as it also seems, that physics‐plus
Page 16 of 18
What is the Physical?
counts as a relevant rival, then M‐physicalism is not better than its relevant rivals. See
Crook and Gillett (2001) for objections along these lines. Melnyk responds that physics‐
plus and variations thereof can count as similar enough to current physics not to be
problematic for his view. As such, Melnyk, as I see it, comes closer to proposing a
compromise view, which I discuss later, than to accepting the first horn of Hempel's
dilemma—though it is a compromise in which physics still plays a role.
(15) Of course, not all properties the physicist appeals to are relevant: when a physicist is
explaining a proposed budget in a grant application or explaining to her supervisor why
she was late for work, she may be appealing to very different properties than when she is
applying her mathematical skills in computing a wave function. But perhaps this
distinction is intuitive enough. See Poland (1994) for discussion of this point.
(16) Terry Horgan also explains physicalism in these terms; as he sees it, physicalism is
the view that ‘humans are, or are fully constituted by, entities of the kind posited in (an
ideally completed) physics’ (1994: 472).
(19) Note that, confusingly enough, in the literature the structural/relational aspects of
the world are sometimes taken to exhaust the physical aspects of the world, as in
Chalmers (1996), while at other times the idea that physics tells us only of the structural/
relational aspects of the world is taken to show that we cannot know the essence of the
physical world (in the sense that the physical world has an intrinsic, non‐relational
aspect). See, for example, Russell (1927/1992) and Unger (1999). Additionally confusing
is that the view that everything we know about the world is structural/relational is
sometimes taken to support idealism. See, for example, Kant (1781–7/1929).
(20) I take it to be merely a terminological matter whether we should call the view a
replacement for the notion of the physical or an interpretation of it.
(21) For views along these lines see Papineau (1993), Poland (1994), and McLaughlin
(2001). Alternatively, one can drop the ‘true and complete’ requirement and turn the
question of physicalism into the question of whether a physics very much like ours but
somewhat improved provides the fundamental dependence base for the mental, as do
Lewis (1983) and Melnyk (2003). But this is uncomfortably close to the first horn of
Hempel's dilemma, because it seems that such a physics will not provide the fundamental
Page 17 of 18
What is the Physical?
dependence base for the mental, since it will, most likely, have neither a complete nor a
correct account of the fundamental properties of nature.
(23) One problem, however, with this formulation is that, as Davidson pointed out, even if
there were non‐physical events, we might be able to identify, or refer to, them in non‐
mental terms (1970).
(24) Brandom (1994) argues that the mental is grounded in the normative (which, as he
sees it, is ultimately grounded in the social).
Page 18 of 18
Idealism
Print Publication Date: Jan 2009 Subject: Philosophy, Metaphysics, Philosophy of Mind
Online Publication Date: Sep 2009 DOI: 10.1093/oxfordhb/9780199262618.003.0011
Arguments specifically for and against idealism are arguments about the nature of the
physical world, not about the nature of mind. This creates a problem for a would-be
author of an entry on idealism, in a collection on the philosophy of mind, if he wishes to
do justice both to the topic of his chapter and to the theme of the book as a whole. This
article tries to resolve this conflict. Arguments for idealism usually take the form of
attempted refutations of physical realism. They therefore tend to take the form of trying
to prove that, in one way or another, the existence of material objects depends directly on
the existence and activity of minds, other than that of the divine creator in his act of
creating.
Keywords: idealism, physical world, nature of mind, philosophy of mind, physical realism, material objects
10.1 Introduction
Page 1 of 20
Idealism
mind is concerned, the idealist need have no different theory from the dualist. Arguments
specifically for and against idealism are arguments about the nature of the physical
world, not about the nature of mind. This creates a problem for a would‐be author of an
entry on idealism, in a collection on the philosophy of mind, if he wishes to do justice both
to the topic of his chapter and to the theme of the book as a whole. In so far as I can
resolve this conflict, how I shall try to do so will emerge at the end of the next section.
Like many terms in philosophy, ‘idealism’ is used in more than one sense, and some of
these are looser than others. The minimal idea is that the physical world is dependent on
mind in a more‐than‐causal sense. I say in a more‐than‐causal sense, because any
orthodox theist would claim that the physical world causally depended on God (p. 190)
(who, for these purposes, can be thought of as a mind), but most theists would not regard
themselves as idealists. This is because they think that, though the world is caused and
sustained in existence by a mind or spirit, the world itself, in its material part, is
essentially non‐mental. By contrast, what is required for idealism is that the physical
world be, in some sense, itself mental. The two main versions of this are: (i) that the
physical world exists only as a complex feature of experience; it exists only ‘in the minds
of’ those who do or might experience it: (ii) that the physical world itself is a mind, or
consists of minds. The idealisms of Berkeley and Kant are examples of the former,
because they have the world as constructed from the sensory contents of individual
perceivers. Leibniz and Hegel are in the second group. For Leibniz it is as if every space–
time point in the physical world were itself a conscious subject: for Hegel, the whole
world is one conscious intelligence developing in its self‐understanding.
The principle contrast with idealism is physical realism. Physical realism is the view that,
putting aside any role God might have as creator, matter and ordinary material objects—
those that are not normally thought to be conscious or to think—could exist, in principle,
without there being any minds. The physical world is mind‐independent, such that there
could be a time when there was matter and no minds, or a physical world in which minds
never evolved.
Arguments for idealism usually take the form of attempted refutations of physical realism.
They therefore tend to take the form of trying to prove that, in one way or another, the
existence of material objects depends directly on the existence and activity of minds,
other than that of the divine creator in his act of creating. It is not surprising, therefore,
that arguments for idealism tend to focus on one of those aspects of the mind which are
held to be difficult for the materialist to accommodate, and then to proceed to show that
the nature and existence of the physical world depend on its relation to one of those
aspects. This, I think, is the connection between arguments for idealism and the main
body of the philosophy of mind.
Page 2 of 20
Idealism
It is a commonplace that there are two features of the mind that create problems for the
materialist. One of them is consciousness, as paradigmatically represented in sensations
and perceptual experiences. This is problematic from a materialist perspective because of
the strangely subjective or private nature of its states. The other is thought, because it
has the mysterious property of intentionality, in virtue of which it is able to be about
something other than itself. ‘Aboutness’ is not a normal physical relation. Arguments for
idealism tend to focus on one of these features of the mind and argue that our conception
of the physical cannot be disentangled from the mental feature in question, in the way
that would be required if physical realism were correct and the physical world were a
genuinely mind‐independent reality. My procedure in this chapter, therefore, will be to
consider, first, the argument that the physical world cannot be made independent of our
sensory consciousness, and, second, that it cannot be set beyond our modes of thinking
about it.
Consciousness
We naturally think of the physical world as possessing the kinds of properties that our
senses perceive it as having. A serious problem arises, however, when we realize that
once those qualities that are sense‐dependent (and, hence, consciousness‐dependent) are
discounted, the resultant conception of matter is attenuated to the point of emptiness.
The argument proceeds as follows.
The traditional list of sensible properties centres on what are usually known as primary
qualities and secondary qualities. The secondary qualities—such properties as colours,
sounds, smells, and tastes—have often been held to be perceiver‐dependent, and not part
of the wholly mind‐independent world.1 But even if they are construed in a naive‐realist
manner, few philosophers would think them suitable to be taken as constituting the core
of matter. The causal powers of matter, when explicated scientifically, are always put
down to the primary qualities of objects. This leaves us with the primary qualities, as the
materials from which to build up our concept of matter.
From the seventeenth‐century atomist tradition, these primary qualities are the spatial
properties, such as shape, size, motion through space, and solidity. In effect, this means
that Newtonian atoms are volumes of solidity. Solidity is essentially the capacity or power
to resist penetration by other bodies. Hume (1739–40: I.iv. 4/1978:229) objected to the
idea that material bodies were volumes of impenetrability on the grounds that this left us
Page 3 of 20
Idealism
with the concept of a body as something that resists penetration by other bodies, and that
this is circular because the concept one is trying to define appears in the definition.
Conceptual criticism of the notion of inelastic, solid atoms, and the increasing scientific
importance of notions such as energy and charge pointed away from an ontology of solid
atoms to one of forces, energy, or fields; that is, to an ontology of powers, or entities
defined in terms of what they are disposed to do.2 Properties of these kinds are the
primary qualities of modern physics. Rom Harré, drawing on the theories of the
eighteenth‐century Croatian priest and diplomat Boscovitch, expresses the general form
of the powers ontology as follows:
(1970: 313)
(p. 192)
The ultimate entities of the world, as we can understand it, must be point sources
of mutual influence, that is centres of power distributed in space
(1970: 308)
Although the ontology of powers can be seen as an alternative to the Newtonian conception of
body, for our purposes the difference is irrelevant, because solidity, conceived of as
impenetrability, is itself a power—the power to resist penetration. Not surprisingly, therefore, an
accusation of vacuity similar to that deployed by Hume can be brought against any pure powers
ontology. In outline the case is as follows.
A power is a power to produce some effect, but if everything is a power, it is the power to
produce another power (presumably by modification of a power entity already present,
not by creation of a power from nothing). Why this leads to a regress can be seen as
follows. Let us call the first power A. We only know what A is if we know what kinds of
thing the actualization of its potentiality gives rise to. In other words, we only know what
A is if we know what it is a power to do, what states would constitute its manifestation.
Let us call the power which A is the power to produce B. So what A is, is the power to
produce B. But this is not informative unless we know the nature of B. B, being a power, is
the power to produce some further power state, call it C. (We could close the circle and
say that it is the power to produce a power of kind A, but then we would be in a circle of
the kind to which Hume objected in his account of solidity.) It seems that we are moving
into a regress. You can understand what a power is only by reference to a further power,
of which you have no specific conception unless you know the power state which would
be its effect, and so on. And though I have stated this argument in terms of what we could
know, the argument is not essentially epistemological. One could equally well say that
what the nature of A is depends on what it is a potentiality for; for what a power is, is
given by what it is a power to do. What it is a power to do is a function of what would
Page 4 of 20
Idealism
constitute its manifestation, and if the nature of this latter can have no determinate
expression, neither can the power which is defined in terms of it.
If the physical realist is committed to the powers conception of matter, and if that
conception is radically defective, then the physical realist is in trouble. Furthermore, his
opponent will point out that, if the pure‐powers conception is vicious, then the powers
that are supposed to constitute matter must produce something which is not itself a
power, but a monadic quality. And such qualities are to be found as sensible qualities in
the sense fields of perceivers. So the physical world is a structured capacity to give rise
to experience, and this is an idealist conception.3 We must look more closely at the
regress argument to see whether it is really vicious.
Disputes about whether a particular regress or circle is vicious tend to be between those
who say that so long as one can always name the next element along the line, the regress
is not vicious, and those who say that unless the process can be completed, (p. 193)
nothing contentful has been proposed. There are regresses of both kinds and the problem
is to decide into which camp a particular regress falls.
The crucial issue seems to be whether the content of the early members of the series
depends on the content of the later. Here is an example of a regress where the content of
the earlier members does not depend on the later: if it is true that p, then it is true that it
is true that p, and so on. No one, so far as I know, thinks that this is a vicious regress. The
reason is that neither the content of nor our understanding of the content of the assertion
that p is true depends on the later elements in the list. No one would suggest that you
cannot understand what it is for p to be true unless you already understood what it is for
it to be true that it is true that p.
Page 5 of 20
Idealism
It might be productive to compare the regress of powers with one which, unlike pi or the
regress of ‘is true’, is genuinely controversial. This is the regress of causes and the kin
issue of whether the world could be infinitely old. There are those who think that the
regress constituted by causes stretching infinitely back into the past would be vicious and
those who think that it would not. Notice that here there is no question of the nature of
any event or entity—say a present one—depending conceptually on the nature of the
events or entities that went before. So suppose that Aristotle were right and the world
were infinitely old and humans had always existed. The essence of what I am—namely, a
human being or a rational animal—would not involve essential reference in a regressive
way to the previous members of the series. The definition can be given complete and well‐
formed for contemporary humans in their own right. The controversy about a regress of
causes relates to the legitimacy of actual infinities, not to anything about the nature of
the things that are supposed to participate in the infinite series. This case is, therefore,
radically different from the (p. 194) regress of powers, and defenders of the pure‐powers
ontology consequently cannot seek any solace by appeal to a comparison with the dispute
concerning an unending regress of causes.4
(1) The most enigmatic is Simon Blackburn's (1990). Blackburn's initial response to the
powers ontology is positive. He says of this picture of the world: ‘Is this the way it has to
work? I believe so’ (1990: 63). But he then considers a version of the ‘regress’ objection
to this theory, accepting that the regress can only be ended because ‘[c]ategoricity in fact
comes with the subjective view’. And he continues: ‘The trouble is that such events,
conceived of as categorical, play no role in the scientific understanding of the world’ (p.
65). This appears to be a complete surrender to the argument. Blackburn's only response
is to lapse into a sceptical detachment. After remarking that ‘I leave the issue in Hume's
hands rather than Berkeley's’ (p. 64 n. 7; which, I presume, is a way of refusing to admit
that it might have idealist implications), he concludes:
It almost seems that carelessness and inattention alone can afford a remedy—the
remedy of course of allowing ourselves to have any idea at all of what could fill in
space.
Page 6 of 20
Idealism
(p. 65)
This sceptical hauteur does not seem to me to help the realist come to grips with the challenge
that the argument presents.
(2) Stephen Mumford thinks that the argument rests on a view of dispositions which ‘is
akin to Ryle's empiricist analysis where the ascription of a disposition is nothing more
than affirming the truth of a subjunctive conditional’ (1998: 33). He is attacking the
argument specifically as it is stated in Robinson (1982), but if the criticism is to have any
general application, it must be an objection to any attempt to define powers or
dispositions in terms of the states that constitute their actualizations, meaning by that the
states brought about when they are activated. Mumford spells out his objection:
We need not look for the actualization of a disposition solely in its manifestation,
however, for we understand dispositions to be actual whenever they are ascribed.
This is the realist alternative that I will be defending. If we were to treat
dispositions as actual properties that (p. 195) play a causal role in their
manifestations, then we can understand why dispositions are actual even when
not currently manifested.
(1998: 33–4)
Mumford's objection only works if essentially characterizing a power in terms of its actualization
is equivalent to adopting the reductive account of powers and dispositions. Surely, however, the
nature or essence of anything which is a pure power must reside in that which it is a power to
do, whether one conceives powers realistically or reductively. I think that Mumford is misled by
a possible equivocation in the sense of ‘actual’ in this context. The fact that a realist believes
that a power is actual, in the sense of actually existing when it is not activated, does not mean
that its nature is not expressed by reference to its actualization, in the sense of the
manifestation of the power. Even when dormant, it is essentially the power or disposition to give
rise to that actualization.
(3) George Molnar (2003: 173 ff.) titles the regress argument ‘always packing, never
travelling’. Although not a defender of the pure‐powers ontology, he does not believe that
it can be refuted a priori, as the regress argument purports to do. He has two main
responses to it. His first objection has some similarity to Mumford's objection considered
above, in that it concerns the connection between the real existence of powers and the
truth of counterfactual conditionals. When Blackburn states his version of the powers
ontology, he does so in a way that seems to imply that powers are either equivalent to or
at least imply the truth of conditionals. (He says, for example: ‘An electrical field can
abide, certainly, but that just means that there is a period of time over which various
counterfactuals are true’; 1990: 63; my emphasis.) Molnar points out that there has been
considerable literature in recent years devoted to casting doubt on whether dispositions
can be analysed conditionally, or even whether they entail conditionals (2003: 176–7). But
the argument as I have stated it above does not make explicit appeal to conditional
statements, only to the essentiality to a power of what would constitute its actualization
or manifestation. No one would deny that the nature of powers and dispositions
Page 7 of 20
Idealism
essentially involves what would constitute their manifestations, and any problem derived
from difficulties in formulating the exact connection between powers and conditionals
that goes beyond this is not relevant.
Molnar has a further objection directed against C. B. Martin's statement of the regress
argument against pure dispositionalism:
Martin has also suggested that leaving out physical qualia at the level of the
fundamental entities results in a Pythagorean ontology in which all is numbers,
quantities, ratios and proportions, but there is no whatness, no quiddity, nothing
that the numbers are numbers of and quantities of. Except of course further
numbers, which returns us to the regress generated by pan‐dispositionalism. I
think Martin's worry can be assuaged. If the property of exerting a certain force is
a definite something that the numbers can measure, so is being the source of that
force. That about the object that makes it a source of a force is a (quantative)
power property. It is open to the dispositionalist to say that this is where the
quiddity lies, this is what the numbers are numbers of.
(2003: 179)
(p. 196) The problem with this response is that it is hard to see what content one is supposed to
attach to the italicized words in the claims that ‘the property of exerting a certain force is a
definite something’ or that there is a ‘[t]hat about the object that makes it a source of a force’, if
all the properties are pure powers. Molnar is talking as if that which lies behind the force is
something other than the force itself, considered as potential or dormant, which on a pure‐
powers ontology it is not. Assigning a numerical value to this potentiality clearly adds nothing
relevant. Of course, it would make a difference if one could consider powers as Janus‐faced, by
giving them some other nature additional to the capacity to act. I now want to turn to
considering various ways that the physical realist might do this.
The dispositions, capacities, and powers of complex objects are grounded in and
explained in terms of the properties of the more minute structures that make up such
complex objects. The problem that concerns us is with the powers of the ultimate
elements of physical reality. If a pure‐powers ontology is not coherent, how should a
physical realist think of basic powers as being grounded?
Solidity is the alleged quality, the possession of which is responsible for the fact
that two material things cannot occupy the same place at the same time and is
logically connected with impenetrability, the power to resist penetration, in that
Page 8 of 20
Idealism
the possession of the former is supposed to account for the manifestation of the
latter.
(1970: 305)
Martin at one time held a similar theory, with any intrinsic quality treated as ‘a two‐sided
dispositional qualitative coin … The dispositional and the qualitative are equally basic and
irreducible; there is no direction for one being basic in a property and the other “supervenient”
’ (1997: 216; citing his previous view). He moved, however, from this Janus‐faced theory to one
according to which the dispositional and the qualitative are not different aspects of a property,
but are identical.
the qualitative and dispositional are identical with one another and with the
unitary intrinsic property itself … What is qualitative and what is dispositional for
any property is less like a two‐sided coin or a Janus‐faced figure than it is like an
ambiguous drawing. A particular drawing, remaining unitary and unchanged, may
be considered one way as a goblet‐drawing and differently considered, it is a two‐
faces‐staring‐at‐one‐another‐drawing.
(1997: 216–17)
The analogy does not look helpful. What is interpreted as a goblet or as two faces is neither of
these things, it is a line or lines on a piece of paper. As there is clearly nothing accessible which
is seen as a disposition and seen as a quality, what is really present must be some mysterious
third unknown.
Such a mysterious unknown plays a part in Harré’s attempt to cope with the regress.
Rom Harré played a major role in popularizing the powers ontology in the (p. 197)
philosophy of science, and in drawing attention to the importance of Boscovitch in the
historical development of this conception of matter (Harré 1970; Harré and Madden
1975). It is interesting to see, therefore, how he copes with the problems that face it.
the dilemma that confronts anyone who tries to universalize the dispositionalist
account of properties. It seems as if one must choose between the inelegant
alternative of grounding science on ungrounded dispositions, and the alarming
prospect of an indefinite regress of groundings.
Given that one is dealing with the ultimate nature of reality, it is not surprising if it should turn
out to be unfamiliar in a disturbing way; so the force of ‘inelegant’ and ‘alarming’ is unclear.
Furthermore, it is hard to see what is inelegant about ungrounded powers, if the notion makes
sense. Nevertheless, Harré does acknowledge it as a problem that he wants to solve. The
immediate solution that he gives is as follows:
Page 9 of 20
Idealism
(1986: 296)
This solution, which he derives from Mach, looks rather like a form of phenomenalism, because
the subatomic world is being grounded, ontologically and not just epistemically, on the
observable. (I put aside the oddity of calling energy fields ‘occurrent rather than dispositional’.)
But later he develops the theory in a Kantian direction. The basic powers of matter are the
product of the interaction between ‘the ur‐stuff of the world’—which he prefers ‘to nickname …
“glub”, to avoid any of the metaphysical temptations that arise from the connotations of the
word “stuff” ’—and our observation apparatus.
Whatever this theory is, it is not straightforward physical realism. Harré compares his
‘glub’ to Kant's noumenal world, having no characteristics accessible to us. The theory,
therefore, seems to have more in common with Kantian idealism than with normal
physical realism.
A more empiricist option is to adopt the neutral monism that Russell once espoused
(1927). This is the line taken by Grover Maxwell (1978), Michael Lockwood (1989, 1993),
and Daniel Stoljar (2001). The bearers of the ultimate powers and the intrinsic qualities
of matter are those qualities that show themselves in conscious experience. These are in
fact the intrinsic nature of the matter of which our brains are made. This theory seems to
imply that the intrinsic nature of, for example, an electron, will be something such as a
blue quale or a slight itch. One might try arguing that it need not be a phenomenal entity
of such a developed kind, but a proto‐phenomenal quality. But what something is that is
not phenomenal but which when combined with others of a similar sort becomes
phenomenal takes us back to (p. 198) mystery. Furthermore, given the difficulty in
formulating a convincing version of the Janus‐faced account of quality and power, the
connection between the qualititive core and the causal powers would seem to be entirely
contingent and, hence, not explanatory.
Once it is acknowledged that secondary qualities are not suitable to constitute the
fundamental nature of matter, the physical realist has a problem in giving an account of
what matter is. The idealist has no such problem, because, on one understanding of
‘matter’, he simply denies its existence. But, nevertheless, he accepts the notion of being
material, which is not, for him, the possession of some elusive nature; rather it signifies
that the entity in question has a certain status within the common‐sense scheme for
understanding experience which we cannot help but apply. The idealist accepts our
common‐sense conceptual scheme for interpreting the world, and being material is,
Page 10 of 20
Idealism
within that scheme, being a mind‐independent occupant of space with causal powers; that
is, from an empirical, as opposed to a metaphysical, perspective, material objects are
extra‐mental.5
Arguments against physical realism based on the nature of thought are much more murky
than those we have just considered and, for historical reasons, are greeted with more
suspicion by analytical and empiricist philosophers. One can trace more than one route to
idealism via the nature of thought, but for our purposes it is best to start from Berkeley's
so‐called master argument in Principles of Human Knowledge:
But say you, surely there is nothing easier than to imagine trees, for instance, in a
park, or books existing in a closet, and no body by to perceive them. I answer, you
may so, there is no difficulty in it: but what is all this, I beseech you, more than
framing in your mind certain ideas which you call books and trees, and at the same
time omitting to frame the idea of any one that may perceive them? But do not you
yourself perceive or think of them all the while? This is therefore nothing to the
purpose: it only shows that you have the power of imagining or forming ideas in
your mind; but it does not shew that you can conceive it possible, the (p. 199)
objects of your thought may exist without the mind: to make this out, it is
necessary that you conceive them existing unconceived or unthought of, which is a
manifest repugnancy.
Page 11 of 20
Idealism
There are many things that can be said about this passage. In sympathetic vein, Christopher
Peacocke (1985) has taken it as making the essentially true point that we cannot imagine an
object simpliciter, but only imagine it as it appears from a certain viewpoint. He does not, of
course, claim that this is enough on its own to prove idealism. John Campbell (2004) also thinks
that Berkeley has identified a deep and serious problem, but one that forces one to direct
realism, rather than to idealism. More critical responses, however, often begin by pointing out
that Berkeley confuses perceiving something with conceiving or thinking about it: from the fact
that you cannot think about an object without thinking about it, it does not follow that one
cannot think about an object unperceived. The criticism of the argument from which I wish to
start is connected to this one. It is that, in one way, Berkeley misses the intentionality of thought.
He confuses the psychological content of the thought episode with the object that the thought is
about. Because, platitudinously, one cannot think of an object without thinking of it, Berkeley
concludes that what one thinks of must be internal to the act of thinking. No doubt it is because
he follows Locke in having an imagist theory of thought that he cannot properly grasp the idea
that a mental episode essentially refers beyond itself. I raise this criticism, however, only to
show its limitations.
Although the intentionality of language shows how thought refers beyond itself to objects
in the world, it does not so easily show how the kinds of facts that we impute to the world
are separable from our ways of thought. I think that many of the most suggestive of the
thought‐based arguments for idealism can be seen as founded on a single idea. This is
that our way of thinking about the world possesses certain logico‐grammatical features
which, on the one hand, could have no correlate in a mind‐independent reality, but which,
on the other hand, are ineliminable from the world as it is represented in our
understanding. A good point to begin to appreciate this line of thought can be found in
Strawson's response to Austin on the correspondence theory of truth:
What ‘makes the statement’ that the cat has mange ‘true’, is not the cat, but the
condition of the cat, i.e., the fact that the cat has mange. The only plausible
candidate for the position of what (in the world) makes the statement true is the
fact it states; but the fact it states is not something in the world [my emphasis]. It
is not an object; not even … a complex object consisting of one or more particular
elements … and a universal element … I can (perhaps) hand you or draw a circle
round, or time with a stop‐watch the things or incidents that are referred to when
a statement is made. Statements are about such objects; but they state facts.
(1971: 195)
Page 12 of 20
Idealism
At first sight it might seem that Strawson is simply being misled by the double sense of ‘fact’,
when it can mean either true proposition or state of affairs, the former being ‘on the side of
language’ and the latter ‘on the side of the world’. Surely the condition of the cat—the property
instances it possesses—must be just as real as the (p. 200) cat. But even thinking in terms of
states of affairs, there are many such states which seem fundamental to our thinking about the
world and which do not seem open to a simple realist construal. There are many true and
important kinds of facts about the world which seem not only to be a function of how the world
coldly and simply in its own right is, but to be a product of how the world is and of a judgement
or assessment of how the world is. A simple illustration of this is as follows. The easiest kind of
fact to take realistically is one that concerns the location of a particular object in space and time.
Thus the thought or proposition ‘There is a pen on the table’, if true, simply reflects a situation
in the world. It is made directly true by what there is in concrete reality. But negative
propositions do not ‘picture reality’ in the straightforward way that some positive assertions do.
The proposition ‘There is no pen on the table’ is not so simple as its positive counterpart.
Absences are not part of concrete reality. A fact of an absence is more like an assessment of a
situation with a certain thought or perspective in mind—in this case, a concern about a pen or
pens. It is grounded in the concrete situation—what the situation positively is on the table's
surface—but it is not a representation of that alone. If one wishes to include negative facts in
one's ontology, a simple physical realism does not seem to be available.
Even the category of states of affairs is more problematic than it may seem. David
Armstrong believes that the world includes states of affairs and that everything in the
world is spatio‐temporal; hence, that states of affairs are spatio‐temporal. Indeed, anyone
who wishes to include states of affairs within his physical ontology, realistically
conceived, would seem to have to deem them spatio‐temporal. But it is not clear that they
have spatial location. Perhaps it is easy enough to assign location to the states of affairs
of macroscopic objects' possessing the traditional primary and secondary qualities. Thus
a ball's being round will be where the circumference of the ball is, and its being red
where its surface is. With more arcane properties, the issue is harder. A particular
electron has location, but where is its having spin or its having mass? Are they
everywhere the electron is? Relations also create further problems. One could draw a
rough outline round the state of affairs of a's being six inches away from b: it would
include a and b, but what of the intervening space would it include? In Timothy Sprigge's
words:
(1983: 164)
Page 13 of 20
Idealism
(1) Our conception of the world, as shown in our language about it, possesses logico‐
grammatical complexity, and certain features of this complexity cannot (p. 201) be
treated straightforwardly as representations of aspects of concrete, mind‐
independent reality.
Therefore
(2) If there is a realist conception of the world, it must be possible to separate those
features of our conception of the world which can be treated in a straightforwardly
realist way and those that cannot. It is the former that constitute the realist
conception of the world, and the states of affairs they represent must be enough to
constitute the realist's world.
(3) It is not possible to make the distinction specified in (2).
Therefore
(4) A realist conception of the world is not possible.
Accepting that there is a prima facie case for (1), there are three ways of responding to the
argument. There are two ways in which one might deny (2). One of them is to deny that a realist
is restricted to having only concrete elements in his reality. Perhaps a physicalist or a naturalist
of a certain kind is committed to this, but a realist per se is not bound to deny that there is a
realm of more or less abstract facts, grounded in but going beyond the concrete core of his
reality. A second kind of objector to (2) regards the argument as wholly misconceived. This is a
response that might come from someone with sympathy for the ‘ordinary language’ school of
philosophy. Of course, this respondent will say, when one describes the world, the descriptions
possess the logico‐grammatical features of the language used. This is a trivial consequence of
using language to describe the world, which is something one cannot avoid. It does not carry
strange consequences for one's conception of the world itself.
The third response is to deny (3) and to take up the challenge to distinguish clearly
between those features of our conception of the world which are to be taken wholly
realistically and those which are not, and to explain how the former sustain the practices
which give rise to the latter. This, I take it, is the point of ‘truthmaker theory’, which tries
to explain how statements are made true by a world conceived as very much sparcer than
the way a literal reading of our statements would make it seem.6
There is not space in this essay to discuss all these responses adequately. I shall make
what I hope are suggestive comments about the first two, and then discuss the
‘truthmaker’ option in slightly more detail. I give it the more careful treatment because I
share the truthmaker theorist's intuition that his theory represents what the realist is
actually committed to.
My reservation about the ‘abstract facts’ strategy is easily stated. The suggestion that a
physical realist can accept abstract objects fails to distinguish between different kinds of
abstract objects. The realist can, at least if he is not a rigorous physicalist, accept
universals, numbers, and propositions, but can he accept that the facts that make certain
propositions about the physical world true are also abstract? It is one thing to allow the
Page 14 of 20
Idealism
proposition that there is no cat on the mat to be abstract, another to treat the fact that
makes it true as abstract.
(p. 202)
The claim that all the phenomena that have been cited merely reflect the trivial fact that
we talk about the world using language and therefore do so in a way that reflects the
properties of language ideally requires an extended discussion. I shall make only one
observation. It could be argued that the very objection brought against Berkeley's
argument works against this strategy. I said that Berkeley fails to give weight to the
intentionality of thought and language. But if one does give this weight, then one has the
resources to distinguish between thought and language on the one side and what they are
about on the other. Once having been allowed this distinction, it is difficult to see what
could prevent one from enquiring further into the relation between the logico‐
grammatical features of language and the question of what there is ‘out there’ to be its
object. If one denies the possibility of separating these two, it looks as if one is in danger
of surrendering to Berkeley's argument as it stands.
According to the truth‐maker theorist, for every statement with a truth‐value there is
something about fundamental reality that necessitates that that statement have the truth‐
value that it does. For a true statement, this is its truth‐maker. It is important that this is
something about the fundamental reality. One cannot generally articulate the truth maker
of a sentence simply by repeating it, as one can when stating its truth‐conditions. So the
truth‐maker for a statement such as ‘There is a hurricane approaching Florida’ will be
some complex state of affairs involving elementary particles moving in a particularly
rapid way over other particles that constitute a section of the Atlantic. The statement
about the hurricane will have the same truth maker as a statement describing the same
event in greater detail: they will both be made true by the same hunk of reality.
This still leaves open a choice between two conceptions of how this theory fits with
realism. They can be stated as follows.
(A) If realism is true, then if, for example, there is no cat on the mat, then it is true
that there is no cat on the mat even if there are no minds.
Therefore
If realism is true, then there must be both truthmakers and truthbearers for all true
statements.
Page 15 of 20
Idealism
The difference is that the latter does not commit one to the existence of truth bearers in the
absence of minds. The point is the following. We are inclined to say that, from the realist
perspective, the facts are what they are whether or not there are any minds to appreciate them.
From the perspective of the second theory, there is no fact that there is no cat on the mat in a
world without minds. In other words, when fact and truth maker are different, the fact will not
exist unless minds do. Does this matter?
(p. 203)
Suppose a certain random event (for example, the decaying of a particle) occurred at t,
but that it could have occurred at t′, and that had it occurred at t′ certain very different
things stretching into the future would, or could, have happened. If you are a realist
about possibilities, you will want to say that it is a fact that these things would, or could,
have happened, whether or not there are any minds. The truth makers for such
statements will be certain immediate dispositions or powers of actual objects. The
extrapolation to more remote future outcomes cannot be thought of as existing in these
actual states of affairs. This is shown to be true by the case in which, had the event
occurred at t′ not t, the future that ensued could itself have gone either of two ways
depending on some further random event. Let us express what actually occurred by p.
What might have occurred is q. If q had occurred, then it might have been followed either
by r or by s. The ‘might’ here is not epistemological, but is based on genuine
indeterminacy. So if p had not been true, but q had, then either q followed by r could have
been the case, or q followed by s. The truthmakers for either of those possibilities in the
actual world would be the same. This is so because if there is a circumstance C in which
there is a set of outcomes that are indeterministically possible, it is the same set of facts
that underlies the possibility or probability of any of those outcomes, because they
determine the range of probabilities as a whole. Nevertheless, it could be argued that the
realist and common‐sense intuition is that it is objectively true that there are the two
possible outcomes, whether or not there are any ordinary minds to appreciate this fact. If
this is the case then factual assessments and, hence, some form of mentality seem to be
presupposed by our common‐sense conception of the world.
Page 16 of 20
Idealism
If the arguments in this section have any force at all—and I think that this is a much more
open question here than it is about the arguments presented in section 10.2— then they
show that we have no notion of a world as it is in itself that is wholly disembrangled from
modes of conception that are definitely mind‐dependent. But what sort of idealism does
this give us? The natural upshot might appear to be that the world is dependent on our
way of thinking about it, as Kant claimed. Out of this idea, by dint of various obscure
moves, came the idealism of the Hegelian tradition.7 (p. 204) But there is an alternative
interpretation of the arguments presented in this section. This is that the existence of a
world depends on its relation to a mind, but not ours. This interpretation is Platonic or,
more precisely, Neoplatonic, rather than Kantian. According to this Platonistic
conception, there could not be an objective world if there were not also an objective
Intellect understanding it. Our intellect merely participates in this Divine Intellect. How
one might set about choosing between the Neoplatonic and Kantian‐Hegelian options is
an issue we have no space here to discuss.
References
Armstrong, D.M. (2004), Truth and Truthmakers (Cambridge: Cambridge University
Press).
Craig, W. L., and Smith, Q. (1993), Theism, Atheism and Big Bang Cosmology (Oxford:
Clarendon).
Foster, J. (1982), The Case for Idealism (London: Routledge & Kegan Paul).
Page 17 of 20
Idealism
Foster, J. (1993), The Succinct Case for Idealism, in H. Robinson (ed.), Objections to
Physicalism (Oxford: Clarendon) 293–313.
Lockwood, M. (1989), Mind, Brain and Quantum: The Compound ‘I’ (Oxford: Blackwell).
Martin, C. B. (1997), ‘On the need for Properties: The Road to Pythagoreanism
(p. 205)
Robinson, H. (1982), ‘Matter: Turning the Tables’, in Robinson Matter and Sense
(Cambridge: Cambridge University Press);
repr. in T. Crane and K. Farkas (eds.), Metaphysics: A Guide and Anthology (Oxford:
Oxford University Press, 2004), 107–19.
—— (1985)’, The General Form of the Argument for Berkelian Idealism’, in J. Foster and
H. Robinson (eds.), Essays on Berkeley (Oxford: Clarendon), 163–86; repr. in T. O'Connor
and D. Robb (eds.), Routledge Companion to the Philosophy of Mind (London: Routledge),
81–102.
Page 18 of 20
Idealism
—— (1993), James and Bradley: American Truth and British Reality (La Salle, Ill.: Open
Court).
Notes:
(1) For a discussion of arguments against realist views of secondary qualities see
Robinson (1994: ch. 3) and, for a more detailed statement of the arguments in the context
of a discussion of idealism, Robinson (1982).
(2) For a lucid account of this development see Harré and Madden (1975: ch. 9).
(3) This is slightly too swift. For a more detailed spelling out of this part of the argument
see Robinson (1982).
(4) For lucid discussion, from both sides, of whether actual infinities are coherent see
Craig and Smith (1993).
(5) There is one very important recent, and, I believe, successful, argument for idealism
which would fall under the rubric of section 10.2 which I have not the space to discuss.
That is John Foster's argument, as found in Foster (1982, 1993) and, in a slightly different
form, in Robinson (1985). I have nothing to add to the argument as found in those places.
(6) For a thorough working out of truthmaker theory see Armstrong (2004).
(7) I have not been able directly to discuss the Kant‐Hegel tradition in the way that might
be expected in a chapter with this title, but the arguments of section 10.3 are an analytic
philosopher's attempt to cast within his own terms what he can understand from that
tradition. Ewing (1934) is a great, but largely forgotten, work by a philosopher with all
the analytical skills and real sympathy with Hegelian idealism, which examines the
arguments that have been brought for that tradition. Sprigge (1983, 1993) are also works
wholly accessible to analytic philosophers, but which make a strong defence of British
Hegelianism.
Howard Robinson
Page 19 of 20
Panpsychism
Panpsychism endorses the co-fundamental status of matter and mind in so far as it allows
there are features of the world which are non-mental. Panpsychism is also not generally a
view in which mentality is taken as ‘substantial’. It is more natural to regard panpsychism
as expressing the view that, roughly speaking, everything exemplifies certain mental
properties. However, it is an important and distinctive claim of many panpsychists that
the ‘object/property’ metaphysics we take for granted is fundamentally mistaken and
must be replaced with another metaphysical vision of the basic structure of reality.
Keywords: panpsychism, status of matter, status of mind, mental properties, metaphysics, structure of reality
Page 1 of 16
Panpsychism
physical terms,1 but must agree that the physical is the ontological foundation of the
world and that mentality is in some way an expression of the physical. Reflection on this
division reveals an obvious third possibility, the ‘inverse’ of physicalism, in which
mentality is ontologically fundamental and the physical is in some way an expression of
mental reality. Such a view would be a kind of panpsychism but not a form commonly
encountered under that (p. 207) label. More frequently, panpsychism endorses the co‐
fundamental status of matter and mind in so far as it allows there are features of the
world which are non‐mental. Panpsychism is also not generally a view in which mentality
is taken as ‘substantial’. It is more natural to regard panpsychism as expressing the view
that, roughly speaking, everything exemplifies certain mental properties. However, it is
an important and distinctive claim of many panpsychists that the ‘object / property’
metaphysics we take for granted is fundamentally mistaken and must be replaced with
another metaphysical vision of the basic structure of reality.
Probably the best way to locate panpsychism within the mind–body problem is to see it as
the counter position to emergentist views of mentality (of which modern physicalist
theories are all examples).2 Modern common sense bridles at the claim that planets,
trees, rocks, atoms, or electrons possess mental attributes. In particular, the modern
scientific world‐view has it that all things are ultimately constituted of fundamental,
simple, and purely physical entities which possess a relatively small number of basic
attributes such as mass, charge, and spin. There is no place for and no need to postulate
these physical building blocks having any mental properties whatsoever. And yet it is
certain that some of the composite objects—such as ourselves—made from quarks,
electrons, and the other fundamental physical entities do exemplify mental properties.
Therefore, the modern scientific viewpoint has to endorse some form of emergentism.
While any conception of emergence requires that composite things have properties which
their components lack, the proper characterization of emergence remains controversial.
Theories range from the highly radical (such as those espoused by the so‐called British
emergentists; see McLaughlin 1992 for an excellent survey; see also Blitz 1992) to benign
accounts that see emergence as nothing more than complexity, perhaps of a high enough
degree to forestall all attempts at practical prediction and explanation. (On such
accounts, the weather is an emergent feature of the physical processes we call the
atmosphere; see Holland 1998.)
From the point of view of the panpsychist, the problem with emergence is that radical
forms seem highly implausible and no less mysterious and opposed to modern common
sense than panpsychism itself, while benign forms seem incapable of explicating the
generation of mentalistic features such as consciousness from the mere complex interplay
of the available physical features.
Panpsychism is a truly ancient doctrine that can probably be traced back to the animism
that seems to have been universally accepted by our distant ancestors, and which
(p. 208) we still see naturally and spontaneously occurring in our children. Philosophical
The rise of ‘Aristotelian Christianity’ through the Middle Ages was not conducive to
panpsychism, until in the Renaissance the doctrine regained its prominence (see Skrbina
2005: ch. 3). But it was the birth of the mechanical world‐view associated with the
scientific revolution that really forced the issue between emergence and panpsychism.
Galileo famously tried to sweep the problem of mind under the rug with his distinction
between primary and secondary qualities, but this could only temporarily put off the need
to integrate mind into the burgeoning scientific picture of the world. Of course,
Descartes's dualism was the most obvious approach: simply and totally hive off mind from
the physical world, permitting only such minor ‘leakage’ between the two realms as
necessary for free human action and sensory consciousness. Cartesian dualism is deeply
unsatisfactory, with its totally mysterious causal interaction between completely disparate
substances, especially as the interaction threatens some of the most fundamental
principles governing the physical world, such as conservation of energy. Philosophers
were quick to find alternative views of mind, and panpsychism figured in two of the most
prominent.
Page 3 of 16
Panpsychism
It was Spinoza's view that both mind and matter were but merely two attributes of an
underlying, infinite, and infinitely complex substance (which Spinoza notoriously
identified with God). Every material thing has its mentalistic aspect, and vice versa. As
Spinoza wrote: ‘[A] circle existing in nature and the idea of the existing circle, which is
also in God, are one and the same thing … therefore, whether we conceive (p. 209) nature
under the attribute of Extension, or under the attribute of Thought … we shall find one
and the same order, or one and the same connection of causes’ (1677/1985: prop. 7,
scholium). In terms of the core principles of panpsychism, Spinoza clearly holds that mind
is fundamental, though not uniquely so, and ubiquitous—there is nothing that is not,
when considered from the appropriate viewpoint, mentalistic in its nature.
Each monad contains within it a complete specification of the entire universe from a
particular viewpoint, expressed with more or less clarity. A monad that was incomplete
could fit into more than one possible world, putting God into an impossible dilemma
about which world to create in so far as He is governed by the principle of sufficient
reason. So why is it that we, for example, find ourselves ignorant about so very much?
To answer this, Leibniz deployed another novel (for the time) and important distinction:
that between conscious and non‐conscious mental states. The mental lives of most
monads are almost entirely unconscious, consisting of petite perceptions. Fully conscious
states are introspectible and form what Leibniz called apperceptions. Even monads with
rich conscious lives, such as ourselves, are aware of only a tiny fraction of our mental
states. From the point of view of articulating the forms of panpsychism, the introduction
of this distinction naturally bifurcates the theory, but also imposes two great conceptual
Page 4 of 16
Panpsychism
difficulties. The first is to give a characterization of the mental which captures its
essential mental aspect without using the notion of consciousness. The second is to
describe the relation between the unconscious and (p. 210) conscious mental states, and
this leads to fundamental problems which threaten to undercut the supposed advantage
that panpsychism has over emergentist views.
It must also be noted that Leibniz did not agree with Spinoza on the co‐fundamentality of
matter and mind. For him, the physical world was a ‘well‐founded illusion’ or logical
construction from the sum total of all monadic points of view. This is a kind of idealism,
but not of the usual sort (such as Berkeley's or the later so‐called absolute idealists). For
Leibniz allows that every thing which we regard as a material being (a speck of dust or a
single unobservable atom) has its corresponding mental aspect which grounds it. More
typical idealists regard objects such as dust motes as mere constructs of the conscious
states of fully‐fledged minds.
The nineteenth century was the heyday of panpsychism, though generally of the idealist
form. Idealism was the ‘received’ metaphysical viewpoint; materialism was but a minor
and disreputable pursuit. The founders of scientific psychology: Gustav Fechner, Wilhelm
Wundt, Rudolf Lotze, and William James, to name a few, were all panpsychists of one
stripe or another. Many embraced a double or multi‐aspect view of the world, in which
everything possessed both a mentalistic and physicalistic side, but the mental aspect was
frequently regarded as the more fundamental. (For pure idealists this goes without
saying.) James, for example, while endorsing a dual‐aspect view he called ‘neutral
monism’, added in a notebook entry written in 1909: ‘[T]he constitution of reality which I
am making for is of the psychic type’ (see Cooper 1990).
The tension between emergentist and panpsychist positions culminated in the early to
mid‐twentieth century with several sophisticated theories of emergence that strove to
integrate the incredible advances that were being made in the physical sciences into a
coherent metaphysical view of the world (see Morgan 1923; Broad 1925; and Alexander
1927; for a survey, McLaughlin 1992). These emergentists clearly made the crucial
distinction between epistemological and non‐epistemological forms of emergence. The
latter involves ‘merely’ the impossibility of our understanding in detail how complex
systems behave, even if we grant that their behaviour is completely determined by the
purely physical properties of the fundamental entities that constitute the world. The
former, often called radical emergence, entails the production of genuinely novel, causally
efficacious features of the world, stemming from the combination of fundamental
components. Writing at about the same time as these emergentists, Alfred North
Whitehead's ‘process philosophy’ (1929) represented the last and most sophisticated
development of panpsychism within the context of a complex, overarching, and
revisionary metaphysics. (For an introduction to Whitehead's panpsychism see Griffin
1998.)
Page 5 of 16
Panpsychism
But emergentism fell rapidly out of favour as the new quantum mechanics promised to
reveal how the heretofore primary example of uncontroversial emergence, chemistry,
could be in principle reductively understood in terms of basic physics. This development
led in turn to a vigorous renewal of materialist views of the mind, so that, somewhat
paradoxically, the death of emergentism was not the victory of panpsychism. Rather, both
accounts were supplanted by a vigorous and fruitful materialist research project in
philosophy seeking to duplicate, at a (p. 211) very abstract level, the successful treatment
of chemistry in the realm of the psychological. Just as chemical properties arise from
entities which entirely lack them, so too would mental properties be seen to arise from
entirely non‐mentalistic physical constituents. This philosophical project began with
‘logical behaviourism’ (see Carnap 1932/1933), proceeded to the psychoneural identity
theory (see Place 1956; Smart 1959) and has led to a host of successor physicalist
accounts of mind (see Kim 2006 for a survey). It has proved surprisingly difficult to
produce an acceptable version of materialism, however, and the problem of consciousness
has loomed recently as especially recalcitrant (see Chalmers 1996). In fact, the so‐called
hard problem of consciousness, the problem of explaining exactly how material systems
generate, realize, or constitute states which have phenomenal character (states for which
there is ‘something it is like’ to be in them) has seemed to some so difficult that a
renewed interest in more radical approaches, such as emergentism and panpsychism, has
appeared; and it is in this light that we ought to consider the arguments for and against
panpsychism.
Notice that this argument does not quite establish panpsychism, since it does not yield the
ubiquity principle.
Page 6 of 16
Panpsychism
Thomas Nagel presented a clear form of the argument, in which the linchpin principle (2)
is stated thus: ‘[T]here are no truly emergent properties of complex systems. All
properties of complex systems that are not relations between it and something else derive
from the properties of its constituents and their effects on each other when so
combined’ (1979: 182). Thus, the only coherent form of emergence is an epistemological
doctrine about the limits of our understanding of complex systems coupled with an
appreciation for the usefulness of high‐level explanatory systems which we (p. 212) must
deploy in the face of intractable complexity. However, it is not altogether clear why Nagel
denies that some of the ‘effects’ which arise from the combination of low‐level physical
entities are not or could not be radically emergent. The classical emergentists would have
agreed with the letter of Nagel's principle, but not the spirit. They allowed that
emergence was a lawlike feature of the world, but denied that it was the effect of
fundamental physical properties working by themselves. Instead, the world exemplifies
underivable ‘laws of emergence’ that govern the combinatory properties of physical
entities. As C. D. Broad put it, such a law ‘would be a statement of the irreducible fact
that an aggregate composed of aggregates of the next lower order in such and such
proportions and arrangements has such and such characteristic and non‐deducible
properties’ (1925: 78). It must be noted though that many if not all of the currently
fashionable views of emergence in the sciences of complexity seem to be consistent with
the purely epistemological reading of emergence (and to resist going beyond it), and to
that extent Nagel's argument carries some weight.
Recently, Galen Strawson (2006) has argued for the incoherence of radical emergence,
roughly on the grounds that the only kind of emergence of Y from X that makes sense is
one in which ‘Y is in some sense wholly dependent on X and X alone, so that all features
of Y trace intelligibly back to X (where “intelligible” is a metaphysical rather than an
epistemic notion)’.3
Perhaps the issue of the coherence of radical emergence comes down to the question of
whether it is possible to articulate a sense of dependence in which emergent features are
dependent on low‐level features but are not merely the product of these low‐level
interactions working ‘by themselves’. Opponents will deny that there is any legitimate
kind of dependence that can be explicated without showing how the low‐level features
have within themselves the power to produce the emergents, thus demoting radical
emergence to the merely epistemological.
Page 7 of 16
Panpsychism
everything’. The point of the thought‐experiment is that we know that computer programs
will provide outputs that depend entirely and only upon the nature of their coded
algorithms. If we code into the system only the principles of state evolution and
interaction of the fundamental physical features (p. 213) then the output will be just what
those features can give rise to according to the theory we are simulating. In a world
containing only epistemological emergence, we would expect that such computer
simulations would exactly mirror the behaviour we observe in the macroscopic world. But
if radical emergence was at work, we would expect an inexplicable divergence of real‐
world behaviour from the behaviour of the simulation.
A scenario in which we have good reason to suppose that our fundamental theories are
correct and in which simulated behaviour diverges from real behaviour seems perfectly
coherent. The fact that computers run the exact code given them obviates the issue that
computers themselves might have emergent properties. It seems possible that we would
opt for accepting a simpler fundamental theory with radical emergence rather than
necessarily declaring that the fundamental theory is mistaken merely because of the
divergence of simulated from actual behaviour. In such a case, we would have to add
certain ‘laws of emergence’ that would impose new constraints on the simulation when
certain complex configurations of the fundamental physical entities arise. We might find
that once the laws of emergence are in place within the simulation, no further divergence
in behaviour between the simulation and actuality ever occurs. Therefore, I think we can
give some real content to the idea that dependence can be preserved within a system that
endorses radical emergence.
However, while this model may legitimate the concept of radical emergence, it does not
go beyond addressing the mere existence and efficacy of emergents. It does not in any
way at all tell us how conscious states—that is, states with phenomenal character—could
emerge from the entirely non‐conscious. In fact, radical emergentists deny that such an
understanding is possible. Emergence must be accepted with, in Samuel Alexander's
phrase, ‘natural piety’ as a pure brute fact. While all theories rest upon some brute facts
(e.g. the values of certain physical constants) it is passing strange that the generation of
something so remarkable as consciousness should be brutely contingent on extremely
special configurations of numerous instances of precisely organized physical matter, such
as we find in the human brain's exquisitely networked 100 billion neurons. If the creation
of states of consciousness is contingent upon the formation of such astonishingly complex
and intricate physical states, it is hard not to believe that they occur because of the
intricate organization and hence can, in principle, be explained by it. But this would mean
that consciousness would turn out to be a merely epistemological emergent. Such a view
runs afoul of the obvious reply that ‘nothing comes from nothing’, and that all that the
intricate organization of fundamental physical entities can produce is intricate
organization and correspondingly intricate patterns of causal relations which are all
ultimately resolvable into the causal connections present at the most fundamental level of
physical reality. Patterns of organization, however, are manifestly not the same as
conscious experience, or, at least, the claim that certain very special and very complex
patterns of material organization are conscious states is precisely the kind of bizarre
Page 8 of 16
Panpsychism
brute fact we are trying to avoid. This line of thought is, in fact, taken to provide a rich
support (p. 214) for their theory by panpsychists. But before turning to that, we should
examine the more empirical version of the genetic argument.
The rapid success of Darwin's theory after its introduction in 1859 transformed debate
about life and mind. The genetic argument for panpsychism here rests on the idea that
evolution is a gradual process which slowly and incrementally modifies pre‐existing
features via natural selection. Yet the advent of consciousness into a world utterly devoid
of experience represents a giant discontinuity. The smallest, vaguest twinge of feeling is
something radically different from the properties of insensate matter. William Clifford put
it thus: ‘[W]e cannot suppose that so enormous a jump from one creature to another
should have occurred at any point in the process of evolution as the introduction of a fact
entirely different and absolutely separate from the physical fact. It is impossible for
anybody to point out the particular place in the line of descent where that event can be
supposed to have taken place’ (1874/1886: 266). Such considerations also moved William
James, who wrote that ‘we ought … to try every possible mode of conceiving of
consciousness so that it may not appear equivalent to the irruption into the universe of a
new nature non‐existent to then’ (1890/1950: 148). The empirical content of this
argument is that Darwinian evolution is the mechanism of emergence (at least within the
biological realm). But then it is clear that all emergence will be of the epistemological
variety, and will not suffice to account for the generation of consciousness.
Turning briefly to the argument from analogy, the basic idea is to find some feature of
matter which suggests—ideally strongly suggests—some fundamental similarity with
mentality. It is unfortunately difficult to come up with anything along these lines which is
at all convincing (see the remarks on Thales above). Some, including Whitehead, have
seen in the indeterminacy of modern quantum physics an echo of freedom of will, but the
pure randomness of quantum indeterminacy seriously weakens this analogy. Perhaps a
more promising avenue is the role of information in quantum physics. The entangled
states which express what Einstein derisively labelled ‘spooky action at a distance’
suggest that it is information rather than causal connection which grounds at least some
of the fundamental constraints at work in the physical world. One might then hope that
there is some kind of proper analogy between quantum information and the intentionality
of mental states. For example, mental intentionality is underived; that is, it does not
depend upon interpretation, convention, or other derivative methods of assigning
meaning. Presumably, the kind of informational connections lurking in the quantum world
are similarly underived. (This does not preclude quantum states having standard derived
intentionality via our interpretation of them, as in the pioneering use of quantum systems
to perform standard computations.) However, it is difficult to provide much real content
to the analogy, at least as things stand now. It is also conceivable that work on the
measurement problem in quantum mechanics will implicate consciousness at the
fundamental level (see Wigner 1962; Lockwood 1989).
Page 9 of 16
Panpsychism
Another argument in favour of panpsychism depends upon the idea of ‘intrinsic nature’.
This argument has close links to the genetic argument, but goes further in attempting to
spell out what makes the emergence of consciousness seem (p. 215) so preposterous.
Although the argument is far from novel, it has been recently revived and advocated in
Rosenberg (2004) and Strawson (2006).4 The argument assumes a distinction between
relational and intrinsic properties, which although intuitively acceptable is notoriously
difficult to spell out precisely, and in particular on the concept of dispositional properties
(which are a species of the relational). For example, we say that an electron has a
negative charge of about 1.6×10−19 Coulombs, but what this means is that the electron is
disposed to move in such‐and‐such a way in an electric field of such‐and‐such a strength.
The intrinsic nature of electric charge remains utterly mysterious. And yet it seems
reasonable to think that every dispositional property stems from underlying intrinsic
properties. Of these, with respect to the fundamental physical constituents of the world,
we know absolutely nothing, since physics deals only with the dispositional properties of
matter. This is a long‐standing position. Both Eddington and Russell, among others, agree
that ‘science has nothing to say as to the intrinsic nature of the atom’. (Russell called the
dispositional properties of matter ‘mathematical properties’.) This led Eddington to assert
further that we know nothing of atoms which ‘renders it at all incongruous that they
should constitute a thinking object’ (1927: 259) and to adopt a panpsychist understanding
of matter.
We might put the argument in another way, as follows. Matter must have an intrinsic
nature to ground its dispositional properties. We know nothing of this nature, and in fact
the only intrinsic nature with which we are familiar is consciousness itself5. It is arguable
that we cannot conceive of any other intrinsic nature because our knowledge of the
physical is entirely based upon its dispositions to produce certain conscious experiences
under certain conditions. Of course, we can assert that matter has a non‐experiential
intrinsic nature which is utterly mysterious to us, but this would seem to make the
problem of emergence yet more difficult. An emergentism which made the generation of
consciousness intelligible would be one that showed how experience emerged from what
we know about matter; that is, from its dispositional properties. But it seems impossible
to see how the dispositions to move in certain directions under certain conditions could
give rise to or constitute consciousness, save by the kind of brute and miraculous radical
emergence discussed above. If granting some kind of experiential intrinsic aspect to the
fundamental physical entities of the world eliminates this problem, it might be worth the
cost in initial uncomfortable implausibility.
Page 10 of 16
Panpsychism
A more serious objection turns the aggregate/unity distinction against the panpsychist. A
natural interpretation of the deployment of this distinction assumes that there is some
combinatory principle by which the simplest psychological features come together to
form the kind of complex minds we are familiar with. Of course, this is the problem of
emergence reappearing. William James, though eventually endorsing a panpsychist
philosophy, presented this objection to good effect:
Take a sentence of a dozen words, and take twelve men and tell to each one word.
Then stand the men in a row or jam them in a bunch, and let each think of his
word as intently as he will; nowhere will there be a consciousness of the whole
sentence. Where the elemental units are supposed to be feelings, the case is in no
wise altered. Take a hundred of them, shuffle them and pack them as close
together as you can (whatever that might mean); still each remains the same
feeling it always was, shut in its own skin, windowless, ignorant of what the other
feelings are and mean. There would be a hundred‐and‐first feeling there, if, when
a group or series of such feelings were set up, a consciousness belonging to the
group as such should emerge. And this 101st feeling would be a totally new fact;
Page 11 of 16
Panpsychism
the 100 original feelings might, by (p. 217) a curious physical law, be a signal for
its creation, when they came together; but they would have no substantial identity
with it, nor it with them, and one could never deduce the one from the others, or
(in any intelligible sense) say that they evolved it
(1890/1950: 160).
This is a powerful objection, but it's worth noting that Leibniz would not have been troubled by
it. For him, each mind is a separate, self‐standing entity which is not composed of sub‐minds,
even though the material object to which it corresponds can be thought of as, in a certain sense,
thus composed. Nonetheless, most panpsychists accept that there is mental emergence.
Whitehead embraced this, as explained by Charles Hartshorne, one of his most prominent
followers and an important panpsychist of the later twentieth century: ‘[I]t is the destiny of the
many to enter into a novel unity, an additional reality’ which means that Whitehead makes the
‘admission not merely of emergence, but of emergent or creative synthesis as the very principle
of process and reality’ (Hartshorne 1972: 162).
The panpsychist, however, can deny that this emergence is of the incoherent radical sort
discussed above. For it is not so hard to see mind as becoming more complex via
organization of already mentalistic features. After all, could we not parody James's
remark thus: Take 1023 molecules of H2O and then jam them in a bunch and let them each
vibrate and move as intently as they will; nowhere will there by any liquidity. Yet we know,
more or less, how liquidity does emerge from the ceaseless jostling of the individual
molecules. This is merely epistemological emergence, and arguably it cannot explicate
the relation between an insensate sort of matter and mind. But the emergence of complex
minds from the joint activity of already experiential components is perhaps easier to
understand, especially in light of the fact that the simple part–whole reductionism which
James seems to be presupposing has been undercut with the introduction of the
superposition principle in quantum mechanics. Some emergentists have seen this as a
way to revive the idea of radical emergence (see Silbertstein and McGeever 1999), but
superpositions seem to be no more than rather special combinations of the pre‐existing
properties of motion, mass, spin, etc. Nonetheless, the idea might point to a way to
understand the combinatory powers of intrinsically mentalistic fundamental features of
the world (see Seager 1995).
Another complaint against panpsychism arises from what is called the causal closure of
the physical (see Kim 1998 for extensive discussion of this concept and its implications).
The physical world seems to be causally complete in the sense that every event has a
purely physical determining cause (or, if indeterminism is allowed, a cause which fully
determines the statistics of possible effects). The addition of mentalistic features to the
fundamental entities of the physical world would thus be causally otiose and consign
mentality to the status of epiphenomena (see n. 3 above). But, in the first place, this
argument threatens to make all emergent features epiphenomenal, since their ‘effects’
can be completely explained in terms of the underlying fundamental physics. One might
reply here that the emergent properties retain efficacy because these properties simply
are compositions of certain fundamental, and efficacious, features. If so, the panpsychist
Page 12 of 16
Panpsychism
can adapt this reply. The (p. 218) fundamental dispositional properties of matter are just a
reflection of its mentalistic intrinsic nature; calling them physical with the implication
that they are entirely non‐mental comes close to begging the question.
But there is no real fear of that, as the empirical sciences of the mind have never been
healthier and, in itself, panpsychism does not in any way undercut these sciences or want
to impose on them any methodological strictures. Philosophy of mind also remains in a
very healthy, extremely unsettled state, which will not suffer from the occasional
expedition into the metaphysical wilderness. Panpsychism reminds us of certain very
difficult problems that beset naturalism, and it stands as a perennially interesting
metaphysical position which may yet turn out to be the best way of understanding the
fundamental nature of mind and matter.
References
Alexander, S. (1927), Space, Time and Deity (London: Macmillan).
Barnes, J. (1982), The Presocratic Philosophers (London: Routledge & Kegan Paul).
Blitz, D. (1992), Emergent Evolution: Qualitative Novelty and the Levels of Reality
(Dordrecht: Kluwer).
Broad, C.D. (1925), The Mind and Its Place in Nature (London: Routledge & Kegan Paul).
Page 13 of 16
Panpsychism
Clark, D. (2004), Panpsychism: Past and Recent Selected Readings (Albany, NY: State
University of New York Press).
Clifford, W. (1874/1886), ‘Body and Mind’, Fortnightly Review, 16: 199–245; repr. in L.
Stephen and F. Pollock (eds.), Lectures and Essays (London: Macmillan), 244–73.
University Press).
Griffin, D. (1998), Unsnarling the World Knot: Consciousness, Freedom and the Mind–
Body Problem (Berkeley, Calif.: University of California Press).
Kim, J. (1998), Mind in a Physical World: An Essay on the Mind–Body Problem and Mental
Causation (Cambridge, Mass.: MIT Press).
Lockwood, M. (1989), Mind, Brain, and the Quantum (Oxford: Oxford University Press).
Nagel, T. (1986), The View from Nowhere (Oxford: Oxford University Press).
Page 14 of 16
Panpsychism
Place, U. (1956), ‘Is Consciousness a Brain Process’, British Journal of Psychology, 47: 44–
50.
Rosenberg, G. (2004), A Place for Consciousness: Probing the Deep Structure of the
Natural World (Oxford: Oxford University Press).
Russell, B. (1927), The Analysis of Matter (London: Kegan Paul, Trench, Trubner).
Seager, W., and Allen‐Hermanson, S. (2005), ‘Panpsychism’, in E. Zalta (ed.), The Stanford
Encyclopedia of Philosophy, Summer 2005 edn. <http://plato.stanford.edu/archives/
sum2005/entries/pan‐psychism/>, accessed 2008.
Sellars, W., and Meehl. P. E. (1956), ‘The Concept of Emergence’, Minnesota Studies in
the Philosophy of Science, 1: 239–52.
Silberstein, M., and McGeever, J. (1999), ‘The Search for Ontological Emergence’,
Philosophical Quarterly, 49: 182–200.
Smart, J. J. C. (1959), ‘Sensations and Brain Processes’, Philosophical Review, 68: 141–56.
Spinoza, B. (1677/1985), Ethics, in The Collected Works of Spinoza, i, ed. and trans. E.
Curley (Princeton, NJ: Princeton University Press), 408–19.
Wigner, E. (1962), ‘Remarks on the Mind–body Problem’, in I. Good (ed.), The Scientist
Speculates (London: Heineman), 284–302. (p. 220)
Notes:
(1) Non‐reductive physicalists deny that there is any explanation of mentality in purely
physical terms, but do not deny that the mental is entirely determined by and constituted
out of underlying physical structures. There are important issues about the stability of
such a view, which teeters on the edge of explanatory reductionism on the one side and
dualism on the other (see Kim 1998).
(2) Save perhaps for eliminative materialism (see Churchland 1981 for a classic
exposition). In fact, however, while eliminative materialism is willing to declare beliefs,
Page 15 of 16
Panpsychism
desires, and other intentional mental states mere fictions of primitive proto‐theorizing,
consciousness itself has never been seriously attacked (but see Dennett 1991: esp. ch.
12).
(3) A similar argument was given by S. Pepper (1926), although Pepper seemed to allow
for radical emergence if the emergent features were physically epiphenomenal. For a
reply see Sellars and Meehl (1956).
(4) Rosenberg argues further that we must accept a more radical revision of our
metaphysical views than stems just from the intrinsic‐nature argument and revamp our
conception of causality itself.
(5) Russell put the point with characteristic bluntness: ‘[E]verything we know of [the
world's] intrinsic character is derived from the mental side’ (1927: 402).
William Seager
Page 16 of 16
Subjectivity
Subjectivity is to many philosophers what the frog is to many biologists: the object of
fascination that first drew them to their discipline. The first part of this article discusses
what we find there in our individual consciousness or subjectivity: experiences of various
sorts, including thoughts and thoughts about thoughts and thoughts about other
experiences. It also briefly considers what these things are doing there. The second part
discusses whether Frank Jackson's ‘knowledge argument’ provides a good reason to
doubt that all these things we find in our minds are events in and states of our brain.
Keywords: subjectivity, individual consciousness, thoughts, experiences, Frank Jackson, mental state
SUBJECTIVITYis to many philosophers what the frog is to many biologists: the object of fascination
that first drew them to their discipline. That we have experiences; that we can't say for sure that
they reflect an external world; that we might be alone in the world; that our mind might be the
plaything of an evil demon; that perhaps to be is to be perceived; that will and idea might
exhaust reality; that even the self might be an illusion—to certain teenagers, who may prefer
books to ponds and introspection to vivisection, these are thrilling thoughts. As the adult
biologist, awash in dissertations about DNA, NSF proposals, and university committee meetings,
needs to return to the pond for a few weeks each summer to rekindle love for the subject, so the
philosopher needs to return periodically to his or her own subjectivity, accept it for what it is, or
at least seems to be, enjoy it, explore it, swim in it, and think about it. If your Chair or your Dean
asks what you are doing, say ‘phenomenology’.
Page 1 of 17
Subjectivity
In the first part of this chapter I discuss what we find there in our individual
consciousness or subjectivity: experiences of various sorts, including thoughts and
thoughts about thoughts and thoughts about other experiences. I also briefly consider
what these things are doing there. In the second part I discuss whether Frank Jackson's
‘knowledge argument’ provides a good reason to doubt that all these things we find in our
minds are events in and states of our brain.
12.1.1 Experiences
Locke calls everything we find in our minds ‘ideas’. Hume calls them all ‘perceptions’.
Neither of these seems felicitous as a general term. I'll just stick with ‘experiences’, and
use ‘perception’ with its ordinary meaning, as the recognition and interpretation of
sensory stimuli, even when talking about Hume.
Hume divides experiences into impressions on the one hand and thoughts and ideas on
the other. Impressions include sensations and passions—what we now usually call
‘emotions’. Impressions come unbidden; we are passive; we have them but do not do
them; they are not the products of thought and will. Ideas and thoughts are less lively and
vivacious than impressions, and we often bring them to mind at will. It's easier to think
about a roast‐beef sandwich than to see or taste one; to do the latter you have to order
one or make it yourself.
Once Hume gets to work, however, he finds things less simple than his dichotomy
suggests. Impressions come intimately associated with ideas, based on previous
experience; what we would ordinarily call a perception involves not just a passive
component, but also the result of various activities on our part: we compare, we
remember, we classify, we anticipate, we predict, all of this rolled up with the having of
sensations. What we would ordinarily call seeing a chair involves not only the sensations
caused by the colour and shape of the chair, but anticipations of visual sensations to be
had by moving this way or that, of experiences we expect to have (being supported) and
not have (falling) if we do try sit on it, and the like. Perceptions are sensations associated
with ideas and thoughts. In Book II of his Treatise Hume also develops a theory of
passions as combinations of impressions and thoughts.
This all suggests that what we need is not so much two disjoint categories of experiences,
but rather two dimensions, along which experiences differ, each dimension being more or
less important in different cases. I'll call these the feel and the content of experiences. A
third dimension is physical painfulness and pleasure. Intensely painful and intensely
pleasant sensations are perhaps as close as we get to the limiting case: all feel, almost no
content. But even these sensations have a bit of content: the sensations are taken as
Page 2 of 17
Subjectivity
‘located’ in various parts of our body, and may give us information about what to do or
not do with those parts to eliminate the pain or sustain the pleasure. By the same token,
thoughts are not all content and no feel. There is something it is like to think a thought,
although we don't think of this feel as definitive of the thought. Thoughts involve words,
often unspoken, images, anticipations of other kinds of experiences, and inclinations to
act in various ways, and are often intimately associated with emotions.
(p. 225)
The most striking feature of a sensation, segregated, as best we can, from the
anticipations and memories and attendant thoughts that flesh it out into a perception,1 is
what it is like to have it, its feel, the aspect philosophers have called the quale, subjective
character, phenomenal aspect, and raw feel (see Chalmers 1996; Nagel 1974; Block 2007;
Feigl 1967). Having a sensation of a red fire hydrant is quite different from having a
sensation of a green patch of grass, even if we bracket off the information we seem to be
getting about hydrants and lawns. Having a sensation of green is quite different from
having the sensation one has when one hears a high trumpet note; the pain of a toothache
is different than the pain of a backache; such unpleasant sensations are quite different
than pleasant ones, like tasting chocolate, or smelling a rose. Although we may be
convinced that there is this aspect of raw feel in experiences, we are hardly able to
categorize without employing the idea of of‐ness, where the far side of the of relation is
not something subjective, but something that is outside the mind—a colour, a sound, a
back, a tooth, a piece of candy, a rose, a fire hydrant, a lawn—or would be outside if it
were real—a unicorn or my (imaginary) red Porsche.
Right now I have a visual field full of shapes and colours, which I take to be a computer in
front of me, a book to my side, a cup full of hot coffee, a table underneath these things,
my own hands perched on the keyboard of my computer, often waiting for inspiration,
occasionally typing, a window, trees, and rooftops outside the house, and a lot more. I am
listening to Johnny Cash sing about poor drunken Ira Hayes, so the sounds that comprise
his words, and the sounds from his guitar, take up a lot of my auditory space. There are
also, now and then, the sounds the keys on the computer make as I tap them, the sounds
made by the cats pursuing various sort of mischief, occasionally the ring of a telephone,
occasionally a request from my wife Frenchie. I feel the computer keys at the end of my
fingers, and the surface on which my hands rest. I occasionally take a sip of coffee, so,
and have that peculiar taste. While having all of these experiences, I think. I think about
what to write, about whether to respond to Frenchie immediately or take a few seconds
to finish what I am doing, whether to go investigate the cat noises or ignore them,
whether coffee is unhealthy like the doctors said five or six years ago or actually not bad
at all like they seem to say these days, and a zillion other things.
Thinking is often an active, purposeful activity. We intend to think about a certain subject
matter: the weather, Calvin Coolidge, prime numbers, the commitments of the day, and
then we manage to do it. But thoughts often come unbidden. While trying to remember
Page 3 of 17
Subjectivity
the year of Calvin Coolidge's election, I find myself thinking about George Bush and Iraq.
While trying to identify the first fifteen prime numbers in my (p. 226) mind, or follow the
arguments of a student in my office, I find myself thinking about a roast‐beef sandwich.
The most striking features of thoughts are their contents rather than their feels. We get
at their contents by saying what things must be like to fit the thought, where the
evaluative property and direction of fit are determined by what kind of thought it is.2 The
content of the thought ‘It's sunny today’ is quite different from the content of the thought
‘Calvin Coolidge was a man of few words’. The direction of fit in the case of such doxastic
or belief‐manifesting thoughts is mind to world, the evaluative properties are truth and
falsity. With wishes, like ‘Would that I receive a big raise’, the dimension of success is
being granted or not; the direction of fit is world to mind; if my wish is not granted, the
world, or at least the Dean's office, is defective, not the thought. Thoughts are about
things and their properties: days, weather conditions, presidents, loquaciousness, money,
and such. The contents are conditions, involving these things—truth‐conditions in the
case of doxastic thoughts, conditions of being granted in the case of wishes.
But still it is like something to think, and it is like something to have thoughts simply
occur to you: thoughts are experiences. A lot of thinking involves something like inner
speech; something similar in feel to rehearsing what one is going to say, or anticipating
saying it, or remembering saying it, or exhorting oneself. Indeed, people often lapse into
talking to themselves when they are really into a piece of thinking. Thinking in English is
different than thinking in German; it's a thrill for the language student when the first
thought formulated in the new language spontaneously makes its appearance. And
thinking is not limited to words; all sorts of ideas, including images corresponding to
various types of sensations are also involved.
The feel and content of experiences differ in their ontological status. The feel of an
experience is an intrinsic quality of it. The sensation may be caused by an external object,
and may be part of a contentful perception of that external thing. But the feel of the
sensation is a fact about what is going on in my mind. A phrase like ‘the sensation of
seeing red’ gets at the feel of the sensation in a roundabout way: it is (roughly) the type
of visual sensation typically caused by seeing red things in favourable light.
The content of experiences, of perceptions and thoughts, is at least not entirely intrinsic
to them. If I see Condoleezza Rice and think ‘That woman is the Secretary of State’, then
my thought is about Rice, and is true because she is the Secretary of State. If it is not
Rice that I see, but Angela Davis, then my thought is about Angela Davis, and is false
because she is not Secretary of State, at least as I write this paragraph. The truth‐value
and truth‐conditions of my thought depend on whom my perception is of, and so are not
intrinsic properties of it. The feel of the thought, however, seems intrinsic. I could have
just those words running through my mind even if I didn't perceive anyone. I'll return to
this issue below.
Page 4 of 17
Subjectivity
Subjectivity does not exhaust what we ordinarily think of as ‘the mind’ or ‘the mental’.
Beliefs and desires are mental, but they are not experiences in the way that sensations
and thoughts are.3 My beliefs and desires guide my thinking, and my thoughts and
experiences in turn affect my beliefs and desires. Thoughts, like actions directed at the
external world, manifest my beliefs and desires. If I believe that Sacramento is the capital
of California, then when the question arises in my mind I will think ‘Sacramento is the
capital of California’; if the question arises in conversation, that's what I'll say.
The word ‘concept’, as I shall use it, stands for cognitive structures involved in my beliefs
and desires and other cognitive states. Using ‘concept’ in this way, concepts are not the
same as ideas, considered as the bits of thinking that make up our thoughts.4 Concepts,
beliefs, and desires are not, like thoughts, transitory by nature; they are acquired at
various times; they may last for years, or they may change after a few seconds. New
experiences and new inferences lead to revisions in our beliefs; desires get changed by
deliberation; sometimes they are satisfied and disappear; and we just forget things we
once knew or at least believed or thought it was important to do or have.
We then must distinguish between that in our minds that we experience, the parts of the
stream of consciousness, and the beliefs and desires that are more like rocks and fallen
trees below the surface of the stream that direct its flow. These we are not directly aware
of but can, in at least a wide range of cases, easily determine, and often alter.
Let me return now to the question of whether any of the intentional or semantic
properties of beliefs and the thoughts to which they give rise are intrinsic. This question
is somewhat vexed, as it is connected with Big Issues like anti‐individualism and narrow
content, and I won't discuss it at great length here. But I will say a bit. There are two
mistakes that are made in discussions on this topic I do want to mention.
Page 5 of 17
Subjectivity
There is no doubt that having the property of perceiving that Condoleezza Rice is
reviewing the troops is not something I have simply on account of what is going on in my
head, the states of my brain, and so too with believing that Condoleezza Rice did review
the troops last week. Those properties require that what goes on in my head be related in
certain ways to Condoleezza Rice, and these relationships require things of the external
world, not just things inside my head. It does not follow from this that the state I am in, in
virtue of which, together with the external situation I find myself in, I have those
properties is not simply a matter of what goes on in my head. It seems, for example, that
there might be a state such that when a person is in it, in a fairly wide range of
circumstances, that person perceives that the person they see is reviewing the troops.
Well, maybe this is too optimistic, for how about the fact that what Rice was reviewing
were troops? We can retreat further, to the state one is in such that, in a wide variety of
circumstances, one who is in that state perceives that the person they are seeing is
examining in an authoritative manner the group of uniformed persons that person is
looking at. Well, one could have further qualms. That is, one might have to work hard to
squeeze the commitments about the external world out of the property, so that the
intentional description that is left, stripped of these commitments, only constrains the
internal properties of the state. But the avenue for getting fewer and fewer commitments
is fairly clear. We start with some intentional characterization of a minded being, such as:
and then we reformulate things in a way that makes the intentional contributions of internal and
external factors clear, relative to some account of the structure of internal states, which for
purposes of illustration can be rather clunky and naive:
Bush is in a belief state with the structure C(n, g), where C is a concept of the activity
of reviewing, n is a notion of Condoleezza Rice and g is a notion of some troops, that is
true only if the person n is of is performing the activity C is of upon the group g is of.
and then we abstract over the specific external factors we can identify:
Bush has a belief b with the structure C(n, g), such that if there is an individual X, and
activity A and a group G, and C is of A, n is of X, and g is of G, belief b is true if and only
if X performs A upon g.
The second mistake infected the once common and often quite worthwhile discussions of
narrow content. This was the supposition that the content that we end up with, when we
have completed this stripping enterprise, will be a proposition P such that the agent has
the property believing P. The model was the transition from de re belief to a supposed
underlying de dicto belief. This supposition is quite groundless. The whole idea of
‘intentionality’ is to describe what goes on inside of us in terms of what is outside of us; it
is things outside of us that we by and large have beliefs about. But the conditions our
stripping operation will leave us with will be existentially quantified conditions on things
inside of us, our perceptions, concepts, notions, (p. 229) and what have you. Bush may
believe that Condoleezza Rice is reviewing the troops because he is in some state C (n, g)
that meets the conditions above. That does not mean Bush believes the proposition that
there is a person X a relation R and a group G and my ideas C, n and g are of respectively
Page 6 of 17
Subjectivity
R, X, and G and X is R‐ing G. What he believes will be that Condoleezza Rice is reviewing
the NATO troops. I doubt very much that Bush has many beliefs about his own concepts
and notions. The narrow content we assign to what goes on inside our heads, as a part of
an account of what we believe, perceive, know, conjecture, and the like, can play its role
without itself being what we believe, perceive, conjecture, and the like (see Perry and
Israel 1981).
While writing some of the paragraphs above I was not only having experiences and
thinking, I was aware of my experiences; I attended to them, thought about them, and
indeed wrote about them. It seems to me that having experiences, and becoming attuned
to the information they carry, is something I share with all sorts of animals. All of these
animals we might ordinarily call ‘conscious’, during those periods when they haven't been
knocked unconscious. But for this use of ‘conscious’ I'll use ‘sensate’.
The way ‘conscious’ is used in the OED definition of ‘subjectivity’ quoted at the beginning
of this paper is different. We are conscious of things. We use the term in this way quite
broadly. I am conscious of the dangers posed by Bush's overspending, I am conscious of
the racket made by the leaf‐blower across the street, and I am conscious of the sensations
I have as I type. Not everything we are conscious of is subjective, but it is consciousness
of the subjective that we now turn to. It seems clear that we not only have experiences
and think thoughts, we can become aware of, focus our attention on, compare, classify,
ponder, admire, detest, describe, and in all sorts of ways think about our experiences and
thoughts.
So in order to be aware of my experiences I have to have them, but having them is not
sufficient. Right now I am having an experience of seeing a computer. It is like something
to have this experience. The experience plays a role in my life; it is involved in my
perception of the computer, and my perception of it is involved in the interactions I am
having with it: supporting it on my lap, typing on the keys, watching letters appear on the
screen, and the like. None of this requires me to be aware of (p. 230) the experience;
having the experience is part of my being aware of the computer. I am attuned to the fact
that when I have this sort of experience, in relatively normal conditions, there is a
computer on my lap, and I can make letters appear on the screen by typing the keys. I am
Page 7 of 17
Subjectivity
able, however, not only to have the experience but to attend to it, to form a concept of it,
to apply concepts to it, and the like. It is a visual experience; it is neither pleasant nor
unpleasant; I didn't directly pay money to have the experience, although I paid good
money for the computer. I can also form a concept of the type of experience I am having,
and of course I can classify the experience in various ways. It is an experience of a
computer; of this particular computer; a visual experience; an experience of a square
white region embedded in an off‐white region, with letters that form words on the white
region and letters on keys in the off‐white region.
When we attend to the experiences involved in perception, the path of least resistance is
to classify them in terms of what they are of. There are two ways the phrase ‘of’ can be
taken here, and often it is not necessary to mark the difference. I see my computer; my
experience is of a computer; that is, a computer causes it. But it is also of a computer in
that I see what I am seeing as a computer, and not, say, a mere computer façade or a
television. I could mistake my television for a computer; then my experience would be of
a television, in the first sense, but of a computer, in the second.
To see the object I am looking at as a computer involves much more than mere sensation.
I take various discontinuities in my visual field as corresponding to the edges of the
screen, for example; I see it as three‐dimensional, which involves expectations of what
would happen if I were to move my head a bit, or stand up and approach the computer, or
reach out and touch it. I expect the pattern in my visual field to remain relatively
unchanged as long as I keep my eyes open and stay still; because I see what I see as a
computer, I don't expect it to suddenly walk away. I actually can see two computers, the
laptop on which I am typing, and, in the distance, a desktop computer. The two
experiences seem similar, but the similarity in the experiences mostly amounts to their
being of the same sorts of things: computers with visible screens, keyboards, and the like.
When we turn from perceptions whose importance is the information they carry about the
external world to those whose dimension of painfulness and pleasure dominates our
interest in them, we still have difficulty squeezing all of the intentionality out of our way
of thinking of them. I have a slight pain in my right wrist. Locating the pain in this way is
to describe it in terms of the information it gives me about the part of my body that is the
source of the pain. I know the pain is in an important sense not located in my wrist but in
my head. I could have a phantom pain like this, without even having a wrist, were the
nerves between wrist and head properly stimulated. The pain is in my wrist in that it
gives me information about my wrist, and it is my wrist I will move in order to deal with
the pain.
Why do we have experiences? In particular, why are we sentient? Why are some of our
brain states like something to be in? There are two questions here. Perhaps the (p. 231)
first is what David Chalmers calls the ‘hard question’ of consciousness (1996). It is pretty
Page 8 of 17
Subjectivity
much the same as Heidegger's question, ‘Why is there something rather than nothing?’,
at least on one interpretation of it. That is, the question is not why there is any universe
at all, rather than nothing, but why there is experience at all, rather than a ‘dark’
universe.
The only way I can imagine of answering this question is by identifying the physical
characteristics that differentiate states it is like something to be in from those that it is
not like something to be in, an enterprise that will doubtless lead to a more complex set
of distinctions than we have adumbrated here, as scientists mine the interplay between
the three ways we have of knowing of such things: examining brains, studying behaviour,
and attending to the states in which we find ourselves. As this knowledge develops, the
question why a given brain‐science‐identified state is like this i (using ‘thisi’ as an inner
demonstrative for experiences and the types they exemplify) may have a clear answer,
against a background of what it is like to be in other brain‐science‐identified states. At
that point, when we can answer each such question against the background of answers to
many others, there will not be, as far as I can see, any ‘explanatory gaps’ left; at least,
none of the sort that ever get closed. Perhaps at some point there will be a moment of
conceptual clarification, where the mysteries of the more general question no longer grip
us. I hope so, for then philosophers may play a role in the Great Day when the answer to
this secret is laid bare.
The second question is this. Given that there are brain states it is like something to be in,
experiences, what purpose does it serve to be in such brain states? What are they for?
What is mother nature trying to do with them?
In his Dialogues on Natural Religion Hume complains that an omnipotent, omniscient, and
indulgent deity could have come up with something better than pain to motivate us to get
out of dangerous situations. One can imagine a sort of permanent semi‐erotic pleasure
that normally suffuses all parts of our body. An injury to the foot or a sprained elbow or a
decayed tooth wouldn't cause the sorts of pain it actually does, but instead merely a
diminution of pleasure in foot or elbow or tooth.
But we're not here to complain. Working within the metaphor of a purposeful mother
nature, this mother is clearly either not omnipotent or not omniscient or not indulgent.
She makes do with what she has, to work her bizarre purpose, which is basically to get
plants and animals to reproduce so that there will be plenty of nutritious stuff,
decomposing, crawling and walking around for other plants and animals to absorb one
way or another so they can last long enough to reproduce and get absorbed in turn. Given
these purposes, and the fact that she has brain states it is like something to be in her
repertoire, what can she do with them? What are experiences for?
Behind Hume's complaint is the idea that our sensations give us information and motivate
us to act in ways that makes sense given that information. Experiences that are
dramatically painful or pleasant provide both motivation and information. When we step
Page 9 of 17
Subjectivity
on a tack we are motivated to do something to bring the experience of pain to an end, and
we know, innately or based on relatively little experience, what to do about it.
(p. 232)
Less dramatic experiences, with no pain or pleasure intrinsic to them, provide more
delicate kinds of information, that can be exploited through mechanisms built up through
evolution, experience, accretions of culture, and the memories and thoughts of a
particular person.
So our experiences are there to give us information about our bodies and the broader
environment, in order to enable us to act in ways that increase the probability of
successful reproduction, and to give us motivation for so acting. The basic way we use the
information that is made available to us is through habit or, more generally, attunement.
As Hume points out, these matters are much too important for nature to leave mainly to
understanding and reason. Attunement means basically that (i) being in some state S
carries the information that we are in some situation E; (ii) being in S causes us to act in
some way that makes sense, given that we are in E. If I step on something sharp, I move
my foot. I may not have the concept of danger, or injury, or survival, or of pain, or of a
foot, or of me. Still, moving the foot makes sense given that not moving it will cause
injury. I am attuned to the regularity or constraint that stepping on sharp things causes
injury.
Most of our visual and auditory sensations aren't unpleasant or pleasant enough or
informative enough on their own to motivate us to do much of anything. But we can learn.
Pavlov's dog learned that the bell meant food; he became attuned to this constraint in his
environment, and started to salivate when he heard the bell. My goats learn that when
they hear the sound I make by pounding the side of an old five‐gallon paint bucket there
will be some alfalfa and sweet cob for them to eat if they walk in the direction of the
sound. Dogs and goats need to discriminate and have the capacity to learn new habits—to
become attuned to new constraints.
An information game is a pair of episodes in which a being gathers the information that P
at some time, and some being, perhaps the same being, perhaps not, uses that
information at some time, perhaps the same time, perhaps not, to do something that
makes sense given the fact that P. If the beings and the times are the same, we have the
‘straight‐through’ information game. Given the pain in my foot, I move it. Given the bell,
the dog salivates. Given the sound of stick on bucket, the goats start to move in the
direction of the sound.
We, and a number of other animals, have the ability to store information for later use. We
detach some of the information from the perception that carries it, and reapply it later,
when we recognize the same object. I call this ‘the detach and recognize’ information
game. This is where beliefs and concepts come in. Perhaps a goat isn't hungry at the time
it hears me pound on the bucket. It wanders off to do some other business for a while.
Then it moves towards the place where the sound came from. The goat picked up the
Page 10 of 17
Subjectivity
information that there was food in such‐and‐such a direction at one time, and then later—
but not so much later that the food wouldn't still be there—it does something that makes
sense given that information. To do this, the goat has to reorient itself, to reidentify the
direction from which it heard the sounds of food, probably by recognizing local
landmarks. It (p. 233) won't suffice to be attuned to general patterns in its sensory field;
it will have to have the conceptual apparatus to separate out trees and people and other
salient objects, and reidentify them.
Suppose, for example, that I see Condoleezza Rice at a Stanford party. Because of the
stress of her years in the Bush administration, and her diplomatic wardrobe, she doesn't
look quite the same as she did while a professor at Stanford, and at first I don't recognize
her. For a while I have two notions of Rice, my long‐standing notion, associated with such
concepts as being brilliant, strong‐willed, and having left Stanford to join the Bush
administration, and a temporary buffer, associated only with the concepts my perceptions
deliver. What do I learn when I recognize her? It isn't hard to say how my doxastic states
change. The two notions have to be of the same person, if my belief, ‘Oh, that person is
Condoleezza Rice’, is to be true. But the change from before to after the recognition can't
be captured by a proposition involving only Rice, without bringing in my buffer. It can't,
that is, be captured by a proposition whose constituents are confined to the subject
matter of my thoughts, the external objects my notions are of. To suppose that all
knowledge can be so captured is what I call ‘the subject‐matter fallacy’. Such
recognitional knowledge is fleeting, for our buffers are usually simply absorbed into the
more permanent notions.6
It is this strategy, developed in many different ways, that has come to dominate human
cognition. We each possess a rich set of concepts, or notions of individual things. We have
two uses for the information associated with each of these notions. We use it to help us to
recognize the objects of which we have notions, and we use it to help us do something
that makes sense with those objects (or, more generally, something that makes sense
given what those objects are like and their relation to us), once we have found them. So,
for example, I have a notion of Michael Bratman that provides me with the ability to
recognize him in favourable circumstances: when I can see him, when I hear his voice on
the phone. But my notion also provides me with facts about him that are useful in
interacting with him. For one thing, I know that his name is ‘Michael Bratman’, so I can
address him with that name. I know he thinks a lot about the philosophy of action, so I
can ask him questions about his work, as a way of being friendly, and as a way of learning
about the topic.
Page 11 of 17
Subjectivity
Suppose, for example, that my goats have language and cellphones and somewhat more
altruistic relations with one another than they actually do. One goat might call a friend,
out of earshot of the signs of food, and tell him that there is food near tree X. This would
be much simpler than providing the goat friend with all the information the latter needs
to get to the food. The responsibility for reattaching the detached information is left to
the goat friend; he can wander around until he sees tree X, then he can apply the
incremental information, that X has food near it.
Being able to attend to our experiences, classify them, note whether we like them or not,
think about their causes and how to avoid having them or increase the chances that we
will have them, adds power to our deliberations. We can plan not only to bring about
certain results in the external world, but also to bring about or avoid certain results in
our own subjectivity. Whether this itself serves any of mother nature's purposes I rather
doubt. It is certainly useful in subverting her plans. We invent ways to have the internal
pleasures of the procreative act without procreating, for example, or ways to link the
tastes we like to substances of no nutritional value. Most likely it is something she didn't
plan on. Human thought, language, and culture have taken the detach and recognize
information game and run amok. That, at any rate, is my picture of human science and
culture, but I'll spare the reader details.7
Page 12 of 17
Subjectivity
12.2.1 Mary
In Frank Jackson's classic statement (1986) the ‘knowledge argument’ has one character
and three steps. Mary is a brilliant woman who, for one reason or another, is raised in a
room where she never sees colours, only black and white and, I suppose, shades of grey
in between. Mary has new knowledge when she finally steps out of the black and white
room and sees a red fire hydrant. But while in the black and white room she could well
have known all the physical facts relevant to colour vision. Conclusion: her new
knowledge is of a non‐physical fact. We need to look closely at Mary.
Mary emerges from the black and white room, sees a fire hydrant, and has her first
colour experience, call it E, of the type qualeRED. She thinks: ‘Thisi experience is the type
I have when I see, in these conditions, the colour of that fire hydrant’. She has a certain
relation to the experience: she has it. A less inquisitive person might have left it at that,
but she also attends to the experience. E is the referent of her thought ‘thati experience’
because of the relations it has to her: it is the one she is having and attending to. She is
having an experience of the colour of the grass beside the fire hydrant too, but ‘thisi
experience’ doesn't refer to that experience, because it is not the one to which she
attends.
Mary forms a concept of the type of colour experience E exemplifies. She notes that E is
similar to the colour experience she has of the fire engine parked nearby, and not similar
to the colour experience she is having of the grass next to the fire hydrant. She can
introduce a term, ‘quale?’ and ask:
Is quale? = qualeRED?
Mary's concept quale? seems like a good candidate for a phenomenal concept, for it is tied to her
current experiences of the fire engine and the fire hydrant; it is the type of colour experience of
which those two colour experiences are instances. Many philosophers put great weight on such
phenomenal concepts, and they are useful in considering the fine structure of Mary's cognitive
states at the moment of liberation. But such concepts are by their nature temporary, like the
perceptual buffers involved in recognizing external objects.
Since quale? is qualeRED, Mary is thinking about the same type of experience in two
different ways when she uses the two terms in thought or language. The referential
relations are quite different; she is related to the quale in two quite different ways. On
the one hand, it is the quale that two of her current experiences exemplify. On the other,
it is the quale that her textbooks referred to, and identified as the type of experience
normal people have in favourable light when they see red objects. Her conceptions of the
two are different. She believes that qualeRED is the one people (p. 236) with normal vision
have when they see red things in favourable light, and she also believes—since my
Page 13 of 17
Subjectivity
Once she draws this inference, she knows something new about qualeRED, that she didn't
know in the black and white room; namely, that it is the type of colour experience
exemplified by her current experiences of the colours of the fire hydrant and the fire
engine. This is a relational fact about particular experiences that hadn't occurred when
she was in the black and white room; it is not knowledge of some new fundamental
property of qualeRED that was of necessity missed by her physicalist texts.
Perhaps Mary is so brilliant that she was able to predict that upon leaving the black and
white room she would see a fire hydrant, grass, and a fire engine, all with their normal
colours, in favourable light. She can introduce terms for the predicted experiences, say
EH and EE. So she had a way of referring to and thinking about the experiences she is
now having and their common quale, qualeRED, before having them. But the referential
relations, in virtue of which she was able to refer to these things and talk about them, are
quite different than the referential relations that enable her to think and talk about them
as ‘thisi colour experience’ (attending to the hydrant), ‘thisi colour experience’ (attending
to the engine), and ‘quale?’. She can predict, while in the black and white room, that she
will have a phenomenal concept, but she cannot yet think of it as one she is having. She
can ask herself whether thisi colour experience is EH, thisi colour experience is EE, and
whether quale? is the one she predicted, qualeRED. She can figure out that they are. So
she still has new knowledge that she didn't have before.
Does anything in all of this give Mary a reason to abandon her physicalist view, that
qualeRED is B52? I cannot see that it does. If her physicalist views are correct, and she is
wearing a new‐fangled autocerebroscope, that produces visual images of the goings‐on in
her brain as she had her experiences, she would have to grant that she was seeing the
very experiences she was having. She might think, ‘Goodness, seeing an experience of
qualeRED through an autocerebroscope (which I've never done before) is certainly nothing
like having an experience of type qualeRED (which I've never had before). But then, why
should it be?’.
Page 14 of 17
Subjectivity
Variations on the Mary story have been proposed, in order to produce the insights that
will lead those of us unconvinced by the original story to dualism. But, actually, most of
the more interesting and dramatic aspects of the various Mary stories (p. 237) seem
irrelevant. The basic point is well made by Feigl's original autocerebroscope fantasy. Here
is a slight variation on it that may prove helpful. Suppose Mary is never imprisoned and
has experienced a normal range of colours by the time she makes it to graduate school to
work on colour vision and brain states. She has good colour vision, good colour memory,
and a good command of the standard names of dozens of colours and shades. Then she
learns as much as you please about the various brain states involved in colour vision.
Suppose she has clearly before her mind what it's like to experience six colours: red,
yellow, blue, orange, brown, and mauve. And she has, on her computer, complete physical
descriptions of the six brain states that correspond to the state one is in when one
perceives those colours. This is a sophisticated hyper‐linked set‐up, so she can zoom in on
pictures, x‐rays, sonograms, or whatever else you want. She can do in a virtual way what
Leibniz imagined doing; she can enlarge the relevant parts of the brain and walk in and
look around them. So far so good. However, the final interface between this sophisticated
program and natural language, the identification of the scientifically described brain
states in terms of ordinary colour words, is not in Mary's native Australian but in
Cantonese, which she doesn't know. Now, can she match up the colours, or, more
precisely, the quale that she associates with seeing the colours, with the physical states
that she is in when she perceives those colours? I can't see how she could. When her
Cantonese room‐mate returns home, and deciphers the names, Mary will learn
something.
Here is the situation as I see it. Let's take mauve as our example. Mary has a memory of
seeing mauve, she knows what it is like to see mauve; and, if you like, she can generate
an actual experience of seeing mauve by pulling her mauve scarf from the drawer. So she
has a concept, as phenomenal a concept as could be wished for, of the state one is in
when one sees mauve. And she has a concept of the brain state people are in when they
see mauve things in ordinary conditions, whose scientific name is BSμ. Between her and
her computer program the whole of brain science is at her fingertips; she can look at
actual pictures of BSμ; she can look at helpful diagrams of it; push a button and the
chemical composition of the various stuff involved in such states will pop up annoyingly;
and so forth. These two concepts are both concepts of the same physical state, according
to the physicalist; and, in addition, if she got her scarf from the drawer it is the state she
is in. Does this mean that the physicalist should expect her to figure out, before her room‐
mate comes home to decipher the Cantonese, that the two concepts are concepts of the
same state?
There is no reason the physicalist should suppose this. What magic would drive a mental
identity sign between two such different concepts as these, even for someone as brilliant
as we suppose Mary to be? Neither concept is defined in terms of the other; neither is
introduced in terms of the other; neither makes reference to the other; they have no
common parts. At this point the knowledge argument says: but then, what does she learn
when her room‐mate comes home? But the answer is clear. It's just the difference
Page 15 of 17
Subjectivity
between having two concepts that are, in fact, of the same thing, and two concepts that
are required by that internal identity sign to be of the same thing. She learns that the
type of experience she is having, and so the type of experience her phenomenal concept
is of, is the type of experience that her other, (p. 238) brain‐science‐based concept is of;
namely, the mauve quale. The contents of her doxastic states change, in that the truth of
her total doxastic state, abstracted from the referential relations of her concepts,
requires that the two concepts are concepts of the same type of experience; that is, BSμ,;
that is, the experience one has when one sees mauve. That is how her beliefs change, and
how her knowledge changes.
References
Bach, K. (1981), ‘An Analysis of Self‐deception’, Philosophy and Phenomenological
Research, 41: 351–70.
Block, N. (2007), ‘Max Black's Objection to Mind‐Body Identity, in T. Alter and S. Walter
(eds.), Phenomenal Concepts and Phenomenal Knowledge: New Essays on Consciousness
and Physicalism (Oxford: Oxford University Press), 249–306.
Feigl, H. (1967), The ‘Mental’ and the ‘Physical’: The Essay and a Postscript (Minneapolis,
Minn.: University of Minnesota Press).
Hall, L. (1993), ‘Individualism, Mental Content and Cognitive Science’, Ph.D. diss.
(Stanford University).
Jackson, F. (1986), ‘What Mary Didn't Know’, Journal of Philosophy, 85: 291–5.
Malcolm, N. (1970), ‘Scientific Materialism and the Identity Theory’, in C. V. Borst (ed.),
The Mind/Brain Identity Theory (New York: Macmillan), 171–80.
Perry, J. (2001), Knowledge, Possibility and Consciousness (Cambridge, Mass.: MIT Press).
—— (2002), Identity, Personal Identity, and the Self (Indianapolis, lnd.: Hackett).
Perry, J., and Israel, D. (1981), ‘Fodor and Psychological Explanations’, in B. Loewer and
G. Rey (eds.), Meaning in Mind (Oxford: Blackwell), 165–80;
repr. in Perry, The Problem of the Essential Indexical and Other Essays, expanded edn.
(Stanford, Calif.: CSLI, 2000), 301–22.
Page 16 of 17
Subjectivity
Notes:
(1) I'll often ignore the ‘success’ or ‘veridicality’ or ‘factive’ implications of words like
‘perception’, and ‘information’. This seems fair in an essay about subjectivity; by
‘perception’ I mean ‘perception or would‐be perception’.
(3) This is a point I did not adequately grasp when I wrote Knowledge, Possibility and
Consciousness (see Bach 1981).
(4) This is a departure from how I have used these terms in the past.
(6) For more on the subject‐matter fallacy see Perry (2001) and Perry (2003).
(7) For more on the topics of this section see the last part of Perry (2002).
John Perry
Page 17 of 17
Higher‐Order Theories of Consciousness
There are several phenomena that constitute what we call consciousness, each of which
gives rise to special problems and puzzles. One is the condition people and other
creatures are in when they are conscious, as against when they are, for example, asleep,
knocked out, or anaesthetized. And there is the related question about what distinguishes
people and other creatures that can be conscious in that way from things that can not,
such as stones and trees. We may call these questions the problem of creature
consciousness. There is also a question about what it is for a person or other creature to
be conscious of something. We may call this phenomenon transitive consciousness.
Keywords: theories of consciousness, creature consciousness, creatures, transitive consciousness, inner sense,
higher-order thoughts
THERE are several phenomena we call consciousness, each of which gives rise to special
problems and puzzles. One is the condition people and other creatures are in when they
are conscious, as against when they are, for example, asleep, knocked out, or
anaesthetized. And there is the related question about what distinguishes people and
other creatures that can be conscious in that way from things that can not, such as stones
and trees. We may call these questions the problem of creature consciousness.
There is also a question about what it is for a person or other creature to be conscious of
something. We may call this phenomenon transitive consciousness. One is conscious of
something when one sees or hears that thing, or senses or perceives it in some other way.
Having a thought about something also sometimes suffices for one to be conscious of that
thing, but not always. Sometimes our thoughts are not about a thing as being present to
us, as with thoughts of Julius Caesar, the number 17, or the planet Saturn, and we are not
in these cases conscious of those things. But one is conscious of something when one has
a thought about it as being present to one, even when one does not also sense or perceive
Page 1 of 17
Higher‐Order Theories of Consciousness
that thing. It is likely that an analogy with sensing leads us to regard this as being
conscious of something, since sensing is arguably the more basic way of being conscious
of things and one always senses things as being present to one.
What these authors do urge is that whenever these states occur in somebody, that
individual is conscious of them. Descartes famously adds that this consciousness is
unmediated; one is immediately conscious of all one's thoughts whenever they occur
(Descartes 1964–1975: vii. 246). It is not until the latter part of the nineteenth century
that theorists come to recognize that some mental states do occur of which we are wholly
unaware. And, because of the resulting need to mark the difference between those
mental states and those we are conscious of, the one‐place predicate ‘conscious’ comes
then to be increasingly applied to mental states themselves.
When we now describe states as conscious, we are presumably saying just what
Descartes and Locke had in mind in saying that one is immediately conscious of a state;
indeed, we often paraphrase those writers using the phrase ‘conscious state’. This
suggests a first step in explaining state consciousness; a state is conscious if one is
conscious of that state in some suitable way. I shall call this the transitivity principle (TP),
since it explains state consciousness in terms of transitive consciousness.TP receives
support from noting that no state is conscious if one is in no way whatever conscious of
being in that state. We can understand the various higher‐order theories as all holding TP,
but as differing in how we are conscious of our conscious states.
Page 2 of 17
Higher‐Order Theories of Consciousness
One might insist that when one is conscious of something, the state in virtue of which one
is conscious of it must itself be a conscious state. But that is not so. Subliminally
perceiving something makes one conscious of that thing, though not in the way we are
conscious of it when we sense it consciously. We sometimes mark this (p. 241) distinction
by saying that in subliminally perceiving something one is conscious, or aware,1 of that
thing, but not consciously aware of it. Only if subliminally perceiving something makes us
conscious of it can we explain the effect subliminal perception has on our behaviour and
on the rest of our mental lives.
Fred Dretske has argued against TP, urging that a state's being conscious consists not in
one's being conscious of it, but in one's being in a state in virtue of which one is conscious
of something else (Dretske 1993; sect. 4). But this view makes problematic the distinction
between mental states that are conscious and those that are not.
D. M. Armstrong and William G. Lycan are the best known contemporary proponents of
the inner‐sense view (Armstrong 1981; Lycan 1996, 2004). Both hold TP, and explain how
we are conscious of conscious states by positing an inner monitoring mechanism that
tracks many of the mental states we are in. This appeal to monitoring allows us, they
urge, to explain the function state consciousness has in our overall psychological
economy; monitoring enhances our ability to think rationally, to make plans, and to
coordinate our actions. As Lycan puts it, such monitoring has ‘the job of relaying and/or
coordinating information about ongoing psychological events and processes’ (2004: 100).
There are, however, difficulties in understanding TP this way. For one thing, sensing and
perceiving always involve some mental quality. States of seeing or hearing things, for
example, always exhibit some visual or auditory qualitative property; otherwise, the
resulting awareness of those things would be indistinguishable from simply having a
thought that those things are present to one. But the only mental qualities that occur
Page 3 of 17
Higher‐Order Theories of Consciousness
when we are conscious of our conscious mental states are qualities that belong to the
states we are conscious of; the higher‐order consciousness of those states exhibits no
mental qualities of its own. So that higher‐order consciousness cannot be a matter of
sensing or perceiving the first‐order states.
Lycan rejects this conclusion, arguing that though such higher‐order consciousness has
no mental qualities, it nonetheless resembles perceiving more closely than it does
thinking (2004: sect. 6). But there is reason to doubt this. Lycan urges that the
monitoring that results in some of our mental states’ being conscious ‘is the functioning
of internal attention mechanisms directed upon lower‐order psychological states and
events’ (2004: 99). And he holds that these attentional mechanisms operate in a quasi‐
perceptual way. Just as we have voluntary control over what we perceive, for example, so
we have considerable voluntary control over which of our mental states are conscious.
But it is unclear that we do have much voluntary control over which of our mental states
are conscious. And even if we do, our higher‐order awareness will not on that account
resemble perceiving more than thinking, since we also have considerable voluntary
control over our thoughts.
Lycan urges several other points in support of higher‐order perceiving. He argues that
higher‐order perceiving better explains the monitoring function of consciousness than
higher‐order thoughts can. But thoughts could monitor first‐order psychological states
equally well, and the absence of any known inner sense organ suggests that such
monitoring very likely occurs that way. Lycan urges that a perceptual model fits better
with the way we focus attentively on some conscious states; but we focus attentively on
things in thought no less than in perception. Lycan holds that the reliability of our higher‐
order awareness of mental states suggests a perceptual model. But the general accuracy
of that higher‐order awareness itself needs explaining, and whatever factors are
responsible for it might figure just as well in a higher‐order‐thought model.3
Many of Lycan's reasons for holding that our higher‐order awareness is more like
perceiving than thinking depend on his seeing that higher‐order consciousness as a kind
of monitoring. But there is reason to question that monitoring model. Not all our mental
states are conscious, nor even all those which matter in some significant way to our
psychological functioning. If our higher‐order awareness is due to some perception‐like
monitoring mechanism, why does it miss so many significant states?
Our higher‐order awareness of our mental states is often affected by our interests,
desires, and fears in ways that resist a model based on perceptual monitoring. Dental
patients are sometimes conscious of themselves as being in pain even though the relevant
Page 4 of 17
Higher‐Order Theories of Consciousness
nerve is anaesthetized or even dead. The usual explanation of this phenomenon is that
fear and the sensation of the drill's vibration lead patients to be conscious of themselves
as being in pain; they are inaccurately conscious of the fear and the sensation of vibration
as pain.
This explanation of the phenomenon, known clinically as dental fear, is confirmed when
giving patients the explanation leads to their no longer being conscious of (p. 243)
themselves as being in pain when drilling resumes. Still, they strikingly remember the
earlier experience as pain, suggesting that what matters to what the experience is like for
one is simply how one is conscious of it. Less dramatically, robust experimental results in
social psychology show that people confabulate conscious beliefs and desires that fit in
with the way they want to see situations (see Nisbett and Wilson 1977; White 1988).
These findings fit at least as well with an appeal to thoughts in understanding our higher‐
order awareness as with an appeal to perceiving.
The perceptual model has actually led to theoretically problematic results. Thus, Hume's
well‐known problem about the self rests on his assumption that any awareness of the self
would be perceptual, since we never perceive the self, as against the states it is in (Hume
1739–1740/1978: app. p. 634). And the perceptual model has led some to reject TP
altogether, on the ground that we do not perceive our mental states (e.g. Searle 1992: 96–
7).
Page 5 of 17
Higher‐Order Theories of Consciousness
We are seldom conscious of ourselves as having such higher‐order thoughts (HOTs), but
that is to be expected. HOTs are themselves mental states, and no HOT is conscious
unless one has a third‐order thought about it. And we can safely assume that this seldom
happens.
There is good independent reason to posit such HOTs.4 One reason rests on the main way
of determining whether a mental state is conscious in the case of humans, which appeals
to whether one can report being in the state. People are in many (p. 244) mental states
that manifest themselves only in non‐verbal behaviour, but non‐verbal behaviour seldom if
ever reveals whether a state is conscious. If an individual is able to report being in some
state, however, that state is conscious.
But every sincere speech act expresses an intentional state with the same content as the
speech act. So a report that one is in some mental state always expresses a
corresponding thought that one is in that state. Verbally expressing one's consciousness
of a mental state is just expressing one's thought about that state.5
On the foregoing argument, what is essential to these cases is that the mental state in
question is accompanied by a HOT that one is in that state. Since there is ample empirical
evidence that human infants and many if not all mammals have the ability to think, it is
open to find out whether they have at least some HOTs of the requisite kind.6
Other features of the connection between consciousness and speech also point toward
HOTs.7 But there are other reasons to posit HOTs as well. When we consciously see
something, we can be conscious of our visual experience in more or less fine‐grained
ways. One may be conscious of one's experience of red, for example, as a relatively
generic experience of red or as an experience of a highly specific shade.
Sometimes this variability is a function of perceptual conditions, but not always; attention
to one's experience may make one conscious of that experience in more detailed and fine‐
grained ways. HOTs offer a credible explanation of this variability in how we are
conscious of our qualitative experiences. The way we are conscious of a (p. 245)
Page 6 of 17
Higher‐Order Theories of Consciousness
qualitative state, on the HOT hypothesis, is a matter of how the accompanying HOT
conceptualizes that state's mental properties. The way we are conscious of our qualitative
experiences varies with the fineness of grain with which our HOTs represent those
experiences.
A vivid example is the way learning new words for qualitative properties sometimes
results in one's experiences coming to be conscious in respect of those very qualities.
When one is inexperienced in tasting wines, for example, the tastes that result from
different wines may be indistinguishable; similarly with the auditory experiences of
distinct musical instruments to one unfamiliar with those sounds. Simple repetition of the
relevant experiences often is not enough to learn to distinguish them. But often learning
words for the different qualitative experiences helps. How can that be?
Coming to have command of new words for experiences means coming to have command
of concepts for those experiences. But that can help only if those concepts figure in the
way we are conscious of our qualitative experiences, which can happen only by way of
intentional states about those experiences. The effect that learning new words for
qualitative states sometimes has on the way we are conscious of those states is evidence
that we are conscious of those states in virtue of intentional states in which those
concepts figure, that is, by way of HOTs.
Some theorists, however, deny that a state can be qualitative, strictly speaking, and yet
not be conscious. On their view, there would be nothing to a state's having qualitative
character at all if that state is not conscious. But being conscious and having qualitative
character are distinct properties. Conscious qualitative states differ in qualitative
character; one may have the mental quality red and another the mental quality blue. So it
may be that states can resemble and differ from one another in just the ways that
conscious visual sensations of red and blue resemble and differ, but without being
conscious. Robust experimental results, moreover, show that states that are not conscious
can nonetheless vary in their effects in ways characteristic of conscious states with
different mental qualities. Such priming effects make it reasonable to conclude that non‐
conscious states resemble and differ in just the ways that conscious qualitative states do,
but without being conscious (see e.g. Marcel 1983a).
(p. 246)
Page 7 of 17
Higher‐Order Theories of Consciousness
It is sometimes thought that we can describe mental qualities only in respect of the way
we are conscious of them.8 If that were so, there would be nothing to a state's having
mental qualities apart from that state's being conscious. But we can describe mental
qualities independently of their being conscious. We can understand a state's having the
mental quality red, for example, in terms of that state's resembling and differing from
other mental qualities of colour in just the ways that perceptible red resembles and
differs from other perceptible physical colours (see e.g. Sellars 1963: chs. 2, 5;
Shoemaker 1975; Rosenthal 2005: chs. 5–7).
And we can fix these similarity relations among perceptible physical colours by reference
to which physical colours we can perceptually discriminate among, whether or not the
relevant perceptions are conscious (see e.g. Rosenthal 2005: ch. 7).
The monitoring model that underlies inner‐sense theories promised help in understanding
the function of consciousness, on the assumption that monitoring one's thought processes
enhances rationality and the coordinating of desires. But it is unclear that this
assumption, however inviting, is correct. Occasionally we do consciously note that our
plans conflict or that our thinking is unsound. And we may then consciously seek to
correct such plans and thinking. But such rationality results mainly from the causal
connections that thoughts and desires have with one another, in virtue of their intentional
contents. And since thoughts and desires have intentional content independently of their
being conscious, their rationality is itself largely independent of their being conscious.
Indeed, there is experimental evidence that we come to be conscious of our decisions only
after those decisions have been formed (see Libet 1985; Haggard 1999); so consciousness
cannot play a role in determining what we decide even when our decisions are conscious.
Plainly we do not have words for all the subtle variations in qualitative character that
occur consciously in us. So it is unlikely that we have concepts for them. This presents a
challenge for the HOT hypothesis, and an apparent advantage of inner sense. If our
concepts cannot capture all the conscious differences among our qualitative states,
neither can our thoughts. Higher‐order sensations, however, would face no such difficulty,
since their qualitative differences could be as fine‐grained as those of our first‐order
qualitative states.
But we need not have individual concepts for each mental quality to capture all their
qualitative differences conceptually. We routinely use comparative terms to describe
colours that differ only slightly from one another: we say that one is brighter than the
other or has more blue in it or is darker. So it may well be that we are conscious of subtle
differences among the corresponding mental qualities partly in those terms. It does not
seem, of course, as though we are conscious of mental qualities even partly in
comparative terms. But, since the relevant HOTs would seldom be conscious, we have
little reason to trust our pretheoretic intuitions about how HOTs would represent the
subtle differences among conscious mental qualities.
Page 8 of 17
Higher‐Order Theories of Consciousness
It may also seem that distinct HOTs in effect repeat the states they are about, and so are
psychologically superfluous, and even that having HOTs in addition to their targets may
tax or exceed our cortical capacity. These considerations have led Peter Carruthers to
adopt a variant HOT theory, on which a state is conscious if a HOT is disposed to occur,
even if the HOT does not actually occur (see Carruthers 1996, 2000, 2004).
A dispositional theory may seem appealing also because a state's being conscious
coincides with its being reportable and its being introspectible, and these are
dispositional properties. We do not introspect or make reports about all our conscious
states, but it seems that we always have some disposition to do so.
But this overlooks a possibility. When one introspects a state, one has a conscious HOT
about that state; so it is open to argue that what happens in the intermediate case of
ordinary, non‐introspective consciousness is that one has a HOT about the state, but that
HOT is not itself a conscious state. Introspection is the special case in which one's HOTs
are conscious thoughts.
Without knowing how mental states are cortically subserved, moreover, it is idle to urge
that we lack cortical capacity for HOTs; indeed, it is likely that we have cortical capacity
to spare. Nor do HOTs repeat the mental properties of their targets; rather, they
Page 9 of 17
Higher‐Order Theories of Consciousness
represent those states as having those mental properties. So occurrent HOTs are not
psychologically superfluous. Most important, a dispositional theory cannot do (p. 248)
justice to TP, since being disposed to have a thought about something does not make one
conscious of that thing.
When a state is conscious, we are seldom conscious of ourselves as being in two states.
But that is because the higher‐order state, in virtue of which we are conscious of the
target, is typically not itself a conscious state. A theory of consciousness must do justice
to what it's like for one to be in conscious states, but the occurrent HOTs that the HOT
model posits are rarely themselves conscious states.
Whatever the merits of this controversial theory of content, it arguably is not desirable to
hold a theory of consciousness hostage to any such theory. But, that aside, the resulting
theory is no longer dispositional, since it ascribes occurrent higher‐order content to each
conscious state. The only dispositional aspect of the theory is that it explains this
occurrent higher‐order content by positing a disposition the relevant state has to cause a
HOT.
Still, the revised theory resembles a purely dispositional theory in positing no occurrent
state distinct from the state we are conscious of. So it conforms in that way with the
phenomenological appearances, on which a state's being conscious involves only one
state, and not two. The higher‐order content that implements TP is intrinsic to the
conscious state itself.
Page 10 of 17
Higher‐Order Theories of Consciousness
One reason to see such higher‐order content as intrinsic to the state in question is to
avoid the problem of what to say about HOTs that occur in the absence of any suitable
target. Just as mental states can occur without HOTs, so distinct HOTs can in principle
occur without suitable targets, and it may seem unclear what it would be like for one in
such a case. If higher‐order content is intrinsic to the target, this difficulty cannot arise.
(p. 249)
But a problem also seems to arise when there is a target, but the higher‐order content
misrepresents its mental properties. Would what it's like for one in such a case match the
higher‐order content or the target it misdescribes?9 Some theorists have urged that even
such misrepresentation cannot occur if the higher‐order content is intrinsic (Natsoulas
1999; Gennaro 2004). But it is unclear why a state's higher‐order content cannot
misrepresent that state's other mental properties.
In any case, TP itself resolves the problem without positing that the higher‐order content
is intrinsic. On TP, a state is conscious if one is conscious of that state; so the mental
properties in respect of which a state is conscious will be those one is conscious of it as
having. If one has a sensation of red and a distinct HOT that one has a sensation of green,
the sensation of red may nonetheless be detectable by various priming effects. But what
it will be like for one is that one has a sensation of green. Similarly if one has that HOT
with no relevant sensation at all.
Mental qualities can be individuated either by reference to the way we are conscious of
them or, as urged in Section 13.4, by way of some tie they have with perceptible physical
properties. Those, like Joseph Levine (2001), who take the first route, find problematic
the possibility of divergence between qualitative states and the way we are conscious of
them. But they generally regard it as at least conceivable that undetectable quality
inversion, which severs the tie between mental qualities and perceptible properties, can
occur. Individuating mental qualities the second way, by contrast, allows mental qualities
to diverge from the way we are conscious of them, but suggests that undetectable quality
inversion is not even conceivable.10
Empirical results also suggest that the higher‐order content that implements TP belongs
to distinct states. As noted earlier, there is strong evidence that decisions occur in
advance of our being conscious of them. We can best explain that on the hypothesis that
our consciousness of our decisions is due to distinct higher‐order states.
Positing intrinsic higher‐order content also makes it harder to explain why particular
states may be conscious at one moment but not another; why, if the higher‐order content
is intrinsic to the state, does that intrinsic property come and go? The explanatory job is
easier if the higher‐order content belongs to a distinct state.
This problem becomes especially pressing in explaining the shift between a state's being
non‐introspectively conscious and our introspectively scrutinizing it. On the HOT
hypothesis, introspection occurs when a HOT becomes conscious. But if the higher‐order
Page 11 of 17
Higher‐Order Theories of Consciousness
content is intrinsic, this shift is a matter of a state's sometimes being conscious in respect
of its higher‐order content and sometimes not. One might seek to explain that shift as a
shift of attention to the higher‐order content.11 But since shifts of attention need not
make the difference between a state's being conscious or (p. 250) not, they cannot by
themselves explain a shift in whether the higher‐order content is conscious.
Whatever the case with qualitative states, higher‐order content cannot be intrinsic to
intentional states. Intentional states differ not only in content, but also in respect of
mental attitude; two states can share intentional content even though one is a belief,
another a doubt, and a third a case of wondering. And though a single state may, by
conjoining contents, have more than one, no state has more than one mental attitude. No
state is at once a case, for example, of both doubting and believing.
When we are conscious of something by being in an intentional state about that thing, the
state's mental attitude is always assertoric; doubting does not make one conscious of
anything. So the higher‐order content that implements TP must occur in connection with
an assertoric mental attitude. Consider, then, a conscious doubt. It is conscious in virtue
of higher‐order content that one has that doubt, but one cannot doubt that higher‐order
content; one must hold an assertoric attitude towards it. Since distinct mental attitudes
require distinct intentional states, the higher‐order content must belong to a state
distinct from that of the doubt itself (see Rosenthal 2005: ch. 2).
The inner‐sense model may seem to have an advantage here; perhaps the qualitative
character of the higher‐order sensing explains why sensing a qualitative state results in
there being something it's like for one to be in that state. But that advantage is illusory,
since the higher‐order sensing would itself seldom be conscious.
HOTs may seem even less promising here than inner sense. But there is reason to think
HOTs can make the difference between there being something it's like to be in a
qualitative state and there being nothing it's like. Recall that learning new concepts for
more fine‐grained mental qualities sometimes results in one's coming to be conscious of
one's qualitative states in respect of those more fine‐grained qualities. That suggests that
higher‐order conceptual states can also make the difference between there being
something it's like for one to be in qualitative states and there being nothing it's like.
Page 12 of 17
Higher‐Order Theories of Consciousness
Some have urged that an explanation of what makes that difference is acceptable only if
it is rationally transparent. But this demand is excessive. Explanations come to seem
rational not on their own, but only against the background of accepted theory. (p. 251)
Whenever a qualitative state becomes conscious, we seem automatically to recognize its
mental qualities. This suggests that a sensation's being conscious involves purely
recognitional concepts, which apply to the sensation solely in virtue of some ability to
recognize that type of sensation, rather than by way of ties that concept has with other
concepts. And some higher‐order theorists have recently urged that such recognitional
concepts figure in the relevant higher‐order states. But it is unlikely that any concepts
are purely recognitional in this way; conceiving and recognizing involve distinct mental
abilities. And without having an informative explanation of such purely recognitional
concepts, appealing to them would simply dissolve the problem by fiat.
If qualitative states resemble and differ in ways that parallel the similarities and
differences among the corresponding perceptible properties, it is natural to construe our
higher‐order concepts as appealing to such similarities and differences. That helps
explain why HOTs about qualitative states arise at all. The concepts that figure in HOTs
about qualitative states will have connections with concepts for perceptible properties,
which figure in perceiving. So once a creature has such concepts for qualitative
properties, perceiving will itself facilitate the occurrence of the relevant HOTs.
References
Armstrong, D. (1981), ‘What Is Consciousness?’ in Armstrong, The Nature of Mind and
Other Essays (Ithaca, NY: Cornell University Press), 56–67.
—— (2004), HOP over FOR, HOT Theory, in R. J. Gennaro (ed.), Higher‐order Theories of
Consciousness (Amsterdam/Philadelphia, Pa.: John Benjamins), 115–35.
Page 13 of 17
Higher‐Order Theories of Consciousness
Descartes, R. (1964–75), ‘Fourth Replies’, in Oeuvres de Descartes, vii, ed. C. Adam and P.
Tannery (Paris: Vrin).
Levine, J. (2001), Purple Haze: The Puzzle of Consciousness (Oxford: Oxford University
Press).
Libet, B. (1985), ‘Unconscious Cerebral Initiative and the Role of Conscious Will in
Voluntary Action’, Behavioral and Brain Sciences, 8: 529–39.
Locke, J. (1700/1975), An Essay Concerning Human Understanding, ed., from the fourth
(1700) edn., P. H. Nidditch (Oxford: Clarendon, 1975).
Page 14 of 17
Higher‐Order Theories of Consciousness
Natsoulas, T. (1999), ‘The Case for Intrinsic Theory, IV An Argument from How Conscious
Mental‐occurrence Instances Seem, Journal of Mind and Behaviour, 20: 257–76.
Nisbett, R. E., and Wilson, T. (1977), ‘Telling More Than We Can Know: Verbal Reports on
Mental Processes’, Psychological Review, 84: 231–59.
Searle, J. R. (1992), The Rediscovery of the Mind (Cambridge, Mass.: MIT Press).
Sellars, W. (1963), Science, Perception, and Reality (London: Routledge & Kegan Paul).
Van Gulick, R. (2000), ‘Inward and Upward: Reflection, Introspection, and Self‐
awareness’, Philosophical Topics, 28: 275–305.
White, P. A. (1988), ‘Knowing More Than We Can Tell: “Introspective Access” and Causal
Report Accuracy 10 Years Later’, British Journal of Psychology, 79: 13–45.
Notes:
(1) I'll speak interchangeably of being conscious of something and being aware of it.
(2) Locke speaks of ‘internal Sense’ (1700/1975: II. i. 4); the term ‘inner sense’ derives
from Kant (Critique of Pure Reason, A22/B37). Aristotle holds that we perceive that we
perceive, though in the case of thinking he holds that we think that we think (see e.g. de
Anima, Γ2, 425b12–20 and Nichomachean Ethics, IX. 9. 1170a29–34).
Page 15 of 17
Higher‐Order Theories of Consciousness
(3) For more discussion of Lycan's argument that our higher‐order consciousness more
closely resembles perceiving than thinking see Rosenthal (2004: sect. 3).
(4) For more on this see e.g. Rosenthal (1986), reprinted, along with other essays on the
HOT hypothesis, in Rosenthal (2005).
(5) For more on this see Rosenthal (2005: ch. 2). For a qualification on the way the
content of speech acts corresponds to that of the intentional states expressed see
Rosenthal (2005: ch. 12).
(6) Some theorists assume that whenever a creature is conscious, whatever mental states
it is in are themselves conscious states. This assumption seems natural, since whenever
people are awake, and hence conscious, many of their mental states are conscious. But,
however alert we are, not all the mental states we are in are conscious; indeed, many
non‐conscious mental states play a significant role in waking behaviour. So there may
well be creatures none of whose mental states are conscious, even when they are fully
awake. The inference from creature consciousness to state consciousness is unreliable.
(7) E.g. only the HOT hypothesis affords an informative explanation of why, in the human
case, verbally expressed intentional states are always conscious (see Rosenthal 2005: ch.
10).
(9) Karen Neander and Joseph Levine both press this difficulty (Neander 1998; Levine
2001: sects. 4.4, 6.4), and both reject higher‐order theories altogether.
(10) On such undetectable inversion see Rosenthal (2005: ch. 7, sect. 7).
(11) See Christopher S. Hill's notion of volume adjustment (1991: ch. 5, sect. 3) and
Robert Van Gulick's related appeal to volume control (2000: sect. III; 2004: sect. III).
David M. Rosenthal
Page 16 of 17
Representationalist Theories of Consciousness
IT is obvious that some mental states have representational content. Take the case of
beliefs, for example. Beliefs are either true or false. They are true if the world is as their
subjects believe it to be and false otherwise. Each belief, thus, has accuracy conditions—it
is accurate in certain circumstances and inaccurate in others—and any state with
accuracy conditions has representational content.
It is also obvious that some mental states are conscious. Recent representationalist
theories of consciousness hold that there is a deep and important connection between the
sort of consciousness that has most puzzled philosophers—phenomenal consciousness, as
it is often called—and representational content.
Page 1 of 17
Representationalist Theories of Consciousness
Mental states that are inherently conscious are said to be ‘phenomenally conscious’ by
philosophers. But just which mental states are these? One not very informative answer is
that they are experiences. More helpfully, we can classify the relevant states into at least
the following categories: (1) Perceptual experiences; for example, experiences of the sort
involved in seeing green, hearing loud trumpets, tasting liquorice, smelling the sea air,
running one's fingers over sandpaper. (2) Bodily sensations; for example, feeling a twinge
of pain, feeling an itch, feeling hungry, having a stomach ache, feeling hot, feeling dizzy.
Think here also of experiences such as those present during orgasm or while running flat
out. (3) Felt reactions or passions or emotions; for example, feeling delight, lust, fear,
love, feeling grief, jealousy, regret. (4) Felt moods; for example, feeling happy, depressed,
calm, bored, tense, miserable.
Some philosophers claim that there are also such experiences as, for example, the
experience of suddenly remembering something or the experience of understanding a
story. Others insist that in so far as there are experiences in these cases they are simply
various perceptual and/or bodily experiences that accompany memory and
understanding.
Page 2 of 17
Representationalist Theories of Consciousness
In defence of (a) the most famous case offered is that of the inverted spectrum. Here is
one version of that case. Suppose that Tom has a very peculiar visual system. His colour
experiences are systematically inverted with respect to those of his fellows. When Tom
looks at red objects, for example, what it is like for him subjectively is the same as what it
is like for other people when they look at green objects, and vice versa. This peculiarity is
one of which neither he nor others are aware. Tom has learned the meanings of colour
words in the usual way and he applies these words correctly. Moreover, his non‐linguistic
behaviour is standard.
Now, when Tom views a ripe tomato, say, in good light his experience is phenomenally,
subjectively different from the experiences you and I undergo. But his experience has the
same representational content as ours. For his experience is the sort that is usually
produced in him by viewing red objects and that usually leads him to believe that a red
object is present. So he, like you and I, in viewing the tomato has an experience that
represents the tomato as red. But the phenomenal character his experience has is the one
tokened in our experiences when we view green objects, not red ones. So phenomenal
character is not the same as, nor does it supervene upon, representational content.
In defence of (b) above, one well‐known example that is often adduced is that of inverted
earth. Inverted earth is an imaginary planet, on which things have complementary
colours to the colours of their counterparts on earth. The sky is yellow, grass is red, ripe
tomatoes are green, and so on. The inhabitants of inverted earth undergo psychological
attitudes and experiences with inverted representational contents relative to those of
people on earth. They think that the sky is yellow, see that grass is red, etc. However,
they call the sky ‘blue’, grass ‘green’, ripe tomatoes ‘red’, etc. just as we do. Indeed, in all
respects consistent with the alterations just described, inverted earth is as much like
earth as possible.
Page 3 of 17
Representationalist Theories of Consciousness
The example of inverted earth is due to Ned Block (1990). In Block's original version of
the tale, one night while you are asleep a team of alien scientists insert colour‐inverting
lenses in your eyes and take you to inverted earth, where you are substituted for your
inverted‐earth twin or Doppelgänger. Upon awakening, you are aware of no difference,
since the inverting lenses neutralize the inverted colors.1 You think that you are still
where you were before. What it is like for you when you see the sky or anything else is
just what it was like on earth. But after enough time has passed, after you have become
sufficiently embedded in the language and physical environment of inverted earth, your
representational contents will come to match those of the other inhabitants. You will
come to believe that the sky is yellow, for example, just as they do. Similarly, according to
Block, you will come to have a visual experience that represents the sky as yellow. For the
experiential state you now undergo, as you view the sky, is the one that, in you, now
normally tracks yellow things. So the later (p. 256) you will come to be subject to inner
states that are representationally inverted relative to the inner states of the earlier you,
while the phenomenal aspects of your experiences will remain unchanged. It follows that
phenomenal character is not the same as representational content.
Those who accept the classic picture of experience typically think of the phenomenal
character of an experience as an intrinsic property of the experience, a property that in
some cases is accompanied by a representational content. Philosophers who accept the
classic view and who are materialists about phenomenal consciousness hold that once all
the internal, physical facts are fixed for any given individual, the phenomenal character of
each experience of that individual is fixed too. On this view, physically identical
individuals located in different environments may well differ with respect to the
representational contents of their experiences but the phenomenal characters of their
experiences must be the same. Philosophers who accept the classic view but who are
immaterialists about phenomenal character are more liberal. They allow that it is
metaphysically possible for phenomenal character to vary even if all the internal physical
facts are the same.
Page 4 of 17
Representationalist Theories of Consciousness
eye and it may be natural to take the ‘feel’ of an experience to be a property of the idea or
picture. But that isn't common sense. It is philosophical dogma—precisely the dogma
which representational theories of phenomenal consciousness oppose.
phenomenally).
It should be clear why the qualifiers ‘meets certain further conditions’ and ‘relevant’ are added.
Not all states with representational content have any ‘feel’ or phenomenal character at all.
Consider, for example, unconscious beliefs or subpersonal representational states involved in
early vision representing light intensity or zero‐crossings.
Reductive strong representationalism is the view both that the representational content
with which phenomenal character is identical can be spelled out in physical or functional
terms and that the further conditions on that content can be spelled out similarly.
Non‐reductive strong representationalism is the view that either aspects of the content
itself or aspects of the further conditions on the content are neither physical nor
functional. On this view, phenomenal character is identified with representational
content, but the content itself, or conditions on it, are irreducibly subjective.
Corresponding reductive and non‐reductive positions are possible with respect to weak
representationalism.
There is also room for disagreement about whether the contents are Fregean or not. For
the Fregean representationalist the pertinent representational contents are individuated
at least in part by conceptual modes of presentation. For the non‐Fregean the contents
are non‐conceptual. Non‐Fregean contents may be Millian (after John Stuart Mill),
individuated exclusively by worldly entities such as concrete objects and properties, or
hybrid, individuated in part by non‐conceptual, sensory modes of presentation as well as
worldy entities (paradigmatically properties).
Page 5 of 17
Representationalist Theories of Consciousness
The hybrid position here faces some difficult questions that the Millian position does not.
If there are indeed non‐conceptual sensory modes of presentation, is there one generic
mode of presentation of this sort? Is there one for each sense? Or are there modes that
vary within a sense? If so, how finely are they to be individuated? And what is it about
these modes that makes them non‐conceptual?
It is worth emphasizing that the Fregean representationalist can allow that the concepts
whose modes of presentation are at least partly individuative of the relevant contents
need not be concepts the subject of the experiences has, however. Thus, in another sense
of non‐conceptual content (under which the content of state S of person P is non‐
conceptual if and only if P need not possess any of the concepts used by theorists to state
the accuracy conditions for S), Fregean representationalism is compatible with the thesis
that the sort of content relevant to phenomenal character is non‐conceptual content.
(p. 258)
Regarding the entities that enter into the relevant representational contents, on the
representationalist view, it should be clear that none of them can be concrete individuals.
These individuals can play no role as far as phenomenology goes; for hallucinatory
experiences have a subjective ‘feel’ to them and in cases of hallucination there are no
appropriate concrete individuals. This much is agreed upon by all representationalists. I
want now to say something more about the reductive and non‐reductive forms of the view.
Intuitively, visual experiences are not themselves beliefs but they are apt for the
production of beliefs. Admittedly, some states that might reasonably be classified as
visual experiences, for example seeing that the table is covered with books, already
involve beliefs or belief‐like states. But such states seem best taken to be hybrid, having a
visual experience proper and a belief or thought as components.
Visual experiences proper are not apt for the production of any old beliefs, however.
Intuitively, each visual experience is the direct basis for the formation of a belief about
Page 6 of 17
Representationalist Theories of Consciousness
The content of the visual experience proper supplies the input to the relevant belief‐
forming processes, where the role of the belief‐forming processes is to generate beliefs of
the sorts just described. But the appropriate beliefs are not always formed, of course; for
introspection can malfunction and, at least in the case of external belief formation, other
background beliefs can interfere. There is also the possibility that attention is not
appropriately directed.
A visual experience has a poised content, then, so long as it is apt for the production in
the right ways of the right beliefs. Likewise for the other senses. In the case of bodily
experiences, desire is also relevant. The experience of pain, for example, is the direct
basis for the desire to protect oneself, to avoid damage.
(p. 259)
Putting these points together, one well known reductive representationalist proposal is
that the phenomenal character of an experience is one and the same as its poised
abstract non‐conceptual intentional (or representational) content or its PANIC, for short
(see Tye 1995: 2000).3
Turning now to the non‐reductive form of representationalism, as noted earlier there are
two ways in which such a view might be developed: either some of the entities out of
which the relevant contents are built are themselves non‐physical or the further
conditions these contents must meet are neither physical nor functional. Of these two
possibilities, only the first has been worked out in any detail.4 Here the most
straightforward proposal is that the content itself is Millian but that it involves non‐
physical qualities—qualities that experiences represent as belonging to things in the
world or part of the body, but qualities that the represented items do not really possess.
These qualities are projected on to the world by experiences; in reality either nothing at
all has them or the experiences themselves do. It is worth noting that the last of these
Page 7 of 17
Representationalist Theories of Consciousness
proposals has features in common with the classic view of experience. For it is now
conceded that experiences have special properties that are relevant to phenomenal
character. What is denied is that these properties are phenomenal characters.
Experiences have phenomenal character, on this view, via their (mistakenly) representing
such qualities as belonging to things other than experiences.5
Suppose that you are standing before a tapestry in an art gallery. As you take in the rich
and varied colours of the cloth, you are told to pay close attention to your visual
experience and its phenomenology. What do you do? The representationalist says that you
attend closely to the tapestry and details in it. You are aware of something outside you—
the tapestry—and of various qualities that you experience as being qualities of parts of
the tapestry, and by being aware of these things, you are aware of what it is like for you
subjectively or phenomenally. But your awareness of what it is like, of the phenomenology
of your experience, is not de re awareness of the experience or its qualities.6 It is de dicto
awareness that you have an experience with a certain phenomenal character or ‘feel’.
Here is another example to illustrate these preliminary points. Suppose that you have just
entered a friend's country house for the first time and you are standing in the living room,
looking out at a courtyard filled with flowers. It seems to you that the room is open, that
you can walk straight out into the courtyard. You try to do so and, alas, you bang hard
into a sheet of glass, which extends from ceiling to floor and separates the courtyard from
the room. You bang into the glass because you do not see it. You are not aware of it; nor
are you aware of any of its qualities. No matter how hard you peer, you cannot discern
the glass. It is transparent to you. You see right through it to the flowers beyond. You are
aware of the flowers, not by being aware of the glass, but by being aware of the facing
surfaces of the flowers. And in being aware of these surfaces, you are also aware of a
myriad of qualities that seem to you to belong to these surfaces. You may not be able to
name or describe these qualities but they look to you to qualify the surfaces. You
experience them as being qualities of the surfaces. None of the qualities of which you are
directly aware in seeing the various surfaces looks to you to be a quality of your
experience. You do not experience any of these qualities as qualities of your experience.
For example, if redness is one of the qualities and roundness another, you do not
experience your experience as red or round.
(p. 261)
Page 8 of 17
Representationalist Theories of Consciousness
If your friend tells you that there are several ceiling‐to‐floor sheets of glass in the house
and that they all produce a subtle change in the light passing through them so that things
seen on the other side appear more vividly coloured than is usually the case, as you walk
gingerly into the next room you may become aware that there is another partitioning
sheet of glass before you by being aware of the qualities that appear to belong to non‐
glass surfaces before your eyes. You are not aware of the second sheet of glass any more
than you were aware of the first; but you are now aware that there is a sheet of glass in
the room by being aware of qualities apparently possessed by non‐glass surfaces before
you.
Visual experiences, according to the representationalist, are like such sheets of glass.
Peer as hard as you like via introspection, focus your attention in any way you please, and
you will only come across surfaces, volumes, films, and their apparent qualities. Visual
experiences thus are transparent to their subjects. We are not introspectively aware of
our visual experiences any more than we are perceptually aware of transparent sheets of
glass. If we try to focus on our experiences, we ‘see’ right through them to the world
outside. By being aware of the qualities apparently possessed by surfaces, volumes, etc.,
we become aware that we are undergoing visual experiences. But we are not aware of the
experiences themselves.7
This is true even if we are hallucinating. It is just that in this case the qualities apparently
possessed by surfaces, volumes, etc. before our eyes are not so possessed. The surfaces,
volumes, etc. do not exist.
Returning now to the first example, the representationalist claims that when you are told
to attend closely to your visual experience, as you view the tapestry, what you actually do
is to attend closely to the tapestry and the various ways it, or parts of it, look to you. Ways
things look to people are typically expressed by predicates (p. 262) (as in ‘looks red’,
‘looks square’, ‘looks close’), and predicates express qualities—in this case, qualities
represented by the relevant experiences of those people, qualities such that if the things
Page 9 of 17
Representationalist Theories of Consciousness
seen have them, the experiences are veridical or accurate. Change the qualities ‘out
there’, more specifically the qualities which are the ways the tapestry looks, and
necessarily the ‘feel’ of your experience changes.
What is true for vision is true for the other senses. Attending to the phenomenology of a
perceptual experience, to its felt character, is a matter of attending to the ways things
look, smell, taste, sound, or feel by touch. In the case of bodily sensations, the object of
your attention is the way a certain part of your body feels.9 With emotions and moods, the
attentional focus is often on things outside—things perceived as dangerous or foul or
pleasing—but there is also attention to ways one's body is changing (pounding heart,
shaky legs, higher blood pressure). More generally, attention to phenomenal character is
a matter of attention to the ways things other than the experience seem; that is, to
qualities that are not qualities of experiences. Change any of the qualities that are the
various ways things look, smell, sound, etc. and necessarily the phenomenal character of
the experience changes. Why should this be? The answer the strong representationalist
proposes is that phenomenal character is identical with a certain sort of representational
content into which the relevant qualities enter.
Some representationalists take a weaker stand. They say merely that once the qualities
that are directly experienced are fixed, the phenomenal character is fixed. Either way, the
appeal to transparency is taken by intentionalists to be a key supporting argument for
their position.
The second argument for the representationalist position I call ‘the many‐property
problem revisited’.10 In the 1970s Frank Jackson presented a problem for the adverbial
approach to visual experience involving multiple after‐images. He called this problem ‘the
many‐property problem’. The problem in essence was this: If what it is to have a sensory
experience of something red is to sense redly, as adverbialists hold, then sensing
something red and round should be a matter of sensing redly and roundly. So having a
red, square after‐image at the same time as a green, round one is to be analysed as
sensing redly and squarely and greenly and roundly. Since the conjuncts here can be
commuted, this is the same as sensing redly and roundly and squarely and greenly. Thus,
the distinction between having a red, square after‐image and a green, round one and
having a red, round after‐image and a green, square one is lost.
Now evidently the phenomenal character of a visual experience of a red, round object and
a green, square one is different from the phenomenal character of a green, round object
and a red, square one. How is this difference to be accounted for? If the phenomenal
character of an experience is an intrinsic property of that experience, then it seems that
the phenomenal character of the first experience must (p. 263) be a conjunctive property
of the experience that is constituted by conjoining the phenomenal characters of
experiences of red things, round things, green things, and square things. This, however,
leaves no room for a phenomenal distinction between the first experience and the second
one (the experience, that is, of a green, round object and a red, square one). To reply that
the first phenomenal character is more complex in that it has a further conjunctive
Page 10 of 17
Representationalist Theories of Consciousness
constituent corresponding to the relative spatial position of the green, round object and
the red, square one is not to solve the problem. For this simply adds a further conjunct
and the problem arises again.
One solution to this problem is to say that each experience has real objects, sense‐data,
that are present whether or not the experience is veridical. The phenomenal character of
the first experience is really a matter of its being an experience that relates its subject to
two sense‐data, one of which is red and square and the other of which is green and round
(in whatever senses of these terms are deemed applicable to sense‐data). Since the
phenomenal character of the second experience involves sense‐data with different
qualities, the phenomenal difference between the two experiences is preserved.
The price paid for this solution is the admission of sense‐data. Since the introduction of
sense‐data via the argument from illusion is not well motivated and there is a whole host
of problems generated by them, this solution seems unacceptable.
So, what is the answer? The representationalist will say that the answer is simple. The
first experience has different accuracy conditions and thus a different representational
content from the second. Since phenomenal character is the same as representational
content (or is metaphysically determined by it), the difference in phenomenal character is
straightforwardly preserved.
The third argument for representationalism points to the intuitive connection between
the phenomenal character of an experience and the way things seem to the person
undergoing the experience. If things subjectively seem the same to you and me, then
intuitively what it is like for me subjectively is the same as what it is like subjectively for
you; and if what it is like subjectively for the two of us is the same then things seem
subjectively the same to us. Since the way things seem subjectively is naturally taken to
be the representational content of the experience just as the way things are believed to
be is standardly taken to be the content of a belief, and sameness of phenomenal
character is no more or less than sameness in ‘what it is like’, phenomenal character
necessarily co‐varies with representational content. Why should this be? The answer,
according to the strong representationalist, is that phenomenal character just is
representational content of a certain sort.
Page 11 of 17
Representationalist Theories of Consciousness
The fourth argument for the representationalist view is in general form extremely simple.
Many very smart philosophers have tried their best to come up with clear, actual cases of
experiences which are identical in their representational contents but which differ
phenomenally.12 They have appealed to blurry vision, double vision, vision with eyes
closed, phosphenes, experiences with one eye open versus two, experiences of tilted
versus untilted coins, gestalt effects of various sorts, the experience of pain, the
experience of elation, experiences of trees at varying distances away, peripheral vision,
seeing versus touching a round object, and many other actual phenomena.
Representationalists have replied to these cases one by one and in detail; and they have
made a plausible case for the view that, in each of these cases, where there is a
phenomenal difference, there is a representational difference too. That leaves the
opponents of representationalism with possible cases. And here the inverted spectrum
and inverted earth loom large. These cases can be handled by the representationalist too
(see Section 14.5). So it seems that no one as yet has managed to describe a clear‐cut
example of representational identity and phenomenal difference among experiences. Why
should this be? The simplest and most straightforward explanation is that phenomenal
character is a certain kind of representational content. On this view, once all the
representational facts are fixed, the phenomenal facts are automatically fixed too. Let us
call this argument ‘the absence of counter‐example argument’.
Obviously, this argument does not prove that the representationalist view is correct, even
granting the assumption (which some philosophers would certainly contest) that as yet no
clear actual or possible counter‐example has been adduced. But few arguments in
philosophy are even close to proofs.
The fifth argument I want to mention I call ‘the appeal to phenomenal indeterminacy’.
This may be illustrated by the case of blurriness in visual experience. As noted above, this
phenomenon is sometimes adduced by opponents of representationalism as creating a
difficulty for the view. But representationalists maintain that this could not be further
from the truth. Consider the example of the visual image of a speckled hen. How many
speckles are there? This question posed an insuperable problem for the sense‐datum
theory. For to the subject of the experience it is indeterminate how many speckles are
present; yet there seems no room for such indeterminacy on the sense‐datum proposal.
Not only does representationalism give a straightforward and (p. 265) satisfying account
of such phenomenal indeterminacy but it also enables us to understand how there could
be a phenomenal difference between seeing blurrily and seeing clearly something blurry.
In the former case there is phenomenal indeterminacy with respect to the number of
speckles, since there is no number of speckles, N, such that the visual experience
represents a hen with N speckles. The visual experience here does not ‘comment upon’
the precise number of speckles. It leaves open how many speckles are present just as
does the linguistic report that there is a speckled hen in the garden. In the latter case one
sees blurrily just in case one undergoes a visual experience that represents fewer surface
details than are represented when one sees clearly. By contrast, one sees clearly
something blurry (e.g. a water‐colour picture with blurry edges) just in case one's visual
experience represents that the boundaries of the blurry thing lie between precisely
Page 12 of 17
Representationalist Theories of Consciousness
located spatial regions A and B without representing exactly where they lie. There is a
phenomenal difference between the two cases, the representationalist will say, precisely
because there is a representational difference. No other theory of phenomenal character
can give a satisfying and plausible account of this difference.13
(p. 266)
The second point to make is that even if it is metaphysically possible for some creatures
to undergo colour experiences that are phenomenally inverted relative to those of others
without there being any difference in narrow functioning or behaviour, this does not
refute the representationalist. For the representationalist can insist that in such a case
there is still a difference in representational content. To appreciate how this can be so,
consider the following case.
Imagine that in a country in Latin America after a revolution car tyres are replaced as
they wear out in a haphazard way and that at some given time in this country there are
forty‐seven cars travelling on the road, all with replaced tyres, and all going at exactly 30
mph. The speedometers in these cars read different things: some read 30, some read 31,
some read 37, some read 28, and so on. Is there any way of telling which of the forty‐
seven speedometers is representing accurately actual car speed just by looking at the
speedometer? Of course not. The speedometers were designed for use with tyres of a
Page 13 of 17
Representationalist Theories of Consciousness
certain size. They give an accurate read‐out of the speed just in case they are operating
as they were designed to operate with tyres of the appropriate size.
Similarly in the case of colour. The colour‐experience system in the brain was designed by
evolution for use with an optic nerve and retina meeting certain further conditions. There
is no way of telling who, among present‐day actual colour perceivers, has the retinal
apparatus and optic nerve meeting the historical design specifications (Dretske 1995).
But whoever does gets the colours of things right in experience. What is true in this case,
the representationalist can say, is true in the metaphysically possible case in which there
is a phenomenal difference unaccompanied by any difference in narrow functioning or
behaviour. Here the perceivers who get the colours of things right are those whose visual
systems are operating in the right historical way. But there is no way of knowing from
current behaviour just which these perceivers are.
The response to the inverted‐earth case is similar. The person who switches to inverted
earth and who is wearing colour‐inverting lenses has the representational contents of his
visual experiences fixed by the historical setting in which his visual system was designed
to be used. Since the experiential state he is in on inverted earth when he views red
things is the state that in his species was designed to track green things on earth under
optimal conditions, it represents green and not red. This remains the case however long
he remains on inverted earth. Furthermore, this difference between Mr Invert and those
who have always lived on inverted earth is one that will show up in subtle behavioural
differences which do not change through time. For example, if Mr Invert is shown two
colour patches, one yellow and one blue, each at maximum saturation, he will classify the
blue patch as brighter, whereas the locals will classify the other patch as the brighter
one.
The upshot is that the representationalist about phenomenal consciousness has plausible
replies to the standard cases of the inverted spectrum and inverted earth.
References
Block, N. (1990), ‘Inverted Earth’, Philosophical Perspectives, 4: 53–79.
Boghossian, P., and Velleman, D. (1989), ‘Colour as a Secondary Quality’, Mind, 98: 81–
103.
Chalmers, D. (2006), ‘Perception and the Fall from Eden’, in J. Hawthorne and T. Gendler
(eds.), Perceptual Experience (Oxford: Oxford University Press), 49–125.
Page 14 of 17
Representationalist Theories of Consciousness
—— (2003), ‘Mind and Illusion’, in A. O'Hear (ed.), Minds and Persons (Cambridge:
Cambridge University Press), 251–71.
Shoemaker, S. (1990), ‘Qualities and Qualia: What's in the Mind’, Philosophy and
Phenomenological Research, 50: 109–31.
Notes:
Page 15 of 17
Representationalist Theories of Consciousness
(3) For detailed discussion and criticisms of this theory, along with replies, see the web
symposium at http://host.uniroma3.it/progetti/kant/field/tyesymp.htm, accessed 2008. See
also the shorter symposium on Tye (2000) in Philosophical Studies (Tye 2003).
(5) Another view not dissimilar to this one is held by Sydney Shoemaker (1990, 1994).
There are two main differences. For Shoemaker, experiences have intrinsic qualities that
they correctly represent things in the world as producing or as being disposed to produce.
Moreover, these qualities have a physical nature. Neither the physical nature of these
intrinsic qualities nor the relational nature of the properties external things are
represented as having (properties such as the property of being disposed to produce an
experience with such‐and‐such an intrinsic quality in normal perceivers) is revealed to the
subjects of the experiences. Even so, according to Shoemaker, the phenomenal ‘feel’ of
the experiences is given by the relevant representational contents.
(6) This claim, it is worth noting, fits well with the linguistic constructions that are
naturally employed in connection with such awareness. To talk of our being aware of the
phenomenology of an experience or of how an experience ‘feels’ is to use a generic
perceptual verb (‘aware of’) followed by an abstract noun (‘the phenomenology’) or an
interrogative nominal (‘how the experience feels’). In cases of this sort, where there is a
perceptual verb, the abstract noun or interrogative nominal typically stands in for a
factive clause so that what is being described is (a species of) awareness of some fact. For
example, if I am described as hearing the answer to your question or as seeing who is at
the door, I do not satisfy the description merely by hearing the sentence that is the
answer or seeing the person who is at the door. I must be aware that the given sentence
is the answer to your question, that the given person is the one at the door. In short, I
must be aware of some appropriate fact. Likewise in the case of awareness of the
phenomenal character of a current experience. For more here see Dretske (1993).
(7) This claim is one that the sense‐datum theorists would have endorsed, although they
would have insisted that the things apparently outside are really immaterial surfaces or
sense‐data rather than physical surfaces. After all, sense‐datum theorists were at pains to
distinguish the act of sensing from the thing sensed. G. E. Moore (1903) is the modern
father of transparency.
(8) There are dissimilarities. In typical cases of seeing‐that, background beliefs play a role
in generating the propositional state of awareness. This is not so in the case of
introspection. For more here see Tye (2000: ch. 3).
(9) I ignore here the case of phantom‐limb sensations, for which there is no relevant limb
to feel any way.
Page 16 of 17
Representationalist Theories of Consciousness
(11) I am assuming here that the opponent of representationalism will say that the person
with phenomenally inverted colour experiences nonetheless is such that ripe tomatoes
and other red things look or seem red to him (or her).
(12) See here McGinn (1982), Peacocke (1983), Searle (1983), Boghossian and Velleman
(1989), Block (1996), among others.
Michael Tye
Page 17 of 17
Sensory Qualities, Sensible Qualities, Sensational Qualities
Keywords: sensory qualities, sensible qualities, sensational qualities, perception philosophy, external objects,
qualitative character
Page 1 of 15
Sensory Qualities, Sensible Qualities, Sensational Qualities
apportioned to the subject matter itself. Clark opens the explanation of his topic as
follows:
A whiff of lilacs presents a particular sweet odour. The warmth of the rising sun
yields certain tactile sensations. Bees' honey has a specific taste. The qualities
that characterize the smell of the lilacs, sensation of the sun, or taste of the honey
are all what I will call sensory qualities.
(1993: 1).
It is not immediately clear from these two paragraphs what sensory qualities are qualities of.
Consider the ‘sweet odour’ of lilacs. We may take the odour of lilacs to be the same as ‘the smell
of the lilacs’, in which case one quality that presumably ‘characterize[s] the smell of the lilacs’ is
sweetness. And what is the smell of the lilacs? One candidate—reading ‘the smell of the lilacs’
like ‘the colour/size of the lilacs’—is a certain odiferous property of the lilacs. On this reading,
sweetness ‘characterizes’ the smell of the lilacs in the sense that it is identical to the smell; it is
identical to an odiferious property of the lilacs. Another candidate for the smell of the lilacs—
leaning heavily on the fact that smells ‘drift’, ‘linger’, and are ‘given off’ by objects like lilacs—is
a certain volume of gas, an ‘effluvium’ (Gibson 1966: 144), or ‘vaporous emanation’ (Lycan 1996:
146). On this rival account, sweetness ‘characterizes’ the smell of the lilacs in the sense that it is
a property of the smell—the effluvium emitted by the lilacs. In any case, there do not seem to be
any mental candidates for the smell of the lilacs: if the smell is anywhere, it is in the lilac bed,
not the mind.
So if we focus on Clark's first example, sensory qualities are not properties of anything
mental. But his second example—the ‘sensation of the sun’—and the tell‐tale ‘what it is
like’ locution suggest exactly the opposite, that sensory qualities attach to mental items
(experiences, probably). However, Clark goes on to say that sensory qualities
‘characterize the way such things [the blooms, sun, and honey] appear’. Since the blooms
appear red, the sun appears hot, and the honey appears sweet (suppose), sensory
qualities ‘characterize’ redness, hotness, and sweetness, for example. Either this means
that sensory qualities are (higher‐order) properties of redness, hotness, etc., or else it
means (a more likely interpretation) that the redness, hotness, etc. are examples of
sensory qualities. Either way, sensory qualities are not—or are not obviously—properties
of experiences.
This is somewhat unfair to Clark: it is perfectly clear from other passages that sensory
qualities are stipulated to be properties of experiences or, as he prefers, ‘sensations’:
Page 2 of 15
Sensory Qualities, Sensible Qualities, Sensational Qualities
for example, the internal state in virtue of which I now see the red rug (p. 270)
and without which I would not see the red rug is (by stipulation) my visual
sensation ‘of’ the rug. Sensory qualities are qualities of sensation, construed in
this broad way.
(1993: 3–4)
A number of questions can be raised about this explanation. First, there is a contrast between
states and events: states are properties or conditions, not occurrences or events, although
frequently philosophers of mind use ‘state’ in a semi‐technical umbrella sense to include events.
Clark might be assuming that sensations are events, and using ‘state’ in the broad sense. If so,
then sensory qualities are properties of particulars; alternatively, if Clark is using ‘state’ in the
narrow sense, they are properties of properties. If it is unclear which option to choose, that casts
some doubt on the idea that ‘sensory quality’ is a label for something we all pretheoretically
recognize.
Second, is it even true that ‘[w]henever any creature perceives something, there is some
internal state of the creature in virtue of which it perceives that thing’? Setting aside the
state/event distinction, much hangs on the interpretation of ‘in virtue of’, which Clark
leaves unexplained. One suggestion would be that seeing the red rug can be decomposed
into an internal state (the ‘sensation’), the red rug, and an appropriate causal relation
connecting the rug to the internal state. But that would require an argument: the
decomposition thesis is considerably stronger than the relatively uncontroversial claim
that a causal connection is necessary for seeing.
Third, are we aware of sensations and their sensory qualities? That might sound like a
silly question—one is frequently aware of one's bodily sensations (‘sensations proper’)
and some of their properties. I am aware of a twinge, its location (my left elbow), and its
achiness. That sort of awareness seems (quasi‐)perceptual—rather like the awareness I
have of a flash of light, its location (the far side of the room), and its brightness. However,
on this conception of (bodily) sensations, they are not experiences; rather, they are
objects of experiences. The twinge is—or at any rate appears to be—some sort of
disturbance in my elbow, not in my mind or head, unlike the presumed ‘experience’ of the
twinge. Clark, however, has a very different conception of bodily sensations, on which the
experience of the twinge, not the twinge, counts as a bodily sensation; hence his
extension of the use of ‘sensation’ to cover paradigmatic perceptual experiences. Without
distinguishing between the two conceptions of sensations it might appear evident that we
are aware of sensations (as Clark conceives of them) and their sensory qualities; in fact, it
is far from evident.
For these reasons, it should not be taken for granted that Clark has succeeded in
circumscribing his purported object of study. For another attempt, consider how Levine
introduces sensory qualities in Purple Haze—a synthesis of his influential work on the
‘explanatory gap’:
Page 3 of 15
Sensory Qualities, Sensible Qualities, Sensational Qualities
Let's take my current visual experience as I gaze upon my red diskette case, lying
by my side on the computer table. I am having an experience with a complex
qualitative character, one component of which is the colour I perceive. Let's dub
this aspect of my experience its ‘reddish’ character … Qualitative character
concerns the ‘what’ it's like for me: reddish or greenish, painful or pleasurable,
and the like. From within the subjective point of view I am presented with these
qualitative features of experience, or ‘qualia,’ [aka sensory qualities] as they are
called in the literature. Reddishness, for instance, is a feature of my experience
when (p. 271) I look at my red diskette case. It is notoriously difficult to explain
this feature by reference to either the physical or formal features of my brain
states.
(2001: 6–7)
Plainly Levine intends sensory qualities (‘qualitative character’, ‘qualitative features’, ‘qualia’) to
be properties of experiences, although he says little about what experiences are supposed to be.
Suppose one is looking at Levine's red diskette case, balanced precariously on the computer
table. One might be aware of a state of the diskette case (its redness), and an event involving the
diskette case (its falling off the table). Is it so obvious that one can also be aware of, or be
‘presented with’, a ‘qualitative feature’ of a mental event or state—one's experience of the
diskette case? Uncontroversially, one is in a position to know that one sees the diskette case, and
that it looks red to one. What is not immediately clear is whether this piece of self‐knowledge
requires any kind of awareness of a mental event or state. Indeed, the very existence of an
appropriate event is a matter of some doubt. ‘One's seeing the diskette case’, for example, is not
a good candidate for such an event—the nominal ‘one's seeing the diskette case’ behaves more
like the name of a fact than of an event (see Bennett 1988: ch. 1). (On the difficulty of squeezing
some philosophical juice out of the word ‘experience’ see Hinton 1973: pt. I.)
Of course, one can easily play along with Levine's terminology, at least to some extent, by
saying ‘My experience is reddish’ iff it looks to one that something is red. But Levine does
not intend ‘reddish experiences’ to be a stylistic variant of ‘looks red’ and similar
expressions (and this chapter will follow Levine's terminological intentions). Levine
evidently assumes that we have some independent handle on reddish experiences, and
that the relation between reddish experiences and how things look with respect to colour
is an open question. The role played by the red appearance of the diskette case is merely
to fix the reference of ‘reddish’; the relation between reddish experiences and the visual
representation of the colour red might well be entirely contingent. And, in fact, that is
exactly what Levine thinks (2001: ch. 4).1
Page 4 of 15
Sensory Qualities, Sensible Qualities, Sensational Qualities
(p. 272)
Levine's sensory quality of reddishness, then, is not redness, the colour. Even if tomatoes
are not in fact red, at least they appear red. And one's experiences do not appear red at
all. Levine himself is quite explicit that redness and reddishness are distinct (2001: 179 n.
5).
Redness can be called a sensible quality—a quality that one perceives, like roundness,
motion, and sweetness. (To say that one ‘perceives redness’ when looking at a tomato is
not to presuppose that the tomato is red.) Clark's A Theory of Sentience distinguishes
between sensible and sensory qualities, using the terminology of ‘phenomenal’ and
‘qualitative’ properties (2000: 2). Although sensible and sensory qualities should not be
identified, presumably they are closely related. But before canvassing one popular
account of their relation (in Section 15.5) it needs to be pointed out that separating the
two is, for the most part, a comparatively recent innovation. Hume, for example, reports
that the ‘fundamental principle’ of the ‘modern philosophy’ ‘is the opinion concerning
colours, sounds, tastes, smells, heat and cold; which it asserts to be nothing but
impressions in the mind’ (1739–40/1978: I. iv). By saying that colours are ‘impressions in
the mind’, and so ‘internal existences’, Hume is (arguably) saying that items ‘in the mind’
are coloured. And if we suppose that the items are experiences (for an alternative, see the
following section), then Hume is apparently identifying sensible and sensory qualities.
[f]or the purposes of this discussion we shall employ the term sensory ‘qualities’ to
refer to all the different attributes or dimensions with regard to which we
differentiate in our responses to different stimuli. We shall thus use this term in a
wide sense in which it includes not only quality in the sense in which it is
contrasted with intensity, extensity, clearness, etc., but in a sense in which it
includes all these other attributes of a sensation. We shall speak of sensory
qualities and the sensory order to distinguish these from the affective qualities
and the other mental ‘values’ which make up the more comprehensive order of
‘mental qualities’.
(1952: 2)
Sensory qualities, then, are mental qualities, specifically ‘attributes of a sensation’. And
sensations, or ‘sense experiences’ are ‘events taking place in some organism’ (Hayek 1952: 3,
16).
Immediately after giving this explanation Hayek says:
Page 5 of 15
Sensory Qualities, Sensible Qualities, Sensational Qualities
(1952: 2)
The sense perceptions have been definitely eliminated from physical acoustics,
optics, and heat. The physical definitions of sound, colour, and temperature are
today in no way associated with the immediate perception of the respective
senses, but sound and colour are defined respectively by the frequency and
wavelength of oscillations, and temperature is measured (p. 273) theoretically on
the absolute temperature scale corresponding to the second law of
thermodynamics.
(Planck 1925: 5)
So the sensory qualities—as Hayek goes on to say—include ‘colours, sounds, odours, feeling of
touch, etc.’ (1952: 3). Science has only turned up correlates of these qualities in our external
environment—‘the frequency and wavelength of oscillations’—not the qualities themselves.
Suppose that Hume and Hayek are right, and that nothing in the external world is
coloured or noisy or tasty. What remains quite unclear is why they think items in the
‘internal’ world—sensations, for instance—have these qualities. Why not say instead that
absolutely nothing has these qualities?
Whatever the answer to this question might be, the introspective psychologists tended to
take it as simply obvious that ‘sensations’ had colours, sounds, and so forth:
Sensations have always been distinguished by their qualities. Every sensation can
be said to have an attribute of quality, which designates it as red or yellow or
bitter or cold or C#.
Returning to Clark's Sensory Qualities, soon it becomes apparent that he is working squarely
within this framework. Examples of ‘a particular sensory quality’ include ‘a red colour or a
camphoraceous odour’ (1993: 128). So the familiar colour solid, with the colours arranged along
dimensions of hue, brightness, and saturation, is taken to be a space of sensory qualities; and the
main task of Clark's book is to supply a broadly physicalistic theory of such spaces.
One way of justifying this surely rather odd talk is to take ‘red’, ‘camphoraceous’, and so
forth to be ambiguous. In one sense ‘red’ refers to a property of tomatoes and
strawberries; in another sense ‘red’ refers to a property of mental items: experiences,
sensations, or whatever. Locke is often supposed to hold such a view. (For discussion see
Stuart 2003.) However, taken as a descriptive semantic claim about English, the
ambiguity thesis is not very appealing. Phrases like ‘red experience’ and ‘red sensation’
cannot be adduced in its support, since these are examples of philosophical jargon, not
Page 6 of 15
Sensory Qualities, Sensible Qualities, Sensational Qualities
ordinary English. ‘An experience of red’ is fairly natural, but here there is no reason to
suspect that ‘red’ has a second semantic value. (The phrase ‘the Taj Mahal’ in ‘an
experience of the Taj Mahal’ refers to the mausoleum, not to anything mental.)2
this is not the question at issue … In order to see what the issue is, it seems to me
to be absolutely essential to see that such words as ‘blue’ and ‘bitter’ may be used
in two very different senses. When we say of such things as a tie or a flag or an
india rubber ball that they are blue, what we are saying about them may be
something very different from what we are saying of an after‐image, which we see
with closed eyes, when we say that it is blue; and when we say of quinine or
wormwood that they are bitter, what we are saying of them is certainly something
very different from what we are saying of a taste which we are tasting or a taste
‘which is in our mouth,’ when we say that it is bitter.
(1942: 655)
Forget for the moment about the wormwood, and concentrate on the alleged ambiguity in ‘blue’.
In one sense it stands for a property of external objects like ties, rubber balls, and so on. In
another sense, it stands for a property of an after‐image ‘or any other sense‐datum’ (Moore
1942: 658). In this second sense ‘blue’ stands for what Moore calls ‘the sensible quality “blue”
’ (1942: 658); to avoid confusion, let us adapt some terminology from Peacocke (1983) and call
this quality instead ‘the sensational quality blue’.3
We now have a third quality to contend with, at least nominally: the sensational quality
blue, in addition to the sensible quality blue and the sensory quality blue (i.e. bluishness,
as Levine would call it). Again borrowing some terminology from Peacocke, call the
sensational quality blue, blue′ (‘blue‐prime’). Blue′ is a property of sense‐data: objects of
experiences, not experiences themselves. According to Moore, when a tie looks blue to
one, one is (or can be) aware of something else, an immaterial (or at least not clearly
material) thing that appears blue′. Even though the tie may not in fact be blue, the sense‐
datum is as it appears, and so is blue′.
Page 7 of 15
Sensory Qualities, Sensible Qualities, Sensational Qualities
Is there such a property as blue′? If ‘blue’ has ‘two different senses’ when we speak of a
‘blue after‐image’ and a ‘blue tie’, then there is at least an initial reason for thinking so.
But there appears to be no ambiguity: ‘My after‐image and this tie are both exactly the
same shade of blue’ would strike us as false if ‘blue’ were ambiguous between ‘the colour
blue’ (not a property of after‐images) and ‘blue′’ (not a property of ties). On the contrary,
this sentence seems unproblematically true. Of course, it might not in fact be true.
Offhand, the sentence is true only if there are such things as after‐images, and arguably
there are no after‐images, there only seem to be (see Smart 1959). But this would trace
the falsity of the sentence to the metaphysics of after‐images, rather than to an ambiguity
in ‘blue’.
After‐images appear to be located on the focused surface (Emmert's law), and can be
mistaken for coloured patches. This would be mysterious if after‐images did not (p. 275)
appear coloured. Further, as explained in any textbook on colour vision, there are
theories of after‐images. These theories explain why, for instance, staring at a yellow
surface will produce a blue after‐image. If after‐images are not blue (or do not appear
blue), the widely accepted characterization of the data to be explained is wrong.
Moore's case for blue′ is uncompelling. Let us briefly turn to his second example, of
‘bitter’. Here the claim of ambiguity (or, better, polysemy) is considerably more plausible.
Wormwood is bitter because it has a bitter taste, just as a skunk is malodorous because it
produces a malodorous smell, and (to borrow Aristotle's example), an apple is healthy (in
the secondary sense) because it is conducive to a healthy body (in the primary or focal
sense). It is at least strained to say that wormwood and the taste it produces are both
bitter.
Have we not found a sensational quality, bitter′, which attaches to gustatory sense‐data?
Not obviously: on the face of it, a bitter taste is some kind of bitter ‘emanation’ produced
by wormwood, just as the sweet smell of lilacs is (on one understanding) a sweet
effluvium produced by lilacs. Notice that Moore can't quite keep his scare quotes off the
phrase, but there is no apparent reason why the taste should not be (spatially) in the
mouth.
Page 8 of 15
Sensory Qualities, Sensible Qualities, Sensational Qualities
conceived as one of working out what the distal stimulus must be like, given that it
produces these effects in us.
Thus Russell:
if we take any common object of the sort that is supposed to be known by the
senses, what the senses immediately tell us is not the truth about the object as it
is apart from us, but only the truth about certain sense‐data which, so far as we
can see, depend upon the relations between us and the object. Thus what we
directly see and feel is merely ‘appearance’, which we believe to be a sign of some
‘reality’ behind. But if the reality is not what appears, have we any means (p. 276)
of knowing whether there is any reality at all? And if so, have we any means of
finding out what it is like?
(1912/80: 6)
This sort of view was set out with particular clarity by Thomas Reid. When a tomato is seen, Reid
thinks, an ‘appearance’ or ‘sensation’ or what ‘Mr Locke calls … an idea’ (1764/1970: 100) of
colour is produced in the mind, and ‘[t]his sensation is followed by the perception of the
object’ (1764/1970: 214). The sensation, whose nature we can know perfectly, ‘suggests the
conception and belief of some unknown quality in the body, which occasions the idea; and it is to
this quality, and not to the idea, that we give the name of colour’. Here Reid takes himself to
have an essentially linguistic dispute with Locke and his followers, whom Reid takes to hold that
‘colour is not a quality of bodies, but only an idea in the mind’. Reid thinks it is absurd to hold
that the vulgar ‘give the name of colour to the sensation’. Instead, colour words denote the
power in bodies to produce our colour sensations, the ‘unknown cause’ of a ‘known
effect’ (1764/1970: 100, 102, 101).
Borrowing Reid's terminology, let us call this picture of perception the natural‐sign
theory. According to it, the only information solely about the colour of a tomato that can
be extracted from the experience of seeing it is that the tomato has some property or
other that causes experiences with the sensory quality of reddishness (the ‘natural sign’
of the sensible quality red).
After quoting Reid with approval (1980: 271), Evans claims that ‘[a]ll it can amount to for
something to be red is that it be such that, if looked at in the normal conditions, it will
appear red’ (1980: 272). Some misguided philosophers deny this:
Page 9 of 15
Sensory Qualities, Sensible Qualities, Sensational Qualities
They have tried to make sense of the idea of a property of redness which is both
an abiding property of the object, both perceived and unperceived, and yet
‘exactly as we experience redness to be’. By concentrating upon one's experience
of colour, one is supposed thereby to know what it is for an object to have this
property: ‘This’, one is to say, referring neither to the experience nor to any
primary property of the thing, ‘this, just as it is, can exist in the absence of any
observer’.
But the leap gets us nowhere, for it inevitably involves an attempt to make sense
of an exemplification of a property of experience in the absence of experience.
Wittgenstein once imagined a world in which there were places which affected
everyone painfully, so that pains (p. 277) were located at places in the way we
locate smells. Suppose this fantasy came true. Would it then make sense to give a
non‐dispositional account of what it is for there to be a pain at such and such a
spot: to suppose a ‘pain as we feel it’ existing in the absence of any observer?
What can the latter form of words mean save that something awful is going on
there, and how can that be, when there is no one who is hurt? To modify a dictum
of Wittgenstein, conceiving of a pain which no one feels on the model of a pain
which one does feel is none too easy a thing to do.
(1980: 272–3)
This passage is not easy to understand, but one interpretation (there may be others) leans
heavily on Evans's remark about ‘an exemplification of a property of experience’ and the analogy
with Wittgenstein's example. So interpreted, the argument takes the natural‐sign theory as a
premise, and runs as follows.
Start with pain, and assume that pains are ‘painful’ experiences. This view, unlike one
that takes pains to be the intentional objects of certain experiences, and not the
experiences themselves, makes pains ‘existing in the absence of any observer’ decidedly
suspect, just as Evans says.5 Further, assume the common view that painful experiences
are ‘mere’ sensations: experiences with no representational content (and so no
intentional objects). Now make the counterfactual supposition that Wittgenstein's fantasy
is true: ‘The surfaces of the things around us (stones, plants, etc.) have patches and
regions which produce pain in our skin when we touch them … In this case we should
speak of pain‐patches on the leaf of a particular plant just as at present we speak of red
patches’ (Wittgenstein 1953: sect. 312). That is, suppose that certain surfaces are
disposed to produce pain sensations (painful experiences) in us, and we use the word
‘pain’ to refer to a property of those surfaces that they possess even when no one is
touching them. ‘Pain’, as used in these counterfactual circumstances, can hardly refer to
a property that our experiences represent certain surfaces as having, because ex
hypothesi painful experiences have no representational content. Instead, the obvious
candidate, given that ‘pain’ is used to refer to an abiding property of surfaces, is that it
refers to the disposition to cause painful experiences in us.
Page 10 of 15
Sensory Qualities, Sensible Qualities, Sensational Qualities
The next step of the argument compares pain to colour. When we look at a tomato we
have reddish sensations, experiences with a distinctive sensory quality, but with no
representational content (no colour content, anyway). Trying to make sense of ‘colour‐as‐
we‐see it’ (Evans 1980: 273) when we don't see it amounts to making sense of
reddishness, a ‘property of experience’, exemplified ‘in the absence of experience’.
However, we use the word ‘red’ to refer not to a property of experiences but to an abiding
property of objects. As with the pain‐patches example, this word plausibly refers to a
disposition to cause reddish sensations in us (a disposition to ‘appear red’, as Evans
unfortunately puts it). Thus, the familiar ‘secondary‐quality’ theory of colours as
dispositions or powers is established.
Once the natural‐sign theory is in place, the secondary‐quality theory is not far behind.
However, the natural‐sign theory is mistaken.
(p. 278)
One difficulty is the assumption of special epistemic access to the ‘effect’, without which
the natural‐sign theory would not get off the ground. And that, although it might be true,
is unobvious and needs arguing for. (Here of course the spatial metaphor is pernicious,
encouraging us to think that sensations are particularly accessible because they are ‘in’
the mind.)
But the decisive defect is this. The natural‐sign theory cannot accommodate the plain fact
that objects look coloured. Imagine that someone is looking at a tomato and sees that it is
red. If we apply the basic outline of the natural‐sign theory to the perception of the
tomato's colour, what is going on? First, a property of the tomato causes a certain effect
in the perceiver: a reddish sensation. Second, she becomes aware that this sensation is
reddish. Third, she forms the belief that the tomato has the power to cause reddish
sensations (i.e. she believes that the tomato is red). End of story.
Of course, the tomato will look red to the perceiver. The natural‐sign theorist can say that
this simply amounts to the fact that the perceiver believes that the tomato is red and has
a reddish sensation. But now imagine that our perceiver does not in fact believe that the
tomato is red; she suspects, say, that the lighting conditions are abnormal. The tomato
will still look red to her. And this cannot be accommodated by the natural‐sign theory.
According to it, the perceiver in this situation is having a reddish sensation, is aware that
her sensation is reddish, and does not believe that the tomato is red. It is not very
plausible that the tomato's looking or appearing red can be reconstructed from such
meagre materials. Adding that the perceiver has some sort of tendency or disposition to
believe that the tomato is red does not help much. If appearances are analysed partly in
terms of dispositions to believe then there are no causal explanations of the latter in
terms of the former. But, intuitively, there are such explanations: a perceiver might be
disposed to believe that the tomato is red because it looks red.6
Page 11 of 15
Sensory Qualities, Sensible Qualities, Sensational Qualities
There are other routes to the secondary‐quality theory, but they are considerably less
direct. Contemporary proponents of one or another sophisticated variant of the
secondary‐quality theory reject the natural‐sign theory, at least in its full Reidian (p. 279)
dress, but the assumption that there are sensory (or sensational) qualities is often crucial
to their arguments (see, for instance, Peacocke 1983: ch. 2, 1984; McLaughlin 2003).
15.6 Summary
Historically there has been a persistent tendency to identify sensible qualities with
sensational qualities or sensory qualities: to take the striking quality of a tomato to be a
quality of one's experience, or a quality of a sense‐datum. Once the three kinds of quality
are separated, it is not so clear that there are any sensory or sensational qualities. And if
there aren't, then a lot of the motivation for secondary‐quality theories of sensible
qualities drains away. More threatens to be thrown out with the bathwater, since the
‘hard problem’ of consciousness arguably presupposes that there are sensory qualities.7
The issue is one of the major fault lines in contemporary philosophy of mind.
Introspection is an unlikely way of making progress. A more promising method is to
proceed on the assumption that sensory and sensational qualities are a myth, and see
how quickly—or whether—one runs into sand.
References
Bennett, J. (1988), Events and Their Names (Indianapolis, Ind.: Hackett).
Brown, D. H. (2006), ‘On the Dual Referent Approach to Colour Theory’, Philosophical
Quarterly, 56: 96–113.
Page 12 of 15
Sensory Qualities, Sensible Qualities, Sensational Qualities
Byrne, A. (2006), ‘Color and the Mind–Body Problem’, Dialectica, 60: 223–44.
Evans, G. (1980), ‘Things Without the Mind—A Commentary upon Chapter Two of
Strawson's Individuals, in Z. van Straaten (ed.), Philosophical Subjects (Oxford: Oxford
University Press), 76–117;
repr. in Evans, Collected Papers (Oxford: Oxford University Press, 1985), 249–90, (pp.
refs. in the text are to the reprint).
Hayek, F. A. (1952), The Sensory Order: An Inquiry into the Foundations of Theoretical
Psychology (Chicago, Ill.: University of Chicago Press).
Levine, J. (2001), Purple Haze: The Puzzle of Consciousness (Oxford: Oxford University
Press).
Page 13 of 15
Sensory Qualities, Sensible Qualities, Sensational Qualities
Reid, T. (1764/1970), An Inquiry into the Human Mind (Chicago, Ill.: University of Chicago
Press).
Rosenthal, D. (1999), ‘The Colors and Shapes of Visual Experiences’, in D. Fisette (ed.),
Consciousness and Intentionality: Models and Modalities of Attribution (Dordrecht:
Kluwer), 95–118.
Smart, J. J. C. (1959), ‘Sensations and Brain Processes’, Philosophical Review, 68: 141–56.
Tye, M. (2000), Consciousness, Color and Content (Cambridge, Mass.: MIT Press).
Notes:
(1) One notable argument for contingency is presented in Block (1990). It should be
emphasized that Clark's views on sensory qualities are generally quite different from
Levine's (see Clark 2000: 22, 35–8).
(2) For arguments in support of the ambiguity thesis see Rosenthal (1999) and Brown
(2006). Reid, who was an ambiguity theorist for ‘smell’ (1764/1970: 39), drew the line at
colour terms (1764/1970: 101–2); see also Section 15.4 below and Clark (2000: 1–2).
(3) Of course, traditionally sense‐data were taken to be either coloured items or else the
colours themselves (see Moore 1953: 28–40; Jackson 1977).
(4) For criticism see Tye (2000: ch. 4). It should be noted that Peacocke's ‘primed’
notation is not quite as simple as our simplified variant of it. ‘Blue′,’ as explained above,
expresses a property of a sense‐datum. However, Peacocke introduces primed
expressions as expressing relations between experiences and regions of an ‘imagined
interposed plane’ perpendicular to the line of sight (1983: 20). Later, though, Peacocke
speaks freely of ‘red regions of the visual field’ (1983: ch. 2; see also Peacocke 1984),
where a ‘region of the visual field’ appears to be very much like a sense‐datum.
(5) The truth may be a confused amalgam of both views: sometimes we think of pains as
intentional objects (‘the pain is in my wrist’) and sometimes as experiences (‘pain is a
feeling’) (see Harman 1990: 40).
Page 14 of 15
Sensory Qualities, Sensible Qualities, Sensational Qualities
(6) Since Evans basically makes this point elsewhere (1982: 123–4), the interpretation of
his argument in the text should be treated with some caution.
(7) For the hard problem see Chalmers (1996). For an argument that the hard problem
vanishes if there are no sensory qualities see Byrne (2006).
Alex Byrne
Page 15 of 15
The Explanatory Gap
This article outlines the strategy for reconciling materialism and the explanatory gap. As
one might imagine, there are a number of complex issues that must be sorted out
concerning the alleged special nature of phenomenal concepts, and their alleged
incommensurability with non-phenomenal concepts, before one can determine whether or
not the strategy works in the end. In particular, it is unclear whether materialists can
deliver an explanation of the special cognitive access afforded by phenomenal concepts
which is itself consistent with its precepts.
Keywords: materialism, explanatory gap, phenomenal concepts, non-phenomenal concepts, cognitive access
16.1
THE term ‘explanatory gap’ was first introduced in Levine (1983), in the context of a
discussion of Kripke's argument (1980) against the mind–body identity thesis. Since then
the discussion surrounding it has developed beyond the context of Kripke's argument.
Along with notions like ‘the hard problem’ (Chalmers 1996) and Nagel's worries about
what it's like to be a bat (1974), the idea that there exists an explanatory gap expresses
the deep puzzle concerning the place of conscious experience in a material world.
Materialism, or physicalism (I will use the two terms interchangeably), is the thesis that
the physical facts metaphysically determine, or constitute, all the facts. Whatever
happens in the world is ultimately a matter of how certain physical objects and properties
are distributed in space–time. It is of course a notoriously difficult matter to specify
precisely what this means, but, despite the growing literature on this controversy, most
Page 1 of 13
The Explanatory Gap
philosophers have a clear enough idea of what physicalism means that they take a stand
pro or con regarding it. For our purposes we'll leave the specification of physicalism fairly
rough, since nothing of substance regarding the nature of consciousness will turn on
making it precise.
With respect to conscious experience, then, physicalist doctrine comes down to this. If
you take a physical object like a brain, composed of cells standing in various physical
relations to each other and instantiating various electrochemical properties, then you
have all it takes for there to be conscious experience. Conscious experience just is
nothing over and above these cells interacting in their various physical (p. 282) ways.
Sure, our descriptions of conscious experience employ a different, conceptually
independent vocabulary, but what they describe are the very same phenomena as the
descriptions employed by neuroscientists.
Before we get to what's puzzling about this it's worth asking why we should believe it in
the first place. Why has physicalism become the dominant (though by no means
universal) position in the philosophy of mind? Despite its large‐scale acceptance, probably
the reasons offered by many philosophers who adhere to the doctrine would differ. Still,
one consideration seems to turn up consistently: the problem of causal interaction. We
know that purely physical, non‐mental events can cause conscious experiences and that
conscious experiences can cause purely physical, non‐mental events. Light bouncing off
objects in my vicinity and hitting my retina causes me to have certain visual experiences,
and my having those visual experiences causally contributes to my arm's extending and
grasping an object in my vicinity. We know that the relevant brain processes intervening
between stimulus and response in this case are causally responsible for the reaction, so if
conscious experience is also playing a role, it had better be somehow constituted by at
least some of these brain processes. Of course, the argument needs refining, but that's
the main idea.
So what's puzzling about this? Well, simply put, it's difficult to see how we can actually
explain the features of our conscious experience by appealing to the physical properties
of the brain processes that allegedly constitute it. This of course requires some
elaboration. So consider a typical conscious sensory experience, such as seeing a red rose
in daylight. There is something it is like to have such a visual experience, one component
of which is the particular way the rose's colour appears. This is a feature of experience
with which one is immediately acquainted. By ‘immediately acquainted’ here I don't
intend anything more than that it is not something we have to conduct any sort of
investigation or inference to determine. We just are aware of it in seeing the rose. Every
sighted person knows what I'm talking about here.
So now let's consider for the moment just this one feature of the experience, the precise
colour the rose appears to have. Of course, we seem to be aware of a feature of the rose,
and one can ask what is responsible for this feature, what explains its colour. But right
now we want to focus not on features of the rose, but on features of the experience.
Seeing the rose as having that precise shade of red, with that precise brightness and
Page 2 of 13
The Explanatory Gap
Page 3 of 13
The Explanatory Gap
16.2
While the foregoing gives the basic idea, there are various ways to elaborate the notion of
the explanatory gap, and connect it with other aspects of the mind–body problem. One
particularly important connection is that between the explanatory gap and the traditional
conceivability argument against physicalism. The conceivability argument—best known
from Descartes's discussion, but revived recently by Kripke (1980) and Chalmers (1996)—
goes like this. It certainly seems possible—or, as we say, it's conceivable—that there be a
brain functioning like ours, and yet there be no conscious experience associated with it.
Put another way, there might be nothing it's like to be that brain.
Of course, one has to understand what's meant by ‘possible’ and ‘conceivable’ here in a
particular way, otherwise this claim will not seem so plausible. Admittedly, given what we
know of the role played by the brain in our mental life, it's not like we think in any
practical sense there really is any possibility that there could be a brain that is in a state
just like mine right now but which has no conscious experience going on in it. However,
the sense of ‘conceivability’ at issue here doesn't demand that level of likelihood. Rather,
by saying that it's conceivable that there be a brain in a state just like mine but there's
nothing it's like to be that brain (or to have one's brain in that state) all we intend is that
there is no logical contradiction or conceptual incoherence involved in the supposition.
That this standard is met can be seen from the fact that it took an empirical discovery to
determine that we think and experience with our brains, and not our hearts, say, or even
our immaterial souls.
Suppose, at least for the moment, we grant this conceivability premise. The next step in
the argument is the claim that identity entails a kind of logical necessity. The idea is this.
If we say one thing is identical to another, what we've asserted is not really (despite the
term ‘another’) that two things stand in an identity relation—after all, what would that
even mean?—but rather that there is just one thing there. Now take any object and ask:
Could it possibly have been something other than it is? Not, mind you, ‘Could it have had
different properties than it has?’ but, ‘Could it have been a different object than the one it
in fact is?’. Well, no, it is what it is, and the idea that it, that very thing, could have been
something else, doesn't really make sense. For then it wouldn't have been it, after all,
would it?
Now, let's take the basic claim of physicalism, that the mind is the brain, or that certain
mental processes are identical to certain physical processes in the brain. This certainly
sounds like an identity claim. If it is, then it expresses a logical necessity, in that it doesn't
make sense—as per our discussion in the paragraph above—to (p. 284) claim that it's
possible that the mind isn't the brain, or that this conscious experience isn't a brain
process; that, again, would be like the possibility that something isn't itself, which is
incoherent. (To be sure, we may have discovered the identity through empirical means,
but that doesn't change the logical import of the identity claim once accepted.) But if we
accept the initial conceivability premise, it seems that we are committed to just this
Page 4 of 13
The Explanatory Gap
possibility. The only way to maintain the conceivability premise and the logical necessity
of identity, it seems, is to deny physicalism; deny that in fact the mind and brain, or
conscious experience and certain brain processes, really are identical.
Now physicalists have a standard reply to this argument (Hill and McLaughlin 1999;
Levine 2001; Melnyk 2001). They point out that the conceivability premise doesn't
establish that it is really possible for there to be brains without conscious experience, just
that it's conceivable. It seems possible because we can't determine a priori that conscious
experiences and brain processes are identical, and so we can imagine a world in which
they aren't. But, in fact, if they are identical, then a world in which there's a brain like
mine that isn't conscious is impossible. While this reply to the conceivability argument
has not gone unchallenged (Kripke 1980; Jackson 1993; Chalmers 1996), it has gained
wide acceptance. The problem for physicalists is that even if we accept this reply, one can
still use the conceivability premise to push the explanatory gap.
The problem is this. Why, after all, is it conceivable that there could be a brain in a state
just like mine but unconscious, or, more precisely, such that there is nothing it's like to be
this brain? Contrast this situation with the following. Suppose I know a brain is in a state
just like mine and the laws of physics and chemistry are working as usual. Is it
conceivable that this brain shouldn't cause the normal behavioural effects that my brain
in fact causes? Well, if these very neurons are firing, and the laws of physics and
chemistry are obeyed, then these other neurons will fire, which will cause these other
neurons downstream to fire, which will eventually cause the relevant muscles in my limbs
to contract, and my body will move as expected. Without an interruption in natural law, I
can't even conceive how this effect could be stopped. However, I still can conceive of all
this going on without there being anything it's like for the creature whose brain it is.
What's the difference?
One might suggest the following. When it comes to the connection between the brain's
state and its behavioural, or motoric, effects, we understand how the effect is brought
about. Once we can explain something, it is no longer conceivable that it shouldn't be that
way. After all, explaining why something happens, or is the way it is, by appeal to certain
facts is a matter of showing why it had to be that way and not some other (relevant) way,
given those facts. So now consider again the case of the brain without consciousness.
According to the argument, the reason that this situation is conceivable is that we don't
understand how the conscious experience is a result of, or constituted by, the relevant
brain process. If we knew why these very neurons firing the way they do amounted to
conscious experience, then we wouldn't still find it conceivable that a brain could occupy
this state and yet lack the conscious experience. The fact that this situation is still
conceivable is a symptom of our lack of the requisite explanation; it's a symptom of the
explanatory gap.
Page 5 of 13
The Explanatory Gap
But now we face another problem, one that also can be seen as a symptom of the
explanatory gap. Suppose we arrange a sequence of creatures starting with ourselves and
ordering them in terms of the degree of difference from human neurophysiology and
functional organization. Functionalists are already committed to the idea that creatures
physically different from us can be mentally, even consciously, like us nevertheless. But
not only are physical differences allowed, it's clear that we can't demand complete and
exact functional identity either. After all, there are many levels of functional organization
underlying human psychological capacities, each implementing the level above it, and
bottoming out with the neural mechanisms that implement the entire hierarchy. There is
also an enormous range of psychological capacities that interact with each other, some of
which can differ from creature to creature while others stay the same. So assuming we
have some sequence of creatures with increasing difference from normal humans, at
which point, even roughly, do we say that the conscious experiences are no longer similar
to our own? What can give us even a clue how to answer this question?
To make the problem a little more concrete, suppose we create an intelligent robot, like
Data on Star Trek, and also meet an intelligent alien. We now ask: When they look at a
red rose, are they seeing it the same way we are? Is their visual experience qualitatively
like ours, or each other's? Suppose we know all there is to know about how they process
colour information, and know that there are some differences from the way we do it and
some similarities. What determines an answer to our question about what it's like for
them? If we had an explanation of what it's like for us in terms of our functional or
physical organization, then it seems we'd have a principle for determining what it's like
for them. We'd look for those very features in them that explain the qualitative character
in us. But it seems pretty clear that we don't have a clue how to answer this question.
Why? The answer seems to be that since we don't really know why our particular
functional and physical organization should be associated with just this type of
Page 6 of 13
The Explanatory Gap
experience, we don't know which features to look (p. 286) for to determine what it's like
for the aliens and robots. Our inability to project our experience on to other types of
creatures is again a symptom of the explanatory gap.
16.4
The brain–mind connection, then, seems mysterious and inexplicable. By appeal to the
physical properties and processes of the brain we seem able to explain how we process
information, how our bodies react to the environment, and even—on analogy with
computers—how we reason. But why any of these processes should give rise to
consciousness, that there should be something it's like for creatures who have these
processes going on inside them, this seems mysterious to us. There are two standard
ways for a materialist to respond to this challenge. On the one hand, some materialists
claim to have theories that actually bridge the gap. On the other hand, other materialists
admit that the gap is unbridgeable, but argue that there is a good reason the gap is so
persistent, and once we understand why we will see that it poses no threat to
materialism.
Straight functionalism itself breaks down into two types, analytic functionalism and
psycho‐functionalism, which differ in how they determine the relevant description of the
system. Analytic functionalism (Lewis 1980) says that associated with each mental
concept—such as ‘pain’ and ‘visual experience of red’—is a set of platitudes concerning
the kinds of stimuli that cause it, the kind of behaviour it causes, and the other mental
states with which it interacts. So, on this view, to be in pain is just to be in whatever
physical state it is that normally results from damage to the body, causes avoidance
behaviour, and causes difficulty in concentrating and a strong desire to stop it. While this
is oversimplified, it gives the general idea. Psycho‐functionalism is the view that the
proper description of pain's (or any other mental state's) functional role will result from
scientific investigation into its nature. Just as we discovered empirically that water is
H2O, so too we will discover that pain is such‐and‐such a functional role.
Neither of these positions really bridges the explanatory gap. Analytic functionalism must
explain why, if our concept of a conscious experience is captured by the relevant
description of a functional role, it is still conceivable that some physical systems satisfy
that role and yet fail to be conscious. The problem for psycho‐functionalism is that it's
Page 7 of 13
The Explanatory Gap
hard to see how scientific investigation alone, without some conceptual breakthrough,
could support the claim that conscious experiences were (p. 287) identical to specific
functional states. The analogy—or rather, disanalogy—with the case of water is telling.
Why, after all, do we take water to be identical to H2O? What have we discovered that
supports this identification? Well, we have taken samples of water, broken them down
chemically, and discovered H2O molecules. What's more, the chemical properties of H2O
explain the superficial properties of water by which we normally identify it, such as its
liquidity at room temperature, its transparency, and the like. If it were not for this
explanatory structure, it would be unclear why we would accept the claim that water is
H2O.
Now contrast the situation with pain, say, and whatever detailed causal/functional role we
find it plays. The psycho‐functionalist wants to claim that pain should be identified with
this causal role, but on what basis? After all, we normally identify pain in our own
experience by how it feels. Can we explain this feeling by appeal to the causal role? If we
could, then it would no longer be conceivable that a creature could occupy a state with
this causal role and yet not really feel pain, but this does seem to be conceivable. In fact,
the order of explanation seems quite the reverse. Pain seems to be the kind of state we
want to avoid because of the way it feels! It's very unclear, then, what could be
discovered from a third‐person perspective that would justify the claim that that causal
role is what pain really is, not just a role pain in fact plays. In other words, we can't
bridge the explanatory gap by appeal to a functionalist identity thesis. On the contrary,
we won't be in a position to accept any functionalist identity thesis until we bridge the
gap.
There are other functionalist views that provide a more specific account of what
constitutes a conscious state than ‘straight functionalism’. Two in particular are the
higher‐order‐mental‐state view (Lycan 1996; Rosenthal 1997) and representationalism
(Dretske 1995; Tye 1995; Byrne 2001). On the higher‐order‐mental‐state view, a state is
conscious when one is conscious of it, the point being that consciousness is really a
relation between a subject and one of her mental states, not an intrinsic property of a
single state. The way a subject becomes conscious of one of her states is that she
occupies yet another mental state that has the first state as its content—that ‘says’, as it
were, that she is in the first‐order state. According to representationalism, the precise
character of a conscious experience is determined by which properties of the
environment presented to the senses the state is representing. To experience redly, as it
were, is just to represent the property of being red in one's visual system.
As one can see from the descriptions just presented, the two views are not really
competitors, since each attempts to address a different aspect of conscious experience.
The higher‐order view provides an analysis of what it is for a mental state to be
conscious, as opposed to unconscious. The conscious mental states, on this view, are
those that are monitored, or detected by other mental states. But this theory doesn't say
why a particular conscious experience is like this—say, the way my current visual
Page 8 of 13
The Explanatory Gap
sensation of a ripe tomato is for me—as opposed to some other way. The particular
experiential content is not explained by this theory; it just attempts to say what it is for
there to be any experiential content at all. Representationalism, on the other hand, while
not addressing the general question of what makes a state conscious at (p. 288) all, does
try to say why a conscious state is like what it's like. It's like this rather than that because
this is what it represents, not that.
So together one might see the higher‐order view and representationalism as constituting
a comprehensive theory of conscious experience (see Lycan 1996 for just such a
combination). Does this combination serve to bridge the explanatory gap? It's hard to say
until one knows what the account of representation is. Notice that both views rely
crucially on the idea that a mental state is representing something: the higher‐order view
has one mental state representing another, and representationalism has the first‐order
state representing sensible properties of external objects.
It's possible to build consciousness into the representation relation in such a way that to
really represent is to be conscious of what one is representing, but then the account is
circular. We still need to know how a purely physical system can represent in this way. On
the other hand, if one's theory of representation follows recent causal/informational
accounts (Fodor 1990; Dretske 1995), then, while there isn't any mystery how a purely
physical device could satisfy such an account, it isn't clear how this bridges the
explanatory gap. The very same considerations described above with respect to straight
functionalism attend this more refined theory as well. This persistence of the intuition of
an arbitrary or brute connection between the states posited by the various theories of
consciousness and one's inner, first‐person conception of conscious experience lends
considerable support to the second strategy. That is, rather than attempt to bridge the
gap, materialists who follow the second strategy try to show that there is something
special about our first‐person access to our experience that makes the existence of an
explanatory gap inevitable. Whether or not this feature of first‐person access is itself
explicable in materialist terms is the challenge.
16.5
A number of philosophers have attempted to reconcile the existence of the explanatory
gap with materialism (Loar 1997; Hill and McLaughlin 1999; Tye 2000; for critiques see
Levine 2001, 2007; Chalmers 2007). Though the details of their various proposals differ in
important ways, we'll sketch the response in a broad manner here. The basic idea is to
locate the problem in the special nature of one's first‐person access to one's own
conscious experiences. That is, there is a way of thinking of one's conscious experience
that is only available to the subject of the experience, and which is sufficiently different in
crucial respects from third‐person, theoretical ways of thinking of it that one can't
Page 9 of 13
The Explanatory Gap
integrate the two in the way necessary for an understanding of how what the latter
describes could be the same phenomenon as what the former describes.
Let's revisit the contrast between the theoretical identification of water with H2O and
pain with some neurological state. (The same considerations apply to functional identity
theses.) As we noted above, the identification of water with H2O is justified largely by the
explanatory work it does. There are these properties of water, such as its (p. 289) freezing
and boiling temperatures, which can be explained by appeal to its chemical structure
(and general laws of physics and chemistry). Similarly, if our only concept of pain had to
do with its causes, behavioural manifestations, and interaction with other internal states
we would have no problem identifying it with the relevant neurological state. We could
say why certain stimuli led to this state, which in turn led to the typical behavioural
responses to pain.
Notice that if we ourselves had never experienced pain, and our only cognitive access to
it was through our observations of others, the explanatory account just outlined would
appear quite adequate; there would be no gap. Our problem is that we know what it's like
to have pain, we know what it feels like, from the ‘inside’. What's more, it seems to be
that what we find from this internal perspective is precisely the aspect of pain that
appears to elude the explanatory reach of neuroscience. If we have powerful reasons to
think that materialism about pain (and all other conscious experiences) must be true, and
it is what we find in the first‐person perspective that seems to be creating the obstacle to
materialism's acceptance, then it's reasonable to see if the blame can be placed not on
materialism itself, but on the peculiar nature of the first‐person perspective.
The reconciliation of materialism with the explanatory gap then proceeds as follows.
Except for the concepts of conscious experiences possessed by the subject herself, all of a
cognitive subject's concepts are of the non‐phenomenal sort. So when she forms
functional and physical concepts to explain human behaviour and information processing,
these are non‐phenomenal concepts. The explanatory gap exists precisely because there
is a kind of ‘architectural’ gap within the human psyche between phenomenal and non‐
phenomenal concepts. To explain conscious experiences in the way that would bridge the
explanatory gap would require taking as explanatory premises—the explanans—
propositions framed with non‐phenomenal concepts and joining them with an explanatory
conclusion—the explanandum—framed with phenomenal concepts. However, our minds
Page 10 of 13
The Explanatory Gap
are just not capable of integrating these two radically different kinds of concept in a way
that would result in a satisfying explanation. Hence, it always seems intelligible to us that
what is described by the phenomenal concepts should turn out to be distinct from what is
described by the non‐phenomenal concepts. This fact doesn't show that they aren't
identical, and that, as far as the metaphysics goes, neurological features don't explain
conscious features. Rather, it's just a sad fact about us that we aren't (p. 290) capable of
appreciating this connection between neurological states and conscious experiences. We
are, as it were, just wired the wrong way to bridge the gap.
The strategy for reconciling materialism and the explanatory gap just outlined represents
a powerful weapon in the arsenal of materialism. As one might imagine, there are a
number of complex issues that must be sorted out concerning the alleged special nature
of phenomenal concepts, and their alleged incommensurability with non‐phenomenal
concepts, before one can determine whether or not the strategy works in the end. In
particular, it is unclear whether materialists can deliver an explanation of the special
cognitive access afforded by phenomenal concepts which is itself consistent with its
precepts. Some have argued that a special mental relation like ‘acquaintance’ (Russell
1912) that is itself inexplicable in materialist terms would have to be posited, and thus
the materialist may have just substituted one gap for another (Chalmers 2007; Levine
2007). Whether or not this is so is the topic of much current research (Kriegel 2006; Alter
and Walter 2007).
References
Alter, T., and Walter, S. (2007) (eds.), Phenomenal Concepts and Phenomenal Knowledge:
New Essays on Consciousness and Physicalism (Oxford: Oxford University Press).
—— (2007), ‘Phenomenal Concepts and the Explanatory Gap’, in T. Alter and S. Walter
(eds.), Phenomenal Concepts and Phenomenal Knowledge: New Essays on Consciousness
and Physicalism (Oxford: Oxford University Press), 167–94.
Fodor, J. A. (1990), A Theory of Content and Other Essays (Cambridge, Mass.: MIT Press).
Hill, C. S., and McLaughlin, B. (1999), ‘There are Fewer Things in Reality than are
Dreamt of in Chalmers' Philosophy’, Philosophy and Phenomenological Research, 59: 445–
54.
Page 11 of 13
The Explanatory Gap
Kripke, S. (1980), Naming and Necessity (Cambridge, Mass.: Harvard University Press).
Levine, J. (1983), ‘Materialism and Qualia: The Explanatory Gap’, Pacific Philosophical
Quarterly, 64: 354–61.
—— (2001), Purple Haze: The Puzzle of Consciousness (Oxford: Oxford University Press).
Lewis, D. (1980), ‘Mad Pain and Martian Pain’, in N. Block (ed.), Readings in Philosophy
of Psychology, i (Cambridge, Mass.: Harvard University Press), 216–22.
(p. 291) Nagel, T. (1974), What is it Like to Be a Bat? Philosophical Review, 83; 435–50.
Putnam, H. (1991), ‘The Nature of Mental States’, in D. M. Rosenthal (ed.), The Nature of
Mind (Oxford: Oxford University Press), 197–203.
Joseph Levine
Page 12 of 13
Phenomenal Concepts
This article is about the special, subjective concepts we apply to experience. These are
called ‘phenomenal concepts’ (PCs) and they are of special interest in a number of ways.
First, they refer to phenomenal experiences, and the qualitative character of those
experiences whose metaphysical status is hotly debated. Conscious experiences strike
many philosophers as philosophically problematic and difficult to accommodate within a
physicalistic metaphysics. Second, PCs are widely thought to be special and unique
among concepts. The sense that there is something special about PCs is very closely tied
up with features of the epistemic access they afford to qualia.
Keywords: phenomenal concepts, phenomenal experiences, qualitative character, metaphysical status, qualia,
epistemic access
Page 1 of 26
Phenomenal Concepts
17.1 Introduction
THIS chapter is about the special, subjective concepts we apply to experience. These are
called ‘phenomenal concepts’ (PCs) and they are of special interest in a number of ways.
First, they refer to phenomenal experiences, and the qualitative character1 of those
experiences whose metaphysical status is hotly debated. Conscious experiences strike
many philosophers as philosophically problematic and difficult to accommodate within a
physicalistic metaphysics. Second, PCs are widely thought to be special and unique
among concepts.2
The sense that there is something special about PCs is very closely tied up with features
of the epistemic access they afford to qualia. When we deploy phenomenal concepts
introspectively for some phenomenally conscious experience as it occurs, we are said to
be acquainted 3 with our own conscious experiences. While philosophers have understood
‘acquaintance’ in various ways, it is generally taken to be a (p. 293) unique
epistemological relation that relates a person to her own mental states directly,
incorrigibly, and, according to some, in a way that reveals the essence of these mental
states. Such a relation has struck many philosophers as deeply puzzling. Accounts of PCs
either have to explain the acquaintance relation, or acquaintance with our phenomenal
experiences has to be denied. The way different accounts of PCs handle these issues will
be the main topic of this chapter.
PCs have received much attention in recent philosophy of mind mainly because they
figure in arguments for dualism and in physicalist responses to these arguments.4 In
Section 17.4 I will briefly explain how features of our epistemic relation to phenomenal
consciousness provide the ground for dualist arguments. In Sections 17.5 and 17.6 I will
discuss some recent accounts of PCs and their role in arguments over physicalism/
dualism. But first, in Section 17.2 I will clarify some background assumptions that it is
important to put on the table, and in Section 17.3 I will elaborate on the epistemic and
semantic constraints on a satisfactory account of PCs.
Throughout the chapter I will assume realism about qualia. I use ‘qualia’ in a minimalist
sense; that is, the sense of there being something it is like to undergo an experience,
something one can normally introspect; for example, the feeling of my fingers flexing that
(partly) characterizes my present bodily sensation. All accounts of phenomenal
experience agree that there are qualia in this sense, though they might disagree on what
Page 2 of 26
Phenomenal Concepts
exactly it means to introspect qualia.5 Philosophers who think for (p. 294) whatever
reason that there are no qualia might still hold that there are concepts that work much
like PCs except that they fail to refer at all.6
Some accounts of PCs explain the unique epistemic role of PCs by the way they relate to
their referent in a uniquely intimate way. The idea is that in their basic uses we apply PCs
to our experiences directly as they occur, merely on the basis of having the experience.
One can object to this by observing that the very same PCs can also be applied in the
absence of the phenomenal experiences they refer to, for example when we apply them to
other people or our own past experiences, and in these cases it is not plausible to say that
the application of the concept is direct or that there is a special, unique relationship
between the concept and its referent. Accordingly, some philosophers locate the
uniqueness of PCs instead in the uniqueness of the properties they refer to, in the
luminosity and transparence7 of qualia themselves. Those who maintain that there is
something unique about the relationship between PCs and their referents emphasize the
difference between basic applications of PCs (closely tied to first‐person experience itself)
and other applications (not so tied), and they count those latter applications as derivative
of, and in some important respect different from, the basic ones.8
Here is a way to explicate the difference between basic and derivative applications of
PCs. We employ PCs in their basic applications when attending to and thinking about
conscious experiences and their phenomenal qualities from the first‐person perspective
(i.e. subjectively). ‘First‐person perspective’ here means that in the basic applications of
PCs their reference is (or is exemplified) in the mind of the thinker, and this fact is crucial
from the point of view of the application of the concept. In a basic application of a PC a
person is aware, for example, that her finger tingles in a particular way and thinks to
herself Here it goes again.9 This thought (though not its public‐language expression)
involves a PC and the PC itself is intimately connected to the very itchy kind of
experience that it is referring to.10 In the case of (p. 295) basic applications of PCs the
token concept and its reference are both occurrences in the person's mind. This is
different from other kinds of concepts (and also from the non‐basic applications of PCs).
For example, one can think the descriptive concept ‘the present king of France’ without—
absurdly—there being any king in one's mind or, for that matter, anywhere at all. But in a
basic application of a PC the reference is in the thinker's mind.
In non‐basic applications of PCs we are thinking from the first‐person perspective (i.e.
subjectively) only in a derivative sense, without there necessarily occurring exemplars of
that qualia. These applications are subjective only via their connections to basic
applications of PCs. Non‐basic applications of PCs can refer both to a person's own
experiences and to those of other people and creatures.11 For example, when Mary thinks
Page 3 of 26
Phenomenal Concepts
that she will later experience the taste of lemon or that Sam is currently experiencing this
taste, she doesn't literally experience a lemony taste herself.
What is the relation between basic and non‐basic applications of PCs? One suggestion is
that a non‐basic application of a PC, as for example in Mary's thought I will soon
experience the taste of lemon, requires some previous basic application of the concept to
an instance of the type of phenomenal experience it refers to (i.e. the lemony taste); or at
least it requires a previous instantiation of the experience itself. There are two ways to
understand this requirement. One is as a claim about how we acquire PCs. The other is as
a claim about necessary conditions for possessing PCs. The former seems correct as a
(contingent) matter of fact, but the latter may be false. Here is an argument that it is
false due to Dennett (2007).
Suppose that Mary possesses the concept lemony taste, having previously tasted a lemon,
but she is not now experiencing a lemony taste. Dennett imagines ‘Swamp Mary’, who
comes into existence and who duplicates Mary's intrinsic physical (and, if there are such,
non‐physical) properties. Swamp Mary doesn't share all of Mary's concepts but it seems
very plausible that she shares Mary's phenomenal concept lemony taste. She can ‘recall’
a lemony taste, anticipate it, compare it with other tastes, and so on. Dennett says that
she has the concept. But if so, the occurrences of previous instances of lemony taste 12
are not constitutive of possession of the concept lemony taste, and non‐basic applications
of lemony taste are possible for a person even before she ever had a basic application of
the concept.
What makes ‘basic’ applications basic then, as opposed to the ‘non‐basic’ ones? It is clear
that to possess a PC lemony taste it is necessary to be capable of recognizing an instance
of lemony taste as falling under the concept; that is, to be able to produce a basic
application of lemony taste. But this is too weak; the capacity in question is a trivial
capacity of any normal human. There could be a person, let's call him ‘Joe’, who has
never experienced a lemony taste, and so can't recall, compare, anticipate, etc. the taste
—in short, doesn't have the concept lemony taste at all; nevertheless, he would also have
this capacity. But there is a capacity Swamp Mary has that (p. 296) Joe doesn't, which
arguably is constitutively necessary for possession of the concept lemony taste; namely,
the capacity of recognizing lemony taste when she encounters it as the same taste she
attributed to others on occasions when she didn't have the experience.13 This is why basic
applications are basic; a concept can be a non‐basic application of a PC only if certain
non‐trivial counterfactuals involving basic applications hold. This condition also explains
the subjectivity of PCs.
Most (perhaps all) phenomenal experiences are representational. For example, a token of
a visual quale blue represents a blue expanse or the sky or perhaps both. If qualia
represent then it is plausible that they represent non‐conceptually. That is, they do not
Page 4 of 26
Phenomenal Concepts
have language‐like structure but rather are akin to pictures and represent in something
along the lines of the way pictures, images, graphs, and other so‐called analogue
representations represent. Exactly how analogue representations work is controversial.14
According to one view of qualia, qualia are not intrinsically representational. A way of
thinking about this is to say that they are aspects of sensation and can occur in the
absence of representational content.15
Some qualia‐realists, for example Loar (2003), hold that phenomenal states and their
qualia are themselves representational, in the sense that there couldn't be non‐
representational qualia. According to his view, however, there is more to qualia than
representational content; in other words, representational content doesn't exhaust all
there is to it.
Page 5 of 26
Phenomenal Concepts
In this framework PCs, being concepts, are particular kinds of words in Mentalese. This
framework will allow us to raise the question of what (if anything) is special about PCs by
discussing what is special about their vehicle, reference, mode of presentation, and
conceptual role.
Page 6 of 26
Phenomenal Concepts
It is important to distinguish PCs from what have been called ‘psychological concepts’.20
Psychological concepts characterize mental states functionally in terms (p. 298) of causal
relations with stimuli, other states, and behaviour and are distinct from PCs.21 For
example, the psychological concept itch* may be (simplistically) characterized as the state
that is caused by tissue irritation and causes scratching.22 In contrast, the PC itch picks
out a certain sensation (itch) directly, without the mediation of a functional or
behavioural mode of presentation. A psychological concept is a third‐person concept in
the sense that the mental state that it refers to does not play any direct role in the mental
machinery associated with the concept nor does the concept contain any reference to the
subject.
Fifty years ago or so most philosophers of mind would have declared that there cannot be
PCs. I have in mind the verificationist and behaviourist views that dominated philosophy
of mind in the mid‐twentieth century and still linger in some places. Wittgenstein (1953),
in his famous private language argument,23 argues that for a term (concept) to have
meaning (or reference) it must be possible to intersubjectively check whether an
application of that term is correct.24 The third‐person psychological concept of itch just
mentioned is like this, since in principle anybody can check whether or not a person is in
a state that satisfies the characteristic causal role. But PCs, if they exist, are not like that.
This led him to the view that first‐person attributions of sensations don't possess truth‐
conditions but rather are used to express pain in something of the way an exclamation
‘ouch’ does. However, few philosophers now would accept this radical conclusion.
Some hold (e.g. Chalmers 1996 and Ch. 18 below) that psychological and phenomenal
concepts are so different (and different in such a way) that they cannot refer to the same
property. Other philosophers, of a physicalist persuasion, think they might, even though
they present their referent in very different ways.25
Phenomenal Concepts
Here is a list of semantic/epistemic features that PCs have been held to have in relation to
phenomenal consciousness. An obvious constraint on the adequacy of an account of PCs
is that it either has to explain them or it has to explain them away; that is, show why
claims about these features seem plausible even though nothing real corresponds to
them.
(a) Acquaintance. We know our conscious states not by inference but by immediate
acquaintance which gives us direct, unmediated, substantial insight into their nature.
(b) Asymmetric epistemology. We are directly aware of our own conscious states in
ways no one else can be. As we have observed above, one can be aware of one's
Page 7 of 26
Phenomenal Concepts
(h) The conceivability of zombies. A scenario in which zombies exist cannot be ruled
out on a priori grounds. Zombies are creatures that are physically exactly like human
beings—they move like us, apparently speak and behave intelligently—but they
completely lack phenomenal experience. There is ‘nothing it's like’ to be them.
Zombies are conceivable, since no amount of information couched in physical and
causal terms is a priori logically sufficient for the application of a basic PC. Whatever
a person may learn about the causal role or neurophysiological nature of what is
going on in her (or anyone else's) brain (or any physically characterized facts) is
obviously not a priori sufficient for her to judge that she is experiencing a particular
qualia. In contrast, some philosophers claim, all other truths, for example the truth
that there is water in the Danube, are a priori derivable from the full physical truth
(Chalmers 1996).
Page 8 of 26
Phenomenal Concepts
(i) The explanatory gap. A closely related issue is what Joe Levine ( 2001 : 76–80)
calls the ‘explanatory gap’ between physical and phenomenal descriptions, i.e.
between a physical description of a person who is having certain experiences,
however detailed and informative, and a phenomenal description of those same
experiences. No current accounts bridge the gap, and the gap appears to be in
principle unbridgeable. The problem doesn't seem to be that we don't know enough
of the functioning of neurons and their interconnections. We certainly will learn a lot
more about all that in the future. One way to explicate the difference between the
usual perspicuous reductive explanations we encounter in science and the putative
cognitive or neurophysiological reductive explanations of phenomenal consciousness
is that whereas the hypothesis that phenomenal consciousness is non‐physical will
always seem comprehensible, the hypothesis that, for example, water is not H2O but
some non‐physical property just doesn't seem intelligible.
Many find (a)–(i) deeply mysterious. Some of these features present a challenge to any theorist
of phenomenal concepts. However, they, and especially semantic stability, the conceivability of
zombies, and the explanatory gap, are particularly worrisome if one is a physicalist, and so
wants to show that both phenomenal experiences and our concepts of them are physical in
nature.
(p. 301)
The other key premise of the argument is that if zombies are conceivable, they are
possible. If, as follows, zombies are possible, then there are some truths that are not
metaphysically necessitated by the complete physical truth about the world, and
therefore physicalism is false.
Page 9 of 26
Phenomenal Concepts
The explanatory gap. The gap argument starts from the premise that there is no
perspicuous explanatory relation between a physical description of a person undergoing
some experiences and a phenomenal description of those same experiences. The problem
is related to the conceivability of zombies, but it can be stated without appealing to the
notion of conceivability, or any thesis linking conceivability and possibility, and so has the
advantage that it doesn't rely on any substantial assumptions about concepts and
conceptual truths. It only relies on a contrast between the comprehensibility of the
hypothesis that phenomenal consciousness is non‐physical and the incomprehensibility of
corresponding hypotheses involving properties figuring in the special sciences; for
example, heat, life, digestion, etc.
The key premise of the gap argument is that if physicalism is true there can be no
explanatory gap between true descriptions of a phenomenon and some physical
description of the same phenomenon. But, the argument goes, since there is an
explanatory gap between phenomenal descriptions and any neurophysiological
description, physicalism is false.
There is a physicalist reply to these arguments that is based on the idea that a zombie
can meaningfully mimic these arguments and arrive at a false conclusion (Balog 1999).
But this reply leaves the puzzling epistemic features the dualist arguments rely on
unexplained. A stronger reply to the dualist arguments would be to show that these
features can be explained physicalistically. Since (a)–(i) are all epistemic/semantic in
nature, and our epistemic relation to phenomenal consciousness is mediated by PCs, it is
plausible that (a)–(i) will be explicable by appeal to the nature of PCs. The insight that is
at the core of what came to be called the ‘Phenomenal Concept Strategy’29 is that to
account for the key epistemic/semantic features of PCs we do not need to invoke the
(p. 302) nature of phenomenal consciousness itself; it is enough to invoke the special
nature of PCs. To what extent this is possible is the key issue both from the point of view
of theories of PCs and from the point of view of the ontology of mind. This is what we are
going to discuss in the next two sections.
But before that I would like to mention another type of physicalist response to the dualist
arguments: analytic functionalism or analytic representationalism (see e.g. Lewis 1966;
Jackson 2003). Analytic functionalism or representationalism is a doctrine about the
meaning of phenomenal terms; it is the doctrine that such meanings can be analysed in
functional or representational terms. Pain, for example, according to analytic
functionalism, has a conceptual role that connects it (in the meaning‐constituting way)
with complex concepts like typically caused by injury, typically causes avoidance
behaviour, typically causes saying ‘ouch’, etc. Analytic functionalism or
representationalism rebuts the conceivability arguments by denying the premise that
zombies are conceivable.30 Anything that has an internal state that plays the appropriate
causal roles/has such‐and‐such a representational profile (and zombies do have such
states) is, by definition, in pain. Analytic functionalism/representationalism, of course, has
to explain why zombies seem conceivable even though they are not.
Page 10 of 26
Phenomenal Concepts
However, this view, just like the analytic behaviourism that is sometimes attributed to
Wittgenstein, runs into problems with the first‐person applications of PCs. In our first‐
person—and usually considered basic—applications of phenomenal concepts functional
role/representational profile is not even in play. We can apply phenomenal terms directly
to the phenomenal states that we are currently aware of, without the mediation of any
functional, representational, behavioural, or physical definition or physical criteria.
Therefore, it doesn't seem that such criteria can constitute the meaning of phenomenal
terms.31 Also, analytic functionalism or representationalism cannot account for many of
the features on our list (a)–(i). Because of these problems, a majority of philosophers have
taken a different approach.
Physicalists, on the other hand, try to explain (a)–(i) in a manner compatible with
physicalism. A satisfactory explanation would be one on which (a)–(i), far from posing a
problem for the physicalist view, turned out to be features the physicalist would expect to
arise in our relation to phenomenal consciousness. On this view, metaphysical dualism is
false; however, there is a dualism of concepts: PCs are unlike other concepts in ways that
have very significant ramifications for how we ordinarily think about the mind;
specifically, they are responsible for our inclination to believe in dualism—even though
dualism is false.
The locus classicus for the Phenomenal Concept Strategy is Brian Loar's paper
‘Phenomenal States’ (1990/97).33 Loar suggested the idea that PCs are direct
recognitional concepts.34 Abstracting from some of the details, what he seems to have in
mind is that when a person is having a particular experience she can deploy a concept
that refers directly to that experience. Loar also suggests that the mode of presentation
of a PC involves the experience itself.
Page 11 of 26
Phenomenal Concepts
We can understand the subsequent discussion of PCs as developing two different strands
in Loar's original proposal. One kind of account elaborates on the idea that PCs refer
directly,35 so it emphasizes the special conceptual role of PCs; the other tries also to make
good Loar's suggestion that the mode of presentation of PCs in some way involves the
experience itself. I am going to sketch the basic ideas underlying these theories in the
rest of Section 17.5.
(p. 304)
Those who focus on the directness of the reference of phenomenal concepts have
proposed causal‐recognitional, demonstrative, and information‐theoretic accounts.
(i) Causal‐recognitional account. Tye (2003), for example, claims that PCs are special
recognitional concepts that refer directly. They have no associated reference‐fixing
descriptions; their mode of presentation is empty, so to speak. According to Tye, PCs
refer via the causal connection they have with their referents. On this account, a
phenomenal concept C
refers to a phenomenal quality Q via C's being the concept that is exercised in an
introspective act of awareness by person P if, and only if, under normal conditions
of introspection, Q is tokened in P's current experience and because Q is tokened.
(Tye 2003: 7)
But since someone could be wired to recognize another person's brain states (that happen to be
phenomenal states), it is clear that to be a PC it is not enough to be a recognitional concept of a
phenomenal state. Tye thinks that for the concept to be a PC it also has to have the right sort of
functional role. This functioning, however, cannot be specified a priori in a way that eschews any
phenomenal language.
(ii) Demonstrative account. A number of philosophers have suggested that PCs are a sort
of demonstrative. On Perry's account (2001) PCs are demonstratives, equivalent to
something like ‘this qualitative character’, where the demonstrative is guided by a
perceptual state to its referent. Levin, on the other hand, suggests that PCs are type
demonstratives without any mode of presentation at all. She thinks physicalists ‘should
reject the claim that phenomenal concepts require some sort of “presence” of, or
“acquaintance” with … the quality denoted, since this claim is backed only by the
intuitions that they have already explained away’ (2007: 105).36
Page 12 of 26
Phenomenal Concepts
All of the above accounts are aimed at explaining the conceivability of zombies and the
explanatory gap in a manner compatible with physicalism. The main point is that there is
nothing in the idea of direct reference appealed to in the recognitional and demonstrative
accounts that is at odds with physicalism—but once you have concepts that refer directly
in the way suggested by these accounts, the conceivability of zombies and the existence
of the explanatory gap will follow. Consequently, the (p. 305) conceivability of zombies
and the existence of the explanatory gap are compatible with physicalism.
These accounts, however, are less successful at explaining some other features on our list
(a)–(i).37 An examination of (a)–(i) suggests that a successful account of PCs will posit an
intimate connection between conscious states and the concepts we form of them. We can
see this by considering that the above accounts conceive of PCs and their referents as
distinct existences related by causation. It seems that this leaves too much of a distance
between, for example, a basic application of the PC pain to a particular pain as it occurs
and the particular pain itself, as on this view their occurrence is independent. In
particular, it is conceivable on this account that a basic application of pain be tokened by
someone in the complete absence of pain. But it seems that this is really inconceivable.
Anybody who tokens a basic application of pain is really in pain.
Loar tried to capture the special intimacy between phenomenal concepts and phenomenal
states by proposing that the mode of presentation of a PC involves the experience itself
that the concept refers to. If thinking about one's own current pain already somehow
involves pain itself then the situations we have been talking about are ruled out. But how
should this idea be best understood? As Papineau (2002: ch. 4) points out, by ‘mode of
presentation’ we cannot mean an associated description that we can already think and
use to refer to an entity which has those properties the description attributes. That would
be presupposing PCs in the explanation of those very concepts. We have to think about
the mode of presentation of PCs in some other way.
Page 13 of 26
Phenomenal Concepts
The Constitutional Account. On Carruthers's view, the special mode of presentation PCs
have is to be accounted for partly by the special nature of phenomenal experiences
themselves (that they represent themselves) and partly by the special nature of PCs (they
are direct recognitional concepts guided by these self‐representing states). There is
another way of thinking about PCs which is silent about the nature of phenomenal states
but still incorporates Loar's suggestion that phenomenal states themselves are involved
in their own presentation by PCs. This view involves variations on the idea that
phenomenal concepts are constituted by the phenomenal experiences they refer to. More
precisely, on this view, every concept token applied to current experience is constituted
by a current token phenomenal experience, and—on most versions of the constitutional
account—this fact is crucial in determining the reference of the concept.
On this account, there is an intimate relation between a phenomenal concept and its
referent, more intimate than any causal or tracking relation. It is also a way of fleshing
out the idea that the experience serves as its own mode of presentation. Metaphorically
speaking, a token of the reference provides the ink in which the token concept is
written.39
In terms of the RTM/LOT model, what is special about PCs then is that in so far as they
have a mode of presentation it has to do with the special vehicle involved in the basic
applications of the concept/mental representation.40 Not only is it the case that a token
state that realizes a token concept is also a token of the referent, but it is because the
concept is so constituted that it so refers. Unlike most concepts—for example the concept
dog, where it doesn't matter exactly what neural configurations constitute a particular
token of dog as long as the requisite causal/informational relations between it and dogs
hold—in the case of phenomenal concepts, for example the concept pain, constitution
matters for reference, both in terms of how the reference is determined, and in terms of
how the concept cognitively ‘presents’ its reference.
Versions of this view have been proposed, on the physicalist side, by Hill and McLaughlin
(1999), Papineau (2002, 2007), Balog (2006), and Block (2007), and David Chalmers
(2003) put forward a variation of this account on the dualist side.
(p. 307)
Page 14 of 26
Phenomenal Concepts
The constitutional account can explain many features of our epistemic relation to
phenomenal consciousness on our list (a)–(i). Take the infallibility of certain kinds of
phenomenal judgements. On the constitutional account, tokens of a phenomenal concept
that refers to a particular type of visual experience, say reddish, are constituted in part
by tokens of that type of experience. Then, for example, in a basic application of the
judgement I am experiencing reddish one's judgement cannot fail to be true.
Semantic stability is also easily explainable on the constitutional account. PCs refer to the
same properties independently of the actual context because—according to the
constitutional account—their reference is determined by their constitution (and
conceptual role), and that is independent of any external factors.
Explanations of the other features on the list (a)–(i) can be constructed on the
constitutional account as well. I will only mention the response to the Conceivability
Argument that this account of PCs makes possible.
(1999: 449)
On their account, phenomenal concepts are special recognitional concepts that are constituted
by their referents:
When one uses a sensory concept to classify one's own current experiences, the
experiences that guide and justify one in applying the concept are always identical
with the experiences to which the concept is applied. Sensory states are self‐
presenting states: we experience them, but we do not have sensory experiences of
them. We experience them simply by virtue of being in them. Sensory concepts are
recognitional concepts: deploying such concepts, we can introspectively recognize
when we are in sensory states simply by focusing our attention directly on them.
(1999: 448)
Block (2007) has a similar view of phenomenal concepts. He discusses the nature of phenomenal
concepts in the context of the many versions of the Property Dualism Argument (see e.g. White
2007), a close relative of the Conceivability Argument:
Page 15 of 26
Phenomenal Concepts
if a token phenomenal feel does double duty … (as a token of an aspect of both the
pain and our way of thinking of the pain), no extra specter of dualism arises. If the
phenomenal (p. 308) feel is a physical property, then it is a physical property even
when it (or a token of it) does double duty.
There are many questions the constitutional account raises, but one is particularly urgent: How
do phenomenal concepts come to refer to experiences that they themselves exemplify? How does
the constitution relation determine or partly determine the reference of a phenomenal concept?
The idea that it does seems strange, since it is not the case for most concepts. The concept dog
is not constituted by dogs, and the fact that the concept atom is constituted by atoms has
nothing to do with why it refers to atoms. The problem of how phenomenal concepts refer is a
pressing one for philosophers across the board, but whereas dualists can appeal to a primitive
relation of acquaintance, physicalists are under a strict obligation to provide a naturalistic
account; that is, an account that appeals only to physicalistically respectable entities and
properties.
The idea of an item partly constituting a representation that refers to that item is
reminiscent of how linguistic quotation works. The referent of ‘—’ is exemplified by
whatever fills in the blank. In a quotation expression, a token of the referent is literally a
constituent of the expression that refers to a type which it exemplifies, and that
expression has its reference (at least partly) in virtue of the properties of its constituent.
Some physicalists have tried to follow up on this idea to explain the reference of
phenomenal concepts.
Papineau (2002) has put forward one of the most elaborate versions of the constitutional
account. He suggests that phenomenal concepts are formed by prefixing perceptual
experiences with the operator ‘the experience …’. He calls this the quotational account of
phenomenal concepts. He hopes to give an answer to questions about the reference of
phenomenal concepts by invoking teleosemantics:
We should also note that phenomenal concepts are compound referring terms
(composed of an ‘experience operator’ and a ‘perceptual filling’). … [A] causal or
teleosemantic account of phenomenal concepts will view the contribution of the
parts to the semantic value of the whole as depending on the systematic
contribution which those parts make to the causes or biological functions of the
wholes they enter into.
(2002: 117)
Papineau (2007), however, apparently in keeping with the teleosemantic account, claims that the
fact that phenomenal concepts are constituted by exemplars of their referent plays no direct role
in explaining why they so refer. This amounts to a repudiation of the idea that phenomenal
concepts work in similar ways to quotation expressions.
Balog (2006), on the other hand, holds that phenomenal concepts are very closely
analogous to quotation expressions and that one must look to the conceptual role of
phenomenal concepts for an explanation of this.
Page 16 of 26
Phenomenal Concepts
These versions of the Phenomenal Concept Strategy are all meant to support physicalism
by invoking the special nature of PCs to explain the epistemic puzzles (p. 309) involving
phenomenal consciousness. However, since these accounts are neutral about the nature
of phenomenal properties, they can be adopted by a non‐physicalist. Chalmers (2003)
himself proposes a version of the constitutional account. On such an account, the
explanations of most aspects of our epistemic relation to phenomenal consciousness will
look much the same, with the exception that phenomenal concepts are constituted by
non‐physical states. However, according to the Phenomenal Concept Strategy, there will
now be two parallel explanations of features (a)–(i). Take the conceivability of zombies.
The dualist says that zombies are conceivable because phenomenal properties are not
physical or functional properties. But this explanation is redundant, since, as we have
seen, there is an explanation of why zombies are conceivable in terms of the special
nature of PCs. Accordingly, the Phenomenal Concept Strategy, if successful, provides
support for physicalism and undermines the rationale for dualism.
In a related vein, Chalmers (2007) poses an intriguing dilemma for the Phenomenal
Concept Strategy. Let C be the physicalist's account of phenomenal concepts, or the
physicalist's account of key features of these concepts responsible for our epistemic
relation to phenomenal consciousness. He argues that if a scenario physically exactly like
ours is conceivable where C is missing, then C is not physically explicable. On the other
hand, if such a scenario is not conceivable then C cannot explain our epistemic situation.
These criticisms provide challenges for the Phenomenal Concept Strategy. There are
other worries for the strategy relating to how, on a physicalist account, determinate
reference to an objective property can be achieved by subjective phenomenal concepts
(see e.g. Papineau 2002: ch. 7). Many of these problems require purely philosophical
treatment; however, one might wonder if in the future psychology and neuroscience will
Page 17 of 26
Phenomenal Concepts
References
Aydede, M., and Güzeldere, G. (2005), ‘Cognitive Architecture, Concepts, and
Introspection: An Information‐theoretic Solution to the Problem of Phenomenal
Consciousness’, Noûs, 39: 197–255.
Block, N. (1994), ‘An Argument for Holism’, Proceedings of the Aristotelian Society, 94:
151–69.
—— (2003), ‘Mental Paint’, in M. Hahn and B. Ramberg (eds.), Reflections and Replies:
Essays on the Philosophy of Tyler Burge (Cambridge, Mass.: MIT Press), 165–200.
—— (2007), ‘Max Black's Objection to Mind–Body Identity’, in T. Alter and S. Walter (eds.),
Phenomenal Concepts and Phenomenal Knowledge: New Essays on Consciousness and
Physicalism (Oxford: Oxford University Press), 249–306.
—— (2007), ‘Phenomenal Concepts and the Explanatory Gap’, in T. Alter and S. Walter
(eds.), Phenomenal Concepts and Phenomenal Knowledge: New Essays on Consciousness
and Physicalism (Oxford: Oxford University Press), 167–94.
Page 18 of 26
Phenomenal Concepts
Hill, C., and McLaughlin, B. (1999), ‘There are Fewer Things in Reality than are Dreamt
of in Chalmers's Philosophy’, Philosophy and Phenomenological Research, 59: 445–54.
—— (2003), ‘Mind and Illusion’, in A. O'Hear (ed.), Minds and Persons (Cambridge:
Cambridge University Press), 251–71.
Kripke, S. (1980), Naming and Necessity (Cambridge, Mass.: Harvard University Press).
Levine, J. (2001), Purple Haze: The Puzzle of Consciousness (Oxford: Oxford University
Press).
Lewis, D. (1966), ‘An Argument for the Identity Theory’, Journal of Philosophy, 63: 17–25.
Page 19 of 26
Phenomenal Concepts
Perry, J. (2001), Knowledge, Possibility and Consciousness (Cambridge, Mass.: MIT Press).
Rey, G. (2007), ‘Phenomenal Content and the Richness and Determinacy of Colour
Experience’, Journal of Consciousness Studies, 14/9–10.
repr. in Russell, Mysticism and Logic (London: Allen & Unwin, 1963), 152–67.
—— (1918/19), ‘The Philosophy of Logical Atomism’, Monist, 28: 495–527; 29: 32–63, 190–
222, 345–80;
repr. in Russell, Logic and Knowledge (London: Allen & Unwin, 1956), 177–281.
Stoljar, D. (2005), ‘Physicalism and Phenomenal Concepts’, Mind and Language, 20: 469–
94.
Sturgeon, S. (1994), ‘The Epistemic View of Subjectivity’, Journal of Philosophy, 91: 221–
36.
Page 20 of 26
Phenomenal Concepts
Tye, M. (2000), Consciousness, Colour and Content (Cambridge, Mass.: MIT Press).
36: 137–51; repr. in B. Gertler (ed.), Privileged Access: Philosophical Accounts of Self‐
knowledge (Aldershot: Ashgate), 31–44.
White, S. (2007), ‘The Argument for the Semantic Premise’, in T. Alter and S. Walter
(eds.), Phenomenal Concepts and Phenomenal Knowledge: New Essays on Consciousness
and Physicalism (Oxford: Oxford University Press), 210–48.
Notes:
(1) I will talk of the ‘ “what it's like” character’, or ‘qualitative character’ of those
experiences, ‘qualia’, and ‘qualia properties’ interchangeably. There are some issues
involving representationalism about qualia concerning a possible distinction between
‘qualitative character’ and ‘qualia’. Later I will mention these issues briefly, but for the
most part this possible distinction will not play a part in the discussion.
(2) There are also philosophers who downplay the uniqueness of PCs—I will come back to
this issue when I discuss particular accounts of PCs.
(3) The term ‘acquaintance’ was introduced in this context by Bertrand Russell. Russell
(1910) developed his famous distinction between ‘knowledge by acquaintance and
knowledge by description’. He then went on, in his (1918/19) lectures on logical atomism,
to argue, in a Cartesian manner, that we are only ever acquainted with ‘sensibilia’;
roughly, our phenomenal experiences.
Page 21 of 26
Phenomenal Concepts
complete description of our universe must include truths about where, when, and which
consciousness properties are exemplified.
(5) Main accounts of the nature of phenomenal character are representationalism (e.g.
Tye 2000), higher‐order monitoring theories (e.g. Rosenthal 2002), the self‐
representational theory (e.g. Kriegel 2004), and different varieties of dualism (e.g.
Chalmers 1996), according to which qualia are either fundamental non‐physical or non‐
physically realized properties or they comprise the categorical bases of physical
dispositional properties. Some physicalists hold that though qualia are physical, it is not
possible to give an explanatorily perspicuous account of them in non‐phenomenal terms.
The accounts of phenomenal concepts we will discuss are non‐committal concerning
these debates about the nature of phenomenal character, though as far as the mind–body
problem is concerned some of them have been used to attempt to defeat anti‐physicalist
arguments.
(6) Rey (2007), for example, denies that phenomenal concepts refer to any properties at
all and so he denies the existence of qualia altogether. He still thinks that phenomenal
concepts play an interesting and unique role in our mental life. Eliminativists, however,
may not want to deny altogether that there is something it is like to see a red rose, or, at
least, that it seems that there is something it is like to see a red rose. I won't pursue this
matter here.
(7) I use ‘transparence’ here not in the usual sense it is invoked in arguments for
representationalism, but rather to indicate the sense that qualia properties reveal their
essence (and the fact that what is revealed is their whole essence) directly to their
subjects.
(8) These are not exclusive options for those who think that there is something unique
about PCs. Papineau (2007), for example, thinks that there is something unique about
phenomenal concepts but it doesn't lie either in a unique reference relation or in some
special feature of the referent; it rather lies in a special feature of the physical vehicle of
the concept. More on this later.
(11) Under certain circumstances even basic applications of PCs can refer to the
phenomenal experiences of others; for example, when one thinks that another person is
Page 22 of 26
Phenomenal Concepts
having the same type of experience one is currently undergoing. Kati Farkas made this
point in conversation.
(13) There are various other accounts of the exact relations non‐basic, derivative
applications of PCs bear to basic ones (see e.g. Papineau 2007).
(15) Block (2003) seems to hold that there can be qualia in the absence of representation.
(16) The originator and foremost exponent of this view is Jerry Fodor; see e.g. Fodor
(1975).
(17) Mentalese is a hypothesized language over whose expressions mental processes are
defined. It is a language in the way that computer languages are languages and of course
is not a language for communication. Some philosophers resist the LOT hypothesis but it
is widely employed in cognitive‐psychological models of mental processes.
(18) For some concepts, e.g. descriptions, the concept's reference is determined by its
mode of presentation. But for other concepts, e.g. names, reference may be determined
by more than, or factors other than, the mode of presentation.
(19) Holists (e.g. Block 1994) say all of the role is individuative, molecularists (e.g.
Peacocke 1992) say just some, and atomists (e.g. Fodor 1987) say none.
(21) There are two ways of understanding functional characterizations of properties: one
as picking out the first‐order‐state type, if there is one, that satisfies the functional
characterization and the other as picking out the second‐order functional property
specified by the functional characterization.
(23) His remarks in §§ 243–315 are often referred to as the ‘private‐language argument’.
Page 23 of 26
Phenomenal Concepts
Let us imagine the following case. I want to keep a diary about the recurrence of a
certain sensation. To this end I associate it with the sign ‘S’ and write this sign in
a calendar for every day on which I have the sensation.—I will remark first of all
that a definition of the sign cannot be formulated.—But still I can give myself a
kind of ostensive definition.—How? Can I point to the sensation? Not in the
ordinary sense. But I speak, or write the sign down, and at the same time I
concentrate my attention on the sensation—and so, as it were, point to it inwardly.
—But what is this ceremony for? For that is all it seems to be! A definition surely
serves to establish the meaning of a sign.—Well, that is done precisely by the
concentrating of my attention; for in this way I impress upon myself the
connection between the sign and the sensation.—But ‘I impress it on myself’ can
only mean: this process brings it about that I remember the connection right in
the future. But in the present case I have no criterion of correctness. One would
like to say: whatever is going to seem right to me is right. And that only means
that here we can't talk about ‘right’.
(1953: § 258)
(25) Papineau (2002: 97–8) makes the point that natural‐language psychological words
like ‘pain’ are ambiguous between phenomenal and psychological concepts of pain or
perhaps express a concept that is an amalgam of the two.
(26) See e.g. Harman (1990), Tye (2000), and Jackson (2004) for transparency arguments.
Representationalists argue that when one attends to one's conscious experience one is
aware only of the representational content of the experience, or, alternatively, only of
features of the objects perceived, and conclude from this that, contrary to common sense,
there are no intrinsic, qualitative, introspectable features of conscious experience. This is
a controversial thesis that runs counter to the acquaintance thesis. I am merely noting
the disagreement; we cannot go into the merits of the argument here.
(27) I will confine myself here to the conceivability and gap arguments. Arguments from
semantic stability are based on variations of the premise that if all the concepts in some
identity claim are semantically stable then its truth or falsity can be determined a priori,
coupled with the claim that psychophysical identities contain semantically stable terms
and yet they are a posteriori. From these premises dualism follows. The exact definition
of semantic stability and the formulation of the corresponding arguments are too
complicated to go into in this discussion. Related arguments are Nida‐Rümelin (2007) and
White (2007); Chalmers's two‐dimensional framework (2006 and Ch. 18 below) also yields
a version of the argument.
(28) Chalmers (2002) distinguishes between two notions of conceivability: positive and
negative conceivability. I am going to rely on the notion of negative conceivability here,
given its greater clarity and simplicity.
Page 24 of 26
Phenomenal Concepts
(30) See also Kirk (2005) for an interesting argument against the conceivability of
zombies on grounds that go beyond analytic functionalism.
(32) Some dualists, e.g. Chalmers (2003), have more to say about acquaintance and
phenomenal concepts. I'll discuss those proposals below.
(33) The idea that the mind–body problem is a product of the special ways in which we
conceive (in the first‐person) of our phenomenal states is first formulated in this paper. A
similar proposal by Scott Sturgeon (1994) appeals to the special epistemology of
phenomenal states.
(34) Loar, like other proponents of the Phenomenal Concept Strategy, focuses his
attention on the basic applications of PCs, and thinks about the non‐basic applications as
somehow derivative of the more central, basic cases. See the earlier discussion of basic
and non‐basic applications of PCs.
(35) Stoljar (2005) argues that there are analyticities involving phenomenal concepts and
that this refutes the idea—integral to the Phenomenal Concept Strategy—that PCs are
direct and unanalysable. However, his examples merely show that there are analytically
necessary conditions on something being a conscious state but not that there are
analytically sufficient conditions, and this is all a proponent of the Phenomenal Concept
Strategy needs to be committed to.
(37) Some philosophers—e.g. Dennett (1991), Levin (2007)—deny these other features of
our epistemic access to consciousness.
(38) By ‘reddish’ I mean the type of phenomenal experience typically caused by seeing
red objects in ordinary light, etc.
Page 25 of 26
Phenomenal Concepts
(39) Notice that this view of PCs is distinct from the self‐representational account of
phenomenal states. The latter is a view of the nature of phenomenal states and is
compatible with constitutional and non‐constitutional accounts of PCs (see e.g.
Carruthers's recognitional account), while the former is a view of PCs and is compatible
with different accounts of the nature of phenomenal states. Of course, one can combine
the self‐representational account of the nature of phenomenal states with the
constitutional account of PCs and claim that the kind of self‐representation that is
essential for a sensory state to be phenomenally conscious is just to be represented by a
PC partly constituted by the sensory state itself. Ouroboros indeed.
(40) That means, among other things, that PCs in their basic applications simply don't
have descriptive modes of presentation.
(41) ‘Sensory concepts’ in Hill and McLaughlin's usage apply to what we have called
phenomenal concepts.
Katalin Balog
Page 26 of 26
The Two‐Dimensional Argument Against Materialism
A number of popular arguments for dualism start from a premise about an epistemic gap
between physical truths and truths about consciousness, and infer an ontological gap
between physical processes and consciousness. Arguments of this sort include the
conceivability argument, the knowledge argument, the explanatory gap argument, and
the property dualism argument. Such arguments are often resisted on the grounds that
epistemic premises do not entail ontological conclusions. This article views that one can
legitimately infer ontological conclusions from epistemic premises, if one is careful about
how one reasons. To do so, the best way is to reason first from epistemic premises to
modal conclusions (about necessity and possibility), and from there to ontological
conclusions. Here the crucial issue is the link between the epistemic and modal domains.
Keywords: dualism, epistemic gap, physical truths, truths about consciousness, ontological gap, physical
processes, consciousness
A number of popular arguments for dualism start from a premise about an epistemic gap
between physical truths and truths about consciousness, and infer an ontological gap
between physical processes and consciousness. Arguments of this sort include the
conceivability argument, the knowledge argument, the explanatory gap argument, and
the property dualism argument. Such arguments are often resisted on the grounds that
epistemic premises do not entail ontological conclusions.
My view is that one can legitimately infer ontological conclusions from epistemic
premises, if one is careful about how one reasons. To do so, the best way is to reason first
from epistemic premises to modal conclusions (about necessity and possibility), and from
there to ontological conclusions. Here the crucial issue is the link between the epistemic
and modal domains. How can one reason from theses about what is knowable or
conceivable to theses about what is necessary or possible?
Page 1 of 26
The Two‐Dimensional Argument Against Materialism
To bridge the epistemic and modal domains, the framework of two‐dimensional semantics
can play a central role. I have used this framework in earlier work (Chalmers 1996) to
mount an argument against materialism. Here I want to revisit the argument, (p. 314)
laying it out in a more explicit and careful form, and responding to a number of
objections.
Here P is the conjunction of all microphysical truths about the universe, specifying the
fundamental features of every fundamental microphysical entity in the language of microphysics.
Q is an arbitrary phenomenal truth: perhaps the truth that someone is phenomenally conscious,
or perhaps the truth that a certain individual (that is, an individual satisfying a certain
description) instantiates a certain phenomenal property. P&∼Q conjoins the former with the
denial of the latter.
If Q is the truth that someone is phenomenally conscious, then P&∼Q is the statement
that everything is microphysically as in our world, but no one is phenomenally conscious.
In this version P&∼Q says that the world is a zombie world. If Q is the truth that a certain
individual instantiates a certain phenomenal property, then P&∼Q is the statement that
everything is microphysically as in our world, but that it is not the case that the individual
in question instantiates the relevant phenomenal property. In this case it will suffice for
the truth of P&∼Q that the world is a zombie world, or simply that the individual in
question is a zombie in a physically identical world. It will also suffice that the individual
in question is an invert, who has an experience that differs slightly from the
corresponding experience of the corresponding individual in our (physically identical)
world.
The first premise of this argument asserts an epistemic thesis, about what can be
conceived. The second premise steps from the epistemic thesis to a modal thesis, about
what is possible. The third premise steps from the modal thesis to a metaphysical thesis,
about the nature of our world.
The third premise is relatively uncontroversial. It is widely accepted that materialism has
modal commitments. Some philosophers question whether materialism is equivalent to a
modal thesis, but almost all accept that materialism at least entails a modal thesis. Here
Page 2 of 26
The Two‐Dimensional Argument Against Materialism
one can invoke Kripke's metaphor: if it is possible that there is a world physically
identical to our world but phenomenally different, then after God fixed the physical facts
about our world, he had to do more work to fix the phenomenal facts.
A familiar complication arises from the observation that physicalism about our world is
compatible with the possibility of dualism in other worlds, and in particular (p. 315) is
compatible with the possibility of a physically identical world that contains extra, non‐
physical phenomenology. This means that if Q is a negative truth about our world—say,
the truth that no one instantiates a certain phenomenal property—then materialism about
our world is compatible with the possibility of P&∼Q. To finesse this point we can
stipulate that in the argument above, Q is a positive truth (one that holds in all worlds
that contain a duplicate of our world: see Chalmers 1996: 40): if Q is a positive truth, then
materialism is incompatible with the possibility of P&∼Q. Alternatively, we can conjoin P
with a ‘that's‐all’ statement T, stating that the world is a minimal world that satisfies P
(see Jackson 1998: 26). Then even when Q is a negative truth, materialism is not
compatible with the possibility of PT&∼Q (where PT is the conjunction of P and T).
The real work in the argument is done by the first and second premises. The second
premise is particularly controversial, as there are a number of examples that have led
many philosophers to deny that there is an entailment from conceivability to metaphysical
possibility. To assess these premises we need to understand the notion of conceivability.
We can say that S is prima facie conceivable for a subject when that subject is unable to
rule out the hypothesis expressed by S by a priori reasoning, on initial consideration. We
can say that S is ideally conceivable when the hypothesis expressed by S cannot be ruled
out a priori, even on ideal rational reflection. The main difference here is that prima facie
conceivability is tied to a subject's contingent cognitive limitations, while ideal
conceivability abstracts away from those limitations.
Some examples: (1) ‘2 + 2 = 5’ is neither prima facie conceivable nor ideally conceivable.
(2) Where S is a highly complex but provable mathematical truth, ∼S will be prima facie
conceivable for most subjects, but it is not ideally conceivable. (3) Where S is ‘There is a
flying pig’, S is prima facie conceivable, and is almost certainly ideally conceivable.
Page 3 of 26
The Two‐Dimensional Argument Against Materialism
In so far as there is a gap between prima facie conceivability and ideal conceivability, it is
ideal conceivability that is a better guide to possibility. This is especially clear in the case
of prima facie negative conceivability: we have seen that the negation of a complex
mathematical truth may be prima facie negatively conceivable, but it is not ideally
conceivable and it is not possible. It is less easy to find cases of prima facie positive
conceivability without ideal positive conceivability (see Chalmers 2002 for some potential
cases), but in so far as there are such cases, there will be little reason to think that they
are possible.
Page 4 of 26
The Two‐Dimensional Argument Against Materialism
There is no reason to believe that both Goldbach's conjecture and its negation are ideally
conceivable, so there is no reason to think that this is a counter‐example to the claim that
ideal conceivability entails possibility. So from here onward talk of ‘conceivability’
simpliciter should always be understood to be talk of ideal conceivability (either positive
or negative).
The other familiar class of purported counter‐examples arises from Kripke's analysis of
the necessary a posteriori. It is often said that sentences such as ‘Water is not H2O’
provide counter‐examples to the claim that conceivability entails possibility: it is
conceivable that water is not H2O, but it is not metaphysically possible.
(p. 317)
Here one has to be careful. There is a sense of ‘conceivable’ in which ‘Water is not H2O’ is
not conceivable (given that water is H2O in the actual world): in this sense, any
conceivable situation in which it seems that water is not H2O (a Twin Earth world, say)
whould better be described as a conceivable situation in which water is still H2O, but in
which there is watery stuff that is not H2O. Using the term ‘conceivable’ this way, one
might say that ‘Water is not H2O’ seems conceivable (or that it is prima facie conceivable,
to one without relevant empirical knowledge), but that it is not really conceivable. We
might call this sense of conceivability secondary conceivability (for reasons familiar from
a two‐dimensional analysis and discussed in Chalmers 2002). Then the Kripkean cases are
compatible with the claim that secondary conceivability entails metaphysical possibility.
But at the same time this claim is not very useful for present purposes, as whether a
sentence is secondarily conceivable will typically depend on a variety of empirical factors,
and an opponent might deny that zombies are secondarily conceivable, on the grounds
that there is an a posteriori identity between phenomenal and physical properties. So a
link between secondary conceivability and possibility does not offer an a priori route to
conclusions about metaphysical possibility.
Instead, what is relevant here is primary conceivability: the sense in which ‘Water is not
H2O’ can correctly be said to be conceivable. The notion of negative conceivability
defined above is a sort of primary conceivability, as it is defined in terms of what can be
ruled out a priori, and ‘Water is H2O’ cannot be established a priori. (One might define a
distinct notion of negative secondary conceivability, but I will set that aside here.) One
can likewise define a notion of positive primary conceivability, so that S is positively
primarily conceivable when a subject can imagine a coherent situation that verifies S,
where a situation verifies S when, under the hypothesis that the situation actually
obtains, the subject should conclude that S. If the subject imagines a Twin Earth situation
with XYZ in the oceans and lakes, and assumes that the situation obtains in their own
environment, then the subject should conclude that water is XYZ rather than H2O. So
‘Water is not H2O’ is positively primarily conceivable, as well as negatively primarily
conceivable.
Page 5 of 26
The Two‐Dimensional Argument Against Materialism
Still, there remains a link between primary conceivability and metaphysical possibility in
these cases. When we conceive that water is not H2O we imagine (for example) a Twin
Earth situation in which the watery liquid in the oceans and lakes is XYZ. This situation is
metaphysically possible, so there is a sense in which our conceiving involves access to a
possible world. Under the usual way of describing possible worlds, this world is not a
world in which water is not H2O. But the world still stands in a strong relation to the
sentence ‘Water is not H2O.’ In particular, if we came to accept that our own world had
the character of this world (with XYZ in the oceans and lakes), we should then endorse
the claim ‘Water is not H2O.’
(p. 318)
This can be put in two‐dimensional terms by saying that while the Twin Earth does not
satisfy ‘Water is not H2O’ (‘Water is not H2O’ is not true of that world considered as
counterfactual), the Twin Earth world verifies ‘Water is not H2O’ (‘Water is not H2O’ is
true of that world considered as actual). Equivalently, we can say that while the
secondary intension of ‘Water is not H2O’ is false at W, the sentence's primary intension is
true there. To a first approximation, a world W verifies S (or S is true at W considered as
actual, or the primary intension of S is true at W) when if we came to accept that our own
world is qualitatively like W we should then endorse S. Strictly speaking, the worlds W
that are relevant to primary intensions are centred worlds: worlds that come with a
marked ‘centre’ consisting of an individual and time. When we consider a centred world
W as actual, we consider the hypothesis that we are currently in the situation of the
individual at the centre. (For much more on these notions see Chalmers 2004.)
We can say that when the primary intension of S is true at some centred world (i.e. when
some centred world verifies S) S is primarily possible, or 1‐possible. When the secondary
intension of S is true at some world (i.e. when some world satisfies S), S is secondarily
possible, or 2‐possible. Then ‘Water is not H2O’ is not 2‐possible, but it is 1‐possible.
The observation that sentences such as ‘Water is not H2O’ are conceivable but not
possible, in these terms, comes to the claim that these sentences are primarily
conceivable (1‐conceivable) but are not secondarily possible (2‐possible). So there is good
reason to believe that 1‐conceivability does not entail 2‐possibility. However, these cases
are entirely compatible with a link between 2‐conceivability and 2‐possibility, and, more
importantly for present purposes, they are entirely compatible with a link between 1‐
conceivability and 1‐possibility.
In fact, it is not hard to argue that all of the standard Kripkean a posteriori necessities
(‘Heat is the motion of molecules’, ‘Hesperus is Phosphorus’, and so on) have this
structure. For each of these necessities one might say that its negation is conceivable but
Page 6 of 26
The Two‐Dimensional Argument Against Materialism
not possible, meaning that it is 1‐conceivable but not 2‐possible. But in each of these
cases the sentence in question is 1‐possible. For example, ‘Heat is not the motion of
molecules’ is verified by a world in which something other than molecules causes
sensations as of heat. ‘Hesperus is not Phosphorus’ is verified by a world in which the
objects visible in the morning and evening skies are entirely distinct. Furthermore, it is
plausible that worlds such as these are just what one is conceiving of when one conceives
that heat is not the motion of molecules, or that Hesperus is not Phosphorus. So in these
cases there remains a strong link between conceivability and metaphysical possibility.
There are two versions of this thesis, depending on whether one interprets the relevant
sort of conceivability as positive or negative.
CP− entails CP+, as ideal primary positive conceivability entails ideal primary negative
conceivability. If S can be ruled out a priori, then no coherent imagined situation will verify S. It
is not obvious whether or not CP+ entails CP−, as it is not obvious whether ideal primary
negative conceivability entails ideal primary positive conceivability. That is, it is not obvious
whether or not there is an S that cannot be ruled out a priori, but such that no coherent
imagined situation verifies S. (In Chalmers 2002 I argue that there is no such S, so that ideal
primary negative conceivability entails ideal primary positive conceivability.) So CP− is at least
as strong as CP+ and is possibly somewhat stronger.
Most importantly for present purposes, however, both CP+ and CP− are compatible with
all the familiar purported counter‐examples to the conceivability–possibility link.
Furthermore, it seems that there are no clear counter‐examples to either thesis (though
later in the chapter I will discuss some potential counter‐examples that have been put
forward). In particular, both theses are entirely compatible with the existence of Kripkean
a posteriori necessities, so while existence of these necessities is often used to case doubt
on conceivability–possibility theses, they cannot be used to cast doubt on CP+ or CP−.
So for now I will take these theses as reasonable conceivability–possibility theses that
might be used in mounting a refined conceivability argument against materialism. Later
in the chapter I will return to the question of their truth.
Page 7 of 26
The Two‐Dimensional Argument Against Materialism
On this reading (1) and (2) are both plausible theses, but (3) is not obviously plausible. The
reason is that materialism requires not the 1‐impossibility of P&∼Q but the 2‐impossibility of
P&∼Q. That is, materialism requires that it could not have been the case that P were true
without Q being true. This is a subjunctive claim about ordinary metaphysical possibility, and so
invokes 2‐impossibility rather than 1‐impossibility.
A materialist might reasonably question (3) by holding that even if there is a world W
verifying P&∼Q, W might be a world with quite different ingredients from our own. For
example, it might be that W does not instantiate true microphysical properties (those
instantiated in our world), such as mass and charge, but instead instantiates quite
different properties; say, pseudo‐mass and pseudo‐charge, which stand to mass and
charge roughly as XYZ stands to H2O. Likewise, it might be that W does not lack true
phenomenal properties, but instead lacks quite different properties: say, pseudo‐
phenomenal properties. If so, then the possibility of W has no bearing on whether true
microphysical properties necessitate true phenomenal properties. And it is the latter that
is relevant for materialism.
Still, it may be that the gap between 1‐possibility and 2‐possibility could be closed. In
particular, when a statement S has the same primary intension and secondary intension,
then a world will verify S iff it satisfies S, so S will be 1‐possible iff it is 2‐possible. If P
and Q both have primary intensions that coincide with their secondary intensions, then so
will P&∼Q, and we could run the following argument:
Page 8 of 26
The Two‐Dimensional Argument Against Materialism
Here the truth of (3) requires that both P and Q have primary and secondary intensions that
coincide. In the case of Q, this claim is quite plausible. As Kripke noted, there does not seem to
be the same strong dissociation between appearance and reality in the case of consciousness as
in the cases of water and heat: while it is not the case that anything that looks like water is
water, or that anything that feels like heat is heat, it is plausibly the case that anything that feels
like consciousness is consciousness. So it is not clear that the notion of ‘pseudo‐consciousness’,
something that satisfies the primary intension of ‘consciousness’ without being consciousness, is
coherent. Likewise for other more specific phenomenal properties. So there is a strong case that
the primary and secondary intensions of phenomenal terms coincide. (For more on this case see
Chalmers 2003.)
However, in the case of P this claim is less plausible. A materialist might reasonably hold
that microphysical terms (such as ‘mass’ and ‘charge’) have primary intensions that differ
from their secondary intensions. In particular, it is plausible that the primary intensions
of these terms are tied to a certain theoretical role. We might (p. 321) say that the
primary intension of ‘mass’ picks out whatever property plays the mass role (e.g. resisting
acceleration in certain ways, being subject to mutual attraction in a certain way, and so
on), and that the primary intension of ‘charge’ picks out whatever property plays the
charge role (e.g. obeying certain electromagnetic principles, being subject to attraction
and repulsion in certain ways, and so on).
By contrast, one might reasonably hold that the secondary intension of microphysical
terms is tied to the property that actually plays the role. For example, if property M plays
the mass role in the actual world, then one might hold that in any world in which mass is
instantiated, mass is M. It follows that if there are worlds in which some other property
M′ plays the mass role, then M′ is not mass in that world (at best, it is pseudo‐mass). If so,
then the primary and secondary intensions of ‘mass’ will not coincide: the primary
intension of ‘mass’ will pick out whatever plays the mass role in such a world, but the
secondary intension will not.
There are other views of the semantics and metaphysics of microphysical terms that may
reject this argument for the distinctness of the primary and secondary intensions of
‘mass’. In particular, the argument will not go through on views according to which it is
necessary that mass is the property that plays the mass role. (These include views on
which ‘mass’ is a non‐rigid designator whose secondary intension picks out different
properties that play the role in some worlds, and views on which it is necessary that M is
the property that plays the mass role, where M is the property rigidly designated by
‘mass’.) Still, the view sketched above is a quite reasonable view—more plausible than
the alternatives, in my opinion—and it is the view that best grounds resistance to an
inference from the 1‐possibility of P&∼Q to its 2‐possibility. So we can suppose that the
view is correct in order to see what follows.
Page 9 of 26
The Two‐Dimensional Argument Against Materialism
On this view, a world may verify P without satisfying P. The secondary intension of P
requires that certain specific properties such as mass, spin, and charge are distributed in
a certain specific way across space–time, with appropriate causal and nomic relations
among them. The primary intension of P requires only that whatever properties play the
mass role, the spin role, and charge role are distributed in this way. If W is a world where
these roles are played by properties other than mass, spin, and charge (we might say that
they are played by ‘schmass’, ‘schmin’, and ‘schmarge’), which are otherwise distributed
in the right way over space–time and have appropriate causal and nomic relations among
them, then W will verify P, but it will not satisfy P. Here we might say that the physics of
W has the same structural profile as physics in the actual world, but that it has a different
intrinsic profile, in that it differs in the intrinsic properties that fill this structure. To
verify P, a world must have the right structural profile, while to satisfy P, a world must
have the right structural and intrinsic profile.
It follows that premise 3 is not guaranteed to be true. Because a world can verify P
without satisfying P, it may be that P&∼Q is 1‐possible but not 2‐possible. However, this
requires that P and Q be related in a certain specific way. In particular, it requires that
some worlds that verify P also verify ∼Q, while no worlds that satisfy P also satisfy ∼Q.
This requires in turn that some worlds that have the same structural profile as the actual
world verify ∼Q, while no worlds that have the same structural (p. 322) and intrinsic
profiles as the actual world satisfy ∼Q. We can assume for the moment that the primary
and secondary intensions of Q coincide. Then we can put all this by saying that the falsity
of (3) requires that the structural profile of physics in the actual world does not
necessitate Q, but that the combined structural and intrinsic profiles of physics in the
actual world do necessitate Q.
This idea—that the structural properties of physics in the actual world do not necessitate
the existence and/or nature of consciousness, but that the intrinsic properties of physics
combined with the structural properties do—corresponds to a familiar view in the
metaphysics of consciousness. This is the view that I have elsewhere called Russellian
monism (or type‐F monism, or panprotopsychism). On this view, consciousness is closely
tied to the intrinsic properties that serve as the categorical bases of microphysical
dispositions. Russell and others held that the nature of these properties is not revealed to
us by perception (which reveals only their effects) or by science (which reveals only their
relations). But it is coherent to suppose that these properties have a special nature that is
tied to consciousness. They might themselves be phenomenal properties, or they might be
proto‐phenomenal properties: properties that collectively constitute phenomenal
properties when organized in the appropriate way.
Page 10 of 26
The Two‐Dimensional Argument Against Materialism
form of physicalism. But because it relies on speculation about the special nature of the
fundamental properties in microphysics, it is a highly distinctive form of physicalism that
has much in common with property dualism, and that many physicalists will want to
reject.1
In any case, we can now close the loophole in the previous argument as follows:
This argument is valid. I discussed the case for premises (1), (2), and (4) earlier, and I
(p. 323)
have just now argued for premise (3). I think that (5) is the proper conclusion of the
conceivability argument. For the reasons given above, such arguments (and also related
arguments such as the knowledge argument and the property‐dualism argument) cannot exclude
Russellian monism, and Russellian monism is arguably a form of physicalism, if a distinctive and
radical kind. So the possibility of Russellian monism needs to be explicitly acknowledged as an
option in the conclusion.
A couple of minor notes on the argument. First, to be fully explicit the argument might
take the truth of Q as a premise. If Q were false, the ground for accepting (4) would
collapse. In the less explicit version of the argument above we can consider the truth of Q
part of the case for accepting premise (4). In fact, for reasons given earlier, the case for
(4) requires that Q is a positive truth about consciousness. Alternatively, one can remain
silent on whether Q is a positive or negative truth, and handle this matter by conjoining P
with a ‘that's‐all’ clause asserting that the world is a minimal world in which P (or,
equivalently, by building such a that's‐all clause into P).
Second, it is worth noting that (contrary to a common supposition), the assumption that Q
has the same primary and secondary intensions is not necessary for the argument for (5)
to go through. To see this we can consider the version of the argument where we adjoin a
‘that's‐all’ clause to P. From (1) and (2) we can derive the conclusion that there is a
minimal world verifying P in which the primary intension of Q is false. If P has the same
primary and secondary intensions, then this world will be a minimal P‐world in which the
primary intension of Q is false. This world must differ from our world, because the
primary intension of Q is true in our world. (There is a small loophole here arising from
the possibility that this world differs merely in the location of the centre of the relevant
centred world. I discuss this loophole in the expanded version of this paper.) It follows
that there is a minimal P‐world that is not a duplicate of our world, so that physicalism is
false of our world. It could be that strictly speaking physicalism will be true of
consciousness, because P necessitates Q, but physicalism will be false of properties
closely associated with consciousness; namely, those associated with the primary
Page 11 of 26
The Two‐Dimensional Argument Against Materialism
intension of Q. We might think of this sort of view as one on which phenomenal properties
are physical properties that have non‐physical properties as modes of presentation.
Alternatively, if P has different primary and secondary intensions, then by the reasoning
given in the earlier discussion of premise (3) one can conclude that either there is a
minimal world satisfying P in which the primary intension of Q is false (which again
entails the falsity of physicalism), or that the primary intension of Q is necessitated by the
structural and intrinsic profiles of physics in our world, but not by the intrinsic profiles
alone. This view can be considered another form of Russellian monism, in that the
intrinsic properties of physics in our world are crucial for constituting the properties
associated with the modes of presentation of consciousness. So if Q has distinct primary
and secondary intensions, then one will have to formalize premises (3) and (4) somewhat
differently, but the argument for (5) will still work just as well.
I will also pass over some hard‐to‐classify objections, including the conditional‐concepts
objection (Hawthorne 2002; Stalnaker 2002; Braddon‐Mitchell 2003), the zombie‐parity
objection (Balog 1999), the indexical objection (e.g. Ismael 1999; Perry 2001), and
objections to two‐dimensional semantics (e.g. Soames 2004). Again, these objections are
discussed at length in the expanded version of this chapter.
Instead, I will concentrate here on objections to the crucial premise (2). It is this premise
that bridges the epistemic and modal domains, and it is this premise and associated
principles that have attracted the most in‐depth philosophical discussion.
Premise (2) says that if P&∼Q is conceivable, P&∼Q is 1‐possible. This premise can be
seen as an instance of the general conceivability–possibility thesis CP:
Page 12 of 26
The Two‐Dimensional Argument Against Materialism
Here ‘conceivability’ should be understood as ideal primary conceivability, of either the negative
or positive variety (I always take ‘ideal primary’ as understood from here on). The two versions
of the thesis that result are equivalent to theses CP− and CP+, discussed earlier. Thesis CP− is
equivalent to the claim that if ∼S is not a priori, S is 1‐possible. The positive version CP+,
holding that if S is positively conceivable, S is 1‐possible, is somewhat weaker than the negative
version, as positive conceivability entails negative conceivability but the reverse is not obviously
the case. Much of my discussion will apply equally to both CP+ or CP−, so I will often just speak
of CP, except where the distinction is relevant.
The case for premise (2) largely derives from the case for CP, and from here on I will
mostly focus on the general principle rather than the specific premise. Of course, if it
turns out that the general principle needs to be restricted to a certain class of sentences
to be plausible, then the question will arise as to whether P&∼Q falls into that class.
Why believe CP? In the first instance, the thesis is plausible because there are no clear
counter‐examples to it. Principles linking conceivability and possibility have been widely
accepted in the history of philosophy, but have more recently been questioned because of
various counter‐examples, such as the Goldbach case (both Goldbach's conjecture and its
negation are conceivable but only one is possible) and especially the Kripke cases
(‘Hesperus is not Phosphorus’ is conceivable but not possible). But CP accommodates
these examples straightforwardly, with the idealization (p. 325) accommodating Goldbach
cases, and the primary/secondary distinction accommodating Kripke cases. If it handles
these cases, then the central sources of resistance to conceivability–possibility principles
is undermined. But, of course, there may be other possible sources of resistance.
It is easy to see that CP− is equivalent to the thesis that there are no strong necessities.
If S is negatively conceivable but not 1‐possible, then ∼S will be a strong necessity. If S is
a strong necessity, then ∼S will be negatively conceivable but not 1‐possible.
Page 13 of 26
The Two‐Dimensional Argument Against Materialism
In so far as CP+ is potentially weaker than CP−, the relationship between CP+ and the
thesis that there are no strong necessities is not as clear. Certainly any counter‐example
to CP+ will yield a strong necessity, but the reverse is not obviously the case. To handle
this we might define two classes of strong necessities, according to whether they provide
counter‐examples to CP+ or merely to CP−. Let us say that a negative strong necessity is
a statement S such that S is 1‐necessary and 2‐necessary but ∼S is negatively
conceivable. The latter condition is equivalent to the requirement that S is not a priori, so
negative strong necessities are equivalent to strong necessities as defined above. A
positive strong necessity is a statement S such that S is 1‐necessary and 2‐necessary
while ∼S is positively conceivable. Then all positive strong necessities are negative strong
necessities, but the reverse is not trivially the case. CP− and CP+ are then equivalent to
the theses that there are no negative strong necessities and that there are no positive
strong necessities respectively.
What would a strong necessity involve? To get an idea, consider a philosophical view on
which it is metaphysically necessary that an omniscient being (e.g. God) exists, but on
which it is not a priori that such a being exists. Then according to this view, ‘An
omniscient being exists’ (or O) is an a posteriori necessity. Like all a posteriori
necessities, O is 2‐necessary, and ∼O is negatively conceivable (and also positively
conceivable, if we add the plausible claim that it is positively conceivable that there is no
omniscient being). If O were an ordinary a posteriori necessity, then O would be 1‐
contingent: there would be a metaphysically possible world verifying ∼O. But if there is
no omniscient being, then it seems that there is no such world. (p. 326) ‘There is an
omniscient being’ does not seem to have any difference in its primary and secondary
intensions, so if a world satisfies O, it verifies O. So given that O is 2‐necessary, O is 1‐
necessary. It follows that if this philosophical view is correct, then O is a strong necessity:
it is at least a negative strong necessity, and given the positive conceivability claim above,
it is a positive strong necessity.
One could put the matter by saying that there is an epistemically possible scenario
verifying ∼O, but no metaphysically possible world verifying ∼O. Here a scenario can be
understood as corresponding to a maximal a priori coherent hypothesis, in the way that
worlds correspond to maximal metaphysically possible hypotheses. (I give a formal
treatment of scenarios in Chalmers 2004 and (forthcoming b), but here I will leave the
notion intuitive. (One might call this sort of scenario a negative scenario, since it
corresponds to a maximal negatively conceivable hypothesis. One could also define a
positive scenario so that it corresponds to a maximal positively conceivable hypothesis.)
The notion of scenarios is not defined in terms of metaphysical possibility, and in
particular it is not assumed that scenarios correspond to metaphysically possible worlds.
But nevertheless it is plausible that there is an intimate relationship.
For any a posteriori necessity S there will be a scenario verifying ∼S. For example, as
‘Water is H2O’ is not a priori, there will be a scenario verifying ‘Water is not H2O’. That is,
there will be some maximal a priori coherent hypothesis H (perhaps involving the
assumption that the watery stuff is made of XYZ, and so on) such that if we accept H, we
Page 14 of 26
The Two‐Dimensional Argument Against Materialism
should accept ‘Water is not H2O’. For ordinary a posteriori necessities these scenarios
will correspond closely to centred metaphysically possible worlds, so that there will be a
centred world verifying ∼S. For example, ‘Water is not H2O’ is verified by a centred XYZ‐
world, where the individual at the centre is and has always been surrounded by clear,
drinkable XYZ in the oceans and lakes. There is little reason to doubt that such a world is
metaphysically possible, and there is an intuitive sense in which it qualitatively matches
the scenario that we imagine when we entertain the hypothesis that water is not XYZ.
When S is a strong necessity, by contrast, there will be a scenario verifying ∼S, but this
scenario will correspond to no metaphysically possible world. (When S is a positive strong
necessity, there will be a positive scenario verifying ∼S; when S is a negative strong
necessity, there will be a negative scenario verifying ∼S.) For example, given the theist
view outlined above, there will be a (negative and positive) scenario verifying ‘There is no
omniscient being’, involving some maximally detailed hypothesis under which there is no
such being. But on this view there is no centred world that corresponds to this scenario,
and there is no centred world that itself verifies ‘There is no omniscient being’. We might
put this intuitively by saying that on this view the space of (centred) metaphysically
possible worlds is smaller than the space of epistemically possible scenarios, at least in
the relevant respect. On this view, there are scenarios that correspond to no world.
To bring this back to the mind–body case: take the paradigmatic type‐B materialist who
holds that premise (1) is true, premise (2) is false, and materialism is true. On this view,
the material conditional P ⊃ Q (which is itself the negation of P&∼Q) is a strong
necessity. The truth of materialism implies that it is 2‐necessary, the truth (p. 327) of (1)
implies that it is a posteriori and its negation is 1‐conceivable, but the falsity of (2)
implies that its negation is 1‐necessary. On this view, there will be a scenario verifying
P&∼Q, including various specific zombie scenarios. But these scenarios will correspond
to no metaphysically possible world.
Note that the analogue of CP with scenarios instead of worlds is close to trivial: if S is
conceivable in the relevant sense, there will automatically be a scenario verifying S (at
least if the notions of a scenario and of verification are unproblematic). Even a type‐B
materialist and a believer in strong necessities can accept that principle. They must
simply deny that all scenarios correspond to worlds. So CP might be seen as equivalent to
the thesis that for every scenario there is a corresponding world.
My view is that for every scenario there is a corresponding world, and that there are no
strong necessities. In The Conscious Mind I gave the following reasons for this: (i) strong
necessities cannot be supported by analogy with other a posteriori necessities; (ii) they
involve a far more radical sort of a posteriori necessity than Kripke's, requiring a
distinction between conceptual and metaphysical possibility at the level of worlds; (iii)
they lead to an ad hoc proliferation of modalities; (iv) they raise deep questions of
coherence; (v) they will be brute and inexplicable; and (vi) the only motivation to posit a
Page 15 of 26
The Two‐Dimensional Argument Against Materialism
strong necessity in the mind–body case is the desire to save materialism. I still accept
most of these reasons, but there is more to say.
In the last decade or so numerous objections to CP have been proposed. These objections
fall into a number of classes. The first, and largest, involves attempts at exhibiting clear
cases of strong necessities. The second involves attempts at explaining how there might
be strong necessities in the phenomenal case (if not elsewhere), by analysing the nature
of phenomenal concepts. The third involves general philosophical objections. I will
discuss the objections in the first class in the section that follows.
It is occasionally proposed that some Kripkean a posteriori necessities are in fact strong
necessities. In particular, it is sometimes proposed that coextensive names (p. 328) such
as ‘Cicero’ and ‘Tully’ may have the same primary intension as well as the same
secondary intension. If so, ‘Cicero is Tully’ is a strong necessity (as it is clearly a
posteriori).
In response: When we entertain the hypothesis that Cicero is not Tully, this hypothesis
corresponds to specific scenarios that we can elaborate. In particular, the relevant
scenarios may involve the hypothesis that the causal chains associated with the names
‘Cicero’ and ‘Tully’ pick out different historical individuals. A scenario like this certainly
corresponds to a centred metaphysically possible world: there are certainly worlds where
the causal chains associated with these words functions in this way. And it seems clear
that if we discovered that our world were such a world we would reject the hypothesis
that Cicero is Tully. So such a world seems to verify ‘Cicero is not Tully’ (although it may
not satisfy ‘Cicero is not Tully’). Worlds like this can be found for any Kripkean a
posteriori necessity S (as Kripke himself pointed out, in effect), and such worlds will
always verify ∼S.
One might resist the claim that the world in question verifies S by rejecting the claim that
there is an a priori entailment (for the speaker in question) from a description of the
world in question to S. In response, I think that the sort of considerations in Chalmers
Page 16 of 26
The Two‐Dimensional Argument Against Materialism
and Jackson (2001) strongly suggest that these entailments are a priori at least in
principle. But even if one rejects this claim, there clearly remains some distinctive
epistemic relation between the world in question and ‘Cicero is not Tully’: in particular, it
remains the case that if one accepts (hypothetically) that the actual world is qualitatively
just like the world in question, and reflects on this hypothesis, then one will reject the
claim that Cicero is not Tully. (Note that this is quite unlike the situation that the theist
thinks obtains in the God case, where there is no world that stands in this inferential
relation to ‘There is no God’.) So if one rejects a priori entailments, one can use this sort
of inferential relation to define primary intensions, and ordinary Kripkean necessities will
always have a primary intension that is false at some world.
Another class of examples comes from the suggestion that consciousness is not alone in
its failure to be a priori entailed by microphysical truths. Block and Stalnaker (1999) have
suggested that many ordinary macroscopic truths, such as ‘Water boils at 100
degrees’ (or W), are not a priori entailed in this way. If so, then P&∼W is at least
negatively conceivable. But P&W is not possible (at least if water and its properties
supervene on the microphysical), and it is not clearly 1‐possible, either. If it is not, then
P&∼W is a counter‐example to CP−.
In response: Chalmers and Jackson (2001) argue that these macroscopic truths are a
priori entailed by P, or at least by PQTI, a conjunction of full physical, phenomenal, ‘that's
all’, and indexical information. Of course P&∼Q, P&∼T, and P&∼I may be negatively
conceivable, but the latter two are clearly also 1‐possible, while the former is precisely
the main topic of dispute. So any failures of entailment by P associated (p. 329) with the
failure of entailment to Q, T, or I give no further support to the existence of strong
necessities. Such support would require at least a truth M such that PQTI does not a
priori entail M, and such that PQTI ⊃ M is 2‐necessary and 1‐necessary. The arguments in
Chalmers and Jackson (2001) suggest that at no ordinary macroscopic truths M are like
this.
In any case, even if one rejects the a priori entailment thesis here, these cases will yield
at best exceptions to CP− (i.e. negative strong necessities), not exceptions to CP+ (i.e.
positive strong necessities). Even if W above is like Q in that P&∼W and even PQTI&∼W
is negatively conceivable, PQTI&∼W is not positively conceivable: ‘water‐zombies’ are not
positively conceivable in the way that zombies are positively conceivable. So this sort of
case leaves CP+ unthreatened.
Page 17 of 26
The Two‐Dimensional Argument Against Materialism
Perhaps the most challenging cases for CP− are mathematical truths M such that M is
true (and so necessarily true and 1‐necessarily true) but not knowable, and so not
knowable a priori. If there are such cases, M is a negative strong necessity (though not a
positive strong necessity, as ∼M is not positively conceivable). Here one might appeal to
unprovable true mathematical sentences, such as those whose existence is entailed by
Godel's theorem.
In response: Unprovability in a given system does not entail non‐apriority. For example,
the consistency of Peano arithmetic is not provable in Peano arithmetic, but is still
plausibly knowable a priori. One can make the case (see Chalmers 2002) that all true
statements of arithmetic are knowable a priori at least under an idealization (i.e. our
failure to know them a priori is due to certain limitations of our cognitive systems). One
might worry about higher set theory, such as the continuum hypothesis, but here it is far
from clear that such sentences are determinately true or false, and it is also far from
clear that they are not knowable a priori under an idealization. So although these cases
provide an interesting challenge to CP−, they do not provide clear counter‐examples to
CP−. And, as in the previous case, there is no challenge to CP+ here.
According to some philosophical views, laws of nature are not just naturally necessary
but are metaphysically necessary. Shoemaker (1998) holds such a view, and suggests that
it may provide a counter‐example to CP.
In response: On some varieties of this view, laws of nature follow a version of the
Kripkean model. That is, if mass in the actual world obeys certain laws, then nothing in
any counterfactual world counts as mass unless it obeys exactly those laws, so any law
involving mass will be necessary. This might hold because (p. 330) of the semantics of
mass (which require that a counterfactual property have the same nomic role as actual
mass in order to qualify as the referent of ‘mass’), or it might hold because of the
metaphysics of mass (according to which properties such as mass have their nomic
profile essentially, as on Shoemaker's view). On these models we need not deny that there
are worlds that correspond to the scenario we conceive when we conceive that mass
obeys different laws: it is just that such worlds will contain ‘schmass’ rather than mass. I
think that it is implausible that the modal profile of ‘mass’ and/or the essential properties
of mass are this precise (see Fine 2002 and Sidelle 2002 for arguments), but in any case,
this model does not provide a counter‐example to CP. In this case, a schmass world may
verify the hypothesis that the relevant law of nature is false, so laws of nature are not
strong necessities.
To yield strong necessities this sort of view must hold that not only are there no worlds
where mass obeys different laws, but there are also no related worlds where something
else, ‘schmass’, obeys those different laws. Here the relevant sort of view is one
according to which the fundamental properties and laws of all worlds are the
Page 18 of 26
The Two‐Dimensional Argument Against Materialism
fundamental properties and laws of our world (and on which these laws are not knowable
a priori). In effect, this restricts the space of metaphysically possible worlds to the space
of naturally possible worlds. If this view is correct, then a fundamental law will be a
strong necessity: there will be no world corresponding to the scenario that we conceive
when we conceive what is false.
I think that there are no good reasons to accept this extremely strong view of laws of
nature, and that there are good reasons to reject it. The best reasons to take seriously the
hypothesis that laws of nature are necessary come from the Kripke and Shoemaker
models above. But nothing in these models supports the strong view, or yields a strong
necessity. Rather, the CP thesis can itself be taken as a reason to reject the view.
Page 19 of 26
The Two‐Dimensional Argument Against Materialism
I have already noted that the existence of strong necessities is entailed by a theist view
on which an omniscient being (or an omnipotent being, or a perfect being) exists
necessarily but on which the existence of such a being is not knowable a priori. If we say
that a god is by definition an omniscient being (or a perfect being, or whatever), then ‘A
god exists’ will be a strong necessity.
In response: I think that theist views of this sort are to be rejected. If the existence of
such a god is knowable a priori, then it may exist necessarily. But if it is not, then one
should conclude that such a being exists at best contingently. I cannot go over the
arguments for believing in a necessary god here, but they all rest on highly contentious
premises, and, once again, CP itself provides an argument against these views. The best
way to defend the existence of a necessary god is to argue that a world without such a
being is not even conceivable.
Even if one believes that the existence of a god provides a strong necessity, it is not clear
that this sort of strong necessity undermines the case against materialism. The debate
over materialism uses necessity as a criterion of ontological distinctness: the question of
whether physical truths necessitate phenomenal truths is relevant (p. 331) precisely in so
far as it bears on the question of whether the phenomenal involves nothing ontologically
‘over and above’ the physical. But a variety of necessity in which the existence of a god is
necessary will not be well suited to this role. On such a view, the existence of a god will
be necessitated by physical truths, but such a god will presumably nevertheless be
ontologically non‐physical. So if the only strong necessities are strong necessities of this
sort, connecting ontologically distinct existences, they are no help to physicalism. Under
this assumption, then, if P&∼Q is conceivable, Q will be something over and above the
physical, either because it is necessitated by the physical or because it is tied to the
physical only by this sort of strong necessity.
Something similar applies to views on which laws of nature are strong necessities. Even
on such views laws presumably connect ontologically distinct properties: if there is a
fundamental law connecting properties A and B, this will not ground any sort of
ontological reduction of one property to the other. Indeed, if this view is correct, then a
dualist view with fundamental phenomenal properties and fundamental laws connecting
them to the physical will itself be a view on which the phenomenal is necessitated by the
physical. So, again, strong necessities of this sort are no help to the physicalist.
Page 20 of 26
The Two‐Dimensional Argument Against Materialism
Yablo (1999) adapts the God case to provide an intriguing argument against CP.
According to Yablo, it is at least conceivable that there is a necessarily existing god. It is
also conceivable that there is no necessarily existing god. So if G is ‘It is necessary that
there is an omniscient being’, then both G and ∼G are conceivable. If so, then by CP both
G and ∼G are 1‐possible. There appears to be no relevant distinction between the primary
and secondary intensions of the expressions involved, so it follows that G and ∼G are 2‐
possible, or (metaphysically) possible simpliciter. So it is possible that it is necessary that
there is an omniscient being, and it is possible that it is not necessary that there is an
omniscient being. But this is a contradiction, at least given S5 as the logic of the
metaphysical modality. If it is possible that S is necessary, then S is necessary, so it is not
possible that S is not necessary.
(p. 332)
The problematic issues here arise because of the double modality: we are conceiving not
just of non‐modal qualitative features of worlds, but also of what is possible or necessary
within those worlds. Conceiving of a god (an omnipotent, omniscient, and benevolent
being, say) is arguably not too hard; but to conceive in addition that the being exists
necessarily, we have to conceive that the space of possible worlds is such that this god
exists in each of them, despite the conceivability of a godless world. That is, we have to
conceive that CP is itself false. This is what does all the work in the example: if it is
conceivable that CP is false, then (by CP!) it is possible that CP is false. CP is surely
necessarily true if it is true at all, so it follows from the possible falsity of CP that CP is
false.
‘cosmoses’ of possible worlds.) But I prefer to hold on to the stronger thesis, by denying
that it is conceivable that CP is false. I hold that CP is a priori, although highly non‐trivial,
like many theses in philosophy (an a priori argument for CP is given in Chalmers 1999,
and in the expanded version of this chapter). If this is correct, then CP is not conceivably
false on ideal rational reflection, and it is not ideally conceivable that a necessarily
existing god exists.
A closely related meta‐modal argument (Marton 1998; Sturgeon 2000; Frankish 2007)
that is specific to the mind–body domain proceeds as follows. (i) It is at least conceivable
that materialism is true about consciousness. So (ii) it is conceivable that P ⊃ Q is
necessary. By CP (and setting aside two‐dimensional structure), it follows that (iii) it is
possible that P ⊃ Q is necessary. But from this it follows (using S5) that (iv) P ⊃ Q is
necessary. Using CP and S5 one can equally infer from the fact that (v) it is conceivable
that P ⊃ Q is not necessary to the conclusion that (vi) P ⊃ Q is not necessary. But (iv) and
(vi) are contradictory. So one should reject CP.
My response here parallels the response in the god case. It may be prima facie negatively
conceivable that materialism is true about consciousness, but it is not obviously
conceivable in any stronger sense. Many people have noted that it is very hard to imagine
that consciousness is a physical process. I do not think this unimaginability is so obvious
that it should be used as a premise in an argument against materialism, but, likewise, the
imaginability claim cannot be used as a premise either. And if I am right that CP is a
priori, then there is an a priori argument that P ⊃ Q is not necessary, so that it will not
even be ideally negatively conceivable that P ⊃ Q is necessary.
It is also worth noting that cases such as these seem to work best as challenges to CP−
rather than to CP+, so that CP+, which is all that is required for the argument against
materialism, is relatively unthreatened. We have also seen at various points along the way
Page 22 of 26
The Two‐Dimensional Argument Against Materialism
that even if one takes certain cases to involve strong necessities, the existence of such
strong necessities will still be compatible with modified versions of CP (say, a version
involving ontological necessities in the law/god cases, or a version involving non‐modal
sentences in the meta‐modal cases) that will be strong enough for the anti‐materialist
argument to go through. So where the consideration of counter‐examples to CP is
concerned, the anti‐materialist seems to be on reasonably strong ground.
Materialists can also object to CP in other ways: for example, by giving theoretical
arguments for why it is false without relying on counter‐examples, or by giving an
explanation of why there should be special counter‐examples in the case of
consciousness. The second strategy has been especially popular, with a number of
theorists arguing that special treatment in the case of consciousness is warranted by the
special nature of phenomenal concepts (e.g. Loar 1997; Hill and McLaughlin 1999;
Papineau 2002). I discuss objections of this sort in the expanded version of this chapter,
and also give a theoretical argument for the truth of CP, arguing that it is grounded in
constitutive links between the modal and rational domains.2
References
Ashwell, L. (2003), ‘Conceivability and Modal Error’, MA thesis (University of Auckland).
Block, N., and Stalnaker, R. (1999), ‘Conceptual Analysis, Dualism, and the Explanatory
Gap’, Philosophical Review, 108: 1–46.
Page 23 of 26
The Two‐Dimensional Argument Against Materialism
—— (forthcoming b), The Nature of Epistemic Space, in A. Egan and B. Weatherson (eds.),
Epistemic Modality (Oxford: Oxford University Press).
Chalmers, D. J., and Jackson, F. (2001), ‘Conceptual Analysis and Reductive Explanation’,
Philosophical Review, 110: 315–61.
—— (2007), ‘Dancing Qualia and Direct Reference’, in T. Alter and S. Walter (eds.),
Phenomenal Concepts and Phenomenal Knowledge: New Essays on Consciousness and
Physicalism (Oxford: Oxford University Press), 195–209.
Hill, C., and McLaughlin, B. (1999), ‘There are Fewer Things than are Dreamt Of in
Chalmers' Philosophy’, Philosophy and Phenomenological Research, 59: 445–54.
Ismael, J. (1999), ‘Science and the Phenomenal’, Philosophy of Science, 66: 351–69.
Kirk, R. (1999), ‘Why There Couldn't Be Zombies’, Aristotelian Society Supplement, 73: 1–
16.
Lynch, M. (2006), ‘Zombies and the Case of the Phenomenal Pickpocket’, Synthese, 149:
37–58.
Marton, P. (1998), ‘Zombies vs. Materialists: The Battle for Conceivability’, Southwest
Philosophy Review, 14: 131–8.
Page 24 of 26
The Two‐Dimensional Argument Against Materialism
Schiffer, S. (2003), The Things We Mean (Oxford: Oxford University Press) (sect.
(p. 335)
Stoljar, D. (2001), ‘The Conceivability Argument and Two Conceptions of the Physical’,
Philosophical Perspectives, 15: 393–413.
Notes:
(1) A related position arises from views on which laws of nature are necessary (e.g.
Shoemaker 1998), and on which there is a lawful connection between physical properties
and phenomenal properties in our world. Such a view may hold that it is essential to
physical properties that they have this nomic profile, so that there is no world satisfying
Page 25 of 26
The Two‐Dimensional Argument Against Materialism
P&∼Q. Some versions of this view will deny (2), and are discussed later in the chapter.
But other versions may accept (2), holding that there is a world verifying P&∼Q, but will
hold that it involves distinct ‘schmysical’ properties that lack this nomic profile. The
resulting view resembles Russellian monism in some respects, but differs from the usual
form in taking the connection between physical and phenomenal properties to be nomic
in the first instance. Because it turns on this nomic connection, this view does not provide
any loophole for materialism: at best, it yields a version of property dualism on which the
laws of nature connecting physical and phenomenal properties are necessary.
(2) In the expanded version of this chapter I also consider the relation between the
arguments I have discussed here and other anti‐materialist arguments, including the
knowledge argument, the property‐dualism argument, Kripke's modal argument, and
Descartes's argument from disembodiment.
David J. Chalmers
Page 26 of 26
Intentional Systems Theory
Intentional systems theory is in the first place an analysis of the meanings of such
everyday ‘mentalistic’ terms as ‘believe’, ‘desire’, ‘expect’, ‘decide’, and ‘intend’: the
terms of ‘folk psychology’ that we use to interpret, explain, and predict the behaviour of
other human beings, animals, some artefacts such as robots and computers, and indeed
ourselves. In traditional parlance we seem to be attributing minds to the things we thus
interpret, and this raises a host of questions about the conditions under which a thing can
be truly said to have a mind, or to have beliefs, desires, and other ‘mental’ states.
According to intentional systems theory, these questions can best be answered by
analysing the logical presuppositions and methods of our attribution practices, when we
adopt the intentional stance toward something.
Keywords: intentional systems theory, folk psychology, mentalistic terms, human behaviour, mental states,
intentional stance
Page 1 of 15
Intentional Systems Theory
INTENTIONAL systems theory is in the first place an analysis of the meanings of such
everyday ‘mentalistic’ terms as ‘believe’, ‘desire’, ‘expect’, ‘decide’, and ‘intend’: the
terms of ‘folk psychology’ (Dennett 1971) that we use to interpret, explain, and predict
the behaviour of other human beings, animals, some artefacts such as robots and
computers, and indeed ourselves. In traditional parlance we seem to be attributing minds
to the things we thus interpret, and this raises a host of questions about the conditions
under which a thing can be truly said to have a mind, or to have beliefs, desires, and
other ‘mental’ states. According to intentional systems theory, these questions can best
be answered by analysing the logical presuppositions and methods of our attribution
practices, when we adopt the intentional stance toward something. Anything that is
usefully and voluminously predictable from the intentional stance is, by definition, an
intentional system. The intentional stance is the strategy of interpreting the behaviour of
an entity (person, animal, artefact, whatever) by treating it as if it were a rational agent
who governed its ‘choice’ of ‘action’ by a ‘consideration’ of its ‘beliefs’ and ‘desires’. The
scare‐quotes around all these terms draw attention to the fact that some of their standard
connotations may be set aside in the interests of exploiting their central features: their
role in practical reasoning, and hence in the prediction of the behaviour of practical
reasoners.
Alarm clocks, being designed objects (unlike the stone), are also amenable to a fancier
style of prediction: prediction from the design stance. Suppose I categorize a novel object
as an alarm clock. I can quickly reason that if I depress a few buttons just so, then some
hours later the alarm clock will make a loud noise. I don't need to work out the specific
physical laws that explain this marvellous regularity; I simply assume that it has a
particular design—the design we call an alarm clock—and that it will function properly, as
Page 2 of 15
Intentional Systems Theory
An even riskier and swifter stance is the intentional stance, a subspecies of the design
stance, in which the designed thing is treated as an agent of sorts, with beliefs and
desires and enough rationality to do what it ought to do given those beliefs and desires.
An alarm clock is so simple that this fanciful anthropomorphism is, strictly speaking,
unnecessary for our understanding of why it does what it does, but adoption of the
intentional stance is more useful—indeed, well‐nigh obligatory—when the artefact in
question is much more complicated than an alarm clock. Consider chess‐playing
computers, which all succumb neatly to the same simple strategy of interpretation: just
think of them as rational agents who want to win, and who know the rules and principles
of chess and the positions of the pieces on the board. Instantly your problem of predicting
and interpreting their behaviour is made vastly easier than it would be if you tried to use
the physical or the design stance. At any moment in the chess game, simply look at the
chessboard and draw up a list of all the legal moves available to the computer (p. 341)
when its turn to play comes up (there will usually be several dozen candidates). Now rank
the legal moves from best (wisest, most rational) to worst (stupidest, most self‐defeating),
and make your prediction: the computer will make the best move. You may well not be
sure what the best move is (the computer may ‘appreciate’ the situation better than you
do!), but you can almost always eliminate all but four or five candidate moves, which still
gives you tremendous predictive leverage. You could improve on this leverage and predict
in advance exactly which move the computer will make—at a tremendous cost of time and
effort—by falling back to the design stance and considering the millions of lines of
computer code that you can calculate will be streaming through the CPU of the computer
after you make your move, and this would be much, much easier than falling all the way
back to the physical stance and calculating the flow of electrons that results from
pressing the computer's keys. But in many situations, especially when the best move for
the computer to make is so obvious it counts as a ‘forced’ move, you can predict its move
with well‐nigh perfect accuracy without all the effort of either the design stance or the
physical stance.
It is obvious that the intentional stance works effectively when the goal is predicting a
chess‐playing computer, since its designed purpose is to ‘reason’ about the best move to
make in the highly rationalistic setting of chess. If a computer program is running an oil
refinery, it is almost equally obvious that its various moves will be made in response to its
Page 3 of 15
Intentional Systems Theory
detection of conditions that more or less dictate what it should do, given its larger
designed purposes. Here the presumption of excellence or rationality of design stands out
vividly, since an incompetent programmer's effort might yield a program that seldom did
what the experts said it ought to do in the circumstances. When information systems (or
control systems) are well designed, the rationales for their actions will be readily
discernible, and highly predictive—whether or not the engineers that wrote the programs
attached ‘comments’ to the source code explaining these rationales to onlookers, as good
practice dictates. We needn't know anything about computer programming to predict the
behaviour of the system; what we need to know about is the rational demands of running
an oil refinery.
Page 4 of 15
Intentional Systems Theory
We just as naturally and unthinkingly extend the intentional stance to animals, a non‐
optional tactic if we are trying to catch a wily beast, and a useful tactic if we are trying to
organize our understanding of the behaviours of simpler animals, and even plants. Like
the lowly thermostat, as simple an artefact as can sustain a rudimentary intentional‐
Page 5 of 15
Intentional Systems Theory
stance interpretation, the clam has its behaviours, and they are rational, given its limited
outlook on the world. We are not surprised to learn that trees that are able to sense the
slow encroachment of green‐reflecting rivals shift resources into growing taller faster,
because that's the smart thing for a plant to do under those circumstances. Where on the
downward slope to insensate thinghood does ‘real’ believing and desiring stop and mere
‘as if’ believing and desiring take over? According to intentional‐systems theory, this
demand for a bright line is ill‐motivated.
Page 6 of 15
Intentional Systems Theory
The contrast between original and derived intentionality is unproblematic when we look at the
paradigm cases from everyday life, but when we attempt to promote this mundane distinction
into a metaphysical divide that should apply to all imaginable artefacts, we create serious
illusions. Whereas our simpler artefacts, such as painted signs and written shopping lists, can
indeed be seen to derive their meanings from their functional roles in our practices, and hence
not have any intrinsic meaning independent of our meaning, we have begun making
sophisticated artefacts such as robots, whose trajectories can unfold without any direct
dependence on us, their creators, and whose discriminations give their internal states a sort of
meaning to them that may be unknown to us and not in our service. The robot poker player that
bluffs its makers seems to be guided by internal states that function just as a human poker
player's intentions do, and if that is not original intentionality, it is hard to say why not.
Moreover, our ‘original’ intentionality, if it is not a miraculous or God‐given property, must have
evolved over the aeons from ancestors with simpler cognitive equipment, and there is no
plausible candidate for an origin of original intentionality that doesn't run afoul of a problem
with the second distinction, between literal and metaphorical attributions.
The intentional stance works (when it does) whether or not the attributed goals are
genuine or natural or ‘really appreciated’ by the so‐called agent, and this tolerance is
crucial to understanding how genuine goal‐seeking could be established in the first place.
Does the macromolecule really want to replicate itself? The intentional stance explains
what is going on, regardless of how we answer that question. Consider a simple organism
—say a planarian or an amoeba—moving non‐randomly across the bottom of a laboratory
dish, always heading to the nutrient‐rich end of the (p. 344) dish, or away from the toxic
end. This organism is seeking the good, or shunning the bad—its own good and bad, not
those of some human artefact‐user. Seeking one's own good is a fundamental feature of
any rational agent, but are these simple organisms seeking or just ‘seeking’? We don't
Page 7 of 15
Intentional Systems Theory
need to answer that question. The organism is a predictable intentional system in either
case.
By exploiting this deep similarity between the simplest—one might as well say most
mindless—intentional systems and the most complex (ourselves), the intentional stance
also provides a relatively neutral perspective from which to investigate the differences
between our minds and simpler minds. For instance, it has permitted the design of a host
of experiments shedding light on whether other species, or young children, are capable of
adopting the intentional stance—and hence are higher‐order intentional systems. A first‐
order intentional system is one whose behaviour is predictable by attributing (simple)
beliefs and desires to it. A second‐order intentional system is predictable only if it is
attributed beliefs about beliefs, or beliefs about desires, or desires about beliefs, and so
forth. A being that can be seen to act on the expectation that you will discover that it
wants you to think that it doesn't want the contested food would be a fifth‐order
intentional system. Although imaginative hypotheses about ‘theory of mind
modules’ (Leslie 1991) and other internal mechanisms (e.g. Baron‐Cohen 1995) have been
advanced to account for these competences, the evidence for the higher‐order
competences themselves must be adduced and analysed independently of these proposals
about internal mechanisms, and this has been done by cognitive ethologists (Dennett
1983; Byrne and Whiten 1988) and developmental psychologists, among others, using the
intentional stance to design the experiments that generate the attributions that in turn
generate testable predictions of behaviour.
Page 8 of 15
Intentional Systems Theory
Page 9 of 15
Intentional Systems Theory
The last example is a paradigm of rational belief, but even in this case the attribution leaves a
great deal of the reasoning inexplicit. Notice, too, that in the other cases of quite unproblematic,
unmetaphorical human belief, such as (E), the swerving case, it is unlikely that anything like an
explicit representation of the relevant beliefs and (p. 346) desires (the propositional attitudes)
occurred in the driver's stream of consciousness. Had it not been for his beliefs about the
relative danger of swerving and being hit by an object, he would not have taken the action he
did; he would not have swerved to avoid a sheet of paper, for instance, and he would not have
swerved had he believed there was a bus in the lane he swerved into, but it is not clear how, if at
all, these guiding beliefs are represented unconsciously; there are so many of them. Similarly,
the cook in (F) may have quite ‘unthinkingly’ opened the tin of baking soda, but she would not
have opened a tin of loose tea or molasses had her hand fallen on it instead. Attributing a large
list of beliefs to these agents—including propositions they might be hard‐pressed to arteculate—
in order to account for their actions is a practice as secure as it is familiar, and if some of these
informational states don't pass somebody's test for genuine belief, so much the worse for the
claim that such a test must be employed to distinguish literal from metaphorical attributions.
The model of beliefs as sentences of Mentalese written in the belief box, as some would have it,
is not obligatory, and may be an artefact of attending to extreme cases of human belief rather
than a dictate of cognitive engineering.
(1) The Martian marionette (Peacocke 1983). Suppose we found an agent (called ‘The
Body’ by Peacocke) that passed the intentional‐stance test of agency with flying colours
but proved, when surgically opened, to be filled with radio transceivers; its every move,
however predictable and explicable by our attributions of beliefs and desires to it, was
actually caused by some off‐stage Martian computer program controlling the otherwise
lifeless body as a sort of radio‐controlled puppet. The controlling program ‘has been given
the vast but finite number of conditionals specifying what a typical human would do with
given past history and current stimulation; so it can cause The Body to behave in any
circumstances exactly as a human being would’ (Peacocke 1983: 205).
This is no counter‐example, as we can see by exploring the different ways the further
details of the fantasy could be fleshed out. If the off‐stage controller controls this body
and no other, then we were certainly right to attribute the beliefs and desires to the
person whose body we have surgically explored; this person, like Dennett in ‘Where am
I?’ (Dennett 1978), simply keeps his (silicon) brain in a non‐traditional location. If, on the
other hand, the Martian program has more (p. 347) than one (pseudo‐)agent under
Page 10 of 15
Intentional Systems Theory
control, and is coordinating their activities (and not just providing in one place n different
independent agent‐brains), then the Martian program itself is the best candidate for being
the intentional system whose actions we are predicting and explaining. (The Martian
program in this case really is a puppeteer, and we should recast all the only apparently
independent beliefs and desires of the various agents as in reality the intended
manifestations of the master agent. But of course we must check further on this
hypothesis to see if the Martian program is in turn controlled by another outside agent or
agents or is autonomous.) What matters in the identification of the agent to whom the
beliefs and desires are properly attributed is autonomy, not specific structures. Of course
a bowl of structureless jelly or confetti is not a possible seat of the soul, simply because
the complex multi‐track dispositions of a mind have to be realized somehow, in an
information‐processing system with many reliably moving (and designed) parts.
(2) The giant look‐up table (Block 1982): Having used the intentional stance to attribute
lots of clever thoughts, beliefs, and well‐informed desires to whoever is answering my
questions in the Turing test, I lift the veil and discover a computer system that, when
opened, turns out to have nothing in it but a giant look‐up table, with all the possible
short intelligent conversations in it, in alphabetical order. The only ‘moving part’ is the
alphabetic string‐searcher that finds the next canned move in this pre‐played game of
conversation and thereupon issues it. Surely this is no true believer, even though it is
voluminously predictable from the intentional stance, thereby meeting the conditions of
the definition.
There are several ways of rebutting this counter‐example, drawing attention to different
foibles of philosophical method. One is to observe that the definition of an intentional
system, like most sane definitions, has the tacit rider that the entity in question must be
physically possible; this imagined system would be a computer memory larger than the
visible universe, operating faster than the speed of light. If we are allowed to postulate
miraculous (physics‐defying) properties to things, it is no wonder we can generate
counter‐intuitive ‘possibilities’. One might as well claim that when one opened up the
would‐be believer one found nothing therein but a cup of cold coffee balanced on a
computer keyboard, which vibrated in just the miraculously coincidental ways that would
be required for it to type out apparently intelligent answers to all the questions posed.
Surely not a believer! But also not possible.
A more instructive response ignores the physical impossibility of the ‘Vast’ (Dennett 1995)
set of alphabetized clever conversations, and notes that the canning of all these
conversations in the memory is itself a process (an R&D process) that requires an
explanation (unless it is yet another miracle or a ‘cosmic coincidence’). How was the
quality control imposed? What process exhaustively pruned away the stupid or
nonsensical continuations before alphabetizing the results? Here it is useful to consider
the use of the intentional stance in evolutionary biology:
Page 11 of 15
Intentional Systems Theory
Francis Crick, one of the discoverers of the structure of DNA, once jokingly
credited his colleague Leslie Orgel with ‘Orgel's Second Rule’: Evolution is
cleverer than you are. Even the (p. 348) most experienced evolutionary biologists
are often startled by the power of natural selection to ‘discover’ an ‘ingenious’
solution to a design problem posed by nature.
When evolutionists like Crick marvel at the cleverness of the process of natural
selection they are not acknowledging intelligent design! The designs found in
nature are nothing short of brilliant, but the process of design that generates
them is utterly lacking in intelligence of its own.
The process of natural selection is a blind, foresightless, purposeless process of trial and error,
with the automatic retention of those slight improvements (relative to some challenge posed by
the world) that happen by chance. We can contrast it with intelligent design. Now how did the
giant look‐up table consisting of all and only clever conversations come to be created? Was it the
result of some multi‐zillion‐year process of natural selection (yet another impossibility) or was it
handcrafted by some intelligence or intelligences? If the latter, then we can see that Block's
counter‐example is a close kin to Peacocke's. Suppose we discovered that Oscar Wilde lay awake
nights thinking of deft retorts to likely remarks and committing these pairs to memory so that he
could deliver them if and when the occasion arose ‘without missing a beat’. Would this cast any
doubt on our categorization of him as an intelligent thinker? Why should it matter when the
cogitation is done, if it is all designed to meet the needs of a time‐pressured world in an efficient
way? This lets us see that in the incompletely imagined case that Block provides it might not be a
mistake to attribute beliefs and desires to this surpassingly strange entity! Just as Peacocke's
puppet does its thinking in a strange place, this one does its thinking at a strange time! The
intentional stance is maximally neutral about how (or where, or when) the hard work of
cognition gets done, but guarantees that the work is done by testing for success. In the actual
world, of course, the only way to deliver real‐time cleverness in response to competitively
variegated challenges (as in the Turing Test) is to generate it from a finite supply of already
partially designed components. Sometimes the cleverest thing you can do is to quote something
already beautifully designed by some earlier genius; sometimes it is better to construct
something new, but of course you don't have to coin all the words, or invent all the moves, from
scratch.
Coming from the opposite pole, Stich and others have criticized the intentional stance for
relying on the rationality assumption, making people out to be much more rational than
they actually are (see e.g. Stich 1981; Nichols and Stich 2003; Webb 1994). These
objections overlook two facts. First, without a background of routine and voluminous
fulfilment of rational expectations by even the most deranged human beings, such
unfortunates could not be ascribed irrational beliefs in the first place. Human behaviour
is simply not interpretable except as being in the (rational) service of some beliefs and
desires or other. And second, when irrationality does loom large, it is far from clear that
there is any stable interpretation of the relevant beliefs (see e.g. Dennett 1987: 83–116;
1994: 517–30).
Page 12 of 15
Intentional Systems Theory
References
Baron‐Cohen, S. (1995), Mindblindness: An Essay on Autism and Theory of Mind
(Cambridge, Mass.: MIT Press).
Brook, A., and Ross, D. (2002) (eds.), Daniel Dennett (Cambridge: Cambridge University
Press).
Byrne, R., and Whiten, A. (1988) (eds.), Machiavellian Intelligence: Social Expertise and
the Evolution of Intelligence in Monkeys, Apes, and Humans (Oxford: Oxford University
Press).
Davidson, D. (1975), ‘Thought and Talk’, in S. Guttenplan (ed.), Mind and Language:
Wolfson College Lectures, 1974 (Oxford: Clarendon), 7–23.
Page 13 of 15
Intentional Systems Theory
(p. 350) Dennett, D. (1991), ‘Real Patterns’, Journal of Philosophy, 87: 27–51.
—— (1995), Darwin's Dangerous Idea: Evolution and the Meanings of Life (New York:
Simon & Schuster).
—— (2006), ‘The Hoax of Intelligent Design, and How It was Perpetrated’, in J. Brockman
(ed.), Intelligent Thought: Science Versus the Intelligent Design Movement (New York:
Vintage), 33–49.
Griffin, R., and Baron‐Cohen, S. (2002), ‘The Intentional Stance: Developmental and
Neurocognitive Perspectives’, in A. Brook and D. Ross (eds.), Daniel Dennett (Cambridge:
Cambridge University Press), 83–116.
Leslie, A. (1991), ‘The Theory of Mind Impairment in Autism: Evidence for a Modular
Mechanism of Development?’ in A. Whiten (ed.), Natural Theories of Mind (Oxford:
Blackwell), 63–78.
Nichols, S., and Stich, S. (2003), Mindreading (Oxford: Oxford University Press).
Premack, D. (1983), ‘The Codes of Man and Beasts’, Behavioral and Brain Sciences, 6:
368.
Ross, D. (2002), ‘Dennettian Behavioural Explanations and the Roles of the Social
Sciences’, in A. Brook and D. Ross (eds.), Daniel Dennett (Cambridge: Cambridge
University Press), 140–83.
Seyfarth, R., and Cheney, D. (2002), ‘Dennett's Contribution to Research on the Animal
Mind’, in A. Brook and D. Ross (eds.), Daniel Dennett (Cambridge: Cambridge University
Press), 117–39.
Page 14 of 15
Intentional Systems Theory
Daniel Dennett
Page 15 of 15
Wide Content
Keywords: wide content, externalist content, physical environment, physical states, intrinsic states, mental state
THOUGHTS are individuated both by their attitude type and by their content. A thinking
subject might bear different attitudes to the same content—she might, for example, both
believe and fear that the conflict in the Middle East will not be resolved—or the same
attitude to different contents—she might believe, say, that the Devils will win the Stanley
Cup and that it will snow tomorrow. Wide or externalist content is individuated in part by
reference to features of a subject's external surroundings, by her physical environment or
the community with which she shares a language. Externalists hold that physically
identical subjects might have thoughts with different, wide, contents. Internalists either
deny that content is wide, claiming that all content supervenes on intrinsic physical states
of the subject, and is hence narrow, or insist, at the very least, that the content that
individuates a subject's thoughts in predictions and explanations of her behaviour must
supervene on her intrinsic properties.
I begin by discussing the arguments for and against wide content. I then argue that
genuinely explanatory content is wide.
Page 1 of 19
Wide Content
Hilary Putnam (1975) argued that the meaning of natural‐kind terms is determined, in
part, by the nature of the subject's physical environment. Putnam asks us to (p. 352)
consider a world, Twin Earth, in which the vocable ‘water’ is used to refer to the liquid
filling this world's oceans and streams, a liquid superficially similar to water, though
chemically quite different. Call this liquid ‘XYZ’. Imagine two subjects, Oscar and his
physically identical counterpart, Twin Oscar, circa 1750, before the development of
modern chemistry.1 Each subject utters the form of words ‘Cold water is refreshing on a
hot day’. According to Putnam, the meanings of the two utterances are different, in virtue
of the fact that they are used to refer to different substances—H2O by Oscar and XYZ by
Twin Oscar. If Putnam is right, the internal, psychological state of the speaker is
insufficient to determine the meaning of natural‐kind terms; the nature of the
environment also plays a meaning‐determining role. As Putnam famously put it,
‘meanings just ain't in the head’. Moreover, on the assumption that a subject's utterances
express his thoughts, the twins' thoughts have distinct contents, in virtue of the distinct
concepts expressed by their use of the vocable ‘water’.
Tyler Burge (1978) argued that the contents of many thoughts are determined, in part, by
community standards of correctness. A medically unsophisticated subject, Alf, believes
that he has arthritis in his ankles, wrists, and, as it happens, his thigh. When informed by
his doctor that arthritis is a disease of the joints, Alf gives up the belief that he has
arthritis in his thigh. We can compare Alf with a second subject, Twin Alf, who has an
identical history, non‐intentionally described. Twin Alf belongs to a linguistic community
that uses the vocable ‘arthritis’ to refer to an ailment that affects the thigh as well as the
joints. Both Alf's and Twin Alf's use accord with this community's use. Burge concludes
that the beliefs that Alf and Twin Alf would express using the form of words ‘Arthritis is
painful’ have different contents. (On the assumption that beliefs have their contents
essentially, Alf and Twin Alf have different beliefs.) The content of thoughts ascribable
using general terms such as ‘arthritis’ is determined, in part, by community‐wide social
and linguistic practice, and is therefore wide.
An argument originally due to Gareth Evans (1982) establishes that thoughts involving
singular reference—so called object‐involving thoughts—have wide content. My belief
that this glass (imagine me pointing to a specific glass in front of me) is empty depends
constitutively upon the existence of a particular glass. My perceptual contact with this
glass puts me in a position to make a singular demonstrative judgement that this glass is
Page 2 of 19
Wide Content
empty, a judgement that my Twin Earth counterpart, pointing to a different glass, is not
in a position to make. Her judgement has a different wide content. The content of object‐
involving thoughts is the least controversial type of wide content and will not be
discussed further.
Page 3 of 19
Wide Content
It has been suggested that the difference in Oscar's and Twin Oscar's physical
environments does not give rise to a difference in the contents of their thoughts, because
both believe something like the following: the clear potable liquid that fills the oceans and
streams is refreshing. The revised belief contents are assumed to supervene on intrinsic
properties of subjects. The issue, more generally, is whether the twins’ concepts include
within their extension not only the liquid that fills their local oceans and streams (H2O
and XYZ respectively) but also anything phenomenally indistinguishable from this liquid.
If, as Crane (1991) and Segal (2000) have argued, the twins’ concepts apply to the
‘watery’ substance in their worlds in virtue of the relatively superficial properties shared
by the two substances, then they have the same concept, and Putnam's argument fails to
establish the existence of wide content.
The fact that nobody in 1750 was in a position to provide a distinguishing description of
the two liquids appears to undermine the claim that the concepts in question express
natural kinds, or kinds distinguished by a hidden nature, as Putnam's argument requires,
rather than more superficial kinds or ‘motleys’, what Locke called ‘nominal kinds’. This,
however, is a mistake. As Putnam (1975) argued convincingly, whether the twins’ terms
express natural‐kind concepts or more superficial kind concepts is an empirical question
turning on the referential intentions of the speaker. Archimedes knew nothing about the
chemical composition of the substance he called χρυϛoϛ, but if he would have rejected as
a genuine instance of the kind something that possessed only the observable properties of
gold—if he would have agreed that all that glitters is not gold—then χρυϛoϛ for
Archimedes indeed expressed a natural‐kind concept (arguably, the concept gold). The
speaker need have only a vague conception of a shared hidden nature or structure
determining the kind; she need not have any idea what that hidden nature in fact is.
Whether a speaker has such a conception in mind can normally be determined by
investigating the speaker's use of the term under a wide range of circumstances. This
process would include querying the speaker about what she would say in various
counterfactual circumstances. (‘If it looked like gold but in fact was made out of a high‐
tech synthetic would it be gold?’) We can imagine a population which, having no interest
in the hidden nature or structure of objects and substances, intends by their use of such
vocables as ‘water’, ‘gold’, and ‘aluminium’ to refer to things that possess only the
observable properties of water, gold, and aluminium. Their use of these terms would be
Page 4 of 19
Wide Content
much like our use of ‘table’ or ‘chair’, expressing relatively superficial kinds (in this case,
functional kinds) or motleys, rather than natural kinds.2
(p. 354)
It is worth noting, however, that a speaker's (or a community's) referential intentions may
not be determinate enough, in advance of theory development and scientific discovery, to
determine whether a particular sample falls within the extension of a given concept. The
notion of hidden nature or structure that a speaker has in mind and that governs her use
of a concept may be too vague to settle whether objects with similar microstructures (say,
isotopes of the same element) would fall under the speaker's concept. But indeterminacy
in the concept's extension does not, as Segal (2000) suggests, turn it into a motley
concept.3 , 4
The externalist interpretation of Burge's thought experiment has also been challenged.
Crane (1991) and Segal (2000) have argued that a single content ascription is true of Alf
and Twin Alf, one that reflects what we might call their personal conception of how things
are. Alf's concept, on this view, is idiosyncratic, relative to his linguistic community; it is
the same concept as Twin Alf's. (Twin Alf is assumed to be a linguistically competent
member of his own linguistic community.) We might call this shared, narrow, concept
‘narthritis’, to distinguish it from arthritis, the concept that applies only to the joints.
Burge's assumption that the community's concept should be attributed to Alf despite his
partial or inadequate understanding of that concept is simply wrong, on this view. Rather
than having a false belief that he has arthritis in his thigh, Alf believes, truly, that he has
narthritis in his thigh, that the narthritis in his ankles is getting worse, that aspirin
relieves the symptoms of narthritis, and so on. He also has the false metalinguistic belief
that the vocable ‘arthritis’, in the mouths of his cohort, expresses the concept narthritis.
As noted above, whether a speaker's use of a term expresses a natural‐kind concept can
normally be determined by investigating the speaker's behaviour under a wide range of
circumstances. It is not clear that the present dispute can be settled in quite as
straightforward a fashion. Burge cites Alf's ‘post‐correction’ behaviour—specifically the
fact that when informed by his doctor that arthritis is a disease of the joints Alf regards
himself as having been mistaken about arthritis—as evidence that his concept, all along,
is the same as the community's. But an internalist interpretation of Alf's post‐correction
behaviour is just as plausible: Alf does intend to deploy in speech and thought the same
concepts as his cohort, but his deference to the doctor should be construed not as
evidence that prior to correction he did in fact deploy the community's concept, but
rather as a change in concept, bringing his own usage into line with his cohort. In any
event, the situation seems quite unlike the natural‐kind case. It is not clear that anything
in Alf's behaviour, or dispositions to behaviour (including his disposition to retract his
claim about his thigh if experts tell him that one can't have that sort of ailment in the
thigh) is sufficient to determine whether, prior to correction, the belief that Alf would
express by his use of the vocable ‘arthritis is painful’ has the same content as it would
Page 5 of 19
Wide Content
have in the mouths of his cohort, or whether it has some idiosyncratic content reflecting
his personal conception of things.
(p. 355)
Some authors have argued that the thought‐experiments do not force a choice between
wide and narrow content; in some circumstances it is correct to ascribe both kinds of
content, each answering to different practical and theoretical interests. According to Loar
(1988), Alf and his linguistic cohort can be described as believing that arthritis is painful;
they share what Loar calls ‘social content’, which is sensitive to environmental and social
factors and is therefore wide. Social content is captured by the content‐ascribing that‐
clauses that we use to attribute propositional attitudes. But content‐ascribing that‐clauses
do not correspond, one‐to‐one, to belief contents, as these are individuated in common‐
sense psychological explanation. So it does not follow that Alf and his cohort have the
same belief contents as these are individuated by common‐sense psychology.
Loar cites a number of examples in support of his claim. He asks us to consider a diary
which we know was written by either an earthling or a Twin Earthling, although we do
not know which. One entry says: ‘No swimming today; the water is too rough’. We would
explain the diarist's behaviour as follows: because she believes that the water is rough,
and that if the water is rough one should not swim, she believes that one should not
swim. If the explanation is challenged on the grounds that we don't know whether she
has any water beliefs at all, then, according to Loar, we would likely produce a
paraphrase of the original explanation that makes no commitment about the nature of the
liquid in question. We might say, for example, ‘She believes that the local sample of the
liquid that fills the oceans and lakes is rough’. In other words, we would produce a single
explanation that subsumes the beliefs of both earthlings and Twin Earthlings, rather than
two explanations, one adverting (p. 356) to water and one to XYZ. On Loar's view, the
original explanation attributes to the diarist only a certain way of conceiving of things; in
particular, it attributes mental states that are interrelated (and related to stimuli and
Page 6 of 19
Wide Content
Noam Chomsky (1992, 1995) has been highly critical of the methodology used to
establish the existence of wide content. The externalist thought‐experiments, he claims,
make illegitimate appeal to speakers' intuitions about such notions as reference,
extension, and true of, technical notions about which speakers can have no theory‐neutral
intuitions (1992: 225; 1995: 42). Moreover, speakers' intuitions about such things as
whether a liquid superficially identical to water counts as the same liquid are highly
sensitive to contextual factors. Change the circumstances or pragmatic presuppositions
of the Twin Earth story, he suggests, and judgements will change accordingly. Chomsky
doubts that ‘any general or useful sense can be given to such technical notions as “wide
content” (or any other notion fixing “reference”) in any of the externalist
interpretations’ (1992: 225).
Chomsky is surely right to point out the lability of the intuitions upon which the
externalist interpretation of the thought‐experiments depends, though his rejection of the
whole externalist programme is an extreme and unwarranted reaction. Care must be
taken in attributing beliefs to subjects simply on the basis of the conventional meanings
of the words they utter.8 The seemingly innocuous assumption appealed to earlier—that a
subject's utterances express her thoughts—does not by itself support the attribution of
wide content to the subject's mental states.9 Moreover, we have seen from the discussion
of Putnam's Archimedes example that empirical investigation of a subject's behaviour in a
wide range of circumstances is sometimes necessary to determine whether a subject has
a natural‐kind concept in mind. The investigation will probe just how a subject's
judgements are sensitive to contextual factors. But the context‐sensitivity of our intuitions
and judgements about such matters does not itself undermine the existence of wide
content or render the methodology deployed in the thought‐experiments problematic.
Page 7 of 19
Wide Content
Various authors have expressed doubts that wide content is well suited to serve the goals
of psychological explanation. Barbara Von Eckart questions the legitimacy of at least
some types of wide content:
How could certain subtle features of the linguistic practices of persons around you
influence the content of your mental representations when those features do not
directly impinge on you in any way?
(1993: 264)10
The argument presupposes that only features that causally affect the subject could be relevant to
mental content. By hypothesis, differences in the environmental or social context of twins make
no difference to their surface stimulations, and so, the reasoning goes, such features could affect
their inner states only by some sort of ‘action at a distance’ (see, for example, Block 1986: 625;
Fodor 1987: 40). At very least, it is claimed, such causal connections as do hold between the
environment or social context and intentional agents do not seem to be of the right sort to affect
their mental states. But, as Burge and others have noted, the presupposition is false. Contextual
or relational properties can affect the type‐individuation of an object or state without affecting it
causally. The birth of a child can make a woman an aunt without the two ever coming into causal
contact.
While there is nothing metaphysically suspect about wide content, the underlying worry
is that it is nonetheless ill suited to play an explanatory role in psychology. Proponents of
narrow content have argued that there is no distinctive explanatory work for wide
content to do. Segal (1989) is concerned explicitly with the content attributed in
computational theories of cognitive capacities. He argues that David Marr's theory (1982)
of early visual processing attributes narrow content to the states and structures it posits.
The argument depends upon general considerations that, if true, would apply to any
account of cognitive mechanisms.11 Moreover, the intuition underlying the argument is
shared by many who think that, whatever the differences in their physical or social
environments, twins are psychologically identical and should be treated as such by any
theory interested in explaining their behaviour or cognitive capacities.
Segal argues that unless subjects differ in their discriminative and recognitional abilities
there is no motivation for positing different contents. Describing a Twin Earth scenario
where distinct environmental conditions, C1 on earth and C2 on Twin Earth, cause
identical internal states, non‐intentionally described, to be tokened in twin subjects, he
says:
Page 8 of 19
Wide Content
ability that could be detected by any tests of the sort employed by the theorist of
vision. For the twins are twins, and will be the same in every testable respect.
(1989: 205)
He goes on to say that positing distinct contents would violate Ockham's razor:
There would just be no point in invoking the two contents, where one would do.
For there would be no theoretical purpose served by distinguishing between the
contents.
(1989: 206)
There is, in fact, an important theoretical purpose served by the ascription of wide content in
theories of cognitive mechanisms. Spelling this out will require a little stage‐setting. I will then
consider whether the point carries over to common‐sense psychological explanation.
The semantic interpretation of a computational mechanism specifies which properties are
tracked by the posited structures when the mechanism is functioning properly in its
normal environment. An interpretation of a computational system is given by an
interpretation function that specifies a mapping between equivalence classes of physical
states of the system—in so‐called ‘classical’ devices these will be symbols or data
structures—and elements of some represented domain. To interpret a device as a visual
system is to specify a mapping between states of the device and tokenings of visible
properties such as changes of depth in the scene. To interpret a device as an adder is to
specify a mapping between states of the device and numbers. The specified states of the
device are thus construed, under the interpretation, as representations of changes in
depth or of addends and sums.
A given computational mechanism would not enhance fitness in every environment. Being
an adaptation is a contingent property of a computationally characterized system. We can
imagine a duplicate of Visua, Twin Visua, appearing spontaneously (perhaps by random
mutation) in a world to which it is not adapted. Imagine that the counterfactual world,
E2, is different enough from the actual world that Twin Visua's states track some distal
property other than changes in depth. Call this property ‘C2’. Since Visua and Twin Visua
are physical duplicates, the two mechanisms have the same discriminative and
recognitional abilities. Visua would track C2 if it were transported to E2. Twin Visua will
contribute to the fitness of the (p. 359) organism containing it only if C2 is a useful
property to track or represent in E2. C2 is some function of surfaces, local light, and local
Page 9 of 19
Wide Content
optical laws, but tracking C2 might not allow the organism containing Twin Visua to
recover what is where in the scene. If it does not, we might wonder whether it is even
appropriate to call Twin Visua a visual mechanism.13
The important question for present purposes is ‘What do Visua's and Twin Visua's internal
states represent?’. It is natural to say that Visua represents C1—changes of depth. It is in
virtue of tracking changes in depth in the scene that Visua contributes to the organism's
successful interaction with its environment. Perhaps it is also natural to say that Twin
Visua represents the distinct property C2. In any case, it would be odd to say that Visua
represents some more general property C3 that subsumes both changes of depth (C1)
and the strange and (from the terrestrial perspective) unnatural property C2.14 In other
words, there is a clear rationale for attributing to Visua and Twin Visua distinct,
environment‐specific, wide contents that make apparent the contribution of the
mechanism to the success of the organism in its normal environment, rather than an
unperspicuous narrow content that does not.
The cognitive theorist sets out to answer questions that are already framed in
environment‐specific terms. If the mechanism's states are interpreted as representing
depth and orientation, rather than more general properties determined by correlations
that obtain in counterfactual environments, then the theory is poised to answer these
questions.
(p. 360)
Page 10 of 19
Wide Content
Let us return to Visua and Twin Visua. Suppose now that the property tracked by Twin
Visua in E2—C2—is a useful property for an organism in E2 to detect. Twin Visua
therefore contributes to the fitness of the organism containing it. Imagine that an
enthusiastic editor on earth (E1), always on the lookout for new markets, asks the
theorist responsible for characterizing Visua to produce a textbook that could be
marketed and sold in both E1 and E2. Since the formal component of the theory that
specifies Visua's basic causal operations will characterize Twin Visua's as well—Visua and
Twin Visua are formally identical mechanisms—the theorist needs only to produce a
single semantic interpretation that specifies what this formally characterized mechanism
will represent in E1 and E2. Since the mechanism does not track C1 in E2 or C2 in E1,
neither C1 nor C2 is a plausible candidate. Rather, an interpretation appropriate to both
worlds would take the mechanism to represent the more general property C3 that
subsumes both C1 and C2.17
Notice, first, that the content C3 is a wide content. The new semantic interpretation
specifies what the mechanism represents in E1 and E2, but not what a physically
indistinguishable mechanism might represent in some third environment E3. (This follows
by an iteration of the reasoning above.) While nonetheless wide, C3 is, in a sense,
narrower than either C1 or C2; C3 prescinds from the environmental differences between
E1 and E2. The explanatory interests served by the new interpretation are less local, less
parochial, than those served by the original interpretation, which was designed to
address questions posed in vocabulary appropriate to earth. Whereas the old
interpretation answered such pretheoretic questions as ‘How is the organism able to
recover the depth of objects in the scene’ by positing representations of depth, the new
interpretation provides the basis for answering this question and an analogous question
framed in vocabulary appropriate to E2—‘How is the organism able to recover
information about C2?’—by positing representations of the more general (p. 361) property
C3, and supplying auxiliary assumptions about how C3 is related to the locally
instantiated properties C1 and C2.
As it happened, the overeager editor was somewhat surprised that sales of the new
interplanetary textbook on earth fell off rather sharply from the first edition, designed
solely for the local market. Besides introducing a new vocabulary containing such
unfamiliar predicates as ‘C3’, the new edition required cumbersome appendices
appropriate to each world, explaining how to recover answers to questions about the
Page 11 of 19
Wide Content
organism's performance in its local environment, questions that motivated the search for
an explanatory theory in the first place. Readers complained that the new edition was
much less ‘user‐friendly’.
The editor was therefore dissuaded from his original idea of commissioning an
intergalactic version of the text, which would provide a genuinely narrow semantic
interpretation that would specify what Visua would represent in any environment.18 He
came to realize that a semantic interpretation of the basic causal processes of a
mechanism is primarily a gloss that allows a theory to address local explanatory interests.
Any gloss that shows that the theory is doing its job will be couched in a vocabulary that
is perspicuous for the local audience with these interests. The moral is that a truly
intergalactic cognitive science would not be representational.
Beliefs and desires must be construed as causally efficacious states of subjects to play the
predictive and explanatory roles required of them in our folk practices. Whereas
computational theories characterize the states they posit in both formal and semantic
terms, folk psychology characterizes propositional attitudes only by their contents.19
Unlike the states involved in the highly modularized, informationally encapsulated
processes characterized by our most promising computational theories, propositional
attitudes have very complex functional roles which include, typically, accessibility to
consciousness. We must rely on content ascriptions to ‘get at’ the complex functional
roles in virtue of which propositional attitudes figure in predictions and explanations of
behaviour. Normally content serves this purpose quite well. The state that is typically
caused by looking out of a window on a rainy day, and that typically causes (in
conjunction with other propositional attitudes) umbrella‐carrying behaviour, can be
referred to as the belief that it is raining. Identifying this state by its content allows us to
infer quite a bit about how it (p. 362) will interact with environmental conditions and
other internal states to produce additional mental states and behaviour.
A moral of Loar (1988) is that content‐ascribing that‐clauses are not perfectly suited to
serve the purposes of psychological explanation. Consider the diarist again. According to
Loar, we would explain the diarist's behaviour by producing a single explanation that
subsumes the beliefs of the diarist and her twin. In this context we don't really care what
kind of liquid the subject is thinking about—information that is typically made available
by a content‐ascribing that‐clause, such as that the water is too rough. What is relevant
for predicting and explaining the diarist's behaviour is the causal role of her belief, its
pattern of actual and potential interaction with stimuli, behaviour, and other mental
states. This narrowly construed causal role, Loar claims, determines a kind of content
Page 12 of 19
Wide Content
that captures the way that the subject and her twin conceive of things, considered
independently of their environments.
The internalist intuition that the twins are identical in psychologically relevant respects is
compelling. But respecting this intuition does not require that they share some narrow
content; it requires only that they share some psychological states. As Loar himself notes
(1988: 105), a ‘more minimal’ lesson to be drawn from the examples is that the predictive
and explanatory schema of common‐sense psychology individuate propositional attitudes
narrowly, by their (narrowly conceived) causal roles, but not necessarily that causal role
determines a second kind of content. Since we don't have a way of specifying the shared
psychological states of the diarist and her twin directly, in ‘intrinsic’ terms that spell out
precisely the psychologically relevant causal potential of the states, we must find a
content ascription that applies to both the diarist and her twin. Such an ascription is not
hard to find: the diarist refrained from swimming because she believed that the ocean
(lake, river, or whatever) was too rough. But there is no reason to think that the content
so attributed is narrow. It is, in a sense, narrower; the content ascription prescinds from
the environment‐specific information supplied by the that‐clause that the water is too
rough, so it attributes a more minimal environmental commitment. But it does not
abstract away entirely from the subject's external surroundings, as it would if it were
genuinely narrow. And, just as in the computational case discussed above, it is unlikely
that contents that applied intergalactically—to subjects in any environment whatsoever—
would serve the complex role required of content. Folk thought and talk about intentional
agents are couched in terms that presuppose the local environments in which they are
situated. We are not interested simply in the bodily movements of subjects considered in
isolation from the objects and properties that make up their world. Wide‐content glosses
on the (narrowly individuated) internal states that cause intentional behaviour allow
environmentally situated creatures such as ourselves to understand these states in terms
of their publicly available normal causes and effects.
The diarist example indicates that wide content can be too fine‐grained to capture what is
relevant for psychological explanation. A second example from Loar (1988) suggests that
it can be too coarse‐grained. Imagine that, some months before we encounter him in
Page 13 of 19
Wide Content
Burge's story, Alf visits Paris and hears of a disease called ‘arthrite’ that affects the joints.
Not realizing that ‘arthritis’ and ‘arthrite’ are intertranslatable, and being a bit of a
hypochondriac, Alf comes to believe that he has two problems with his ankles, one called
‘arthritis’, which he later comes to believe has spread to his thigh, and a second ailment
called ‘arthrite’. Loar claims that Alf has two beliefs with distinct causal powers. Alf's
‘French’ belief that he has arthritis in his ankles does not interact in the appropriate way
with his ‘English’ belief that if he has arthritis in his ankles he should take aspirin,
although his ‘English’ belief that he has arthritis in his ankles does. There are two distinct
states with different interactive potential, but only one belief ascription—the belief that
he has arthritis in his ankles—available to pick out the distinct states. Loar concludes that
Alf's ‘arthritis’ and ‘arthrite’ beliefs have distinct ‘psychological’ contents, determined by
their distinct causal roles.
Once again, a more minimal conclusion is preferable. Alf does seem to have two distinct
states with different causal roles. It is not hard to find distinct content‐ascribing that‐
clauses that characterize the difference in the causal roles and hence serve to pick out
the underlying states. We might say ‘He thinks that the original problem with his ankles
can also occur in muscles and bones, but that the new disorder affects only the joints’.
There is no reason to think that the content ascribed thereby is narrow. We can say, with
Loar, that the distinct content ascriptions capture ‘the way the subject himself conceives
of things’ (1988: 108). The subject conceives of himself as having two distinct problems
with his ankles. But ‘the way the subject conceives of things’ is not a name for narrow
content.
As several authors have noted (see, for example, Field 1978; Loar 1981, 1988; Schiffer
1981), we make use of the fact that people's beliefs are often reliable indicators of the
way the world is. As Field (1978) puts it:
Page 14 of 19
Wide Content
[W]e are constantly using our opinions about other people's beliefs in forming
opinions about the world. The fact that a child believes that he has done
something I won't like (a fact that can often be inferred from his behaviour) gives
good reason to think he has done something I won't like; the fact that most
physicists believe that there are gravitational waves … is good reason for me to
believe that there are gravitational waves; and so forth.
(1978: 103)
On occasion, when the point of a belief ascription is to convey information, a speaker might
ascribe a content that she knows does not capture the way the subject conceives of things. One
might say, for example, that Perry White, the editor of the Daily Planet, thinks that Superman
has real literary talent. White, of course, does not know that his star reporter, Clark Kent, is
Superman, but the belief ascription serves its purpose—conveying information about Clark Kent/
Superman—as long as the audience knows that Clark Kent is Superman.
Burge (1978) identifies another use of content ascribing that‐clauses. They constitute a
standard by reference to which a subject's grasp of the commitments of public discourse
may be evaluated, and her intellectual responsibilities with respect to that discourse
assessed. Thus, we can hold an individual answerable to the community standard, and
subjects can hold themselves so answerable, even when they have only incompletely
mastered the elements within the standard. Alf, in Burge's original thought‐experiment,
can take himself to have been wrong about arthritis, and to have been corrected by his
doctor.
Notice that these two uses of content presuppose local interests. The information
conveyed by a belief attribution is often about the local environment. In the above case, it
is about Superman, not his twin in some other world. A subject's competence (p. 365) will
be evaluated by the standards of the local linguistic community. When the goal of content
ascription is the prediction and explanation of behaviour, I have argued, content ascribing
that‐clauses are used to pick out the causal roles of our beliefs and desires, properties
shared by twins. All of these purposes are served by wide content. We are sometimes
interested explicitly in how the subject would behave if the world were different in
various ways, and the relevant that‐clauses may prescind from some features of the
environment, as they do in the diarist case. So content can be wide in varying degrees.
Page 15 of 19
Wide Content
Genuinely narrow content, content that supervenes on the intrinsic states of the subject
and so would apply in any environment, is an idealization along this dimension. But it is
too general and too abstract to bear an explanatory burden.22
References
Bach, K. (1987), Thought and Reference (Oxford: Oxford University Press).
Brown, J. (1998), ‘Natural Kind Terms and Recognitional Capacities’, Mind, 107: 275–303.
Burge, T. (1978), ‘Individualism and the Mental’, Midwest Studies in Philosophy, 4: 73–
121.
Crane, T. (1991), ‘All the Difference in the World’, Philosophical Quarterly, 41: 1–26.
—— (1999), ‘In Defense of Narrow Mindedness’, Mind and Language, 14: 177–94.
—— (1988), ‘Social Content and Psychological Content’, in R. Grimm and D. Merrill (eds.),
Contents of Thought (Tucson, Ariz.: University of Arizona Press), 99–110.
Page 16 of 19
Wide Content
Putnam, H. (1975), ‘The Meaning of “Meaning”’, in Putnam, Mind, Language, and Reality:
Philosophical Papers, ii (Cambridge: Cambridge University Press), 215–71.
Segal, G. (1989), ‘On Seeing What Is Not There’, Philosophical Review, 98: 189–214.
—— (2000), A Slim Book about Narrow Content (Cambridge, Mass.: MIT Press).
Von Eckart, B. (1993), What Is Cognitive Science? (Cambridge, Mass.: MIT Press).
Notes:
(1) As many commentators have noted, we must ignore the fact that humans are largely
composed of water.
(2) These people would have little interest in basic science, which, of course, seeks to
explain the observable properties of things by appeal to underlying nature or structure.
We might imagine them to be served by a distinct population responsible for developing
and maintaining the useful objects that keep the society running. This group would speak
a different language, with genuine natural‐kind terms.
(3) It is likely that the shapes of most of our natural‐kind concepts are in flux, as
empirical discoveries and developments in theory force a revision of their extensions.
(4) For further discussion of the issues in the last two paragraphs see Brown (1998).
(5) See the discussion in Brown (2003) of Segal's suggestion (2000: 77) that Alf's
concepts have undergone a ‘natural evolution’.
(6) A radically holist view of concept individuation, according to which any change of
belief involving the vocable ‘arthritis’ would count as a change in concept, is, of course,
highly controversial. But the view under consideration here is not radically holist. It is not
implausible to think that a subject who believes that ‘arthritis’ refers to a disease of the
joints and the bones has a different concept.
Page 17 of 19
Wide Content
(8) This is one moral of Loar (1988), but the point is also made by Bach (1987), Patterson
(1990), and Crane (1991).
(9) Chomsky himself is sceptical about the notion of a ‘shared public language’ and so
would reject the idea that the subject's utterances have wide meaning (see Chomsky
1995).
(10) Von Eckart raises the worry, but it is not clear that she actually endorses the
sentiment expressed in the quote.
(11) I argue in Egan (1995) that Marr's theory attributes wide contents to the posited
computational structures. Nothing in the present discussion will turn on the
interpretation of Marr's theory.
(13) If Twin Visua is not a visual mechanism, then of course Visua is a visual mechanism
only contingently. I argue in Egan (1999) that computationally characterized mechanisms
subserve cognitive functions such as vision or parsing only contingently.
(14) An internalist might claim that Twin Visua represents C1, but this seems
unmotivated, since Twin Visua's states may never co‐vary with C1 in E2. E2 is not a world
where Twin Visua (or Visua, for that matter) sees depth.
(15) See Egan (1999) for the argument that computational characterization is
individualist or narrow.
(17) C3 may be understood as the disjunction of C1 and C2, or as a determinable that has
C1 and C2 as determinates.
(21) The formal component of computational theories, rather than the content
component, specifies the relevant functional architecture of the system.
Page 18 of 19
Wide Content
(22) Thanks to Brian McLaughlin and Robert Matthews for helpful comments on an
earlier version of this paper.
Frances Egan
Page 19 of 19
Narrow Content
Narrow Content
Gabriel Segal
The Oxford Handbook of Philosophy of Mind
Edited by Ansgar Beckermann, Brian P. McLaughlin, and Sven Walter
The notion of narrow content arises from Hilary Putnam's well-known article ‘The Mean
ing of “Meaning” ‘. Putnam raised the question of whether the meaning of a word in a giv
en subject's mouth is fixed by the subject's psychological states in (what he termed) ‘the
narrow sense’. According to Putnam, a psychological state is narrow if a subject's being
in that state does not presuppose the existence of anything outside the subject. The idea
is that a narrow psychological state is intrinsic to the subject: it does not in any essential
way require the subject to stand in any particular relations to anything in her environ
ment.
Keywords: narrow content, Hilary Putnam, meaning of a word, psychological states, subject, narrow sense
To address his question about meaning Putnam introduced his famous Twin Earth experi
ment, which I will briefly recapitulate. Putnam invited us to conceive of a planet that is
exactly like earth in 1750, except that the transparent, odourless, tasteless, wet stuff that
fills oceans and rains from the sky is not H2O but has a different chemical composition
that we can call ‘XYZ’. We imagine a typical 1750 Earth subject, Oscar. And we imagine
that he has a ‘twin’ on Twin Earth, an exact duplicate in all intrinsic respects, a molecule‐
Page 1 of 15
Narrow Content
for‐molecule replica. We now consider whether the meaning of the word ‘water’ is the
same in the two Oscars' vocabularies.
(p. 368)
Putnam argued that it is not. When Oscar says ‘water’ he is referring to water; that is,
H2O. When Twin Oscar says ‘water’ he is referring to twin water (‘twater’), which is not
H2O, hence not water. And this difference of reference entails a difference in meaning. By
hypothesis, the twins' narrow psychological states are the same. So meaning is not deter
mined by narrow psychological states.
The Oscar example was designed to show that the meaning of natural‐kind terms, such as
‘water’, depends in part on the underlying nature of samples that the term is used to talk
about. Putnam then provided a second Twin Earth thought experiment designed to show
that the meaning of words can also depend on social factors in a subject's environment.
Putnam can't tell the difference between elms and beeches. He knows of nothing that dis
tinguishes the two, bar their names. We imagine a twin Putnam on a twin Earth that is
just like Earth except that there elms are called ‘beeches’ and beeches are called ‘elms’.
When Putnam says ‘elm’ he means elm. When Twin Putnam says ‘elm’ he means beech.
So what the subjects mean by ‘elm’ is not fixed by their narrow psychological states, but
depends in part on how other people in their environments use their words.
Putnam summed up his conclusion with a slogan: cut the pie any way you like, meanings
just ain't in the head. Catchy as the slogan is, it is not particularly accurate. What the
thought experiments are supposed to show is that meanings are partly individuated by re
lational factors. They may be compared to sunburn, as Donald Davidson (1987) points
out. Subjects cannot be sunburned unless they have been affected by the sun. But it
wouldn't be appropriate to conclude that sunburn is not a condition of the skin.
Putnam himself did not extend his line of reasoning to psychological states themselves.
But others did (e.g. Burge. 1982; McGinn 1982). Consider the belief that Oscar expresses
when he says, for example, ‘It is good to drink cold water on a hot day’. What he believes
is that it is good to drink cold water on a hot day. When Twin Oscar utters the same word
forms he is not expressing a belief about water, a substance he has never encountered or
heard about. Rather, he is expressing the belief that it's good to drink cold twater on a
hot day.
Page 2 of 15
Narrow Content
Alf in all intrinsic respects. According to Burge, Twin Alf does not worry that he has
arthritis in his thigh. When Twin Alf says ‘arthritis’ his word does not express the concept
of arthritis. Rather, it expresses a concept that applies to the wider group of ailments.
The content of (p. 369) the twins’ ‘arthritis’ concepts and the beliefs they express when
they use the term ‘arthritis’ is partly determined by their social environment.
To make the idea vivid, consider a simple thought experiment involving a singular con
cept. John and Mary are friends. When John says ‘Mary likes thought‐experiments’ his
concept refers to Mary and what he says is true if and only if Mary likes thought experi
ments. On Twin Earth, Twin John and Twin Mary are friends. When Twin John says ‘Mary
likes thought experiments’ he expresses a concept that refers to Twin Mary and what he
says is true iff she likes thought experiments.
There are three basic positions that one might hold in respect of broad and narrow con
tent. Externalism is the view that at least some concepts have broad content and do not
also have a narrow content. Two‐factor theories hold that all concepts have narrow con
tent, and at least some have broad content as well. Thus, for example, Oscar's and Twin
Oscar's ‘water’ concepts have different referential content, which is broad, but also have
a shared narrow content. (Two‐factor theories are sometimes classified as internalist, and
I shall adopt that terminological convention for the sake of convenience.) Radical inter
nalism is the view that all concepts have narrow content and none has broad content.
Each of the three basic positions has a number of variants. Thus, externalists divide over
the question of whether the only kind of broad content is referential. Nathan Salmon
(1986) and Scott Soames (1987) have defended the claim that singular concepts have on
ly referential content. And Jerry Fodor (1994) (having previously argued for a two‐factor
theory: Fodor 1987) has applied a version of the view to both singular and general con
cepts, the latter being held to refer to properties. John McDowell (1977) and Gareth
Page 3 of 15
Narrow Content
Evans (1982) have defended the view that Fregean sense is often broad. And Burge (1979)
has argued that the Fregean sense of (at least (p. 370) some) general concepts (such as
arthritis) is broad. Internalists also divide over what kind of thing narrow content is.
For the remainder of this chapter I will be concerned with arguments for narrow content
(Section 21.2) and theories of narrow content (Section 21.3).
The following argument is adapted from Fodor (1987). Arguments about the nature of
content should be constrained by the role that content plays or would play in scientific
psychology. Taxonomy in science always respects causal powers, in the sense that items
with the same causal powers are classified together. The psychological states of twins
have the same causal powers. Therefore, they would be classified together. Therefore
they would be ascribed the same content. So content is narrow.
There is something immediately suspicious about that argument. For it appears to entail
that scientific taxonomies never recognize relational properties. Thus, any two items that
are intrinsically exactly similar have the same causal powers. So any two items that are
intrinsically exactly similar should be classified together, even if they differ in their rela
tional properties. But plenty of scientific taxonomy recognizes relational properties. Tax
onomy by species in biology, for example, is sensitive to a creature's ancestry. If my
neighbour's cat has a twin on a Twin Earth, that twin is not a cat, even though the two an
imals have identical causal powers. If the planet Mars has an exact duplicate floating in a
void somewhere far, far away, that object is not a planet, because, by definition, planets
have to be gravitationally bound to a star.
Page 4 of 15
Narrow Content
you are a planet affects your causal powers, which is all that individualism [i.e. internal
ism] asks for’ (Fodor 1987: 43).
(p. 371)
But when do relational properties affect causal powers? Fodor returned to this question
in Fodor (1991). At that point he thought that just about any contingent relational proper
ty could affect causal powers because you could build a detector for it. Consider Fodor's
example of the properties of being an H‐particle and being a T‐particle. Something is an
H‐particle at a time t, iff it is a particle and Fodor's dime is heads up at t. Something is a
T‐particle at time t iff it is a particle and Fodor's dime is tails up at t. One wouldn't initial
ly think that being an H‐particle affects a thing's causal powers. But according to Fodor
(1991) it does. We could build a machine that detects H‐particles. An H‐particle would
cause the machine to register ‘H’ while a T‐particle would not.
Analogously one could build a detector that could detect water thoughts: a water thought
would cause it to register W, say, and a twater thought would not. So, Fodor argued, the
question that is relevant to internalism is thus not ‘Do the twins' thoughts have different
causal powers?’, but rather ‘Do they have different causal powers in respect of the effects
that matter to psychology?’.
Now ‘water’ thoughts and ‘twater’ thoughts do have different effects. For example, if
someone believes that it's good to drink cool water on a hot day and that the 25th of July
will be a hot day, then, if they consider the matter, these beliefs will cause them to believe
that it will be good to drink cool water on the 25th of July. If someone has the correspond
ing beliefs about twater, then they will be caused to believe that it will be good to drink
cool twater on the 25th of July. Similarly, if someone wants to drink some water this might
cause them to try to find some water, whereas if someone wants to drink some twater this
might cause them to go looking for some twater.
What we need to know is whether the difference between the effects is due to a differ
ence between the causal powers of water thoughts and twater thoughts or due to some
thing else. By way of comparison: since I have a brother I have the power to have a son
who has the property of being a nephew. My siblingless twin lacks that power. His son
would not be a nephew. But, intuitively, the difference between these effects is not due to
a difference between my and my twin's causal powers.
Reflection on that sort of case led Fodor to propose that a difference between properties
of causes that correlates with a difference between effects is a difference of causal power
Page 5 of 15
Narrow Content
only if it is not conceptually necessary that causes which differ in that way have effects
which differ in that way. Thus, the difference between me and my twin that correlates
with my having a son who is a nephew and him having a son who is not a nephew is not a
difference between our causal powers, because it (p. 372) is conceptually necessary that
someone who has a brother can have a son who is a nephew and someone who lacks a
sibling cannot.2
Fodor then argued that the difference between water thoughts and twater thoughts is like
the difference between me and my siblingless twin. It is conceptually necessary that wa
ter beliefs cause further water beliefs rather than twater beliefs and that water desires
cause water‐seekings rather than twater‐seekings. Hence, the difference doesn't count as
a difference of causal power.
The first objection begins with the observation that the differences among the effects of
water thoughts and twater thoughts are not confined to intentional ones. If Oscar seeks
water and Twin Oscar seeks twater, then Oscar is likely to find water and Twin Oscar is
likely to find twater. Fodor dismisses this sort of case quickly on the grounds that it is ac
counted for just in terms of context. If Oscar were in his twin's context (i.e. on Twin
Earth) then his seeking water would eventuate in his finding twater, and if Twin Oscar
were in Oscar's context then he'd find water.
But even if it is true that if the twins switched contexts they would have the same effects,
it could still be the case that the different effects that the twins have in their different
contexts is due to a difference between the twins' causal powers. To explain this, it will
help to back up a bit. It is in fact not obvious that just about any relational property can
affect causal powers just because one could build a detector for it. Indeed, Fodor no
longer believes this (and he has persuaded me, pace Segal (forthcoming)). What causes
an H‐particle detector to register an ‘H’? First, the particle's being a particle. And sec
ond, Fodor's dime being heads up. And that is it. The particle's being such that Fodor's
dime is heads up is not a causal power of the particle, because it makes no separate con
tribution. The causal story is fully told in terms of the particle's being a particle and the
dime's being heads up.
But it is important to notice that relational properties can affect causal powers. Consider,
for example, a ship's being anchored to the seabed. Clearly, this affects its causal powers:
it can resist being moved by the currents. A twin ship that was not anchored could not do
this. And notice that there is no way of decomposing the property of being anchored into
a conjunction of intrinsic properties of the ship and ship‐independent properties of the
environment. (The example and the point are due to Williamson 1998.)
It may be that being about water is more like being anchored to the sand than like being
an H‐particle. It certainly involves causal relations between a mental state and water, and
thus fails to decompose into a conjunction of intrinsic properties of the mental state and
mental‐state‐independent properties of the environment (see Williamson 1998). Thus, for
Page 6 of 15
Narrow Content
all of Fodor's clever argumentation, it remains possible that differences of broad content
make for differences of causal powers (and ones that are relevant to psychology).
(p. 373)
The second objection to Fodor's argument is that the premise which says that taxonomy
in science respects causal powers may well be false. It seems plausible, for example, that
there is no difference between the causal powers of a cat and those of a twin cat. But
they would still count as members of different biological species. In general, one would
expect some sciences to be exclusively concerned with the causal structure of their do
mains, and one would expect other sciences to have other concerns. Science is not a
monolith.
The third objection goes like this. If we accept the standard view of the Twin Earth exper
iments, then we should accept that folk psychology attributes wide contents. Fodor's view
was that a scientific psychology would have to attribute narrow contents, because of the
‘taxonomy by causal powers’ constraint. But it is not obvious either that there could be a
scientific psychology that respects the constraint or that, if there were such, it would tax
onomize by content. Maybe the only content is broad content, and scientific accounts of
cognition will appeal to, say, neurological properties or non‐intentional computational
ones (as Stephen Stich (1983) argued).3
A second argument for narrow content centres on the existence of empty concepts: con
cepts that lack an actual extension. To explain this argument it will help to bring in a new
example. In the eighteenth century George Stahl (extending a theory of combustion origi
nated by Johann Becher) developed a theory of the calcination and corrosion of metals
which proposed that metals were made of substances called ‘calx’ and ‘phlogiston’. The
theory turned out to be wrong: there are no such substances as calx and phlogiston.
Now let us consider two Twin Earths, T1 and T2, each inhabited by a twin Stahl (T‐Stahl1
and T‐Stahl2). Let us suppose that each of these T‐Stahls proposes a theory that looks just
like Stahl's actual theory and that these theories are true of their worlds. Finally, let's
suppose that different substances play the phlogiston roles on T1 and T2, phlogiston1 and
phlogiston2 respectively. It is reasonable to assume that the difference between phlogis
ton1 and phlogsiton2 is of the sort that a follower of Putnam would take to be a difference
of kind identity, like the difference between XYZ and H2O. So according to the standard
externalist line, ‘phlogiston’ has different meanings on T1 and T2 and expresses different
contents in the mouths of Stahl1 and Stahl2.
The argument proceeds as follows. Stahl's ‘phlogiston’ concept has a cognitive content.
Some radical externalists (such as Ruth Millikan; see Millikan 1984) might deny this. But
the denial flies in the face of good psychology. One simply cannot explain Stahl's sophisti
cated theorizing behaviour without attributing to him a concept of phlogiston. (Indeed
Millikan 2004 concedes that while he doesn't have (p. 374) a concept of phlogiston he does
Page 7 of 15
Narrow Content
have a ‘conception’ of it. I would argue that if a conception is not a concept, then it can
not do the explanatory work it needs to do.)
But then the possibility arises that T‐Stahl1 and T‐Stahl2 have concepts with the same
content as Stahl's actual concept. And, indeed, a psychological theory that attributed the
same cognitive content to all three Stahls' concepts would have an advantage over one
that did not, since it would have greater generality and could account for the cognitive
and behavioural similarities among the twins. Good methodology therefore suggests that
the content of the three Stahl's concepts is the same.4
The third argument for narrow content is due to Brian Loar (1987) (and was developed
and defended in Segal 2000). It is designed to show, pace Burge, that the content of psy
chological states is not individuated by the linguistic environment. Recall Alf, who ap
peared to believe that he had arthritis in the thigh. Suppose that prior to visiting his doc
tor and learning the correct definition of ‘arthritis’ Alf travels to France. There he learns
of a condition called ‘arthrite’, coming to know that it is an inflammation of the joints and
the joints alone. He does not realize that ‘arthrite’ and ‘arthritis’ are synonymous, and
continues to believe that arthritis can affect the thigh.
Given his epistemic state, Alf might be disposed to assert ‘I have arthritis in my thigh’ and
to deny ‘I have arthrite in my thigh’. In order to explain this, we have to suppose that Alf
associates two concepts with different cognitive contents with the two words ‘arthrite’
and ‘arthritis’.5 Yet the concepts' socially individuated Burgean content is the same.6
The functional role of a psychological state is the causal role it plays in its owner's psy
chology, such as might be specified by the Ramsey sentence of some psychological theory
that talked about the state.7 To be a functionalist about some domain of psychological
properties is to hold that properties in that domain can be identified with functional roles.
Some functionalists are externalists who identify broad content with functional role (Har
man 1987). And some internalists are functionalists about narrow content, and hold that
Page 8 of 15
Narrow Content
narrow contents can be identified with functional roles (e.g. Block 1986). We might call
the view ‘narrow functionalism’.
There are two very different forms narrow functionalism might take. The first would fol
low the standard form of a functional reduction. A standard functional reduction of a
range of properties appeals to a theory that already has a vocabulary that specifies the
properties and details a range of causal relations among their possessors. For example, a
functionalist about chemical properties appeals to the theory of chemistry. This theory
has terms for the chemical properties and tells us about how chemicals interact. Func
tionalism then says that chemical properties can be identified with the causal roles as
cribed to the possessors of those properties by chemistry—properties that could be speci
fied, say, by Ramsey's method. Run‐of‐the‐mill narrow functionalism would take the same
form: it would refer to a theory that specifies narrow contents and details a range of
causal relations among their possessors and claim that narrow contents could be identi
fied with the functional roles given by that theory. The problem for run‐of‐the‐mill narrow
functionalism is that we don't seem to have the required theory to appeal to. The second
form of narrow functionalism—named directly reductive narrow functionalism (or
‘DRNF’) in Segal (2000)—attempts to finesse this difficulty.
The idea behind DRNF is that we could abstract an account of narrow functional roles
from a psychology that attributes broad contents. On the face of it, this seems plausible.
Suppose for example that the following are fragments of psychological theories that apply
to 1750 earth and Twin Earth subjects:
If x is thirsty then x will desire to drink some water and if x desires to drink some wa
ter and x believes that there is some water in a glass in his hand then x will try to drink
the water.
If x is thirsty then x will desire to drink some twater and if x desires to drink some twa
ter and x believes that there is some twater in a glass in his hand then x will try to
drink the twater.
Then we might abstract some narrow functional roles by, for example, replacing the con
(p. 376)
The narrow content of, for example, the thought that there is some water in a glass in one's hand
would then be identified with the role of the thought characterized by P2 in the schematic theo
ry.
Segal (2000) made two objections to DRNF, neither of which is persuasive.
The first objection was that there is no way to distinguish aspects of a concept's function
al role that go with its content from those that don't. So, for example, we can imagine
near‐twin subjects who have exactly the same propositional attitudes, but one of whom
has more memory‐buffer space than the other. The functional roles of their concepts will
then differ.
Page 9 of 15
Narrow Content
And indeed there are indefinitely many features of a cognitive system that might affect
the functional role of a concept without affecting its content. In order to specify the en
tire functional role of a concept, one would have to specify its role in every possible cog
nitive system and this would seem to be beyond the scope of any proper psychological
theory.
But the objection is not persuasive because the narrow content of a concept could per
haps be identified not with its total functional role but with some appropriate and natural
sub‐part of that role, such as its role in inference.
The second objection was that a concept's content is what explains its functional role, so
the content cannot be identical with that functional role. But no argument was given for
this claim, and it is highly contentious. It is a common and respectable view that many
theoretical terms are functionally defined by their role in the relevant theory. And there is
no particular reason to suppose that the terms of psychology are not like that.8
So perhaps DRNF is true. But it doesn't help much because we have no vocabulary for as
cribing functional roles and theorising about them.
David Chalmers (2002) has offered an account of narrow content as ‘epistemic intension’.
This account can be seen as the most developed of a family of proposals all of which en
deavour to adapt the traditional notion of an intension to internalist purposes. An inten
sion is a function from possible worlds to extensions. So, for example, the intension of
‘the British Prime Minister in the year 2005’ is (roughly speaking) a function that maps
each possible world on to the particular individual who serves as the British Prime Minis
ter in 2005 in that world (and otherwise on to nothing). The intension of ‘the British
Prime Minister in the year 2005 is (p. 377) male’ is a function that maps each possible
world on to ‘the true’ iff the British Prime Minister in the year 2005 in that world is a
male in that world, otherwise on to ‘the false’. Classical intensions are not narrow. The in
tension of the earth term ‘water’ maps each possible world on to H2O, while the intension
of the Twin Earth term ‘water’ maps each possible world on to XYZ.9 But there have been
various attempts to account for narrow content in terms of some non‐classical kind of in
tension—one that would, for example, assign the same function to ‘water’ and Twin ‘wa
ter’.10
Hesperus is not Phosphorus. There is also a set of epistemically possible worlds that ‘fal
sify’ the thought that Hesperus is not Phosphorus: roughly speaking, those worlds, includ
ing the actual one, in which a single heavenly body appears both in the morning and the
evening and is the referent of your two concepts. Thus, we can see the epistemic inten
sion of a thought as a function from epistemically possible worlds to truth‐values.
In fact, Chalmers's story is a little more complicated than that, involving possible worlds
as observed from particular points in space and time. Entertain the possibility that water
is not H2O. There is a possible world in which some planets are rife with H2O and others
are rife with XYZ. If you were on an H2O planet, then the world would verify the thought
you are entertaining. If you were on an XYZ planet, then that same world would falsify the
thought. So we need a notion of a ‘centred’ world, a world plus a spatio‐temporal location
within it. We can call these centred worlds ‘scenarios’ (roughly following Chalmers's own
terminology).
For each thinker there is a set of scenarios that constitutes their epistemic space: the set
of possibilities that is not ruled out for them a priori. Every a posteriori belief of the
thinker's rules out some possibilities a priori, while leaving others a priori open. The for
mer set of scenarios verify the thought believed, the latter set falsify it. Intuitively, to see
whether a given scenario, S, verifies a thought, T, one considers whether the hypothesis
that S is actual is a priori consistent with T. If it is, then S verifies T, if it is not, then S fal
sifies T.
Chalmers argues that narrow content is epistemic intension. This account has some intu
itive appeal. The narrow content of a thought ought to correspond to (p. 378) how things
are presented by the thought to the subject's subjective point of view. Prima facie, epis
temic intensions appear to be decent candidates to play that role. But the account be
comes less plausible when the technical details are fleshed out a little.
What exactly is it for a scenario to verify or falsify a thought? Chalmers's answer goes as
follows. Every scenario has a canonical description, D. To consider the hypothesis that a
scenario, S, is actual is to consider the hypothesis that D is the case. D consists in an ob
jective description of S along with an indexical description of the thinker's location in S.
An objective description contains only ‘semantically neutral’ terms. Roughly speaking, a
semantically neutral term is one that is not subject to Twin Earth experiments with cor
rect externalist interpretations. Thus, it is one that would mean the same in the lan
guages of any twins, no matter how their environments differed. More precisely, a seman
tically neutral term is one with narrow extension conditions; that is to say, extension con
ditions that are not fixed by anything outside the subject's skin. (According to Chalmers,
most names, natural‐kind terms, indexicals, and terms used deferentially are not semanti
cally neutral.) If D is a canonical description of S, then S verifies T if the hypothesis that
D is the case implies S, and S falsifies T if the hypothesis that D is the case contradicts S.
Page 11 of 15
Narrow Content
ulation that all the descriptions in the sets must be semantically neutral is required if
epistemic intensions are to be narrow.
The problem with this account is that it is implausible that the requisite semantically neu
tral descriptions exist. Consider a modest and trusting subject, Angelica. Angelica has a
thought she would express by saying ‘Hesperus is Phosphorus’. What is the epistemic in
tension of her thought? To a first approximation it is verified by those scenarios satisfying
the description: ‘The heavenly body that first appears in the evening is identical with the
heavenly body that brings light in the morning’. But suppose that Angelica uses ‘heavenly
body’, ‘first’, ‘evening’, ‘identical with’, ‘light’, and ‘morning’ deferentially. Then this de
scription is not semantically neutral. But then what semantically neutral descriptions
might fit the bill? On the face of it, there aren't any.
In fact, Chalmers stipulates the descriptions are formulable in an idealized language. But
there is no guarantee that any idealized language will contain the required descriptions.
Remember that the scenarios at issue are those that are a priori consistent with
Angelica's thought that Hesperus is Phosphorus. It is reasonable to suppose that no set of
semantically neutral descriptions that she herself understands would precisely delimit the
epistemic intension of her thought. The alternative would be that the descriptions are
largely formulable in terms that she does not understand. But we would not expect those
to capture the space of possibilities delimited by Angelica's own thoughts and concepts.11
The problem for this view is that it does not provide a satisfactory account of Putnam's
‘elm’ and ‘beech’ concepts. According to the standard externalist account of the
‘elm’/‘beech’ case, Putnam's ‘elm’ concept extends over elms and elms only, while Twin
Putnam's ‘elm’ concept extends over beeches and beeches only; mutatis mutandis for
their ‘beech’ concepts. If we are to treat this case on the suggested model for ‘water’ we
would have to conclude that both concepts extend over both elms and beeches and any
other elm‐like trees. But that just seems wrong.
It thus looks as though some non‐classical sort of intension would be a better candidate
for narrow content, some kind of function from possible worlds to extensions.
21.4 Conclusion
There are, then, powerful arguments that cognitive content is not individuated in terms of
the natural and social environment in the way that broad content is, according to stan
Page 12 of 15
Narrow Content
dard externalist lines. There is therefore reason to think that cognitive content is narrow.
There is, however, at the time of writing, no satisfactory account of narrow content. It is
possible that narrow content could be identified with functional role. There is however, at
the time of writing, no account of what a psychological theory that attributes functional
roles would be like.
References
Blackburn, S. (1984), Spreading the Word: Groundings in the Philosophy of Language
(Oxford: Oxford University Press).
Block, N. (1986), ‘Advertisement for a Semantics for Psychology’, Midwest Studies in Phi
losophy, 10: 615–78.
Burge, T. (1979), ‘Individualism and the Mental’, Midwest Studies in Philosophy, 4: 73–
121.
—— (1982), ‘Other Bodies’, in A. Woodfield (ed.), Thought and Object: Essays on Inten
tionality (Oxford: Clarendon), 97–120.
Dennett, D. (1982), ‘Beyond Belief’, in A. Woodfield (ed.), Thought and Object: Essays on
Intentionality (Oxford: Clarendon), 1–96.
—— (1990), A Theory of Content and Other Essays (Cambridge, Mass.: MIT Press).
—— (1991), ‘A Modal Argument for Narrow Content’, Journal of Philosophy, 88: 5–26.
—— (1994), ‘The Elm and the Expert (Cambridge, Mass.: MIT Press).
Kripke, S. (1972), ‘Naming and Necessity’, in D. Davidson and G. Harman (eds.), Seman
tics of Natural Language (Reidel: Dordrecht), 253–355.
Page 13 of 15
Narrow Content
Loar, B. (1987), ‘Social Content and Psychological Content’, in R. Grimm and D. Merrill
(eds.), Contents of Thought: Proceedings of the 1985 Oberlin Colloquium in Philosophy,
99–139.
McDowell, J. (1977), ‘The Sense and Reference of a Proper Name’, Mind, 86: 159–85.
McGinn, C. (1982), ‘The Structure of Content’, in A. Woodfield (ed.), Thought and Object:
Essays on Intentionality (Oxford: Clarendon) 207–58.
Segal, G. (1997), ‘Content and Computation: A Critical Notice of Jerry Fodor's The Elm
and the Expert, Mind and Language, 12: 491–501.
—— (2000), A Slim Book about Narrow Content (Cambridge, Mass.: MIT Press).
Stich, S. (1983), From Folk Psychology to Cognitive Science: The Case against Belief
(Cambridge, Mass.: MIT Press).
White, S. (1982), ‘Partial Character and the Language of Thought’, Pacific Philosophical
Quarterly, 63: 647–65.
Williamson, T. (1998), ‘The Broadness of the Mental: Some Logical Considerations’, Philo
sophical Perspectives, 12: 389–410.
Notes:
(1) It is actually very difficult to formulate the notion of narrowness in a satisfactory way.
All kinds of issues come up. For example, what is the modal force of ‘presuppose’ or ‘in
any essential way require’ (or ‘individuated in terms of’ as used below)? What are the
boundaries of the subject? (The skin is not the right boundary, since it imposes a distinc
Page 14 of 15
Narrow Content
tion between say, the subject's heart and her husband that is not wanted in this context.)
Unfortunately, there is no space to discuss these issues here.
(2) Actually I am oversimplifying here: Fodor's condition is more complicated than that.
But the details don't matter for present purposes.
(3) Fodor himself gave up his arguments for narrow content in Fodor (1994), I think
largely because he came to believe that a science that attributed broad contents could
formulate good, predictive generalizations. For a critique of Fodor (1994) see Segal
(1997).
(4) The conclusion of the argument is that the twins' phlogiston concepts have a shared
content. It does not quite follow that the content is narrow. But at least it undermines the
idea that the cognitive content of a natural‐kind concept depends partly on its extension.
(5) That is to say we need to adopt a roughly Fregean criterion of difference for concepts:
If a subject sincerely asserts a sentence, S, and denies a sentence S′, then he expresses
different thoughts by S and S′. And since ‘I have arthritis in my thigh’ and ‘I have arthrite
in my thigh’ differ only in the words ‘arthritis’ and ‘arthrite’, we can infer that Alf asso
ciates different concepts with those particular words. I defend the roughly Fregean crite
rion in Segal (2003).
(6) The conclusion is that cognitive content is not individuated partly by social factors in
the way Burge describes. Again it does not quite follow that cognitive content is narrow.
But again a reason for thinking that it is broad is undermined.
(7) More properly: the functional role of a psychological state that matters in this context.
A psychological state will have other sorts of functional role as well.
(8) See my ‘The Causal Inefficacy of Content’ (forthcoming) for some considerations in
favour of a functional account of cognitive content.
(9) At least according to many, who follow Kripke (1972) in holding that ‘water’ is a rigid
designator.
(10) For proposals in this ballpark see Dennett (1982), White (1982), Fodor (1987), and
Loar (1987).
(11) For more extensive discussion of Chalmers's proposal see Chalmers (2006).
Gabriel Segal
Page 15 of 15
Information‐Theoretic Semantics
Informational semantics takes the primary — at least the original — home of meaning to
be the mind: meaning as the content of thought, desire, and intention. The meaning of
beliefs, desires, and intentions is what it is we believe, desire, and intend. The sounds and
marks of natural language derive their meaning from the communicative intentions of the
agents who deploy them. As a result, the information of chief importance to informational
semantics is that occurring in the transactions between animals and their environments.
So for informational semantics the very existence of thought and, thus, the possibility of
language depends on the capacity of (some) living systems to transform information
(normally supplied by perception) into meaningful (contentful) inner states like thought,
intention, and purpose.
If symbols are the bearers of meaning, the things in the world that have meaning,
information‐theoretic semantics locates the primary source of this meaning in the
relations these symbols bear to the world they are, or purport to be, about. These symbol–
world relations can be described in explicitly information‐theoretic terms (source, signal,
noise, receiver, etc.), as occurs in Dretske (1981), or they can be described in more
general causal terms (Stampe 1977, 1986; Matthen 1988; Fodor 1990; Israel and Perry
1990). However it may be expressed, though, the basic idea is that the primary
Page 1 of 15
Information‐Theoretic Semantics
determinant of a symbol's meaning is those situations in the world about which symbols
carry, or are supposed to carry (more about this important qualification in a moment),
information, those situations in the world of which the symbols are (in normal conditions)
reliable signs. Since, according to some (e.g. Dretske 1986; Israel and Perry 1990), the
information an event carries is what its occurrence indicates, informational semantics is
sometimes referred to as indicator semantics: what a symbol means is what, in certain
central cases, it indicates, or is supposed to indicate, or normally indicates, about other
parts of the world. Theories of this sort are to be contrasted with conceptual‐role,
procedural, or (p. 382) consumer semantics—theories that locate the meaning of a
symbol not (or not only) in upstream causes, but also (and sometimes primarily) in the
downstream effects these symbols have on other symbols and on the behaviour of the
system in which they occur (Block 1986; Papineau 1987; Millikan 1989).
22.1 Information
The word ‘meaning’ is ambiguous. Two of its possible meanings (Grice 1989) are: (1) non‐
natural meaning—the sense in which the English word ‘fire’ stands for or means fire; and
(2) natural meaning—the way in which smoke (not the word ‘smoke’, but smoke itself)
means (indicates, is a sign of) fire. Non‐natural meaning, the kind of meaning theories of
meaning (like information‐theoretic semantics) are supposed to be theories of, has no
necessary connection with truth. The words ‘Jim has the measles’ mean that Jim has the
measles whether or not Jim has the measles. Natural meaning, on the other hand,
requires the existence of the condition meant: if Jim does not have the measles, the red
spots on his face do not mean (indicate) that he has the measles. Perhaps all they mean is
that he has been eating too much candy. Natural meaning, what an event indicates, what
it is a sign of, is a relation between the sign and what it signifies that does not depend on
anyone recognizing or identifying what is meant. It does not depend on conscious beings
at all. Expanding metal (e.g. the mercury in a thermometer) indicates or means that the
temperature is increasing whether or not anyone knows that this is what it means. A
particular cloud formation means a cold front is approaching even if (because no one is
aware of its meaning) it does not mean this to anyone. We found out that this is what it
Page 2 of 15
Information‐Theoretic Semantics
means. Unlike non‐natural meaning, what things indicate, what they mean in a natural
sense, is something we learn about events by patient investigation. We learn what tracks
in the snow (or a cloud chamber) mean; we do not, as we do with words and other
conventional symbols, assign them their meaning.
Thinking of information in terms of natural signs and what they indicate makes it clear
that information depends on a system of stable relations existing between signal and
source. It isn't enough, for instance, that s, an object at a source, always happens to be
moving whenever a particular type of event, e, occurs at the receiver. That isn't enough to
make e indicate that s is moving. No, for e to mean (in the natural sense) that s is moving,
for e to carry this piece of information, e must depend on s's movement in a particularly
reliable way. Circumstances must be such that in these circumstances only s’s movement
will result in e. If, in these conditions, something else can result in e, then the occurrence
of e does not mean that s moved. The ringing of my telephone doesn't indicate, doesn't,
therefore, carry the information, that you are calling me simply because, in point of fact,
you are the only person who ever calls me. What is relevant is not whether anyone else
ever does call me, but whether anyone else might call me—even if only someone dialing a
wrong number or a bothersome telemarketer. If it might be someone else, then the
ringing phone does not indicate that you are calling me. At most it means that it is
probably you.
Page 3 of 15
Information‐Theoretic Semantics
Another way of expressing this relation between source and receiver, the relation on
which the flow of information depends (one I adopted in Dretske 1981 in order to make
clear its connection with the mathematical theory of information and cognitive science in
general), is to say that e, some event at a receiver, carries information about an object, s,
at the source—the information, say, that s is F—only if conditions are such that e raises
the probability of s's being F to 1. It isn't enough to raise the probability to something less
than 1—to 0.99 for instance. That isn't good enough. The reason it isn't good enough is
that the relation in question (whether called information, indication, or natural meaning)
is a transitive relation: if a indicates (means, carries the information) that b, and b
indicates (means, carries the information) that c, then a must indicate (mean, carry the
information) (p. 384) that c. If force on the restraining spring means that an electrical
current is flowing in the circuit, and current flow in the circuit means there is a voltage
difference, then force on the spring indicates a voltage difference. If Sally's expression
indicates she is interested, and interest indicates (at least partial) understanding, then
Sally's expression indicates (at least partial) understanding. This is why no signal can
carry the information that water is freezing without carrying all the information carried
by freezing water—that, for example, the temperature is at or below 32 °F. Setting the
probability (required for the transmission of information) at anything less than 1 would
not preserve this transitivity. It is not (in general) true that if the probability of b, given a,
is (at least) 0.99, and the probability of c, given b, is also (a least) 0.99, then the
probability of c, given a, is also (at least) 0.99. It may be less than 0.99. So if we are going
to use probability at all to express the relations in question, those relations between
source and receiver that enable events at the receiver to carry information about the
source, we must express these as probabilities of 1. Nothing less than 1 maintains the
transitivity of this indicator (natural meaning, informational) relation.
It may be thought that this sets the bar too high. In our messy unpredictable world
probabilities seldom, if ever, reach a value of 1. If information requires a probability of 1,
then we seldom, if ever, get information. This worry betrays a misunderstanding. The
probabilities in question, whether they are probabilities of 1, 0.99, or 0.5, are assigned
against a background of stable circumstances, circumstances that can, but are assumed
not to, change for purposes of fixing what can happen. In normal circumstances the ring
of my doorbell carries information: it indicates or means that someone is at my door. It
does so despite the fact that in different conditions—when there is a short circuit in the
wiring, when there is an infestation by poltergeists, when squirrels have been trained to
push doorbell buttons—a doorbell can be made to ring without anyone (a person) pushing
the doorbell button. But in normal circumstances, circumstances that include the actually
existing wiring and the absence of poltergeists and trained squirrels, there is nothing else
besides someone pressing my doorbell button that can make my doorbell ring. So in these
circumstances the ring indicates that someone is at my door. It carries this information. It
makes the probability of someone's being at my door 1. It makes the probability 1
whether or not we know it is 1. If it doesn't make the probability 1, if there is, unknown to
us, a small probability that my doorbell is ringing when no one is at my door (there is a
Page 4 of 15
Information‐Theoretic Semantics
trained squirrel that ranges freely in the neighbourhood), then the ring does not indicate
that someone is at my door. All it indicates is that someone is probably there. It might be
the squirrel.
Page 5 of 15
Information‐Theoretic Semantics
22.2 Meaning
Semantics is the study of meaning in so far as meaning is understood to be non‐natural
meaning: the kind of meaning in which ‘smoke’ means smoke (not fire); the (p. 385) kind
of meaning in which a prophet can say, and thus mean, that the end is near without the
end actually being near; the kind of meaning in which something in a person's head, for
instance a perceptual experience or a belief, can mean (represent) one line to be longer
than another even though the lines are of equal length. Information, as we have just
described it (i.e. as indication or natural meaning), obviously cannot be equated with
meaning in this (non‐natural) sense. Nothing can carry the information that one line is
longer than another, nothing can (naturally) mean or indicate that this is so, unless, in
fact, the line is longer than the other. So an information‐theoretic semantics, if it is to do
the job, if it is to provide a plausible theory of non‐natural meaning, must supplement or
combine information as this is presently being understood with some other ingredient to
capture the targeted concept.
There is some disagreement about what this additional idea or ingredient might be, but,
whatever it is, it must somehow manage to convert natural signs of F, things that carry
information about the F‐ness of things, into things, symbols, which do not (or need not)
carry such information. The favoured way of doing this is by appealing to the sort of
‘norms’ generated by things having a certain function. The basic idea is modelled on the
way we give certain objects—measuring instruments, for instance—the power to mean, in
a non‐natural way, that something is so‐and‐so by giving these objects the job or function
of ‘telling us’ (carrying the information) that something is so‐and‐so. A speedometer can
say or represent that a car is going 60 mph when it isn't. What gives the instrument the
power to mean, non‐naturally, that the car is going 60 mph is the fact that it has a certain
informational function (a function we—designers, makers, and users—give it): the
function of telling us, providing us with information, about the speed of the car. When it is
doing its job, it tells us how fast the car is going. If, because of mechanical difficulties, the
instrument ceases to do its job, it nonetheless still ‘says’ (represents) that the car is going
a certain speed, but it now misrepresents the speed. It (by pointing at the numeral ‘60’)
says something false. It exhibits a form of non‐natural meaning.
No one, however, gives our brains an information‐carrying function in the way that we
give measuring instruments their information‐carrying functions. So if this is to be a
model for how our brains acquire (in the form of experience and thought) the power to
mean that something is so (whether or not it is so), the power to think that one line is
longer than another when it isn't, the information‐carrying functions must be (what we
may call) natural functions—perhaps biological functions—functions that do not depend
on or derive from us in any way. To smuggle in functions that, in whatever way, depend on
our purposes in the way the functions of speedometers depend on them is to smuggle in
at the very beginning of the analysis the very thing—the kind of meaning associated with
Page 6 of 15
Information‐Theoretic Semantics
Setting aside for the moment questions about the existence of natural functions (I return
to this in the next section), the idea would be to equate meaning or representational
content with an element's information‐carrying function. The functions convert natural
signs into non‐natural symbols. If event type E has tokens, (p. 386) e 1,e 2,… which, in
normal circumstances, are natural signs that s is F, and if E acquires, through an
appropriate history, the function of carrying this information, then tokens of that type
thereafter mean, in a non‐natural way, that s is F even when s isn't F. They mean (non‐
naturally) this because that is what they are supposed to mean (naturally). The norm
invoked by speaking of this as something they are supposed to mean is given by their
information‐carrying function. They are supposed to mean this in the same way hearts are
supposed to pump blood.
Just as a cashier, someone whose job is to operate the cash register, can be asked to
sweep floors, a symbol for F, something whose job it is to indicate F, can be pressed into
kinds of service unrelated to its information‐carrying function. It might appear, for
instance, in a desire for F, a fear of F, or just idle thought about F. Once we have a word
for F in the language of thought, this word can now occur in mental questions (I wonder
whether that is an F), commands (Give me an F), and hopes (I hope that is an F), as well
as assertions (judgements that something is an F).
I spoke above of natural functions, the kind of function a thing has that is independent of
anyone's recognition and/or attribution of that function. Stop signs and speedometers
have functions, but they are not natural functions. As we all know, what stop signs and
speedometers are supposed to do, or what we are supposed to do with them, derives from
the collective purposes, desires, and beliefs of their designers, makers, and users.
Information‐based semantics needs something else. It needs natural functions, functions
that (to avoid circularity) are independent of collective (or individual) purposes and
beliefs. Where are these functions? What is it in nature that gives the visual systems of
animals the function of providing information about optical surroundings? What makes
Page 7 of 15
Information‐Theoretic Semantics
the provision of information something the eyes, ears, and nose are for? Granted they do
supply information; we couldn't survive without it. The question is whether there is any
sense in which they are supposed to supply it.
The presumption is, of course, that the functions have their source in natural selection
and learning. If the heart is supposed to pump blood and the liver is supposed (p. 387) to
clean it, if these are the functions of these bodily organs created by a process of natural
selection, why can't the eyes and ears and associated neural systems also have duties—in
particular, information they are supposed to provide? Or if during operant conditioning, a
pervasive form of learning, internal indicators of those stimulus conditions with which
reinforced behaviour is to be coordinated come to exercise control over behaviour (the
only way this kind of learning can be successful), why isn't it thereafter the job, the
function, the purpose of these indicators (e.g. perceptual states) to provide information
about the stimulus conditions on which the acquired behaviour depends?
It isn't important that the word ‘function’ be used to describe what these organs or
systems develop to do. Some people seem to think that the word ‘function’ (in the sense
of ‘purpose’, not in the sense of ‘causal role’) is only really appropriate when applied to
systems to which human beings, given their special interests, have assigned a purpose.1
According to this view, there is, independent of our interests, nothing the eyes, ears, and
nose are for, nothing that they are supposed to be doing. We needn't, however, quarrel
over the word ‘function’. What is important for the purposes of information‐theoretic
semantics is that there be a set of circumstances, or perhaps a kind of history, that,
independent of human interests, grounds descriptions of animals and their parts as ill,
sick, broken, damaged, injured, diseased, defective, flawed, infected, contaminated, or
malfunctioning. If the truth of these descriptions is independent of our interests and
purposes,2 then there is a way natural systems are supposed to be, or supposed to
behave, that is independent of how we conceive them. There would, therefore, be a
perfectly natural sense in which, given appropriate development, the information‐
processing systems in animals would be able to make mistakes—able, that is, to mean
that something was so when it wasn't. Misrepresentation, the sine qua non of non‐natural
meaning, would thus become possible in a perfectly natural way. It is hard to see why this
much isn't available.
In order to mean, in the non‐natural way we are trying to understand, that s is F, it must
be possible for something to mean this when s isn't F. It doesn't have to be a dog for me to
think and say it is. Symbols (linguistic or mental) that mean dog don't have to be caused
by dogs. They can be caused by wolves or (Fodor's example3) cats‐on‐a‐dark‐night—things
one might mistake for a dog. But anything that means or indicates, anything that carries
the information, anything that is a sign, that s is a dog (p. 388) has to be caused by a dog.4
This is the fundamental difference between natural and non‐natural meaning.
Page 8 of 15
Information‐Theoretic Semantics
This seems to create a problem. How can you convert something that is a sign of F,
something that indicates that F, something that (carrying information about F‐ness) can
only be caused by an F, into something, a symbol for F, that can be caused by non‐Fs?
Fodor (1990: 60) was certainly right that almost all attempts to solve this problem rely on
distinguishing two contexts—one (call it the pure or normal context) in which the symbol
for F is only caused by Fs (thus carrying information about the F‐ness of a source) and
one, call it the impure context, in which it can be caused by a variety of non‐Fs (those, for
instance that look like Fs), thus making misrepresentation (and, therefore, non‐natural
meaning) possible. The symbol type gets its information‐carrying function, and thus its
meaning, in the first context, when it is carrying information. This gives it a meaning that
it takes to the impure context, a context in which, given its acquired function, it now
means F (what it has the function of indicating) without anything necessarily being F.
The problem is to understand how ‘pure’ cases are possible, the conditions in which
something can indicate dog while, at the same time, being capable of being caused (in the
impure condition) by a wolf. If in impure conditions the symbol can be caused by a wolf
(or anything else one might mistake for a dog), then why wouldn't it have been caused by
a wolf in the pure case if, contrary to fact, a wolf had appeared there? If it would have
been caused by a wolf had a wolf appeared in the pure condition, then how, in the normal
or pure case, can the symbol indicate, how can it carry the information, that something is
a dog? All it really seems to indicate, given the constraints on information described
earlier, is a disjunctive (hence the ‘disjunction’ problem) fact, the fact that s is either a
dog or a wolf. If that disjunctive fact is, indeed, the information it really carries, then the
information the symbol acquires the function of carrying must be this disjunctive
information. According to information‐theoretic semantics, then, this disjunctive fact is
what the element comes to (non‐naturally) mean. If that is what it non‐naturally means,
however, then instances of the symbol caused by wolves in the impure condition are not
misrepresentations. The cause—a wolf—actually is what the symbol's meaning says it is;
namely, a wolf or a dog. So we have not achieved what we were after; that is, a symbol,
something that can mean dog when it is caused by, and thus applied to, a wolf.
A lot of ink has been spilt on this problem, and nothing I can say in the space of a few
paragraphs is going to settle matters, but I think it worth emphasizing that natural
meaning, indication, and, therefore, information (of the relevant sort) are all context‐
dependent notions. Something can indicate F in one set of circumstances, (p. 389) not in
another. Certain coloration and markings on the bird indicate that it is a blue jay, but
these signs can (be made to) occur on an object that is not a blue jay. If circumstances are
such that fake jays (non‐blue jays that have the same colour and markings) are a genuine
possibility, then coloration and markings of that sort do not indicate that the bird is a blue
jay. They only indicate something disjunctive in character—that it is either a blue jay or a
fake jay. Readers familiar with recent developments in epistemology will recognize these
points as part of the ‘relevant‐possibility’ theme in theories of knowledge. Recognizing a
blue jay at the bird feeder does not require that one be able to distinguish real blue jays
from fake jays—not if fake jays are not, as they usually aren't, relevant possibilities under
Page 9 of 15
Information‐Theoretic Semantics
normal viewing conditions. That is why knowing there is water (i.e. H2O) in the creek
does not require being able to distinguish H2O from XYZ (a molecularly distinct substance
found on Twin Earth that looks like water). Not if XYZ is not, as it clearly is not if it is
found only on Twin Earth, a relevant possibility. If, on the other hand, XYZ is actually
found in some lakes and streams on earth, then you don't know the water you see is water
even if it is water. It might, for all you can tell, be XYZ.
Quite independently of theories of meaning, then, we must recognize that something can
carry information about dogs (blue jays, water) in one context, not in another. There are
pure contexts, contexts in which dogs have to be taken as the only relevant cause of (the
symbol) DOG, and impure contexts, contexts in which wolves, cats‐on‐a‐dark‐night, and
perhaps even drugs in the bloodstream might be causing DOG. This being so, I see no
problem in appealing to this distinction in a theory of meaning. A symbol acquires the
function of carrying information in one context, a pure context, and it is used in contexts,
impure contexts, in which it may or may not carry the information it has the function of
carrying, contexts in which it can, therefore, get things wrong. My own view is that the
pure contexts are those in which an information‐carrying element is incorporated into a
control system, a context defined by that time in which the information (relevant to an
organism's adaptive behaviour) ‘gets its hand on the steering wheel’ (see Dretske 1981,
1988). But there are other possibilities.
Not only do information‐carrying functions give physical systems (and, therefore, the
central nervous system) the power to misrepresent the world—thus serving as one way
natural meaning is converted into non‐natural meaning—they also hold promise of solving
a related problem that any naturalized semantics faces: the problem of grain. Non‐natural
meaning individuates propositions in a very fine‐grained way. Thus, for instance, ‘b is 32
°F’ means (non‐naturally) something different from ‘b is 0 °C’, even though these are
merely two different ways of picking out the same temperature: the freezing point of
water. In saying that b is 32 °F one does not say that b is 0 °C. One who believes that b is
32 °F does not necessarily believe that b is 0 °C. These are two distinct beliefs, distinct
meanings, distinct mental contents. Yet (p. 390) if something is a sign (indicates, carries
the information) that b is 32 °F it must be a sign (indicate, carry the information that) b is
0 °C. We may not know it carries this information, of course, but it nonetheless carries it.
If the probability, given the signal, of b being 32 °F is 1, then the probability that b is 0 °C
is also 1. Natural meaning does not carve things up as finely as does non‐natural
meaning. This appears to be a problem for theories like information‐theoretic semantics
that propose to analyse non‐natural meaning in terms of natural meaning. Non‐natural
meaning and natural meaning have different grains.
And so they do, but information‐theoretic semantics is not making the mistake of
identifying non‐natural meaning with natural meaning. Non‐natural meaning is being
identified with an object's information‐carrying function, and it is reasonably clear that
Page 10 of 15
Information‐Theoretic Semantics
22.4 Epiphenomenalism
If there really are semantic engines, systems whose behaviour (some of it anyway) is
driven (explained) by the semantic properties of their internal states, a theory of meaning
should reveal just how meaning gets its hands on the steering wheel. How does meaning
achieve this causal relevance? If the neurobiological properties of internal events are not
enough to explain why we do some of the things we do—the purposeful actions—how do
the semantic properties, the content or meaning of internal events (i.e. what we believe,
intend, and desire), figure in causal explanations of behaviour?
(p. 391)
The problem of finding a genuine explanatory role for meaning in the behaviour of
intentional (i.e. semantic) systems is especially difficult for semantic theories that locate
the source of meaning in the extrinsic or relational properties of the internal events that
presumably have meaning. If a theory says that the meaning or content of internal events
derives from the way these symbols are related—functionally, informationally, causally, or
whatever—to (usually) external affairs, then it should be possible for physically
indistinguishable events, symbols that have all their intrinsic properties in common, to
have quite different meanings. This should be possible in the same way it is possible for
identical twins, say, to have different histories. According to orthodox thinking about
causality, however, the causally relevant properties of an object are local or intrinsic.
Physically indistinguishable objects (a real $100 bill and a perfect forgery, say), placed in
the same circumstances, will have exactly the same effects on the rest of the world. They
will have different histories, but these historical (extrinsic, relational) differences will be
masked or screened off by their current physical constitution. And so it is with meanings
in so far as meaning supervenes on history. What will be causally relevant to the effects of
Page 11 of 15
Information‐Theoretic Semantics
symbols in a system will not be their history—and, therefore, their meaning. It will be
their physical (intrinsic) properties.
Information‐theoretic semantics has this problem in spades, since it locates the meaning
of a symbol in the history of that symbol, in what kind of information it was, at some point
in the past, developed to carry, and history is obviously not a local, not an intrinsic, and,
therefore, not (according to orthodox theory) a causally relevant property. Every other
semantic theory (and this seems to be most theories) that locates the meaning of a
symbol in its extrinsic or relational properties has the same problem. This problem is
most often dramatized by Davidson's swampman example. Lightning strikes a swamp
creating a duplicate of Donald Davidson, a being that is physically indistinguishable from
Davidson but one having a completely different history. Does one's theory of meaning
assign the same meanings, the same content, to the internal states of this duplicate being
as it does to Donald Davidson? If so, then history is irrelevant to the assigned meanings.
If not, then the duplicate has no thoughts. It is a zombie. The second choice seems to be
inconsistent with scientific materialism. The first choice is inconsistent with information‐
theoretic semantics. Conclusion? Information‐theoretic semantics is inconsistent with
scientific materialism. This wouldn't be so bad for some theories, but for information‐
theoretic semantics it is a disaster, since it was proposed as a materialistically acceptable
account of meaning.
Some philosophers (e.g. Paul Churchland 1981; Patricia Churchland 1986) take this
problem to be insurmountable and construe semantic explanations of behaviour, folk‐
psychological explanations couched in terms of the content or meaning of internal events,
as, at best, a façon de parler, a convenient verbal heuristic that will eventually (when we
know enough) be replaced by scientific (neurobiological) explanations of behaviour.
Others (e.g. Dennett 1987) acknowledge the instrumental or heuristic nature of semantic
explanations, but take the heuristic—the intentional stance—to be unavoidable. Still
others argue for a genuine relevance (p. 392) for meaning in a variety of different ways:
Fodor (1987) identifies an alleged kind of meaning, narrow meaning, that is intrinsic to
the events that have it; Dretske (1988) argues that the behaviours to be explained by
meaning are not the bodily movements that are best explained by neurobiology, but,
rather, causal processes that result in bodily movements (which are best explained by the
extrinsic properties of the internal causes); Kim (1996) identifies a kind of causality,
supervenient causality, that extrinsic properties (and, therefore, meanings) can
participate in; Burge (1989) defends the causal relevance of semantic properties by
arguing that genuine causal explanations are those that are actually given and accepted
in our daily explanatory practice (a practice loaded with semantic explanations).
Perhaps, though, the best defence against the charge of epiphenomenalism is to say that
although the problem is real enough, it is not just a problem for information‐theoretic
semantics. It is a problem for any theory of meaning—and this seems to be all theories of
meaning—in which meaning resides in a symbol's extrinsic or relational properties. In
philosophy everybody's problem isn't anybody's problem.
Page 12 of 15
Information‐Theoretic Semantics
References
Ariew, A., Cummins, R., and Perlman, M. (2002) (eds.), Functions (Oxford: Oxford
University Press).
—— (1990), A Theory of Content and Other Essays (Cambridge, Mass.: MIT Press).
Grice, P. (1989), Studies in the Way of Words (Cambridge, Mass.: Harvard University
Press).
Israel, D., and Perry, J. (1990), ‘What Is Information?’, in P. P. Hanson (ed.), Information,
Language, and Cognition (Vancouver: University of British Columbia Press), 1–19.
Page 13 of 15
Information‐Theoretic Semantics
Notes:
(1) For an excellent collection of essays debating this issue see Ariew et al. (2002).
(2) Is an animal sick or diseased, or is its leg broken, only if we, human beings, deem it
so? That doesn't sound right. If these conditions really are independent of our (or, indeed,
anyone's) beliefs and interests, then there is a way things are supposed to be that is, in
the appropriate sense, natural or objective in the way required by information‐theoretic
semantics.
(3) It was Fodor who first raised the disjunction problem (1984). See also Fodor (1990).
(4) This isn't strictly true, since a signal can carry the information that s is F without
being caused by s's being F. There are causal arrangements (e.g. common causes) in
which e carries the information that s is F without being caused by s's being F. The claim
in the text, though, is close enough to being true for the purposes at hand. Nothing can
indicate or mean (naturally) that s is a dog, nothing can carry this information about s,
unless, in fact, s is a dog.
(5) See Dretske (1981: 215–22) for a similar discussion of how to distinguish, on
information‐theoretic grounds and compositional considerations, the concept (and,
therefore, belief that something is) WATER from the concept (belief that something is)
H2O despite the (metaphysically) necessary equivalence of water with H2O.
Fred Dretske
Page 14 of 15
Biosemantics
The term ‘biosemantics’ has usually been applied only to the theory of mental
representation. This article first characterizes a more general class of theories called
‘teleological theories of mental content’ of which biosemantics is an example. Then it
discusses the details that distinguish biosemantics from other naturalistic teleological
theories. Naturalistic theories of mental representation attempt to explain, in terms
designed to fit within the natural sciences, what it is about a mental representation that
makes it represent something. Frequently these theories have been classified as either
picture theories, causal or covariation theories, information theories, functionalist or
causal-role theories, or teleological theories, the assumption being that these various
categories are side by side with one another.
Keywords: biosemantics, mental representation, teleological theories, mental content, naturalistic theories,
information theories
‘BIOSEMANTICS’ was the title of a paper on mental representation originally printed in the
Journal of Philosophy in 1989. It contained a much‐abbreviated version of the work on
mental representation in Language, Thought, and Other Biological Categories (Millikan
1984). There I had presented a naturalist theory of intentional signs generally, including
linguistic representations, graphs, charts and diagrams, road‐sign symbols, animal
communications, the ‘chemical signals’ that regulate the functions of glands, and so forth.
But the term ‘biosemantics’ has usually been applied only to the theory of mental
representation. Let me first characterize a more general class of theories called
‘teleological theories of mental content’ of which biosemantics is an example. Then I will
discuss the details that distinguish biosemantics from other naturalistic teleological
theories.
Page 1 of 14
Biosemantics
Roughly, the idea is this. You tell the teleologist what you think makes some item in the
head, some facet or activity of the brain, into a representation of some facet of the world,
say, into a belief that it is raining. The teleologist may well agree with your theory about
this. But then she will go on to point out (typically this is so) (p. 395) that your theory is
really, at root, a story only about what it is for a mental state or activity to represent truly
or correctly. You need to add a story about what a representation is like that represents
falsely. And she will claim that this is very easy to do. Assume that the brain was
designed, by evolution or learning, to make or to learn to make representations of the
kind you have described. But what it was designed to do will not always be what it in fact
does. Difficult environmental circumstances, even circumstances that merely fail to be
ideal, often cause temporary failures for biological systems. Systems designed to produce
representations will sometimes fail to produce them correctly. Sometimes they will
produce items that behave in the mind/brain as though they represented something, but
that in fact do not represent anything. These are false representations. They are ‘false’ in
the dictionary sense of ‘not genuine or real’, ‘resembling but not accurately or properly
designated as such’, the sense in which false faces and false fronts are false. That is,
teleological theories are best understood as denying that there IS any state of affairs or
occurrence being represented when one thinks falsely or that there IS any object at all
that is being represented when one thinks emptily, say, when seeming to think about
‘phlogiston’ or ‘the ether’.
Similarly, there IS no object, not even an inner one, being seen when one has a
hallucination. Mistaken representations, rather than representing peculiar objects, things
called ‘intentional contents’, are just representations that are failing to represent. False
representations are representations yet fail to represent in the same way that something
can be a can opener but be too blunt, hence fail to open cans or—and this is a better
analogy—something can be a coffee maker yet fail to make coffee because the right
ingredients were not put in or it was not turned on. They are ‘representations’ only in the
sense that the biological function of the cognitive systems that produced them was to
make representations. Thus, falsehood is explained by the simple fact that biological
purposes often go unfulfilled, and the ghostly realm of intensions, reified meanings, non‐
Page 2 of 14
Biosemantics
But much work still remains. The teleologist must give an account of biological functions
and of functions derived from learning that can support the view that the mature brain
has the production of representations, representations of quite specific kinds, as one of
its functions. The account I have given ultimately rests these functions on a variety of
different kinds of selection, the most fundamental being natural selection. That story can
be found in Millikan (1984: chs. 1–2; 1993: chs. 1–2: 2002). In Millikan (2004: ch. 1 n. 2) it
is defended against the claim that human intelligence could have been a genetic accident
rather than a trait selected for.
The other main work that remains is to explain what it is for some facet or activity of the
brain to be a representation of some affair in the world. What teleological theories do not
have in common is any agreed‐on description of what representing is. They do not agree
on what an organism that is representing things correctly, actually representing things, is
doing, hence on what it is that an organism that is misrepresenting is failing to do.
Teleological theories, just as such, are not theories of mental (p. 396) content. Failure to
grasp this last point has led many to take a dismissive attitude toward teleological
theories. How, they ask, could the question whether my current thought is the thought
that grass is green rather than the thought that aardvarks bark be a matter that is settled
in part by reference to evolutionary history or to my past learning history? But a
teleological theory, just as such, makes no attempt to explain what makes your thought be
a representation that grass is green or that aardvarks bark. A prior theory of what
(correct) representation is is needed for that. To the shell that is ‘teleosemantics’, then, it
is necessary to add a description of what successful representing, actual representing, is
like. The main work of this article is to explain what representing is according to
‘biosemantics’.
Above I suggested that the teleologist can take any naturalistic theory of representation
at all and turn it into a teleological theory of content. But there is a catch. If the
teleologist anchors the notion of function in selection, the theory of representation
adopted must allow us to explain how producing inner representations might sometimes
benefit an organism. Otherwise it will be a mystery why any organism would contain
systems designed by selection to make representations. Surely such a requirement is
reasonable, but on careful consideration it turns out to be quite confining.
Page 3 of 14
Biosemantics
representations. Peirce spoke of the interpretant of a sign as being another sign, but
taking this at face value would produce a regress. The interpreter of an inner sign cannot
be supposed merely to translate the sign into another inner sign which is again
translated, and so forth. ‘Interpreting’ a sign must ultimately consist in some
independently useful activity.
According to biosemantics there are several different kinds of process that use
representations. Most theories of representation deal with descriptive representations
only—with representations that are designed to represent facts. But directive
representations are certainly equally important—representations that tell what to do. And
the most primitive and fundamental kind of representation, I believe, faces both ways at
once, saying at the same time what the case is and what to do about it. For example, the
dance of the honey bee tells where the nectar is and at the same time where the watching
bees are to go. I call this last a ‘pushmi‐pullyu’ representation after Hugh Lofting's
charming two‐headed creature by that name (Millikan 1996). Pushmi‐pullyu
representations simultaneously describe and direct. Better, since the term
‘representation’ suggests to many people something more fancy than the simplest
examples I have in mind (in particular, it may suggest symbolic forms that are ‘calculated
over’) I prefer to call the inner signs that I describe ‘intentional icons’ (Millikan 1984)—I
will explain why in a minute. Representations that (p. 397) are calculated over—that
participate in inference processes—are just one very fancy kind of intentional icon. Below
are diagrams of each of the three basic kinds of intentional icon. I will discuss them in
turn.
In each of the diagrams there is a producer and an interpreter or ‘consumer’. These have
been designed by natural selection or by learning to cooperate with one another. Perhaps
each resides in a separate organism; for example, one is part of a dancing bee, the other
part of a sister watching bee. Or perhaps they correspond merely to two different
functions performed within the same brain. What the producer does helps the consumer
to perform functions that loop back to make both the producer and the consumer more
likely to survive or to maintain their current settings (selection through learning) or to
proliferate. The presence of each is part of the normal mechanism by which the other
helps itself to survive or proliferate, and this cooperation is no accident but the result of
past selection or learning that operated on both together.
Page 4 of 14
Biosemantics
Page 5 of 14
Biosemantics
functioning normally, they will picture in accordance with these rules, and they are then
said to be ‘true’ or to be ‘satisfied’.
I had best add to this that no limit need be placed on the complexity of the semantic‐
mapping functions that might map intentional icons on to the states of affairs they
represent. Isomorphisms can be defined by functions that are as bizarre, as grue‐like, as
you please. A bizarrely coded secret message from a CIA agent can be as much an ‘icon’
or ‘picture’ that maps on to a certain world affair in accordance with a definite semantic‐
mapping function as any bee dance, sentence, or diagram. Intentional icons must be
things apt for use by icon users, but icon users can be very idiosyncratic in their habits.
For example, if mental representations are systems of brain happenings or brain states
that map on to represented world affairs, no a priori limitation on the kinds of brain
happenings or states involved or on the complexity of the mappings employed is implied.
Every representation is in some kind of code. The complexity of the code is irrelevant. On
the other hand, any intentional icons in the brain would of course have to come with inner
interpreters that knew how to read them; that is, interpreters that could be guided by
them reliably to fulfil their own functions. Simple codes relying on only a few principles, if
they were also highly productive, tapping into rich natural isomorphisms between the
domains of the signs and the signifieds, would seem much the most likely to be preferred
by natural selection.
The semantic rule associated with a descriptive intentional icon determines a condition or
state of affairs that must obtain if the consumer is to perform its tasks, whatever they
may be, in the normal way. The consumer varies its activities systematically according to
variations in the icons presented to it. The result is that the consumer's activity conforms
or is adapted to the condition or state of affairs represented by the icon so that it can
perform its functions properly given that condition. Descriptive intentional icons are
designed to stand in for world affairs, typically affairs outside the organism or organisms
involved, and to vary according to these world affairs, controlling internal or external
behaviour as needed to adapt to these affairs. Thus, the bee dance is, in part, a
descriptive intentional icon because if the watching bees are to achieve their function of
finding nectar by reacting to the dance in the normal way it needs to correspond by a
certain rule to a fact about nectar location. (We need not assume it takes any thought on
a watching bee's part to react appropriately; so we need not assume that the dance is
interpreted by translating it into another sign.) Similarly, consider my belief that there is
yogurt in the refrigerator. Which are my belief‐using systems? Some, at least, are the
systems that make practical inferences and turn the conclusions into practical action. My
belief that there is yogurt in the refrigerator can help these systems to perform their
function of guiding me into activities helping to fulfil my desires and plans (to eat yogurt,
to make breakfast for a guest, etc.) in a normal way (not by serendipitous accident) only if
there is yogurt in the refrigerator.
(p. 400)
Page 6 of 14
Biosemantics
Notice that the content of the descriptive icon is not determined by reference to the kind
of tasks that it (with the help of its consumer(s)) normally performs. The icon may be
called on to promote a whole series of functions, but each of these will be fulfilled
normally only if the icon is ‘true’. If the consumer works in part by making inferences, the
content is not determined by any particular set of inferences its consumer is disposed to
make. It is determined by the fact that whatever inferences its consumer makes when
functioning properly, the result will be (non‐accidental) true belief or successful action
only if it actually represents according to a certain correspondence rule. This means that
its content is not determined by its conceptual role.
Note also that in the case of descriptive icons the producer's job is primarily to make an
icon that corresponds by the right rule to a state of affairs. If the producer succeeds in
this task, whether through the normal mechanisms of icon production or by freakish
accident, the intentional icon is still true. Truth does not rest on whether the means of
production was normal. For example, a belief can be true without being knowledge.
Taking a classic case first introduced by Dretske (1986), consider the magnetosome of a
northern‐hemispheric ocean bacterium that normally works by being pulled toward
magnetic north, hence toward geomagnetic north, hence away from aerated surface
water which is lethal to these bacteria. The magnetosome's pointing in a certain direction
is an intentional icon of the direction of lesser oxygen because that is what it needs to
correspond to for the bacterium's resultant motion to serve its proper function. If the
magnetosome points in the direction of lesser oxygen because, serendipitously, a bar
magnet under it is pointed in the right direction, then it tells the truth by accident rather
than through normal mechanisms, but it still tells the truth. On the other hand, if a bar
magnet attracts the bacterium to its destruction by pointing it towards oxygen, what the
magnetosome says is false despite its being produced through a normal mechanism.
Attributions of truth and falsity do not rest on whether descriptive icons are produced by
normal mechanisms, but only on normal mechanisms associated with their use. (An icon's
function cannot be to have been produced by something! Functions are effects, not
causes.) Of course, Dretske is right that the magnetosome that directs the bacterium in
the wrong direction because a bar magnet is held over it is not broken or malfunctioning.
In that sense it is functioning ‘perfectly properly’ (Dretske 1988). But it doesn't follow
that it is succeeding in performing all of its functions, any more than a perfectly
functional coffee maker is performing all its functions when it is turned on but no coffee
has been put in it. Very often things fail to perform their functions, not because they are
damaged, but because the conditions they are in are not their normal operating
conditions.
rule to the direction of flight of the watching bees (p. 401) if their dance‐interpreting
apparatuses succeed in serving their functions normally. Suppose that you have a desire
to eat yogurt. The systems designed to be moved by your desires—the ‘consumers’ of
your desires—are the systems that make practical inferences eventually turning the
conclusions into practical action. If your desire to eat yogurt affects these systems as
designed, it will guide them to effect the fulfilment of your desire to eat yogurt. The chief
function of a desire is to get itself fulfilled. This is not to say that, on average, desires do
get themselves fulfiled. On average they generally get eaten up by bigger opposing
desires first, or perhaps no means are known to fulfil them. It is very common for a trait
or capacity to have been selected because it sometimes performs a useful function,
occasional performance being better than none. The point is that people would not have
the capacity mentally to represent various states of affairs as desirable unless these
desires were sometimes fulfilled. The capacity would otherwise be useless.
Thus, the cooperation between producer and consumer in the production and use of
intentional icons can be accomplished in any of three basic ways. It may be that the
producer is the one primarily responsible for making the icon correspond to a certain
state of affairs; it may be that the consumer is the one primarily responsible; and in the
case of pushmi‐pullyu icons it is the responsibility of the producer to make the icon
correspond to one kind of affair and the consumer's job to make it correspond to another.
In each of these basic cases, granted the cooperation between producer and consumer
comes about through the normal causal mechanisms, the intentional icon will also be a
‘local natural sign’ carrying ‘local natural information’ about the affair or affairs to which
it corresponds (Millikan 2004: chs. 3–4). Local natural signs are, in part, abstract pictures
of what they represent. Although there is not room to unpack the notion of local natural
information here, I mention this because it follows that when intentional systems are
functioning normally in accordance with normal explanations, intentional icons represent
both by being pictures of what they represent and by carrying natural information as to
what they represent. For example, a bee dance is often a local natural sign both of where
there is nectar and of where the watching bees will go. In the fundamental sense, actually
representing involves both picturing and carrying natural information. As such it is not a
matter determined by a history of selection. Representing intentionally, however, is a
matter of having a certain kind of history. Also, attributions of truth or falsity and of
satisfaction or unsatisfaction make sense only by reference to function, hence by
reference to a history of selection. The terms ‘true’ versus ‘false’ and ‘satisfied’ versus
‘unsatisfied’ do not apply to natural signs, the most basic kind of representations.
equilateral. These problems do not arise for the theory of intentional icons. Information
carried by normally operating intentional icons is a form of natural information. It does
not follow that all of the natural information carried by an intentional icon is carried
intentionally. The information that a natural sign carries intentionally is only the
information it is its function to carry, the information that its cooperative interpreters
know how to use. This information may be very abstract, and it may be about very distal
affairs. If its consumers are so designed that they can use only the information that
something is triangular, then that is all the information that the icon carries intentionally.
If they are so designed as to use only the information that a predator is near, then the
intentional icon will not intentionally carry information about any more proximal affairs,
such as patterns on the retina.
Similarly, not every stimulus that an organism discriminates on the way to producing an
intentional icon is represented intentionally. Nor must an organism be capable of
infallibly discriminating the distal objects, properties, or kinds that it intentionally
represents from all others that are similar. It needs only a fallible capacity to recognize
some natural signs or other of these things under some local conditions. Possibly it even
gets things wrong a large part of the time because a large part of the time supporting
conditions on which its mechanisms of icon production rest are absent. Similarly, the
rabbit's danger thump may be elicited more commonly when rabbit danger is absent than
present. (What matters is the converse—that when danger is present it should usually be
elicited.)
The basic theory of representing on which biosemantics rests is a picture theory and an
informational theory, but equally it is a functionalist theory. The basic idea is that what
makes something into a representation, for example into a mental representation, is not,
of course, what it is made of, but what functions it performs and/or how it performs these
functions. Items that function in certain ways are representing, and if they have been
designed to function in these ways they are representing ‘intentionally’. They are
representing, that is, in accordance with natural purposes and such that they can be said
to be true or false, satisfied or unsatisfied. (The intimate connection of functions that
have been selected for with purposes is argued in Millikan 1984; 2004: ch. 1.)
Page 9 of 14
Biosemantics
hen's food call to her chicks descriptively shows the time and place of food and directively
tells the time and place her chicks are to come. But it is evident that there is no sense in
which a particular time stands for itself or a particular place for itself outside the context
of some such signal. Similarly, in telling what direction the bacterium is to go the
direction in which its magnetosome points stands for itself, but there is no prior sense in
which a direction stands for itself.
The biosemantic account also implies that there are not and could not be intentional icons
that lacked attitude. Rejected is the Fregean idea that first a proposition is represented,
then an intentional attitude added. Intentional icons always have, as such, functions and
their functions automatically create attitudes. Hypothetical thinking, for example, or just
thinking of possibilities, is an extremely sophisticated activity, and one that is only
possible for a creature that sometimes uses the results in the production of ordinary
descriptive and directive representations. It is because thoughts ‘of possibilities’ have the
function of sometimes turning into more basic kinds of representations that they exist as
representations at all. Similarly, desires are intentional icons only because they are
designed to turn, under certain conditions, into full‐blown intentions, whose functions are
to effect their own fulfilment more directly. There would be no benefit in the capacity to
have desires if desires did not sometimes travel the whole route through intention into
action.
Intentional icons represent complete states of affairs. This implies that they represent not
only properties but also the things that have those properties. When produced normally,
intentional icons also carry natural information corresponding to what they represent.
This implies that there can be natural information as to what things have what properties
—including what individuals have what properties. Contrast Dretske's description of the
natural information carried by signals. As Dretske describes the matter, although a signal
can carry the information that an individual x is F, there is no part or aspect of the signal
that carries the information that it is x that is F. For example, the petrol gauge on your car
may carry the information that your petrol tank is half full, but no aspect of the signal
indicates which tank is half full (Dretske 1981, 1988). You have to know that
independently. The petrol‐gauge reading does not represent its subject, nor could it on
Dretske's theory of natural information, because there are no natural laws that pertain to
any individuals just as such. A necessary and central feature of the theory of local natural
information (Millikan 2004) is that it explains how a natural sign can signify which
individual it carries information about. The result is an explanation of how intentional
representations of individuals are possible, something for which, to my knowledge, no
other naturalized theory of intentionality accounts.
Page 10 of 14
Biosemantics
produces a likeness of that pattern. Similarly, of course, (p. 404) one's eyes are not
designed to see any particular object but, roughly speaking, to see whatever object lies in
front of them. If a particular person is in front of your open eyes, it is a function of your
eyes to help produce an accurate perception of that person. Similarly, the function of an
adding machine is not to give any particular answer, but to give the sum of the numbers
put into it. Natural selection has designed cognitive systems not to turn out particular
products, say particular beliefs and desires, but to turn out quite different beliefs and
desires depending on environmental circumstance. In the case of human beliefs and
desires, however, the matter is considerably more complicated than with the camera. In
order to turn out beliefs that will vary depending on states of affairs in the environment
and in order to tune the systems that use these beliefs during practical and theoretical
deliberation and in the production of useful action, humans must first develop adequate
empirical concepts. Indeed, to complete the biosemantic programme a rather long story
needs to be told about conceptual development. We must explain how the producers and
consumers of beliefs and desires can learn or be tuned to employ empirical concepts
cooperatively without actually practising together through the production of concrete
actions. We must explain how their representation, production, and use dispositions can
be tailored in advance to fit one another. That story is told in Millikan (1984: chs. 15–19;
2000: esp. ch. 7; 2004: p. IV, esp. ch. 19).
Many critical questions about the biosemantic theory first presented in 1984 and 1989
have come and gone, but there are three that have been especially tenacious. I will say a
word about each.
A second question concerns the possibility of biological systems whose jobs are to
produce false representations. For example, people who are overconfident may be more
successful at performing certain tasks than people who evaluate their skills correctly.
Notice first, however, that it will not be the falseness per se, but only the high confidence
Page 11 of 14
Biosemantics
that contributes to success. If one is completely and perfectly competent (p. 405) at a
task, there certainly will be no gain in believing one is not! Notice second that there are
many biological systems that ride piggyback on systems developed earlier for quite
different purposes. If there actually were systems whose jobs were to distort certain
beliefs they would have to ride on more general systems whose basic jobs were to
produce true beliefs. Otherwise there would be no semantic rules in place according to
which the distorted beliefs would be false. The various concepts out of which any beliefs
are formed are designed to serve purposes in arbitrary belief contexts. The systems
responsible for concept development tune these concepts and the systems that normally
use them for general purpose use, not for any one specific use such as increasing one's
confidence. The semantics of mental representation is productive. That is what the
‘picturing’ or ‘mapping’ guarantees.
Third, consider Pietroski's tale about the kimus and the snorfs (1992). The kimus are
attracted by the red sunrise glowing over their local mountain so that they climb up it
each morning. Thus, they conveniently avoid their chief predators, the snorfs, who pass
by each morning below. Moreover, this is how the attraction to red light got selected for
in kimus. Those not attracted by red light got eaten. On the biosemantic view, Pietroski
claims, ‘kimus climb the hill because they believe the hill is snorf‐less’, and when they
approach red things that are not snorfs ‘they are acting on the belief that the area in
question is snorf‐free’ (1992: 276). Given that kimus ‘can't reliably discriminate snorfs
from non‐snorfs’, (ibid.), it is implausible, he claims, that the kimus have any beliefs about
snorfs. In summary, on the biosemantic account
[a] system can have the belief that P is instantiated without having any systematic
ability to tell whether P is instantiated (in a given region at a given time). Indeed,
instantiations of P can be completely irrelevant to the system's tokening of the
belief that P is instantiated. The corresponding intentional explanations of such a
system's behaviour will … be very implausible.
What Pietroski describes in the kimus seems to be a simple tropism. They are attracted to red
the way a moth is attracted to light. Apparently they have neural pushmi‐pullyu intentional icons
or signals that tell where the snorfs are fewer and hence where to go. Of course these icons are
not at all like beliefs. Beliefs are formed only after the acquisition of concepts, which generally
rest on multiple ways of recognizing. Further, the functions of beliefs involve participation in
inference. (For a discussion of empirical concepts and of inference see Millikan 2000.) Pietroski
seems to assume that an ‘intentional explanation’ of an animal's behaviour must not only be a
belief–desire explanation but must also be a straightforward causal explanation. Why intentional
explanations are causal but not straightforward is explained in Millikan (1993, 2007). Moreover,
full intentional explanations do not begin with the presence of intentional representations, but
explain also how the intentional representations get formed. The red light is definitely involved
in an ‘intentional explanation’ of how the kimus avoid snorfs.
Page 12 of 14
Biosemantics
Finally, there are no distal objects or stimuli that any organism has the capacity to
discriminate under all conditions. All successful discrimination of distal affairs (p. 406)
depends on merely local natural information. Local natural information rests on
correlations that are not perfect but that are not accidental either, for they must persist
throughout a spatio‐temporal region for a reason (Millikan 2004: chs. 3–4). On this
analysis the kimus do get local natural information each morning about the direction of
fewer snorfs. Similarly, although there is no causal connection, the correlation between
magnetic north and lesser oxygen used by the anaerobic bacteria persists in the northern
hemisphere for a reason. It carries local natural information about the location of lesser
oxygen. The intentionality that characterizes pushmi‐pullyu icons responsible for simple
tropisms of this kind is the limiting case of intentionality. It is intentionality in the way
zero is a number. If your theory doesn't count in these cases you will find that it fails to
account for any of the obvious cases either.
References
Dretske, F. (1981), Knowledge and the Flow of Information (Oxford: Blackwell).
—— (1986), ‘Misrepresentation’, in. R. Bogdan (ed.), Belief: Form, Content, and Function
(Oxford: Clarendon), 17–36.
repr. in Millikan, White Queen Psychology and Other Essays for Alice (Cambridge Mass.:
MIT Press); and in C. Macdonald and G. Macdonald (eds.), Philosophy of Psychology:
Debates on Psychological Explanation (Oxford: Oxford University Press, 1995), (pts VII–
X).
repr. in L. May and M. Friedman (eds.), Mind and Morals (Cambridge, Mass.: MIT Press,
1996), 145–61.
Page 13 of 14
Biosemantics
—— (2004), Varieties of Reference: The Jean Nicod Lectures 2002 (Cambridge Mass.: MIT
Press).
Page 14 of 14
A Measurement‐Theoretic Account of Propositional Attitudes
Keywords: attitudes, relational conception, attitude predicates, propositional attitudes, monadic properties,
relational predicates
Page 1 of 16
A Measurement‐Theoretic Account of Propositional Attitudes
Beginning in the late 1970s, a number of philosophers began to argue that our difficulty
understanding the nature of the attitudes stems from our misunderstanding the natural‐
language predicates by which we attribute propositional attitudes. In particular, they
argued that propositional‐attitude predicates function very much like the numerical‐
measure predicates by which we attribute physical magnitudes (e.g. has a mass of 5 kg),
that in fact propositional‐attitude predicates are a kind of measure predicate (see
Churchland 1979: 105; Field 1981: 113–14; Stalnaker 1984: 8–11; Dennett 1987: 123–5;
Davidson 1989; Matthews 1990, 1994, 2007; Beckermann 1996).1 Seen in this way, the
sentences by which we attribute propositional attitudes are indeed relational in form, and
hence do express relations, but the relations that such predicates express are not
relations that are in any way constitutive of the possessor's possession of the
propositional attitudes attributed by these predicates. To say that a subject has a certain
propositional attitude is no more to say that the subject stands in a substantive
psychological relation to the particular that is the referent of the that‐clause than is to say
that an object has a temperature of 30 °C to say that the object stands in a substantive
physical relation to the number 30. Rather, it is to attribute to that subject a certain
psychological state or property which is specified by means of its location in a
representational domain, in just the way that we specify the temperature of an object by
means of its location on a numerical scale. The referents of that‐clauses, like numbers,
are simply abstract entities used to represent measurement‐theoretically the
Page 2 of 16
A Measurement‐Theoretic Account of Propositional Attitudes
propositional attitudes of those to whom these attitudes are attributed. They are not, in
any sense, the psychological ‘objects’ of these states.
Philosophers attracted to the idea that attitude predicates might be a kind of measure
predicate have been so attracted for a variety of reasons. Churchland (1979) sees in the
idea a way of avoiding what he regards as a metaphysically vexing commitment in
psychology to propositions. Field (1981) finds in it not simply a way of avoiding any
commitment in psychology to propositions, but also the possibility of a representationalist
solution to Bretano's problem of the intentionality of propositional attitudes. Stalnaker
presents what he calls the ‘measurement analogy’ as a way of blunting the intuitive force
of what he terms the ‘linguistic picture’ of propositional attitudes (defended by Field,
Fodor, and others), a view which he describes as ‘undertak[ing] to explain thought by
speech’ (1984: 5). As Stalnaker sees it, the measurement analogy makes conceptual room
for the ‘pragmatic picture’ of the attitudes that he favours, according to which
propositional attitudes ‘should be understood primarily in terms of the role that they play
in the characterization and explanation of (p. 409) action’ (1984: 4). Dennett (1987) and
Davidson (1989) see in the measurement idea not simply a way of avoiding a vexing
commitment in psychology to abstract objects but also a way of avoiding altogether a
relational account of the attitudes. Both find in it the possibility of a non‐relational, non‐
representationalist account of the attitudes. Dennett points out (1987: 125n.) that while
we attribute a particular mass to an object by relating that object to a particular real
number, we don't suppose that the object's having the mass that it does is a matter of its
being related to some particular. Having the mass that it does is an intrinsic (monadic)
property of the object; relating the object to some particular number is just a way of
specifying the property in question. Perhaps, Dennett suggests, the same is true for
propositional attitudes: we attribute them to an individual by relating that individual to a
particular, but propositional attitudes are not themselves relations that their possessors
bear to particulars. Davidson, for his part, insists that it does not follow from the fact that
we specify what a thinker thinks by relating him to a certain object that the thinker bears
any psychological relation to that object. We are, he says, ‘free to divorce the semantic
need for content‐specifying objects from the idea that there must be any objects at all
with which someone who has an attitude is in psychic touch’ (1989: 9; my emphasis).
According to Davidson, the analogy then between measuring weight and attributing
states of belief is this:
Page 3 of 16
A Measurement‐Theoretic Account of Propositional Attitudes
state of mind do not have to play any psychological or epistemological role at all,
just as numbers play no physical role.
(1989: 11)
The chief virtue of a measurement‐theoretic account of the attitudes, besides getting right the
semantic nature of attitude predicates, is precisely that, as Davidson suggests, it enables one to
embrace the very plausible relational conception of attitude attributions without thereby having
to embrace the much more dubious relational conception of the attitudes themselves. To be a
non‐relationalist about the attitudes one does not, as Quine, Fodor, and others have assumed,
have to be a non‐relationalist about attitude predicates. It is entirely possible that propositional
attitudes are monadic properties of their possessors, even though they are predicated of
individuals by means of relational predicates.
In this brief chapter it is not possible to present a detailed exposition and defence of the
idea that attitude predicates are measure predicates, nor of the import of this idea for our
conception of the attitudes themselves.2 I will instead attempt only to sketch the general
outlines of a measurement‐theoretic account, explaining in only (p. 410) the briefest
terms why the account seems plausible, how the account might go, and what its
implications might be for our conception of propositional attitudes. I begin with some
brief comments about numerical‐measurement theory upon which the account is based.
Page 4 of 16
A Measurement‐Theoretic Account of Propositional Attitudes
Numerical‐measurement theory traces its origins to work in the late 1800s and early
1900s on extensive magnitudes such as mass and length.3 This early work focused almost
exclusively on what has come to be called the ‘representation problem’: the problem of
specifying the conditions that an empirical system must satisfy in order for there provably
to exist an assignment of real numbers to objects that respects the empirical relations
among quantities of these magnitudes. Numerical‐measurement theory does this by
proving a so‐called representation theorem, typically by proving that the numerical
relational structure on the reals is homomorphic (in a sense of this term that is stronger
than that employed in set theory) to the empirical relational structure of the represented
quantities of the measured physical magnitude. But the assignment of numbers to
quantities is not unique: there will be different numerical assignments, that is different
scales (e.g. metric and avoirdupois scales for weight, the centigrade and Fahrenheit
scales for temperature), each (p. 411) of which faithfully represents all the relevant
empirical facts about the magnitude being measured. Measurement theory proves that a
certain set of distinct assignments of numbers to quantities are all equally faithful
representations of the empirical facts by proving a so‐called uniqueness theorem, where
this theorem provides a formal characterization of the class of homomorphisms all of
which specify relation‐respecting representations of physical magnitudes on the reals.
Page 5 of 16
A Measurement‐Theoretic Account of Propositional Attitudes
Page 6 of 16
A Measurement‐Theoretic Account of Propositional Attitudes
Finally, both sorts of predicate seem to be quite similar extensions, each within their
respective domains, of a pervasive common‐sense predication scheme. On this scheme we
individuate the capacities (skills, traits, and dispositions) of persons, animals, and even
inanimate objects along two orthogonal dimensions. One dimension has to do with the
object or event that is the focus or target of the individual's capacities; the other
dimension has to do with the particular relation that the individual bears to that object or
event. (Thus, people can be flag wavers, firefighters, stock traders, and so on; animals
Page 7 of 16
A Measurement‐Theoretic Account of Propositional Attitudes
can be anteaters, nest builders, and disease carriers, etc.; and inanimate objects can be
can openers, typewriters, dishwashers, and the like.) Numerical‐magnitude predicates
and propositional‐attitude predicates preserve this two‐dimensional relation‐to‐an‐object
character of the common‐sense scheme, but the specific nature of both the object and the
relation is changed. In the common‐sense scheme we attribute a capacity (skill, etc.) to
an individual (say, firefighting) by (p. 413) specifying (i) the object or event towards which
exercises of this capacity are directed (fires), and then (ii) the sort of relation that the
individual bears to this object or event in the course of exercising the capacity in question
(fighting). In the extensions of this scheme to physical magnitudes or propositional
attitudes we again attribute a property to an individual by specifying an object and a
relation to that object. But the object now serves as a representative of the particular
property being attributed, and the relation serves to type‐identify the property being
attributed (e.g. as a quantity of mass, or as a kind of belief). The common‐sense scheme
has become a kind of indexing scheme for specifying and individuating certain internal
states and properties of individuals.
attitude attributions manage to be informative. Indeed, without such an account, even the
claim to find support for the relational conception in the logical form of attitude
attributions is in trouble, since relationalists lose all justification (p. 414) for reading the
logical form of attitude attributions back on to the attitudes themselves. So it looks as if
relationalists and non‐relationalists alike need a measurement‐theoretic account of
attitude predicates.
There are other, more specific, reasons why both relationalists and non‐relationalists
might want such an account. First, it might provide a ‘solution’ to the problem of the
intentionality of propositional attitudes by showing us that intentionality is no more a
problem for propositional attitudes than is ‘numerality’ a problem for physical
magnitudes: intentionality might just be a feature of our particular way of representing
propositional attitudes rather than a property of the attitudes themselves. Second, it
might explain what Fodor (1987) calls the ‘striking parallelism’ between the causal
relations among propositional attitudes and the semantic relations among their contents:
on a measurement‐theoretic account, the causal relations respect the semantic ones, not
because cognitive architecture somehow contrives to enforce such respect, but simply
because of the measurement‐theoretic representation relation between propositional
attitudes and their natural‐language representatives. The parallelism that Fodor finds so
striking might turn out to be no more remarkable than the strict parallelism that obtains
between the empirical relations among quantities of physical magnitudes and the
numerical relations among their numerical representatives on the reals. The
measurement‐theoretic account might also explain why this parallelism holds not only for
what Fodor (1987) calls ‘core cases’ of propositional attitudes but even for those
‘derivative cases’ which admittedly don't satisfy his so‐called Representational Theory of
Mind.
Page 9 of 16
A Measurement‐Theoretic Account of Propositional Attitudes
empirical relational structure into the representing relational structure which is its
image; and (iv) a proof of a representational theorem and (p. 415) uniqueness theorem for
this set of morphisms so characterized. And because much of our puzzlement focuses on
attitude attributions, the account should also provide an explanation of (v) why we should
have developed the particular attribution practices we have, and (vi) why we should
expect the various puzzles that have so dominated philosophical theorizing about the
attitudes. We might even go on to speculate, as some philosophers have done, as to (vii)
the intrinsic nature of propositional attitudes that would explain why they have the
empirical structure that they do, understanding that such speculation would not, strictly
speaking, be part of a measurement‐theoretic account of the attitudes.
Page 10 of 16
A Measurement‐Theoretic Account of Propositional Attitudes
There is a lot to be said about the properties and relations defined over the IUFs that are
the representatives of propositional attitudes, but basically IUFs have all the properties
and relations of utterances of the sentences for which they are IUFs, including semantic,
inferential, syntactic, pragmatic, phonological, phonetic, and even orthographic
properties and relations. These properties and relations are exploited differentially in the
individuation of particular attitudes depending upon the context of utterance of the
attitude attribution.
Page 11 of 16
A Measurement‐Theoretic Account of Propositional Attitudes
representations of the attitudes, on the assumption that the relations defined on the
domain of the latter ‘respect’ the empirical relations defined over the psychological states
that are the attitudes, and from the explanatory role of propositional attitudes in
common‐sense explanations of behaviour and thought.
The objects that constitute the empirical domain could, as proponents of Fodor's
Representational Theory of Mind would have it, be mental representations. But there is, I
believe, little empirical reason to suppose this to be the case. If we ask ourselves what
minimally must be true of these psychological states that are the attitudes if they are to
be the sorts of states that both find an image in our natural‐language representations of
the attitudes and also exhibit the causal properties that they do in the production of
behaviour and thought, it seems enough that these attitudes be certain psychological
‘aptitudes’ of their possessors. By this I mean simply that they must be states apt to
produce certain characteristic effects, both in the sense that they are the sort of states
that can produce such effects and in the sense that in the appropriate context they do
produce such effects, where these effects include both the (p. 417) behaviour and the
other mental states that we take to be the characteristic effects of propositional attitudes.
There might be more to these states than just being aptitudes for the characteristic
effects that we associate with propositional attitudes, but nothing about common‐sense
propositional‐attitude explanatory practice or the nature of our natural‐language
representations of the attitudes would seem to require it, not even, I would argue, the
intentionality or semantic evaluability that we philosophers associate with the attitudes.
Assuming that these aptitudes which are the attitudes are simply states that involve the
instantiation of a property at a time, then the mapping of propositional attitudes into their
natural‐language representatives will be (like the case of the mapping of physical
magnitudes into their real‐number representatives) a trans‐type mapping of properties
into objects, in this case of aptitudes into IUFs. The relations defined over this domain of
aptitudes will consist of various causal and material constitutive relations, relations that
find their image primarily in the inferential relations defined over IUFs.
There is obviously a lot more that needs to be said to motivate this account of the
empirical structure of the attitudes. Most crucially, we need at very least informal
sketches of both a representation theorem and a uniqueness theorem to convince us that
this account captures what we intuitively take to be the conceptually salient features of
propositional attitudes, especially given that the account does not attribute to these
aptitude states any intentional or semantic properties of the sort presumed by the
Representational Theory of Mind. We also need to be convinced that this account can
handle all the usual semantic puzzles about propositional attitudes, puzzles that are the
focus of so much philosophical discussion. We can get some sense of why the proposed
measurement‐theoretic account might be able to handle these puzzles by noting that it
has all the explanatory and individuative resources of the traditional accounts, but like
the relationalist accounts mentioned above that distinguish between the semantic and
psychological objects of the attitudes, this account apportions those resources differently
Page 12 of 16
A Measurement‐Theoretic Account of Propositional Attitudes
The proposed account is clearly realist about the attitudes, inasmuch as on this account
propositional attitudes are aptitudes for the behavioural and cognitive effects that we
characteristically associate with the attitudes. But the account is not realist about the
semantic/intentional contents of propositional attitudes. The semantic properties that we
intuitively associate with the attitudes turn out to be properties of our natural‐language
representations of the attitudes, and not of the attitudes themselves. Propositional
attitudes turn out to have semantic properties in just the way that physical magnitudes
have numerical properties: given our canonical way of representing propositional
attitudes by means of an IUF that specifies a state of affairs to which the possessor of the
attitude is behaviourally related in certain characteristic ways, we are no more able to
conceive of propositional attitudes in non‐linguistic terms than we are able to conceive of
physical magnitudes in non‐numerical terms. And yet propositional attitudes no more
have the linguistic properties of the IUFs that are their linguistic representatives than do
physical magnitudes have the numerical properties of their numerical representatives.
Although propositional attitudes turn out not to have the semantic properties that we
intuitively conceive of them as having, they do turn out to have a kind of intentionality,
but again not the sort of intentionality that we might associate with linguistic entities.
Rather, theirs is a kind of behavioural intentionality: they have intentionality in the sense
that they are aptitudes for, among other things, behaviours that are directed towards (in
the case of desire‐like states) or sensitive to (in the case of belief‐like states) the states of
affairs in terms of which we represent and specify them.
Page 13 of 16
A Measurement‐Theoretic Account of Propositional Attitudes
Because the account remains realist about propositional attitudes, it in no way impugns
Gricean intention‐based semantic programmes that propose to account for the
intentionality of natural language in terms of the possession by speakers of certain
propositional attitudes. Nor does the account impugn the broad outlines of recent
translationalist accounts of linguistic competence, including semantic competence,
according to which knowing a language is knowing how to translate between sentences in
a natural language and sentences in a language of thought, provided that such accounts
are stripped of their commitment to any particular representationalist account of the
attitudes.
The proposed account is also fully consistent with the idea that representations play an
essential role in all sorts of cognitive processing. It is even consistent with the idea that
as a matter of empirical fact possession of propositional attitudes requires the possession
of various sorts of representations. The account only denies (p. 419) that propositional
attitudes are to be identified with any such representations, that any such
representations are constitutive of the attitudes. If this is correct, then to the extent that
cognitive psychology is computational, and as such traffics in representations of one sort
or another, then propositional‐attitude psychology is not, as some claim, proto‐cognitive
(or proto‐computational) psychology. And precisely because propositional‐attitude
psychology turns out not to be proto‐computational, it entails nothings as regards human
cognitive computational architecture, classical or otherwise, beyond of course the very
minimal requirement that, whatever that architecture, it must be capable of supporting
those aptitudes that the account identifies with the attitudes. But this, it seems, is a
requirement that can seemingly be met by many different sorts of computational
architecture, so that possession of propositional attitudes turns out to be compatible with
a wide variety of architectures.
Where the proposed account diverges most sharply from received relational conceptions
of the attitudes is first and foremost in its unwillingness simply to read back on to the
attitudes properties of their natural‐language representations. And from this
unwillingness arises the stark minimalism that characterizes the measurement‐theoretic
account of the attitudes. Of course, it might turn out as a matter of empirical fact that
there is much more structure to propositional attitudes than this account concedes, but if
this is so, then this fact cannot be discerned either in our natural‐language
representations of the attitudes or in the common‐sense explanatory practice in which
propositional attitudes play their role. Nor, I would argue, is it to be discerned in the role
that propositional‐attitude attributions play in cognitive‐psychological theorizing, though
this is a claim that can't be defended here (see Matthews 2007).
References
Beckermann, A. (1996), ‘Is there a Problem about Intentionality?’, Erkenntnis, 51: 1–23.
Page 14 of 16
A Measurement‐Theoretic Account of Propositional Attitudes
—— (1989), ‘What is Present to the Mind?’, Grazer Philosophische Studien, 36: 3–18.
Larson, R., and Ludlow, P. (1993), ‘Interpreted Logical Forms’, Synthese, 95: 305–55.
Larson, R., and Segal, G. (1995), Knowledge of Meaning (Cambridge, Mass.: MIT Press).
Matthews, R. (1990), ‘The Measure of Mind’, report no. 57/1990, Research Group on
Mind and Brain, ZiF (Centre for Interdisciplinary Research) (Bielefeld).
(p. 420) Matthews, R. (1994), ‘The Measure of Mind,’ Mind, 103: 131–46.
—— (2007), The Measure of Mind: Propositional Attitudes and their Attribution (Oxford:
Oxford University Press).
Suppes, P., and Zinnes, J. (1963), ‘Basic Measurement Theory’, in R. Luce, D. Krantz, and
E. Galanter (eds.), Handbook of Mathematical Psychology, i (New York: Wiley) 1–76.
Notes:
Page 15 of 16
A Measurement‐Theoretic Account of Propositional Attitudes
(1) Suppes and Zinnes (1963: 7) were perhaps the first to suggest the possibility of a
measurement‐theoretic account of ‘attitude statements’. Swoyer (1987: 281–3) offers a
sketch of how the idea that attitude predicates are measure predicates might be
developed.
(4) See Larson and Ludlow 1993 and Larson and Segal 1995.
Robert J. Matthews
Page 16 of 16
The Normativity of the Intentional
Many philosophers have claimed that the intentional is normative. (This claim is the
analogue, within the philosophy of mind, of the claim that is often made within the
philosophy of language, that meaning is normative.) The first two sections of this article
give a brief clarification of what this claim that ‘the intentional is normative’ actually
means. The third section considers a number of arguments that philosophers have
advanced in favour of this claim; as the discussion reflects, many of these arguments are
inconclusive at best. However, the last section gives a sketch of a different argument,
which the author regards as a persuasive argument for this claim.
Keywords: philosophy of mind, normativity, philosophy of language, meaning, intentional, persuasive argument
MANY philosophers have claimed that the intentional is normative. (This claim is the
analogue, within the philosophy of mind, of the claim that is often made within the
philosophy of language, that meaning is normative.) In the first two sections of this
chapter I shall give a brief clarification of what this claim that ‘the intentional is
normative’ actually means. In the third section I shall consider a number of arguments
that philosophers have advanced in favour of this claim; as we shall see, many of these
arguments are inconclusive at best. However, in the last section I shall give a sketch of a
different argument, which I regard as a persuasive argument for this claim.
Page 1 of 19
The Normativity of the Intentional
What is meant here by ‘the intentional’? For present purposes we may take the
intentional to be a subset of the mental. Many mental states—indeed, some philosophers
would say all mental states—are about something; they are ‘concerned with’ or ‘directed
towards’ something. For example, some of my beliefs are about Iraq; some of my
intentions are about travelling to Ireland; and some of my fears are about snakes. As
many philosophers put it, these mental states all have intentional content, and so may be
called ‘intentional states’ (see Searle 1983: esp. ch. 1).
(p. 422)
What does ‘normative’ mean? For the purposes of this chapter I hope that it will be
sufficient if we just start with the assumption that the paradigmatic normative terms are
‘ought’, ‘should’, ‘right’, ‘wrong’ and their closest equivalents in other languages. In
addition to terms like ‘ought’, ‘should’, ‘right’, and ‘wrong’, I shall assume that terms like
‘correct’, ‘incorrect’, ‘rational’, and ‘irrational’ are also normative terms.
What then does the claim that the intentional is normative amount to? The answer to this
question may depend, in part, on what is the correct sort of semantics for normative
concepts. First, suppose that a broadly truth‐conditional or factualist semantics for
normative concepts is correct. According to this sort of semantics, whenever propositions
involving these normative concepts are true, they are true in virtue of corresponding
normative facts; and it is convenient to suppose that these normative facts must involve
corresponding normative properties and relations.3 Then we can interpret the claim that
the intentional is normative as the claim that in giving an account of the essence or
nature of intentional states we must mention these normative properties or relations.4
Equally, in giving an account of the nature of an intentional fact—an account of what the
intentional fact consists in, as some philosophers would say—we must state some
normative fact. In that sense, intentional facts are partially constituted by normative
facts.5
On this understanding, then, the claim that the intentional is normative is a metaphysical
claim concerning the nature or essence of intentional states, or concerning (p. 423) the
constitution of intentional facts. It is not a semantic thesis about the meaning of
intentional terms, or a conceptual thesis about what is built into our concepts of
intentional states. Such semantic or conceptual theses could be true even if the
Page 2 of 19
The Normativity of the Intentional
The claim that the intentional is normative may have to be understood a bit differently if
an expressivist account of normative concepts (instead of a broadly truth‐conditional or
factualist account) is correct. According to the expressivist, although it is perfectly all
right for everyday conversation to talk about ‘normative facts’ or ‘normative properties
and relations’, no such normative facts, properties, or relations are capable of playing a
real explanatory role in an account of the essence or nature of any part of reality itself.
So, on this expressivist understanding of normative concepts, the claim that the
intentional is normative would seem to express a sort of anti‐realism about intentional
mental states. This sort of anti‐realism about the mental raises some very delicate and
difficult issues.6 In order to avoid getting embroiled in these issues, I shall just assume, at
least for the purposes of the rest of this chapter, that a broadly truth‐conditional or
factualist account of normative terms and concepts is correct.
If the claim that the intentional is normative is true, then it must be possible, at least in
principle, to give an account of the nature of our intentional mental states in partly
normative terms. In fact, such normative accounts of the nature of intentional states
could take many forms. For example, some of these accounts may claim that the
intentional is actually reducible to the normative; that is, that there is a correct reductive
account of the nature of intentional states, and any such account must mention normative
properties or relations (compare e.g. Brandom 1994).7 Other accounts may claim merely
that the normative and the intentional must each be mentioned in giving any account of
the other, so that the two domains of properties and relations are essentially
interdependent, without either of the two being reducible to the other. These normative
accounts could also differ in many other ways as well. In what follows, however, I shall
abstract from all these differences between these normative accounts of intentional
states. I shall also not take the time to explore the metaphysical and methodological
implications of the claim that the intentional is normative. Instead, I shall focus on
examining the reasons that philosophers have put forward in favour of this claim.
Anyway?
In the last two sections of this chapter I shall consider various arguments in favour of the
claim that the intentional is normative. But before doing that I shall consider a rather
abstract objection that can be made against this claim. Why should there be any
informative ‘account’ of the intentional at all? Perhaps intentional mental states are
wholly irreducible properties of thinkers, of which no informative ‘account’ can be given.
Some philosophers accept a strong form of naturalism, according to which it must be
Page 3 of 19
The Normativity of the Intentional
It seems to me that there is a problem with the doggedly quietist approach. This is
because of a plausible point about the relation between the intentional states that we are
in and our dispositions. (When I refer to our ‘dispositions’ here, I do not mean to focus
exclusively on our behavioural dispositions; I mean to include our mental dispositions as
well, such as our dispositions to revise our beliefs, intentions, and other attitudes, in
response to various conditions.) This point is an analogue of the point that is sometimes
expressed by the slogan ‘meaning is use’: just as the meaning of a linguistic expression is
determined by the ways in which the members of the relevant linguistic community are
disposed to use the expression in discourse, so too the identity of a concept is determined
by the ways in which we are disposed to use that concept in thought.
For example, suppose that a community has a concept in their repertoire that they are all
disposed to apply, quite systematically, to cats and only to cats; the experiences on which
they base their judgements involving this concept are precisely the sorts of experiences
that we would use to recognize something as a cat; and they treat everything that falls
under this concept as belonging to a single natural kind. Now consider the supposition
that in this case, this concept is in fact the concept of a bicycle, so that almost all of these
thinkers' beliefs involving this concept are irrational and false. This supposition appears
absurd (compare Fricker 1991). In general, it seems that if there is a community the
members of which possess a concept that they are disposed to use in their reasoning in
exactly the same ways as we are disposed to use a certain concept in our repertoire, then
their concept must be the same as ours. So it seems that there must be something about
the thinker's dispositions—together perhaps with certain facts about his environment and
the dispositions of other thinkers in his linguistic community—that determines the
identity of the concepts that figure in the thinker's thoughts.
(p. 425)
Similarly, if there is a community that has a type of attitude that they are disposed to form
and revise in just the same way as we are disposed to form and revise our beliefs, it
would be absurd to suppose that this attitude is in fact the attitude of disbelief—so that
almost all of their attitudes of this type are incorrect and irrational. Again, something
about our dispositions with respect to an attitude type must determine exactly which
attitude type it is that figures in our thoughts.
However, if it is true that facts about the thinker's dispositions determine which concepts
and attitude types figure in the thinker's thoughts in this way, then there surely must be
some account of the precise way in which these dispositions determine this (see
Boghossian 1991: 66). The bare claim that the thinker's dispositions determine which
Page 4 of 19
The Normativity of the Intentional
concepts and attitude types figure in the thinker's thoughts is compatible with infinitely
many different precise ways in which these dispositions might determine this. Surely
there must be some general principle that captures the specific role that these
dispositions play in determining this. Otherwise it will remain utterly mysterious why
these dispositions determine this in the specific way in which they do, rather than in any
of the infinitely many other possible ways.
So it seems that there must, at least in principle, be some explanatory account of how the
thinker's dispositions determine what intentional states he is in. There is no reason why
this account should take the fully reductive form of giving an account of intentional states
in terms of dispositions that are characterized in wholly non‐intentional, non‐normative
terms. On the contrary, there is no objection to characterizing these dispositions in
general intentional terms—that is, precisely as mental dispositions, dispositions to form
certain attitudes involving certain concepts in certain circumstances—so long as our
characterization of these dispositions does not presuppose the identity of the particular
concepts and attitude types that they involve. It may well be a completely primitive and
irreducible feature of me that I use concepts and have attitudes at all. However, it cannot
be an irreducible feature of me that I am using a concept that refers to cats (as opposed
to referring to dogs or to cows, for example); that must be explained in terms of the
dispositions that I have with respect to this concept.8
So the explanatory account that seems to be called for could take the following form. It
could start out by presupposing that we are dealing with a thinker who has attitudes
towards contents that are composed of concepts.9 But it would start by treating these
attitudes and concepts as initially uninterpreted. That is, it would start with a ‘neutral’
characterization of the thinker's intentional dispositions—a characterization of the
thinker's dispositions with respect to these attitudes and concepts that does not in any
way presuppose the exact identity of these attitudes and (p. 426) concepts. Then this
account could articulate certain principles concerning the nature of various attitudes and
concepts, which would enable us to identify the thinker's attitudes and concepts—say, as
the attitude of belief, or as the concept of water—on the basis of such a neutral
characterization of the thinker's intentional dispositions (along with further facts about
the thinker's natural and social environment). In what follows I shall assume that all
adequate accounts of the nature of our intentional mental states take this form.
Certain normative accounts of the intentional could certainly accommodate this point. For
example, such an account might consist of two parts. The first part would give an account
of the different attitudes and concepts in terms of certain normative principles governing
them. The attitude of belief, for example, might be identified as that attitude that it is
correct or appropriate to take towards a proposition just in case that proposition is true;
the concept ‘if’ might be identified as that concept for which certain sorts of inference
are rational (such as inferences that are instances of modus ponens); and similar accounts
could be given of other concepts and types of attitudes. Second, there would have to be
an account of what has to be true of a thinker if she is to possess those concepts or to be
capable of having those attitudes. Here the account could say, very roughly, that the right
Page 5 of 19
The Normativity of the Intentional
way to interpret the thinker must be the most charitable reading of her dispositions, the
way that makes thinkers' dispositions emerge as most rational and most sensitive to the
norms that apply to the intentional states that they are in.10
So a normative account of the intentional could satisfy the demand for an explanatory
account of how exactly thinkers' dispositions (together with their environment) determine
which intentional states they are in. But what reason is there to think that any such
account is true?
Some philosophers try to argue for the claim that the intentional is normative by pointing
out that facts about one's intentional states entail normative facts. Then they argue that
this entailment shows that the intentional is in itself essentially normative.11
(p. 427)
For example, the argument might be elaborated as follows. Necessarily, anyone who
believes p is committed to believing ‘p or q’. That is, it is a necessary truth, applying to
everyone, that one ought not simultaneously to believe p and consciously to refuse to
believe ‘p or q’ (that is, to consider whether or not ‘p or q’ is the case, and yet to fail to
believe ‘p or q’). But how else are we to explain why this is a necessary truth, unless this
truth logically follows from principles that are essential to or constitutive of all beliefs
whose content is of the form ‘p or q’? And if a normative truth follows from these
principles, mustn't these principles themselves mention normative properties or
relations?
One might start by questioning whether it really is true in general—let alone necessarily
true—that one ought not simultaneously to believe p and consciously to refuse to believe
‘p or q’. What if a powerful demon will destroy the world unless you simultaneously
believe p and consciously refuse to believe ‘p or q’ (see Byrne 2002)?
The best response to this sort of objection, I think, is to insist that there are different
sorts of oughts involved here. It is still a ‘mistake’ of some kind simultaneously to believe
p and consciously to refuse to believe ‘p or q’—even if the demon will destroy the world
unless you commit this mistake. But now the proponent of this argument must say more
about what exactly this sort of ‘ought’ means. It may be that it just means: ‘ought, in
order to achieve the goal of believing the truth and nothing but the truth (about any
question that one has actually considered)’. If so, then we can give a simple explanation
Page 6 of 19
The Normativity of the Intentional
Moreover, even if the sort of ‘ought’ that is involved here is not susceptible to any such
deflationary interpretation, we still do not have a sound argument for the normativity of
the intentional. As Paul Horwich writes:
[A] property may have normative implications without itself being normative. For
example, killing is prima facie wrong; none the less one can presumably
characterize ‘x kills y’ in entirely non‐normative terms. … It might be possible for
our most basic normative principles to have the form
(x)(Dx → Nx)
where ‘D’ describes some state of the world and ‘N’ specifies what ought or ought not to
be done in that situation. Thus the normative implications of a meaning property leave it
entirely open that its nature is completely non‐normative.
(1998: 188–93)
In this way, the ‘entailment’ of normative facts by facts about one's intentional states could be
explained by the existence of certain fundamental normative axioms. But there is no reason to
suppose that these normative axioms are constitutive of or (p. 428) essential to the intentional
states that they apply to. So the entailment argument seems not to be a persuasive argument for
the normativity of the intentional.
Some philosophers argue for the claim that the intentional is normative on the basis of
considerations concerning interpretation. This argument is based on two premises.
According to the first premise, the central or canonical way of ascribing mental states to
others is to attempt to ascribe those mental states that make best sense of them and of
their behaviour—where making best sense of them requires following a principle of
charity, ascribing those mental states that give the most charitable possible
interpretation of them and their behaviour (see esp. Davidson 1980: 221; 1984: 153).
Furthermore, according to this first premise, this principle of charity must be understood
in normative terms: to interpret someone charitably is to interpret her as forming beliefs
and intentions in a rational way, and as having at least a disposition to have experiences,
desires, and emotions that really are correct or appropriate, and so on.13
Page 7 of 19
The Normativity of the Intentional
According to the second premise, the canonical method of ascribing mental states to
others must proceed by attempting to track those facts about them that really make it the
case that they have the mental states in question. The first premise appears to imply that
this canonical method attempts to track certain normative facts about the subjects who
are being interpreted; namely, normative facts about what is the most charitable
interpretation of them. So it must be these normative facts that make it the case that they
have those mental states; and so a version of a normative account of the intentional is
true.
Both premises of this argument can be questioned. The second premise seems too quick
to assume that what is fundamental to our canonical method of coming to know some fact
must also be fundamental to the nature of that fact itself. The ‘canonical method’ could
still provide us with true beliefs about the mental states of others, even if it were only a
contingent fact about us that we tend to form beliefs and intentions in rational ways, and
to have experiences and desires that are correct or appropriate. In that case, the
normative facts that guide our interpretation would generally lead us to the truth about
the mental states of thinkers in the actual world, but they would not be in any way
constitutive of the mental states themselves.
The first premise could also be questioned. Perhaps the principle of charity can be
formulated in entirely non‐normative terms. For example, perhaps we can define which
interpretation is most charitable in purely logical terms, as the interpretation that makes
the subjects emerge as much as possible as believing the truth, and conforming to certain
standards of consistency and probabilistic coherence in their beliefs and preferences. Or
perhaps interpretation should follow not a principle of charity but a principle of humanity,
ascribing those mental states that make the subjects (p. 429) emerge as much as possible
like statistically normal human beings (see Grandy 1973). No doubt, if we have
independent reasons for accepting that the intentional is normative, then it will become
much more plausible that the canonical method of interpretation involves following a
normative principle of charity. But in the absence of such independent reasons we have
not been given much reason to accept the first premise of the argument. So the
interpretation argument seems dialectically ineffective as an argument for the claim that
the intentional is normative.
Many of the philosophers who claim that the intentional is normative try to support their
claim by arguing against the obvious alternative—what I shall call the ‘non‐normative‐
dispositionalist’ view, according to which the nature of our intentional states is
determined by features of our dispositions that can be specified in wholly non‐normative
terms.
Page 8 of 19
The Normativity of the Intentional
Here I shall just discuss the two most challenging objections that any non‐normative
dispositionalist theory must face. Both objections stem from the point that if the nature of
a given concept is determined by our dispositions to form beliefs and other attitudes
involving that concept, then those dispositions must surely also fix the extension or
reference of that concept. But it may seem that there are at least two reasons why our
dispositions cannot fix the extension of our concepts.
First, as Kripke famously insisted, it appears, at least offhand, that our dispositions are
finite, whereas the extension of many concepts—at least if we include their extensions
across all possible worlds—is infinite (1982: 26). For example, the extension of the symbol
‘+’ is a function whose domain and range are both infinite. In consequence, the
arguments and values of this function include numbers that are so vast that I have no
dispositions whatsoever with respect to them. So how can my finite dispositions for using
this symbol fix on a unique infinite function—the addition function—out of all the
infinitely many such functions that there are?
Second, we all have dispositions to make mistakes. Our mistakes are not random; on the
contrary, our mistakes are typically quite systematic. The non‐normative dispositionalists
face a dilemma here. On the one hand, they could be ‘holists’ and claim that all our
dispositions for using a concept are equally involved in determining the concept's
extension. Alternatively, they could reject holism and try to draw some distinction
between those dispositions that fix the concept's extension and those that do not. But
according to Kripke, the first horn of the dilemma is hopeless. It is natural to assume that
if one's dispositions for using a concept determine the concept's extension, then they
must determine it in such a way that one's disposition to form beliefs involving the
concept in certain circumstances guarantees that those beliefs are true in those
circumstances. But if all one's dispositions for using a given concept are equally involved
in determining the concept's extension, then one cannot be disposed to form false beliefs
at all. Indeed, if it is (p. 430) true, as some philosophers assume, that however one uses a
concept, one must have been disposed to use it in that way, then as Kripke put it this
‘holistic’ dispositional account amounts to ‘an equation of performance and
correctness’ (1982: 21). Thus, the first horn of the dilemma seems to lead to the absurd
Protagorean claim that all beliefs are true.
However, the second horn of the dilemma looks almost equally daunting. How exactly is
the non‐normative dispositionalist to distinguish between the extension‐determining
dispositions and all the other dispositions that we have with respect to a given concept?
Perhaps if we could appeal to a normative distinction—say, between rational and
irrational dispositions—we could answer this challenge. But it is quite unclear how to
formulate this distinction in wholly non‐normative terms, without simply presupposing
what the extension of the concept is.
Page 9 of 19
The Normativity of the Intentional
Until recently many philosophers assumed that dispositions can be analysed by means of
counterfactuals: according to this sort of analysis, to be disposed to ϕ in circumstances C
is to be such that if one were in circumstances C one would ϕ. But recently this view has
been shown to have many counter‐intuitive consequences (see for example Bird 1998;
Fara 2005). Dispositions are more plausibly regarded as being closer to ceteris paribus
laws, where laws may be regarded as nomic connections between properties. For an
object to have a disposition is for it to fall within the domain of the ceteris paribus law in
question. If the law is a connection between properties, then the disposition is just as
‘infinite’ as those properties are. So, for example, the relevant disposition with respect to
the concept ‘+’ may be the disposition to accept an inference precisely in response to the
fact that one has considered an inference of a certain general form. For example, in the
case of the concept ‘+’, perhaps the relevant dispositions are as follows. First, one is
disposed to respond to the state of considering any proposition of the form ‘n + 0 = n’ by
accepting that proposition. Second, one is disposed to respond to the state of considering
any inference of the form ‘n + m = k; therefore, n + m′ = k′’ (where n′ is the successor of
n) by accepting that inference. Now these ‘forms’ are exemplified not just by those
propositions and inferences that one is capable of considering but by infinitely many
propositions and inferences.14 It is, one might claim, these general forms, to which one's
disposition responds, that fix the extension of the concept.
Regarding the second objection, several different responses are possible. First,
proponents of the ‘holistic’ form of non‐normative dispositionalism may reject Kripke's
assumption that if our dispositions to use a concept determine a concept's semantic
value, then our disposition to form a belief involving the concept in certain (p. 431)
circumstances must guarantee that the belief is true in those circumstances. Holists
influenced by Davidson might say that the correct interpretation of the concept is the one
that ‘maximizes’ the true beliefs that the thinker is disposed to form using that concept;
but it does not have to guarantee that every belief that the thinker is disposed to form
using that concept is true (see Davidson 2001: 137).
Page 10 of 19
The Normativity of the Intentional
these reasons, then, this Kripke‐inspired argument also seems not to be a persuasive
argument for the normativity of the intentional.
Suppose that one of your dispositions with respect to the use of a concept F is a
disposition to use the concept in an irrational way, such as a disposition to engage in
some fallacious form of reasoning. Then the holist would have to admit that it is partly in
virtue of this disposition that you possess the concept F. This disposition is part of what it
is about you that makes it the case that you possess the concept F. In effect, this
disposition is an essential part of what realizes your possession of F.16
(p. 432)
Thus, it seems to me that if one possesses a concept, the dispositions in virtue of which
one possesses it must be rational dispositions. In general, the dispositions that determine
the identity of the concepts (and of the types of attitude) that figure in one's thoughts
must be rational dispositions. More specifically, it may be that the disposition that
determines the identity of a concept that figures in one's thoughts is whatever is causally
the most basic of one's rational dispositions with respect to the concept. Thus, if we are
told that there is a concept C that figures in a thinker's thoughts, and we are also told (in
terms that do not presuppose the identity of C) what is the most basic of the rational
dispositions that the thinker has with respect to C, then this information will be enough to
determine which concept C is.
This makes it plausible, it seems to me, that for each concept there is some specific
rational disposition that is essential to possessing the concept. For example, in the case of
the concept ‘if’, it seems impossible for there to be a thinker who possesses the concept
‘if’ without having any disposition for rationally accepting modus ponens inferences
Page 11 of 19
The Normativity of the Intentional
(whereas it certainly seems possible for there to be superhuman thinkers who possess the
concept ‘if’ but have no disposition to engage in any fallacious forms of reasoning
involving this concept). If this disposition really is essential to possessing the concept,
then it will be impossible for any thinker to possess the concept without possessing it in
virtue of having this disposition. As I have argued, the dispositions in virtue of which we
possess a concept must be rational dispositions. So if a disposition is essential to
possessing a concept, it must be impossible for this disposition not to be a rational
disposition: it must be an essentially rational disposition.
Now some philosophers may think that this still does not show that an account of what is
essential to a given concept will have to mention normative properties or relations (such
as the properties or relations that are referred to by the term ‘rational’). According to
these philosophers, it is possible for such an account to single out the relevant essentially
rational disposition by some purely non‐normative feature—where this other feature
guarantees that the disposition in question is a rational disposition, but is not equivalent
to its being a rational disposition.
(p. 433)
However, there is a reason for thinking that these dispositions cannot in fact be specified
without mentioning normative properties or relations.18 It seems plausible that the
relevant rational dispositions are all dispositions to form or revise one's mental states in
some rational way. In a broad sense of the term, they are dispositions for certain forms of
rational reasoning. For example, one sort of rational reasoning might lead one from
having a visual experience that presents an object in a certain distinctive way (in the
absence of any positive reason to believe one's experiences to be unreliable) to one's
forming a belief that predicates the concept ‘… is yellow’ of the object in question. A
disposition for this sort of reasoning might be essential to possessing the concept ‘… is
yellow’. Other sorts of rational disposition might be dispositions to accept rational
deductive or inductive inferences, or to engage in rational sorts of practical reasoning.
For example, one simple sort of rational practical reasoning might be the kind that leads
from having a rational belief of the form ‘All things considered, I ought to ϕ’ to forming an
intention to ϕ. It may be that a disposition for this sort of reasoning is essential to
possessing the concept that is expressed by the relevant kind of ‘ought’.
As these examples suggest, there seems not to be any way to specify the relevant sorts of
rational reasoning without mentioning normative properties or relations. The
specification that I have just given of the sort of reasoning that leads from visual
experiences to beliefs of the form ‘x is yellow’ included a proviso ‘in the absence of any
positive reason to believe one's experiences to be unreliable’; in specifying this sort of
reasoning in this way, I mentioned the normative relation of having a reason to believe
something. Similarly, my specification of the sort of reasoning that leads to an intention to
ϕ is restricted to cases in which one has a rational belief that all things considered one
Page 12 of 19
The Normativity of the Intentional
ought to ϕ; in specifying this type of reasoning in this way, I have mentioned the
normative relation of rationally believing something.
The reason for specifying these sorts of reasoning in this normative way is clear. These
sorts of reasoning, like most if not all others, are defeasible. That is, even if one is in the
‘input’ mental states for this sort of reasoning, certain defeating conditions must be
absent if it is to be possible to form a rational belief or intention by means of this form of
reasoning. But the nature of defeating conditions is precisely that they are those
conditions that make it irrational in the circumstances to form the belief or intention in
question in response to these input mental states. There seems to be no way of specifying
these defeating conditions without mentioning normative properties or relations.
For a sort of reasoning to be essentially rational, then, it seems that its input conditions
must include the absence of defeating conditions as such. It would not be enough if its
input conditions included only the absence of certain defeating conditions, but not the
absence of defeating conditions as such. Then there would be cases (p. 434) in which this
sort of reasoning would be irrational: in these cases, this sort of reasoning would involve
treating certain conditions as if they constituted a reason for a certain revision to one's
beliefs or other attitudes when in fact, on account of the presence of defeating conditions,
they did not. Equally, it would also not be enough if this sort of reasoning were markedly
oversensitive, so that its input conditions included not just the absence of defeating
conditions but also the absence of various additional factors that were in fact quite
irrelevant to the rationality of the sort of reasoning in question. Then this sort of
reasoning would be irrational in a different way: it would involve treating certain
conditions as relevant when they are not in fact relevant at all.
For this reason it seems that these essentially rational forms of reasoning cannot be
specified without mentioning the absence of defeating conditions as such—which in effect
involves mentioning certain normative properties and relations. But, as I have argued, the
dispositions that are essential to possessing concepts, and to being capable of the various
types of attitude, must be dispositions to engage in such essentially rational forms of
reasoning. So it seems that the dispositions that are essential to possessing concepts (and
to being capable of attitude types) cannot be specified without mentioning normative
properties or relations.
Some philosophers might suggest that one could still be disposed to reason in accordance
with these essentially rational forms of reasoning even if the normative properties or
relations do not have to be mentioned in specifying these dispositions themselves. For
example, one might have a host of separate dispositions, such that the net effect of all
these separate dispositions is that one forms and revises one's beliefs and other attitudes
in just the same way as one would if one had a disposition that actually responded to the
normative facts themselves. However, if all these dispositions really are quite separate
and causally independent of each other, then it seems that it could just be a fluke that the
net effect of all these dispositions is that one forms and revises one's attitudes in a
rational way. But in that case it actually seems to me quite doubtful whether the
Page 13 of 19
The Normativity of the Intentional
manifestation of these dispositions can really count as rational thinking at all. Indeed, this
case seems in effect to be the case that was originally described by Ned Block and has
since come to be known as the ‘Blockhead’, which seems not be engaged in rational
reasoning at all.19
So it seems most plausible that if there are any dispositions that are essential to
possessing concepts (and to being capable of attitude types), these dispositions must
respond precisely to normative facts; hence, these dispositions cannot be specified
without mentioning such normative properties or relations.
Some other philosophers might insist that it is still bound to be possible, at least in
principle, by using some highly complex psychological terms, to specify these essentially
rational forms of reasoning in wholly non‐normative terms. But I do not have to deny
this.20 I have said that we cannot specify these rational dispositions without (p. 435)
mentioning normative properties or relations; I have not said that we cannot specify these
rational dispositions without using normative terms. It may well be possible, in principle,
to mention these normative properties or relations by means of some complex
psychological terms. Even if these normative properties or relations can be specified in
purely psychological terms, it would still be an interesting and surprising fact that any
adequate account of the nature of intentional mental states has to mention these
normative properties or relations—so long as these normative properties or relations can
only be specified by means of highly complex psychological terms. This is because it
would still be an interesting and surprising fact if the property or relation referred to by
this particular complex psychological term has to be mentioned in any adequate account
of the nature of intentional mental states.
It certainly seems that if there are any purely psychological terms that refer to the
absence of defeating conditions, they will be immensely complex. Defeating conditions,
after all, can come from anywhere: from experience, from memory, from the testimony of
others, or even from elaborate deductive reasoning or scientific theorizing. So the only
simple way of specifying what these conditions all have in common is in normative terms
—as conditions that make it irrational to reason in certain ways. Even if there is also
some more complex way of specifying these defeating conditions in non‐normative terms,
my argument would not be undermined if the only simple way to specify the dispositions
that are essential to possessing certain concepts, or to being capable of certain attitude
types, is by using normative terms.
As I argued in Section 25.2 above, there must be some account that explains how the
thinker's dispositions determine which concepts and attitude types figure in his thoughts.
As I have argued in this section, the prospects for a purely non‐normative
dispositionalism are dim. So since some sort of dispositionalism must be true, then (at
least so long as it seems that a normative sort of dispositionalism could avoid the
problems that other theories face) we have a good reason to accept the claim that I have
been considering here: the claim that the intentional is normative.
Page 14 of 19
The Normativity of the Intentional
References
Bird, A. (1998), ‘Dispositions and Antidotes’, Philosophical Quarterly, 48: 227–34.
Fodor, J. A. (1990), A Theory of Content and Other Essays (Cambridge, Mass.: MIT Press).
Grandy, R. (1973), ‘Reference, Meaning, and Belief’, Journal of Philosophy, 70: 439–52.
Page 15 of 19
The Normativity of the Intentional
Kratzer, A. (2002), ‘The Notional Category of Modality’, in P. Portner and B. Partee (eds.),
Formal Semantics: The Essential Readings (Oxford: Blackwell), 289–323.
Kripke, S. (1980), Naming and Necessity (Cambridge, Mass.: Harvard University Press).
Rosen, G. (1997), ‘Who Makes the Rules Around Here?’, Philosophy and
Phenomenological Research, 57: 163–71.
Notes:
(1) For example, Robert Brandom argues that normative truths are in some sense
‘imposed’ by ‘human practices’ in his book Making It Explicit (1994: 46–50), whereas I
would myself insist that there must be some fundamental normative principles that are
simply necessary truths, and so wholly independent of all contingent facts about human
practices. But I shall have to remain neutral about these philosophical controversies here.
(2) Indeed, I have attempted to give such a unifying account elsewhere. Roughly, my idea
is that what unifies all these normative concepts is that it is a constitutive feature of these
concepts that they play a regulative role in certain rational practices (see Wedgwood
2002: esp. 268–9); for a more detailed account of a particular normative concept see
Wedgwood (2001).
Page 16 of 19
The Normativity of the Intentional
of the intentional that we are focusing on here. For our purposes, speaking about
properties and relations is useful for achieving generality; but it could in principle be
dispensed with if it were necessary to do so.
(4) To fix ideas, we may think of an ‘account of the essence or nature’ of something in
roughly the same way as Kit Fine would conceive of a statement of the ‘constitutive
essence’ of the thing (1995: 56).
(5) Compare the interpretation of this claim given by Gideon Rosen (2001).
(7) At least, this is how Gideon Rosen interprets Brandom in Rosen (1997). See also
Morris (1992).
(8) In effect, an account of what it is for a thinker to be using the concept cow may
employ both the term ‘concept’ and the term ‘cow’, but it may not employ the phrase ‘the
concept cow’ or ‘refers to cows’ or any variant thereof. It should conform to Christopher
Peacocke's constraint, that an adequate account of a concept should take ‘the A(C)
form’ (1992: 5–16). Compare Saul Kripke's insistence that an account of why a given
name refers to a particular individual must not be ‘circular’ (1980: 68–70).
(9) It may also be necessary to start out with presuppositions about the syntactic
category that the various concepts belong to—whether they are general or predicative
concepts, or singular concepts, or propositional operators, and so on.
(10) For the idea that interpretation must be guided by a principle of charity see Davidson
(2001). A normative theory could avail itself of this idea without accepting Davidson's full‐
blown ‘interpretivism’. As David Lewis suggests (1974), the reference to interpretation
could just be taken as a way of dramatizing what is objectively constitutive of the
intentional states in question.
(11) This is very close to an argument that Kripke used to argue for the normativity of
meaning (1982: 11, 21, 24, 37).
(12) This objection, like the next, is due to Horwich (1998: ch. 8).
(13) One philosopher who insists that the principle of charity (at least as it applies to the
ascription of preferences) must be interpreted in evaluative terms is Hurley (1989: ch. 6).
Page 17 of 19
The Normativity of the Intentional
(14) For this response to the ‘finitude’ objection see Peacocke (1992: ch. 5). It may
genuinely be a certain property that one's disposition is responding to, even if one's
dispositions are not perfect and do not respond to that property in absolutely every case.
(15) This seems to me a much more promising route than the route that Paul Boghossian
assumes the dispositionalist will take, of focusing on the thinker's dispositions under
‘optimal conditions’ (see Boghossian 1989).
(16) More precisely, if the holist is right, it is essential to possessing the concept F that
one's possession of the concept F supervenes on one's total set of dispositions for using
the concepts in one's repertoire; and the minimal supervenience base for possessing F
that you actually instantiate is one that includes your having this irrational disposition.
This seems to imply that your having this irrational disposition is part of what realizes
your possession of this concept.
(19) Block's original example was of a machine that passes a ‘Turing Test’ simply by
means of a huge ‘look‐up table’ (1978: 294–6).
(20) I do not have to deny this; but there may still be reasons for denying it. For an
argument for the claim that the conditions of rational belief and rational choice are
‘uncodifiable’ see Child (1993).
Ralph Wedgwood
Page 18 of 19
Concepts and Possession Conditions
The first virtue of the rough characterization of concepts in terms of ways of thinking of
objects and properties, and their role in that-clauses, is that it highlights the relation
between concepts and reference. The second virtue of the initial characterization is that
it establishes the prima facie relevance of what has come to be called Frege's Principle in
the individuation of concepts. A third virtue of the initial characterization is that it brings
out a phenomenon whose significance is insufficiently appreciated. There is a
phenomenon of productivity for thought about mental states that is just as striking as the
original phenomenon of the productivity of conceptual thought about the non-mental
world.
Keywords: thinking of objects, concepts and reference, Frege's Principle, individuation of concepts, conceptual
thought, mental state
26.1 Introductory
THE large‐scale questions about concepts are the same as those one would ask about any
other apparent category of apparent entities, to wit:
Page 1 of 22
Concepts and Possession Conditions
Unlike other domains, however, almost every philosophical claim about concepts is
controversial. Perhaps the only two uncontroversial propositions in this territory are (a) that a
good theory of concepts is fundamental to a wide range of philosophical issues and (b) that
proposition (a) is about the only philosophical proposition about concepts on which there is
general agreement. The disagreement exists because the nature of concepts is so closely tied to
fundamental issues in the theory of thought, in metaphysics, and in epistemology. In this area
the implied costs of error are high; and so are the potential rewards of correctness. I will
highlight a number of open questions in the philosophical theory of concepts. To say that they
are open is consistent (p. 438) with important contributions having been made to their
resolution: it is just to say that they remain far from settled. To avoid a thicket of numbers,
years, and references, all bibliographic information and citations are given at the end of the
paper, section by section.
Concepts are commonly introduced as ways of thinking of objects, properties, or anything
else on whose identity turns the truth or falsity of attributions of attitudes in that‐clauses.
It can be true that
Hence, it is said, under this mode of introducing concepts, the concepts elephant and long‐
trunked pachyderm are distinct concepts. They are different ways of thinking of the same
animals. The same applies, by the parallel reasoning, to establish the distinctness of the pairs of
concepts pain and C‐fibre activation; 3 p.m. and now (even when used at 3 p.m.); and so forth.
This mode of introduction via ways of thinking and that‐clauses needs several
refinements, but it also has important virtues. The first refinement it needs is something
that takes account of the fact that there can be quite a loose connection between the
concept someone is actually employing in thought, and the meaning of the words (even in
context) that are used in a true attribution of propositional attitudes to the thinker.
Sometimes the words in the language do not obviously express a specific concept, and
have only something at the level of reference associated with them. Almost everyone
believes this for proper names of persons and of cities, for instance. Such words can still
be used in attribution of attitudes. Many recent writers, from quite varied theoretical
backgrounds, have held that
x believes that Fa
means roughly
There is some concept α of a, and there is some concept ϕ of F, such that x believes ϕα.
Further refinement is needed to take account of the fact that we commonly use that‐clauses in
attributions of states where the content is plausibly non‐conceptual. It trips off the tongue to say
that
The cat sees that the mouse has gone behind the sofa.
Page 2 of 22
Concepts and Possession Conditions
But if we draw a distinction between conceptual and non‐conceptual content, and think that cats
are capable only of states with non‐conceptual content, we have some work to do. Introducing
concepts via that‐clauses will not relieve us of the non‐trivial, and interesting, task of saying
what is distinctive of conceptual content, as opposed to non‐conceptual content. I call this ‘the
characterization problem’. When we have a solution to the problem, we can use it to cut down
the class of that‐clauses that are pertinent to the introductory explanation of concepts. If we do
well in addressing the characterization problem, we may be able to bypass mention of that‐
clauses altogether. After all, the characterization in terms of that‐clauses is a (p. 439) linguistic
characterization of something that is not at all obviously fundamentally linguistic. One may hope
to dispense with any reference to language to public language at least—when we have a more
theoretical characterization of conceptual content before us.
The first virtue of the rough characterization of concepts in terms of ways of thinking of
objects and properties, and their role in that‐clauses, is that it highlights the relation
between concepts and reference. On this intuitive, initial characterization, it is part of the
nature of a concept that there is an answer to the question: What relation has to hold
between the concept and something for the latter to be the former's reference? On
certain neo‐Fregean views, the answer to this question actually individuates the concept.
But on any view that has this mode of introduction as its starting point, the question must
have an answer, whether individuative of the concept or not. The issue of the correct way
of conceiving of the relation between a concept and its reference is, I will be arguing
later, of enormous consequence for the philosophical theory of concepts.
The second virtue of the initial characterization is that it establishes the prima facie
relevance of what has come to be called Frege's Principle in the individuation of
concepts. Concept A is distinct from the concept B iff a thinker can rationally judge some
complete (truth‐evaluable) content containing the concept A without judging the
corresponding content containing B. Were this not so, we could not use distinctness of
concepts to explain, or at least to characterize, the difference between the belief with the
content concerning elephants and the belief with the content concerning long‐trunked
pachyderms. If concepts slice at least this finely, it will, apparently, be an obligation of a
substantive theory of concepts to explain certain facts about informativeness and
uninformativeness concerning particular concepts.
A third virtue of the initial characterization is that it brings out a phenomenon whose
significance is insufficiently appreciated. There is a phenomenon of productivity for
thought about mental states that is just as striking as the original phenomenon of the
productivity of conceptual thought about the non‐mental world. If a thinker possesses the
concepts for forming the Thought that p, and has the concept of the propositional attitude
ψ, then he is capable of judging contents of the form
x ψ s that p.
This holds for arbitrary contents p and attitudes ψ, whatever the constituents of p, and however
complex p may be. This is something that calls for a correspondingly general explanation. It
could be called ‘the second‐level productivity challenge’.
Page 3 of 22
Concepts and Possession Conditions
The intuitive way‐of‐thinking and that‐clause mode of introducing concepts are very far
from determining what concepts are. Very different kinds of entity, and in some cases
fundamentally none at all, have been proposed as concepts by theorists who would accept
these modes of introducing concepts. There are broadly three very different kinds of
theory consistent with such approaches.
(1) On theories of the first kind, concepts are abstract objects. This was the classical
Fregean thesis: the concepts are amongst the inhabitants of the third realm, distinct,
according to Frege, from both the physical and the mental. Frege said very (p. 440) little
about what it is for a thinker to stand in a relation to an element of the third realm. The
view of concepts as abstract objects does not need to have them standing in such proud
isolation from thinkers as, apparently, Frege did (a). A concept considered as individuated
by the condition a thinker must meet to possess the concept can still be an abstract
object; but manifestly under that conception concepts are not individuated independently
of their possible relations to thinkers (b). If there are such things as object‐dependent
concepts, or property‐dependent concepts, by whatever reasoning, these will be what
Quine called impure abstract objects, individuated in part by their relations to what they
are concepts of (c). A theorist could also reject the idea that concepts are individuated by
possession conditions, but still think they are individuated by some other relation to
thinkers. You do not have to believe that concepts are individuated by possession
conditions to reject Frege's conception of the third realm as utterly independent of
thinkers.
(2) According to theories of the second kind, concepts are mental representations. Mental
representations are particulars located in time and space. These representations may
have syntactic properties, and will have a variety of semantic properties, including
conceptual content. Theorists of this second stripe are rarely concerned only with
syntactic properties. They are commonly eager to acknowledge that two thinkers may
have different mental representations of different syntactic types, but with the same
conceptual content.
as
John stands in the belief‐relation to the Thought elephants can swim.
Page 4 of 22
Concepts and Possession Conditions
But, on this view, there is no more to concepts than is legitimized by the underlying true
sentences that legitimize talk of concepts. Talk of concepts is, to that extent, like talk about the
average resident of New York and that resident's properties. It is a legitimate grammatical
fiction; it is nothing more than that.
What is the relation between these three styles of approach? Now certainly there is
nothing that is at once both an abstract object, and a mental particular, and a
grammatical fiction. All the same, these three kinds of view should not be seen as in every
case competing with the others. A theorist can recognize both the abstract and the
mental representation views as legitimate classifications, each appropriately cited in
explaining different phenomena. When, for example, we are concerned with timeless
facts about what is informative in what circumstances, that is best explained by (p. 441)
concepts conceived at the abstract level. The properties of particular mental
representations are irrelevant to their explanation. When we are concerned with the
contingencies of the interaction of particular mental representations in a particular type
of mind, or in a particular species, or in a given individual, those contingencies will not be
present in all possible mental representations with a given conceptual content.
Correspondingly, it would be quite inadequate to try to cite only conceptual content
considered on the abstract model in their explanation.
The third, pleonastic, conception is a radically different animal from the other two. It
seems clear that the idea of concepts as having any explanatory power is one that the
pleonastic conception must in a wide range of cases reject. This is acknowledged by its
proponents, and I will return to consider this position separately.
Page 5 of 22
Concepts and Possession Conditions
In the reverse direction, Frege's criterion is not a general law applicable to all entities. It
is something specific to concepts that if C and D are distinct concepts, then there must be
some circumstances or other and some frame A( ) or other, such that (p. 442) in those
circumstances a thinker can rationally judge A(C) without rationally judging A(D).
For any given concept there is a distinctive pattern that captures certain facts about
rationality that involve the concept. It is distinctive of the concept red that a shade in a
certain range, when experienced by the thinker in ordinary circumstances as the shade of
a perceived object, makes it rational for the thinker to judge ‘That thing is red’ without
any intervening inference. Experiencing a pain can make it rational for a thinker to judge
‘I'm in pain’ without any inference; whereas the same circumstances do not by
themselves, without reliance on an inference, make it rational for the subject to judge
‘The person living at 27 Elm Street is in pain’ (not even if he is the person living at 27
Elm Street, for that rationally requires the further premise ‘I am the person living at 27
Elm Street’). If the thinker already accepts that p and that q, that can make it rational
without further premises to conclude that p & q. That does not make it rational, without
further inference, for the thinker to conclude that ∼(∼p v ∼q). We can summarize the
distinctive facts about a concept's involvement in rationality in what we can call the
concept's rationality profile. Take a concept C, and consider all the pairs <S, A(C)> such
that in circumstances S it is rational for the thinker to judge A(C), without any
intermediate inferences or transitions, if he possesses the concept C. The set of all such
pairs I call the rationality profile of the concept C. We can also speak of the rationality
profile of sets of concepts. The rationality profile of the pair of concepts C, D is the set of
all pairs <S, A′(C, D)> such that in circumstances S it is rational for the thinker to judge
A′(C, D), without any intermediate inferences or transitions, if he possesses the concepts
C and D. So the rationality profile of the pair of concepts pain and the first‐person concept
I includes the pair <the thinker is in pain, the Thought I am in pain>. The point just made
earlier was that the rationality profile of pain and the person living at 27 Elm Street does
not include the pair <the thinker is in pain, the thought the person living at 27 Elm Street
is in pain>.
Is there an explanation of why a given concept, or pair of concepts, has the rationality
profile it has? Is there an explanation of why sets of concepts have the rationality profile
they do? It is better to have some explanation of such facts if we can attain it. Besides the
virtue of achieving explanation of the eternal, timeless facts of rationality profiles, there
is also a gain in psychological explanation if we can explain the rationality profile. The
rationality profile is an open‐ended, and arguably infinite, set. If we can provide an
Page 6 of 22
Concepts and Possession Conditions
explanation of the profile, we will be on the way to explaining how finite thinkers can
attain rationality (in so far as they can). We will also be on the way to explaining how
thinkers can reliably classify some judgements as rational in the relevant circumstances
and others as not.
Can all aspects of rationality specific to a given concept or to a given type of concept
be explained in terms of the nature or that concept, or of that type of concept?
Here I mention open question 1, which is concerned with exhaustion, partly just to distinguish it
from the goal of inclusion, according to which there are some facts about rationality specific to a
concept that we can explain from the nature of a concept, even if not all such facts are thus
explicable. To aim for inclusion seems to be a modest, reasonable enterprise.
A content is known only if it comes to be accepted by rational means. If concepts have a
link with rationality, they are thereby pertinent to epistemology too. Open question 2 is:
Some theorists go so far as to say that some concepts should be individuated in terms of the
conditions under which certain contents containing them are known. One way of elaborating
that approach is to hold that outright judgement aims at knowledge, so that individuating
concepts in terms of certain judgements is necessarily individuating them in terms of
contributions to conditions for knowledge. Open question 2 must include consideration of
whether the connections between concepts and knowledge are deep or superficial, and, in either
case, what their source is.
When we talk about an explanation of facts about rationality or knowledge drawing upon
the nature of a concept, what do we, or what should we, mean? The nature of a particular
concept will be given by a theory that says something specific about that concept, not
Page 7 of 22
Concepts and Possession Conditions
For empirical concepts, we can consider how the model might be developed for Frege's
example of Afla and Ateb. ‘Afla’ is introduced to refer to a mountain seen from one
direction in one location, ‘Ateb’ for a mountain seen from an entirely different location.
The thought ‘Afla = Ateb’ is informative; it would not be in the rationality profile for
either the concept Afla, or the concept Ateb, or for the pair of them. ‘That's Afla’ would be
in the rationality profile for Afla when the demonstrative expresses a perceptual‐
demonstrative concept of a mountain seen from the location at which the name was
introduced. By a series of individually obvious steps, in this case using empirical
information, one way of coming rationally to reach the truth that Afla = Ateb is to keep
track of the mountain while travelling. At successive times the thinker makes judgements
of the form ‘This is still Afla’ as he travels. Eventually our thinker reaches a judgement
‘This is still Afla’, where this most recent demonstrative expresses a perceptual‐
demonstrative concept for which ‘This is Ateb’ is in the rationality profile of the concept
Ateb. If our thinker has come to learn ‘This is Afla’ and ‘This is Ateb’ (for the same
demonstrative concept and context), he is in a position to infer ‘Afla = Ateb’.
For any case to which the model of primitive and derived rules is to be applied, one must
identify some plausible primitive rules, and explain how they account for those parts of
the rationality profile of the concept that are subject to the explanation. If the concept
has primitively rational uses beyond, for instance, the perception‐based cases, the
possession condition must be expanded to cover them too (and their relation to the other
primitive cases must be satisfactorily explained). Various observational concepts, logical
concepts, and certain sensation concepts have been accorded treatments of this kind.
Page 8 of 22
Concepts and Possession Conditions
Even for the more limited goal of explaining inclusion, rather than exhaustion, there are
at least three problems with the primitive/derived model of the nature of concepts as so
far described, and they all raise deep issues.
First, there are many examples of concepts which, it seems, cannot be fitted into the
primitive/derived‐rule mould. Some concepts, and perhaps even some logical concepts,
are such that all the primitive rules that are correct for them seem also to be informative.
It seems that a thinker has to reflect, to work out that they are correct. This seems to be
true even for so fundamental a concept as that of negation. The classical natural‐
deduction systems give an introduction rule for negation, commonly some form of
reductio, and some form of double‐negation elimination as the elimination rule. Both of
these are principles one has to reflect upon to appreciate their correctness. It is not at all
like the acceptance of ‘That's Afla’ in the canonical (p. 445) circumstances of the name's
introduction, or like ‘That's red’ where the perceptual demonstrative picks out an object
given as having what is definitely a shade of red. Whatever the nature of this transition to
a rational appreciation of these primitive laws about negation, in making the transition
the thinker certainly seems to draw on his grasp of the concept of negation. It follows
that understanding of negation should not be identified with willingness to accept
instances of those rules. At the very least, grasp of negation explains rational acceptance
of those rules (and nothing can explain itself).
Second, there is what I have called ‘the phenomenon of new principles’. It seems that
sometimes a thinker can come, by rational reflection involving a concept he grasps, to
accept principles that do not follow from primitive principles he already accepts, by
inferential principles he already accepts. That no natural number has infinitely many
predecessors is arguably one such principle. Eventual acceptance of the epsilon‐delta
style definition of the limit of a series of fractions is another. This definition was
informative for those who had been using the notion of a limit for decades. The definition
was not even know to Newton and Leibniz, the originators of the notion of the limit of a
series. When there is such rational acceptance of a new principle, we need quite a
different account of the identity and possession of a concept than is given in the
primitive/derived‐rule model. One resource here is the idea that a thinker's grasp of a
concept consists in tacit knowledge of some condition governing the concept. The
reflective thinker draws on this tacit knowledge, or implicit conception, in classifying
cases. But to articulate the content of the tacit knowledge explicitly, the thinker must
Page 9 of 22
Concepts and Possession Conditions
engage in abductive inference from cases. This is equally what we have to do when we
succeed in articulating syntactic and semantic rules of which we have tacit knowledge.
The third problem with the primitive/derived‐rule model as so far described is that it fails
to address a question that any positive theory of concepts must address. The primitive/
derived‐rule model is attempting to account for facts about rationality. But what is it for a
judgement to be rational in given circumstances? What is the aim that is furthered by
judging in accordance with primitive and derived rules for a concept? Even if there are
some concepts for which the model of primitive and derived rules fully captures the facts
about rationality specific to each of those concepts, we are still lacking an explanation of
why this is sufficient for judgements made in accordance with those rules to be rational.
It is here that we have to face the fundamental question of the relation between concepts
on the one hand and truth and reference on the other.
(1) There is the classical view that a substantive conception of truth and reference plays
an ineliminable part in individuating conceptual content. On this view, if a concept is
individuated on the primitive/derived‐rule model it is a real, non‐trivial task to show that
judging in accordance with those rules furthers the goal of judging only what is true.
(2) There is the view that although it is correct to accept the proposition that rational
judgement aims at truth, this states no substantive constraint on a theory of conceptual
content. There are many different varieties of this non‐substantive view. One variety holds
that rationality consists in judging in accordance with certain rules or conceptual roles, of
one sort or another. Under views of this sort there is no genuine task of showing that
judging in accordance with the rules leads to true judgements. This type of view is
commonly combined with a minimalist theory of truth in general, one that aims to say
that there is no more to truth than can be explained in terms of a careful elaboration of
the disquotational schema. (If more than that were involved in truth, there would still
remain a substantive task of explaining the rationality of the conceptual roles and rules
that are used in individuating concepts.) Conversely, of course, those who accept
minimalist theories of truth and reference will agree that constraints involving reference
and truth cannot play the same role in ratifying rules as rational that they do on the
classical view. Variants of such a minimalism are held by such different thinkers as
Brandom, Field, Harman, and Horwich.
Page 10 of 22
Concepts and Possession Conditions
Not all theorists accept even that truth is an aim of judgement. Some have said that the
aim of assertion is warranted assertibility. A corresponding position in the theory of
thought holds that judgement aims at warranted judgement. Sometimes this position has
been held in combination with substantive theories of conceptual content (thus for
example Davidson) and sometimes not. The problem for this position is that in general,
warrant cannot be elucidated independently of reference and truth. For it is not plausible
that we can unify the specific conditions that give warrant for judging a given content
without mentioning the references of the concepts in the content. In general it is an
empirical matter what warrants judgement of a given content. The empirical condition
that must be met for something to be a warrant for the content is that it gives evidence
that objects referred to in the content have the (p. 447) properties and relations specified
in the content. I discuss this crux further in the next paragraph but one.
Two types of challenge exist for those who hold that reference and truth have no
substantial role to play in the individuation of concepts. One challenge for those who
think that concepts can be individuated purely in terms of rules is to explain which sets of
rules determine concepts. It is widely agreed that not all sets of rules determine
concepts. Some theorists have argued that only rules that satisfy a conservative‐extension
requirement succeed in determining concepts. If we formulate it at the level of language,
the conservative‐extension requirement states that, after the introduction of a new
expression and rules governing it to an old language and old set of rules, one cannot
prove anything formulated purely in the old language that one could not prove without
the new expression and new rules. The necessity and sufficiency of conservative
extension as such a requirement has been disputed by those who argue for a more
substantial role for reference and truth in the individuation of concepts. The issues are in
part technical, and I will not discuss them in detail here.
The other type of challenge is not technical. When we consider contents about the past,
or about other places, or predications about objects currently perceived that concern the
way they were at other times, we face a challenge if we do not invoke reference and truth
in the explanation of concept possession. It seems that we understand such predications
without thereby knowing what would be evidence for them, or what the consequences of
their truth might be. Told that it rained here yesterday, or that it is raining in the north of
Scotland now, or that this computer was made in Taiwan, or that some other creature is
in pain, you have to work out what might be evidence for these various conditions. You
also have to work out what would be consequences of their holding. Simply grasping
these contents does not tell you what such evidence or consequences would be. It follows
that no particular evidential conditions or conditions about consequences should be
written into an account of what it is to grasp these conditions. This makes it a problem
Page 11 of 22
Concepts and Possession Conditions
what a conceptual‐role account of these contents that does not make reference to
reference or truth could possibly be. By contrast, if we are allowed to use reference and
truth in our account of content, we can give a highly intuitive account of grasp of content
in all of these cases. For it to be raining here yesterday is for yesterday to have the same
property as today has to have when it is raining here today; that is, the property referred
to in ‘It's raining here today’ must be instantiated yesterday for the past‐tense thought to
be true. For ‘This computer was made in Taiwan’ to be true, the same object referred to
by ‘this computer’ must also have been made in Taiwan. And so forth. If we are allowed to
use reference, truth, and identity in the explication of these contents, we have a simple
explanation of what has to be grasped in understanding them. It is a prediction of this
account that it is an empirical matter what would be evidence for the holding of these
conditions. For it is an empirical matter what would be evidence that the same property
was instantiated yesterday at this place; an empirical matter what would be evidence that
this object was made in Taiwan; and so forth. This point incidentally also shows that we
cannot even elucidate the conditions of warranted assertibility or judgement, at least for
these contents, without involving reference and truth.
(p. 448)
What is the role of reference and truth in the individuation of concepts? Are the points
we have just made for concepts of other times, other places, and other subjects of
experience universal, applicable to all concepts at some point, or do they apply only to
a special subclass? In either case, what is their epistemic significance?
Page 12 of 22
Concepts and Possession Conditions
structures of language ‘mirror’ the tree‐like structure of Thoughts built up from concepts
(Frege's senses).
Given this difference between perceptual content and conceptual content, it is natural to
try to solve the characterization problem by saying that what is distinctive of complete
conceptual contents is that their mental representations have sentence‐like structure.
This suggestion seems however to be necessary but not sufficient for the content in
question to be conceptual. In microbiology, it is not difficult to conceive of molecular
structures that either command the building of, or contain information about, more
complex molecules, where the commanding or informational molecules themselves have
sentence‐like structure. But these would not be mental representations with conceptual
content (although they certainly have some kind of content). So what more do we need to
mention to attain a sufficient condition for the content's being conceptual? It cannot be a
matter of the content's having objective correctness conditions. That is something that is
present also in non‐conceptual content. The distinguishing feature cannot be the
possibility in conceptual thought of one and the same object, or one and the same
property, being given in different ways. There seem to be analogues of that distinction in
non‐conceptual perceptual content. We can conceive of an animal that is not capable of
conceptual thought seeing a shape as a square rather than as a regular diamond. We have
also already noted that we (p. 449) equally use that‐clauses in ascribing perceptual
content (a fact which, incidentally, means that the view that a concept can be extracted
from any use of a that‐clause is incorrect).
I suggest that conceptual contents are rather distinguished by the fact that their
constituents are individuated in part by reference to the reasons that the subject may
have for being in states involving them. The reasons may be described in terms of their
relations to tacit knowledge of reference‐conditions. A thinker does not have reasons for
being in perceptual states. A perception is not a mental action of the subject, unlike a
judgement. A subject can intend to bring about a certain state of affairs specified by non‐
conceptual content, and of course may have reasons for that intention. But the content of
his intention exists only because it also has a life, by which it is individuated, in the non‐
conceptual content of perceptual states. For perceptual states, the question of the
subject's reasons for being in them does not arise.
Page 13 of 22
Concepts and Possession Conditions
Viewed from sufficient distance, one should not be at all surprised that concepts are
externally individuated. Consider what is explained by intentional states with conceptual
content, the states of intention, emotion, values, judgement, and desire. What is explained
by such intentional states is a set of relational properties of a person. As the person goes
through his day, these states will explain his relations to people—whom he seeks out,
what he says to them; to locations, to buildings, to institutions; to objects, events, and
states of affairs. Such relational explananda require relational explaining conditions.
States that are not externally individuated would not by themselves be suitable to
explaining relational facts. Since a subject could, empirically, be in such non‐externally
individuated states but not bear any particular relations to external locations, people, and
things, they could at best (p. 450) explain his actions described in environmentally neutral
terms. They would not be up to explaining actions in relation to locations, people, and
things our subject encounters. Correspondingly, they would not by themselves support
counterfactuals or predictions about the subject's relations to these things,
counterfactuals and predictions that we rely upon every hour as a result of our
attributing intentional states with particular conceptual contents to our friends (and to
our enemies).
There are several ways in which external individuation of a concept can show up in a
concept's possession condition. One of the most basic ways is for the concept to be
individuated by its relations to perceptual states with non‐conceptual content, where that
non‐conceptual content is itself externally individuated. That is arguably the case for
basic spatial, temporal, and other observational concepts such as bent, spotted, yellow,
preceded, and the like. The properties picked out by these concepts enter the
representational content of spatial and temporal experience. What makes these
properties the ones that enter the content of experience is, arguably, that when the
perceiver is properly connected to the world around him, it is the instantiation of these
properties that causes experiences with these contents, by and large. If an observational
concept is individuated in part by the fact that someone who possesses it must be willing
to apply it to an object presented in an experience with the corresponding non‐conceptual
content, the external individuation of the non‐conceptual content will be transmitted to
external individuation of the observational concept.
Page 14 of 22
Concepts and Possession Conditions
This is not the only way external individuation may arise. Another famous way is for the
possession condition for a concept to involve a commitment to its instances being of the
same underlying natural kind as instances commonly encountered, even if that underlying
kind is not yet known (Kripke, Putnam). Yet another is for a thinker's use of a concept, as
ordinarily ascribed, to be deferential to whatever is meant by the term for the concept in
his linguistic community (Burge). There may also be cases that we can describe as
externalism within the mind itself. What makes the concept pain the concept it is is partly
that the thinker is willing to apply the concept to himself in rational response to his own
pains. Whether a thinker is in pain is something that supervenes on what is going on
inside the boundaries of his body, including his brain. But if we imagine an internal
boundary, between states with intentional contents, and the subjective conscious
sensations like pains, it seems clear that what gives those intentional states contents
involving the concept pain is their relations to sensations at the other side of the
boundary. That is, we have what we can call a kind of internal externalism within the
mind itself.
There are many ramifications of the external individuation of concepts. One ramification
concerns the nature of a subpersonal psychology that explains how thinkers come to be in
conceptual states. If those states are externally individuated, so must at least some of the
subpersonal states that explain how thinkers come to be in these conceptual states be
externally individuated. This applies in particular to a subpersonal computational
psychology.
What is the correct account of the relation between external individuation of mental
states, epistemic norms for judging given conceptual content, and the identity of the
concepts in that content?
is just an equivalent of
Page 15 of 22
Concepts and Possession Conditions
Lassie is a dog.
The introduction of talk explicitly about Thoughts gives, according to Schiffer, a conservative
extension of a theory that does not talk about them (with various technical qualifications).
‘Merely adding propositions does nothing to disturb the pre‐existing causal order; as properties
are shadows of predicates, so propositions are shadows of sentences’ (2003: 71). If the
pleonastic account is correct, much of the previous discussion of this chapter is misconceived. It
is simply a mistake to think either that there can or that there need be a substantive theory of
the nature of concepts. They no more have a substantive nature than does the average man.
According to this pleonastic conception, when someone utters a sentence stating that a
subject stands in the belief relation to a Thought or proposition, then, to quote Schiffer:
(2003: 74–5)
Schiffer also argues that we know that propositions p and q are distinct because we know that it
is possible for someone, for instance, to doubt p without doubting q. Our knowledge proceeds
thus, rather than moving from the distinctness of the propositions to the possibility of doubting
one without doubting the other (2003: 77).
The pleonastic conception faces the following objections.
(p. 452)
The first objection flows from considerations of finitude. There are infinitely many
sentences of the form ‘a believes that p’ that we understand, for each of the infinitely
many possible substitutions for ‘p’. How do we finite beings have this infinitary capacity?
If the contents expressed by the substituends for ‘p’ are, after they have been
contextually determined, built up from finitely many conceptual constituents by finitely
many modes of combination, we have a finitary basis for the infinitary capacity. But there
is no such substantive composition by conceptual constituents on the pleonastic view. If
what Schiffer calls the ‘contextually determined criteria for truth evaluation’ of a belief
report do not rely on such substantively compositional Thoughts, they have not been
finitely specified.
One can compare the situation with the position of someone who says that there are
contextually determined criteria for determining what it is that someone is saying when
he utters the embedded sentence p. However much contextual determination goes on—
and a lot does, I agree—if what is determined is not a structured Thought or proposition,
Page 16 of 22
Concepts and Possession Conditions
we have the classical problem of explaining how finite resources allow us to understand a
potential infinity of sentences. The problem recurs for intentional contexts on the
pleonastic view. To be fair to Schiffer, I should note that his view is that there is no
problem for unembedded sentences either, as he argued, controversially, two decades
ago.
The second objection is that intuitively the criteria for evaluating the truth of, to take
Schiffer's example, belief reports do plausibly rely on the attributee's meeting the
possession condition for certain concepts, and his employing those concepts in his
attitudes. If we attribute a first‐person belief to a thinker, the thinker must meet the
possession condition for the first‐person way of thinking, and that way of thinking must
be used by the thinker if, for instance, it is true that he believes ‘I am hungry’. It is the
possession condition for the first‐person concept, and its role in his belief, that makes the
difference between the criteria for evaluating the truth of the utterance ‘Peter believes he
himself is hungry’ and the criteria for evaluating the truth of the utterance ‘Peter believes
Smith is hungry’.
Third, it should be agreed that one way, even in a sense a fundamental way, of coming to
know that p and q are distinct Thoughts or propositions is to appreciate that someone can
believe or doubt one without believing or doubting the other. But this fact can be
explained on theories of conceptual content other than the pleonastic conception. The
fact is a consequence of possession‐condition theories of concepts, when these are
constrained by considerations of epistemic possibility, and take themselves as answerable
to the goal of explaining examples of Frege's informativeness criterion. Possession‐
condition theories individuate a concept F by some condition of the form ‘F is that
concept C to possess which a thinker must meet condition A(C)’, possibly with one or
another favoured restriction on the condition A( … ). The concepts C quantified over here
are already in the ontology being employed, and are conceived of as constituents of other
complex contents of various kinds that may also be mentioned in the condition A( … ).
Such complex contents would be mentioned if the possession condition for a logical
concept speaks of the concept's role in certain logical transitions between complete
contents of (p. 453) certain forms containing the concept, or speaks of tacit knowledge of
contribution to a truth‐condition. So what individuates the concepts, on this approach, is
not a condition which meets the requirement of the pleonastic conception. There is no
involvement here of a something‐for‐nothing transformation. But still it follows from this
kind of possession‐condition theory, and the constraints to which it is answerable, that a
fundamental means of coming to know that thoughts or propositions are distinct, and a
constraint on what individuates them, is that it is possible to judge or doubt one without
judging or doubting the other. It is not only the pleonastic approach that offers an
explanation of this fact.
Space prevents a full consideration of the interesting issues about conservative extension.
But I do briefly observe that if the considerations of the earlier section about external
individuation of conceptual content and the role of conceptual states in explanation is
correct, then it is not at all obvious that talk about mental states with conceptual content
Page 17 of 22
Concepts and Possession Conditions
conservatively extends talk which does not mention concepts. States with conceptual
content allow the explanation of events under relational descriptions, descriptions under
which they would not otherwise be explained; and similarly for environment‐involving
counterfactuals about thinkers and their conceptual states.
References
Introductory
Page 18 of 22
Concepts and Possession Conditions
Frege, G. (1892/1993), ‘On Sense and Reference’, repr. in A. Moore (ed.), Meaning and
Reference (Oxford: Oxford University Press), 23–42.
For emphasis on the idea that a concept or sense is individuated by the fundamental
condition for something to be its reference see
For theories that in one way or another quantify over concepts or modes of presentation
in their account of the attribution of attitudes in language see
Crimmins, M., and Perry, J. (1989), ‘The Prince and the Phone Booth: Reporting Puzzling
Beliefs’, Journal of Philosophy, 86: 685–711.
Fodor, J. (1998), Concepts: Where Cognitive Science Went Wrong (Oxford: Oxford
University Press).
Johnston, M. (1988), ‘The End of the Theory of Meaning’, Mind and Language, 3: 28–42.
Page 19 of 22
Concepts and Possession Conditions
(p. 455) Williamson, T. (2000), Knowledge and Its Limits (Oxford: Oxford University
Press).
Genzten, G. (1970), The Collected Papers of Gerhard Gentzen, ed. M. Szabo (Amsterdam:
North‐Holland).
Frege gives the example of Afla and Ateb in an apparently undated letter to Jourdain,
which is reprinted in translation in
A. W. Moore (ed.), Meaning and Reference (Oxford: Oxford University Press, 1993).
For the view that truth and reference play an ineliminable role in concept possession see
For the view that truth and reference play no fundamental role in concept possession and
understanding, in addition to the works of Johnston and Schiffer listed above, see
Field, H. (1977), ‘Logic, Meaning and Conceptual Role’, Journal of Philosophy, 74: 379–
409.
Page 20 of 22
Concepts and Possession Conditions
McDowell, J. (1994), Mind and World (Cambridge, Mass.: Harvard University Press).
Burge, T. (1979), ‘Individualism and the Mental’, Midwest Studies in Philosophy, 4: 73–
121.
The principal sources are the books by Schiffer and the paper by Johnston cited above.
For Schiffer's views on the explanation of the understanding of unembedded sentences
see
Page 21 of 22
Concepts and Possession Conditions
See the reply by Jerry Fodor in the same volume; and also the exchange between Schiffer
and Peacocke in Mind and Language, 1 (1986); and for similar views see
Burge, T. (2005), Truth, Thought, Reason: Essays on Frege (Oxford: Oxford University
Press), in particular the essay ‘Frege and the Hierarchy’ and the important ‘Postscript’.
Notes:
(*) An earlier version of this chapter formed part of an overview presented at the joint
Columbia–Rutgers Seminar on Concepts I gave jointly with Jerry Fodor in the fall of 2005;
I thank him for comments and advice. For further development of some of the views in
this chapter see my more recent book Truly Understood (Oxford: Oxford University Press,
2008).
Christopher Peacocke
Page 22 of 22
The Distinction Between Conceptual and Nonconceptual Content
Page 1 of 19
The Distinction Between Conceptual and Nonconceptual Content
Page 2 of 19
The Distinction Between Conceptual and Nonconceptual Content
explanations should be understood. Some theorists (e.g. Fodor 1975) have argued that
the representational states attributable to non‐linguistic creatures do not differ in any
significant respect from those attributable to language‐using creatures, while others
maintain that the explanatory practices should not be taken at face value (e.g. McDowell
1994). There is room, (p. 459) however, for a middle position, on which non‐linguistic
creatures are genuinely capable of representing the world and acting in virtue of how
they represent the world, even though there are significant differences between the types
of representation available to non‐linguistic creatures and those available to language‐
using creatures. This middle position can in turn be developed in a number of ways. One
of these identifies the significant differences at the level of content, claiming that only
states with nonconceptual content are available to non‐linguistic creatures, with
conceptual content reserved for language‐using creatures (although language‐using
creatures may also be in states with nonconceptual content). This approach is adopted in
Bermúdez (1998).
The first question that one might ask, in light of the different appeals that have been
made to the conceptual/nonconceptual distinction, is where exactly the distinction is to be
located. Is it a distinction between two different types of content? Or is it a distinction
between two different types of content‐bearing state? We can, following Heck (2000),
term the first interpretation the content view and the second the state view.1
Suppose that we are thinking about what distinguishes perceptual experiences from the
propositional attitudes (the first domain of application for the conceptual/nonconceptual
distinction). According to the content view, there is a fundamental difference between the
type of content that perceptual experiences can have and the type of content that beliefs
and other propositional attitudes can have. According to the state view, in contrast, there
is only one type of content. Perceptual experiences and propositional attitudes have
contents of the same type. What distinguishes them is that states of the latter type are
concept‐dependent while states of the former type are concept‐independent. A state type
is concept‐dependent just if it is content‐bearing and it is impossible for a thinking and
perceiving subject to be in a token state of that type without possessing the concepts
required to specify the content of that token state. In contrast, a subject can be in states
of a concept‐independent type even if she lacks all or some of the concepts required for
an accurate specification of the relevant contents.
Page 3 of 19
The Distinction Between Conceptual and Nonconceptual Content
Many of the arguments offered to support the thesis that perception has a nonconceptual
character are arguments in support of the content view. So, for example, (p. 460) it has
been argued on a number of occasions that perceptual experiences have a richness of
grain that cannot be captured at the conceptual level—and there has been an extensive
debate about whether, for example, demonstrative concepts such as that shade can
capture the fine‐grained representational content of my experience of a colour swatch.2
The debate here (pace Byrne 2003) is surely about whether or not perceptions can have
the same kind of content as beliefs. The issue is not about whether or not perceivers must
have the conceptual abilities corresponding to the demonstrative concept that shade. All
the questions about fineness of grain persist even if we stipulate that the perceiver in
question possesses (and deploys) any potentially relevant demonstrative concept.
The state view proposes a principled distinction between concept‐dependent state types
and concept‐independent state types. Plainly, proponents of the distinction owe us an
account of where it comes from. Why is it the case that beliefs do, while perceptions do
not, respect the conceptual constraint? The most obvious answer is that the concept‐
independence of perceptual states is a function of their distinctive type of content. This
answer is of course not available to the state‐view theorist, who only appears to have two
possible lines of response. On the one hand, she can turn to the different functional roles
of the respective state types. Alternatively, she can look to their different
phenomenologies for an explanation of the difference between concept‐dependent and
concept‐independent state types.
The appeal to functional role appears to be restating the problem rather than providing
an explanation. Since it is part of the functional role of belief that it is concept‐dependent
(so that, for example, one cannot form beliefs about matters that are beyond one's
conceptual grasp), and part of the functional role of perception that it is concept‐
independent (so that one can react to objects that one perceives even though one cannot
conceptualize them), there seems to be no prospect of explaining concept‐(in)dependence
in terms of functional role.
(p. 461)
Page 4 of 19
The Distinction Between Conceptual and Nonconceptual Content
I conclude, then, that the state view is unlikely to help us make sense of the conceptual/
nonconceptual distinction. Nonconceptual theorists are committed both to the thesis that
perception is concept‐independent and to the claim that this concept‐independence is a
function of the distinctive type of content enjoyed by perceptual experiences. In the
remainder of the chapter I will be discussing different versions of the content view.
To say that a mental content is nonconceptual is to say that its subject need not
possess any of the concepts that we, as theorists, exercise when we state the
correctness condition for that content.
(2000: 62)3
There are two points to note. First, the characterization is purely negative (nonconceptual
content is simply content that fails to satisfy a constraint upon conceptual content). Second, the
characterization is indirect, proceeding via conditions upon the ascription of content, rather
than by placing constraints upon the nature of nonconceptual content itself. As the examination
of the state view in Section 27.2.1 makes clear, the standard characterization cannot be the
whole story. We need an explanation of why nonconceptual content is not concept‐dependent.
The first step is to clarify the determinable of which both conceptual and nonconceptual
content are determinates. How is the generic notion of content being understood? In
particular, what is it to attribute content to a state? If we start from the basic principle,
shared by both proponents and opponents of the general idea of nonconceptual content,
that the content of belief is a paradigm of conceptual content, then it is easy to see that
we are dealing with a notion of content that is richer than purely truth‐conditional
content, since there are indefinitely many ways of giving the truth‐condition of a belief
that fail to be concept‐dependent. Suppose I believe truly that my car is currently in the
drive. Since there is presumably a complete microphysical description of that region of
space–time, this description can be used to give a specification of the truth‐condition of
my belief. As such it will count as nonconceptual. Plainly this is not the canonical way of
specifying the content of my belief.
Page 5 of 19
The Distinction Between Conceptual and Nonconceptual Content
With the perspectival constraint in play the point of the standard characterization is
clearer. Non‐conceptual content, if there is such a thing, is content specifiable in a way
that respects the perspectival constraint without being constrained by the conceptual
capacities of the thinking and perceiving subject. It is also clearer what the opponent of
nonconceptual content is claiming; namely, that the concepts a thinker or perceiver
possesses fully determine how she apprehends the state of affairs she is representing.
The best way to see how the perspectival constraint might be satisfied in a concept‐
independent manner is through an example. The notion of scenario content developed by
Christopher Peacocke is the most developed account available.
Peacocke's account of the content of perceptual experience is based on the idea that the
way a perceiver represents the environment should be specified in terms of all the ways
of filling out the space around the perceiver consistent with that perceptual
representation. We organize the space around the perceiver by specifying origins and
corresponding axes based on the perceiver's body. There is a range of such origins,
varying according to the modality in question. In the case of touch, for example, the
possibilities include origins based on the centre of the palm of each hand, as well as
certain other origins that capture the sensitivity to touch of the remainder of the body
surface. Once we have selected a particular origin and defined a set of axes in terms of it
we can identify points (or, rather, point types) in terms of their distance and direction
from the origin. We can now complete the account of how space is ‘filled out’ (in the
visual case) by first specifying for each point type whether there is a surface there and, if
there is, then characterizing its hue, saturation, brightness, degree of solidity, orientation,
and so on.
Completing this characterization for a given origin and set of axes determines what
Peacocke calls a scenario—namely, a class of ways of filling out space, any of which could
be instantiated in the real world. The content of perception is given by positioning the
scenario in the physical environment—that is, by assigning real‐world values (p. 463) to
Page 6 of 19
The Distinction Between Conceptual and Nonconceptual Content
the origin and axes and determining a time. The positioned content is correct just if the
volume of perceptually discriminable space around the perceiver at a specified time is a
member of the class specified by the scenario.
Prima facie the notion of scenario content satisfies the standard characterization. The
basic idea behind scenario content is that capacities for perceptual discrimination are not
dependent upon conceptual capacities, and the machinery of scenario content is designed
purely and simply to specify what it is that a perceiver is capable of perceptually
discriminating, in a manner that is completely independent of the perceiver's conceptual
capacities (if any).4 A theorist might employ complicated differential equations to capture,
for example, how perceived brightness changes at particular points in the scenario. There
is no requirement that the perceiver be capable of understanding those equations for
them accurately to reflect how she perceives the environment (any more than a
parachutist needs to understand the basic principles of gravitational acceleration in order
to be capable of a competent free‐fall). And nor, one might think, need a perceiver have
any idea of what saturation is in order to be able perceptually to discriminate and
represent different degrees of saturation.
Page 7 of 19
The Distinction Between Conceptual and Nonconceptual Content
John McDowell (1994) has been the most vocal and influential critic of the claim that
perception has nonconceptual content. He offers two principal objections. The (p. 464)
first is that the notion of non‐conceptual content is epistemologically problematic. If
perceptual experiences lack conceptual content, McDowell argues, then they can have no
role to play in justifying, as opposed simply to causing, perceptual beliefs. Attributing
nonconceptual content to perceptual experiences is, according to McDowell, a
sophisticated way of falling into the myth of the given. The second objection is that it is
unnecessary to appeal to nonconceptual content, because there is no reason to think that
capacities for perceptual discrimination do outstrip conceptual capacities.
It is true that we do not have ready, in advance of the course our color experience
actually takes, as many color concepts as there are shades of color that we can
sensibly discriminate. But if we have the concept of a shade, our conceptual
powers are fully adequate to capture our color experience in all its determinate
detail.
It is not easy to see why this line of argument should trouble the nonconceptual theorist, who is
not committed to denying any of the claims made in these two passages. The question at issue is
whether capacities for perceptual discrimination are constrained by the concepts possessed by
the perceiver. It would be an objection to the nonconceptual theorist if, for example, it turned
out to be the case that perceivers lacking the concept of saturation were unable to discriminate
differences in saturation. But the simple fact (if fact it be) that a perceiver possessing the
concept of saturation can give verbal expression to the discriminations that they are capable of
making says nothing about the sources and origins of that discriminatory capacity. As far as I
can see, the extensive literature stemming from McDowell's remarks has no bearing on the
merits or otherwise of the case for nonconceptual content. The nonconceptual theorist is not
committed to the claim that perceptual experience is ineffable.
McDowell's second point is more subtle and less discussed. It is, he thinks, a constraint
upon any theory of perceptual experience that it explain how perceptual experiences can
fall within what he terms ‘the space of reasons’. That is, an adequate theory of perceptual
content must explain how the content of perception can stand in justificatory and
inferential relations to perceptual beliefs and other propositional attitudes. Yet, McDowell
argues, the type of experience without concepts envisaged by nonconceptual theorists
Page 8 of 19
The Distinction Between Conceptual and Nonconceptual Content
How should we cash out the image of blindness? To say that an experience is not
blind is to say that it is intelligible to its subject as purporting to be awareness of a
feature of objective reality: as a seeming glimpse of the world. And Evans himself
insists that this can be so only against the background of an understanding of how
perception and reality are related, (p. 465) something sufficient to sustain the idea
that the world reveals itself to a perceiving subject in different regions and
aspects, in a way that depends on the subject's movement through the world. Such
a background can be in place only for a subject with a self‐conscious conception of
how her experiences relate to the world, and we cannot make sense of that in the
absence of conceptual capacities in a strong sense, a faculty of spontaneity.
(1994: 54)
Page 9 of 19
The Distinction Between Conceptual and Nonconceptual Content
and the perceptual discrimination of objects and properties must be distinct from the
process of applying concepts if some applications and judgements are to be warranted
and others not. Roughly speaking, a thinker is warranted in applying the concept of, say,
a particular shade (p. 466) of colour to a property just if her perceptual experience makes
that shade available in the appropriate manner. By extension, perceptual judgements are
warranted just when they are appropriate responses to the world as it is presented in
perception. The notion of warrant cannot get a grip here if we have not left open the
possibility of a gap between how the world is presented in perception and the concepts
that are deployed to think about it. Nonconceptual theorists maintain that there can be
no such gap unless we allow that perceptual experiences have nonconceptual content.
We can be briefer in discussing the case for nonconceptual content in the second domain
of application. Even implacable opponents of the idea of nonconceptual content at the
personal level are generally happy with ascriptions of nonconceptual content at the
subpersonal level. McDowell, for example, remarks that ‘it is hard to see how cognitive
psychology could get along without attributing content to internal states and occurrences
in a way that is not constrained by the conceptual capacities, if any, of the creatures in
question’ (1994: 55).
Neither of the two lines of objection to claims about the nonconceptual content of
perceptual experience can get any grip in this domain. The epistemological concerns
raised by McDowell are completely beside the point in this area, since it is standardly
taken to be characteristic of subpersonal information states that they are inferentially
insulated from personal‐level states (whether perceptual experiences or propositional
attitudes). There is a number of relations that might hold between the information‐
processing states of, say, the early visual system and personal‐level visual experiences—
such as the relations of causation and realization—but no one has ever maintained that
the reason‐giving relation might be one of them. Nor is there any temptation to try to
Page 10 of 19
The Distinction Between Conceptual and Nonconceptual Content
There are two substantive issues that arise in this domain, however. First, there is the
question of how content attributions at the personal and subpersonal levels (p. 467) are
related. Unsurprisingly, McDowell is keen to make a sharp distinction between
attributions of content at the personal and subpersonal levels. The latter involve, he
thinks, only an ‘as‐if’ form of content that needs to be sharply distinguished from
‘genuine’ content, which McDowell takes to be content that falls within the ‘space of
reasons’.6 An opposing view is taken by Bermúdez (1995), who sketches out a number of
generic ‘marks’ of content and argues that they are satisfied by nonconceptual content no
less than by conceptual content.
The second issue is related. Suppose that we think that there are states with
nonconceptual content at both the personal and the subpersonal level. Should they both
be specified in the same way? Certainly, in the case of perceptual processing at the
subpersonal level it would be easier to understand how these types of processing are
related to personal‐level perceptual experience if there were at least structural parallels
between the relevant content specifications. The machinery of scenario content is suited,
as Peacocke has noted (1992: 65–6), for giving the content of the representations posited
by theorists of early visual processing. Marr suggests, for example, that visual processing
involves the construction of what he terms a 2.5D sketch, representing the orientation
and depth of visible surfaces on retinocentric coordinates (1982). The content of the 2.5D
sketch lends itself to specification in terms of values on certain specified dimensions at
point types in a positioned scenario. Such a specification would be far more impoverished
than a specification of the corresponding personal‐level perceptual experience but
nonetheless continuous with it in a way that opens up the possibility of exploring
connections between personal‐level experience and the underlying subpersonal
information processing.
On the other hand, however, the notion of nonconceptual content seems applicable to a
much broader range of subpersonal computational states than those involved in
perceptual processing. The processing involved in linguistic understanding is a case in
point. Let us suppose, as is standardly assumed by linguists and many philosophers, that
linguistic understanding involves tacit knowledge of some form of syntactic and/or
semantic theory. It is plain that the tacitly known theories we might posit to explain
linguistic understanding are completely unconstrained by the conceptual repertoire of the
language user. And yet it is plausible to think that the axioms and rules of such theories
correspond to the contents of subpersonal computational states. (Without such an
assumption it is hard to see how we might develop a theory of language processing.) In
order to characterize the content of such states we need to move beyond scenario
content, making connections between the theory of nonconceptual content and accounts
of tacit knowledge proposed by philosophers such as Evans (1981).7
Page 11 of 19
The Distinction Between Conceptual and Nonconceptual Content
In a sense the case discussed in Section 27.3.1 for thinking about the perceptual states of
language‐using creatures in terms of nonconceptual content carries over
straightforwardly to non‐linguistic creatures. However, thinking about non‐linguistic and
pre‐linguistic creatures raises a new set of issues.
As mentioned in Section 27.1.3, the pressure for attributing states with nonconceptual
content to non‐linguistic creatures comes primarily from the explanatory practices of
disciplines such as cognitive ethology and developmental psychology. This gives a very
different emphasis from the domains of application that we have already discussed. When
we are interested primarily in representations that explain behaviour, issues to do with
the fine‐grained character of perception recede into the background (except, of course,
where the behaviour to be explained exploits capacities for fine‐grained perceptual
discriminations). What becomes important instead is representations of objects and
properties in the distal environment. This raises interlocking questions. First, can there
be nonconceptual perception of objects and properties in the distal environment? Second,
are there any reasons to attribute to non‐linguistic creatures non‐perceptual
representational states with nonconceptual content? In other words, do we need to go
beyond accounts of nonconceptual content, such as scenario content, that are most
naturally applicable to perceptual states (whether at the personal or at the subpersonal
levels)?
The first question immediately raises an issue that has been in the background up to now.
No theorist has ever seriously denied that our perceptual experience outstrips our
conceptual capacities with respect to such phenomena as colours, pitch, shape, and so on
in the minimal sense that we do not have concepts corresponding to all the colours,
pitches, and shapes we can perceptually discriminate (although of course philosophers
such as McDowell have argued that this does not prevent perceptual experience from
being brought within the realm of the conceptual). Yet it might much more plausibly be
argued that perceivers have concepts for the different types of object and property that
they are capable of perceptually discriminating. There is room for a minimal account of
concept possession at the non‐linguistic level, on which possession of the concept F
amounts to no more than the capacity appropriately to distinguish Fs from non‐Fs and to
act appropriately in the presence/absence of Fs. On the minimal account of concept
possession, the principal types of evidence for attributing psychological states to non‐
linguistic creatures would ipso facto be evidence for attributing to them conceptual
abilities.
Page 12 of 19
The Distinction Between Conceptual and Nonconceptual Content
concept (1994, 2000). In fact, for theorists (p. 469) persuaded by arguments to the effect
that any sort of higher‐order thought is unavailable at the non‐linguistic level (e.g.
Bermúdez 2003: chs. 8–9), placing any sort of reflective requirement upon concept
possession would effectively make all non‐linguistic content nonconceptual. Nonetheless,
there are intermediate positions available. One might, for example, make it a requirement
upon possession of the concept F that a thinker be capable of making certain canonical
judgements and inferences in the presence of Fs, where those judgements and inferences
are of a type available to some but not all non‐linguistic creatures whose behaviour might
be explained in psychological terms.8
Suppose, then, that we are operating with an account of concept possession that allows
us to attribute nonconceptual contents at the non‐linguistic level. Are these contents
necessarily going to be perceptual contents? According to what Bermúdez (2003: ch. 3)
terms the minimalist account of non‐linguistic thought, in giving psychological
explanations of the behaviour of non‐linguistic creatures we do not need to go outside the
sphere of perception. So, for example, Dummett has argued that the types of thinking of
which non‐linguistic creatures are capable are all fundamentally perceptual in form,
involving the manipulation of spatial images essentially tied to the perceptual context and
to the possibilities that the environment affords for action. According to Dummett, the
vehicles of what he terms proto‐thoughts are ‘spatial images superimposed on spatial
perceptions’ (1993: 123). Similar views have been proposed by John Campbell (1994). If
something like the minimalist view is correct then we do not need to go outside the realm
of nonconceptual perceptual content in order to accommodate psychological explanation
at the non‐linguistic level. We could, for example, make use of Peacocke's notion of
scenario content (perhaps with the modifications suggested in Section 27.4 below).
On the other hand, it might turn out to be the case (as I argue in Bermúdez 2003: ch. 3)
that the requirements of psychological explanation at the non‐linguistic level take us
beyond the resources of the minimalist conception—particularly when we are dealing
with behaviours that involve instrumental reasoning about the likely consequences of
actions and/or the tailoring of means to ends. Such an extension might involve explaining
how there might be analogues of propositional attitudes (proto‐beliefs, say, or proto‐
desires) with nonconceptual content. Alternatively, it could be the case that the failure of
the minimalist conception of non‐linguistic thought maps exactly on to the distinction
between conceptual and nonconceptual content, so that the types of behaviour that resist
explanation in terms of something like Dummett's ‘spatial images superimposed on
spatial perceptions’ are precisely those that require explanation in terms of conceptual
contents.
Nonconceptual Content
Page 13 of 19
The Distinction Between Conceptual and Nonconceptual Content
In this section I consider how the account of scenario content sketched out in Section
27.2.3 might need to be developed and modified in the light of the discussion in the
previous section.
Peacocke is aware of the limitations of scenario content and posits a further layer of
nonconceptual content, which he terms proto‐propositional content. He writes:
The contents at this second layer cannot be identified with positioned scenarios,
but they are also distinct from conceptual contents. These additional contents I
call protopropositions. These protopropositions are assessable as true or false. A
protoproposition contains an individual or individuals, together with a property or
relation. When a protoproposition is part of the representational content of an
experience, the experience represents the property or relation in the
protoproposition as holding of the individual or individuals it also contains.
(1992: 77)
The example that Peacocke gives of proto‐propositional content is the difference between
perceiving a figure as a square and perceiving it as a diamond. This difference lies, he plausibly
claims, in which axes of symmetry a perceiver finds most salient. When the salient axes of
symmetry are those bisecting the sides then the figure is perceived as a square. When the
salient axes of symmetry bisect the angles then the figure is perceived as a diamond.
Peacocke does not go into much detail as to what it is for an individual to be represented
at the proto‐propositional level. However, the nonconceptual theorist clearly needs to
explain what it is to perceive an object without conceptualizing it as an object, or as an
object of a particular type. The nonconceptual theorist also needs to be able to do justice
to differences in how objects are perceived by different types of perceiver (e.g. by pre‐
linguistic infants as opposed to by adult language users).
What is it to perceive a portion of, say, the visual field as an individual? At a minimum it
seems to involve perceiving that portion as a bounded segment and having certain
expectations about how that bounded segment will behave. These expectations can, in the
case of adult perceivers, be expected to track some of the (p. 471) basic properties of
Page 14 of 19
The Distinction Between Conceptual and Nonconceptual Content
Many objects lack one or more object‐properties, and certain object‐properties may be reducible
to others (any object that presents resistance is likely to be impenetrable). Some object‐
properties might not be expected to apply to all objects at all times. Objects that are lighter than
air, for example, do not fall when unsupported, nor do flying birds. Nonetheless, the properties
on this list correspond to higher‐order physical regularities that by and large govern the
behaviour of physical objects.
The perceiver's implicit understanding of these regularities can be thought of as a naive
physics. At a minimum we can see this implicit understanding as consisting in a set of
perceptual expectations, manifestable in action and in surprise. Perceivers typically act in
ways that reflect their naive physics, and show surprise when physical regularities are
breached and their expectations thwarted. By working backwards from these
manifestations it is generally possible to identify the object‐properties to which
perceivers are perceptually sensitive and that determine how they parse the perceived
environment.9 This sensitivity is both nonconceptual and part of the content of
perception.10
This basic idea of perceptual sensitivity to higher‐order physical regularities can be used
in conjunction with Peacocke's model of scenario content to give a fuller picture of the
nonconceptual content of perception—a picture that does justice to the range of
explanatory tasks for which the notion of nonconceptual content has been invoked.
References
Bermúdez, J. L. (1994), ‘Peacocke's Argument Against the Autonomy of Nonconceptual
content’, Mind and Language, 9: 402–18.
Page 15 of 19
The Distinction Between Conceptual and Nonconceptual Content
Campbell, J. (1994), Past, Space and Self (Cambridge, Mass.: MIT Press).
Dretske, F. (1981), Knowledge and the Flow of Information (Cambridge, Mass.: MIT
Press).
Evans, G. (1981), ‘Semantic Theory and Tacit Knowledge’, in S. Holtzman and C. Leich
(eds.), Wittgenstein: To Follow a Rule (London: Routledge), 118–32; repr. in Evans,
Collected Papers (Oxford: Oxford University Press), 322–42.
Fodor, J., and Lepore, E. (1994), ‘What is the Connection Principle?’, Philosophy and
Phenomenological Research, 54: 837–45.
Gopnik, A., and Meltzoff, M. (1997), Thoughts, Theories, and Things (Cambridge, Mass.:
MIT Press).
Page 16 of 19
The Distinction Between Conceptual and Nonconceptual Content
McDowell, J. (1994), Mind and World (Cambridge, Mass.: Harvard University Press).
Miller, A. (1997), ‘Tacit Knowledge’, in B. Hale and C. Wright (eds.), Companion to the
Philosophy of Language (Oxford: Blackwell), 146–74.
Slater, A., Morison, V., Somers, M., Mattock, A., Brown, E., and Taylor, D. (1990),
‘Newborn and Older Infants’ Perception of Partly Occluded Objects’, Infant Behaviour and
Development, 13: 33–49.
Tye, M. (2000), Consciousness, Colour and Content (Cambridge, Mass.: MIT Press).
Page 17 of 19
The Distinction Between Conceptual and Nonconceptual Content
Notes:
(1) For further discussion of the distinction between the state and content views see
Byrne (2003).
(3) Similar formulations can be found in Cussins (1990), Martin (1992), and Bermúdez
(1995).
(4) With regard to the parenthesis, Peacocke originally maintained that states with
scenario content were only available to creatures possessing a certain minimal
conceptual capacity (see Peacocke 1992: ch. 3 and his 1994 reply to Bermúdez 1994). He
has since changed his view and accepted the possibility of what he terms autonomous
non‐conceptual content (see Peacocke 2002).
(5) Evans himself held that a subject's awareness of an objective world is a function of
their possessing what he termed a ‘simple theory of perception and action’, which is not
itself part of the content of perception. For further discussion of Evans's views in this
area see Campbell (2005).
(6) A similar conclusion can be found in Searle (1990). Searle maintains a connection
principle to the effect that a state can only be genuinely representational if it is (at least
potentially) accessible to consciousness, which clearly rules out subpersonal
computational states. For critical discussion see Fodor and LePore (1994).
(7) See also Davies (1989), Peacocke (1989) and, for a general overview of theories of
tacit knowledge, Miller (1997).
(8) The account of concept possession developed in Peacocke (1992) fits this description.
Peacocke requires concept possessors to make judgements and inferences for reasons,
but does not in general require that those reasons be subject to reflective scrutiny.
Page 18 of 19
The Distinction Between Conceptual and Nonconceptual Content
individuals are spatially connected bodies that retain their connectedness when they
move. But this often leads them to take to be a single object what is really a composite
(Gopnik and Meltzoff 1997).
(10) Note that this is not the same as the simple theory of perception mentioned in n. 5.
José Luis Bermúdez, Dean of the College of Liberal Arts, and Department of
Philosophy, Texas A&M University
Page 19 of 19
Intentionalism
In recent years there has been considerable debate over whether all mental states are
intentional; in particular, over whether all conscious mental states are intentional or
entirely intentional. This article uses the term intentionalism for the general thesis that
the nature of a conscious mental state is determined by its intentionality. (Intentionalism
is sometimes called representationalism; the difference is purely terminological.) There
are a number of ways of developing this general thesis; this article examines two of them.
One is the view that the conscious character of a state of mind is determined by its
intentional or representational content. The other is the view that the conscious character
of a state of mind is determined by (what the article calls) its entire intentional nature.
The article argues for the superiority of the second view over the first.
Keywords: mental states, intentionalism, representationalism, state of mind, conscious character, philosophy of
mind
Page 1 of 23
Intentionalism
28.1 Introduction
THE central and defining characteristic of thoughts is that they have objects. The object of
a thought is what the thought concerns, or what it is about. Since there cannot be
thoughts which are not about anything, or which do not concern anything, there cannot
be thoughts without objects. Mental states or events or processes which have objects in
this sense are traditionally called ‘intentional’, and ‘intentionality’ is for this reason the
general term for this defining characteristic of thought.
Under the heading of ‘thought’ we can include many different kinds of mental
apprehension of an object, among them relatively temporary episodes of contemplating or
scrutinizing, as well as persisting states like beliefs and hopes which are not similarly
episodic in character. These are all ways of thinking about an object. But even construing
‘thought’ in this broad way, it is clear that not all mental states and events are thoughts:
sensations, emotions, and perceptual experiences are not thoughts, but they are also
paradigmatically mental. Do these mental states and events have objects too? Or are
there mental states and events which have no objects?
The view that all mental phenomena have objects is sometimes called ‘Brentano's thesis’
or the thesis that intentionality is the ‘mark’ of the mental.1 Sometimes the name
‘Brentano's thesis’ is given to certain other views too; for example, to the view that only
mental phenomena are intentional, or that all and only mental phenomena are intentional,
or that nothing physical is intentional. These views are, however, distinct from the view
that all mental phenomena are intentional. For holding that (p. 475) all mental
phenomena are intentional does not imply that nothing non‐mental is.2 And holding that
all mental phenomena are intentional does not imply (pace Dennett 1969) that nothing
physical is intentional, since if physicalism were true, then the mental itself would be
physical. What I am concerned with here, however, is the idea that all mental states are
intentional, regardless of whether anything else is, or whether anything physical is.
In recent years there has been considerable debate over whether all mental states are
intentional; in particular, over whether all conscious mental states are intentional or
entirely intentional. I will use the term intentionalism for the general thesis that the
nature of a conscious mental state is determined by its intentionality. (Intentionalism is
sometimes called representationalism; the difference is purely terminological. I prefer
‘intentionalism’.3) There are a number of ways of developing this general thesis; in what
follows I shall examine two of them. One is the view that the conscious character of a
state of mind is determined by its intentional or representational content. The other is the
view that the conscious character of a state of mind is determined by (what I shall call) its
entire intentional nature. I shall argue for the superiority of the second view over the
first. But before doing this I need to explain what I mean by intentional nature and
intentional or representational content.
Page 2 of 23
Intentionalism
For some states of mind which are intuitively intentional or representational—in the
sense that they intuitively seem to concern things other than themselves—the question
‘What is this intentional state about?’ is awkward or makes little sense. It is awkward, if
not ungrammatical, to say ‘What is your desire for a bottle of inexpensive (p. 476)
champagne about?’. A desire is a desire for something, not a desire about something. But
clearly my desire concerns something other than itself, just as my thought about
Odysseus concerns something other than itself. And it is the nature of this ‘concerning’
which is the focus of the study of intentionality. Hence, we should not become too
attached to the words ‘about’ and ‘aboutness’ in describing intentionality. If we do, then
we might find it hard even to understand the view that conscious states are intentional.
For example, intentionalists say that a headache is an intentional state, but the question
‘What is your headache about?’ makes little sense. Here we can avoid this superficial
problem by thinking of intentional states in a more general, abstract, and semi‐technical
way to begin with: intentional states are ‘directed upon’ objects, and ‘objects’, by
definition, are what intentional states are directed on. Sometimes intentionality is
described even more abstractly as ‘self‐transcendence’: the idea being that intentional
states are concerned with what transcends the state itself. Given what I mean by ‘object’,
this is just another way of saying that intentional states have objects.
Page 3 of 23
Intentionalism
to such objects must be explained away in the style of Quine (1948). Here I simply note
the fact and do not attempt to explain it.
The second characteristic is related to the first. As indicated above by my talk of ‘self‐
transcendence’, intentional objects are not immanent in intentional states. That is, if an
intentional object exists at all, its existence ‘transcends’ any intentional state which has it
as its object.4 I mean this transcendence of the intentional object to imply that it is never
essential to an intentional object that it is the object of any particular state of mind. (This
applies even to those intentional states which are the objects of ‘higher‐order’ intentional
states: my thought that p is independent from, and inessential to, the thought that I think
that p.) Externalists about mental content may say that it is essential to certain states of
mind that they involve a real relation to some real thing (e.g. water). But they do not
think that it is essential to water that it is the object of any particular mental state.
(p. 477)
Saying what the intentional object of a state of mind is does not yet tell us what the state
of mind itself is, since the same intentional object can be the object of many different
states of mind. In order to fully characterize different states of mind, we need to make
two further distinctions. One is that the same object can be the object of a desire, a
thought, a hope, and so on. This is what I call a difference in intentional mode.5 The other
distinction is that the same object could be presented to the mind in different ways even
when the mode is the same: my bottle of inexpensive champagne could also be thought of
as a bottle of inexpensive famous sparkling wine from France. This kind of difference in
the way the intentional object is presented is what I call a difference in intentional
content.
Every intentional state must have an intentional content in this sense. This is because the
intentional object of a state is what the state is directed on; but a state cannot be directed
on something without that thing being represented in one way or another. There is no
such thing as a state of mind which has an object represented in no particular way. What
could this possibly mean? The idea of representation itself implies representation in a
particular way: in representing something in language or in pictures one has to choose
some particular way of representing it. The particular way in which the intentional object
is represented is what I call the content of the state. So for a state to have a content is for
it to have an object represented in a particular way. Intentional content is
representational content.
Intentional states, then, involve both intentional mode and intentional content. But what
is the relationship between the intentional content and the intentional object of a state?
Different theorists have different views of this relationship. Some think that the content
of an intentional state must determine its object; in other words, that states with the
same content must have the same object. Someone who thinks this might then be led to
think that even though its existence transcends the intentional state, the intentional
object is nonetheless an essential part of the state, since states with different objects are
different in their nature.6 Since I believe that many intentional states have the same
Page 4 of 23
Intentionalism
content regardless of whether their objects exist, I must reject the thesis that intentional
content determines intentional object. On the conception of intentionality I favour, the
intentional object is never part of the state. The nature of the intentional state is
exhausted by its mode and content.
The phrase ‘what is ϕ‐ed’ picks out a different intentional element for different intentional
modes. In the case of fear, for example, we can say that what you fear is the object of your
fear, not its content. If you are frightened of the dog around the corner, that is the object
of your fear, that is what you are frightened of. Matters are different for hope, belief, and
wonder, for example. What is believed, hoped, or wondered is the content of the belief,
hope, or wonder. The lesson is that we cannot derive a systematic conception of
intentional content and object from reflection on phrases like ‘what is believed/feared/
etc.’ alone.
This whole way of thinking about intentionality allows for a number of different
understandings of content. On a dominant view, for example, the contents of all
intentional states are assessable as true or false; in other words, the content is
propositional. There is an ongoing debate about the nature of propositions (see Salmon
and Soames 1988; Schiffer 2003). Here all I will mean by ‘proposition’ is the content of an
intentional state that is true or false. As we shall see, some philosophers think that all
intentional content is propositional. Yet it seems to me that there are many intentional
states whose contents are not assessable as true or false; for example, the object‐directed
attitudes of love and hate (see Crane 2001: sect. 34; Montague (forthcoming)). Hence I
reject the thesis that all intentional content is propositional. Although I will rely on this
rejection later in the chapter, I will not defend it further.
Page 5 of 23
Intentionalism
Taken literally, strong pure intentionalism is of dubious coherence. It says that the
phenomenal character of an experience is identical with its representational content. The
representational content of an experience, according to pure intentionalism, is a
proposition. The phenomenal character of an experience is what it is like to have that
experience. But how can what it is like to have an experience be identical to a
proposition? Propositions are abstract objects—maybe Fregean thoughts, maybe
modelled by sets of possible worlds or intensions or ordered n‐tuples of objects and
properties—and what it is like to have experience is (arguably) a feature of an experience,
a concrete event. How can a feature of a concrete event be identical to an abstract
proposition? If we are to take strong intentionalists at their word, their claim makes little
sense.7
However, we should interpret the claim with a charitable pinch of salt. Let's instead
understand strong pure intentionalists as saying that for an experience to have a certain
phenomenal character simply is for it to have a certain intentional content. Strong pure
intentionalists should not say that the phenomenal character is identical with the content;
Page 6 of 23
Intentionalism
rather they should say that an experience's having phenomenal character is identical with
its having a certain content. Weak pure intentionalism can then be expressed as follows:
any two experiences which share intentional content will share phenomenal character.
This is consistent with the identity claim made by strong pure intentionalists but does not
imply it, since it allows that an experience could have non‐representational phenomenal
properties which (for some reason) supervene on the experience's having the
representational content it does. It seems to me that the distinction between strong and
weak pure intentionalism is of little significance in the present context, and in what
follows I will accordingly understand pure intentionalism as the weaker, supervenience
claim.
As noted, pure intentionalists tend to hold that all intentional content is propositional.
Since I reject this thesis, this is one reason for rejecting pure intentionalism as (p. 480)
usually understood. But maybe this thesis is not essential to pure intentionalism; it does
not seem that there is anything in the essential idea behind pure intentionalism which
implies that content must be propositional. So I want to concentrate here on a problem
which is closer to the heart of pure intentionalism.
In a recent paper (2003) David Bain defends the pure‐intentionalist view that the
experience of pain represents damage to the body. Part of his defence involves addressing
what he calls the ‘distinctiveness’ problem: how a pure intentionalist should characterize
what is distinctive about pain sensations. At one point he expresses this problem in terms
of how one can distinguish between (for example) seeing that one's body is disordered
and somatosensorily feeling the same thing:
(2003: 516)
Some pure intentionalists (e.g. Dretske 2000: 458) respond by insisting that such experiences
must in fact differ in what they represent. Others (e.g. Tye 1995) say that the contents of the
experiences themselves have different natures (for example, the content of a somatosensory
experience is non‐conceptual). Bain's own view is that what is represented does not differ in
these two cases, but that there is a difference in ‘modes of presentation’ (i.e. their Fregean
content) (2003: 517–18).
These responses are really very similar: they all attempt to locate the differences in these
experiences in their intentional contents. Yet it seems to me that there is a much simpler
response available, which is unavailable to pure intentionalists like Bain and Tye
(although it is quite within the spirit of their view). This simple response is that the
difference between feeling one's leg to be damaged and seeing it to be damaged is just
the difference between feeling and seeing. In other words, it is a difference in what Searle
and I call mode, and what others would call attitude. We already know that sameness of
content does not suffice for sameness of mental states in general; a belief and a hope
Page 7 of 23
Intentionalism
might have the same content. So why should we expect that it suffices for sameness of
phenomenal states, states which are distinguished by their phenomenal character?
Therefore, I think we should reject pure intentionalism and deny that the phenomenal
character of an experience is determined by its representational content alone.9 The
alternative way of developing the view that all mental states are intentional is what I call
(following Chalmers 2004) impure intentionalism.10 This says that the phenomenal
character of an experience is determined by its entire intentional nature; in particular, by
its mode and its content. The phenomenal character of an experience therefore
supervenes on its intentional nature. There cannot be two experiences which are identical
in their intentional nature but differ in their phenomenal character.
Page 8 of 23
Intentionalism
Note that both pure and impure intentionalism are actually strengthenings of the original
claim (which I called ‘Brentano's thesis’) that all mental states are intentional. For,
someone could hold that all mental states are intentional but nonetheless reject
intentionalism and representationalism if they were to hold that two experiences could be
identical in their intentional nature but differ in their non‐intentional (p. 482) properties
(sometimes called ‘qualia’). Call this the ‘qualia theory’. The qualia theory might be
considered a version of the view that all mental states are intentional, but one that stops
short of intentionalism by holding that there are non‐intentional properties which
contribute to phenomenal character. Impure intentionalism, by contrast, insists on the
determination of the phenomenal character of an experience by its intentional nature
alone, and therefore implies the rejection of the qualia theory.11
So far we have only said what the doctrine of impure intentionalism is; we have not yet
said anything about reasons for believing it. This will be the task of the next section.
What positive reasons are there for believing that conscious states are entirely intentional
at all? Pure intentionalists sometimes argue by appealing to what has come to be called
the transparency of perceptual experience: the idea that introspective reflection on a
perceptual experience reveals the external objects of experience and their features, and
does not reveal intrinsic features of experiences themselves (Harman 1990; Tye 1995).
There has been much discussion of the transparency of experience in recent years (see
Martin 2002; Siewert 2003; Stoljar 2004). Here I shall understand it as the combination of
these two claims: (i) that reflection on a perceptual experience reveals aspects of the
external objects of experience; and (ii) that reflection on a perceptual experience does not
reveal aspects of the experience itself. The first claim is relatively uncontroversial,
although it does raise the question of what a defender of transparency should say about
objects of experience which do not exist (see Johnston2004; Crane 2006a). It is the
second claim which is normally disputed by those who reject transparency. These
philosophers claim that in certain cases—e.g. blurry vision—one can come to be aware of
something more than how one is representing the world to be, or of the objects of
experience. When one sees something in a blurry way, one does not necessarily represent
Page 9 of 23
Intentionalism
the world as blurry. Hence, the second transparency claim (ii) is false, and in so far as
pure intentionalism implies this claim, pure intentionalism is false too.
There are, of course, pure‐intentionalist responses to this kind of objection (Tye 2000).
But given that we have already rejected pure intentionalism (Section 28.3), I will not
pursue them here. Let's consider instead what someone might say who rejects pure
intentionalism and the second transparency claim. There seem to be at least two
possibilities. The first is the impure‐intentionalist response: that blurry vision and other
such phenomena are part of the intentional character of the experience, since phenomena
like this can result from the particular intentional mode of the experience: it is the
particular way of being aware of the world, the particular acuity of one's visual
perception, which (in addition to its content) determines the phenomenal character of the
experience. After all, individuals differ in their perceptual acuity; it is hardly surprising
that this should affect the phenomenal character of the experience. The second view is
the qualia theory: that blurry vision and other such phenomena result from experiences
having certain non‐intentional qualia, understood as properties which can vary
independently of the intentional nature of an experience.
It can be rather hard to see what is at issue in the debate between these two views. One
reason for this is that it is not always clear exactly what non‐intentional qualia are
supposed to be (for discussion see Dennett 1988; Martin 1998). But what does seem to be
clearly in dispute is the truth of the supervenience thesis. Does fixing the intentional
nature (mode and content) of the experience fix its phenomenal character? The impure
intentionalist says yes, and the non‐intentionalist says no (Block 1990, 1996). To sharpen
the debate, then, perhaps we do not need a positive account of qualia. Rather, we only
need treat non‐intentionalism as a negative claim: the denial of the supervenience of
phenomenal character on intentional nature. This puts the impure intentionalist on the
defensive: they need to explain why intentional mode and content determine phenomenal
character.
One thing which is inadequate about the mere appeal to transparency and introspection
is that it does not help us understand why it is that experience is wholly intentional. If we
want to know why it is that phenomenal character is determined by intentional nature,
then we don't just need a statement of what some philosophers take to be obviously true,
we want some kind of understanding of the intrinsic connection between the idea of the
intentional and the idea of phenomenal character. An analogy. It may well be that
universal suffrage is a feature of all liberal democracies. Hence, being a state with
universal suffrage supervenes on being a liberal democracy. But merely stating the
supervenience connection will not explain why this is so; what we need is some
explanatory connection between the idea of liberal democracy and the idea of universal
suffrage.
(p. 484)
Page 10 of 23
Intentionalism
It will be helpful to return for a moment to pure intentionalism. Alex Byrne (2001) has
provided a direct argument for pure intentionalism which does not simply appeal to
either of the intuitions about transparency. Byrne's discussion is instructive, but it seems
to me that his argument is ultimately unsuccessful. Nonetheless, a brief examination of
the argument will indicate an underlying reason for believing in impure intentionalism.
(1) If a (suitably idealized) subject has two consecutive experiences which differ in
phenomenal character, then the subject will notice the difference.
(2) If a subject notices a change in the phenomenal character of these experiences,
then the way things seem to the subject will be different in each case.
(3) But the way things seem to the subject is the content of the experience: if two
experiences share a content then the way things seem is exactly the same in each
experience.
(4) Therefore if experiences differ in phenomenal character they differ in content.
The key move in this argument is the step between (2) and (3). The idea is that if the way things
seem to the subject changes so does the content, since the way things seem to the subject just is
the content of the experience.
Now it is certainly plausible to say that ‘the way things seem to the subject’ can be read
as a synonym for the content of an experience. But someone who doubts the pure‐
intentionalist supervenience claim can say that the ‘way things seem to the subject’ can
be read in another way too: it can pick out aspects of the phenomenal character of the
experience itself. It could be said that when one experiences something the world seems
a certain way to the subject; but, in addition, having the experience also seems a certain
way: ‘things’ covers both aspects of this seeming. Hence, an opponent of the
supervenience claim can say that the argument equivocates between stages (2) and (3):
in (2) the phrase ‘the way things seem’ picks out the content and aspects of the
experience itself, while in (3) it picks out simply the content. Byrne's appealing argument
therefore fails.
Nonetheless, there is something we can learn from the failure of this argument. What
Byrne's argument relies on is that in the case of perceptual experience at least the
notions of how things seem to the subject and how an experience represents the world to
be are intimately intertwined. Suppose I am asked how things perceptually seem to me
now, and I give a description of this. Then suppose I am asked how my perceptual
experience represents the world to be. It is reasonable to expect that I might give the
same or a similar description (see Strawson 1979). Or, rather, if I did then I wouldn't
obviously be failing to carry out the task asked of me. After all, how things seem to me in
perceptual experience is at least a matter of how the world (p. 485) around me is
Page 11 of 23
Intentionalism
Someone might think that the descriptions must differ because ‘how things seem to me’
must refer to the ‘seeming’ itself and not what seems. But this would simply be a
stipulation about how to use the phrase ‘how things seem’, and would therefore be
deliberately ignoring the ambiguity I have drawn attention to. I am not saying that when
you describe how things seem to you, you should only describe the objects of experience.
If I say, because I am short‐ (or near‐) sighted, that a distant building looks blurred, I
need not be describing the object as blurred. Rather, I am describing the object and how
it seems to me. My point here is that describing accurately how things seem to you in
perceptual experience will involve describing the objects of experience, even if it also
involves other things as well. Byrne's argument trades on this fact. (The point is not just
about the phrase ‘how things seem’: the same could be said for ‘what it is like’ and ‘how it
is with me’; see Martin 1998.)
Like Byrne, I have been talking about perception, and it may be thought that the situation
is different where bodily sensations and some emotions are concerned. I will return to
this question in the next section. But for the moment I would like to draw out what I see
to be the moral of this discussion of Byrne's argument.
I have claimed that Byrne's argument fails because the notion of how things seem to the
subject may be read in two ways: as describing how the world seems to be and how the
experiencing ‘seems’ to be, or what reflection on the experience itself yields. Hence, the
step from (2) to (3) equivocates. But the fact that the notion has these two readings is
very significant for impure intentionalism. For what we are trying to describe when we
describe an experience is the subject's perspective on the world, the subject's point of
view. A description of the subject's point of view is not a description of (e.g.) the
arrangements of some ‘blank’ or ‘blind’ intrinsic properties; it is a description of a point
of view on something. Already contained within the idea of how things seem to the
subject is the idea of a perspective or point of view on ‘things’. The same is true for the
idea of what it is like to have an experience. A description of what it is like to experience
something visually is inevitably a description of what it is for this thing to be experienced.
If you leave this out, you leave out part of what it is like for the subject, part of what
makes the experience have the phenomenal character it does. This is why it is so easy for
Byrne to move from talk about how things seem to the subject (‘what it is like’) to talk
about the experience's representational content (‘how it represents the world to be’). Of
course, the move is very natural, for few these days will deny that what it is like to have a
perceptual experience is partly a matter of how the experience (p. 486) presents or
represents the world to be.14 Many of Byrne's opponents will accept this, but they will
nonetheless deny the supervenience thesis.
However, the best explanation of the fact that the notions of how things seem and how
things are represented to be are so close to one another, it seems to me, is that they have
a common core. They describe what is captured at a more abstract level by the idea of
Page 12 of 23
Intentionalism
There is a further reason for adopting impure intentionalism, which has been proposed by
a number of writers recently and which deserves brief discussion here. This is the view
that a (pure‐ or impure‐) intentionalist conception of mental phenomena will render them
more amenable to a physicalist reduction (see Byrne 2001; Hellie 2002). According to a
widespread view, there are two challenges which any physicalist account of mind faces:
the reduction of content and the reduction of consciousness. An aspect of this widespread
view is that while progress has been made in the attempt to reduce representational
content—for example by explaining it in terms of information or causal covariation—little
progress has been made in reducing consciousness. While content is widely considered to
be based on causal relations to the environment, there is almost no consensus about the
physical basis of consciousness. Some see the advantage of (pure or impure)
intentionalism as offering a way out of this deadlock (Byrne 2001: 7).
I am sceptical of this kind of motivation for intentionalism, for two reasons. First, I am
sceptical about the prospects of reduction of intentionality. The fundamental (p. 487)
doubts about misrepresentation which arose in the discussions of naturalized
intentionality in the late 1980s have still not been answered (see Chalmers 2004 for a
recent expression of scepticism). Second, even if prospects were better for a reductive
account of intentionality than they actually are, this would not help with the problem of
reducing consciousness, unless it were explained what it is about certain intentional
states that made them conscious. Certain reductive pure‐intentionalist theories (like Tye's
so‐called PANIC theory 1995) do attempt to address this question, by drawing attention to
specific features of the intentional content of conscious states. But the difficulty here is
that to the extent that there was an original worry about consciousness and its
Page 13 of 23
Intentionalism
explanation, this worry will carry over to the pure‐intentionalist account in terms of
representational content. For example, if the worry was that zombies seem to be possible,
then (given the manifest possibility of non‐conscious intentionality) this worry will arise
even on a (pure‐ or impure‐) intentionalist approach.
I doubt, then, whether intentionalism can really help a reductive account of mind. But, as
I see it, its appeal for reductionists is not the main reason for believing in intentionalism.
The line of thought developed earlier in this section is a defence of intentionalism as a
phenomenological thesis: a thesis about what it is like to be a subject of experience. As a
phenomenological thesis, intentionalism offers the prospect of a unitary account of states
of mind; an account which explains the unity of our concept of mind. Our concept of
mind, on this account, essentially involves the concept of a subject's point of view on the
world. The essence of intentionality is having such a point of view. There are those who
are sceptical about the unity of the concept of mind. But given the connections just drawn
between phenomenal character and intentionality, the burden is now upon them to
explain why intentionality is not the mark of the mental. In the final section of this
chapter I will consider what some claim to be the strongest objections to this thesis.
A non‐intentionalist account of pain is one which says either that pain has no
intentionality or that it is not exhausted by whatever intentionality it may have. On these
views, the phenomenal character of pain is wholly or partly characterized by its non‐
intentional qualitative properties or qualia. Many non‐intentionalists have come to accept
that in so far as they are felt to have a location, pains exhibit some (p. 488)
intentionality.15 But they insist that there is more to these experiences than their
intentionality: they also have their characteristic qualia.
Pure and impure intentionalists resist this, normally on the grounds that the notion of
qualia is obscure and ill‐defined. But it is not enough to say this; they also have to explain
what it means for a pain or a mood to be an intentional state.
Some pure intentionalists (Tye 1995; Bain 2003) have argued that the representational
content of a pain in a part of one's body is that the part of the body is damaged or
otherwise disturbed or disordered (the view derives from Armstrong 1968). The view has
a number of advantages, not least of which is the way it connects the representational
content of pain to its manifest function of alerting an organism to harm which has been
Page 14 of 23
Intentionalism
done to its body. But nonetheless it seems to me that the view cannot be correct. The
main reason is that it is phenomenologically implausible, and the task we are engaged in
here is a phenomenological one. Although there might well be cases where an experience
of pain in a part of the body seems also to be an experience of damage to that part, there
are also many cases where it does not. There is nothing in an experience of a headache
which connotes damage to the subject, and it would be entirely irrational for the subject
to conclude that his head was damaged purely on the basis of a headache. (To insist that
nonetheless the headache is experienced as a disorder in the head may be true, but it
hardly helps the thesis unless we know what kind of disorder it is, which is what the idea
of ‘damage’ was meant to do for us. The mere idea of disorder is surely too unspecific to
single out what is distinctive of the phenomenal character of pain.)
To this kind of criticism Tye (1995) responds that the content of the pain experience is
‘non‐conceptual’ and this is why an experience of pain does not connote damage to the
subject. But the non‐conceptual character of this experience does not seem to be relevant
here. For suppose the experience does have a non‐conceptual content in the way that it
has been claimed that visual experience has a non‐conceptual content. In the visual case
the differences between colours which are supposed to be non‐conceptually represented
are themselves phenomenologically salient (Evans 1982). But our criticism of the damage/
injury theory is that damage is often not phenomenologically salient at all, conceptually
or not. Tye might respond that the ‘damage’ proposal is meant to be one about
‘subpersonal’ information processing below the level of consciousness. But this would
then be changing the subject; he would not in this case be offering an account of the
conscious character of the experience.
We should, I think, reject the pure‐intentionalist idea that the content of pains must be
characterized in terms of damage. And, as we saw above, the pure‐intentionalist view that
phenomenal character is wholly determined by the content of an experience is
independently implausible. So an intentionalist account of pain should rather explain the
phenomenal character of pain in terms of three things: intentional mode, intentional
content, and intentional object, where these do not involve a representation of damage to
the body.
(p. 489)
On the picture I recommend, the intentional object of a pain is the felt location of the
pain, the part or region of the body which hurts. Treating the location or apparent
location of a pain as its intentional object allows us to say, as in other cases of
intentionality, that the intentional object of a pain transcends the experience itself, and
might in certain cases not exist, or not be real. (For example, this is how we should treat
the case of a phantom limb.)
The intentional content of the pain is the representation of its felt location. This
representation is always representation in a certain way or under a certain aspect. For
example, if one feels that one has a pain in one's arm, the arm is the intentional object of
the experience, and its content is the representation of one's arm as one's own. This is the
Page 15 of 23
Intentionalism
In some cases, although it might not be immediately obvious what the intentional object
of a mood is, it may have an object which is revealed by further examination. So it is with
those moods whose objects are their causes. For example, you might feel generally
irritated and it not be clear to you what you are irritated about, but only on reflection do
you realize that it is the presence of your aged relative who is both the cause and the
object of the irritation. So it is, too, with another kind of case: those moods whose objects
can only be characterized in the most generalized way. (p. 490) A mood of depression can
have as its object ‘things in general’ or ‘everything’—a point underlined by the important
commonplace that the depressed person and the non‐depressed person live in different
worlds.17
Further reflection on this kind of case opens up the possibility of a slightly different
approach. One way of spelling out what is meant by saying that the depressed and the
non‐depressed ‘live in different worlds’ is to say that a mood like depression can be
something which affects a subject's entire mental condition: not just every conscious
episode of thought and experience, but the subject's motivation, imagination, and action
are all permeated with the mood. In this kind of case, then, perhaps we should not think
of the mood as an individual mental state in its own right, so to speak, but rather as a
Page 16 of 23
Intentionalism
commonality among all the mental states of someone experiencing such a mood. In
general, it is plausible that every conscious episode of thought, perception, or desire has
a certain affective ‘colouring’ to it. Objects are presented to us as meaningful in various
ways, and part of this meaning is their affective significance: objects can seem lovable,
valuable, in need of care, frightening, or nauseating. This can be, depending on the case
in question, an aspect of their content or an aspect of their intentional mode. If this is so,
then we can see how it might be that an aspect might be common across many different
mental states, how such an aspect might give a thoroughly negative colouring to all one's
experiences, and how we might generalize across this aspect by calling it a mood of
depression. A mood like this might therefore be a general way in which experience might
be modified, a way which can be common across many different kinds of mental episode.
To what extent each recognizable mood fits into one of the intentionalist classifications
discussed is something which will need further investigation. The category of the
emotions (including moods) is a category in which it is crucial to pay attention to
individual differences between the mental phenomena in question. It may be unlikely that
one general theory will apply to all the things we recognize as emotions. But nonetheless
we have found no good reason to think that there are emotions which lack intentional
objects altogether.
28.6 Conclusion
The general idea that intentionality is the ‘mark’ of the mental can be developed in a
number of ways. Different developments will differ in what they say about intentionality,
but also in what they say about consciousness since, as I argued above, the general idea
only becomes controversial where consciousness is involved. In this chapter I have
considered two such developments: the pure‐intentionalist thesis (p. 491) that the
phenomenal character of a conscious state of mind is determined by its representational
or intentional content, and the impure‐intentionalist thesis that the phenomenal character
of a conscious state is determined by its intentional content and its intentional mode. I
argued for the phenomenological superiority of impure intentionalism. In addition, there
seems no good motivation for pure intentionalists to insist on their conclusion. The
guiding idea behind pure intentionalism—captured in Dretske's slogan that ‘all facts
about the mind are representational facts’—can be better developed by the impure
intentionalist than by the pure intentionalist.18
References
Armstrong, D. M. (1968), A Materialist Theory of the Mind (London: Routledge & Kegan
Paul).
Page 17 of 23
Intentionalism
—— (2003), ‘Mental Paint’, in M. Hahn and B. Ramberg (eds.), Reflections and Replies:
Essays on the Philosophy of Tyler Burge (Cambridge, Mass.: MIT Press), 165–200.
Davidson, D. (1970), ‘Mental Events’, in L. Foster and J. Swanson (eds.), Experience and
Theory (Amherst, Mass.: University of Massachusetts Press), 79–101; repr. in Davidson,
Essays on Actions and Events (Oxford: Oxford University Press, 1980), 207–25.
Dennett, D. C. (1969), Content and Consciousness (London: Routledge & Kegan Paul).
(p. 492) Evans, G. (1982), The Varieties of Reference (Oxford: Clarendon Press).
Page 18 of 23
Intentionalism
Grahek, N. (2001), Feeling Pain and Being in Pain (Oldenburg: Hanse Institute for
Advanced Study).
Horgan, T., and Tienson, J. (2002), ‘The Intentionality of Phenomenology and the
Phenomenology of Intentionality’, in D. Chalmers (ed.), Philosophy of Mind: Classical and
Contemporary Readings (Oxford: Oxford University Press), 520–33.
Page 19 of 23
Intentionalism
Moran, D. (1996), ‘Brentano's Thesis’, Proceedings of the Aristotelian Society, suppl. vol.
70: 1–27.
Parsons, T. (1980), Non‐existent Objects (New Haven, Conn: Yale University Press).
Putnam, H. (1975), ‘The Meaning of “Meaning”’, in Putnam, Mind, Language and Reality:
Philosophical Papers, ii (Cambridge: Cambridge University Press), 215–71.
Salmon, N., and Soames, S. (1988) (eds.), Propositions and Attitudes (Oxford: Oxford
University Press).
Stainton, and C. Viger (eds.), New Essays in the Philosophy of Language and Mind, Suppl.
vol. of Canadian Journal of Philosophy (Calgary: University of Calgary Press), 341–90.
Strawson, P. F. (1979), ‘Perception and Its Objects’, in G. Macdonald (ed.), Perception and
Identity: Essays Presented to A. J. Ayer with His Replies (London: Macmillan). repr. in A.
Noë and E. Thompson (eds.), Vision and Mind: Selected Readings in the Philosophy of
Perception (Cambridge, Mass.: MIT Press), 91–110.
Notes:
Page 20 of 23
Intentionalism
(1) See Moran (1996) for a historical discussion of this thesis and Crane (1998) for
further elaboration of the thesis; for Brentano's views see Brentano (1874/1995); for
some discussions of their contemporary relevance see Crane (2006b).
(2) Some philosophers think that certain non‐mental states and events exhibit a kind of
intentionality: for example, the rings of a tree can represent or indicate their age (see
Dretske 1981).
(4) In saying this I side with Husserl (1901/2001) as opposed to Brentano (1874/1995).
Brentano originally thought that the objects of intentional acts were phenomena (i.e.
appearances); Husserl rejected this and emphasized the transcendence of the object.
(5) Here I follow John Searle (1983). Husserl (1901/2001) calls this difference a
difference in intentional ‘quality’; David Chalmers (2004) uses the term ‘manner’. Those
(like Davidson 1970) who think that all intentional states are propositional attitudes
would call it a difference in ‘attitude’.
(6) Hilary Putnam (1975) employs a similar principle (‘intension determines extension’) in
his argument for the claim that ‘meanings ain't in the head’. See Farkas (2006) for
discussion of such principles in connection with externalism and internalism.
(8) Note: (i) mode is not Fregean mode of presentation, which is an element of the
content of the state. Chalmers (2004) prefers his term ‘manner’ to my ‘mode’ because of
the risk of confusion with ‘mode of presentation’. (ii) To make this claim about modes or
manners is not to agree with Lycan's view (1996: 11) that the functional role of a state of
mind contributes towards its phenomenal character. To get here one would have to add
the further claim that differences in mode are explained by differences in functional role
—a plausible thesis but not one which follows from intentionalism. Here I disagree with
Block (2003); my sympathies are more with McLaughlin (2003), Jackson (2004) and
O'Dea (forthcoming).
(9) See Chalmers (2004) for an additional plausible argument against pure intentionalism.
(10) Terminology has really started to grow wild here: Byrne (2001) calls this view
‘intermodal intentionalism’, while Block (2003) calls it ‘quasi‐representationism’.
Page 21 of 23
Intentionalism
(11) For the qualia theory see Block (1990, 2003). In Crane (2003), written in 1998, I
called this version of the qualia theory a version of intentionalism (‘weak’ intentionalism).
Since this theory does not say that phenomenal character is determined by intentionality,
it now seems to me misleading to call this a form of intentionalism at all.
(13) Disjunctivists such as Martin (2002) will disagree; for them, the objects and
properties we experience are not represented but instantiated in veridical perception. In
Crane (2006a) I discuss the difference between disjunctivism and intentionalism in terms
of whether their proponents think experience is representational. Here I am not providing
an argument against disjunctivism, but describing the options for someone tempted by
intentionalism.
(14) Hence it is very unnatural, as many recent philosophers have noted (Siewert 1998;
Horgan and Tienson 2002; McCulloch 2003; Chalmers 2004) to think that the notion of
‘what it's like’ only has application to the qualia of mental states, conceived of as non‐
intentional intrinsic properties.
(15) Even Searle (1992: 251) accepts this. For further criticism of the non‐intentionalist
views of pain see Crane (1998: sect. 2) and Grahek (2001).
(17) For more on the intentionality of moods see Crane (1998: sect. 3) and Goldie (2000:
143–51).
(18) Thanks to Katalin Farkas, Hong Yu Wong and Dan Zahavi for discussion, and to
Mario De Caro and Alfredo Paternoster for their helpful comments at the first meeting of
the Italian Society for Philosophy of Mind and Cognitive Sciences at the University of
Calabria in June 2006. This chapter was written with the support of the EU Sixth
Framework NEST (‘new and emerging science and technology’) project, REFCOM (‘the
origins of referential communication’).
Tim Crane
Page 22 of 23
The Content of Perceptual Experience
Print Publication Date: Jan 2009 Subject: Philosophy, Epistemology, Philosophy of Mind
Online Publication Date: Sep 2009 DOI: 10.1093/oxfordhb/9780199262618.003.0030
Our waking life involves a constant stream of perceptual experience. But what is it,
exactly? What is the content of perceptual experience? This is a very difficult question to
answer satisfactorily, and it has been made a great deal more difficult by the fact that
almost all the key terms in the debate have been used in different ways, and sometimes in
mutually incompatible ways. This article begins by establishing some basic points and
terms of discussion, and asking readers to pay careful attention to the use the article
gives to terms, which may not be the same as their own preferred use. It treats the notion
of perceptual experience as non-factive, unlike the notion of perception, and lacking the
causal implication of the notion of perception.
Keywords: perceptual experience, waking life, notion of perception, phenomenology, sensory experience,
intentionality
29.1 Introduction
OUR waking life involves a constant stream of perceptual experience. But what is it,
exactly? What is the content of perceptual experience? This is a very difficult question to
answer satisfactorily, and it has been made a great deal more difficult by the fact that
almost all the key terms in the debate have been used in different ways, and sometimes in
mutually incompatible ways. I will therefore begin this brief account by establishing some
basic points and terms of discussion, and asking my readers to pay careful attention to
the use I give to terms, which may not be the same as their own preferred use. I will have
nothing to say directly about the rather technical debate that has recently arisen between
those who hold that the content of perceptual experience is wholly conceptual, and those
Page 1 of 22
The Content of Perceptual Experience
who reject this view (see e.g. McDowell 1994; Peacocke 2001), because I believe it
construes the content of perceptual experience far too narrowly.
The verb ‘perceive’ is what philosophers commonly call a ‘factive’ or ‘success’ verb, as
are all the more specific verbs of perception, such as ‘see’, ‘smell’, ‘hear’, ‘taste’, and
‘touch’. Thus, if John perceives—sees—a cow, it follows that there is a cow that he sees,
and if he sees that the cow is brown, then she is brown.1 It is also (p. 495) overwhelmingly
plausible to say that for any verb of perception ϕ and object O, if John ϕs O then there is a
causal relation between John and O of the kind specified by the verb.2 In this article,
however, I am not going to work primarily with the notion of perception, but, rather, with
the wider notion of perceptual experience. I am going to treat the notion of perceptual
experience as (a) non‐factive, unlike the notion of perception, and (b) lacking the causal
implication of the notion of perception. In my terms, then, one can have a perceptual
experience of a cow—i.e. a perceptual experience as of a cow—without there actually
being a cow to hand, let alone a cow with which one is in causal contact.
Plainly all perceptions are perceptual experiences, on the present terms, but not all
perceptual experiences are perceptions. A wholly realistic hallucination of a cow is a
perceptual experience, on the present terms, but it is not of course an actual or ‘veridical’
perception of a cow, although it is by definition subjectively indistinguishable from an
actual or veridical perception of a cow (for it is ‘wholly realistic’).3 Given these terms, one
traditional approach to the problem of perception may be said to focus on the question of
what makes a perceptual experience a (veridical) perception, and I will say something
about this in due course. The philosophical discussion of perception has, however, been
excessively entangled with attempts to deal with the problem of scepticism about the
external world, and this has caused a great deal of conflation of issues that can be largely
avoided by focusing on the question of how to characterize the content of perceptual
experience, properly and broadly construed.4
29.2 Phenomenology
The very first thing to say, perhaps, is that all perceptual experiences are experiences.
They are conscious experiences that represent or present 5 things as being a certain way;
and an experience is something that has, essentially, a certain phenomenology; that is, a
certain phenomenological qualitative character, a character which is such that there is, in
a familiar phrase, ‘something it is like’ for the subject of experience to have it. Some, it
seems, define perceptual experience in such a way that it need involve no conscious
experience at all, and allow that non‐conscious robots can (p. 496) perceive things,
motivated by the fact that a creature can receive informational inputs that serve all the
practical purposes of perception without having any experience. Some participants in the
current rather narrow debate, in fact, want to define all perception merely functionally, in
this way, and do not really admit that it involves any phenomenology at all, as just
Page 2 of 22
The Content of Perceptual Experience
defined, even in our own case. In this chapter, however, I will take it to be true by
definition that perceptual states have, necessarily and essentially, a phenomenological
qualitative character.
What more may be said about phenomenology? All perceptual experiences involve
sensory experience, whether exteroceptive (visual, olfactory, auditory, gustatory, tactile)
or interoceptive or somatosensory (e.g. experience of pain, hunger, nervousness, bodily
movement, and so on), and many take it that the word ‘phenomenology’ only covers the
qualitative character of sensory experience. It is, however, clear that there is more to
perceptual experience, qualitatively or phenomenologically, i.e. experientially, than
merely sensory phenomenology. It is clear that there is more to perceptual experience
(as) of a cow lying beside a beach ball than there is to having merely sensory experience
of the array of colours that would be recorded by a photograph taken from where one is. I
will express this point by saying that in any perceptual experience there is non‐sensory
phenomenology, which I will call cognitive phenomenology, as well as sensory
phenomenology.6
Some philosophers seem to deny that there is any such thing as phenomenology as I
understand it, i.e. experiential ‘what‐it's‐likeness’. There is an old programme in analytic
philosophy of mind that seeks, in one way or another, to give a reductive analysis of the
mental in non‐mental terms (‘to reduce the mental to the non‐mental’), and hence seeks
to give a reductive analysis of phenomenology in non‐phenomenological terms (‘to reduce
the phenomenological to the non‐phenomenological’). This inevitably leads to the denial
of the existence of consciousness or phenomenology, as I understand it, because its
nature cannot be fully specified in non‐phenomenological terms, any more than the
nature of the physical can be fully specified in non‐physical terms. It is true that to
reductively identify X with Y is not, on the face of it, to deny the existence of X; it is
simply to say that although X exists, it is really nothing over and above Y. And it is this
fact that gives those who seem to be denying the existence of (real) phenomenology a
way of denying that that is what they are doing. ‘Phenomenology exists’, they say, ‘but it
is just this’, giving some functional specification of it that involves no essential reference
to qualitative character. In the case of phenomenology, however, to reductively identify it
with something non‐phenomenological is to eliminate the existence of (real)
phenomenology while denying that that is what one has done. This is obvious to some but
it is of course disputed. The issue lies outside the scope of this chapter, however, and I am
going to take it as read.
Page 3 of 22
The Content of Perceptual Experience
Perceptual experiences are also said to have content. The notion of content has, however,
been understood in many different ways, so let me begin with the simple proposal that
the content of a perceptual experience is whatever is given to one in having a particular
perceptual experience. It is whatever is given to consciousness, however this is further
characterized. Since I take the phenomenological character of the experience to be part
of what is given in the experience, it is part of the content of the experience, on my view.
One might call it the ‘phenomenological‐character content’ or ‘purely phenomenological
content’, for it is the (sensory and cognitive) phenomenological character of a perceptual
experience considered just as such, and hence considered entirely independently of
anything in the world that the perceptual experience purports to be about. Plainly this is
a fundamental part of the content of a perceptual experience.
Some philosophers now use the word ‘content’ in a way that prevents them from
recognizing the existence of purely phenomenological content as just defined; and since
different people also understand the word ‘phenomenology’ in different ways, as
remarked above, it will be useful to have a neutral term to refer to purely
phenomenological content. I will call it ‘A‐content’. To say that A‐content is part of the
content of an experience is not yet to say that it is part of the intentional content of the
experience, although I will later claim that it is.
What else, other than A‐content, should be included as part of the content of perceptual
experience, or indeed experience in general? To answer this question, consider first what
one might call the ‘content/attitude’ approach to the general notion of content. Suppose
Jane is thinking (happy, disappointed) that the computer in front of her is square.
According to the content/attitude approach, this phenomenon has the structure Rab,
where a and b are names for things and R is the name of a relation. There is (1) the
thinker, a, (2) the thing designated by the that‐clause, b, and (3) the thinking (etc.)
relation in which a stands to b, which I will call the ‘intentional mode’ or ‘manner of
representation’.7 On this view, the thinking (etc.) is a two‐place relation between the
subject and whatever is designated by the that‐clause,8 which is (p. 498) taken to be a
state of the world wholly distinct from the thinking (etc.) itself: that the computer in front
of me is square, for example. Now many philosophers hold that the content of the
Page 4 of 22
The Content of Perceptual Experience
thinking (etc.), by which they mean the intentional content, is simply whatever is
designated by the that‐clause; that is, a state of the world wholly distinct from the
thinking. I will call this ‘B‐content’: B‐content is defined as states of the world wholly
distinct from the thinking (etc.) in question.
When the content/attitude account is then carried over to the case of perceptual
experience, we get the following: When Jane sees that the computer is square, the seeing
is the ‘intentional mode’, the (visual) ‘manner of representation’, and that the computer in
front of me is square is understood to be the intentional content of the perception. The
question now is whether the intentional content of the perception (of perceptual
intentional modes in general) is or is not more than B‐content as just defined.
I said earlier that the content of a perceptual experience is what is presented or given to
one in having a particular perceptual experience. Otherwise put, the idea is that the
content of a perceptual experience is everything that you are aware of in having that
perceptual experience, whether it is in the focus of awareness or at the periphery. I am
now going to argue, with Brentano, that it follows from the fact that you are (by
definition) aware of all the content of a perceptual experience (a) that all the content of a
perceptual experience is intentional content and (therefore) (b) that all its A‐content is
intentional content. Since many philosophers take it that all the intentional content of a
perceptual experience is B‐content, they may read this as saying that all supposed A‐
content is really nothing but B‐content. But this is not at all what I am proposing.
To put the point in other words: given the currently dominant understanding of the words
‘intentional’ and ‘content’, I may be thought to sound like a ‘representationalist’ about
perception, or an advocate of the ‘transparency’ thesis, or a certain sort of extreme
‘externalist’ or extreme ‘direct realist’. But in fact I am none of these things. This is what
I will try to explain.
(1) a representation of things being a certain way in reality that can be wholly
conveyed in terms of B‐content.
If we allow (1), as almost all philosophers do, I think we have no reason to object to picking out
for philosophical purposes
(2) the purely sensory A‐content of the perception.
Page 5 of 22
The Content of Perceptual Experience
(p. 499)(2) is often attacked as an illegitimate abstraction, and it is true that I cannot isolate
(2) and hold it before my mind, in having the perception, in the way that it seems that I can hold
(1) before my mind.9 The sensory A‐content of the perception is, however, fully real, and it is no
less part of what I am aware of than the B‐content. (If there were no sensory‐phenomenological
aspect to my experience, it would not be a perception at all.) And the present claim is that it is
ipso facto part of the intentional content of a perceptual experience, simply because it is part of
what I am aware of in having the perceptual experience, although it is not in the focus of
attention (it is ‘non‐thematic’ or ‘non‐thetic’, in the phenomenologists' terminology). Plainly,
then, the present conception of intentional content is broader—softer—than any conception of it
that requires that anything that is to count as intentional content must in some sense be in the
focus of attention, or must be ‘aimed at’.
We can put (1) and (2) back together again in
(3) the thick, overall phenomenological way reality is represented, both sensory and
cognitive. In the present case this is, on the sensory side, specifically visual, rather
than say tactile, and it has, as visual, further specificity in its details, including those
which stem from my spatial position in relation to the computer, the lighting, and so
on.
This too is part of the content of the perception, part of what is given to awareness, and is
therefore part of the intentional content, on the present view.
The crucial claim, from the present perspective, is that there is always also
(7) some sort of awareness of the experience or experiencing (see Searle 1983;
Siegel 2006).
Page 6 of 22
The Content of Perceptual Experience
It follows from this claim, which is strongly endorsed by Aristotle as well as Brentano
(p. 500)
and many others more recently, and which I will shortly discuss further, that the A‐content of a
perceptual experience is not just part of what is involved in the overall phenomenon of the
occurrence of a perceptual experience, but is also and necessarily part of the intentional content
of the experience.
I am happy to say this. I am happy to say that a visual experience is about its visual
character, among other things, simply because its visual content is, necessarily, part of
what is given to one in having it. But many are closely attached to a narrower use of
‘about’, and I will therefore be sparing with ‘about’ in the remainder of this chapter, in
which I want to clarify my position by briefly considering its relations (p. 501) to a
number of closely connected and currently popular theses and terms already mentioned:
‘transparency’, ‘externalism’, ‘direct realism’, and ‘representationalism’. All of these have
many forms, and I will have to be very selective in the features I pick out.
Page 7 of 22
The Content of Perceptual Experience
29.6 Transparency
According to G. E. Moore, writing in 1903,
the moment we try to fix our attention upon consciousness and to see what,
distinctly, it is, it seems to vanish. It seems as if we had before us a mere
emptiness. When we try to introspect the sensation of blue, all we can see is the
blue; the other element is as if it were diaphanous.
(1903: 450)
Here his suggestion seems to be that the A‐content of an experience is no part of what is given
in the experience, and this is the sort of point stressed by the advocates of ‘transparency’ or
‘diaphanousness’. In ordinary life, they say, when we see a chair we are conscious only of the
chair and not of any experience of the chair. Moore goes on to grant that the sensation ‘can be
distinguished, if we look attentively enough, and know that there is something to look for’ (1903:
450). However, it is not as if it has to be extracted in this way to be part of what is given in
awareness. That is, in my view Moore is wrong if he thinks that there is any sense in which the
‘sensation of blue’ is not already fully given, with the blue, in the experience of blue in which ‘all
we can see is the blue’. So, too, advocates of transparency are wrong if they think that there is
any sense in which the chairish visual phenomenological content is not fully given, with the
chair, in ordinary perceptual experience of a chair in which all we see (and are thinking about) is
the chair.14 Such A‐content is therefore not only part of the existence of the perceptual
experience, it is also part of what is given in the perceptual experience. It is therefore part of
what it is ‘about’, in my broad terms, and it is therefore part of the intentional content of the
experience.
In saying this I endorse the view, which is well expounded and defended in the
phenomenological tradition, that—roughly—awareness always involves some sort of
awareness of awareness. I agree with Aristotle when he writes that ‘if we perceive, we
perceive that we perceive, and if we think, that we think’ (Nicomachean Ethics, IX. 9.
1170a29–b1), although I don't think one should lay too much stress on his explicitly
propositional formulation ‘perceive that we perceive’, since I take it that a creature can
perceive without possessing the concept of perceiving.15 Aristotle notes what is true in
the transparency position, remarking that ‘knowledge (p. 502) and perception and opinion
and understanding have always something else as their object, and themselves only by
the way’ (Metaphysics, X. 9. 1074b35–6; my emphasis). The awareness of the awareness
is in the ordinary case ‘non‐thematic’, in the phenomenologists' terms: it is not explicitly
in the focus of attention.
Aristotle also noted the threat of infinite regress in the claim that conscious awareness
always involves awareness of awareness (De Anima, III. 2). He rightly found it
unthreatening, although the details of his solution are still debated.16 Here I can note
only that one can either hold that this awareness of awareness involves no higher‐order
Page 8 of 22
The Content of Perceptual Experience
operation at all, experience being in some way essentially ‘self‐intimating’, or that the
regress stops at the second order.17 The no‐higher‐order view is well put by Gurwitsch:
(1941/1966: 330)18
in the same mental phenomenon in which the sound is present to our minds we
simultaneously apprehend the mental phenomenon itself. What is more, we
apprehend it in accordance with its dual nature in so far as it has the sound as
content within it, and in so far as it has itself as content at the same time. We can
say that the sound is the primary object of the act of hearing and that the act of
hearing itself is the secondary object, so that it is intentional with respect to itself.
(1874/1973: 127–8)19
Page 9 of 22
The Content of Perceptual Experience
29.7 Representationalism
Many today adopt a position called ‘representationalism’, but they disagree about what it
amounts to. I will take the basic claim of representationalism to be that the
phenomenological character of an experience is its representational content, or, more
weakly, that the phenomenological character of an experience supervenes (p. 503) on its
representational content. On this view, then, any difference in A‐content has to be
reflected in a difference in intentional or ‘representational’ content, the A‐content of an
experience being wholly fixed by or determined by its representational or intentional
content. (From now on I will simply use ‘intentional’ and not ‘representational’.) In
claiming with Brentano that all perceptual‐experience content is intentional content and
that all phenomenological content is (therefore) intentional content I may sound, to
present‐day ears, like a ‘representationalist’. Really, though, we are very far apart.
On the view I favour, the A‐content of a perception is itself an irreducible part of its
intentional content, essentially over and above its B‐content.21 To attempt to provide any
reductive account of it, functional or otherwise, is to wholly misconstrue what one is
aware of, and so misconstrue part of the intentional content of the perception. This is
where the issue of the understanding of the word ‘phenomenology’ discussed above
becomes acute. In attempting a reduction of phenomenology, in particular A‐content, to
B‐content these strong representationalists have really eliminated it so that it can no
longer feature in the perceptual content, and hence in the intentional content, in the way
it actually does.
I don't know whether anyone endorses representationalism in such a strong form, and
there are ‘weak’ forms that seem to genuinely acknowledge the existence of
phenomenology. However, I also disagree with those weaker representationalists who,
while seeming to acknowledge the existence of phenomenology, claim that it features in
experience only as the ‘manner’ or ‘mode’ (e.g. visual, auditory, etc.) in which the
intentional content is presented and is not itself part of the intentional content of the
perception (see e.g. Chalmers 2004; Crane, Ch. 28 above).
Page 10 of 22
The Content of Perceptual Experience
Consider two subjects, X and Y, who are having veridical perceptual experiences of a boy
playing a trumpet. They are, we may suppose, internally—molecule‐formolecule—
identical, and they are therefore also identical in respect of any phenomenological
properties. According to standard accounts of externalism, their experiences may, in spite
of this, have completely different content, typically called ‘wide content’. X may see and
hear Tom (Tom1), while Y, located, perhaps, on a perfect copy of earth, sees and hears a
quite different but qualitatively indistinguishable boy, also called Tom (Tom2). Now what
fixes the content of an experience, according to externalism, is wholly a matter of
external relations (causal and functional) between the subject and the environment, and X
and Y are in completely different environments (numerically speaking). So their
experiences have completely different content.22 According to internalist accounts, by
contrast, content is fixed, at least in part, by the internal features of experiencers, so that
X and Y will have at least some content—typically called ‘narrow content’—in common,
even if their environments differ.23
Page 11 of 22
The Content of Perceptual Experience
Direct realism states that what we perceive, and directly (or ‘immediately’) perceive,
when we have perceptual experience of a boy playing a trumpet that is (p. 505)
appropriately caused in us by a real boy and a real trumpet, and is therefore a case of
perception, is a boy and a trumpet, period. Indirect realism, by contrast, holds that what
we directly or immediately perceive is only an internal phenomenologically propertied
representation of a boy and a trumpet. To that extent we perceive an actual boy and an
actual trumpet only indirectly or (‘mediately’). Indirect realists point out that a veridical
perception of the boy and a completely realistic hallucination can be completely
indistinguishable, and argue that this shows beyond doubt that what we immediately
perceive—what is directly given to us—is only some sort of psychological entity, an inner
representation, and not a boy.24
In both these debates the truth lies between the two opposed parties. The present
account of perception is direct realist in so far as it holds that what we perceive, in this
case, is a boy playing a trumpet, period. But it does not in any way deny the existence of
the internal phenomenological state that makes this so. Being in the phenomenological
state just is directly perceiving the boy, so long as the phenomenological state is
appropriately caused by the boy: for nothing gets in the way, nothing at all. No realist,
however direct, can plausibly deny that there is a causal process involved in seeing the
boy, that perception is a causally mediated process. Nor can any sensible direct realist
deny that there is indeed a phenomenological state involved in perception, at least in the
case of creatures like ourselves. The existence of the phenomenological state can be
made vivid precisely by the point that a realistic hallucination could in principle be
completely indistinguishable from the veridical perception of the boy. We may say, then,
that the perception of the boy is experientially direct, perceptually direct, judgementally
direct; but it is also, of course, causally mediated, and the causal mediation involves, in
addition to light waves, sound waves, and so on, a richly propertied, richly contentful,
internal representation.
The Brentanian proposal that the internal phenomenological aspect of the perception is
part of its intentional content, in addition to the boy, does not interfere with the direct
realism about the boy. The internal phenomenological aspect is obviously something of
which one is directly aware, but it does not follow that the boy is not directly perceived.
This is because being in the internal phenomenological state just is directly perceiving
the boy (given that the external connections are right).
The current proposal is also comfortably externalist in that it takes the boy and the
trumpet to be part of the content of what is perceived, part of the content of the
perception (just as it takes the boy and the trumpet to be part of the content of what is
believed or thought about, if belief or thought is in question). It differs from extreme
externalism in that it denies that everything about the content of a perception is
determined by external relations. It simply sees no reason to deny that whatever the
perceptual experiences of X and Y and the hallucinator have in common (p. 506) (and it is
Page 12 of 22
The Content of Perceptual Experience
ex hypothesi the case that they have something fundamental in common) can be called
part of the content of their perceptual experience.
29.9 ‘Disjunctivism’
Some of those who call themselves direct realists in the current debate deny that the
veridical perception and the hallucination have any content in common. They do this by
defining content in a wholly externalist way, with the result that the hallucination has no
content at all (or has a very different kind of content from genuine perception). This line
of thought prepares the way for a view sometimes known as disjunctivism. In John
Foster's terms, disjunctivists define perceptual experiences that are veridical perceptions
as essentially ‘physically perceptive’; that is, as psychological states whose identity
depends essentially on the existence of the physical object that they are perceptions of
(see Foster 2000). By contrast, perceptual experiences that are not veridical perceptions,
like hallucinations, are not ‘physically perceptive’. This then allows these extreme direct
realists to say that the types of psychological states involved in veridical perception and
hallucination are (radically) different—they are fundamentally different psychological
kinds. In the case of veridical perception we have a psychological state of the physically
perceptive kind, whereas in hallucination we don't have a psychological state of that kind
at all. This, however, seems an uneasy and wholly unnecessary move, because admitting
the commonality between hallucination and veridical perceptions does not interfere in
any way with the ‘direct access’ to objects sought by direct realism.
The present position is thus anti‐disjunctivist. That is, it rejects the idea that a
hallucination and a qualitatively indistinguishable veridical perception can be said to be
psychological states of completely different kinds. Some have tried to identify anti‐
disjunctivism with indirect realism, but we have already seen that the first does not entail
the second, and in considering some of the other terms in which these matters have been
debated I will from now on use the expression ‘anti‐disjunctivist’.
Page 13 of 22
The Content of Perceptual Experience
Some may now ask whether modes of presentation are to be understood as object‐
dependent or object‐independent.26 This is in effect just another way of expressing the
difference between psychological states that are ‘physically perceptive’, in Foster's terms,
and those that are not (although those who accept the terminology of physically
perceptive states may reject the terminology of modes of presentation). According to the
object‐dependent view, a mode of presentation of an object depends on the existence of
the relevant object for its existence; it is essentially object‐involving. If the object in
question does not exist, then neither does the relevant mode of presentation. This view of
modes of presentation, then, goes hand in hand with externalism and disjunctivism. If in
the case of genuine perception the mode of presentation depends on the existence of the
particular object being perceived, then that same mode of presentation cannot occur in
the absence of that object. Thus, the modes of presentation that feature in veridical
perception, and that constitute its content, simply will not feature in hallucination.27
Object‐independent modes of presentation, by contrast, are essentially general in nature.
They are ‘ways in which things can appear’ that are not tied to particular objects. Plainly
the same chairish mode of presentation can feature in a hallucination and a particular
veridical perception.28 Such a view, therefore, may be anti‐disjunctivist.
Page 14 of 22
The Content of Perceptual Experience
McGinn (1989, 1997) accounts for the particularity of perception in terms of the causal
relations that exist between perceptual experiences and the objects they are causally
related to. Perceptual experiences are anchored in the world in virtue of being causally
related to unique objects.29 But one consequence of McGinn's view, which he recognizes,
is that the answer to the question of which particular object a perceptual experience is
about depends only on whatever causes it in the relevant way, even if its internal
components quite radically misrepresent the object in question.
McDowell, following Evans, objects that McGinn's ‘two‐factor’ view does not really
explain the particular‐object‐directedness of perceptual experiences, or indeed thought in
general. On McGinn's view, it seems, perception is not in itself object‐directed. Modes of
presentation are perfectly general and so cannot on their own isolate particular objects.
How can we then characterize perception, or thought—in themselves—as object‐directed?
McDowell concludes by embracing disjunctivism, with its essentially object‐dependent
modes of presentation, on the grounds that it is a consequence of giving a satisfactory
explanation of the object‐directedness of perceptual experiences.
Disjunctivism also has severe troubles, however, as already noted, in insisting that there
is no sense in which subjectively indistinguishable veridical perceptions and
hallucinations can be said to be mental states of the same psychological kind. Some
disjunctivists attempt to capture this commonality with the following description: a
perception of a purple table and a hallucination of a purple table are both subjectively
indistinguishable from a perception of a purple table (see McDowell 1986). What one
wants, however, is some theoretical account of why this is the case, and of why and how it
is that one cannot in a particular case know for sure (even if one is a disjunctivist)
whether one is really perceiving a table or, under hypnosis, merely hallucinating a table.
It is natural, at this point, to say the profound commonality of the two states is due to
their both being of a single psychological kind.
Page 15 of 22
The Content of Perceptual Experience
To some this may seem like begging the question. But whether or not one accepts their
idea, there are surely limits on what the (internal) phenomenological content of
perceptual experience caused in me by the impacts of light waves reflected from a tree
can be like if it is to be true that I am seeing a tree, and indeed seeing that tree. In
conclusion I will briefly sketch a view that seeks to capture some of what is intuitively
correct on both sides. The ‘matching view’, as I will call it, gives an essential role to
internal components of perceptual experience in determining which object is being
perceived, because it holds that in order for a perceptual experience to be about an
object there must be a certain degree of matching between the properties an object has
and the properties the experience represents the object as having. But it doesn't invoke
object‐dependent modes of presentation, or risk disjunctivism.
Suppose one is in causal, sensory, and indeed visual contact with a garden shed.
However, due to a disorder in one's visual system, or perhaps a hypnotic command, when
one looks at the shed one has an experience as of a pink elephant. Despite one's
inaccurate conception one can locate and track the shed. But does one see the shed? It is
arguable that one does not, because one's conception of it is simply too inaccurate. On
this view, to stand in a causal, sensory, and indeed visual relation to an object is not
enough to guarantee that one sees it, and this is so even if this relation allows one to
locate and track the object: to truly see an object, one's experience (conception) of the
object must not be wildly inaccurate.
The general principle, then, is that if there is to be perception of the object then there
must be a certain degree of match between the properties represented in the (internal)
experience and the properties of the object, even if it is impossible to determine the lower
bound on matching with any precision. The claim is that there can be a fatal failure of
match even when the external conditions of perception are sufficiently met. If this is
correct, internal components of perceptual experiences play an essential role in
determining which object (if any) one sees.
The matching view therefore meets McDowell's desideratum that internal components of
perceptual experiences play a necessary role in determining the object of perception, and
the internal component it appeals to is precisely the overall sensory‐cognitive
Page 16 of 22
The Content of Perceptual Experience
phenomenological state, considered now in all its fullness, the very thing in virtue of
whose existence it is true to say that one directly perceives the object. It requires a
sufficient degree of match, however exactly this is to be determined. (The same goes for
thought.) There are, no doubt, many difficulties with this view. My present claim is only
that it seeks to occupy a place that any satisfactory theory of perceptual experience—of
perception—must occupy.31
References
Anscombe, E. (1965), ‘The Intentionality of Sensation: A Grammatical Feature’, in
Anscombe, Metaphysics and the Philosophy of Mind: Collected Papers, ii (Minneapolis,
Minn.: University of Minnesota Press), 3–20.
Burge, T. (1979), ‘Individualism and the Mental’, Midwest Studies in Philosophy, 4: 73–
121.
—— (1991), ‘Vision and Intentional Content’, in E. Lepore and R. Van Gulick (eds.), John
Searle and His Critics (Oxford: Blackwell), 195–214.
Dewey, J. (1896), ‘The Reflex Arc Concept in Psychology’, Psychological Review, 3: 357–
70.
Gibson, J. (1979), The Ecological Approach to Visual Perception (Boston, Mass.: Houghton
Mifflin).
Page 17 of 22
The Content of Perceptual Experience
Horgan, T., and Tienson, J. (2002), ‘The Intentionality of Phenomenology and the
Phenomenology of Intentionality’, in D. Chalmers (ed.), Philosophy of Mind: Classical and
Contemporary Readings (Oxford: Oxford University Press), 520–33.
Kripke, S. (1980), Naming and Necessity (Cambridge, Mass.: Harvard University Press).
McDowell, J. (1984), ‘De Re senses’, in C. Wright (ed.), Frege: Tradition and Influence
(Oxford: Blackwell), 98–109.
—— (1986), ‘Singular Thought and the Extent of Inner Space’, in P. Pettit and J. McDowell
(eds.), Subject, Thought, and Context (Oxford: Clarendon), 137–68.
Pitt, D. (2004), ‘The Phenomenology of Cognition, or What Is It Like to Think That P?’,
Philosophy and Phenomenological Research, 69: 1–36.
Page 18 of 22
The Content of Perceptual Experience
Putnam, H. (1975), ‘The Meaning of “Meaning”’, in Putnam, Mind, Language, and Reality:
Philosophical Papers, ii (Cambridge: Cambridge University Press), 215–71.
Siegel, S. (2006), ‘Subject and Object in the Contents of Visual Experience’, Philosophical
Review, 115: 355–88.
Notes:
Page 19 of 22
The Content of Perceptual Experience
(1) Some philosophers argue that there are non‐factive uses of perceptual verbs (see e.g.
Anscombe 1965).
(2) The causal claim has been challenged, e.g. by Snowdon (1980).
(4) The ‘problem’ of scepticism seems insoluble but also quite unimportant for the
present discussion.
(7) Following Crane (2003) and Chalmers (2004). (It is generally called an ‘attitude’.)
Chalmers distinguishes between manners of representation and (4) modes of
presentation.
(8) Some think that it is three‐place relation between (1), (3), and (4). For a good
statement of this see Schiffer (1995).
(9) In fact, though, it is not at all clear that I can really be said to continue to have the
perception, rather than just a thought, if I am focusing just on the B‐content.
(10) Here I have in mind everything that is true in the ‘enactive’ theory of perception
found, for example, in Dewey (1896), Heidegger (1927/1962), Gibson (1979), Noë (2004),
and others.
(12) Aristotle puts the point by saying that ‘we perceive that we perceive’ (see below).
The former kind of perceiving is obviously different from the latter. In short, the claim is
not that we see that we see. What the first occurrence of perception amounts to is a very
controversial and big issue. See Caston (2002) for some discussion. The present claim is
Page 20 of 22
The Content of Perceptual Experience
simply that whatever its full account, it will involve appeal to phenomenological
properties.
(14) Reid (1785/2002: essay II, ch. 16) gives a very good diagnosis of this error and notes
how natural it is.
(18) It is important that ‘aware of my being conscious of it’ is not read as involving any
consciousness of self considered as such. Gurwitsch's central point in this paper is that
‘the subject in his dealing with the object, aware as he is of this dealing, is nevertheless
in no way aware of his ego, much less of his ego's involvement in his dealing’ (1941/1966:
327).
(19) Many who agree with Brentano's basic position reject his use of ‘object’ in the
phrase ‘secondary object’, but the broad use of ‘object’ amounts to no more than ‘what is
given to the subject’, and in that use I think it can be accepted. For other examples of
non‐higher‐order views see Thomasson (2000), Kriegel (2003), Smith (2005), and Zahavi
(2006).
(20) Note that ‘representationalism’ is used to mean the opposite of what one would
expect it to mean. It is natural to expect it to be used for a theory that distinguishes
sharply between the object of perception and the representation of the object of
perception, even if it is not used, as it might well be, as a name for a full‐blown
representative theory of perception of a Lockean kind, according to which what we
directly perceive, when we perceive a chair, is only a representation of a chair, by virtue
of which we can be said to perceive (but only indirectly) the chair itself.
(22) See Putnam (1975), Burge (1979), Kripke (1980) for classic accounts of externalism.
(23) How much content is internalist is controversial. See Segal (1991) for a wholly
internalist view and Chalmers (2004) for a partial internalist view.
(24) The debate between direct and indirect realism assumes realism about material
objects, as its name suggests. For contemporary accounts of idealism—the view that
Page 21 of 22
The Content of Perceptual Experience
material objects are not logically independent of the human mind—see Robinson (1994)
and Foster (2000).
(25) For present purposes I focus on modes of presentations of physical objects although
the term has a wider use.
(26) For the first see e.g. Evans (1982) and McDowell (1984); for the second see e.g.
McGinn (1989, 1997). It is unfortunate that some use ‘mode of presentation’ and the
Fregean term ‘sense’ almost interchangeably in this debate, and I will stick to ‘mode of
presentation’.
(27) In the absence of material objects, Evans (1982) seems to vacillate between saying
the psychological state in question has no content and the content is different.
(28) There is another alternative in the literature (see e.g. Burge 1991; Bach 1994), which
I will not discuss here, according to which what can be common to a hallucination and a
genuine perception is a gappy entity similar to an open sentence, e.g. x is a purple table.
Only genuine perceptions have object values for the gap.
(29) Of course there are well‐known problems for stating the appropriate causal link
between a perception and the relevant object.
(30) See Evans's criticisms of the ‘photograph model’ (1982: ch. 3).
(31) Thanks to: George Bealer, Brian McLaughlin, Susanna Siegel, Dave Chalmers, Dan
Zahavi, David Smith, and especially Galen Strawson.
Michelle Montague
Page 22 of 22
Phenomenology, Intentionality, and the Unity of the Mind
Phenomenology, the movement, has contributed and continues to contribute much to the
study of phenomenal consciousness as well as to understanding the role of intentionality
in our conscious lives. This article, however, is not about intellectual history or
methodological movements. So it is not about such contributions. There is nothing in this
article about the movement (Phenomenology) but much about the property
(phenomenology). The word ‘intentionality’ is a technical word for the feature of a mental
state in virtue of which it is directed at or is about or represents something other than
itself.
Keywords: phenomenology, intentionality, unity of the mind, mental state, phenomenal consciousness, conscious
lives
‘Phenomenology’? ‘Intentionality’? We have just deployed two technical words. These two
words also occur in the chapter's title. So before taking a single step, or even stating our
intentions in this chapter, something must be said about them.
Page 1 of 30
Phenomenology, Intentionality, and the Unity of the Mind
Phenomenology, the movement, has contributed and continues to contribute much to the
study of phenomenal consciousness as well as to understanding the role of intentionality
in our conscious lives. This chapter, however, is not about intellectual history or
methodological movements. So it is not about such contributions. There is nothing in this
chapter about the movement (Phenomenology) but much about the property
(phenomenology).
The word ‘intentionality’ is a technical word for the feature of a mental state in virtue of
which it is directed at or is about or represents something other than itself. We persons
think about or represent all sorts of things: red sunsets, pains in the neck, the Empire
State Building, and the sun in Arizona.
Page 2 of 30
Phenomenology, Intentionality, and the Unity of the Mind
device with consciousness … I don't think we even know how to begin’ (1998: 102).
However, Kim adds, if asked to design a structure with intentionality ‘it seems to me that
we can go about designing’ such things (1998: 103).
Is Kim right? Can the one (intentionality) occur or be created without the other
(phenomenology)? If so, that would be one‐way independence or separateness of
intentionality from phenomenology. Others say that the phenomenological is independent
of the intentional. John Searle (1983: 2), for example, offers the examples of forms of
elation, depression, or anxiety in which one is not ‘elated, depressed, or anxious about
anything’. ‘Many conscious states’, he says, ‘are not intentional.’ If so, that would be
independence in the other direction: of phenomenology from intentionality.2
Given theoretically different dimensions along which, or sorts of mental states in which,
either the separability or inseparability of the phenomenological and intentional may be
said to occur, various versions of separatism and inseparatism have been proposed or
categorized in the literature (see Siewert 2003; Chalmers 2004).5 (p. 515) Our concern in
this chapter is to spell out separatist and inseparatist theses as well as key terms used in
their description. We say something about why one might hold, as we now do, that
despite its relative unpopularity within Anglo‐American philosophy inseparatism is true.
We mention some prominent objections to inseparatism and say something about how an
inseparatist of the specific variety or type that we favour might reply to each.
Our strategy here is not merely expository, but positive or positional. We aim to defend a
position. Elements of the inseparatist position defended here are described and defended
by us in a series of earlier papers.6 The present chapter is complementary to the papers
in that series. Although entirely self‐contained, it develops the position beyond what is
said there by further exploring ideas of unity of mind, determinacy of intentional content,
and the phenomenal and conceptual components in our conscious intentional lives (see
especially Graham et al. 2007; Horgan and Graham forthcoming).
Page 3 of 30
Phenomenology, Intentionality, and the Unity of the Mind
(p. 516)
Page 4 of 30
Phenomenology, Intentionality, and the Unity of the Mind
understand the meaning of words spoken in your mother tongue. Compare, however,
what it is like to listen to a conversation in your mother tongue (English, let us assume)
with what it is like to listen to a conversation in a language you completely and utterly fail
to comprehend (Swedish, let us assume). The English utterances will immediately sound
different to you than the Swedish utterances, not just as sounds, but as comprehended in
meaning in the first case, and as uncomprehended in the second. Meaning
comprehension, writes Barry Dainton, can be ‘as much a phenomenal feature of what
[you] hear as the timbre and pitch of [a] voice’ (2000: 12).
(p. 517)
On our view there is a difficulty philosophers get into (like Kim, for example) when they
refer to conscious experience period as states with qualitative character but do not
distinguish between qualia in the narrow and broad sense, all the while assuming that
conscious experiences possess qualia in the narrow sense. The difficulty is that some
states (such as consciously thinking of Arizona) typically do not possess anything
narrowly qualitative. They are not endowed with constitutive narrow qualia. Meanwhile
defences of separatism sometimes presuppose that phenomenal character just is narrow
Page 5 of 30
Phenomenology, Intentionality, and the Unity of the Mind
A person can think in a manner that is utterly devoid of uniformly type‐identifiable qualia
of the narrow sort. Of course, some intentional states actually do have immediately
apparent qualities in the narrow (and not just broad) sense and not as mere
accompaniments but as constitutive of the state. If I tell you that I am thirsty and desire a
drink, it may make sense for you to ask, ‘How desirous do you feel?’; and it would not
suffice for me to say, ‘Oh, I just have the desire without feeling anything’. When thirsty
you may feel desirous of a drink and you may report how much you desire it. Desire for a
drink is (as will be appreciated after the discussion of intentionality to follow) an
intentional state. Felt intensity is an essential feature of some instances of desiring.
For inseparatists such as the three of us, fortunately, talk of conscious experience does
not commit one to the view that narrow as opposed to broad qualitative character is
essential to conscious experience. The spectrum of conscious experience, as Galen
Strawson remarks (1994: 4), ranges from ‘the most purely sensory (p. 518) experiences to
the most abstractly cognitive experiences’. These last include ‘the experience of
consciously entertained thought, of reading and understanding, of unplanned fantasy, and
of directed imagining’. Some experiences, especially those on the sensory, mood, and
emotion side of the spectrum, involve proprietary and distinctive feelings (or images), but
others, including those on the cognitive side (such as the experience of consciously
entertained thought), do not. Alvin Goldman writes as follows: A person can be
immediately aware of ‘having a determinate thought‐content’. ‘Entertaining [such a]
conceptual unit has a phenomenology, just not a sensory phenomenology’ (1993: 365).
‘Thoughts’, Owen Flanagan writes (1992: 65) referring to consciously entertained
thought, ‘seem in a certain way’, though, he says, ‘they don't have the sort of robust
qualitative feel that sensations have.’ They needn't, we hasten to add by way of
clarification, have any feel proper to them, robust or not.
Consider the kinds of evidence to which I might appeal in establishing the truth of the
following two statements.
Page 6 of 30
Phenomenology, Intentionality, and the Unity of the Mind
(1) There is a tree over there [said when pointing] that's dying.
(2) There is a tree over there [said when pointing] in front of me.
The evidence that might support the truth of (1) might include the following: what a tree expert
told me about the tree, time‐delayed photographs of the gradual decline of the tree, whether the
tree fails to support bird‐life, and so on.
Let's call the evidence just cited for (1) ‘indirect’ or ‘inferential’ (treating these
synonymously). We are calling the evidence ‘indirect’ (etc.) because we are supposing
that in saying that a tree is dying, and indeed, by extension, in observing that a tree is
dying I infer from the evidence to something about a tree: that it is dying. Given the
indirectness there is inferential room for epistemic error. In such a case I might be
misled. Experts, photographs, and the behaviour of wildlife can mislead in the sense that
they might encourage or permit drawing incorrect conclusions. I might believe that a tree
is dying when it is not.
Let's call the fact or purported fact (in the hypothetical context of the example) that a
tree is dying presentationally mediated (or indirectly evidenced). This fact or purported
fact (a tree dying) is presented to me, as the perceiver, only indirectly, that is, through
such evidence as the testimony of a tree expert, photographs, and so on. It is not, as it
were, inherent in the perceptual experience.
Let's compare the hypothetical evidence for (1) with that for (2). Suppose, in the full
context of the example, that I am a normal perceiver with semantic competence with
concepts like those of ‘tree’ and ‘being in front of’. If so, the evidence that supports the
truth of (2) is, we might suppose, not indirect or inferential. Suppose (p. 519) that there
being a tree in front of me is directly, self‐evidently, non‐inferentially, or immediately (we
are treating these locutions as synonymous) evident or presented to me. After all, to see a
tree as being in front of me I have merely to open my attentive eyes under normal
viewing conditions. I don't have to talk to experts, examine photographs, or birdwatch.
Having conscious content as of a tree in front of me is, as it were, inherent in the
experience or intrinsic to its presented character.
Indeed, to emphasize the immediacy or inherence of the content of the experience, note
that I might have the very same conscious content without there actually being a
presented objective scene of a tree in front of me. When I dream, for example, I might
have the same sort of experience as of a tree being in front of me. There can be such an
experience without there being some suitable kind of causal or evidential connection
between what is going on in my head when I dream and the wider environment.
So in the case of (2) (as described here) there apparently being a tree in front of me is
manifest or presents itself to me without evidential mediation. Tree‐in‐front‐of‐me
immediately appears to me. Four points about such presentational immediacy should be
noted and are germane to the type of inseparatism we favour.
Page 7 of 30
Phenomenology, Intentionality, and the Unity of the Mind
(i) In general, to be immediately appeared to precludes being misled about the content of
the experience, although it does not preclude being misled or mistaken about whether
things truly are as they appear. It might immediately seem to me as if I am in front of a
tree without it's being the case that I am, as it happens, in such a position. I might be
dreaming or hallucinating. There‐seeming‐to‐be‐a‐tree‐in‐front‐of‐me might not be caused
or environmentally associated in the right way (by my being facing the tree) to count as
veridical or as a reliable indicator of the world independent of my experience.8
(ii) Immediate appearance or presentation has two dimensions: what it is of and whom it
is to. Colin McGinn (1988/1997: 298) aptly refers to these components as ‘Janus‐faced’
features of conscious experience (see also Kriegel 2003). Thus, the perceptual experience
of a tree is of something other than the subject (it is of a tree) but it also presents a
subjective face to the subject. There apparently being a tree in front of oneself is like this
to the subject having the experience. Because conscious experience possesses these two
dimensions, we refer (above) to our concept of the content of conscious experience as the
self‐presentational immediacy conception. Conscious content is not just immediate. It also
presents itself to a self (subject) immediately or without evidential mediation.
To different degrees and in different aspects, in the words of William Seager, the entire
realm of conscious experience is ‘ “parsed” by [our] conceptual apparatus and presented
to … consciousness as a world of trees, cars, houses, people’, and so forth (1999: 188).
(iv) One notion of consciousness that may seem close to that of conscious experience as
immediate appearance or self‐presentational immediacy is what Ned Block (1995) calls
‘access consciousness’. A state of a subject is access conscious, says Block, just when it
can be introspectively reported by the person and can govern her motor movements or
behaviour. For example, I can report perceiving as if being in front of a tree and this
perception may govern my behaviour of looking for Tabby the cat, who, I believe, has run
up the tree in Professor Malcolm's garden. Block says that access consciousness is not
one and the same as phenomenal consciousness but is a different sort of consciousness.
We are sceptical. If conscious experience consists of self‐presentational immediacy, this
fact helps to explain why phenomenal content—that which is immediately apparent—is
reportable and can govern behaviour. Other things being equal, access to content is
immediate given that content evidentially is immediate. Or, more precisely: so‐called
access consciousness is not a different type of consciousness at all, but rather is a
reportorial and behavioural competence afforded to a neurologically healthy subject
Page 8 of 30
Phenomenology, Intentionality, and the Unity of the Mind
We are not going to pause for extended discussion of each of the above four points. They
are intended to clarify the notion of immediate self‐presentation. That notion is crucial for
the brand of inseparatism we favour, that is intended to be free of allegiance to the
proposition that conscious states qua conscious possess qualities in the narrow sense of
sensory‐imagistic qualia. We make three general assumptions about conscious experience
culled from the above discussion. (1) Conscious experiences are immediate appearances
or direct presentations. (2) Some conscious experiences are narrowly qualitative in
character; others are not. All conscious experiences, however, are qualitative in at least
the broad sense of being constituted by distinctive and proprietary immediate
appearances or presentations. (3) Conscious experiences are informed by a subject's
semantic or conceptual competence.
Page 9 of 30
Phenomenology, Intentionality, and the Unity of the Mind
30.2.3 Intentionality
(p. 521)
Here are some other examples (among dozens that may be mentioned) of types of states
or conditions with intentionality: memory, fear, hope, love, hate, expectation, anger,
shame, and disappointment. Intentional‐state types are sometimes called ‘attitudes’.
Being intentional means that these states or attitudes are of or about something. I am
angry that such‐and‐such, disappointed about so‐and‐so, etc.
Being directed at something means that intentional states or attitudes have intentional
objects and intentional content (sometimes also known as propositional or
representational content). Suppose, for example, that you believe that a picture in your
living room is hanging crooked. The intentional object is a photograph; the content is that
the photograph is hanging crooked.
The distinction between object and content can be characterized as follows. The
ascription of an intentional state to a person typically involves the use of a that‐clause, as
in ‘I believe that a picture in my living room is hanging crooked’. To the right of the that‐
clause is mentioned the content or proposition on which the intentional state is directed
(‘that a picture in my living room is hanging crooked’). The so‐called (logically atomic)
content includes the topic(s) of the that‐clause (‘the picture in the living room’) as well as
the comment made about the topic (‘is hanging crooked’).
Intentionality though relational is not an ordinary physical relation like that of touching
or kissing. (Whether it is a non‐ordinary physical relation of a sort that physical science is
committed to and can discover is discussed below.) I cannot touch something unless it
exists, but I can think of something (a unicorn) even if it doesn't exist. The intentional
object might not really exist. I can desire to ride a unicorn even though no unicorn exists.
Page 10 of 30
Phenomenology, Intentionality, and the Unity of the Mind
backyard, I can desire that a unicorn be in my backyard, etc. One and the same type of
attitude can harbour different contents. I can believe that a unicorn is in my backyard, I
can believe that the picture is hanging crooked, etc.
Page 11 of 30
Phenomenology, Intentionality, and the Unity of the Mind
What is the disagreement between separatists and inseparatists about? At one level of
description, as Charles Siewert notes (2003: 25), it's a disagreement about ‘how to
conceive of the mind or domain of psychology as a whole’. It is about how to solve the
problem of the unity of mind. Is there something that deserves to be called the unity of
the mental? Inseparatism says yes. Separatism says no.
Separatism says that the intentional mind and phenomenal mind are not intimately
related. One exists on the intentional side of the intentionality/non‐intentionality divide.
The other exists on the phenomenological side of the phenomenal‐character/non‐
phenomenal‐character divide. Meanwhile the phenomenological is non‐intentional and
the intentional is non‐phenomenological.
Inseparatism says that the intentional mind and phenomenal mind are intimately related.
We are not of two distinct minds. Each is, in some sense, co‐present in the other.
Intentionality is, in some sense, proper to phenomenology. Phenomenology is, in some
sense, proper to intentionality. Together they unify the domain of the psychological and
provide the concept of mentality with a univocal sense or meaning (at least in
uncontroversial cases of mentality; see below).
The above statements of separatism and inseparatism are fairly imprecise. A major
reason for this imprecision is that attempts to define the core notions of intentionality and
consciousness traverse essentially contested theoretical territory. So, for example, if it is
assumed that intentionality is more scientifically tractable than consciousness, then one
will believe that intentionality is independent of conscious experience. Or if one assumes,
for another example, that consciousness necessarily is qualitative in the narrow sense
and contains proprietary images or feelings, then, again, one may claim that this means
that consciousness is independent of intentionality. Can such background assumptions be
pruned from initial descriptions of consciousness and intentionality; that is to say, from
descriptions offered before embrace or rejection of either separatism or inseparatism? It
would be helpful and ecumenical to understand concepts like those of intentionality and
phenomenology so as not to beg the question in favour of separatism or inseparatism. We
have attempted to do just that in Sections 30.2.1–3.
Page 12 of 30
Phenomenology, Intentionality, and the Unity of the Mind
perceptual states such as colour experiences, pains, and emotions) have intentional
content that is inseparable from their phenomenal character.
The phenomenology of intentionality (Thesis PI): Mental states that are
uncontroversially intentional (e.g. cognitive states such as belief and conative states
such as desire) have phenomenal character that is inseparable from their intentional
content.
In terms of these two theses we will now state the form of inseparatism that we favour and
which we call moderate inseparatism: Mental states that uncontroversially have intentional
content have phenomenal character (i.e. are conscious), and mental states that uncontroversially
possess phenomenal character have intentional content directed at something (i.e. are
possessed of intentionality). More or less equivalently: In uncontroversial (i.e. paradigmatic,
exemplary, or incontestable) cases of mental intentionality, the fact that a mental state is
intentional entails facts about the presence of phenomenal character; whereas in
uncontroversial cases of phenomenal consciousness, the fact that a mental state is phenomenal
entails facts about the presence of intentional content.
As we are putting things, to say that a mental state is uncontroversially X (mental‐
phenomenal, mental‐intentional) is to say that its being X is incontestable. It's a paradigm
or exemplar of X. It is a state that competent speakers would not question or doubt as
being of the X sort. Pains, for example, are mental states that are uncontroversially
phenomenal; they are incontestable instances of the phenomenally conscious sort. The
desire that a unicorn run across a lawn, for example, is a mental state that is
uncontroversially intentional; it is an incontestable instance of the intentional sort. So
understood, the distinction between uncontroversial/incontestable, on the one hand, and
controversial/contestable, on the other, is a broadly epistemic distinction. It is a
distinction between cases of robust (paradigmatic or exemplary) warrant for classifying a
state as phenomenological or intentional, respectively, and cases in which the warrant
falters or can reasonably be challenged by competent speakers.
Moderate separatism is quite compatible with the claim that if a state can be shown to be
phenomenological or intentional, even if not incontestably so, it, too, might also be
intentional in content or phenomenological in character, respectively. Such a stronger or
less moderate inseparatism, however, would require much more detailed discussion and
examination than we have the space for in this chapter. On our view the inseparability of
consciousness and intentionality, though framed by us as moderate, nonetheless is
pervasive and estimable, raising deep questions not just about the scientific tractability of
intentionality (to be considered later) but about (p. 524) a variety of different issues in the
philosophy of mind including the issue (to be described momentarily) of whether
intentional content is constitutively determined by something ‘inside the head’ or
partially ‘outside the head’.
Page 13 of 30
Phenomenology, Intentionality, and the Unity of the Mind
The first thesis of moderate inseparatism, namely Thesis IP, has advocates—of a sort. So‐
called representationalist theories of phenomenal consciousness are a currently
influential departure from separatism about phenomenology (see Dretske 1995; Tye
1995). According to representationalism, everything in the category of mental states
paradigmatically cited as conscious is included in the category of states that are
intentional or directed at something. Our perceptual experiences inform us about the
local environment. Emotions have intentional objects (as when I feel ashamed for having
committed a cowardly act). According to representationalism, even what John Searle calls
(1983: 2) being ‘simply elated, depressed, or anxious without being elated, depressed or
anxious about anything’, contrary to Searle, is intentional. Objectless depression is a kind
of representation of everything (not of any one particular thing or circumstance) as
negative. It's a way of being conscious of things in general: everything is perceived to be
worthless or pointless. Or, as a depressed character in Ingmar Bergman's Scenes from a
Marriage puts it: ‘Everything's getting meaner and grayer, with no dignity’ (see Church
2003: 175>). Sensations such as pain are experienced as having a (represented) location
(as phantom‐limb pains demonstrate). (See Seager 1999: 132–77 for a generally
sympathetic discussion of representationalism.)
Most extant versions of representationalism, including those of Dretske and Tye, say that
the intentional content of phenomenal states is constitutively determined by certain
objective relations (e.g. certain kinds of correlational relations) between the occurrence
of these states and the presence, in a cognitive agent's surrounding environment, of
certain external properties or features, such as the red surfaces of external objects. This
means, in effect, that phenomenal character ‘isn't in the head’. In some sense,
phenomenal character is constitutively determined in part by what is in the outer world.
The phenomenology of perceptual experience: the enormously rich and complex what‐it's‐
like of being perceptually immediately presented with a world of apparent objects,
apparently instantiating a rich range of properties and relations—including (p. 525) one's
own apparent body, apparently interacting with other apparent objects which apparently
occupy various apparent spatial relations as apparently perceived from one's own
apparent‐body‐centred point of view.
Page 14 of 30
Phenomenology, Intentionality, and the Unity of the Mind
An important epistemological by‐product of the fact that these and other aspects are
intrinsic or internal is that a person can just tell or subjectively discern what these
aspects are—what her conscious attitude content is, for example. The reason for being
able to ‘just tell’ is that conscious aspects qua conscious and qua mentally intrinsic
immediately present themselves. (Remember, for us conscious content is self‐
presentationally non‐inferential or immediate.) Some aspects (say, pains) are like this.
Others (say, thoughts of Paris) are like that. Moreover, being able to just tell is impervious
or resistant to sceptical doubts about such things as, for example, whether one is, as
Descartes envisioned in the First Meditation, the victim of an enormously powerful,
enormously clever deceiver who ensures that conscious contents are radically non‐
veridical, and that one's conscious beliefs are massively false (see Horgan et al. 2006).
Such radical scepticism gets no grip because self‐presentational phenomenology
undergirds the experiencing subject's capacity to just tell what conscious state she is
currently undergoing—although arguably such radical scepticism should get a grip if (as
representationalists typically claim) the constitutive determinants of phenomenal
character are partly ‘outside the head’.
If the picture of conscious content that we prefer is on the right track, then most extant
versions of representationalism are on the wrong track in misconstruing its intrinsicness
as extrinsic. They also underestimate the richness and pervasiveness of the
phenomenological in our mental lives, since most representationalists tend to think of the
domain of the distinctively phenomenological as that of narrow qualia (feelings, images,
etc.) rather than broadly self‐presented (including cognitive as well as narrow) qualia. We
shall assume, however, that representationalism is on the right track at least in saying
that paradigmatic or uncontroversial phenomenological mental states are intentional.
What of the converse proposition that fully‐fledged intentional mental states are
conscious? Has any theorist or philosopher advocated that proposition (Thesis PI)?
Descartes said ‘[t]here can be nothing in the mind … of which it is not aware’ (1993: 171).
On his view, the intentional (being mental) is conscious. Not many theorists agree with
Descartes, though some come close.
(p. 526)
Page 15 of 30
Phenomenology, Intentionality, and the Unity of the Mind
Alvin Goldman claims that ‘a plausible looking hypothesis is that mental states are states
having a phenomenology, or an intimate connection with phenomenological events’ (1993:
24). The phrase ‘intimate connection’ is not clarified by Goldman.
John Searle (1990, 1992) has advocated the connection principle (his expression) that
‘unconscious intentional states are in principle accessible to consciousness’ (1992: 156).
Galen Strawson has said that ‘one cannot have intentionality unless one is an
experiencing being’ (1994: 208). Both Searle and Strawson make the existence of
phenomenology in a subject a precondition for its having intentional states. For Searle a
mental state must be conscious or the person must be able to access the state
consciously, although this is a notion of access about which Searle is none too clear. For
Strawson, the person must be a conscious subject, a requirement that Strawson asserts
but about which, again, at least on our reading, he is none too clear. Strawson writes:
‘[T]here is no … intentionality in a world in which there is no consciousness or
experience’ (1994: 208).
In the effort to defend Thesis PI we might press the texts for clarifying interpretations of
Goldman, Searle, and Strawson. Searle's connection principle has already received a
great deal of attention (see the peer commentary in Searle 1990). Perhaps such
interpretations would uncover convincing arguments for the thesis that uncontroversial
intentional mental states are conscious. We shall not explore these interpretative
possibilities.
Page 16 of 30
Phenomenology, Intentionality, and the Unity of the Mind
Consider the case of what the frog's eye tells the frog's brain (Lettvin et al. 1959/2000;
see also Dennett 1987, esp. pp. 106–16, 302). Suppose that whatever the frog's eye tells
the frog's brain is not consciously processed by the frog. It is not ‘like anything’ for the
frog's brain to be told something by the frog's eye. Suppose, also, for the sake of
hypothesis, that as a result of what the eye tells the brain, the frog comes to have a non‐
conscious perceptual belief that helps to govern the behaviour of catching and
swallowing flies.10 Now suppose the frog is provoked into catching and swallowing a fly
which darts in front of it. If we interpret the signal coming from the eye as ‘telling’ the
frog that there is a fly coming towards it, then we might also interpret the frog's non‐
conscious perceptual belief as the belief that there is a fly coming towards it. However,
now suppose that we provoke the frog into catching and swallowing a lead pellet we toss
at it. If we interpret the signal as ‘telling’ the frog that there is a lead pellet coming
towards it, then we might interpret the frog's perceptual belief as the belief that there is
a lead pellet coming towards it. If we are robustly charitable in our belief attribution,
when situations like those of the fly and the pellet arise we might interpret the frog as
never making a mistake. Every darting object in the relevant pathway that the frog
catches and swallows can always be interpreted by saying that the frog believes that
such‐and‐such a particular sort of object is coming towards it: fly or pellet or moth or
dark moving spot or insect or whatever. This would be excessively inflationary, of course,
endowing the frog with concepts of fly, pellet, moth, and so on. It would also make it
impossible for the frog to believe falsely that a fly is coming towards it, when, as it
happens, a pellet is coming towards it, as it catches and swallows the pellet. For when a
fly darts, the frog (we shall assume) believes that it is a fly. Whereas when a pellet comes,
it believes (we shall assume) that it is a pellet. And so on. In principle, it would seem that
the frog's behaviour can be described and explained with the invocation of any number of
belief contents: that there is an insect flying off to the right, that there is some food now
there, that something yummy is nearby, that there is a pellet, etc.
We will not develop the narrative with other potentially relevant hypothetical details.
What, as told, is the lesson we wish to extract from the story? The lesson is a worry about
intentional content. We might worry not just about how to avoid over‐endowing the frog
with conceptual competence, attributing a more conceptually specific or sophisticated
meaning to the frog's perceptual system than the case warrants. We might also worry
whether the frog ever misrepresents or misinterprets what is coming towards it. Whether
it believes ‘thing of kind F (fly) or thing of kind P (pellet) or kind (etc.) is here now’, its
belief (as we might interpret it given that we have nothing but its behaviour to go on)
might always turn out to be true.
And then we might have a related worry, one closely connected to the defence of Thesis
PI. If the frog never misrepresents its environment, is its perceptual belief (p. 528) really
possessed of intentional content? That is, is the state of the frog that we have called its
believing really a state with content? Remember earlier we presupposed that a defining
mark of intentionality is intensionality or the failure of substitution in the idioms deployed
to characterize intentional content. If the frog believes that a fly is coming towards it, this
must not be the same as believing that a pellet is coming towards it even if, as it happens,
Page 17 of 30
Phenomenology, Intentionality, and the Unity of the Mind
a pellet is coming towards it and the trajectory of the pellet causes the frog to believe
that a fly is coming towards it. (Recall: If Lois Lane believes that she kissed Clark Kent,
this is not the same as believing that she kissed Superman, even if, as it happens, Clark is
Superman and even if kissing Superman caused her to believe that she kissed Clark.)
Intentional content must be sufficiently determinate or individual in its represented
aspect so that there is a genuine difference between, for example, believing that a fly is
approaching and believing that a pellet is approaching and thus between representing
and misrepresenting the environment. It is one thing to represent the pellet as a pellet,
another to (mis)represent it as a fly.
Perhaps the content that the frog employs is indeterminate or non‐aspectual, lacking
determinate identity. Perhaps there is no (specific) fact of the matter as to what the frog
believes. So as long as the frog catches and swallows whatever visible and digestible
object comes towards it, it makes no sense to say that the frog ‘misbelieves’ or
‘misperceives’ or ‘misrepresents’ something that it believes is kind F (say, a fly) but, as it
happens, is kind P (say, a pellet). What explains the frog's catching and swallowing a
pellet may well be the fact that the frog has the information that it does (whatever that
is). But the information fails to constitute real‐honest‐to‐goodness intentional content
because it is non‐aspectual or intractably indeterminate. Given that the frog's information
lacks intentional content, the frog is not in an incontestable or paradigmatic (intentional)
belief state. (More on whether such content even deserves to be called ‘information’
below.) Or more generally: The frog's ‘information’ is not intentional (or at least not
incontestably intentional); not unless, as we shall argue, it is conscious. Not unless, to
anticipate some remarks to follow, consciousness is essential to determinate content and
thus to uncontroversial intentionality.
What is it about a thinking creature that individuates or fixes or settles the question of
what it is thinking about? What determines intentional content? We assume,
hypothetically, that there are three and only three theoretically possible ways of
answering this question. First, something physical might individuate content—either
something physical about a thinker considered in environmental isolation, or something
physical about a thinker and its environmental context or environmental history. Second,
something phenomenal might individuate the content of a thought. Descartes, for
example, notoriously bemoaned being unable to tell which of his perceptual beliefs is
veridical, but he claimed to be certain of what he believed (p. 529) (i.e. of the intentional
content of his beliefs). This is because Descartes conceived of intentionality and
consciousness as forming a single package, presupposing that a belief's intentional
content is part of its cognitive phenomenology. Or third, one could rest with (what we
might call) a nihilist dissolution of the problem of content individuation: nothing
individuates or determines the intentional content of thought. There is no such content.
Thought, strictly speaking, is thoughtless (i.e. contentless). (Of course this may also mean
Page 18 of 30
Phenomenology, Intentionality, and the Unity of the Mind
that there are no such things as thoughts, or events such as thinking, which is a prospect
we shall not explore here, since we are not attracted to nihilism.)
In short, we are faced with the following dialectical situation: Assuming that intentional
states possess intentional content, and that intentional content is sufficiently determinate
to distinguish representing from misrepresenting and to satisfy the constraint of
intensionality in its description, then content is individuated either physically or
phenomenally. So how might physical individuation work? Physical individuation would
consist in physical patterns or relationships that determine the identity of the intentional
object and content of thought. These might be the causal physical antecedents of the
thought, systematic correlations between environmental activities (stimuli) and
responses, configurations of brain states and environmental affairs, adaptive functions of
neural states in ecological niches, features of an internal neurocomputational mechanism,
millions of interactions among subatomic particles, activations of receptor cells,
neurochemical or cell‐biological transfers, or whatever. A schematic attempt at physical
individuation was at work in our brief discussion of the example of the frog and the fly. As
suggested, however, by that discussion, physical individuation threatens to fail unless
some principled means can be found to determine the intentional object qua intentional
content of the frog's belief.
We will not try to discuss candidate principles for physical‐content individuation here. In
contemporary philosophy an active cottage industry has been trying to produce such
principles for years; some philosophers believe that this work has been successful,
others, like ourselves, are sceptical. (For some background discussion see Adams 2003.)
Here, however, we will suppose that what Daniel Dennett (1987) says about all such
attempts at physical individuation—which is a direct extension of Quine's thesis (1960) of
the inscrutability of reference—is true. This is that when these physical principles, no
matter what they are, ‘fall short of perfection, as they always must, there will be
uninterpretable gaps … so that no further [physical] fact could settle what the [thinker] in
question really [thinks]’ (Dennett 1987: 40; see also Dennett 1987: 106–16, 312). No
physical fact about the frog, including its causal relation to the environment or its
learning history, can individuate what the frog believes. What satisfactory grounds can we
have for saying that the thinker is thinking about this? No definite answer, says Dennett,
can be extracted from the physical manner of intentional‐content individuation.
Dennett's Quinean nihilism about physical individuation is not accepted by those who
hold one of the currently popular versions of ‘naturalistic’ theories of intentional content.
These include some popular extant ‘outside the head’ versions (p. 530) (such as those of
Dretske and Tye) of representationalism about phenomenal content which hold that
conscious content is intentional and is itself ‘outside the head’. All sorts of worries have
been raised for such theories, some for particular versions, and others for the whole
externalist brand of theory. These worries include: (i) worries about names (as in beliefs
about George Washington) and demonstratives (as in beliefs about this or that); (ii)
worries about uninstantiated properties or individuals (as in beliefs about unicorns and
Superman); (iii) worries about disjunctive content (as in whether the frog's content locks
Page 19 of 30
Phenomenology, Intentionality, and the Unity of the Mind
to fly or pellet); (iv) worries about over‐endowment (too much conceptual sophistication);
(v) worries about proximal projections (why a frog's thought might mean its distal cause
[fly] rather than a more proximal cause such as a retinal projection of a fly); (vi) worries
about instantiations of content in organisms without environmental histories (so‐called
swampman worries) or without normally embodied embedment (brains in vats) or normal
waking states (persons in protracted dreaming and hallucinatory conditions)—to name
some of the many sorts of worries that have been raised about outside‐the‐head or causal/
historical/environmentally embedded accounts of content individuation. (See Adams 2003
for a helpful survey from the perspective of someone sympathetic to ‘outside the head’
theories of intentional content. For criticism specifically of Dennett see Keeley 1999.)
Although we certainly cannot defend the assumption here, as noted above we share
Dennett's nihilism about the prospects for the physical individuation of intentional
content. So, too, does John Searle (1992: 163–4). However, whereas Dennett vigorously
denies that intentional content can be phenomenally individuated—for Dennett is a
nihilist about intentional‐content individuation period (whether physical or phenomenal)
—Searle advocates a phenomenal mode of individuation. We do, too, although not in
Searle's manner.
To illustrate: Suppose that you are travelling in New Hampshire and you consciously
wonder whether rabbits in New Hampshire have long tails. If you contrast consciously
wondering whether New Hampshire rabbits have long tails with consciously believing
that New Hampshire rabbits have long tails, you will notice, we claim, that there is
something common phenomenologically that remains the same in consciousness as you
pass from, say, wondering whether New Hampshire rabbits have long tails to believing
that New Hampshire rabbits have long tails. It is the distinctive phenomenal content New
Hampshire rabbits have long tails. It seems patently false—and false for reasons having
to do with the like‐thisness or just‐tellness of conscious content—that there is
indeterminacy as to whether you are thinking that New Hampshire rabbits have long tails
or thinking that (p. 531) New Hampshire squirrels have long tails or thinking that
Alabama rabbits have long tails or thinking that New Hampshire collections of
undetached rabbit parts have long‐tail subsets. It is false because the immediately self‐
presenting phenomenal intentional content New Hampshire rabbits have long tails is just‐
tell (subjectively) distinguishable from those other contents. (See Horgan and Tienson
2002: 523 for additional discussion, as well as Horgan et al. 2006.)11
Of course, if we are wrong, and Dennett is right that the identity of intentional content
cannot be settled by any means, then either (a) intentional content is not determinate or
aspectual but also determinacy/aspectuality is not necessary for intentionality (and
descriptions of intentionality fail to possess the property of failure of substitutivity or
Page 20 of 30
Phenomenology, Intentionality, and the Unity of the Mind
Assuming, contrary to Dennett, that some facts really do settle the matter of questions
about the identity of intentional content and that there is one right answer as to whether
I am thinking this, that, or some other thing entirely, but agreeing with his Quinean
misgivings about the physical determination of content, then all that's left as a mode of
individuation is phenomenological individuation. Subjects can ‘just tell’ what they are
thinking (see also Horgan and Graham, forthcoming). We hasten to add that for us this
does not mean that nothing physical or physically outside the head plays any role
whatsoever in some individuating features of some intentional objects and contents.
Important lessons about the relevance of things ‘outside the head’ to some questions of
content have emerged from the work of ‘outside the head’ theorists. One of these lessons
concerns so‐called wide truth conditions of thought contents that incorporate the actual
physical‐worldly referents (if any) of the relevant thought contents. An adequate overall
philosophical position about intentionality should accommodate various lessons of the
externalists, we claim, though without abandoning the thesis of phenomenal
intentionality. Just how this can be done is a complex exercise which we will not even
attempt to outline here (see Horgan et al. 2004 for detailed discussion).
Things are getting rapidly complicated, so let's bring the themes of the last several pages
to a close. We have been trying to defend Thesis PI. Let's briefly sum up where we are in
defending Thesis PI and moderate inseparatism.
If some states of mind are intentional, incontestably so, there must be something that
distinguishes their contents from each other. What counts as the same or different
content can depend in theory upon something physical or upon something phenomenal.
Physical individuation does not work (this is the Quine‐inspired Dennett nihilism). When
intentional states, however, are consciously intentional, they (p. 532) are immediately
self‐presenting in such a way that subjects can ‘just tell’ what their contents are. Their
contents are therein determinate or distinguished one from the other. Then and only then,
we claim, are they incontestably or uncontroversially intentional. So there is no
incontestable intentionality without phenomenology. Perhaps this is what Galen Strawson
means when he says: ‘I think that mere behavioral intentionality can never amount to
true intentionality’ (1994: 208). He continues: ‘There is a clear and fundamental sense in
which … intentionality exists only in the conscious moment’ (1994: 209).
Page 21 of 30
Phenomenology, Intentionality, and the Unity of the Mind
First set of questions: What about unconscious mentality including tacit and implicit
intentional states? Doesn't excluding them from the domain of the incontestably
intentional impoverish our ability to predict and explain behaviour?
Our response? For starters, we do not deny that it is useful for various purposes to
ascribe unconscious intentional states to a creature (including oneself). Instead, we deny
that such attributions are of genuinely uncontroversial or paradigmatic intentional states.
As long as we are careful to distinguish uncontroversial belief from useful belief‐
ascriptions, we may readily admit that we might have good grounds for employing
ascriptions of non‐conscious beliefs even though it is contestable whether the states
ascribed really are beliefs. Likewise, mutatis mutandis, for ascriptions of unconscious
intentional states other than those of belief, such as unconscious desires, memories, and
so on.
(p. 533)
Page 22 of 30
Phenomenology, Intentionality, and the Unity of the Mind
(and shouldn't) need to credit the fly with unconscious beliefs, since similar dispositions
to behaviour might be structurally implemented in non‐conscious information states
which are indeterminate as between being locked onto flies, pellets, or whatever and
therein fail to incontestably qualify as beliefs. Or, more precisely, since some theorists are
apt to think that information, by its very nature, is determinate, and thus that the idea of
‘indeterminate information’ is an oxymoron, let's refer to the relevant non‐conscious
states as quasi‐information states that are indeterminate between being locked onto flies,
pellets, or whatever. What explains the frog's behaviour may well be the fact that the
frog's brain state carries the indeterminate quasi‐information that it does, ‘information’
that never rises to the level of conscious attitude but functions similarly to a conscious
belief. (It may move, as we have put it elsewhere (Graham and Horgan 2002), along the
same causal grain as a conscious experience.) If so, similarly indeterminate ‘information’
might underlie the behavioural dispositions whose observation gives us good reason to
attribute unconscious beliefs to creatures such as ourselves. After all, as Quine and
Dennett would claim, the behaviour of the frog does not exhibit behavioural sensitivity to
the difference between flies and pellets. So the same internal ‘information’ state can be
assumed to explain the animal's responses to either sort of object. Meanwhile just as the
frog has ‘information’ states which are explanatory of its behaviour, so, too, we may
assume does the ‘unconsciously believing’ person. A person has no paradigmatic beliefs
unless these are conscious beliefs, so says inseparatism of the moderate sort, but a
person might have non‐conscious ‘information’‐processing states that undergird similar
dispositions as those presumably unconscious beliefs that are attributed to them. They
might behave as if they had beliefs without these being beliefs or intentional states of an
incontestable sort.12
Much more may need to be said in response to the first set of questions, but let's turn to
the second.
(p. 534)
Second set of questions: Ask separatists why they prefer separatism, and what are they
likely to say? Why favour separatism? We suspect that many would offer, among primary
reasons, the following.
Treating the mental as parsed or divided into the intentional and phenomenological
provides a methodological basis for conceptual and empirical advances in the physical
science of mind (see Wilson 2003). This is because (so the reasoning goes) while there
aren't principled obstacles to the explanatory analysis of intentionality in physical‐science
terms, there are principled obstacles to the explanatory analysis of consciousness in such
terms. David Chalmers (1996: 203), for instance, calls the problems associated with the
physical‐scientific analysis of the intentional ‘easy problems’ in the sense that we can
picture how neurobiological science might offer the analysis. Understanding
consciousness, he says, as a physical phenomenon is ‘the hard problem’ in the sense that
we are unable to picture how brain or physical science might propose the analysis.
Page 23 of 30
Phenomenology, Intentionality, and the Unity of the Mind
While there has been much optimism about the prospects of success in accounting
for intentionality [in broadly physical terms], pessimism about explaining
consciousness has deepened progressively. We can, it is felt, explain what makes a
mental state have the [intentional] content it has; at least there is no huge barrier
of principle in the way of our doing so. But, it is commonly conceded, we have no
remotely plausible account of what makes a mental state have the
phenomenological character it has; we do not even know where to start.
(1988/1997: 295)
We agree with Chalmers that what he calls the hard problem of consiousness is indeed a very
difficult problem. Standard physical accounts of consciousness do not close the ‘explanatory
gap’ (as it has been named by Joseph Levine 1983, 2001) between conscious states and whatever
brain states supposedly subserve them; rather, physical analyses neglect a defining feature of
the conscious content of mental states. This feature we have called their self‐presentational
immediacy or like‐thisness. We believe, too, however, as card‐carrying inseparatists, that the
scope of the hard problem is much broader or more extensive than it is often said to be, for the
expanded or whole hard problem incorporates the fact that the intentional, in a sense, is
pervasively conscious. The hard problem is not limited to conscious states, in some narrowly
parsed, intentionality‐excluding way. The explanatory gap ramifies throughout the realm of the
phenomenally intentional.
What about the physical science of mind? Doesn't inseparatism make that impossible?
Much depends upon what is meant by the physical science of mind. An important task for
a physical science of the mental, where the mental is assumed to possess moderately
inseparable features, would be to provide a way of mapping phenomenal intentionality on
to the types of entities or processes studied in neuroscience; that is to say, of correlating
types of phenomenal intentional states including their contents with types of brain states
and neural activities. This might be called the search for the neural correlates of
phenomenal intentionality (NCPI). Arguably, though, the discovery of modes of physical
correlation or NCPI will not tell (p. 535) us why certain brain states are about that or
possess intentional content. If Dennett is warranted in his Quinean nihilism about
physicalist individuation, as we assume that he is, we aren't going to get that question
answered in physical‐scientific terms. Hopes for a cell‐biological alphabet of intentional
content would appear stymied by the union of uncontroversial phenomenology and
intentionality in our mental lives. The unity of mind, although moderate in its inseparable
parts, may humble certain physicalistically immodest aspirations of its physical science.
References
Adams, F. (2003), ‘Thoughts and Their Contents: Naturalized Semantics’, in S. Stich and
T. Warfield (eds.), The Blackwell Guide to the Philosophy of Mind (Malden, Mass.:
Blackwell), 143–71.
Page 24 of 30
Phenomenology, Intentionality, and the Unity of the Mind
Church, J. (2003), ‘Depression, Depth, and the Imagination’, in J. Phillips and J. Morley
(eds.), Imagination and Its Pathologies (Cambridge, Mass.: MIT Press), 175–86.
—— (2007), ‘What Robo Mary Knows’, in T. Alter and S. Walter (eds.), Phenomenal
Concepts and Phenomenal Knowledge: New Essays on Consciousness and Physicalism
(Oxford: Oxford University Press), 15–31.
—— (2006), The Primacy of the Subjective: Foundations for a Unified Theory of Mind and
Language (Cambridge, Mass.: MIT Press).
Graham, G., and Horgan, T. (2000), ‘Mary Mary, Quite Contrary’, Philosophical Studies,
99: 59–87.
Page 25 of 30
Phenomenology, Intentionality, and the Unity of the Mind
Horgan, T., and Tienson, J. (2002), ‘The Intentionality of Phenomenology and the
Phenomenology of Intentionality’, in D. Chalmers (ed.), Philosophy of Mind: Classical and
Contemporary Readings (Oxford: Oxford University Press), 520–33.
Horgan, T., Tienson, J., and Graham, G. (2003), ‘The Phenomenology of First Person
Agency’, in S. Walter and H.‐D. Heckmann (eds.), Physicalism and Mental Causation: The
Metaphysics of Mind and Action (Exeter: Imprint Academic), 323–40.
—— (2004), ‘Phenomenal Intentionality and the Brain in a Vat’, in R. Schantz (ed.), The
Externalist Challenge: New Studies in Cognition and Intentionality (Berlin, New York: de
Gruyter), 297–318.
Kim, J. (1998), Mind in a Physical World: An Essay on the Mind–Body Problem and Mental
Causation (Cambridge, Mass.: MIT Press).
Lettvin, J., et al. (1959/2000), ‘What the Frog's Eye Tells the Frog's Brain’, Proceedings of
the Institute of Radio Engineers, 47: 1940–51;
repr. in R. Cummins and D. Cummins (eds.), Minds, Brains, and Computers: The
Foundations of Cognitive Science (Malden, Mass.: Blackwell, 2000), 82–96.
Levine, J. (1983), ‘Materialism and Qualia: The Explanatory Gap’, Pacific Philosophical
Quarterly, 64: 354–61.
—— (2001), Purple Haze: The Puzzle of Conscious Experience (Oxford: Oxford University
Press).
Page 26 of 30
Phenomenology, Intentionality, and the Unity of the Mind
Raffman, D. (2005), ‘Even Zombies Can Be Surprised’, Philosophical Studies, 122: 189–
202.
(p. 537) Strawson, G. (1994), Mental Reality (Cambridge, Mass.: MIT Press).
Notes:
Page 27 of 30
Phenomenology, Intentionality, and the Unity of the Mind
(5) Among philosophers who have recently been working on the topics of consciousness
and intentionality with, in occasionally quite particular respects, inseparatist sensitivities
are Barry Dainton, Owen Flanagan, Nicholas Georgalis, Alvin Goldman, Uriah Kriegel,
Brian Loar, Colin McGinn, David Pitt, and Galen Strawson, as well as Siewert and
Chalmers (among others). Some of this work is cited or discussed in the rest of the
chapter as well as in the papers mentioned in n. 6.
(6) The general position defended here was first introduced in general terms in Horgan
and Tienson (2002), extended in Horgan et al. (2003), and refined in Horgan et al. (2004)
and (2006). It was distinguished from representationalism in Graham and Horgan (2000).
It has elicited criticism in one or another respect from Georgalis (2003), Wilson (2003),
Raffman (2005), and Dennett (2007), with a reply to Raffman in Graham and Horgan
(2005) and a reply to Dennett in Graham and Horgan (2008).
(7) Compare with Georgalis (2006), a friend of the first‐personal or subjective but not of
the language of the broadly phenomenal or qualitative.
(8) What's said in the text is also intended to leave open the possibility of being mistaken
in applying certain linguistic labels to one's experiences, by virtue of lacking a full grasp
of the meanings of certain words in a natural language.
(9) Thus, nothing essential will turn on the fact that we will describe the kind of state we
are focusing on—a non‐conscious, putatively intentional state—as a putative belief. We
will use belief talk mainly for vividness. The real issue concerns the purported intentional
content of such a non‐conscious state.
Page 28 of 30
Phenomenology, Intentionality, and the Unity of the Mind
(10) Again, it does not really matter whether or not one calls this internal state a belief.
See the preceding note.
(11) Conceptually infused phenomenal content of the sort involved in thinking that New
Hampshire rabbits have long tails shares in whatever vagueness attaches to its
associated concepts. But vagueness as such is one thing. Indeterminacy of intentional‐
content identity is another. Just what counts as a rabbit may be debated (a vagueness
issue) without debating whether I can just tell that I am thinking of rabbits (an issue of
determinate content).
George Graham
Terence Horgan
John Tienson
Page 29 of 30
The Self
This article aims to give some sense of the conceptual geography of the notion of the self,
and the article can do this without requiring the reader to accept any of its assumptions
or conclusions. It talks about historical figures, but only because they provide by far the
best, most direct, most vivid, and most searching discussions of the problem of the self.
Keywords: notion of the self, historical figures, problem of the self, subject of experience, semantic intention,
existence of self
31.1 Introduction
‘I know that I exist; the question is, what is this “I” that I know?’ (Descartes 1641/1984:
18). I am, I answer, a human being, a product of evolution by natural selection, a living
organism, a physical object—a wholly physical object. I will take all this for granted.
I should admit, though, that I don't fully know the nature of the physical. No one does.
Nearly all of us take it that the physical is essentially spatio‐temporal, for example, but no
one expert in these matters claims to know for certain what space and time are, or
whether they are really fundamental features of reality as we standardly conceive them.
Many are sure that they are not, and speak instead of ‘space–time’, but the same doubts
arise about the unified space–time of relativity theory, and even if our best theories are
right about the nature of space–time, the overall nature of the physical remains in many
respects profoundly obscure to us, especially given that consciousness itself is something
physical. So even if I know—take it for granted—that I am a wholly physical thing, much
remains unknown.1
Page 1 of 28
The Self
(p. 542)
I am a subject of experience, then. Am I also a self? Only if selves exist. Do they exist? It
depends what one means by the word ‘self’. In what follows I will take the word ‘self’ in a
certain specific way, make a number of assumptions, and answer Yes to the question:
selves do indeed exist.
Many, however, will think my Yes amounts to a No, because they will not be prepared to
take the word ‘self’ in the way I do. And perhaps I would do better to answer No. In many
ways my position is closer to the No‐sayers than the Yes‐sayers, and in saying Yes I am
likely to lose the sympathy of both sides, each objecting that I concede too much to the
other. The pro‐self party will object that the selves I claim to exist don't really deserve the
name, and that by using ‘self’ in this way I am obscuring the fact that there are other
things that do deserve the name. The anti‐self party will agree with the pro‐self party that
the selves I claim to exist don't deserve the name, and object that by using ‘self’ in this
way I am obscuring the fact that nothing deserves the name. This will, I hope, be helpful.
My present aim is to give some sense of the conceptual geography of the notion of the
self, and I can do this without requiring the reader to accept any of my assumptions or
conclusions. I will talk about historical figures, but only because they provide by far the
best, most direct, most vivid, and most searching discussions of the problem of the self.
Some analytic philosophers think that this conclusion is inevitable, if one is going to talk
of selves at all (see, paradigmatically, Kenny 1988, 1999). They infer that it is much better
not to. It is enough, they say, to talk of human beings. Talk of selves is superfluous and
intensely misleading.
Page 2 of 28
The Self
I disagree. I agree that I am a human being, and I think that I am a self, but I don't think
that the self that I am is the human being that I am. I don't reject the logic of the identity
relation. That would be silly. I reject the assumption that the word ‘I’ is univocal—the
assumption that it has only one possible meaning or reference in thought or speech. The
reference of ‘I’—both its actual reference and its explicitly or implicitly intended or
presupposed reference—standardly shifts between two different things in my thought and
speech and in the thought and speech of others. Sometimes ‘I’ is used to refer to a human
being considered as a whole, sometimes it is used (p. 543) to refer to a self. These are two
things that have quite different identity‐conditions, if indeed selves exist, and that stand—
I will argue—in some sort of straightforward part–whole relation.3
To say this is not to assume that selves exist. It is only to report a fact about how the word
‘I’ is used. If it turns out that the best thing to say about selves is that there are no such
things, then the best thing to say about ‘I’ may well be that it is univocal after all, and
that the apparent doubleness of reference of ‘I’ is just the echo in language of a
metaphysical illusion. If this is right, then ‘I’ is not in fact used to refer to selves as
distinct from human beings even when its users intend to be making some such reference
and believe that they are doing so.4 On this view, the semantic intentions of ‘I’‐users
sometimes incorporate a mistake about how things are.
I disagree. I think that we do at different times successfully use ‘I’ to refer to different
things: to human beings considered as a whole and to selves. In this respect the word ‘I’
is like the word ‘castle’. Sometimes ‘castle’ is used to refer to the castle proper,
sometimes it used to refer to the ensemble of the castle and the grounds and associated
buildings located within the perimeter wall, sometimes it can be taken either way. The
same goes for ‘I’, but ‘I’ is perhaps even more flexible, for it can sometimes be taken to
refer both to the self and to the whole human being, indifferently.5 Our thought (our
semantic intention) is often unspecific as between the two; it's all pretty relaxed.6
My claim, then, is that there is a fully correct use of ‘I’ in which it is true to say that I am
not a human being, i.e. a human being considered as a whole.7
That's fine, you may say, so far as the word ‘I’ is concerned. But wouldn't it be better,
even so, just to stop using the word ‘self’ altogether, in philosophy? It really doesn't do
any good, and it causes a great deal of confusion. Shouldn't we try to get everyone to stop
using it? The notion of a self is so hopelessly unclear (see e.g. Olson 1998; Kenny 1999).
The notion of a self is extremely unclear, but this proposal is unrealistic. People will
always talk of the self, both in and out of theoretical contexts, and there are good reasons
why this is so. The idea of the self is very compelling. James Trefil speaks for common
sense when he writes ‘No matter how my brain works … one (p. 544) single fact remains
… I am aware of a self that looks out at the world from somewhere inside my skull … this
is … the central datum with which every theory of consciousness has to grapple’ (1997:
181).8 More importantly, I believe we get something right in thinking and talking in terms
of the self, although I also think we get something wrong. I think that there really are
things that are subjects of experience and that are not the same things as whole human
Page 3 of 28
The Self
beings and that are good candidates for being called ‘selves’. It's true that the use of the
definite article ‘the’ in the phrase ‘the self’ is very misleading in some contexts, and
directly question‐begging in others, and for this reason I will often speak of selves, or a
self, rather than of the self. But the self will often be the most convenient expression, and
when it is I will use it freely.
(1) not the same thing as a whole human being—a human being considered as a
whole.
Otherwise there would be no point in introducing the term ‘self’. I also take it that anything
properly called a self must be
(2) a subject of experience.
I take this to be true by definition, and no one, I think, will wish to disagree.
(2) does, however, raise a problem. For we standardly say that human beings considered
as a whole are subjects of experience, and it follows from (1) that a subject of experience
that is a self cannot be the same thing as a subject of experience that is a whole human
being. So what can it be?
I assume that the inner subject of experience is wholly located in the brain, but this is not
essential, and ‘inner’ can also be taken more loosely to cover the possibility that the existence of
the subject essentially involves the existence of a non‐physical or ‘immaterial’ soul.9
(3) implies (2) and (1), but now we face a question about the relation between the self
that one is—for one must surely count as a self if there are such things as selves at all—
and the human being that one is.
(p. 545)
Page 4 of 28
The Self
My next condition is that any candidate for being a self must count as a single thing or
object or substance of some sort. Or, rather, since the notions of object and substance are
surprisingly unclear,10 and may be thought metaphysically dubious, it is that a self must
be
(4) at least as good a candidate for being a substance or object as anything else.
The idea that the self is a thing or substance of some sort is wholly conventional in the long
recorded history of discussion of the self, from the early Hindu texts onwards, and it is obviously
accepted by those who think that selves are immaterial soul‐substances, as these are
traditionally conceived, although it equally obviously does not require any belief in any such
entities. Condition (4), then, ensures that we are addressing the traditional metaphysical issue
directly, and it will be valuable in structuring the discussion even if it comes under considerable
pressure. For my own part I think there is a sense in which it must be correct, given only (2): the
fact that a self must be a subject of experience.11
Condition (4) imposes a clear and highly substantive constraint on any attempt to answer
the question whether selves exist. It rules out, for one thing, common present‐day uses of
‘self’ to refer merely to the personality or character of a human being, where the
personality of a human being is seen as just a set of properties of that human being.12 It
also brackets whatever it is that people have in mind when they talk of the ‘social’ self, or
the ‘ecological’ self, or the ‘narrative’ self (see e.g. Mead 1913; Gibson 1993; Phillips
2003), for these phrases are typically used to denote modes or styles or aspects of human
behaviour and experience, properties or characteristics of human beings that can be fully
characterized without supposing that there is such a thing as a self distinct from a human
being considered as a whole. The present concern (at least so far) is with the good old‐
fashioned metaphysical question ‘Do selves (p. 546) exist?’ and this question is taken to
be equivalent to the question ‘Are there objects or substances that are correctly called
“selves” ’?
I don't, however, have a good old‐fashioned view of what an object or substance is, so far
as most of the western philosophical tradition is concerned. My sympathies lie rather
with the philosophers like Nagarjuna, Nietzsche, Whitehead, and others. I think, for
example, that the hugely natural distinctions between objects and processes, on the one
hand, and between objects and their propertiedness, on the other hand, are hopelessly
superficial when one gets metaphysically serious (see Section 31.6 below). I also take it
that Berkeley's description of the self as a ‘thinking active principle’ passes the ‘object’
test comfortably, and I see no reason to think it will exclude Fichte's conception of the
self as a Tathandlung, a ‘deed‐activity’ (Berkeley 1713/1975: 116; Fichte 1794–1802/1982:
96).13
Page 5 of 28
The Self
31.4 Assumptions
To raise doubts about the standard understanding of what objects and substances are is
not to say that they don't exist, and I am going to assume for argument that they do, that
i.e. that we can acceptably talk in terms of a plurality of objects or substances, when we do
metaphysics, and also that
(ii) a human being is an object or substance if anything is.
physicalism (materialism) being, unsurprisingly, the view that every real, concrete phenomenon
in the universe is … physical.14
I don't, however, mean what most philosophers today mean by ‘physical’ or ‘physicalism’.
I mean real physicalism, serious physicalism, i.e. physicalism that is fully realist about
conscious experience, i.e. fully realist about the existence of the subjective qualitative
character or ‘what‐it's‐likeness’ of experience, where this is conceived of in the most
ordinary everyday way. Such conscious experience (from now on I'll simply call it
‘experience’) is in fact the only thing we know for certain to be physical, on the
assumption that physicalism is true, because it is the only concrete phenomenon we know
for certain to exist.
(p. 547)
Experience, then, is the fundamental given natural fact. It is what any remotely serious
version of physicalism or naturalism must start from. (Here I make the common
assumption that physicalism and naturalism are the same thing.) Some self‐styled
physicalists and naturalists use ‘physicalism’ and ‘naturalism’ to denote a view that
involves the denial of the existence of the most certainly existing natural thing there is—
experience, or what‐it's‐likeness—but they are unreal physicalists, unnatural naturalists,
defenders of the silliest view ever espoused in the history of humanity.15 Real physicalism
or naturalism, by contrast, is defined by the fact that it fully acknowledges the existence
of the most certainly existing thing there is, experience, and takes it—considered as such
—to be a wholly physical phenomenon. Real physicalism can't treat ‘physical’ and
‘mental’ as terms that are in any way opposed (it must rather distinguish between mental
and non‐mental aspects of the physical), because it holds that everything mental,
including everything experiential, is wholly physical. Nor can it have anything to do with
physicsalism, the view that the fundamental nature of everything concrete can be
accounted for in the terms of physics, for if physicalism is true then nothing is more
certain than that there is more to the physical than what is or can be described or
Page 6 of 28
The Self
accounted for by the discipline of physics, which has no terms for the phenomena of
experience considered just as such.
(iv) human beings are physical objects, wholly physical objects or substances
(5) something that standardly persists for life, or at least over a reasonably long
period of time,
and
(7) something that has a certain character or personality and is not a mere locus of
consciousness and cognition
I think, in fact, that all of (5)–(8) can be cogently challenged, when one tries to work out the
absolutely minimal case in which one may reasonably speak in terms of a self—an ‘inner’ subject
of experience or locus of consciousness, an inner someone, an inner mental presence.17 In this
chapter however, I am only going to challenge (5).
31.6 Objects
— You say you want to show that there is such a thing as the self, and that you use
the word ‘thing’ to mean object or substance, but you've been fooled by the word
‘self’—by the fact that it is grammatically substantival, a noun. There is no such
thing as ‘the self’.
Page 7 of 28
The Self
The objection is ancient, but it misunderstands my aim, which is precisely to defend the self's
claim to be an object or substance of some sort, if only as a way of articulating the general
structure of the notion of the self.
Suppose I can't show that there are things that are worthy candidates for the title ‘self’
on these terms. Suppose I can't even show that a view that takes the self to be a thing or
substance of some sort is better than a view according to which it is best seen as a
property of some other thing like a brain or a human being. Suppose I find it ‘quite
impossible’, in Kant's words,
to determine the manner in which I [i.e. the self or soul] exist, whether it be as
substance [or thing, or object] or as accident [or property].
(1781–7/1933: B420)18
In that case I may have to conclude that the self is guilty as charged: non‐existent. I may have to
conclude that all the undoubtedly real self‐phenomena—all the phenomena that lead us to talk in
terms of the self, whether or not there is such a thing—just do not provide adequate grounds for
the claim that selves actually exist. For the moment, though, my aim is to argue that selves exist,
and that they are objects or substances of some sort, and hence, given physicalism, physical
objects.
This last suggestion strikes many as absurd. The reaction stems, I believe, from a failure
to think through what it is to be physical, on any realistic physicalist view, and, equally,
from a failure to think through what it is to be a thing or object. It is (p. 549) as if one has
to solve for three inadequately conceived quantities—self, object, physical—
simultaneously, using each to get leverage on the others.
This is a large task, but the claim that selves are physical objects may start to look a little
less strange once one begins to acquire a proper understanding of what it is to say that
something is an object, or physical. One simple point, already touched on, is that even the
most cursory consideration of what we ordinarily think of as physical objects, from stones
to brains, shows them to be fabulously unlike our everyday conception of them. The most
basic lay acquaintance with science shows physical objects to be inconceivably
insubstantial by common‐sense standards. They are patterns of energy, diaphanous
process‐entities, entities whose being involves a constant interchange with the quantum
vacuum given which it is literally correct to say that they are partly constituted by the
vacuum. The first thing to do, I think, in order to see things more accurately, is to think of
physical objects as, literally, processes—where this requires, at the least, abandoning the
standard metaphysic according to which a process requires an object or substance
distinct from itself in which to occur. If this is to commit a ‘category mistake’, don't blame
me. Blame the collision of ordinary thought and language with the world.
The same collision occurs in the case of the routine but ultimately wholly untenable
ontological distinction between an object, considered at any given time, and its properties
or qualitative characteristics, at that time, where by its properties, its propertiedness, I
mean whatever it is about it concretely that makes it true to say that it is the particular
way it is. One can't, when one gets metaphysically serious, suppose that one has, at any
Page 8 of 28
The Self
given time, on the one concrete ontological hand, the object, and, on the other concrete
ontological hand, its propertiedness, the totality of its qualities or accidents, in any way
that allows that they can be said to be ultimately ontologically distinct, in concrete reality.
When it comes to the ancient problem of particular and universal, the trick is not to say
too much: to say no more than Kant when he remarks that ‘in their relation to substance,
accidents [or properties] are not really subordinated to it, but are the mode of existing of
the substance itself’ (1781–7/1933: A414/B441).19 There is no ontological subordination
of the total qualitative being of the object to the object considered just as such, no
ontological subordination of its nature to its existence. The whole object/process/
property/state/event cluster of distinctions, unexceptionable in everyday life, is fatally
superficial from the point of view of science and metaphysics. Our most fundamental
thought categories simply do not get the world right, and when we think hard we can see
a priori that this is so.20
So far, perhaps, so good. But if, now, you ask me how long I think human selves exist, I
will answer ‘No more than a second, if that’—flatly rejecting (5), the view that anything
that can count as a self must be something that persists for life or at least over a
reasonably long period of time.
It seems plain that if a being qualitatively identical to an ordinary adult human being
sprang suddenly into existence, enjoyed two seconds of experience just like two ordinary
seconds of your or my experience, and then ceased to exist, a self would most certainly
exist in that two‐second period; if, that is, selves exist at all. But one could fully accept
this while rejecting as absurd the claim that selves (assumed now to exist) last only a
second or two—or less—in our case. Many, in fact, will take such a claim to be equivalent
to the claim that there are no such things as selves, because they take it to be true by
definition that selves persist for a long period of time, a lifetime for example, if they exist
at all. The inclination to think of the inner subject or self as a persisting thing lies deep in
most ordinary (non‐pathological) thought.
Page 9 of 28
The Self
And yet I do not think we do well to conceive the self as some sort of persisting—sleeping,
waking—inner inhabitant of a human being. If we must have something that endures in
this way, I think we do best to stick to the whole human being, as many analytic
philosophers have wished to do.
— Each normal human being contains some complex persisting brain structure Q
which is (obviously) not the same thing as the whole human being and which can be
roughly characterized as follows: Q is that which supports the consciousness and
personality of a human being. Imagine a human being losing limbs, body, reduced by
surgical miracle to a head or just a brain or even something less than a brain,
without any fundamental disruption of Q or of psychological life or character. In this
case a subject of experience plainly persists, and what this shows is that talk of a
persisting inner self doesn't involve any bad inner inhabitant's picture. It's just that
Q, the self‐organ, as it were, is wholly located in the brain. It's simply part of the
brain, a brain structure, as you say, and there can be no objection to thinking of it as
a persisting thing, like a hand or eye, although it is not a single part of the brain
physiologically speaking. One might call this the ‘self‐system view’ (see Flanagan
2003: ch. 6).
(p. 551) Granted. I certainly can't show that such a use of the word ‘self’ would be wrong, and
this is one simple way to give an unqualified Yes in answer to the question whether the self
exists. It is a view that is fully available to physicalists, and most will think it obvious that selves
exist, so understood, for they exist as surely as hearts do.21 Why do I hesitate? Because I doubt
that selves can satisfy condition (4) in an adequate fashion, when they are understood in this
way. I lack space to justify this view here, however (see Strawson 2009), and propose instead to
give some further substance to the idea that human selves are transient. I think this is a good
way to articulate thought about the notion of the self even if it is not in the end the best thing to
say about selves.
(a) the thick conception according to which it is human beings and other animals
considered as a whole that are subjects of experience.
Page 10 of 28
The Self
I take it that (a) and (b) both build in the assumption that a subject may and standardly does
continue to exist even when it is not having any experience: whether you think that human
subjects are whole human beings or whether you think they are inner loci of consciousness you
are likely to allow that they can continue to exist during periods of complete experiencelessness
—in periods of dreamless sleep, say.
It is this that creates the need for a third conception of the subject,
(c) the thin conception according to which a subject of experience, a true subject of
experience, does not and cannot exist without experience also existing, experience
which it is having itself.22
The thin conception stands opposed to both (a) and (b) precisely because they both contain the
natural assumption that a subject of experience can be said to exist in the absence of any
experience, but it doesn't by itself offer any support to the idea that selves are transient entities.
The claim that human selves or thin subjects are short‐lived, given (c), is an independent,
empirical claim.
(p. 552)
There is a problem of exposition here, because most are so accustomed to (a) and/or (b),
and to the idea that they exhaust the options, that they cannot take (c) seriously. And yet
(c) simply makes a place for a natural use of the term ‘subject’ according to which it is a
necessary truth, no less, that
there cannot be a subject of experience, at any given time, unless some experience
exists for it to be a subject of, at that time.
On this view, there can no more be a subject of experience without an experience than there can
be a dent without a surface. (c) requires that the subject be ‘live’, so to say, in order to exist at
all. I believe it is crucial to have (c) in play when taking on the metaphysics of the self.
Page 11 of 28
The Self
sort whose existence is taken for granted by almost all philosophers in his time.23 So far
as the empirical data are concerned, he observes, we find only transient selves: ‘our
thought is fluctuating, uncertain, fleeting’; it is ‘in a perpetual flux’, in which different
thoughts and experiences ‘succeed each other with an inconceivable rapidity’ (Hume
1779/1947: 194; 1739–40/2000: 165; 1739–40/1978: 252).24
Descartes also endorses the thin conception of the subject, as far as I understand him,
although his thin subjects, being possibly immortal,25 are far from transient. (p. 553) His
most fundamental claim is that the soul or self or subject does not and cannot exist at all
in the absence of conscious experiencing (‘thinking’, in his terminology). A subject
without conscious experiencing for it to be the subject of is as impossible, for Descartes,
as a physical object without extension, and his root—radical—idea about the nature of the
subject of experience or soul is that it is wholly and literally constituted of conscious
experiencing. That, it seems, is what a res cogitans—a soul—is. It is nothing like an
immaterial soul as traditionally conceived. Its being is, rather, a matter of occurrent
experiencing—so that it certainly can't exist when there isn't any. At one point Descartes
talks of ‘our soul or our thinking’ as if the two terms were strictly interchangeable
(Descartes 1644/1984: 184). At another he writes, seemingly unequivocally, that in being
the essential attribute of thinking substance ‘thinking … must be considered as nothing
else than thinking substance itself …, that is, as mind’ (Descartes 1644/1984: 215).
Page 12 of 28
The Self
These are good quotations, but they do not undermine the attribution of (P3) to Descartes, who
for reasons of caution makes a particular point of continuing to use accepted terminology even
though it no longer plays any active role in his view.26 It must be stressed again that (P3) cannot
be understood to incorporate any standard substance–property distinction, for while Descartes
grants, as just noted, that ‘we call the thing in which [mental attributes] inhere a substance’, he
does not think that the notion of substance has any meaning or intelligible reference or
explanatory force, in so far as it is supposed to be something that is in any way distinct from its
(‘its’) attributes or properties.27 Speaking of the fundamental attributes of thought and
extension, he says that ‘the distinction between these notions and the notion of substance itself
is a merely conceptual distinction’, and by this he means that although it is a distinction that can
indeed be made in thought it is not a ‘real’ distinction, where to say that there is a real
distinction between two things is simply to say that each can exist in reality without the other
existing, that each can exist ‘separately’, in complete ontological separation from the other
(Descartes 1644/1984: 215).28 In his conversations with Burman he confirms this view: ‘[T]he
attributes [of a substance], when considered collectively, are indeed identical with the
substance’ (1648/1976: 23). It is only in so far as we agree to employ the language of ‘inherence’
and ‘substance’, and agree to speak of thought and extension as ‘inhering in a substance’, that
we must—given the way in which the properties of thought and (p. 555) extension are on his
official theory essentially mutually repellent—say that there are two substances. The word
‘substance,’ however, is doing no work at all, and towards the end of his life Descartes the
indefatigable dissector of brains, in every other pore of his philosophy a materialist, admits in
effect that he does not know enough of matter to be sure of this repulsion.29
There is still a serious difficulty in the radical position as so far characterized: the
difficulty of finding a ground or ‘place of residence’ for mental faculties like will and
understanding and ‘intellectual memory’30 (not to mention innate ideas), given that the
thinker is, metaphysically, simply a process of conscious experiencing. Nor is this
difficulty diminished by Descartes's view that these faculties are ‘potentialities’;31 for
potentialities, it seems, also need a place of residence beyond what can be supplied by an
occurrent process of conscious experiencing if the whole being of such a process,
considered at any particular moment, is only the entertaining of whatever content is then
occurring. It seems plain that Descartes conceives of the process of conscious
experiencing as something that is somehow inherently active. This, however, is a matter
for another time. It raises wider issues not only in Descartes's wildly misrepresented
Page 13 of 28
The Self
philosophy but in metaphysics generally (see e.g. Strawson 1994: ch. 5; 2006: sect. 8).32
For now the point is that (c), the thin‐subject view, is not a particularly eccentric view.
Leibniz also accepts it, taking thin subjects to be long‐lived. Kant writes that ‘the thinking
or the existence of the thought and the existence of my own self are one and the
same’ (1772/1967: 75). Fichte and Nozick (following Fichte) can also be read as accepting
it, taking thin subjects to be short‐lived.33 Buddhists also accept short‐lived thin subjects
of experience, in spite of the fact that they are often supposed to deny the existence of
subjects of experience altogether. So also does William James, the greatest modern
philosopher of mind.
31.10 James
James takes over Descartes's standard use of the word ‘thought’ to cover all types of
conscious episodes,34 and explicitly appoints ‘Thought … with a capital T’ as a name
(p. 556) for ‘the present mental state’ (James 1890/1950: i. 186–7, 338). He then argues
that ‘the I’ or self ‘is a Thought’. ‘The thoughts themselves are the thinkers’, he says; each
‘ “perishing” … pulse of thought’ or ‘ “section” of consciousness’ is a (short‐lived) thinker
or self or subject of consciousness (James 1890/1950: i.400–1, 371, 337).35 As a
materialist, he takes it that these pulses are occurrences in the brain.
‘The thoughts themselves are the thinkers’ sounds hopelessly odd at first. One can recast
it as ‘the conscious episodes themselves are the subjects of consciousness’, for the best
translation of Descartes's ‘cogito, ergo sum’ is ‘I am conscious, therefore I am’, but this
hardly improves matters. Slightly less indigestible, perhaps, is: ‘the existence of each
conscious episode or experience consists in the existence of a subject of consciousness
entertaining a certain mental content’, this being, once again, a wholly physical event.
There is a way of reading this in which it doesn't obviously dilute James's claim, and it
may sound just a little less unpalatable to some: it may have a slightly better chance of
delivering the odd mental shock that comes from grasping (even if only briefly) truths
that have no non‐counter‐intuitive expression.
One can only go so far with such rephrasings, however. The idea remains very difficult on
first acquaintance, given the ordinary background of thought. It is I believe correct, given
that one is operating with (c), the thin conception of the subject, but it has to grow on
one.
I … is a Thought [present mental state], at each moment different from that of the
last moment, but appropriative of the latter, together with all that the latter called
its own. All the experiential facts find their place in this description,
Page 14 of 28
The Self
unencumbered with any hypothesis save that of the existence of passing thoughts
or states of mind.
There are many subjects, on this view, in the life of a human being like Louis. Each one, each
‘pulse of thought’, is an ‘indecomposable unity’ (James 1890/1950: i. 371) in every sense in
which it needs to be for distinct thought‐elements like grass, green, and wet to be able to be
genuinely bound together (in the sense of the ‘binding problem’) in whatever way they must be if
a thought‐unity like the thought the grass is green and wet is genuinely to occur, or if a single
pulse of experience is to be, simultaneously, experience of swimming, thinking about Vienna, and
hearing Australian magpies carolling.
What is the metaphysical status of these subjects relative to each other? James seems
quite clear on the point. They are numerically distinct—albeit short‐lived — substances. A
brain, no doubt, is a single continuing thing, but ‘the same brain may subserve many
conscious selves, either alternate or coexisting’, that have ‘no substantial identity’ (James
1890/1950: i. 401).
(p. 557)
At one point James puts the point by saying that the self or subject consists in ‘a
remembering and appropriating Thought incessantly renewed’ (1890/1950: i. 338, 362–
3),37 and some may take this as evidence that he thinks selves do have some sort of long‐
term continuity. However, his more careful statement of his view cancels any such
suggestion:
Successive thinkers, numerically distinct, but all aware of the past in the same
way, form an adequate vehicle for all the experience of personal unity and
sameness which we actually have.
(1892/1984: 181)
(1890/1950: i. 360)
Page 15 of 28
The Self
How does this fit with James's doctrine of the ‘stream of consciousness’, the apparent continuity
of conscious experience that occurs in the case of an individual human being like Louis, and
leads him (i.e. one short‐lived Louis subject or another) to think of himself, in so far as he is a
self, as a single, persisting—continuously existing—entity? The first thing to note is that James
rightly doesn't claim to know that the stream of consciousness is continuous.38 Second, he has a
ready explanation of the appearance of continuity, for on his view each subject ‘appropriates’—
makes its own, takes to itself, inherits, has access to—the content of the experiences of its
numerically and substantially distinct predecessors. The notion of appropriation is loose but it is
clear enough what James has in mind and it supplies all the apparent (phenomenological)
continuity and developmental coherence of experience.39 A further reason why the appearance
of continuity is unsurprising, on his view as on mine, is that the subjects (subjects‐or‐
experiences) in question arise successively, in a single person's brain, from brain conditions that
have considerable similarity from moment to moment even as they change.40 (p. 558)
— James doesn't really believe in numerically distinct short‐lived metaphysical
subjects. His ‘pulses of thought’ are really just convenient theoretical abstractions
from the fundamental given fact, a metaphysically single flow. At one point he writes
that ‘as the brain‐changes are continuous, so all these consciousnesses melt into
each other like dissolving views. Properly they are but one protracted consciousness,
one unbroken stream’ (1890/1950: i. 2478)41
This quotation doesn't I think conflict with James's endorsement of short‐lived, numerically
distinct subjects, if only because numerically distinct pulses of thought, riding on continuous
brain‐changes, can succeed each other in a seamlessly joined and indeed partly temporally
overlapping (‘coexisting’) way to form an unbroken stream. James has already denied that we
can know there to be an unbroken stream of consciousness, even if we assume there to be an
unbroken stream of brain‐changes, and this passage gives no reason to doubt that he takes it
that the subject‐constituting pulses of thought are genuinely metaphysically distinct entities
rather than theoretical abstractions from a single flow that is single in some way that goes
beyond the singleness or uninterruptedness of (possibly overlapping) seamlessness. He thinks
that there are many Is or selves or subjects, ‘indecomposable unities’ that are not mere
theoretical abstractions.42
These selves, these thoughts, are of course physical unities, on his view, and they have
the seemingly unique distinction of being physical unities that we can know for certain to
exist (given the assumption that physicalism is true) and to be true unities. For we know
(1) that particular thoughts, for example, occur: thoughts in the narrower sense,
particular events of proposition‐comprehension. And we know (2) that thoughts must
have distinguishable parts or elements, in order to be genuine thoughts. And we know (3)
that these distinguishable parts must be held or bound together in the unity that is the
comprehension of the proposition by a subject, if there is to be a genuine thought at all.
So we know (4) that there are actually existing physical entities that concretely realize a
certain sort of unsurpassable unity, a ‘logical’ unity, in Kant's terms, the ‘absolute …
logical unity of the subject’, the ‘logical unity of any thought’ (Kant 1781–7/1933: A356,
A398).43 In knowing, then, that thoughts exist, and in knowing that any thought must
have a certain sort of absolute unity, i.e. the unity that makes it truly the thought that
(e.g.) grass is green, we know, to put it in terms that will seem provocative to some
Page 16 of 28
The Self
Kantians, although not I believe to Kant, when I have sufficiently explained myself, that
there exist entities that are genuine, concrete metaphysical unities of an unsurpassable
sort.
(p. 559)
The existence of one of these fundamental metaphysical unities may well involve a large
number of neurons acting in concert, as we ordinarily suppose; it may well essentially
involve a plurality of ‘different substances acting together’, to use Kant's formulation in
his discussion of the Second Paralogism (1781–7/1933: A353). And we can't ground any
claims about the non‐materiality and hence possible simplicity and hence incorruptibility
and hence immortality of the soul on the knowable existence of these ‘absolute’ concrete
unities. This, after all, is the line of thought Kant demolishes in his Paralogisms, in which
he takes it as common ground that ‘the assertion of the simple nature of the soul is of
value only in so far as I can thereby distinguish this subject [i.e. the soul] from all matter,
and so can exempt it from the dissolution to which matter is always liable’ (1781–7/1933:
A356; my emphasis).44 This demolition is, however, wholly compatible with, and indeed
derives part of its force from, the fact that we can know that these fundamental and
unsurpassable unities exist even though they are no help in establishing the existence of a
persisting immaterial soul.45 Kant allows that thoughts (experiences) really, concretely
exist, and it is he himself who makes it most vivid that the fact that ‘the I in every act of
thought is one, and cannot be resolved into a plurality of subjects … lies already in the
[very] concept of thought’ (1781–7/1933: B407).
So when I propose that these real concrete unities are or essentially involve selves, and
propose to dignify these selves with the title ‘object’ or ‘substance’, I am not in
fundamental conflict with Kant, because I am not trying to lay down a path to immortality
via incorruptibility via simplicity via non‐materiality.46 I am, though, proposing that there
are in the universe no better candidates for being physical objects or substances than
subjects of experience. For what is a physical object? This question receives many
different answers in metaphysics (see e.g. Dorr 2005: 234–6), but a physical object is, at
least, and most basically, a certain kind of physical unity. The primordial criterion for
something's counting as an object, in fact, is its having a certain sort of unity; and it now
appears that there is no more certain or absolute unity in nature than the unity of the
subject of an experience in having a thought,47 the unity of subjectivity which is
necessary to the thought's genuinely occurring at all. The fundamental particles of
physics are perhaps the best rival candidates for being unsurpassable physical unities,
but photons and electrons as described by physics cannot rival selves or subjects so far as
the certainty of their ontological unity is concerned, if indeed they exist at all as
irreducibly separate unities, as now seems extremely doubtful. (p. 560)
— Won't there be two subjects in the L‐reality if temporal ‘overlap’ of the sort you
describe does actually occur in the L‐reality? And isn't this consequence of your view
completely unacceptable?
Page 17 of 28
The Self
Yes to the first question—if there can be such overlapping then there will for a time be two
experiences‐with‐subjects in the L‐reality, one rapidly waning and the other rapidly waxing—No
to the second. The idea of two selves or loci of consciousness coexisting in a single brain at a
single time is another idea that may offend our ordinary intuitions and emotions, but there is no
theoretical difficulty in it, and there is reason to believe that it sometimes happens, and may
indeed be routine. If there is such temporal overlapping, neither of the two neuronal‐
recruitment‐constituted thin subjects that will then be numerically distinguishable at a given
time t will be aware of there being two subjects at t, any more than Louis the whole human being
considered as a (thick) subject of experience will be. However we count subjects in this case, we
may take it that the experience of each one of them will have, for that subject, the character of
being part of a continuous process of experience on the part of a single subject.48
It's like a word running in lights along a news ticker in Times Square. We naturally think
of the word as a single persisting thing, but really it is a rapid series of numerically
distinct illuminated things each of which is made up of numerically distinct parts, and
would be seen as such by a creature with a higher flicker‐fusion rate than our own. If we
now imagine that the word is conscious, and feels that ‘it’ has a continuous existence,
although there is no continuing ‘it’ at all, only a rapid series of numerically distinct
illuminated words each of which individually has the feeling that it has continuous
existence, we have a passable model of the ordinary human experience of a continuing
self, such as it is.50 It bears comparison with the experienced constancy and steadiness of
vision across constant saccades.
(p. 561)
The persistence belief is at the heart of the ordinary conception of the self, and it is not
hard to see how it arises. It is supported, directly or indirectly, by many things; most
fundamentally, our deep, innate, often thoroughly sensible and extremely general
tendency, so ingeniously analysed by Hume, to posit the existence of continuing single
things, long‐term continuants, when confronted by certain sorts of diachronically
sustained patterns of resemblance between phenomena—experiences, or in his
terminology ‘perceptions’—that are in fact numerically distinct. This tendency leads us,
as Hume observed, to assume or posit (‘suppose’) a continuously existing table or chair as
the cause—and thus explanation—of certain sorts of resemblances among our successive
Page 18 of 28
The Self
experiences just as surely as it leads us to posit a continuing self as the common locus of
these experiences and part of the cause (and thus explanation) of their resemblances,
which are resemblances of content.51
Fundamental among these resemblances of content from one experience to the next are
deep, regularly repeated—constantly revived—similarities of interoceptive,
proprioceptive, somatosensory bodily awareness, sustained both from moment to moment
and from day to day,52 and equally deep similarities of mental feeling‐tone that are in fact,
although we may not know it, part of this overall bodily awareness (see e.g. Damasio
1999). To this we may add all the similarities and flow coherencies of development of
perceptual and cognitive content from experience to experience, as we move around in
the world and think our thoughts. Further support comes in from our exteroceptive and
cognitive awareness of our evident singleness and continuity as whole human beings. To
this we may add the fact that our thought about ourselves moves very freely between our
apprehension of ourselves as persisting human beings and our apprehension of ourselves
as mental subjects, in such a way that our awareness of ourselves as persisting whole
human beings feeds the persistence belief about the self even though conceiving of
oneself as a self is precisely conceiving of oneself as something that is not the same thing
as the whole human being. Tied in with all this is the powerful support the persistence
belief receives from one's natural apprehension of others as single continuing persons,
and one's vivid awareness of others' apprehension of oneself as a continuing single
person. All this feeds through to fix the idea of a specifically inner persisting self, (p. 562)
and further support comes from specific forms of persistence‐presupposing moral
emotion (guilt among them); from the instinct of self‐preservation; from the natural
‘narrativity’ of much human thought (see e.g. Dennett 1991: ch. 11; Strawson 2005); from
fear of death; from religious commitments; from the sense that the process of
consciousness is stream‐like; from practical planning purposes: from experience of
memory (backward‐looking) and intention (forward‐looking); and from the intrinsic
continuities of action sequences.
If you believe in an inner self at all, then, you are likely to believe in a persisting inner
self and feel yourself to be one, in so far as you experience yourself as a self; and I have
as remarked on page 551 no decisive argument against a physicalist version of this view
according to which the self is whatever anterior‐insula/precuneus/medial‐prefrontal‐
cortex/reticular‐activating‐system/etc.‐involving thing it is in the brain that supports the
consciousness and personality of a human being and is still there when the human being
is in dreamless sleep. Of the four conditions laid down in 31.3 as crucial to being a self, a
persisting self conceived of in this way can fulfil the first three, and can also be argued to
fulfil the fourth. It is not as if I can show it to be a mistake to say that such a thing is a
subject of experience, a respectable candidate for being called an object or substance,
and not the same thing as the whole human being.
I have no such experience of being a persisting inner self, although my sense of being an
inner mental presence is as strong as anyone's, and this, no doubt, is one reason why I
prefer my way of putting things—although I agree with William James that such features
Page 19 of 28
The Self
References
Berkeley, G. (1713/1975), ‘Three Dialogues between Hylas and Philonous’, in
Philosophical Works, ed. M. R. Ayers (London: Dent).
Bradley, F. (1893–7), Appearance and Reality, 2nd edn. (Oxford: Oxford University Press).
Damasio, A. (1994), Descartes's Error: Emotion, Reason, and the Human Brain (New York:
Avon).
—— (1999), The Feeling of What Happens: Body and Emotion in the Making of
Consciousness (New York: Harcourt Brace).
Page 20 of 28
The Self
Fichte, J. G. (1794–1802/1982), The Science of Knowledge, trans. and ed. P. Heath and J.
Lachs (Cambridge: Cambridge University Press).
Gibson, E. (1993), ‘Ontogenesis of the Perceived Self’, in U. Neisser (ed.), The Perceived
Self (Cambridge: Cambridge University Press), 25–42.
—— (1892/2001), ‘The Self’, Psychology: The Briefer Course (New York: Dover), 43–83.
—— (1999), ‘Body, Soul, and Intellect in Aquinas’, in M. James and C. Crabbe (eds.), From
Soul to Self (London: Routledge), 33–48.
Page 21 of 28
The Self
McLaughlin, B., and McGee, V. (2000), ‘Lessons of the Many’, Philosophical Topics, 28:
129–51.
David (eds.), The Self in Neuroscience and Psychiatry (Cambridge: Cambridge University
Press), 319–35.
—— (1999), ‘The Self and the SESMET’, Journal of Consciousness Studies, 6: 99–135.
—— (2002), ‘Postscript to “The Self”’ in R. Martin and J. Barresi (eds), Personal Identity
(Oxford: Blackwell), 363–70.
—— (2008b), ‘The identity of the categorical and the dispositional’, Analysis, 68/4.
—— (in preparation), The Evident Connexion: Self, Mind and David Hume (Oxford: Oxford
University Press).
Trefil, J. (1997), Are We Unique?: A Scientist Explores the Unparalleled Intelligence of the
Human Mind (New York: Wiley).
Page 22 of 28
The Self
van Inwagen, P. (1990), Material Beings (Ithaca, NY: Cornell University Press).
Notes:
(1) Greene (2004: 471–93) lists some striking reasons for thinking that reality is not
ultimately spatiotemporal—spacetimelike—in nature.
(2) The ‘am’ in ‘I am undone’, ‘I am English’, or ‘I am here’ is not the ‘am’ of identity.
(3) I defend this view against Wittgensteinian and other philosophical‐logical objections in
Strawson (2009: sect. 2.2).
(4) Such intentions and beliefs are common even though they may not be explicit.
(5) ‘Whole human being’ is short for ‘human being considered as a whole’. It doesn't rule
out amputees.
(6) The reference of ‘I’ doesn't expand outwards in a continuous fashion, like the
reference of ‘now’ and ‘here’, with which it is often compared. Instead it moves between
two fixed positions.
(7) Objection. Suppose I say [P] ‘I am conscious and I was conscious yesterday too’, using
‘I’ in both instances to refer to a self (rather than to the human being that I am). We want
[P] to turn out to be true, but it's false, on your view. Reply. I do think [P] is false—it is
quite plain to me that the ‘I’ who is conscious now was just not there yesterday. Still,
someone sympathetic to my view who nonetheless wanted [P] to be true could embrace a
‘counterpart semantics’ for past and future sentences: the second ‘I’ would be evaluated
in terms of there being a counterpart of me that existed yesterday and was conscious (see
Sider 2001; thanks to Brian McLaughlin).
(8) Compare Dawkins (1998: 283–4): ‘each of us humans knows that the illusion of a
single agent sitting somewhere in the middle of the brain is a powerful one’.
(9) In this case the subject is no longer ‘inner’, strictly speaking, but it is still not the
same thing as the human being, and this is the crucial idea. Note that ‘immaterial’ is a
merely negative word—it has no positive meaning at all—and simply means ‘non‐
physical’.
(10) I use the two words interchangeably, aware that the word ‘object’, with its chair‐like
feel, causes greater resistance when applied to the self.
(11) Kant himself makes the relevant point as he attacks the metaphysics of the self
popular in his time (1781/1996: A350), see p. 549 below.
Page 23 of 28
The Self
(12) If that is what a self is then we know there are such things and there is no
philosophical issue.
(13) ‘Principle’ is useful inasmuch as a principle manages to sound like a thing of some
sort, in this old use of the word, without sounding anything like a chair or a dog. Plainly
selves, so conceived, are not things that can exist only as a ‘predicate of something else’,
to put it in Aristotelian/Kantian terms.
(14) If immaterial souls existed they would of course be concrete phenomena, for ‘x is a
concrete entity’ simply means ‘x is not an abstract entity’, where an abstract entity is
something like a number, or a proposition.
(15) All religious claims are by comparison only a little less sensible than the claim that
grass is green.
(16) Note that (1)–(8) are just as likely to be accepted by those who think that selves are
immaterial souls as by anybody else.
(17) ‘Mental’ does not imply ‘non‐physical’, and here means only ‘mentally propertied’.
(18) A strong distinction between substances and their properties is not metaphysically
tenable, as Kant knows (see e.g. Kant 1783/1953: sect. 46), but it serves here to make a
point.
(19) ‘Mode of existing’ cannot here mean the particular way a substance is, where the
substance is thought to be somehow independently existent relative to its mode of
existing; for that would be to take accidents or properties to be somehow ‘subordinate’
after all.
(20) See further Strawson (2008a). Note that our natural counterfactual talk has the
classical object–property metaphysic essentially built into it—it is in fact just an
expression of this metaphysic—and may very easily mislead us.
(21) Van Inwagen holds that selves so understood cannot be said to exist any more than
hearts can, even when it is assumed that human beings are wholly material entities (van
Inwagen 1990: sect. 15; his reasons for saying this are of great interest). Note that those
who identify selves with whole human beings can agree that the brain or self‐system
alone is the self in the case in which the brain or self‐system is all that is left of a human
being, because it is in that case the whole human being; but not otherwise.
(23) ‘Some philosophers … imagine we are … intimately conscious of what we call our
SELF: that we feel its existence and its continuance in existence; and are certain both of
its perfect identity and simplicity.’ But we do not ‘have … any idea of self, after the
manner it is here explain'd’ (Hume 1739–40/2000: 164; my emphasis).
Page 24 of 28
The Self
(24) For a discussion of Hume's widely misunderstood position on personal identity and
the self, see Strawson (2001, in preparation).
(25) Descartes does not think one can prove immortality, and when the second edition of
the Meditations appeared he took care to delete the words ‘in which the immortality of
the soul is demonstrated’ from the subtitle. (They had been added in the first edition
without his knowledge or consent; see Clarke 2006: 202–3.)
(26) In January 1642 he writes to Regius: ‘I wish above all that you would never propose
any new opinions but, while retaining all the old ones in name, only offer new arguments.
No one could object to that, and anyone who understands your new arguments properly
will conclude immediately from them what you mean. Thus, why did you need to reject
substantial forms and real qualities explicitly?’ (quoted by Clarke 2006: 224; see
Descartes 1619–50/1991: 205).
(27) He thereby rejects a fundamental assumption of his time (and ours?). See e.g. Clarke
(2003: chs. 1, 8, 9).
(28) See also Clarke (2003: 215): there is ‘no real distinction, in the Cartesian sense,
between a thing and its properties’.
(30) Memory that cannot on Descartes's official theory be stored in the brain.
(31) ‘The term “faculty” denotes nothing but a potentiality’ (Descartes 1648/1984: 305).
(33) ‘The I exists only in so far as it is conscious of itself … The self posits itself, and by
virtue of this mere self‐assertion it exists; and conversely, the self exists and posits its own
existence by virtue of merely existing. It is at once the agent and the product of
action’ (Fichte 1794–1802/1982; 98, 97). ‘There is no preexisting I; rather the I is
delineated, is synthesized around the reflexive act. An entity is synthesized around the
reflexive act and it is the “I” of that act’; ‘the self which is reflexively referred to is
synthesized in that very act of reflexive self‐reference’, ‘an entity coagulates’ (Nozick
1981: 87, 91, 88).
(34) ‘Thought. I use this term to include everything that is within us in such a way that we
are immediately aware of it. Thus all the operations of the will, the intellect, the
imagination, and the sense are thoughts’ (Descartes 1641/1984: 113).
(35) A common and spectacular misreading of James has him taking the self to ‘consist
mainly of [muscular] motions in the head or between the head and throat’ (James
1890/1950: 301).
Page 25 of 28
The Self
(36) He makes the same point in other terms in James (1890/1950: i. 338–42).
(37) Compare Damasio (1994: 240): ‘[A]t each moment the state of self is constructed,
from the ground up. It is an evanescent reference state, so continuously and consistently
reconstructed that the owner never knows that it is being remade unless something goes
wrong with the remaking’.
(39) Kant makes the same point in his famous footnote in the Third Paralogism (1781–
7/1933: A363–4). The notion of appropriation would of course be much too strong if it
suggested that no content is ever lost.
(40) Given short‐term or ‘working’ memory, the content of the immediately preceding
experience standardly forms part of the overall experiential context in which the new
experience arises in every sense in which features of the external environment do. This is
true both of content in the focus of attention and background content, somatosensory
content, mood‐tone content, and so on (somatosensory and mood‐tone contents are
continually refreshed from sources that usually undergo only slow change—from pulse to
pulse, subject to subject—even as salient contents may turn over rapidly). It may be
added that the features of the brain that constitute a human being's character, beliefs,
general epistemic and cognitive outlook, general conative and emotional outlook, and so
on generally remain highly stable from moment to moment.
(41) I owe this objection to Mark Sacks. Note that when James uses the phrase ‘melt into
each other like dissolving views’ a second time (1890/1950: i. 279) he is talking about
transitions within a single pulse of thought.
(42) The claim that they are unities is wholly compatible with the view that they involve
many neurons and many parts of neurons. Note that a ‘temporal slice’ of the subject of
experience cannot be said to be itself a subject having that experience, or a part of that
experience, or any experience.
(43) ‘That … the I of … thought … designates a logically simple subject—this lies already
in the very concept of thought’ (Kant 1781–7/1933: B407).
(44) This qualification of his attack on the idea that we can know that the soul is a simple
substance—that we can't know it to be a simple substance in such a way as to support the
hopes and ends of religion—is crucial but little noticed. Kant goes on to say that ‘this is
indeed, strictly speaking, the only use for which the above proposition is intended’ (1781–
Page 26 of 28
The Self
7/1933: A356), ‘the only case in which this concept can be of service’ (1781–7/1933:
A360).
(45) Even if we do not know them as they are in themselves, given that they have for us
an irreducibly temporal character.
(46) I expand on this claim, which may be doubted, in Strawson (2009: sects. 2.9, 4.3,
8.5).
(49) Compare Dennett's compelling account of the persisting self as an illusion, a ‘center
of narrative gravity’ (1991: 418). See also Dennett (1988).
(50) A substantial minority of human beings have no particular experience of there being
a persisting, continuing self.
(51) For Hume's general discussion of this process see his (1739–40/2000: 133–5, 165–9
[1739–40/1978: 200–4, 253–9]). For his discussion of it with specific regard to the self see
Hume (1739–40/2000; 169–71 [1739–40/1978: 259–62]). There is reason to think that the
process in the case of the self might be more different from the process in the case of
chairs than Hume allows, even on the terms of his own scheme. For important recent
improvements on the general point, which can be put by saying that the concept of
objects as continuants is innate, see e.g. Spelke (1994).
(52) Wundt: ‘the most speculative of philosophers is incapable of disjoining his ego from
those bodily feelings and images which form the incessant background of his awareness
of himself’ (1874/1911, quoted in James 1890/1950: i. 303 n.). Bradley: ‘there seems … no
doubt that the inner core of feeling, resting mainly on what is called Coenaesthesia, is the
foundation of the [sense of the] self’ (1893–7: p. 68). James: ‘we feel the whole cubic mass
of our body all the while, it gives us an unceasing sense of personal existence’; ‘the
nucleus of the “me” is always the bodily existence felt to be present at the
time’ (1890/1950: i. 333, 400). The idea is old but has recently been treated as new.
Page 27 of 28
The Self
Galen Strawson
Page 28 of 28
Unity of Consciousness
When one focuses on words on a monitor and, say, feels a twinge of pain, one is not
conscious of the words and, separately, of the pain. One is conscious of the words and the
pain together, as aspects of a single experience. At least since Kant, this phenomenon has
been called the unity of consciousness. A variety of approaches to characterizing unified
consciousness have been tried by different theorists. Some start from the idea that a
unified conscious experience is a composite of other experiences. Others assert or
assume that, while a unified conscious experience will have a complex object or content,
it has no experiential parts. This article returns to this disagreement. The first two ways
of characterizing the unity of consciousness that are examined here are within the
experiential-parts approach.
WHEN one focuses on words on a monitor and, say, feels a twinge of pain, one is not
conscious of the words and, separately, of the pain. One is conscious of the words and the
pain together, as aspects of a single experience. At least since Kant, this phenomenon has
been called the unity of consciousness.
32.1 Approaches
Page 1 of 16
Unity of Consciousness
What characterizes the unity of consciousness? Note that this question can be asked and
answered whether there actually is any such thing as unified consciousness or not.
Indeed, we need to know what we are talking about even to address the question about
existence. That said, it should also be noted that it is difficult to say much about the unity
of consciousness that is both non‐question‐begging and more than a thinly disguised
synonym, a point that Dainton (2000) emphasizes. Even as great a theorist of the subject
as Immanuel Kant threw up his hands. He observed that this unity is ‘not the category of
unity’ (1781–87/1998: B13)—and said no more.
(p. 566)
The first two ways of characterizing the unity of consciousness that we will examine are
within the experiential‐parts approach.
32.1.1 Subsumption
32.1.2 Co‐consciousness
Page 2 of 16
Unity of Consciousness
Some theorists combine the two approaches. In fact, many of the people who use the
term ‘co‐consciousness’ seem to have subsumption in mind as the underlying notion.
Dainton, for example, embraces the language of co‐consciousness, but cashes the notion
in terms such as ‘being experienced together’ (2000: 236). Lockwood (1989: 88)
describes co‐consciousness as ‘the relation in which two experiences stand, when there is
an experience of which they are both parts’. Similarly Shoemaker: ‘The experiences are
co‐conscious … by virtue of the fact that they are components of a single state of
consciousness’ (2003: 65).
Like subsumption, co‐consciousness would also require that there be experiential parts.
At any rate that is true of co‐consciousness as a relationship among experiences. James
held that it is a relationship among objects of a single experience, which is compatible
with unified conscious experiences not having parts—a view to which he in fact adhered,
as we will see.
At least two approaches to unified consciousness are compatible with the no‐experiential‐
parts view. One we find in Tye (2003). In unified conscious states, (p. 567) the things that
we experience are ‘experienced together’, ‘enter into the same phenomenal
content’ (2003: 36)—which phenomenal content could be the content of a single non‐
composite experience. Another approach has been advanced by Brook and Raymont
(Brook 1994: 38; 2000; Brook and Raymont 2006, forthcoming). The key idea here is joint
consciousness: If an experience that one is having provides consciousness of any item,
then it provides consciousness of other items and of at least some of the items as a group.
Likewise for consciousness of acts of experiencing. This notion is related to the
phenomenal side of the Bayne and Chalmers notion of subsumption—there being
something it is like to be in two conscious states simultaneously (Bayne and Chalmers
2003: 32)—but in a way that does not require subsumption.
32.2 Taxonomies
Most writers agree that unified consciousness comes in a number of forms. There
agreement stops. There are almost as many ways of dividing it up into sub‐kinds as there
are theorists. Tye (2003: 11–15), for example, distinguishes object unity,
neurophysiological unity, spatial unity, subject unity, introspective unity, and, finally,
phenomenal unity. Bayne and Chalmers (2003: 24–7) distinguish ‘objectual’ unity (a
matter of two conscious experiences of one object, so different from Tye's object unity),
spatial unity, subject unity, and subsumptive unity—the last a matter of two or more
conscious states becoming aspects of a single conscious state. Then within subsumptive
unity they distinguish between access unity and phenomenal unity. Their definition of the
Page 3 of 16
Unity of Consciousness
latter, as with Tye their main interest, we just examined: two conscious states are
phenomenally unified ‘if there is something it is like for the subject to be in both
conscious states simultaneously’. And so on.
Interestingly, Kant and philosophers in his tradition break phenomenal unity itself down
into sub‐kinds, usually starting with the division of the mind into subject, representation,
and object or content and assigning to each a special form of unified consciousness. Few
contemporary theorists break phenomenal unity down at all, so this division is of some
interest. Kant's division results in at least four sub‐kinds.
The first three can be expressed in terms of the notion of joint consciousness just
introduced. There is unified consciousness of individual objects, Tye's object unity; Bayne
and Chalmers's objectual unity. The most common name for the process at work here is
binding (Hardcastle 1998; Revonsuo 1999). Binding is the process of tying various
features of a visual scene such as colour, shape, edges, and contours, features detected in
various places in the visual cortex, together into an experience of a unified, three‐
dimensional object. Binding may be necessary for consciousness of individual objects but
it does not seem to be sufficient. We must, it seems, also be jointly conscious of the
various elements to have unified consciousness of an object.
Kant and many Kantians have also spoken of a fourth sub‐kind of the unity of
consciousness; namely, unified consciousness of oneself. Here one is or certainly seems to
be (Rosenthal 2003 says merely seems to be) conscious of oneself not just as subject but,
in Kant's words (1781–7/1998: A350), as the ‘single common subject’ of the many aspects
Page 4 of 16
Unity of Consciousness
of the unified experience that one is now having and of a number of similar experiences
past and, in anticipation, still to come. Such consciousness of self is a sub‐kind of unified
consciousness.
Hume (1739–40/1962: 252) and Rosenthal (2003) deny that we have such unified
consciousness of self and Dennett (1991, 1992) says at minimum that we have much less
of it than we think. We will return to this issue.
Some theorists hold that there is also a fifth form of unified consciousness. We might call
it unity of focal attention. It consists in our ability to pay unified, focused attention to
things. It differs from the last three sub‐kinds of unified consciousness. In them,
consciousness ranges over either many represented items, or complex representings, or
oneself as subject of the representings. Unity of focus picks out something within these
unified ‘fields’. Wilhelm Wundt (1874/1893: ii. 67) captured what we have in mind in his
distinction between the field of consciousness (Blickfeld) and the focus of consciousness
(Blickpunkt). Unity of focal consciousness is more like unified consciousness of a single
item than it is like any of the other three.2
The first four sub‐kinds are clearly Kantian. Kant did speak of attention, too—not very
often but he did speak of it (e.g. Kant 1781–7/1998: B156 n.).
Page 5 of 16
Unity of Consciousness
As an example, consider a classical argument for mind–body dualism. It starts like this:
When I consider the mind, that is to say, myself inasmuch as I am only a thinking
thing, I cannot distinguish in myself any parts, but apprehend myself to be clearly
one and entire
Descartes then asserts that if the mind is not made of parts, it cannot be made of matter,
because anything material has parts. He adds that this by itself would be enough to prove
dualism, had he not proved it already. Notice where it is that I cannot distinguish any parts. It is
in the unified consciousness that I have of myself. The central claim is then that unified
consciousness could never be achieved by a system of components.
Arguments of this sort were familiar to Kant (chiefly from similar reasoning in Leibniz
and Mendelssohn). However, Kant held that unity tells us nothing about what sorts of
entity minds are, including whether or not they are made out of matter (1781–7/1998:
chapter on the paralogisms of pure reason). His argument is that the achievement of
unified consciousness by a system of components acting together would be no less
mysterious than its being achieved by something simple—that is, that has no components
(Kant 1781–7/1998: A352).
The notion that consciousness is unified was also central to one of Kant's own arguments:
his transcendental deduction of the categories. There Kant claims that in order to tie
various objects of experience together into a single unified conscious representation of
the world, we must be able to apply certain concepts to the items in question. In
particular, we have to apply concepts from each of four fundamental categories of
concept: quantitative, qualitative, relational, and what he called ‘modal’ concepts. We
cannot explore this argument, but it does illustrate how central the unity of
consciousness was in Kant's thought.3
As we saw earlier, unified consciousness played a major role in James's work, too, under
the label ‘co‐consciousness’ (James 1909/1967: 221).4 Early in the twentieth century the
notion disappeared from the work on the mind of all but a few philosophers. Logical
atomism in philosophy and behaviourism in psychology had little (p. 570) need for such a
notion. A partial exception is Carnap's choice (1928) to avoid any commitment to atoms of
experience as the elements of his system, opting instead for ‘total experiences’. His
notion of irreducible experiential wholes can, as we will see, be usefully developed.
Page 6 of 16
Unity of Consciousness
Carnap's actual motivation seems to have been a desire to make his system compatible
with gestalt psychology (Smith 1994: 23). However, while gestalt unity does indeed
appear in some structures of which one is conscious, it is different from the unity that
characterizes one's consciousness of various objects, objects that need not include unified
gestalt structures.
Brain‐Bisection Operations
Commissurotomies are the best‐known situations in which unified consciousness seems to
split but otherwise remains intact. Often referred to as brain‐bisection operations, the
abnormality is that the person seems to have been split into two unified centres of
consciousness within one body. In a commissurotomy, the corpus callosum, a large strand
of about 200,000,000 neurons running from one hemisphere to the other, is cut. When
present, it is the chief channel of communication between the hemispheres. (These
operations are a last‐ditch effort to control certain kinds of severe epilepsy by stopping
the spread of seizures.)
(p. 571)
In special laboratory conditions it is natural to hold that the original single instance of
unified consciousness has been fissioned into two such instances. Here are a couple of
examples of the kinds of behaviour that prompt this assessment. The human retina is split
Page 7 of 16
Unity of Consciousness
vertically in such a way that the left half of each retina is primarily hooked up to the left
hemisphere of the brain and the right half of each retina is hooked up to the right
hemisphere of the brain. Now suppose that we flash the word TAXABLE on a screen in
front of a brain‐bisected patient in such a way that the letters TAX hit the left side of the
retina, the letters ABLE the right side and we put measures in place to ensure that the
information hitting each half of the retina goes only to one lobe and is not fed to the
other. If such a patient is asked what word is being shown, the mouth, controlled usually
by the left hemisphere, will say TAX while the hand controlled by the hemisphere that
does not control the mouth (usually the left hand and the right hemisphere) will write
ABLE. Or, if the hemisphere that controls a hand (usually the left hand) but not speech is
asked to do arithmetic in a way that does not penetrate to the hemisphere that controls
speech and the hands are shielded from the eyes, the mouth will insist that it is not doing
arithmetic, has not even thought of arithmetic today, etc.—while the appropriate hand is
busily doing arithmetic!
Note that in none of these situations do we have much inclination to say that the unity
has been destroyed or even much damaged. It is just ‘parcelled out’ differently. The same
may be true of the next condition that we will examine.
In contrast to the cases that we just considered, there are phenomena in which unified
consciousness seems more to shatter than to split.
Schizophrenia
In some particularly severe forms of schizophrenia the victim seems to lose the ability to
form an integrated, interrelated representation of his or her world and self altogether.
The person speaks in ‘word salads’ that never get anywhere, indeed sometimes never
become complete sentences. The person is unable to put together or act on plans even at
the level necessary to obtain sustenance, tend to bodily needs, or escape painful irritants.
Here it seems more correct to say that the unity of consciousness has shattered rather
than split. The behaviour of sufferers seems to express what (p. 572) we might call mere
experience fragments, the contents of which are so narrow and unintegrated that the
subject is unable to cope and interact with others in the ways that even split‐brain
subjects can.
Dysexecutive Syndrome
Page 8 of 16
Unity of Consciousness
In schizophrenia of the severe sort just described the shattering of consciousness is part
of a general breakdown or deformation of mental functioning: affect, desire, belief, even
memory all suffer massive distortions. In another kind of case the normal unity of
consciousness seems to be just as absent but there does not seem to be general cognitive
or affective disturbance. This is what some researchers call dysexecutive syndrome
(Dawson 1998: 215). What indicates breakdown in the unity of consciousness is that these
subjects are unable to consider two things together, even things directly related to one
another. For example, such people cannot figure out whether a piece of a puzzle fits into a
certain place even when the piece and the puzzle are both clearly visible and the piece
obviously fits. They cannot crack an egg into a pan. And so on.5
Simultagnosia
A disorder presenting similar symptoms is simultagnosia, or Balint's syndrome. In this
disorder patients see only one object located at one ‘place’ in the visual field at a time.
Outside a few ‘degrees of arc’ in the visual field, patients say they see nothing but an
‘undifferentiated mess’ and seem to be receiving no information about objects
(Hardcastle 1998: 562).
What is common to dysexecutive disorder and simultagnosia is that subjects appear not
to be conscious of two items in a single conscious state. They cannot, for example,
compare the objects. If the person has any representation of the second item at all, it is
not unified with consciousness of the first one. Unlike the split‐brain case, the second
representation is not just missing from one conscious state. It is not unified with any
other conscious states, too. Rather than consciousness being split into two unified
parcels, there is little unity or none at all.
Page 9 of 16
Unity of Consciousness
Brain‐bisection cases are a contrast in the other direction. The ability to tie objects
together is intact, especially in the left lobe. What opens the way to the idea of a split in
consciousness is that in some situations whatever is conscious of some items being
represented in the body in question is not conscious of other items being represented in
that same body at the same time. (See the examples of the word TAXABLE and of doing
arithmetic.)
Then there are two claims about what characterizes unified consciousness. One urges
that some kind of relationship among contents is generally required, logical coherence for
example (Hurley 1994, 1998).6 Unified conscious states cannot be inconsistent with one
another. But that cannot be right. Suppose that one sees a stick immersed in water as
being bent but knows that this is an illusion. Here one's conscious perception conflicts
with one's conscious belief that it is not bent. Yet one's consciousness of them is unified.
The other we will call the strong unity thesis. This is the thesis that all the conscious
states of a subject are unified; every conscious state is unified with every other conscious
state. Thus Bayne and Chalmers (2003: 24): ‘Necessarily, any set of conscious states of a
subject at a time is unified’. Against this thesis, cases of dissociation may be adduced,
particularly split‐brain cases, where it looks as if a subject has parallel streams of
consciousness between which there is no unified consciousness.
(p. 574)
Even though there is reason to question these requirements on unity, the idea that unified
consciousness goes with such exotic properties may be behind some of the recent
scepticism about whether there is any such thing. This is the fourth claim. Davidson
(1982) and Dennett (1991) have both urged that the mind's unity has been overstated,
citing the everyday incoherence of our beliefs, perceptions, etc. From the fact that not all
of one's conscious contents are unified it does not follow, of course, that none are.
Page 10 of 16
Unity of Consciousness
Proponents of EP include Lockwood (1989: 88), who introduces the notion of co‐
consciousness as ‘the relation in which two experiences stand, when there is an
experience of which they are both parts’, and Dainton (2000: 88), who says: ‘The relata of
co‐consciousness are experiences’ and speaks of co‐conscious experiences as ‘component
parts’ of the ‘total experience’ that results from their linkage (2000: 214). Similarly,
Shoemaker (2003: 65) says: ‘The experiences are co‐conscious … by virtue of the fact that
they are components of a single state of consciousness’.7
James was the first to champion NEP in an unmistakable way. He endorsed it in the
course of repudiating the ‘mind‐stuff theory’, according to which ‘our mental states are
composite in structure, made up of smaller states conjoined’ (1890: 145). Against this
James says that while our experience is complex, this complexity is not a matter of there
being several experiences (or ‘feelings’) present in an encompassing experience. This is
because ‘we cannot mix feelings as such, though we may mix the objects we feel, and
from their mixture get new feelings’ (1890: 157). If one's (p. 575) experience appears to
become more complex, that is a matter of a single experience's content being more
complex, and is not the addition of more experiences (of the diversity of content).
In the tradition of James, NEP is now advocated by Searle and Tye among others. Searle
ventures the hypothesis that ‘it is wrong to think of consciousness as made up of parts at
all’. For I have a ‘single, unified, conscious field containing visual, auditory, and other
aspects’, but ‘there is no such thing as a separate visual consciousness’ (2002: 56). Do
‘visual experiences stand to the whole field of consciousness in the part/whole relation?’.
No, says Searle (2002: 54). Tye offers a similar view, which he dubs the ‘one‐experience
view’ (2003: ch. 1). Considering the polymodal nature of our experience, he says: ‘There
are not five different … experiences somehow combined together to produce a new
unified experience’. Instead, ‘there is just one experience here’ (2003: 27).
Page 11 of 16
Unity of Consciousness
As James saw it, EP faces a difficulty that does not confront NEP. His well‐known example
of the twelve‐word sentence is designed to expose it. Suppose each word in the sentence
is known by one and only one of twelve people. It is hard to see, James says, how these
twelve thoughts could be combined to yield a unified consciousness of the sentence. As he
says:
Take a sentence of a dozen words, take twelve men, and to each one word. Then
stand the men in a row or jam them in a bunch, and let each think of his word as
intently as he will; nowhere will there be a consciousness of the whole sentence.
(1890: 160)
What EP needs is a way of combining experiences that yields more than just an experiential
aggregate, for a mere combination of experiences is not the experience of a combination. EP
needs, then, a way of putting together experiences that also puts together their contents.
Without any specification of how this combining of contents is to be achieved, we are left with a
mere aggregate of experiences, each member of which is oblivious to the contents of the other
states in the aggregate. As James says: ‘Idea of a+idea of b is not identical with idea of
(a+b)’ (1890: 161).
Tye links NEP to the idea that conscious states are transparent. ‘Transparent’ here means
that while I am conscious via conscious states, I am not conscious of them. I ‘see through’
them; hence ‘transparency’. What might seem to be qualities of conscious states are
really qualities that things are represented as having. ‘Phenomenal unity is a relation
between qualities represented in experience, not between qualities of experiences’ (Tye
2003: 36).
However, proponents of NEP need not affirm transparency. One alternative is the idea
that conscious states are self‐representing states, states of which one becomes conscious
just by having them. On this approach, some of the qualities of which one becomes
conscious in having an experience belong to the experience itself. This idea seems to be
entirely consistent with NEP. If it is, NEP carries no commitment to the transparency
thesis.
References
Bayne, T. (forthcoming), The Unity of Consciousness (Oxford: Oxford University Press).
Page 12 of 16
Unity of Consciousness
Brook, A. (1994), Kant and the Mind (Cambridge: Cambridge University Press).
Brook, A., and Raymont, P. (2006), ‘The Unity of Consciousness’, in E. Zalta (ed.),
Stanford Encyclopedia of Philosophy, <http://plato.stanford.edu/entries/
consciousness‐unity/>, accessed 2008.
Carnap, R. (1928/1967), Logical System of the World, trans. R. George (Berkeley, Calif.:
University of California Press).
—— (1992), ‘The Self as the Center of Narrative Gravity’, in F. Kessel, P. Cole, and D.
Johnson (eds.), Self and Consciousness: Multiple Perspectives (Hillsdale, NJ: Erlbaum),
101–15.
Page 13 of 16
Unity of Consciousness
Simulation, and the Unity of Consciousness (Oxford: Oxford University Press), 49–78.
Kant, I. (1781–7/1998), Critique of Pure Reason, trans. and ed. P. Guyer and A. Wood
(Cambridge: Cambridge University Press).
Nagel, T. (1971), ‘Brain Bisection and the Unity of Consciousness’, Synthese, 22: 396–413.
Page 14 of 16
Unity of Consciousness
Smith, B. (1994), Austrian Philosophy: The Legacy of Franz Brentano (Chicago, Ill.: Open
Court).
Tye, M. (2003), Consciousness and Persons: Unity and Identity (Cambridge, Mass.: MIT
Press).
White, S. (1990), The Unity of the Self (Cambridge, Mass.: MIT Press).
Notes:
(1) Kant was the first to articulate this distinction between unified consciousness of
individual objects and of multiples of objects (Kant 1781–7/1998: first division, bk. 1, ch.
2; see Brook 1994: 123).
(2) If focus is closely related to attention, as we suggested, attention need not be part of
at least some forms of consciousness (Hardcastle 1997).
(3) One aspect of Kant's strategy continues to capture the imaginations of philosophers:
his attempt to connect the unity of consciousness to the structure of knowledge.
Arguments of this form can be found in P. F. Strawson (1966), Hurley (1994, 1998), and
Cassam (1996). For a critical analysis of moves of this kind see Brook (2005).
(4) Parfit reintroduced the term into contemporary literature (1984: 250).
Page 15 of 16
Unity of Consciousness
(6) Hurley stops short of applying this constraint universally, but says that it applies ‘in
many cases’ (1998: 115).
(7) Tye (2003: 21) numbers Bayne and Chalmers among the adherents of EP (without
using that label). However, while Bayne and Chalmers do indeed speak of the
encompassing conscious state as involving ‘at least a conjunction of each of many more
specific conscious states’ (2003: 27), and of a ‘complex phenomenal state and a simpler
state that is intuitively one of its components’ (2003: 40), they also caution that thinking
here in terms of ‘a mereological part/whole relation among phenomenal states’ should be
regarded only as an ‘aid to intuition rather than as a serious ontological proposal’ (2003:
40). So it is not clear how their view should be classified.
(8) An earlier version of some portions of this chapter appeared in Brook and Raymont
(2006). Material carried over to this chapter has been extensively rewritten.
Paul Raymont
Andrew Brook
Page 16 of 16
Personal Identity and Metaphysics
Print Publication Date: Jan 2009 Subject: Philosophy, Metaphysics, Philosophy of Mind
Online Publication Date: Sep 2009 DOI: 10.1093/oxfordhb/9780199262618.003.0034
This article tries to situate the recent discussions of personal identity in their historical
context. It begins with a brief discussion of Locke's account of identity as such, turning
then to a more detailed account of his views on personal identity in particular. It then
discusses a pair of problems that arise for any sort of neo-Lockean account, and surveys
the sorts of responses that have been offered to them. Finally the article briefly presents
three recent accounts, each of which can be seen as relying on a subtle re-
characterization of the traditional problem of personal identity.
Keywords: personal identity, metaphysics, neo-Lockean concept, judgements, living things, imaginary cases
THE traditional philosophical problems surrounding the issue of personal identity arise
from trying to answer the following series of questions in a systematic way. Given a
person X, we want to know:
Page 1 of 17
Personal Identity and Metaphysics
The questions themselves appear to be straightforward. But it has proved challenging to answer
them in a philosophically satisfying way. It is surprisingly difficult to come up with an account of
personal identity that simultaneously respects plausible commitments about the metaphysics of
identity while preserving the platitudes about personhood that seem to characterize our intuitive
grasp of the concept. As a result, most sophisticated contemporary accounts are both subtle and
complicated. Indeed, there is little agreement even about what sort of concept person is—a
substance concept or a phase sortal? a natural‐kind concept or a conventional one? an ethical
concept or a metaphysical one?—and hence little agreement about what sort of analysis we
should aim to provide for it. Without some sense of how the debate came to have the form it
does, recent discussions of personal identity are likely to seem bewildering.
(p. 579)
In the remainder of this chapter I will try to situate these recent discussions in their
historical context. I will begin (Section 33.1) with a brief discussion of Locke's account of
identity as such, turning then to a more detailed account of his views on personal identity
in particular. I will then (Section 33.2) discuss a pair of problems that arise for any sort of
neo‐Lockean account, and survey the sorts of responses that have been offered to them.
Finally (Section 33.3), I will briefly present three recent accounts, each of which can be
seen as relying on a subtle re‐characterization of the traditional problem of personal
identity.
Contemporary discussions of personal identity take their cue from John Locke's
discussion ‘Of Identity and Diversity’ in his Essay Concerning Human Understanding.
Locke presents his analysis in the context of defending the view that identity is ‘suited to
the Idea’ (1690; II. xxvii, sect. 7); that is, that criteria for identity (over time) are criteria
for identity as an X (over time).1 These criteria, he suggests, can be grouped into three
categories:2 one for material substances (atoms and masses of matter); one for living
entities (plants and animals); and one for self‐conscious beings (persons).3
Page 2 of 17
Personal Identity and Metaphysics
identical to an atom at another if they are connected by a continuous path in space and
time. (Locke 1690: II. xxvii. sect. 2).4
Masses of matter are objects where ‘two or more atoms [are] joined together into the
same mass’ (Locke 1690: II. xxvii. sect. 26); that is, collections of contiguous atoms with
some sort of force connecting them.5 A given mass of matter consisting of some set of
contiguous atoms is identical to any future (or past) mass of matter consisting of those
same contiguous atoms so long as those and only those atoms remain properly united
throughout. As with atoms, our evidential basis for such judgements rests on our ability
to track objects that trace continuous paths through space–time (though we are
presumably insensitive to minor changes in constitution, and hence to the identity of stuff
over time); few particular practical judgements seem to rest on (re)identifying such
complexes, though again our general practices presuppose the absence of massive error
(Locke 1690: II. xxvii. sect. 3).
In sum, atoms and masses of matter are individuated on the basis of their constitution
properties and nothing else—an atom or mass at one time is not (identical to) an atom or
mass at another unless they have exactly the same contiguous matter; facts about how
that contiguous matter is organized, or about properties that emerge as a result of that
organization, are irrelevant. By contrast, objects that fall under the sortals that Locke
discusses next are individuated not by the matter that composes them but by the various
properties that their organized matter manifests.
Page 3 of 17
Personal Identity and Metaphysics
33.1.1.3 Persons
The final category that Locke discusses is that of person: ‘a thinking intelligent being,
that has reason and reflection, and can consider itself as itself, the same thinking thing,
in different times and places’. A person at one time is identical to a person at another
time if they are the same ‘rational being’: ‘as far as this consciousness can be extended
backwards to any past action or thought, so far reaches the identity of that person’.6 In
ordinary cases we make judgements about personal identity both on the basis of physical
properties (as sameness of human animal is generally indicative of sameness of person)
and on the basis of manifest psychological characteristics (evidenced through speech and
non‐verbal behaviour).7 And these judgements are of great practical significance; for
‘person’ is ‘a forensic term’ (Locke) and legitimate attributions of praise and blame
require that the individual who is punished or rewarded be the same person as the
individual who performed the commendable or condemnable deed. Moreover, it is identity
of personhood that matters to us subjectively, ‘happiness and misery being that for which
every one is concerned for himself, and not mattering what becomes of any substance,
not joined to, or affected with that consciousness’ (Locke 1690: II. xxvii. sects. 9–26).
In short, Locke insists, it is ‘one thing to be the same substance,8 another the same man,
and a third the same person’ (1690: II. xxvii. sect. 7): the first is a question about
sameness of matter; the second is a question about sameness of living being; the third is
a question about sameness of consciousness. And, he insists, it is the third of these
questions that is of primary importance to our daily practices, both with regard to
ourselves and with regard to others: our judgements of praise and blame and our basis of
self‐concern track personal rather than animal or substantial identity.
Page 4 of 17
Personal Identity and Metaphysics
that developing an adequate theory of personal identity will require careful sorting out of
our intuitions, which may well be coloured by our tendency to conflate distinct notions.
It is in this context that he introduces the now famous example of the prince and the
cobbler, as follows: ‘Should the soul of a prince, carrying with it the consciousness of the
prince's past life, enter and inform the body of a cobbler, as soon as deserted by his own
soul, every one sees he would be the same person with the prince, (p. 583) accountable
only for the prince's actions: but who would say it was the same man?’ (1690: II. xxvii.
sect. 15).12 That is, if the psychology that we associate with the prince came to be united
with the body that we associate with the cobbler, we would say (speaking loosely) that the
resulting being was the same person as the prince, but that he was the same man as the
cobbler.13 Moreover, the ‘every one’ who ‘sees [that] he would be the same person with
the prince’ includes the prince himself; for on the basis of his subjective experiences it
would seem quite apparent to him that what has happened is simply that he has changed
bodies.
Page 5 of 17
Personal Identity and Metaphysics
satisfactory criterion for identity over time for persons that is reflexive,
(p. 584)
symmetric, transitive, and intrinsically determined, and that obeys the law of
indiscernability of identicals.14
33.2.1 Coincidence
The first cluster of problems includes the following. On one natural reading, Locke is
committed to a sortal‐relative view of individuation according to which there are at least
three things sitting where you are right now—a mass of matter, a human animal, and a
person—each a material object, each composed of many (or all) of the same parts, and
each occupying the same space.15 Moreover, if the human animal is capable of thinking
(and it is hard to see why it would not be), then there are at least two thinking things
sitting where you are now, a person and a human animal, each contemplating whatever
you are contemplating.16 And the problems do not end here. For each of these entities is a
distinct and viable candidate for reference by your name: when I spoke of something
‘sitting where you are right now’ or ‘contemplating whatever you are contemplating’,
what exactly did ‘you’ refer to—the person, or the animal, or the mass of matter, or
perhaps all three at once—or was it instead ambiguous?
Among those who take this cluster of worries seriously in the specific context of personal
identity, three sorts of responses have been offered. Some think the problems so severe
that they should cause us to reconsider the adequacy of any sort of materialist account of
personhood. (For discussion see Zimmerman 2003; see also Swinburne 1984; Merricks
1998.) Others insist that for these and other reasons we should accept an ontology that is
four‐dimensionalist; that is, an ontology according to which the material world is
composed of temporal as well as spatial parts (see Sider 2001). The final group responds
by limiting the number of contenders, for example by denying that one or the other of the
candidate objects exists,17 or by insisting that our referential practices allow us to latch
on properly to the correct entity (see, for example, Noonan 1998).
Even if we set aside these issues, difficulties remain. For the appealing feature of the
Lockean analysis is that it seems to capture why we care about tracking persons over
(p. 585) time; namely, that we care about tracking the clusters of psychological properties
that they manifest. But in so far as these property clusters are qualitatively specified,
they could not serve as criteria for identity, since there is no guarantee that a property
specifed qualitatively will be uniquely instantiated;18 and in so far as they are specified in
some other way, the initial appeal of the psychological account is lost.
33.2.2.1 Teletransportation
Page 6 of 17
Personal Identity and Metaphysics
Two families of descendents of Locke's prince and cobbler reveal these difficulties. The
first family of cases exploits the assumption that two structurally identical individuals
composed of qualitatively identical matter will manifest the same characteristics, both
physically and (assuming materialism) psychologically. If so, there is no in‐principle
barrier to teletransportation—in which information about the microstructural properties
of an entity provides a blueprint for producing one or more duplicates. Whereas certain
psychological‐continuity criteria would credit each duplicate with being identical to the
original, their (presumed) non‐identity with one another renders this interpretation
untenable: if personal identity is an equivalence relation, then mere causal continuity of
consciousness is not sufficient for personal identity. So Locke cannot be right that an
individual at one time is the same person as an individual at another time if they share
the same consciousness.
33.2.2.2 Fission
Suppose, then, that we retreat from a pure neo‐Lockean view, and insist that personal
identity requires not only psychological continuity but also sameness of sustaining
substance. Will this psychology‐plus‐substrate view do the trick? No: unless we require
that that sustaining substance be indivisible (in the sense that no two disjoint proper
parts of it could each be sufficient for sustaining the requisite consciousness), parallel
problems will arise.
We can see this by considering a slightly modified version of Locke's prince and cobbler
case. Assume for the sake of discussion that it is the prince's brain that plays the role of
sustaining the prince's consciousness. Assume further (idealizing somewhat) that the
features required for sustaining consciousness are distributed redundantly, so that if
either half of the brain were transplanted into an appropriate environment (such as the
recently decerebrated body of the cobbler) we would be inclined to follow Locke in saying
that ‘the resulting being was the same person as the prince’.19 But now suppose that,
instead of a single transfer taking place, fission occurs: the brain is divided and each of
the lobes is suitably resituated in one of two decerebrated cobbler bodies.
(p. 586)
Now there are two individuals—each with the prince's psychology, each with enough of
the prince's brain such that were it not for the presence of his competitor we would say
that he was the same person as the prince.20 But even if we would say that each resulting
being is the same person as the prince, we surely cannot say it of both (since they are not
identical to one another, and hence cannot both be identical to the same thing). Nor do
we wish to say it of one of the resulting beings as opposed to the other (since by
hypothesis each on its own would be an equally good identity candidate, so there are no
non‐arbitrary grounds for choosing between them). So unless we assume that there were
two persons there all along (see Lewis 1976), the only remaining possibility is that neither
is the same person as the prince. But if we maintain our initial assessment of the half‐lobe
prince and cobbler case, then this means that a process (say, the transfer of the left lobe)
that is entity‐preserving under one set of circumstances (the single transplant case) is
Page 7 of 17
Personal Identity and Metaphysics
33.2.2.3 Responses
Since reflexivity, symmetry, and transitivity are non‐negotiable, two sorts of retreat seem
possible. The first involves maintaining a commitment to intrinsic determination and
holding that person is a phase sortal (something that can apply to a given entity at some
but not other times) rather than a substance concept (something that applies to an entity
at every moment of its existence) (see Wiggins 2001: 30).22 The cost of such a view is that
persons turn out to be much less fundamental than we might have thought: a person is
not something that a given individual could fundamentally be; instead, person is a status
that a given individual (who belongs fundamentally to some other kind) could have.23
A second retreat involves abandoning the commitment to the view that identity is
intrinsically determined.24 This move is problematic for two reasons. First, (p. 587) there
are intuitive grounds for thinking that the facts that determine whether an entity persists
over time should be facts about the entity itself, not facts about (the existence or non‐
existence of) some other entity. So unless one thinks that there are no interesting criteria
of identity over time that depend only on intrinsic relations between the candidate
entities (along with the obtaining of certain background facts), the extrinsic
determination of same person raises the spectre that personal identity is identity manqué.
The second reason for worrying about the abandonment of intrinsic determination is that,
if one accepts a series of individually quite plausible assumptions, it implies that
‘personal identity is not what matters’.25 The argument goes as follows. Suppose the
prince's left lobe is transferred according to the procedure described above, and that one
accepts the psychology‐plus‐substrate view. Regardless of whether the right lobe is also
transferred (in which case he will go out of existence and be replaced by another
individual), or destroyed (in which case he will continue to exist as the new occupant of
the cobbler's body), all of the things that matter to the prince about his relation to some
future being (for instance, that it be psychologically continuous with him for the right
sorts of reasons) will be present: they depend only on the intrinsic relations between him
and his continuer(s), and not on facts about any of their relations to anything else. So, in
contrast to personal identity, which is extrinsically determined, the thing we care about
when we worry about our future is something that is intrinsically determined—so
personal identity cannot be the thing we care about.26
Page 8 of 17
Personal Identity and Metaphysics
interested in things that matter to us we are interested in something other than identity.
It looks as if we can talk about strict identity; or we can talk about persons; or we can talk
about something that matters—but we can't talk about them at the same time.
The problem, of course, is that we would like to do all three at once. The reason Locke's
analysis is so appealing is that it seems to allow us to pick out a class of entities that we
care deeply about, whose identity‐conditions are a matter of their psychological features,
and which persist over time in a ‘strict and philosophical’ sense. The reason the debate
about personal identity is so complicated is that—even setting aside the larger
metaphysical issues about coincident identity and reference—it looks as if we just can't
have everything we want.
33.3.1 Olson
In a series of papers culminating in his 1997 The Human Animal Eric Olson considers
arguments akin to those in Section 33.2, leading him to conclude that what each of us is,
strictly speaking, is a human animal whose persistence conditions are independent of
facts about our psychology: ‘no sort of psychological continuity, with or without further
physical qualifications, is either necessary or sufficient for us to persist through
time’ (1997: 4).
This might seem like a wholesale rejection of the Lockean account. And in some sense it
is. But by reconceptualizing what sort of thing persons are, Olson's framework leaves
room for Locke's insights. Olson suggests that even if our persistence conditions are
those of animals, the animals that we are may also be persons, where ‘being a particular
person’ is ‘a sort of role or office that a human organism (or any other appropriate object)
might fill at a particular time’ (1997: 66). This allows for the possibility that ‘being the
same person is a practical or moral relation’—characterized by just the sorts of
psychological‐continuity relations that neo‐Lockeans have identified. And it likewise
allows that our legal and emotional practices be predicated on personhood, just as Locke
maintained.
Page 9 of 17
Personal Identity and Metaphysics
33.3.2 Rovane
According to Rovane, the core notion of personhood is the notion of agency: on her view,
persons are beings who are capable of acting and bringing others to act on the basis of
reasons (as opposed to non‐rational causes); they are all and only those beings who are
‘commit[ted] to achieving overall rational unity’ (1998: 8; italics omitted). As a result,
Rovane's view allows that persons may be something radically distinct from individual
human animals: groups of human animals might together form a single person (‘group
persons’), or numerous persons (such as the ‘alters’ of multiple personality disorder)
might together exist within a single human being (‘multiple persons’). Moreover, since
‘persons are nothing but certain sorts of episodes standing in certain sorts of relations’,
relations which may obtain to a greater or lesser degree, ‘the boundaries between
persons can be vague … [both] over time and at a particular time’ (1998: 134). So on
Rovane's view as well we might say (using Olson's terminology) that being the same
person is not a metaphysical relation.
33.3.3 Sider
Page 10 of 17
Personal Identity and Metaphysics
while, on such a view, being the same person might be, in so far as anything is, a
metaphysical relation, being a metaphysical relation is, to put the point slightly
tendentiously, not one (see Sider 2001).
In short, though they disagree about much else, each of our contemporary authors holds
that the question of personal identity (as traditionally understood) is not a metaphysical
question (as traditionally understood)28—though the question of whether it is such a
question is a metaphysical question par excellence. Though there (p. 590) are ways of
resisting this picture, it is nonetheless striking how readily the Lockean legacy pushes us
down one of the many paths towards it.29
References
Alston, W. P., and Bennett, J. (1988), ‘Locke on People and Substances’, Philosophical
Review, 97: 25–46.
excerpt repr. in J. Perry (ed.), Personal Identity (Berkeley, Calif.: University of California
Press, 1975), 99–105.
—— (1980), Reference and Generality. 3rd edn. (Ithaca, NY: Cornell University Press).
Hudson, H. (2001), A Materialist Metaphysics of the Human Person (Ithaca, NY: Cornell).
Page 11 of 17
Personal Identity and Metaphysics
Lewis, D. (1976), ‘Survival and Identity’, in A. Rorty (ed.), The Identities of Persons
(Berkeley, Calif.: University of California Press), 17–41;
repr. in Lewis, Philosophical Papers (Oxford: Oxford University Press, 1983), 55–72.
excerpt repr. in J. Perry (ed.), Personal Identity (Berkeley, Calif.: University of California
Press, 1975), 33–52.
Merricks, T. (1998), ‘There Are No Criteria of Identity Over Time’, Noûs, 32: 106–24.
Olson, E. T. (1997), The Human Animal: Personal Identity Without Psychology (Oxford:
Oxford University Press).
—— (2002a), ‘An Argument for Animalism’, in J. Barresi and R. Martin (eds.), Personal
Identity (Oxford: Blackwell), 318–34.
—— (2002c), ‘Personal Identity’, in S. Stich and T. Warfield (eds.), The Blackwell Guide to
the Philosophy of Mind (Oxford: Blackwell), 352–68.
(p. 591) Parfit, D. (1984), Reasons and Persons (Oxford: Oxford University Press).
Rea, M. (1997) (ed.), Material Constitution: A Reader (Lanham, Md.: Rowman &
Littlefield).
excerpt repr. in J. Perry (ed.), Personal Identity (Berkeley, Calif.: University of California
Press, 1975), 107–18.
Rovane, C. (1998), The Bounds of Agency (Princeton, NJ: Princeton University Press).
Page 12 of 17
Personal Identity and Metaphysics
—— (1999), ‘Self, body, and coincidence’, Proceedings of the Aristotelian Society, suppl.
vol. 73: 287–306.
Unger, P. (1980), ‘The Problem of the Many’, Midwest Studies in Philosophy, 5: 411–67.
Van Inwagen, P. (1990), Material Beings (Ithaca, NY: Cornell University Press).
Wiggins, D. (2001), Sameness and Substance Renewed (Oxford: Oxford University Press).
Notes:
(1) Note that this sort of commitment to sortal‐relative individuation need not bring with
it a commitment to (Geachean) relative identity (the view that (objects) x and y both could
be the same (kind) F without being the same (kind) G, while both being F and G). (See
Geach 1967; for discussion see Wiggins 2001; Deutsch 2002; Hawthorne 2003.) Avoiding
it, however, requires either accepting some type of dominant sortal view (that each object
belongs fundamentally to one particular sort), or adopting some sort of temporal‐parts
view (that entities have temporal as well as spatial parts), or embracing coincidence (that
more than one entity (of the same kind) might be at the same place at the same time).
These issues are discussed briefly in Section 33.2 below.
(2) The taxonomy is not exhaustive; for example, Locke does not discuss identity criteria
for non‐living natural objects (such as rocks and mountains), though he does discuss (as I
will not) identity‐conditions for God and for finite intelligences (which he holds to be akin
to those for simple substances), and offers a few brief remarks concerning identity‐
conditions for artefacts (which he holds to be akin to those for living entities).
(3) Each of the main paragraphs in the next three subsections is organized around
answering analogues of (1)–(4) for the sortal in question. That is, for each of Locke's
categories I present synchronic and diachronic identity‐conditions for Cs, articulate the
Page 13 of 17
Personal Identity and Metaphysics
basis on which we do and could make judgements about C‐identity, and say something
about the practical and emotional significance of C‐hood.
(4) Indeed, these matters seem so fundamental that the prospect of massive error (say,
that atoms indiscernibly swap locations every few seconds) seems best met by re‐
characterizing our referential intentions.
(5) Locke does not say whether such objects are ‘nested’, so that a (suitably configured)
four‐atom clump would contain four three‐atom clumps each containing three two‐atom
clumps, or whether only maximal discontinuous clumps count.
(6) Locke's text seems to suggest that he is committed to a ‘memory’ criterion for
personal identity: P2 is identical to P1 if and only if P2 remembers P1’s experiences. This
account faces a number of well‐known problems, including the problem of non‐
transitivity: that identity over time is transitive, but memory is not (an objection generally
credited to Reid 1785), and the problem of circularity: that the concept of memory
presupposes the concept of personal identity and hence cannot be used to explicate it (an
objection generally credited to Butler 1736). In order to finesse these and other
objections, contemporary accounts appeal instead to some broader requirement, such as
psychological continuity, according to which P2 is identical to P1 only if P2’s
psychological states are connected to P1’s psychological states by the right sort of causal
chain. Such views are generally referred to as neo‐Lockean views.
(7) Locke explicitly notes our fallibility in this regard (see Locke 1690: II. xxvii. sect. 22).
(8) For discussion of Locke's use of the term ‘substance’ see Alston and Bennett (1988).
(9) The locution ‘Socrates the mass of matter’ may be taken as shorthand for ‘the entity
with mass‐of‐matter persistence conditions that (is one of the things) we refer to by
“Socrates” ’. For discussion of some of the problems associated with such a notion see
Section 33.2. Strictly speaking, it is unlikely that we ever use the term ‘Socrates’ to refer
to some mass of matter; its exacting identity‐conditions render its practical interest
minimal. We do, however, sometimes seem to use personal names to refer to something
like ‘stuff’, as when we say ‘Socrates is buried outside Athens’ or ‘Poor Socrates: he's all
over the universe by now’. Whether this is a literal use of the term—some might contend
that ‘Socrates’ here is used as shorthand for ‘the matter that once composed Socrates'
body’—and whether the ‘stuff’ in question is a suitable candidate for reference are
matters for consideration elsewhere.
(10) Again, we need to be a bit careful here. The memories of Socrates the person are
things of which he is conscious; the memories of Socrates the human animal may be
things of which the person with which he coincides is conscious. For discussion of some
of these issues see Noonan (1998) and Olson (1997, 2002a).
Page 14 of 17
Personal Identity and Metaphysics
(11) Cf. Unger: ‘[W]hy not stick only to actual cases? … The reason is that this extremely
conservative methodology is apt to incur great costs … In attempting to ascribe beliefs to
ourselves on the basis of quite limited data, we might wrongly describe our own
attitudes’ (1990: 11). Similarly Shoemaker: ‘What Mackie and Perry have done is to
indicate how personal identity (or co‐personality, or psychological unity) are realized in
us, i.e. in members of our own species. And this does not answer the question ‘What does
personal identity consist in?’ at the level of abstractness at which we want it
answered’ (1984: 127).
(12) Twentieth‐century renderings of this story generally cast brain in the role of soul. In
Sydney Shoemaker's version (1963), for example, the brain of one Brown is transferred to
the body of one Robinson, carrying with it all of Brown's psychological characteristics.
The resulting character, whom Shoemaker dubs Brownson, ‘display[s] all the personality
traits, mannerisms, interests, likes and dislikes, and so on that had previously
characterized Brown, and … talk[s] and act[s] in ways completely alien to the old
Robinson’. ‘What would we say if such a thing happened?’, asks Shoemaker. ‘There is
little question that many of us would be inclined, and strongly inclined, to say that while
Brownson has Robinson's body, he is actually Brown’ (1963: 23–4). The substitution
introduces a number of subtle differences whose significance is exploited in the
contemporary literature: unlike the soul, the brain is part of the body; and unlike
(presumably) the soul, the brain could be divided without loss of function. (For an
additional difference between Shoemaker and Locke n. 14)
(13) ‘Speaking loosely’ because this way of putting things suggests a commitment to
relative identity, which presumably we wish to avoid (see Sect. 33.1.1). One might handle
this problem by privileging one of the sortals (as Shoemaker does), by explicitly
acknowledging a commitment to coincident identity, or by re‐characterizing the
ontological situation so as to allow such relativity. For further discussion of these issues
see Section 33.2.
(14) The law of indiscernibility of identicals says that if x is identical to y, then x has all
and only the properties that y has.
(15) Indeed, the problems do not end here. Since each of the entities is (presumably) a
material object, each itself confronts the gauntlet of puzzles that have shaped
contemporary discussions of material constitution, including the problem of undetached
parts (Tib/Tibbles) (see Geach 1980) and the problem of the many (see Unger 1980). (For
recent discussion of these and related issues see the essays collected in Rea 1997.)
Ironically, the generality of the conundrum may make the particular problem of sortal‐
relative identity seem less pressing.
(16) Olson (1997, 2002a, 2000b) makes much of this problem; for one response see
Shoemaker (1999).
Page 15 of 17
Personal Identity and Metaphysics
(17) For various versions of this approach see van Inwagen (1990), Olson (1997);
Merricks (2001), and Wiggins (2001).
(18) One might insist that clusters of psychological properties are necessarily uniquely
instantiated. But justifying this insistence would require either denying the possibility
that two sufficiently complex sets of qualitatively identical microparticles could be
configured in identical ways, or rejecting the supervenience of the mental on the physical.
(19) Some have denied this, insisting that the required physical substrate involve at least
51 per cent of the original matter; most consider this move to be ad hoc (see Unger 1990;
Wiggins 2001).
(20) For discussion of some of the subtleties involved in presenting this case in a non‐
question‐begging way see Unger (1990: 259–60).
(23) On the view in question, substance concepts are constrained by a requirement that—
in so far as we are concerned with constitutive (question 3) criteria, as opposed to
evidential (question 4) criteria—the facts that determine identity over time are facts
about a and b and the relations between them, and not facts about anything else (the
‘only‐a‐and‐b rule’) (see Wiggins 2001: 96 ff., 231 ff., 237 ff.). It is a subtle question
whether there is really sense to be made of the only‐a‐and‐b rule as a metaphysical (as
opposed to an epistemic) principle; discussing this issue would take us beyond the scope
of this chapter.
(24) Sydney Shoemaker considers and endorses such a move in his (1984).
(25) The main advocate of such a view has been Derek Parfit. For one widely discussed
formulation of this argument see Parfit (1984).
(26) For one way of resisting this line of reasoning see Gendler (2002); one might also
contend that ‘what matters’ is itself extrinsically determined (see Johnston 1987, 1992;
Sosa 1990).
(28) Did Locke take himself to be asking a metaphysical question when he set discussion
on its current course? Answering this seems a tricky exegetical issue. But regardless of
what Locke thought he was asking, it is clear that many of his followers set out to answer
such a question.
Page 16 of 17
Personal Identity and Metaphysics
(29) For comments and suggestions I am grateful to Ansgar Beckermann, Tim Crane,
Brian McLaughlin, Ted Sider, and Zoltán Gendler Szabó.
Page 17 of 17
Imagination
The topic of imagination is broad and multifarious: we are dealing with an extended
family of interrelated concepts and capacities. Imagining covers everything from the
simple mental image, to the philosophical counter-example, to the most sublime act of
creation. It takes in both dreaming and logical thinking, reverie and hypothesis formation.
To bring order to this range of mental phenomena it helps to make an initial distinction
between what the article calls sensory and cognitive imagination — roughly, between
forming a mental image and entertaining a possibility in the conceptual style.
Keywords: imagination, mental image, logical thinking, hypothesis formation, cognitive imagination, sensory
imagination
THE topic of imagination is broad and multifarious: we are dealing with an extended
family of interrelated concepts and capacities.1 Imagining covers everything from the
simple mental image, to the philosophical counter‐example, to the most sublime act of
creation. It takes in both dreaming and logical thinking, reverie and hypothesis formation.
To bring order to this range of mental phenomena it helps to make an initial distinction
between what I shall call sensory and cognitive imagination—roughly, between forming a
mental image and entertaining a possibility in the conceptual style. Thus, I may form a
visual image of the colour blue or of a friend's face, on the one hand, and I may consider
the possibility that a skyscraper might at some future time be built on this very spot, on
the other. The ascription of the former type of imaginative act has the form ‘x forms an
image of y’, the latter takes the form of a that‐clause: ‘x imagines that p’. Sensory
imagination, as the label suggests, partakes of the phenomenology of the sense
modalities in some way (in which way we shall have to examine)—we ‘see’ with the mind's
eye as well as with the ordinary eye—while cognitive imagination is a conceptually based
propositional attitude, in the same general family as belief. What we might call ‘the
Page 1 of 13
Imagination
imaginative faculty’ includes both sensory and cognitive imagination, and of course these
may interact with each other. I shall accordingly divide my discussion into two parts.
Two extreme views have been advanced. One view, associated with Hume, maintains that
the image is really just a faded percept:2 the senses present the mind with a sensory
impression of some specific type, and this impression undergoes a process of degradation
as it lingers in memory, becoming fainter as time passes. Even an invented image—as,
say, of a fire‐breathing dragon—is just the bleached‐out version of a possible percept.
Images and percepts differ purely quantitatively, in their degree of ‘force and liveliness’.
And this, it is supposed, is why we intuitively judge that images and percepts have
something in common—the image just is a percept, though one that has lost a lot of its
initial sparkle. On this view the percept is the basic item of the pair, with the image being
just a special case of it: there can be percepts without images, but not vice versa. The
relation between percept and image is that of an original to an indistinct copy.
According to the second extreme view, associated perhaps with Kant, it is the other way
about:3 it is imagination that is the basic item of the pair, with percepts being derivative
from imaginative acts. Thus, it is held that all perception is a kind of imaginative
construction, in which stimuli from the world are processed and interpreted by the
imaginative faculty. The mind must contribute to the bare input of the senses, it is held, in
order for anything like a conscious percept to occur, and this active ingredient in
perception is the work of the imagination—a constructive, creative faculty. Perception is
really just the imagination tied down to an impinging perceptual stimulus. Accordingly,
there is no genuine perception without imagination—imagination suffuses perception.
The reason for the similarity between image (p. 597) and percept is therefore quite
simple; in effect, the percept is an image, though one that is anchored to a sensory input.
Page 2 of 13
Imagination
The first step in getting clear about imagination is to see that both of these assimilationist
positions are incorrect. Neither percept nor image is a special case of the other, despite
the manifest similarity between them. Each is sui generis, though they are intimately
related. I shall now enumerate some of the salient points of difference, which preclude
both assimilations.
When you perceive something, you are passive before your sensory experiences. The
course of your visual experience, say, is outside the direct control of your will; it is
dictated by the barrage of stimuli that assail the sense organs. You cannot simply decide
to see something and conjure up the appropriate visual experience. This is as true of
hallucinating as it is of veridical perceiving: you cannot, at will, cause yourself to
hallucinate pink rats, say. Suppose you are seeing a swelling ocean and would rather be
seeing a tropical jungle; this wish cannot be converted into the desired experience, save
by taking an appropriate trip. But it is quite otherwise with images: they are ‘subject to
the will’ (to use Wittgenstein's term (1981: 621, see also 627, 632)).4 If I don't like what I
am seeing I can indeed conjure up an image of something more to my liking—in fact, I do
this all the time. I can also order you to form an image of your father, and you can
voluntarily comply. Even when an image comes to you unbidden, you can control its
course and termination; at least, it makes perfect sense to try to do these things. Forming
images is something that belongs on the side of action, not passion: it can be motivated,
exhibit weakness of will, satisfy desire, and be morally evaluated. Here, then, is a marked
difference between the perceptual faculty and the imaginative faculty: the latter is
subject to executive control, being part of our active nature, while the former operates
outside the domain of the will. This point already shows that both assimilations have to be
mistaken: images cannot be a type of percept, pace Hume, or else they would be as
passively received as percepts; and percepts cannot be externally elicited images, or else
they would be in principle subject to the will. Thus, the passivity of perception and the
activity of imagination show that neither is a special case of the other.
Second, percepts and images relate differently to attention.5 I can attend to what I am
seeing or I can fail to do so (and similarly with the other senses). But it is not that I am
blind with respect to what I am not attending to—there are unattended sightings of
things. My sensory experiences proceed in unattended form as I focus my attention on
other matters. Shifts of attention away from a percept do not obliterate it. However, the
same does not seem to be true of images: they do not persist in unattended form as I
focus my attention on other matters. I cannot form an image of the colour blue, say, and
then leave that image in suspension in my consciousness as I turn my attention to lunch.
My images do not subsist as a background to my (p. 598) conscious life over which my
attention selectively plays, as my percepts do; rather, they are products of my attentive
acts. When I form an image of something I thereby attend to it in the mode of
imagination, and I cannot keep it in my imaginative sights while I switch attention
elsewhere. In this respect images are like thoughts: I also cannot keep thinking of
something as I turn my attention to other matters. It would be nice if I could, because
then I could think of many things at the same time, but, unfortunately, thinking
intrinsically takes up attentional space. Images are likewise hungry for attention, and wilt
Page 3 of 13
Imagination
away from its spotlight. But, then, images must differ fundamentally from percepts: if
images were a type of percept, they should be attention‐independent; and if all
perception were an exercise of imagination, then percepts would not enjoy their freedom
from the limits of attention. Imagining is something we do by means of our attentive
faculty, but perceiving is something that befalls us whether we wish to focus on it or not.
Fourth, percepts come with a corresponding sensory field, but images do not. Thus, we
have the visual field, with its periphery and centre, its fixed perimeter, its sense of a
spatially proximate world; but there is nothing comparable in the case of visual images.
To be sure, visual images display spatial extent in some way, but they don't have a centre/
periphery distinction, have no anatomically fixed scope, and do not present the imaged
object as within the surrounding space. When I see something I see it as here/there, in
the vicinity of me, as looming or receding, as in the centre of view or off to the side, as
about to vanish from my visual field or stay put. But when I visualize something none of
this applies: the object is given in a spatially neutral way, not located in my own
egocentric space, and it does not occupy more or less of the fixed field of my inner eye, or
threaten to disappear behind me. Imagined objects appear in a kind of limbo, in a space
all their own, not as located beside other objects and in spatial relation to me and my
body.9 Phenomenologically, imagining is not felt as a kind of orienting to the environment,
an immersion within it; instead, it reaches beyond the impinging world, bringing absent
Page 4 of 13
Imagination
objects within its scope. So perception and imagination do not occur within the same kind
of matrix of intentionality, despite the frequent identity of their intentional objects, and
hence cannot be reduced one to the other.
Fifth, and finally, it is easy to see that images and percepts do not compete with each
other for mental space: while seeing one thing, I can imagine another.10 There are two
sensory streams of consciousness here, and neither usurps the other: the stream of
percepts and the stream of images. On a typical day both streams will be found happily
flowing together, in choreographed simultaneity, as you shift your attention from what is
being currently perceived to the image that momentarily seizes your fancy; in particular,
the exercise of imagination in no sense blinds you to what is going on, nor even dulls your
sensory acuity. And it is fortunate that this is so, or else sensory breakdown would
accompany every act of imagination—no reveries while driving! But notice how hard it is
to square this obvious fact with the assimilationist theories: if an image were merely a
faint percept, rather like an after‐image, then it would interfere with (p. 600) ordinary
perception, since I cannot have different percepts occupying the same region of my visual
field; and if perception were merely externally stimulated imagination, then percepts
would not interfere with other percepts, since a mere image does not prevent the
occurrence of a simultaneous percept with a different content.11 But it is of the essence of
the imaginative faculty that it can operate concurrently with the perceptual faculties,
while the perceptual faculties themselves exclude competing elements—as when that
hovering after‐image momentarily blinds me to what is before my eyes.
None of this is to deny that images and percepts are in some important way similar; it is
just to say that they are not merely versions of each other, differing only in degree. Nor is
it to deny that perception is a constructive process, in which prior knowledge is brought
to bear upon a fragmentary stimulus—so long as this is not exaggerated into the claim
that the very same faculty of imagination that is involved in paradigm cases of imagining
is employed to generate percepts. Perceiving is not a form or mode of imagining, and
imagining is not a form or mode of perceiving. The right thing to say here is rather that
the two are analogous, or even homologous. Ordinary seeing and visualizing are
variations on the theme of visual experience, with neither of them primary or
paradigmatic. It is not that visualizing can only count as visual because of its reduction to
ordinary seeing, nor that ordinary seeing is unified with visualizing by being an instance
of it. It is rather that the concept of visual experience is more abstract than either of
these two species of it, so that it is inherently neutral between them. That neutral concept
of the visual then gets specialized into these two very different exemplifications of it,
marked by the deep contrasts we have enumerated. Visual experience—the genus—comes
in these two very different modes; the mistake is to think that conceptual unification here
requires conceptual assimilation.12 So: some types of visual experience are subject to the
will, attention‐dependent, non‐observational, devoid of visual field, and tolerant of other
types of visual experience. These are the distinctive characteristics of what we aptly call
the mind's eye, and that eye has just as much of a title to the name ‘vision’ as the other
Page 5 of 13
Imagination
kind—once we resist those assimilationist urges. But the broader lesson here is to
acknowledge the affinities between perception and sensory imagination while not
neglecting the deep differences between them. In particular, the autonomy of the
imagination, with respect to perception, needs to be respected.
Having said all that, we must enter an important caveat: it is not that perception and
imagination can never interact, join together, and even interpenetrate. That occurs in the
phenomenon usually labelled seeing as, though this phenomenon might better be called
imaginative seeing.13 This is itself a large subject, but some brief remarks may be in
order. When I see an ambiguous figure, such as the duck/rabbit drawing, in a particular
way, say as a picture of a duck, this is aptly described as imagination (p. 601) making
contact with perception. It is as if the perceptual stimulus evokes an image, in this case of
a duck, which then becomes attached to the percept, thus combining to form an episode
of imaginative seeing. Normally, images float free of the impinging perceptual world, but
on occasion they may be ‘hooked’ by a stimulus—attracted to the stimulus as to a magnet.
Then we see the world under an imagined aspect, which is contributed by the faculty of
imagination, rather than being contained in the stimulus itself. And this imaginative
aspect has all the marks of an image that we have just listed, notably subjection to the
will. This might well encourage the idea that all perception is imaginative, but this is
really just a conflation of disparate phenomena. When I see an object as square in the
ordinary way, there is no sense in which I have imposed an imagined aspect upon it: the
object really is square and this information has forced itself on my senses—even though
much processing of the stimulus must occur for the percept to result. But when at night I
imaginatively interpret a tree as a monster, say, this is a very different matter, since the
tree is not in fact a monster and my seeing it that way really is an exercise of imagination:
I may stop seeing it as a monster once my fear retreats, while nothing like that will stop
me seeing an object as square. Seeing as is a joining or mingling of perception and
imagination, in which the two elements can always be distinguished, and this
phenomenon does not in any way compromise the distinctions drawn above.
34.1.1 Dreams
Dreaming poses a challenge to what we have said so far. If we agree that dreaming is a
form of imagining, then it is not clear that the mental images that compose dreams obey
the conditions that characterize daytime images; specifically, there is the question of their
subjection to the will. Normally, when I dream I seem to myself to be passive before the
onslaught: I cannot manipulate the course and content of the dream, but occupy the role
of helpless (and hapless) recipient. How then can dreaming be a form of imagining, if
imagining is essentially active and willed?
Two ways out suggest themselves. The first is to deny that dreaming is imagining—why
not say that it is hallucinatory in character? That is, we might try suggesting that dreams
consist not of images but of objectless percepts, like the hallucinations of the
dipsomaniac or the brain in a vat. Then, of course, they will not be subject to the will,
Page 6 of 13
Imagination
since percepts in general are not. But there are a number of objections to this suggestion.
In the first place, dreams regularly have a narrative structure that is hard to square with
the percept theory of their contents. Daydream images are suited for manipulation into
narrative sequences precisely because of their subjection to authorial (hence intentional)
control, but if nocturnal dreams consist of percepts they will not be suitable resources for
narrative manipulation. How can percepts be formed, deformed, strung together, made
coherent (or incoherent), according to affective and narrative demands? How can the
mind work on them if they are outside the control of narrative agency?
Third, it is actually not true that all dreaming is resistant to the will: so‐called lucid
dreaming, in particular, offers examples of dreams that are wholly under the dreamer's
voluntary control—which is not feasible if all dreams consist of passively received
percepts. In a lucid dream the subject can determine the course of the dream, and is well
aware that it is a dream, just as in daydreaming; but if the dream experiences are held to
be percepts, which are passive in nature, then lucid dreaming ought to be impossible.
And there are lesser examples of dream control, as when you terminate a particularly
appalling nightmare and wake up.
The second way out abandons the criterion of subjection to the will as definitive of the
imagination: dreams are simply a counter‐example to this, it maintains, so we should
modify our account of what marks images off from percepts. This is certainly an option,
but it is an unattractive one. It introduces an ugly division within the class of images, with
nothing to recommend it except its ability to retain the imagistic account of dreams.
Moreover, we start to lose our grip on the distinction between percepts and images once
we relax the conditions that characterize them. How many distinctive conditions could we
disavow and still retain a viable distinction between images and percepts? Could we also
say that dream experiences carry a visual field, are attention‐independent, are
observational, and still maintain they are images. Surely it would be much more
Page 7 of 13
Imagination
satisfactory to stick to our original list of differentia and try to account for the felt
passivity of dreams in some other way. That is what I propose we do.
The key move is simple: shift agency to an unconscious author. Here I can only sketch the
way this can solve our problem, without going into independent reasons for favouring the
idea.14 If we suppose that the dreaming mind factors into two components, one active and
one passive, where the passive component is unaware of the actions of the active
component, then we can see how the felt passivity of the dreamer is only a partial
account of the matter—for the unconscious component can be as active as you like
compatibly with the felt passivity of the receiving component. That is, we need to
distinguish the author of the dream, a submerged narrative agency, from the audience of
the dream—you, as you experience the dream. Then we can (p. 603) insist that all images
are subject to the will of some agency, but make room for the passivity of the audience of
the dream. There is, in the dreaming mind, a kind of ‘psychic split’, wherein the agency
behind the dream image is sealed off from the recipient of the dream, thus engendering
an illusion of total passivity, which mimics (but is not a case of) the passivity of the
perceptual state. Of course, this is a substantial theoretical move, and would need to be
independently motivated, but my limited point here is that it solves our conceptual
problem, as well as being not entirely ad hoc or unprecedented. It enables us to include
dreams within the domain of the imagination while retaining our earlier conception of
how imagination distinguishes itself from perception. It also provides a neat account of
lucid dreaming, which now emerges as simply a reunification of the psychic split that
normally accompanies dreaming: the author and audience of the dream are one in the
lucid dream, and the susceptibility of the dream to voluntary control is fully evident to the
dreamer.
Page 8 of 13
Imagination
other. What we have here are two kinds of ‘thinking’, irreducible to each other, analogous
to the two kinds of sensory experience we explored earlier.
I think this connection between imagination and possibility leads naturally to a certain
conception of what knowledge of linguistic meaning is; namely, it is the sentence‐directed
imaginative apprehension of possibility. Suppose someone says ‘snow is green’ and I
understand his utterance to mean that snow is green. What has happened is that his
utterance has caused me to construct, from the structure and composition of the uttered
sentence, a certain possibility, which I then associate with the processed utterance: I
envisage the possibility that snow is green by means of my imaginative faculty, and this
envisaging is what my understanding of the utterance consists in. I work out which
possible state of affairs would make the sentence true—its truth‐condition—and the result
of this process is understanding the sentence to represent that very possibility. None of
this would be possible unless I had a faculty that put me into contact with the realm of
possibility—and the name of this faculty is ‘imagination’. Sentences represent
possibilities, and understanding them is imaginatively grasping which possibility is in
Page 9 of 13
Imagination
question (and how this is determined by the structure of the sentence). Knowing a
sentence's truth condition rests upon an imaginative act.16
This conception nicely fits the fact that meaning is essentially freedom from the actual: a
sentence can mean something that has no counterpart in the real world. More precisely,
the falsity of a proposition is no impediment to its being meant by a sentence. Meaning is
a matter of alternatives—regions of logical space. If a sentence is meaningful, its negation
always is. But this kind of freedom from the actual is exactly what imagination trades in:
the contemplation of what is not the case. Also, and connectedly, meaning and
imagination share a combinatorial passion: meanings (p. 605) can separate and join,
recombining with other meanings, and imagination is itself expressly designed to arrange
concepts into new combinations, novel and unheard of. The productivity of meaning is
thus aligned with the productivity of imagination. Just as language can join together
words whose referents are not found together in the world (as with ‘snow is green’), so
imagination can associate elements that are pure fancy. Both are combinatorially
unbounded (within the limits of sense). Thus it is—to put it grandly—that language and
the mind break free of the tyranny of the actual.
This is not to say that meanings are sensory images in the style of old empiricist theories
of understanding—cognitive imagination is not sensory imagination. But it is to draw a
connection of sorts between meaning and images; for images, too, are massively
combinatorial, fluently productive, novel, and stimulus‐free. So although it would be
wrong to identify meanings with mental images, the idea that there is an important
affinity here is not as wide of the mark as much recent philosophy has maintained. The
affinity, however, has more to do with the structure of the image‐generating faculty than
with its products. It is not implausible to conjecture that the imagery faculty pre‐dates the
capacity for imagining‐that, providing it with a foundation of cognitive operations—in
which case, the roots of meaning lie with imagery, if not meaning itself. The old image
theorists were not as wildly in error as we have come to suppose.
The general point I would make is that the imagination is a basic faculty of mind,
distinctive, irreducible to other faculties, which runs through a number of different
mental capabilities, from simple images, to dreams, to language itself. In its furthest
reaches, which we have not discussed—I mean, the creativity of the arts and sciences—it
must take its rise from these more basic imaginative modes. Indeed, in my view, the
creativity of even the humblest modes of imagination far exceeds what we tend to
suppose, especially when we fully acknowledge the deep divisions between perception
and imagination. When a percept is transformed into a memory image, undergoing a
genuine metamorphosis—let alone when radically novel images are generated—the mind
is already operating at a high level of creativity, as measured by the distance between the
percept as input and the image as output. The image is no more a faint copy of an original
percept, its pale simulacrum, than a butterfly is a faint copy of a caterpillar. Images are
things with creativity built right into them.
Page 10 of 13
Imagination
References
Brann, E. (1991), The World of the Imagination (Boston, Mass.: Rowan & Littlefield).
Kant, I. (1950), Critique of Pure Reason, trans. N. K. Smith (New York: Humanities).
(p. 606) McGinn, C. (2004), Mindsight: Image, Dream, Meaning (Cambridge, Mass.:
Harvard University Press).
Wollheim, R. (1980), Art and Its Objects (Cambridge: Cambridge University Press).
Notes:
(1) Strawson writes: ‘The uses, and applications, of the terms “image”, “imagine”,
“imagination”, “imaginative”, and so forth make up a very diverse and scattered family.
Even this image [!] of a family seems too definite. It would be a matter of more than
difficulty exactly to identify and list the family's members, let alone establish their
relationships of parenthood and cousinhood’ (1970: 31). A useful historical and
contemporary survey is provided by Brann (1991). I discuss imagination from several
different angles in McGinn (2004). This chapter is basically a précis of that book, and
states theses more than defends them.
(2) Thus Hume: ‘Those perceptions, which enter with most force and violence, we may
name impressions … by ideas I mean the faint images of these in thinking and
Page 11 of 13
Imagination
reasoning’ (1739–40/1985: 49); and later: ‘That idea of red, which we form in the dark,
and that impression, which strikes our eyes in sunshine, differ only in degree, not in
nature’ (1739–40/1985: 51).
(3) Thus Kant (1950): ‘Psychologists have hitherto failed to realize that imagination is a
necessary ingredient of perception itself’. I take this quotation from Strawson (1970),
who discusses the way Kant brings imagination into perception; in fact, the view became
quite common among psychologists and is regarded by many as a virtual truism.
(4) These remarks are discussed by Budd (1989: ch. 5); see also O'Shaughnessy (2000:
ch. 11) and McGinn (2004: ch. 1). The point needs further elaboration and qualification
than I have supplied in the text.
(5) I discuss this neglected point at length in McGinn (2004: ch. 1).
(6) This feature of imagination is stressed by both Wittgenstein and Sartre, whose
treatments of imagination are remarkably similar. Wittgenstein (1981: 621) writes:
‘Images tell us nothing, either right or wrong, about the external world’. Sartre (1966:
25) writes: ‘[T]he image teaches nothing, never produces an impression of novelty, and
never reveals any new aspect of the object’.
(7) Nothing I say in the text is incompatible with a ‘causal theory’ of the image, to the
effect that it is at least a necessary condition of an image being of an object that that
object should be the cause of the image. Even if we accept that claim, it does not follow
that the object continually updates the image. Compare reference: maybe an object needs
to be at the origin of a causal chain leading to the use of a name for it to be a name of
that object, but it doesn't follow that users of the name are continually receiving causally
mediated information from the object.
(8) Here it might be thought that recent empirical findings cast doubt on this claim, since
one can use an image to answer questions about an object, as when I ‘consult’ an image
of my living room to say how many windows it has. (See Kosslyn 1996 for a survey of
these findings.) However, the point to stress is that such images issue from information
that is already stored in memory, instead of acting as conduits for new information from
the environment. I discuss this question in McGinn (2004: chs. 1, 5).
(9) See Casey (2000: pt. 3) for a discussion of this and other germane matters. I also
discuss the point in McGinn (2004: ch. 4), in which the notion of a distinctive imagistic
space is explored.
(10) Wittgenstein (1981: 621) makes the point that I cannot form an image of an object
while I am looking at that object, though there is in general no problem about forming
images of objects distinct from those one perceives. I discuss this also in McGinn (2004:
ch. 1), entering various qualifications.
Page 12 of 13
Imagination
(11) Of course, if all percepts were images, then forming one image might interfere with
forming another image, given the limits of attention, and this will exclude simultaneous
percepts with incompatible contents. But I take it that it is obvious that this is not the
same phenomenon as perceptual exclusivity, in which it is impossible to have
incompatible perceptual contents in the same region of the visual field. For one thing, this
is a not a phenomenon of attention at all; and the interference between two images is not
a matter of how one's sensory field can present objects; that is, how things can
perceptually seem.
(13) There is a large literature on seeing as; for example, Scruton (1974: ch. 8); Wollheim
(1980: sects. 11–14); Budd (1989: ch. 4).
(15) The idea that people are distinguished from the brutes primarily through their
imaginative capacities is an attractive one, as such ideas go: homo imaginans is not such
a bad label. Certainly, we owe much of our distinctive pattern of life to our imaginations—
even our linguistic life (see below).
Colin McGinn
Page 13 of 13
Thinking
Human language is not the only naturally occurring symbol system. There are many
animals other than human beings that communicate by means of signs or signals; vervet
monkeys, for example, have specialized warning cries for different kinds of predators.
And some animal-communication systems even have a rudimentary syntax: the dances
performed by certain honey bees have structural elements that tell other bees the
direction and distance from the hive of a nectar source. But what's distinctive of human
language — and the feature that Descartes was highlighting — is that the syntax of
human language permits us to take parts of signs and recombine them with parts of other
signs.
Keywords: symbol system, human language, animal-communication systems, reflex behaviour, human syntax
Perhaps the first thing to be said about thinking is that it is a mental activity. But what
does that mean? Most people, in speaking of the ‘mental’, have in mind some kind of
contrast with the physical. Physical activities all involve our bodies, or parts of them, in
some conspicuous way. Physical actions that we perform—walking, writing, playing a
musical instrument—all involve the voluntary movement of our limbs. Physical processes
like perception and respiration are clearly dependent upon the well functioning of our
bodily organs. Thinking seems different from both those things. I can think without
Page 1 of 27
Thinking
moving any part of my body; I can think with my eyes closed, and my ears plugged. The
only thing that seems to be necessary in order for me to think is that I be conscious.
Could it be that thinking doesn't require a body at all?
[A]lthough these machines might do several things as well or perhaps better than
we do, they are inevitably lacking in some other, through which we discover that
they act, not by knowledge, but only by the arrangement [disposition] of their
organs. For, whereas reason is a universal instrument which can serve in all sorts
of encounters, these organs need some particular arrangement for each particular
action. As a result of that, it is morally impossible that there is in a machine's
organs sufficient variety to act in all the events of our lives in the same way that
our reason empowers us to act
Page 2 of 27
Thinking
To understand the argument, we must first appreciate the distinction that Descartes is drawing
between two ways in which seemingly intelligent behaviour might be produced: it might be
simply necessitated by the machine's antecedent physical state—by the ‘disposition of its
organs’—or it might be produced ‘by knowledge’. Only in the second case, Descartes says, is
there genuine intelligence at work, for it is only in this case that the appropriateness of the
behaviour to its circumstances is due to an understanding of the situation. In the first case the
appearance of intelligence behind the behaviour can be explained by the fact that the physical
structure that produces it happens to be (or, in the case of a machine, was designed to be)
specialized to produce precisely that kind of behaviour, in precisely those circumstances. We can
be confident that this is the explanation, Descartes argues, because the apparently skilled
behaviour displayed by animals or machines is always limited to one or a few domains. In
contrast, Descartes says, our faculty of reason, which enables us to act from knowledge, is ‘a
universal instrument which can serve in all sorts of encounters’.
Let's suppose for the moment that Descartes has established a categorical difference
between two types of behaviour production. (Let's call the first type ‘brute causation’ and
the second type ‘rational causation’.) How does he get from this premise to the
conclusion that rational causation cannot transpire within a material medium? He says of
brutes' ‘souls’ as compared with human beings’ that ‘once one knows how different they
are, one understands much better the reasons which prove that the nature of our souls is
totally independent of the body’. How does the rest of (p. 609) the argument go? We need
to understand better what Descartes takes to be involved when behaviour is ‘due to
knowledge’.
In order for us even to suspect that a piece of behaviour is genuinely intelligent, the
behaviour has to be appropriate to the circumstances in which it is displayed. Contrast,
for example, the following two cases: in the first, my doctor strikes my knee with her little
mallet and my foot kicks out; in the second, I kick my foot and hit an assailant who has
just struck me with a mallet (and seems about to do so again). In the first case there is
nothing that makes my kicking appropriate to the circumstance of having been struck by
a mallet; it is simply what my foot does when my knee has been struck in that way. In the
second case, however, it makes sense for me to kick, because doing so may well drive
away my assailant. The first case is a clear example of reflex action—hardly an ‘action’ at
all—and the second a clear example of rational action. What's the difference?
In the second case, as we know, Descartes would say that my behavioural response to the
circumstances I face is mediated by knowledge. That means that one of the antecedents
of my behaviour is a representation of the situation I face. In fact, there are several
representations involved. I need to represent not only my circumstances, but the goal I
wish to achieve, and the means by which I might achieve it. These representations stand
in rational relations to each other. Because I want to drive away my assailant, and
because I believe that kicking her is apt to drive her away, I conclude that kicking her
would be a good thing to do. If these representations are all accurate—if they correctly
represent my situation, my goals, and my options—then the behaviour that results from
this process will be appropriate to the situation: it will further my goal.
Page 3 of 27
Thinking
Here, then, are two surpassingly important features of thoughts: First of all, they
represent—they display what the Austrian philosopher Franz Brentano called
‘intentionality’: the feature of standing for, or being about, something else (Brentano
1874/1973). Second, they possess some kind of rational structure in virtue of which they
can stand in logical or rational relations to each other. It is likely that Descartes had both
these features well in mind when he spoke of actions being due to knowledge.
Returning to the first of my two illustrative cases, the case of reflex ‘action’, we can now
see that the arbitrariness of the kicking response relative to the circumstance of being
struck by the hammer is due to the fact that the kicking response is not in any way
sensitive to my representation of my situation. We can better appreciate the difference by
considering counterfactuals in the two cases. In the second case, if I had not represented
the situation as one in which I was being assaulted then I would not have kicked. But in
the first case the ‘disposition of my organs’ would have determined me to kick regardless
of how I viewed the situation. It is this feature of rationally caused behaviour that makes
for flexibility. In so far as my thoughts accurately represent my situation, goals, etc., and
in so far as my behaviour is counterfactually dependent on my thoughts, my behaviour
will tend to be appropriate to my circumstances whatever they are.
(p. 610)
This, then, is the categorical distinction Descartes needs: between beings that can and
beings that cannot condition their behaviour to representations of their circumstances,
goals, and options. But why think that non‐human animals and machines fall into the
second category rather than the first?
Thus, the fact that they do better than we do does not prove that they have a
mind, for, if that were the case, they would have more of it than any of us and
would do better in all other things; it rather shows that they have no reason at all,
Page 4 of 27
Thinking
and that it's nature which has activated them according to the arrangement of
their organs—just as one sees that a clock, which is composed only of wheels and
springs, can keep track of the hours and measure time more accurately than we
can, for all our care.
So the argument for dualism proceeds: since beings that lack reason can only produce
appropriate behaviour through the operation of specialized organs, they would need a separate
specialized organ for every distinct task they face. It would be impossible—morally impossible,
Descartes says, and this is important—to outfit a purely material being with enough of these
organs to enable the being to respond appropriately to any circumstance whatsoever. Since we
human beings do possess this kind of general flexibility, the faculty that enables it—‘reason’, in
Descartes's terms—must inhere in an immaterial substance.
But there's still a piece missing: why is Descartes so confident that we possess the degree
of flexibility that would require an impossibly large inventory of specialized organs in a
purely material being? What makes him think that reason is truly a ‘universal
instrument’? The answer comes in the first test he gives for telling whether one is dealing
with a person or a mechanical simulacrum, and has to do with language. Machines,
Descartes says,
Notice that when Descartes asserts that no machine could use language, he is very specific
about the limitation he has in mind. He does not doubt that one could build a machine that
would produce what sound like meaningful utterances, and produce them in circumstances for
which they appear appropriate—a machine, for example, that produces the sound ‘Please don't
do that’ if we strike it. The thing Descartes thinks a machine could not do is to ‘arrange words in
various ways to reply to the sense of everything said in its presence’. That is, no machine can
recombine the parts of its ‘utterance’ so as to compose new utterances that make sense relative
to new circumstances.
Descartes is here pointing to a strikingly distinctive feature of the human communication
system: its compositionality. We've already remarked that thought is representational,
that it has intentional content, or meaning. Language does, too: it is composed of symbols
that stand for other things. The set of relations between symbols and the things they
represent is called the semantics of the symbol system. While all symbol systems have
semantics, only some have syntax—rules that govern the ways in which primitive signs
can be combined so as to produce more complex signs.
Page 5 of 27
Thinking
Human language is not the only naturally occurring symbol system. There are many
animals other than human beings that communicate by means of signs or signals; vervet
monkeys, for example, have specialized warning cries for different kinds of predators.
And some animal‐communication systems even have a rudimentary syntax: the dances
performed by certain honey bees have structural elements that tell other bees the
direction and distance from the hive of a nectar source. But what's distinctive of human
language—and the feature that Descartes was highlighting—is that the syntax of human
language permits us to take parts of signs and recombine them with parts of other signs.
The vervet monkey can utter a shriek that tells its troupe that an eagle is approaching
overhead, and a different kind of shriek that tells them that there's a snake nearby in the
grass. But each of these signs is an indissoluble unit—the monkey cannot recombine parts
of the two shrieks to express, for example, the thought that there's an eagle in the grass
(Cheney and Seyfarth 1990). In contrast, any human being who can say both that there's
an eagle overhead and that there's a snake in the grass can also say that there's a snake
overhead or that there's an eagle in the grass—this is guaranteed by the way sentences of
human languages are composed. Jerry Fodor and Zenon Pylyshyn call this property
systematicity: your communication code is systematic if in possessing the ability to
express one thought you are automatically equipped to express many (Fodor and
Pylyshyn 1988).
But there's another important property of human syntax: it is recursive. This means that
we can take the expression that results from the application of some rule of combination
and reapply that same rule to the new expression. For example, I can form a complex
noun phrase by conjoining two nouns: ‘Bob’ and ‘Carol’ → ‘Bob and Carol’. But I can also
reapply that rule to the noun phrase I just constructed: ‘Bob and (p. 612) Carol’ and ‘Ted’
→ ‘Bob and Carol and Ted’. Compositionality plus recursion give us a guarantee that
language has the resources to generate an unlimited number of distinct, meaningful
signs. This property is called by Chomsky creativity (1975) and by Fodor and Pylyshyn
productivity (1988).
But here lies the great irony. In highlighting the systematicity of thought, Descartes
inadvertently laid the groundwork for a materialist theory of mind. If it is the mind's
systematicity—rather than, say, its subjectivity—that is the sticking point for a materialist
Page 6 of 27
Thinking
account of mentality, then the Discourse argument for dualism could be neutralized if
someone could show how systematicity could be mechanized. And this, it turns out, is
exactly what Alan Turing did, more on this soon.
As we've seen, Descartes presumed that the structure of language mirrored the structure
of thought—that language was, indeed, merely the outward expression of thought. The
most important aspect of this isomorphism was rational structure, secured by
compositionality. This emphasis on rational structure was one of the things that
distinguished philosophical rationalism from the empiricism that emerged in Great Britain
about half a century after the publication of Descartes's Discourse (Locke 1690/1975). So
let's be more explicit about what compositionality involves. We'll necessarily go beyond
Descartes's brief remarks in the Discourse.
Language is compositional: its primitive parts are words, which can be combined to
create phrases and sentences. Does thought, similarly, have parts? It appears that it does.
Consider the thought that HERSHEY'S KISSES ARE MADE OF CHOCOLATE.2 This
thought seems to contain, as components, several elements: one that is about Hershey's
kisses, one that is about chocolate, and one that is about being made of something. In this
respect, thought contrasts interestingly with another kind of mental state, a ‘qualitative’
state like the one I would experience were I to taste a Hershey's kiss. The experience of
tasting the candy doesn't seem at all to be segmented in the way the thought is—I would
be hard pressed to distinguish a HERSHEY'S KISS part of a candy‐eating experience from
a CHOCOLATE part.3
We have a name for thought parts: we call them ‘concepts’, and we individuate thoughts
according to the concepts that compose them. The thought that lobster is (p. 613) high in
cholesterol is a different thought from the thought that the most expensive thing on the
menu is high in cholesterol, because the concept LOBSTER is different from the concept
THE MOST EXPENSIVE THING ON THE MENU even if, in a given circumstance, the
most expensive thing on the menu happens to be lobster. In other words, it's one thing to
think about lobster as lobster, and another thing to think about it as the most expensive
thing on the menu. Thoughts enable us to do this, and they do so by permitting us to pick
out the same thing under different concepts.
Page 7 of 27
Thinking
of Superman in two different ways: as SUPERMAN, and also as CLARK KENT. Because
she has two distinct concepts of the same individual, her SUPERMAN thoughts are
distinct thoughts from her CLARK KENT thoughts.
Brentano held that intentionality was the ‘hallmark of the mental’—that what made a
state a mental state was its being representational. It is currently a hot topic in
philosophy whether this claim is true. It is clear that sensations and emotions are part of
our psychologies, but not so clear that they have representational content; some
philosophers say they do, and some say they do not.5 But if such states do represent, they
do it in a different way than thoughts do. I do not have available to me different ways of
experiencing the taste of the candy, one a ‘Hershey's kiss’ way and one a ‘chocolate’ way.
There is one distinctive way that a Hershey's kiss tastes, and that one way defines the
qualitative character of the experience. Relatedly, my experience of the taste of a
Hershey's kiss does not depend on my having any knowledge or understanding of
‘Hershey‐ness’. I need the concept HERSHEY'S KISS in order to think the thought ‘A
Hershey's kiss is made of chocolate’, but I don't need it to experience its chocolate‐y
goodness. We can capture these differences by saying that thoughts are conceptual while
qualitative mental states are non‐conceptual.
Not all concepts are the same. The concepts HERSHEY'S KISS and CHOCOLATE are
different in a crucial way from the concept IS MADE OF. Roughly, the first (p. 614) two
concepts are about things, while the third is about a property of, or a relation among,
things. These different kinds of concept combine in specific ways, giving thought its
structure.6 We noted above that concepts can combine and recombine in lots of different
ways, provided they are of the right type to work together. To a first approximation, we
can say that any concept of the ‘thing’ type can combine with any concept of the
‘property’ type. (In language, the mirror distinction is between subject terms and
predicate terms.) This is, obviously, not all there is to conceptual structure, but it may be
the fundamental feature of conceptually structured representation (Evans 1982; Fodor
2007).
The version of computationalism that I'll present below is Cartesian or rationalist, in that
it takes for granted that thought is structured compositionally. I'll have a few words to say
at the end about the empiricist alternative, and the kind of computational model
consonant with it.
Page 8 of 27
Thinking
What Turing showed was that for any step‐by‐step procedure that could be used to
calculate the value of a function for a given argument, that procedure (or algorithm)
could be captured by some Turing‐machine table. Hence, if a function is computable (p. 615)
at all—that is, if there is an algorithm guaranteed to produce a value for any argument—
there will be a Turing machine that can compute it. Thus, for example, there is at least
one Turing machine that embodies an algorithm for addition function, and at least one for
subtraction, and so on. Showing this much is an achievement dazzling enough, but Turing
also went on to show something utterly breathtaking. He proved that one single machine
—a ‘universal Turing machine’—could do anything and everything that could be done by
any simple Turing machine. The trick was to show that one could effectively code and
then list all the different special‐purpose (or ‘simple’) Turing machines; one could then
build a Turing machine that took the list positions and coded instructions as part of its
input. In this way, a single machine could, in effect, look up the simple machine that
embodied the function to be computed, and then just follow the instructions associated
with that machine.
The universal Turing machine's ability to simulate the activity of any simple Turing
machine affords it precisely the kind of flexibility that Descartes thought was beyond the
capacities of ordinary matter: the flexibility to make apposite responses to an unlimited
number of distinct problems. Moreover, the machine's flexibility is the result of the
systematicity of the symbol system that encodes its inputs, outputs, and instructions. The
Page 9 of 27
Thinking
same two symbols—‘1’ and ‘0’—are combined and recombined in rule‐governed ways to
represent a potential infinitude of meanings. Yet the machine's behaviour—no matter how
complex—is always decomposable into small step‐by‐step reactions to primitive elements
of the complex symbols.
Turing's mathematical results were quickly applied to physical reality, resulting in the
construction of concrete ‘thinking machines’: digital computers. These were not literally
Turing machines—they incorporated some architectural innovations due to John von
Neumann, for example data stores, or ‘memory’—but still realized the essential feature of
Turing's model: computation performed on structured symbols. The philosophical import
of these developments was also rapidly exploited.7 Hilary Putnam noted that the relation
between the machine table of a computer and the ‘hardware’ that implemented it offered
a useful model for representing the relation of mind to body (Putnam 1960, 1967a,
1967b). Mental states, on this model, would be abstractly characterized states of a human
brain, just as machine states were abstractly characterized states of an electronic
mechanism. Because the defining features of these abstract states were the functional
relationships among individual states, inputs and outputs, this general view of mentality
became known as functionalism.8
If we take the machine analogy completely at face value, we get the view of thinking
known as the computational model: thinking is computation. That is, a thought process
consists in a series of transitions from brain state to brain state, in a way that mirrors an
explicit inference or computation, just as the electronic states of a digital (p. 616)
computer each embody steps in an inference or computation. Jerry Fodor pressed the
analogy a step further. He argued that if thinking was computation, then there needed to
be a medium of computation—a symbolic language over which the computational
processes that constituted thought were defined. Thoughts, then, would be tokenings of
sentences in an internal language. This ‘language of thought’, he argued, was analogous
to a computer's ‘machine language’, the system of symbols to which the machine's most
basic computational operations are sensitive. Electronic computers can translate machine
language strings back and forth into more ‘user‐friendly’ programming languages by
means of special translation programs (compilers). Analogously, we can regard the
acquisition of human natural languages as the building of a translation program that
allows us to move back and forth between ‘Mentalese’ and ordinary public language
(Fodor 1975). We'll return to the idea of a mental ‘machine language’ below.
Before leaving this point, however, let me just note that Fodor's view raises a possibility
that Descartes would not have allowed; namely, that ‘dumb brutes’ might be thinkers.
Descartes, as we saw, presumed that the ability to use language tracked the ability to
think—that all and only language users possessed the faculty of reason. (This is so even
though his argument requires only that language use be evidentially sufficient for the
attribution of reason to a creature.) But if Fodor is correct, the possession of an internal
language of thought—a biologically specified ‘machine language’—is independent of the
capacity for speech. We can thus make sense of attributions of genuine thought to non‐
linguistic creatures—infrahuman animals, as well as pre‐linguistic human children; to say
Page 10 of 27
Thinking
that such creatures think is to say that their behaviour is mediated by an internal
compositional system of representation. This is indeed the position taken by
contemporary ethologists and developmental psychologists, who have evidence of
behavioural flexibility and creativity in animals and infants that might have changed
Descartes's assessments (Gallistel 1990; Carey and Spelke 1994).
Although Turing did not develop a computational theory of human thought, he did commit
himself explicitly to the proposition that a machine could—at least in principle—think.
Ironically (again), Turing endorsed the very same criterion of thinking that Descartes
appealed to in his argument for dualism (Turing 1950). Turing, like Descartes, thought
that the sustained ability to respond appropriately to a variety of circumstances was
sufficient for being a genuine thinker. And like Descartes he saw verbal interaction as
providing the crucial test. It's just that, unlike Descartes, Turing thought that this ability
could be embodied in a machine.9
It's noteworthy that Turing's test of intelligence imposes no conditions on the means by
which the machine is allowed to accomplish its deception. Most philosophers have taken
this to mean that Turing was proposing a behaviourist definition of ‘thinking’: that he was
essentially saying that thinking is whatever might turn out to be responsible for
intelligent‐appearing behaviour. If this is what Turing had in mind, it's clear that his
‘definition’ is inadequate, for it's easy enough to describe a hypothetical case in which an
obviously dumb device manages to pass the Turing test.
To see how this might work, let's consider a hypothetical machine devised by Ned Block
(1995). The machine runs a simple program, constructed in the following way. First a
giant list is made of all possible strings of symbols that could be typed by a human
interrogator within some particular period of time—say an hour. The list will include a
vast number of what we might call ‘conversations’: strings that look like transcripts of
some sensible conversation between two human beings. The next step, therefore, is for a
team of programmers to comb through the list of strings, pulling out every conversation,
and throwing away all the other strings, the ones that do not resemble actual human
Page 11 of 27
Thinking
Now the instructions to the machine are trivial. All it has to do is to match strings of
remarks with conversations on the list. It can do this step by step. When it receives its
first remark as input, it finds some conversation that begins with that remark, and then
copies the second remark in that conversation as its output. (Remember that every
remark that is intelligible and that can be typed in the space of an hour will appear as
part of some—indeed many—conversations on the grand list.) The human interrogator
will, of course, make some reasonable response to the machine's remark. The machine
now finds a conversation containing the string consisting of the first remark, the
machine's own remark, and the interrogator's next remark, and outputs the fourth
remark in that conversation. (Again, every possible three‐remark conversation will be on
the grand list.) This process is repeated until the hour is up.11
(p. 618)
Will this work? Many actual working programs operate on essentially these principles.
Think of (the now ubiquitous) phone menus. Suppose you want to renew a
pharmaceutical prescription. You dial the number of the pharmacy, and a pleasant voice
answers, welcoming you to the wonderful world of Acme Drugs, and requests that you
‘please listen carefully to your options’, adding that ‘you may enter the number of your
choice at any time’. When the voice finally gets to the option you're interested in—‘to
refill a prescription, press “4” now’—you follow the instruction: you press ‘4’. The voice
then says, ‘please enter your prescription number’, and so on and so forth. What is
actually going on, of course, is that the computer running the menu is simply using a
look‐up table like the one described above. The computer takes your typed signals as
input, and activates a particular taped message in response. When all goes well, the
whole experience is not unlike a conversation with a real person. A naive user might even
think that she actually is having a conversation with a real person.
But if you are an experienced user of such technology you know that things often do not
go well. Sometimes the menu of options is too limited, and doesn't include one that covers
your needs. In that case, the voice may just stupidly—but pleasantly!—keep repeating the
inadequate list over and over until you give up. Slightly more user‐friendly systems will,
after a round or two, have the pleasant voice announce that ‘your call is being forwarded
to an operator’ (i.e. a real person). Another problem that can arise is when your inputs
don't meet the ‘expectations’ of the program—you enter too many digits, or too few, or
you pause too long in between entering. If there is voice‐recognition software involved,
the machine may find your accent or pronunciation pattern too unlike its stored
exemplars to be able to find a match. In such cases, the machine may generate (in a
pleasant, pleasant voice!) the message ‘I'm sorry—I didn't understand. Could you re‐
Page 12 of 27
Thinking
Returning to our master machine, we can now ask what would happen if the interrogator
typed a question that was completely off the wall, or badly mistyped? Block points out
that the basic programming strategy outlined above can be both simplified and improved
(for purposes of passing the Turing test) if the programmers of our master machine keep
a real person in mind as they craft the machine's responses. Block suggests his Aunt
Bubbles: someone who apparently took a slightly dim view of academic exercises, and
who thus would be prone to respond to strange or ill‐formed questions with a throwaway
remark like, ‘My nephew told me you might try to confuse me’. The programmers could
take a few remarks such as this, and have the machine cycle through them at random
whenever it received an input that did not match any of the anticipated intelligible
remarks. This expedient not only allows the programmers to give the machine a plausible
and coherent ‘personality’, but provides them with a way of coping with inputs that might
otherwise expose the program and expose the deception (lock 1995).
Here, too, there have been working programs that run on such principles. A program
written by Joseph Weizenbaum, called ELIZA, was designed to mimic the (p. 619)
behaviour of a Rogerian psychotherapist (Weizenbaum 1976). Therapists in this tradition
are ‘non‐directive’—they do not offer their patients analysis or advice; instead they try to
facilitate their patients' self‐discovery by asking open‐ended questions prompted by
themes in the patients' own comments. This fact about Rogerian therapeutic practice
makes it ideal for mimicry by Aunt Bubbles‐type programs. After a brief opening question,
‘Why have you come to see me?’ or ‘How are you feeling today?’ the machine can let the
‘patient’ take the lead, responding with highly stereotypical responses, sometimes picking
up on words in specified grammatical positions. Here's a typical example:
Page 13 of 27
Thinking
Although literature and lore are full of anecdotes about people who stumbled on to ELIZA
unawares (and this is easier to do now in the age of email and the internet than it was back in
the 1970s) and were fooled into thinking they were communicating with a real person, the
program was pretty easy to unmask.12 But with Block's Aunt Bubbles machine we have the
guarantee of a much more effective deception. As long as the interrogation lasts no longer than
an hour, the machine will pass the Turing test. And of course, following Block's principles, we
could design a machine that would last even longer by expanding the size of the original list. But
while this machine would readily fool a human interrogator, no one who knows how it is
accomplishing the deception would say that the machine is thinking. What Block's example
shows is that we are not willing to identify thinking with whatever happens to be responsible for
apparently intelligent behaviour. We want more than just intelligent‐looking behavioural
responses—we want the responses to be caused in a certain way.
But this, of course, is just the point that Descartes emphasized in his Discourse argument.
Behaviour can appear to be the result of thinking, without there really being thinking
behind it. Notice that Block's machine is a version of the kind of physical device
Descartes thought was ‘morally’ impossible. That is, the Aunt Bubbles machine
essentially has a special‐purpose ‘organ’—in this case, a specialized subroutine—available
for every envisioned circumstance. What's important to notice is that such a machine is
not literally impossible; it is only practically impossible.13 Merely writing down the
program we've been talking about would take more time (p. 620) and material than is
available to any human being in the space of a lifetime. The amount of time it would take
a physical device to search the relevant lists would also be prohibitively large. What all
this means is that the look‐up table is not a plausible explanation of the intelligent
behaviour we observe in other human beings. While Descartes may have been wrong
about the capacities of matter in general, he was absolutely right in thinking that no
human head—indeed, no physically possible being, period—could produce the range and
amount of intelligent behaviour that we do by relying on a bag of special‐purpose tricks.
To see what's missing in our account of thinking, let's look back at the Aunt Bubbles
machine. Imagine that in place of a machine that's programmed to match remarks we
instead achieve our effect by employing a human being to sit in a room and follow the
instructions that constitute the machine's program. So our Aunt Bubbles operator will go
through the procedure we imagined the machine doing automatically: she will look up the
input strings, match them to conversations, and output the next appropriate strings. This
is, of course, a rather silly procedure for a human being to go through, given that she
could just read the input remarks and then think of appropriate replies. But now let's
change the case a bit. Imagine that the conversations, inputs, and outputs are all in
Chinese. If our operator is a monolingual speaker of English then she will not be able to
understand any of the remarks she is dealing with, but—and this is the significant thing—
Page 14 of 27
Thinking
she will still be able to manipulate the strings of Chinese characters in accordance with
the instructions, so as to simulate just as effectively as before the behaviour of an
intelligent person.
The set‐up I've just described is—except for one wrinkle that we'll add in a minute—
identical to the set‐up devised by John Searle in his landmark essay, ‘Minds, Brains, and
Programs’ (1980). His thought‐experiment, now known as the ‘Chinese Room’, was
devised to show that the limitations of machines go beyond those identified in the Aunt
Bubbles example. According to the line of thought we were pursuing, the problem with
the machine, the thing that kept it from being a genuine thinker, was the fact that it
relied on the wrong kind of program: it used a ‘stupid’ look‐up algorithm instead of a
program that exploited the compositional structure of the remarks it was manipulating.
But according to Searle, the problem is the fact that the machine is running a program,
period. A program is a set of instructions for the manipulation of symbols. Crucially, the
instructions treat the symbols as mere formal objects, things that can be identified purely
by their shapes. As far as the instructions go, the symbols may mean anything, or
nothing. Consequently, the instructions can be followed by someone—or something—who
has no idea whatsoever what the symbols mean.
(p. 621)
Much earlier we noted that thoughts have contents, that they represent or are about
things, that they possess the property philosophers call intentionality. But of course
thoughts are not the only things that can represent. Ordinary objects can stand for
things, as a red light stands for the command ‘stop’, and gestures can too, as a thumbs‐up
gesture is a symbol of approval. And, of course, words stand for things. But there's a
difference. Objects, gestures, and words, it appears, derive their intentional contents,
their meanings, from the thoughts they function to express. Because we are thinkers, we
can impose meaning on things—light fixtures, gestures, vocalizations, and squiggles on
paper—that are otherwise completely insignificant. Because of conventions adopted by
our ancestors, the word ‘horse’ is used by some of us to refer to horses; other people,
because of different conventions adopted by different ancestors, use the word ‘cheval’ for
the same purpose. The intentionality of words and other public symbols is thus a derived
intentionality. The source from which this intentionality derives is the original
intentionality of thought.
Now of course we can impose derived intentionality on anything we like, including the
symbols processed by computers. And we do. We speak of computer data as ‘code’, in
recognition of the fact that we can construe the strings of ones and zeros that we type
into the machines as representations of any information we want to store or utilize. And
we can write programs—instructions for the manipulation of code—that mirror in form
the thought processes that we want to apply to that information. That's what Turing
showed us we could do, remember: devise a mechanical device the operation of which
mirrors any thought process that can be carried out in a step‐by‐step way. But to say that
the mirroring is formal is to say that the operations are sensitive only to the formal
Page 15 of 27
Thinking
properties of the symbols, that they would work the same way no matter what the
symbols mean. Indeed, as far as the operation of the machine is concerned, the symbols
mean nothing at all, just as the Chinese symbols manipulated by our Aunt Bubbles
operator mean nothing at all to her.
This is the wrinkle that I referred to above. Go ahead, says Searle, give the person in the
Chinese Room whatever set of instructions you like, including whatever program it is you
think characterizes the competence of a Chinese speaker. The result, he contends, will be
the same. The instructions will still always treat the words of Chinese as meaningless
squiggles; the operator will always be able to follow them successfully (p. 622) without
understanding a thing. Indeed, Searle adds, we can even imagine that the operator
memorizes the program, so that her instructions are completely internalized, as is,
presumably, the program ‘running’ in the head of a Chinese speaker. The intuition
persists: our operator does not understand Chinese.
Another response made by defenders of the computational model concerns the way in
which the operator in the Chinese room interacts with the outside world—or, rather, the
way in which she does not interact with the outside world. Actual thinkers can gather
information about the external world by means of their various sense organs, and can
affect that world by moving their limbs. It's this causal interaction with the external world
that enables a thinker to attach meaning to the symbols it manipulates in thought. But
once again Searle is happy to modify the set‐up to remedy the claimed deficiencies. Give
the operator a camera, through which she can see the inputs being delivered, and let the
symbols she outputs be wired up so as to cause movements in artificial limbs. This is
analogous to attaching a computer to a robot, says Searle, and makes no difference at all
to the crucial intuition. We would have, in this case, a monolingual English speaker
causing the movements of a robot in ways that are appropriate (we may assume) to
sentences in Chinese, but not via an understanding of the Chinese characters.
But if Searle is right, that nothing we can add to the operator's situation will make the
manipulated symbols come alive with meaning, where does intentionality come from?
What makes it the case that thoughts represent things outside themselves? Searle's
answer to this question is notoriously obscure. Intentionality, he says, is a biological
Page 16 of 27
Thinking
product of the human brain. In saying this, he insists that he is not ruling out the
possibility of an artificial brain; that is, one made of non‐biological materials. Such a
‘brain’, he says, could even be made out of electronics, provided only that it possesses the
same causal powers as a human brain (1980, 1992). But it's difficult to understand what
Searle has in mind. The brain has many causal dispositions; an artificial brain would
presumably need to possess only some of these. Which causal powers would be the ones
essential to the ‘production’ of intentionality, and which ones could the designer of an
artificial brain neglect? In the case of artificial hearts, it is the causal powers that have to
do with the heart's function that we want preserved. Those that have to do with its
reaction to, say, serum cholesterol are not properties that we want preserved; indeed,
they're ones we'll take care to avoid duplicating. So it's presumably those properties of
the brain relevant to the function of thinking that are the important ones. But if we use
functional criteria to single out those ‘causal powers’ of the brain that are crucial for
intentionality, then it's not clear how Searle's view differs from the functionalist's after
all.14
But in any case there is a point about the Chinese Room thought‐experiment that we are
neglecting, one that's key to understanding how thinking might be computation after all.
Consider the fact that while the Chinese Room operator does not understand the Chinese
symbols she is manipulating, she does understand something; namely, the English
instructions. This suggests that the Chinese Room paradigm (p. 623) may not provide a
fair test of the computationalist's hypothesis. Since the operator's relation to the English
instructions she understands is obviously functionally different from her relationship to
the Chinese symbols she does not understand, it's open to the computationalist to say that
the thought‐experiment does not capture the relevant relationships between operator and
language—that whatever program the operator is running, it is not the ‘program’ that
human beings ‘run’ in the course of understanding language.
While we're at it, let's look back at the Aunt Bubbles machine. Here again, on a second
look, we have some reason to revise our original assessment of the situation. For the
machine to carry out the ‘stupid’ operations associated with the look‐up table, operations,
that is, that require no understanding whatsoever of the English strings it is receiving as
input or producing as output, it must be running some program that enables it to respond
appropriately—non‐stupidly—to the symbols in which the program is written. We have
already encountered the concept of a ‘machine language’. This is the code in which the
most basic computer commands are written; it's the symbol system that directly triggers
changes of states in the computer. Perhaps if we could get clearer on what it is for the
machine to figuratively understand its own machine language, we could get an idea what
it is, in computational terms, for the Chinese Room operator to really understand her
instructions. When we figure that out, perhaps we'll be able to characterize the
difference, in computational terms, between the operator's relationship to the English
instructions and her relationship to the Chinese characters that she does not understand.
Page 17 of 27
Thinking
Let's pretend that the Aunt Bubbles machine not only stores an enormous list of
conversations expressed in English, but that its machine language is a simplified form of
English.15 In that case, we could give the machine the following instruction: ‘If FIND
“FIND” THEN SAY “OK” ’. There are two occurrences of the English word ‘FIND’ in this
command, but they differ importantly in the effects they will produce in the computer.
The first ‘FIND’ will cause the machine to initiate a sequence of state changes that
constitute—because of the programmer's cleverness—a search routine. The second
‘FIND’, however, because it is enclosed in quotation marks, will not instigate such a
sequence. The computer's program will treat this occurrence of the symbol as a mere
object, and not as the expression of a command. That's the effect of the quotation marks:
they cordon off the symbol and present it to the computer as a meaningless, inert shape.
The fact that the shape is the same as the shape of the symbol that, when unquoted,
prompts the computer to begin a search is invisible to the computer, and irrelevant to its
operation. The machine's program uses the symbol in one case, and merely mentions it in
the second.
Functionally speaking, we can mark this distinction by saying that the unquoted symbol
possesses, for the machine, a conceptual role, a distinctive pattern of causal relations to
the computer's internal states, to its other potential inputs and outputs, and, in particular,
to the other symbols that constitute the machine's machine (p. 624) language. The quoted
symbol, on the other hand, has no such distinctive causal profile (although the quotation
marks themselves do). This, then, is the difference between the Aunt Bubbles machine's
relation to the symbols that compose its instructions and its relation to the symbols that it
is instructed to manipulate: the former, but not the latter, have a conceptual role in the
machine's functional architecture. It can use the symbols that compose its instructions,
but it cannot use the symbols that are merely mentioned in those instructions. The same
distinction can be drawn, arguably, in the case of the Chinese Room. The symbols that
compose the operator's instructions (whether written down or vocalized to her) must be
integrated into the operator's overall psychology in much the same way that the ‘FIND’
symbol is integrated into our machine's internal functioning in order for her to
understand them. The fact that the Chinese symbols are merely mentioned, and thus the
fact that they have no conceptual role within the operator's psychology, is sufficient to
explain why she does not understand them.
But is conceptual role all there is to meaningfulness? To put the question another way,
must we now say that the Aunt Bubbles machine really understands its instructions? We
do not. For there is a difference between the kind of conceptual role possessed by
symbols in a machine's machine language and the kind possessed by mental
representations in the minds of real thinkers. The difference has to do with the way in
which the symbols are connected—or not connected—to states of affairs in the external
world. In the ‘conceptual economy’ of a mere computer the internal relations among the
internal states of the machine are independent of events occurring in the external world.
All of these states are interconnected with the inputs and outputs the machine receives,
Page 18 of 27
Thinking
but which inputs the machine receives, and the effects of the outputs the machine
produces, are unrelated to anything but the intentions and interests of the person using
the machine.
In contrast, the inputs that a genuine thinker receives are information‐bearing: they co‐
vary reliably with specific external states of affairs. Similarly, the outputs produced by a
genuine thinker—typically, actions of one sort or another—have consequences in the
external world. These patterns exist independently of the will or intentions of any thinker;
they are matters of natural law. Because the conceptual roles of symbols in the internal
representational system of a genuine thinker necessarily involve a coordination of these
‘naturally’ meaningful inputs and these consequential outputs, the ‘interpretation’ of the
symbols will be naturally constrained. I can impose whatever interpretation I want on the
symbols on which a computer is operating, treating the one and the same configurational
state, perhaps, as a position in a chess game on one day and a depiction of erosion in
Iceland the next. In so far as such a state has a meaning at all, it is a meaning that derives
from my intentions to use it in a certain way. I am not similarly at liberty with respect to
the meanings of my mental symbols. DOG will mean ‘dogs’ in my psychological economy
just in case the occurrence of that symbol is connected in lawful ways, via its connections
with information‐bearing input states and consequential output states, to environmental
(p. 625) circumstances involving dogs.16 The intentionality, or meaningfulness, of my
To say that the mental symbols deployed by real thinkers possess original intentionality is
not to say that the intentional contents of the symbols affect the operations of the
computational device that realizes the thinker's mind. Fodor has argued that
computationalism is committed to a principle he calls methodological solipsism: the
principle that computational mechanisms are sensitive only to intrinsic, formal features of
the symbols they manipulate, and a fortiori not to the relational properties that determine
their meanings.17 But methodological solipsism, Fodor insists, does not entail that the
manipulated symbols have no meanings. He charges that Searle makes exactly this
erroneous inference, conflating formal operations on symbols with operations on formal
symbols. To say that the operations that a computer (or, as computationalism would have
it, a brain) performs are formal is to say that the causally relevant features of the
representations over which the operations are defined are formal features. This is just the
doctrine of methodological solipsism, and is indeed something to which computationalism
is committed. But it is no part of computationalism to say that the symbols on which the
mental operations work are formal symbols; that is, symbols devoid of meaning (Fodor
1991).
Page 19 of 27
Thinking
I will leave matters here. In doing so, I acknowledge that I leave many objections
unanswered, and many questions unaddressed. In particular, I have said nothing about
the relationship between thinking and consciousness, a lacuna many philosophers—
including both Descartes and Searle!—will find unforgivable. Let me just assert—without
argument, and in the face of many contemporary defences of his view—that Descartes's
fundamental mistakes about the capacities of matter carry over, and call into question his
arguments for the connection between thinking and consciousness. I do not mean to cast
stones at Descartes, however. My own characterization of thinking was developed on the
basis of philosophical argument (p. 626) and low‐level empirical observation—the very
kind of armchair theorizing that Descartes undertook in his own Meditations. Still, I
believe that my account comports nicely with our emerging scientific understanding of
the mind, with work in cognitive psychology, linguistics, ethology, information sciences,
and neuroscience. Just as Descartes would have expected.
Page 20 of 27
Thinking
As I remarked earlier, the conception of thought I've been presuming is the Cartesian,
rationalist conception, which treats inference as the paradigm mental process. But it
must be acknowledged that much of our mental life does not fit this conception. In
particular, a good portion of our thinking involves non‐rational trains of thought. For
example, I hear news on the radio of the Iowa caucuses. Thinking of Iowa makes me think
of my favourite musical, The Music Man, which is set in Iowa. The phrase ‘music man’
makes me think of the phrase ‘confidence man’, which reminds me of a magazine article I
read about confidence men in the mid‐nineteenth century, which makes me think of the
bill sitting next to the magazine that I need to pay, which reminds me that paying the bill
will trigger an overdraft.
In such cases as this the thoughts composing the train of thought are not bound by logical
relations among the thought contents, but rather by loose relations of some kind of
similarity. They are bound by association. Against Descartes and other rationalist
theorists, the British empiricists—John Locke, George Berkeley, David Hume in the
eighteenth century and John Stuart Mill in the nineteenth—held that all thinking,
fundamentally, had this character. In their theories they sought to characterize the
principles that explained why certain ideas would be associated with others, principles
like resemblance (Iowa caucuses and the Iowa setting, ‘music man’ and ‘confidence
man’), contiguity (the magazine and the bill), and causation (bill paying and overdrafts).
Not only did the empiricists think that associations bound thoughts together in the mind,
they held that association bound thoughts to their contents. According to these
philosophers, individual ideas got their contents through association with things in the
outside world. My idea of a dog was a dog‐idea, an idea that meant ‘dog’, because it was
an idea of the type that was regularly occasioned in my mind by perceptions of dogs.
(This view about the origin of semantic content represents an unusual point of
convergence between empiricists and contemporary rationalists like Fodor.) Because of
this idea–world connection, relations like similarity and contiguity among objects in the
world would be tracked by the ideas thinkers associated with them. In this way, they
thought, we get an account not only of thinking, but of knowledge. Properties that are
frequently co‐instantiated, and event types that are (p. 627) ‘constantly
conjoined’ (Hume's phrase) in our experience, and hence in nature, will be mirrored by
strongly associated ideas in our minds.
Associative processes do seem to be part of our mental life. Memory, for example,
appears to be organized in accordance with the sorts of principles posited by the
empiricists. And it's clear that the content of our stereotypes—mental templates that we
rely on for ‘quick and dirty’ conclusions—are built out of experience in much the way the
empiricists held that we constructed general ideas from particular experiences.
But empiricism fails as a general account of thinking in offering no account at all of the
kinds of thought processes the rationalists focused on; that is, rational processes. From
the thoughts ‘Clinton accepts money from big insurance companies’ and ‘No one who
accepts money from big insurance companies will implement real health‐care reform’ we
Page 21 of 27
Thinking
might infer ‘Clinton will not implement real health‐care reform’. Trains of thought such as
these cannot be explained just by association. The last thought certainly is similar enough
to each of the preceding thoughts for it to have been triggered by association with them—
that's not the problem. The problem is that there are any number of different thoughts
that could be equally strongly associated with the previous two, including the negation of
the last thought: ‘Clinton will implement real health‐care reform’. The relation of
association simply places no constraints on the kinds of thoughts that can be associated
with each other. (Since contiguity is one of the relations that gives rise to association, we
can easily forge an association between any two thoughts whatsoever, just by
juxtaposition.) Logic, however, does impose constraints; not just any thought follows—in
the logical sense—from any other.
The inability of empiricist models to account for logical inference is connected with the
empiricists' failure to recognize the existence, or at least the importance, of the
compositional structure of thought. Most importantly, they failed to apprehend what
Frege noticed—that the component parts of thought differ in character from each other.
This meant that they had no materials with which to build conceptual structure other
than appending and abstracting, and those are not enough.
Just as the von Neumann machine provides a natural model for the rationalistic
conception of thought, a different kind of computer, called a connectionist network, offers
a way of realizing the empiricist conception. A connectionist network consists in layers of
interconnected nodes. Each node can send signals to any node in the upper layer to
which it is connected. Every node has a firing threshold, which means that the node will
fire when the total strength of the signal reaching it (got by adding the strengths of all
the signals) reaches that threshold. Connection strengths can change over time, partly as
a function of the frequency with which a connected node fires. The first layer of nodes is
called the ‘input layer’, and the last is called the ‘output layer’. Connection strengths
between nodes can change in response to the frequency and patterns of firings from the
nodes at the level below. If and when a network reaches the point at which new inputs fail
to produce any change in connection strengths, the network is said to ‘settle’.
(p. 628)
Page 22 of 27
Thinking
References
Antony, L. (1997), ‘Feeling Fine About the Mind’, Philosophy and Phenomenological
Research, 57: 1–8.
Cheney, D. L., and Seyfarth, R. M. (1990), How Monkeys See the World: Inside the Mind
of Another Species (Chicago, Ill.: University of Chicago Press).
Page 23 of 27
Thinking
Fodor, J. (1975), The Language of Thought (Cambridge, Mass.: Harvard University Press).
—— (1991), ‘Yin and Yang in the Chinese Room’, in D. Rosenthal (ed.), The Nature of Mind
(Oxford: Oxford University Press), 524–5.
Fodor, J., and McLaughlin, B. (1990), ‘Connectionism and the Problem of Systematicity:
Why Smolensky's Solution Doesn't Work’, Cognition, 35: 183–205.
Fodor, J., and Pylyshyn, Z. (1988), ‘Connectionism and Cognitive Architecture: A Critique’,
Cognition, 28: 3–71.
Harman, G. (1982), ‘Conceptual Role Semantics’, Notre Dame Journal of Formal Logic,
28: 242–56.
Heil, J., and Mele, A. R. (1993) (eds.), Mental Causation (Oxford: Oxford University Press).
Page 24 of 27
Thinking
Putnam, H. (1960), ‘Minds and Machines’, in S. Hook (ed.), Dimensions of Mind (New
York: New York University Press), 148–79.
—— (1967b), ‘The Nature of Mental States’, in W. H. Capitan and D. D. Merrill (eds.), Art,
Mind, and Religion (Pittsburgh, Pa.: Pittsburgh University Press), 37–48.
Ramsey, W., Stich, S., and Garon, J. (1991), ‘Connectionism, Eliminativism and the Future
of Folk Psychology’, in W. Ramsey, D. Rumelhart, and S. Stich (eds.), Philosophy and
Connectionist Theory (Hillsdale, NJ: Erlbaum), 199–228.
Rumelhart, D. E., McClelland, J. L., and the PDP Research Group (1986), Parallel
Distributed Processing, i–ii (Cambridge, Mass.: MIT Press).
Searle, J. (1980), ‘Minds, Brains, and Programs’, Behavioural and Brain Sciences, 3: 417–
24.
Weizenbaum, J. (1976), Computer Power and Human Reason (San Francisco, Calif.:
Freeman).
Notes:
(3) Which is, of course, not to say that I cannot distinguish Hershey's‐kiss‐chocolate from
other kinds of chocolate, like for instance Ghirardelli‐chocolate. I'm no philistine.
(4) It's an unfortunate fact, and the source of much confusion, that this term is
homophonic with Brentano's term ‘intentionality’, especially since there's almost
complete overlap between the things that are intentional, and the things that are
Page 25 of 27
Thinking
intensional. For a comprehensive discussion of the meanings of these terms, and the
connections between them, see Crane (2001).
(5) Several of the chapters in this volume address the issue of whether qualitative mental
states such as sensations have intentional contents, so I will not consider it here.
(6) Frege may have been the first philosopher to recognize this fact, or at least the first to
understand its crucial contribution to the compositionality of thought (1892/1970).
(7) For a detailed history of the development of computation and its implications for the
study of the mind see Harnish (2002).
(8) There are different varieties of functionalism; one variety, called ‘analytical
functionalism’ makes no appeal to a machine analogy. For discussion see Block (1978).
(9) Turing's one doubt about this came, bizarrely, from his reading of the evidence for
extra‐sensory perception, which he found, ‘at least for telepathy’ to be
‘overwhelming’ (see Turing 1950: 453–4).
(10) Turing's actual set‐up is more complicated, but in ways that add nothing to the
structure of the experiment (see Turing 1950: 433–4).
(11) For a complete description of the machine, with some helpful illustrations of its
operation, see Block (1995: esp. 381–4).
(12) To visit ELIZA yourself, just google ‘Eliza computer therapist’ and go to any of the
dozens of sites that come up.
(13) Depending on the amount of time we want to allow the interrogation to run, it might
become physically impossible—we could quickly get to the point where the number of
characters needed to write the program would be larger than the number of electrons in
the universe.
(14) For fuller development of this criticism of Searle's account of intentionality see
Antony (1997).
(15) Actual machine languages are cumbersome and unperspicuous. Since simple
commands written in English can be readily translated into machine languages, though,
our pretence does no harm.
(16) I should point out that while many philosophers of mind agree that original
intentionality involves ‘natural meaning’ in some way or other, they disagree violently
about precisely what way this is. Fodor thinks that a distinctive pattern of covariance
between symbol and world is sufficient, and views the symbol's conceptual role as simply
one way that the covariance can be causally sustained (1987). Harman (1982) and Block
(1986) contend that it's the conceptual role itself that constitutes a symbol's meaning.
Both Millikan and Dretske think a symbol's meaning is determined by its function, but
Page 26 of 27
Thinking
differ in their accounts of what determines a symbol's function (Millikan 1984; Dretske
1988).
(17) Many philosophers disagree that the considerations Fodor cites really count in
favour of methodological solipsism, and contend that relational properties, like semantic
properties, can be causally efficacious. For discussion see the papers in Heil and Mele
(1993). Other philosophers, like Tyler Burge, disagree that mental states can be
individuated without reference to their intentional contents (Burge 1986; for discussion
see Egan 1991).
Louise Antony
Page 27 of 27
Language and Thought
This article aims to explore a pattern of reasoning about language and thought that
seems to virtually guarantee a distortion of what precisely constitutes thinking. The
guiding idea is a simple one, an idea officially embraced by many philosophers, but an
idea the implications of which are too easily neglected. The article states a belief in the
thesis that something functions as a representation only in so far as it is given a use by a
representing agent. This thesis — that representation requires a representing agent —
applies in equal measure to non-linguistic ‘pictorial’ and to linguistic representations.
Keywords: language and thought, representation, representing agent, linguistic representations, conscious
thinking, non-linguistic thoughts
A dog believes his master is at the door. But can he also believe his master will
come the day after tomorrow?
36.1 Introduction
FOR human beings thought and language go hand in hand. But must they? This is most
often put as a question concerning the relation of language to thought. Could thought (or
‘genuine thought’) occur in the absence of language (or ‘genuine language’)? Could a
creature who lacked a language (a ‘genuine language’) entertain thoughts (‘genuine
thoughts’)? Even if thought of a rudimentary sort is possible without language, are there
kinds of thought that are thinkable only linguistically, thinkable only in language?
Page 1 of 20
Language and Thought
Before venturing into deeper waters, a word on terminology. I take the question whether
thought requires language to boil down to the question whether conscious thinking, or
some conscious thinking, is inevitably linguistic. And I take ‘linguistic’ thinking to involve
thoughts ‘in’ some natural language: English, Portuguese, Urdu. I do not mean to exclude
the possibility that some thinking is unconscious or ‘subpersonal’. This is a topic to which
I shall return in Section 36.8. Meanwhile, the focus will be on conscious thought and
reflection.
What of the language of honey bees, computer languages, Ockham's sermo interior,
Fodor's ‘language of thought’ (1975)? These raise interesting questions, but I shall set
(p. 632) them aside here. The plan is to concentrate on the clear case of natural language
and see where this leads. It might then be possible to make sense of honey bees,
computing machines, and the language of thought.
My broader aim is to explore a pattern of reasoning about language and thought that
seems to me virtually to guarantee a distortion of what precisely constitutes thinking. The
guiding idea is a simple one, an idea officially embraced by many philosophers, but an
idea the implications of which are too easily neglected. What I have in mind is the thesis
that something functions as a representation only in so far as it is given a use by a
representing agent. Let me hasten to add that I shall employ ‘agent’ in a broad sense that
encompasses not merely intelligent human beings but intelligent systems generally. In
this sense a digestive system—the ‘brain in the gut’—could constitute an intelligent
system.
It is easy to lose sight of the imagistic character of conscious thought when we start
philosophizing about language and thought. We suppose that although there is no
problem at all as to how interior utterances could mean what they do, there is a special
problem as to how pictorial images could have determinate meanings. Philosophers on all
sides agree that meanings are not built into representations. Representations acquire
significance by being put to use by representing agents. The apparent fact that pictorial
representations lack ‘intrinsic meaning’ is simply a reflection of a wholly unsurprising
fact about representations generally. Interior utterances—verbal images—are no less
bereft of intrinsic meaning than images of sugarplums.
Page 2 of 20
Language and Thought
Regarded in this light, discussions of language and thought take on a new colouring.
Thinking—conscious thinking—is not the having or entertaining of images, verbal or
otherwise. Thinking is a matter of an agent's using such images, putting them to work.
And whatever it is to put images to work, it is not solely a matter of entertaining further
images. Nor is thinking something occurring behind the scenes when you deploy
representations: it is the deployment of those representations—in your head or otherwise.
Your thinking the cat is on the mat might be a matter of your saying to yourself ‘the cat is
on the mat’ or it might involve your conjuring a visual image of a cat on a mat. In either
case, however, you do not do two things: (1) entertain an image; (2) think. Thinking
occurs in your deploying the image. This point about use was made forcefully by
Wittgenstein, but it has been endorsed by philosophers as (p. 633) different as Locke and
Andy Clark. Although philosophers routinely pay lip service to the idea, its implications
have not been widely appreciated. Or so it seems.
This is not an argument but the statement of a theory for which Davidson supplies
elaborate support. I think it best to read Davidson not as denying that mute creatures are
capable of representing their surroundings mentally and acting on those representations,
but as contending that there is a stark discontinuity between representations deployed by
such creatures and the kinds of representation found in agents equipped for higher‐order
representation (see Bennett 1964). This is why, according to Davidson, ascriptions of
beliefs, desires, and intentions to mute creatures are at best figurative and at worst flatly
misleading.
You can see what Davidson is driving at by noting that ascriptions of propositional
attitudes to mute creatures are ‘semantically transparent’. Spot barks at an intruder
lurking beneath the window. Which of the following is true:
Page 3 of 20
Language and Thought
Such questions seem puzzling, according to Davidson, not because we lack ready access to
Spot's inner life but because, so long as Spot remains incapable of expressing his inner state
linguistically, there is no fact of the matter as to which of these characterizations is correct.
Genuine belief ascription is ‘semantically opaque’. Substitution of co‐referring terms in such
ascriptions can alter their truth‐value. The belief that there is an intruder beneath the window
differs from the belief that there is someone beneath the window. Thoughts about particular
objects or states of affairs are thoughts about objects or states of affairs under some description.
Attributions of states of mind to Spot, in contrast, identify those states of mind wholly by
reference (p. 634) to objects on which they are directed. Because Spot lacks a language, there is
no question of his representing a given object, the intruder, say, under one description rather
than another.1
Davidson encapsulates all this in a slogan: only a creature possessing the concept of
belief could have beliefs. A creature could be furry without possessing the concept of
being furry; a creature could fly without possessing the concept of flight. Why should the
mere possession of beliefs require possession of the belief concept?
Davidson is thinking about it this way. Beliefs can be true or false. Possession of the
concept of belief includes possession of a conception of something—a representation—
that purports to fit the facts but could fail to do so. This is the concept of a truth bearer.
Such a concept could be deployed only by a creature in a position to ascribe beliefs. Why?
Consider your place in the world. As you move through the world, you come to represent
your surroundings in various ways. In so far as you are observant, changes in your
surroundings will be mirrored by changes in your representations of those surroundings.
You represent the water in a pond as deep. You subsequently step into the pond and
discover it is shallow. You now represent the pond as shallow. None of this requires you to
distinguish what you believe to be the case from what is the case.
A slightly more tendentious way to put this point is to note that a concept of truth comes
into play only when you are in a position to distinguish truth bearers—representational
states with definite, truth‐evaluable contents—from truth makers.2 This amounts to the
idea that your possession of a truth concept requires possession of a capacity to
represent representations and their contents—requires, that is, that you be capable of
higher‐order representation. What is striking about Davidson's argument is his contention
that your deploying representations with definite contents requires your possessing a
truth concept, hence (if the reasoning above is sound) your being in a position to
comprehend higher‐order representations. This is something you do when you ascribe
beliefs to others.
You invoke higher‐order representations in reflecting on your own beliefs, but Davidson's
idea seems to be that in the absence of an opportunity to ascribe representational states
of mind to others agents are in no position to ascribe such states of mind to themselves.3
Page 4 of 20
Language and Thought
Given just the world and your experiences of it, you have no use for a concept of
representation, no grounds for a distinction between the world as it is and the world as
you represent it as being.
These are deep waters, but pretend for the moment that something like this is right: the
concept of representation finds purchase only when you are in a (p. 635) position to
ascribe representational states to others. Once the concept takes hold, you are in a
position to extend it to your own case. Indeed, first‐ and third‐person recognition go hand
in hand.4
What has any of this to do with our having beliefs, ordinary first‐order representations of
states of affairs? Here the argument becomes, if anything, even denser. Imagine a solitary
creature who represents the world in particular ways. The creature's representations are
dynamic: they ‘update’, reflecting changes in what the creature encounters. What are we
to say about the contents of such a creature's representational states of mind? Return to
Spot's belief about an intruder outside the window. We saw that there are many ways to
characterize the content of this belief. Davidson would say that there is no reason to
prefer any of these characterizations to any other: there is no fact of the matter as to how
the creature represents its world. We can grant that the creature is in a state that
changes in ways that reflect changes in the world, and we can grant that the creature's
behaviour is guided by this state, without supposing that the state has any definite
content. It's ‘content’, such as it is, can be described indifferently so long as it includes
the pertinent object. Because the creature has no use for distinctions of the sort
appearing in (1)–(4), the distinctions play no role in the creature's mental life.
Suppose this were right, or at least that it is not obviously wrong. The final move is to tie
a capacity for language to the capacity for higher‐order representation. Let us grant that
an agent's entertaining higher‐order representations requires that the agent be in a
position to ascribe representations with definite contents to other agents who are
themselves ascribers. How is this to be brought off, Davidson asks, without the kind of
definiteness we find in language? Sentences (1)–(4) differ, and so are apt for the
ascription of distinct representational states, states possessing determinate contents. The
idea is that only a linguistic representation could be a representation of a
representational state with a determinate content.
Page 5 of 20
Language and Thought
There you have it. The possession of representational states with determinate contents
requires a capacity for higher‐order representation. Higher‐order representation requires
possession of a language. So the possession of states with definite contents requires
possession of a language.
You are likely to remain unmoved by all this, but the aim is not to convince you of
Davidson's view, only to convince you that the view is not entirely mad. Many (p. 636)
philosophers have been attracted to the idea that there is something special about the
capacity for higher‐order representation and to the thesis that a capacity for higher‐order
representation requires a capacity for language. Philosophers have been attracted to
such views for reasons that appear to have nothing to do with Davidson's arguments.
What is important here is that Davidson does make explicit one line of reasoning to a
popular, but unevenly defended, conclusion.
I shall argue that all this is off the mark, but before going on the offensive it will be useful
to look at another, independent line of argument that appeals to higher‐order thoughts en
route to the conclusion that certain prominent species of thought must be linguistic in
character.
To see how Bermúdez's argument works, consider our capacity for what Andy Clark calls
‘second‐order cognitive dynamics’:
Page 6 of 20
Language and Thought
thinking about the conditions under which we think best and trying to bring them
about. The list could be continued, but the pattern should be clear. In all these
cases, we are effectively thinking about our own cognitive profiles or about
specific thoughts.
The characteristically human capacity for complex thoughts about thoughts could thus be tied to
our capacity for language. This, according to Clark, is due not to the distinctiveness of the
contents of linguistic thoughts but to the way they are embodied: as mental utterances or
inscriptions of sentences. Imagine a particular state of affairs, the cat's being on the mat. This
state of affairs might be represented pictorially or sententially. Your thought about the cat's
being on the mat could take the form of (p. 637) an image of a cat on a mat or an interior
utterance of a sentence ‘The cat is on the mat’. In either case, I have suggested, you make use of
imagery: the ‘vehicle’ of your thought might be a visual image of a cat on a mat or a verbal
image—what Bermúdez calls an ‘imaged sentence’ (2003: 160).
Against this background you can see Bermúdez's argument proceeding in two stages.
First, he notes that thoughts on which second‐order thoughts are directed (‘target
thoughts’) ‘must be at the personal level’ (2003: 159); they must be kinds of thought
concerning which we could be conscious. This apparently uncontroversial proviso is
meant to set to one side Fodor's language of thought, sentences of which are not
consciously available, hence evidently unsuitable as ‘targets’ for higher‐order thoughts.
The language of thought aside, might there be kinds of thought possible only for
creatures possessing a natural language? Trivially, thoughts about linguistic
representations (that include a grasp of the significance of those representations) are
possible only for creatures possessing a language.5 The interesting question is whether
there might be other kinds of thought that would require linguistic competence.
Bermúdez's contention is that there are. Clark's ‘second‐order cognitive dynamics’ would
be possible only for creatures linguistically endowed.
This is stronger than the thesis defended by Clark. Clark argues that it is a contingent
fact about human beings that much of our reasoning involves consideration of thoughts
expressed sententially. Bermúdez, in contrast, holds that certain kinds of thought are
flatly impossible in the absence of language:
There are certain types of problem that we solve by manipulating mental images
and exercising the visual imagination. And we are, of course, conscious of bodily
sensations, emotional feelings, and other such qualitative states (although these
are not properly described as types of thinking at all). But we are not, I think, ever
conscious of propositional thoughts that do not have linguistic vehicles.
(2003: 160)
Page 7 of 20
Language and Thought
Next Bermúdez advances an ‘argument by elimination’ to the conclusion that in the case of such
thoughts ‘the only available vehicles at the personal level are public language sentences’ (2003:
159). Representational ‘vehicles’ can be of two sorts. On the one hand, ‘representation might be
secured symbolically through the complex symbols of a natural language’. In this case ‘a thought
would be represented … through its linguistic expression and would appear as an object of
thought qua linguistic entity’. On the other hand, ‘representation might be secured in an
analogue manner, through some kind of pictorial model’ in which ‘the vehicle of thought is a
pictorial representation of the state of affairs being thought about’ (2003: 160).
Although first‐order thoughts can be ‘vehicled’ pictorially, such thoughts, Bermúdez
contends, cannot serve as objects of second‐order thoughts. This is a surprising claim,
given that this is exactly what anyone does who reads Bermúdez's (p. 638) colourful
accounts of non‐linguistic thoughts.6 Consider your own case. In the right circumstances
you can entertain a ‘pictorial’ thought of the cat's being on the mat and you can reflect—
as you are now reflecting—on that thought. Bermúdez is after bigger game, however.
Representation, he contends, evidently requires the holding of a ‘structural isomorphism’
between representation and represented state of affairs. There are, however, at least two
senses in which a representation could be said to share a structure with what it
represents.
Page 8 of 20
Language and Thought
Consider cognitive ‘maps’ of the sort discussed by David Braddon‐Mitchell and Frank Jackson
(1996) and consider a ‘canonical example of second‐order cognitive dynamics’: recognizing ‘the
evidential basis for a particular belief and then evaluating the inferential transition made on that
basis’ (Bermúdez 2003: 161).
It is perfectly easy to see how there could be some very basic forms of inferential
transition between maps. Such transitions might be modeled on broadly
associationist lines, and it is the possibility of such transitions that enables maps
to serve as guides to action. What is not possible, however, is for such transitions
to be understood and evaluated in terms of either deductive validity or
probabilistic support.
Such evaluations require the interpretation of the maps ‘in broadly propositional
(p. 639)
terms’. Thus, ‘we must interpret one map as expressing one proposition and the second as
representing a further proposition, and then evaluate the inferential relations … between these
two propositions’. The problem is that ‘our only understanding of how to do this rests on the two
propositions being linguistically formulated’ (Bermúdez 2003: 162).
What of ‘mental models’ (see e.g. Craik 1967; Johnson‐Laird 1983, 1999; Rips 1994)?
Mental models have been proposed as alternatives to ‘syntactic’ models in accounts of
reasoning. The idea is that given an argument from premises to conclusion we ‘construct’
models of the premises and judge that the conclusion is implied when the conclusion is
true in each of these models.
Mental models, Bermúdez argues, are constructed on the basis of ‘sententially encoded
propositions’. The theory
In this regard,
Bermúdez's contention is not that agents do not use mental maps and models, nor that agents
cannot think about such things.
Page 9 of 20
Language and Thought
The point is that we cannot use mental maps or mental models for thinking about
thoughts in the manner demanded by second‐order cognitive dynamics. Natural
language sentences are the only proxies that will permit thoughts to function as
the objects of thought in this manner.
This makes it clear that, despite suggestions to the contrary, Bermúdez's argument applies only
to certain kinds of higher‐order thought: thoughts pertaining to evidential and inferential
relations among representations. Thus construed, the thesis is that we can make sense of
evidential and inferential reasoning only when the relata are sentential in form.
In summarizing his conclusion, however, Bermúdez endorses a stronger thesis: ‘[T]here
can be no intentional ascent without semantic ascent. We think about thoughts through
thinking about the sentences through which those thoughts might be expressed’ (2003:
164). This makes it sound as though higher‐order thought, quite generally, requires a
capacity for linguistic thought. In what follows I shall offer a challenge to those
sympathetic to either a weak or a strong reading of Bermúdez's (p. 640) contention. The
message is simple: where cognition is concerned, there is nothing special about
language.7
Here are two reasons you might think that thought requires language. First, persuaded
by Fodor, you might imagine that thoughts must themselves be linguistic. To think is to
engage in something like inner speech: the medium of thought is linguistic.8 Non‐verbal
imagery lacks the kind of determinate content required for genuine thought. Second, you
might want to connect the having or entertaining of thoughts with actual or possible
manifestations of those thoughts. You might have verificationist motives here: ‘an “inner
process” stands in need of outward criteria’ (Wittgenstein 1953/1968: sect. 580). Or you
might simply take it to be part of the nature of a thought that it manifest itself in
particular ways in particular circumstances. An unmanifestable thought would be one the
possession of which made no difference to what its possessor does or might do. This,
Page 10 of 20
Language and Thought
coupled with the idea that the only unambiguous manifestation of thoughts—or some
thoughts—is linguistic, implies that only linguistically endowed creatures could have
thoughts—or thoughts of certain kinds.
Consider this second reason more closely as it applies to Spot. The idea is that thoughts
require manifestations (actual or possible), and only linguistic utterances could count as
unequivocal manifestations of tensed thoughts. Spot could, the day after tomorrow, look
expectantly at the door. This, however, would be most naturally taken to be a
manifestation of Spot's belief that his master is at the door, not a manifestation of a
tensed belief.
Perhaps, however, the problem is not that Spot lacks a capacity for the linguistic
manifestation of tensed beliefs, but that Spot lacks a use for tensed representations.
(p. 641) One sort of use of such representations would be the conveying of tensed
The suggestion on the table is that the idea that Spot is linguistically challenged hence
unable to entertain thoughts about temporally remote states of affairs is a red herring. To
be sure, Spot lacks a language. But it is not this that makes it unlikely that Spot could
entertain tensed thoughts. Spot lacks a use for tensed representations, hence a capacity
for tensed thinking.9 One prominent sort of use might be communicative—a fact that
points to language. What is doing the work here, however, is not language, but use.
You might be sceptical that there could be anything like a tensed representation in the
absence of language. If you thought that, you could accept what I have said about use,
but note that in many cases, including Spot's, only linguistic use is available. Maybe. But
maybe not.
36.5 Thought
Suppose we take seriously the venerable idea that language is a tool. Intelligent agents
employ language in the way you might employ a screwdriver to achieve various ends. If
language is a tool, it is a multi‐purpose tool, a Swiss Army knife. (To be sure, even a
‘single‐purpose’ tool—a screwdriver, for instance—can have multiple uses.) If this is right,
it is a mistake to regard thoughts as being in language. Thinking is linguistic when we
think with language. If a linguistic episode—an inner utterance, for instance—is
analogous to a screwdriver, thinking is analogous to driving a screw with a screwdriver.
Regarded in this light, the question whether thought, or some thought, requires language
reduces to the question whether language might be the only tool that could be used for
certain tasks. And just as it seems unlikely that any tool is irreplaceable, so it seems
unlikely that language is, for any particular task, irreplaceable.
Page 11 of 20
Language and Thought
What exactly is it to deploy language as a tool? And what other intellectual tools might be
available to intelligent agents? Just as you make use of utterances—written or spoken—to
articulate ideas to others, you can use inner utterances in the articulation of ideas to
yourself. You can talk through a problem, recall some earlier event, or plan a course of
action by listing steps to its completion in your head. In these cases inner utterances are
not manifestations or copies of thoughts; you are thinking with language just as you
might open a can with a can opener. There is no mystery here any more than there is
mystery in your spontaneously engaging in intelligent conversation. You need not first
think—rehearse—then speak. This is so whether speech is overt or covert.
Inner utterances (I say, siding with Bermúdez) are a species of mental imagery, where the
images are images of what their audible, visual, or tactile counterparts (p. 642) sound,
look, or feel like. There is no logical or conceptual gulf between linguistic (‘propositional’)
imagery and imagery of other sorts. Conscious thought in general is imagistic.10
Not all thoughts incorporate linguistic imagery, however. Much of our thought involves
non‐linguistic visual, auditory, tactile, or olfactory imagery. Indeed, your thought about a
particular person might include verbal imagery (an inner utterance of a name, for
instance) accompanied by a visual image of the person and perhaps other imagery as
well.
The association of imagery with thought is not a matter of reducing thoughts to images.
Thinking is a matter of using imagery. Bare images, what C. B. Martin (1987) calls the
materials of thought, are devoid of significance. You can grant this, grant that images
themselves lack intrinsic intentionality, without thereby giving up on the idea that
conscious thinking is a matter of deploying images. There is nothing special about verbal
imagery—what you might conceive of as linguistic thought. Interior utterances are no
different in this regard than any other imagery. Intentionality arises when images (or
signs, or representations generally) are put to work by intelligent agents. Without use,
images or signs are empty; severed from use, representations fail to represent.
It is easy to lose sight of this simple idea so long as we persist in the illusion that interior
utterances are not themselves images altogether on a par with images of sunsets and
romantic Greek isles. Debates over whether imagery is ‘pictorial’ or ‘propositional’ are
beside the point (see Kosslyn and Pomerantz 1977; Pylyshyn 1981, 1984; Rollins 1989;
Tye 1991). Linguistic thoughts are as ‘pictorial’ as any others in the sense that they
include the use of imagery: verbal imagery.
Page 12 of 20
Language and Thought
feels (did feel, will feel, or might feel), tastes (did, will, or might taste), sounds (did, will,
or might sound), or smells (did, will, or might smell). One species of such imagining is
verbal: you imaginatively utter, or hear yourself utter, words.
In considering such cases we philosophers are apt to be misled in at least two ways. First,
we are likely to regard verbal imagery as privileged. On the one hand, there is ordinary
‘pictorial’ imagery; on the other hand, there is the entertaining of ‘sentential’ thoughts.
This is the kind of distinction Peter Carruthers (1996) appears to have in mind in
distinguishing ‘imagism’ and ‘sententialism’. An ‘imagist’ regards (p. 643) thoughts as
invariably pictorial. ‘Imagists’ hold that ‘thoughts inherit their representational
properties from the representative powers of the images that constitute
them’ (Carruthers 1996: 32).
Carruthers mentions Locke, Hume, and Russell as proponents of imagism. ‘Imagists’ hold
that thinking consists of a succession of picture‐like ideas parading across the mind.
Carruthers contrasts ‘imagism’ with ‘sententialism’: the view that thoughts, or some
thoughts, are linguistic in character. ‘Tokens’ of such thoughts are mental sentences,
which are to be distinguished sharply from ‘pictorial’ images. But what are mental
sentences if not images—verbal images? The contrast here, if there is one, is not between
images and non‐images, but between two species of image: one non‐sentential, one
sentential. There is no sense in which verbal images—mental ‘sentence tokens’, interior
utterances or inscriptions—are less imagistic than nightmares or the remembered taste
of a madeleine.
A second way in which we philosophers are likely to be misled in thinking about verbal
and non‐verbal imagery is more pernicious. Consider a passage from Carruthers:
An image will always carry with it excess content beyond mere entailment
relations. The sentence, ‘A cat is on a mat’ carries no more, and no less, content
than the proposition [that a cat is on a mat]. So it will contain the content, [that a
mammal is on a mat], since this is entailed, but it will not contain the content,
[that a tabby cat is on a mat], nor the content, [that a cat is sitting on a mat]. Not
so with an image. Although an image can be in various respects indeterminate or
vague, it must always be more determinate than a proposition. My image of a cat
on a mat must always be an image of some particular kind of cat (fluffy or short‐
haired, tabby or black) in some particular position (sitting or standing, facing or
turned away) on some particular sort of mat.
Page 13 of 20
Language and Thought
Berkeley's own account of general ideas begins with the observation that images
inevitably exhibit a kind of definiteness (Carruthers's ‘excess content’) lacked by
linguistic representations. A pictorial representation of a cat on a mat will be a
representation of a cat sitting or lying on a mat, for instance. This, Carruthers thinks,
makes it ill suited as a vehicle for the thought that the cat is on the mat. But Berkeley—
and Locke—note what Carruthers does not: generality is a product of our deployment of
images. An image of a reclining tabby could be used to represent a cat's being on a mat.
Think of road signs depicting a car skidding on a wet road. The image is of a particular
kind of car, but sign‐makers count on our grasping the image as representing vehicles
generally.
(p. 644)
The point applies across the board: the key to understanding the nature of thought is the
recognition that thinking is something we do when we use imagery. Thinking is not the
having or entertaining of images, sentential or otherwise. Thinking is something we do
imagistically.
36.7 Proto‐language
Philosophizing is a largely verbal enterprise. It is scarcely surprising, then, that when we
engage in philosophical reflection—when we think about philosophical topics—we do so
in a thoroughly verbal idiom. There is nothing special about words, however, nothing
magical about utterances and inscriptions. Words take on significance when they are put
to use by intelligent agents or intelligent systems. If it is doubtful that Spot could
entertain the thought that his master will arrive home the day after tomorrow, this is not
because Spot lacks words to express this thought. It is because it is hard to see how Spot
could have a use for a representation with this content.
(1987: 277)
Page 14 of 20
Language and Thought
He has also noticed that they eat different things at different times. When he
catches a fish, he opens its stomach to see what it has been feeding on, so that he
can use it as bait.
On one unsuccessful day's fishing he notices an approaching storm that looks like
spoiling the fishing for a long time. Frustrated, he intends not to return to the
fishing hole until the weather changes. He picks up his fishing gear and starts for
the cave. He happens to frighten a mink eating a fish. His curiosity overtaking
him, he opens the stomach to see what the (p. 645) fish had eaten and takes out
some grasshoppers. This is a procedural action whose projected outcome is
information about the past. It also has the point of finding out what would have
helped him to catch fish if he had used it as bait.
Page 15 of 20
Language and Thought
The idea that non‐conscious thought might be imagistic is not something I want to pursue
here. I do not want to pursue it because I believe there is a better way of understanding
the nature of at least some non‐conscious thought. Let me begin with an example. If you
are like me, you can sometimes resolve nagging intellectual problems by sleeping on
them. You go to bed puzzled, wake up enlightened. Something similar can happen when
you are trying to remember a forgotten name: you put it out of your mind and discover
that it subsequently comes to you. How should we understand such cases?
One possibility is that something is going on backstage altogether analogous to what goes
on when we consciously reflect. Call this the shoemaker‐and‐elf model: when the
shoemaker retires for the night, elves materialize, performing just as the shoemaker
would, but secretly.
Another possibility is that our unconscious mental life differs structurally and
qualitatively from its conscious counterpart. Think of the brain as a complex, dynamic
dispositional system. The system manifests itself in myriad ways: in bodily (p. 646)
behaviour, in speech, and, often enough, in conscious thought. The system manifests itself
as well in the regulation of various bodily functions, thermal regulation, the maintenance
of equilibrium, and the like. The finely tuned, focused dispositions constituting the
nervous system are not static but dynamic: they evolve continuously over time. The
system that, a few minutes ago, was unable to produce the name of an acquaintance has
resettled itself and can now, in concert with other, reciprocal, systems, yield the name.
36.9 Conclusion
What is the upshot? First, thinking—ordinary conscious thinking—is imagistic. This is so
for ‘linguistic’, or ‘sentential’, thoughts as well as for patently non‐linguistic thoughts.
Second, thinking does not consist in merely having or entertaining images, but in putting
those images to work. Third, verbal and non‐verbal imagery are representationally on a
par. The significance of any sort of representation lies in the use to which it is put by
intelligent representers. Finally, what we regard as non‐conscious thought need not
resemble conscious thought occurring out of sight. Non‐conscious thought could turn out
Page 16 of 20
Language and Thought
Creatures lacking a capacity for language could well be unable to entertain thoughts of
various kinds. But this might be so, if it is so, not because thoughts or some kinds of
thought are necessarily linguistic. An inability to use language and an inability to think
certain thoughts could have a common cause. A creature could lack the kind of make‐up
required for the pursuit of projects the satisfaction of which would afford opportunities
for the deployment of representations of particular sorts. This, I think, puts language in
its place.12
References
Angell, J. R. (1897), ‘Thought and Imagery’, Philosophical Review, 6: 646–51.
Davidson, D. (1975), ‘Thought and Talk’, in S. Guttenplan (ed.), Mind and Language
(Oxford: Clarendon), 7–23;
repr. in Davidson, Inquiries into Truth and Interpretation (Oxford: Clarendon, 1984), 155–
70.
Page 17 of 20
Language and Thought
repr. in E. Lepore and B. P. McLaughlin (eds.), Actions and Events (Oxford: Blackwell,
1985), 473–80.
Heil, J. (1992), The Nature of True Minds (Cambridge: Cambridge University Press).
Kosslyn, S. M., and Pomerantz, J. R. (1977), ‘Imagery, Propositions, and the Form of
Internal Representations’, Cognitive Psychology, 9: 52–76.
Pylyshyn, Z. W. (1981), ‘The Imagery Debate: Analogue Media Versus Tacit Knowledge’,
Psychological Review, 88: 16–45.
Rollins, M. (1989), Mental Imagery: On the Limits of Cognitive Science (New Haven,
Conn.: Yale University Press).
Page 18 of 20
Language and Thought
Notes:
(1) This is not to say that there might not be more‐ or less‐perspicuous ways of describing
Spot's states of mind, only that pertinent factors fall well short of those constraining
belief ascription to creatures endowed with a language.
(2) The concept of a truth maker and the concept of a truth bearer are correlative
concepts: grasp of one requires grasp of the other.
(4) So it would be misleading to imagine that you reason from the third‐ to the first‐
person case. The relation is, rather, reciprocal.
(5) The parenthetical rider is important. A cat might entertain thoughts about your
utterances without having any idea what they meant or even that they have meaning.
Your thoughts about sentences in an unfamiliar tongue might be no better in this regard
than the cat's.
(6) Perhaps what Bermúdez has in mind is that in reading about them we do not
represent non‐linguistic representations but merely linguistic descriptions of non‐
linguistic representations. His discussion of non‐linguistic representations purports to be
a discussion of the representations themselves, however, and not merely descriptions of
them. In discussing such representations Bermúdez's own thoughts must be directed on
them.
(7) I do not deny that, as a matter of contingent empirical fact, human beings lean heavily
on language. My interest is in the question whether what we accomplish by means of
language could only be accomplished through the use of language.
(8) Although the discussion focuses on spoken language, I intend for it to encompass
other forms of linguistic expression—signing, for instance. Vonnegut (1973: 58–9)
imagines an interesting possibility.
(10) This, perhaps, is what lies behind the old debate about the possibility of ‘imageless
thought’. See, for instance, Angell (1897), Woodworth (1906), Danzinger (1980), Thomas
(1999, 2001).
(11) Perhaps such creatures produce, as well, silent imagistic counterparts of these
utterances!
Page 19 of 20
Language and Thought
(12) The central thesis developed here was impressed upon me twenty years ago by C. B.
Martin; see Martin (1987, 2008). Since then I have learned much about imagery from
Nigel Thomas (see his 1999, 2001) and Mark Rollins (see his 1989). I am grateful to José
Bermúdez, and to audiences at the University of Otago and the University of Geneva who
suffered through preliminary versions of this chapter.
John Heil
John Heil: Professor of Philosophy, School of Philosophy and Bioethics, Building 11,
Monash University VIC 3800. AUSTRALIA. And: Professor of Philosophy, Washington
University in St Louis, Department of Philosophy, Wilson Hall, One Brookings Dr.,
Campus Box 1073, St. Louis, MO 63130–4899. USA. Email:
john.heil@arts.monash.edu.au; jh@wustl.edu
Page 20 of 20
Consciousness and Reference
Russell was the first to articulate the connection between consciousness and a
recognizably modern conception of reference, using his notion of acquaintance. The idea
of ‘acquaintance’ is the idea of a kind of epistemic contact with a thing or property. Many
theorists would agree that reference to something requires a particular kind of epistemic
contact with it. Yet you might agree that the notion of ‘epistemic contact with a
particular’ is needed to characterize reference, while arguing that epistemic contact is to
be explained as possession of a body of propositional knowledge.
Page 1 of 17
Consciousness and Reference
subject and the things experienced; and given such a conception, we can make visible the
link between consciousness and reference.
Russell was the first to articulate the connection between consciousness and a
recognizably modern conception of reference, using his notion of acquaintance. The idea
of ‘acquaintance’ is the idea of a kind of epistemic contact with a thing or property. Many
theorists would agree that reference to something requires a particular kind of epistemic
contact with it. Yet you might agree that the notion (p. 649) of ‘epistemic contact with a
particular’ is needed to characterize reference, while arguing that epistemic contact is to
be explained as possession of a body of propositional knowledge. Minimally, you might
say that what it takes to have epistemic contact with b is this: you must have a piece of
propositional knowledge whose content can be specified using a term referring to b, such
as ‘b’. There are many ways in which to pursue the programme of explaining ‘epistemic
contact with b’ in terms of propositional knowledge relating to b. For instance, you might
introduce the notion of a ‘dossier’, as follows. Suppose you have two pieces of knowledge
whose contents are given by: ‘b is F’, and ‘c is G’. On the face of it you are not yet in a
position to argue that anything is both F and G, even if b and c are in fact identical. To
draw that conclusion you need a further premise, the identity statement ‘b is identical to
c’. Suppose, in contrast, that you have two pieces of knowledge, ‘b is F’, and ‘b is G’. Here
it looks as though you are in a position immediately to draw the conclusion, ‘b is both F
and G’. As we might put it, in the second case, but not in the first, you are in a position to
trade on the identity of the object referred to in those two pieces of knowledge. You might
have a body of information relating to the same object b, and be in a position to trade on
the identity of the object referred to in any two of the pieces of knowledge in that body. In
such a case we can say that you have a ‘dossier’ on the object. So you could expand on
the initial characterization of ‘epistemic contact with an object’ by saying that it requires
a dossier of information on it, rather than just a single piece of propositional knowledge.
There are a number of conditions you might think it important to impose on the body of
information in the dossier. For example, Evans held that the sheer volume of detail is
important, as are the reasons why you are interested in the referent in the first place.
Kaplan suggested that a dossier should constitute a major part in a narrative concerning
those ‘who fill major roles in that inner story which consists of all those sentences which
[the subject] believes’ (1971: 136). There are other distinctions we might apply. The
important point about all these variations and refinements is that they are variations and
refinements internal to the project of explaining the notion of epistemic contact with an
object or property in terms of possession of propositional knowledge about it.
Russell had a dramatic alternative to any such approach. Russell argued that our
knowledge of things cannot, in general, be explained in terms of our knowledge of truths.
Russell thought that there were two sorts of knowledge: knowledge of truths, and
knowledge of things. Knowledge of truths depended on knowledge of things. In particular,
it depended on the kind of knowledge of things he called acquaintance:
Page 2 of 17
Consciousness and Reference
(1912: 46)
(1912: 46)
What the subject is acquainted with fixes the references of simple referring terms, and that
reference fixer is also the basis of the subject's knowledge of truths.
You might try to interpret these remarks in terms of causal chains that fix reference to
physical objects. Perhaps those very causal chains are the source of knowledge of truths
about the objects. The ‘right kind’ of causal chain to fix reference would be what
Sainsbury (1979), for example, called an ‘epistemic chain’. When we ask what an
epistemic chain is, the natural answer is that it is one which produces knowledge of
truths concerning the object referred to. So the relevant notion of a causal chain is being
explained in terms of propositional knowledge. This loses Russell's idea of acquaintance
as a knowledge of things that is more fundamental than knowledge of truths.
One way of trying to sustain a causal approach here is to shift away from the focus on
proper names, and emphasize the arguably more basic case of demonstratives referring
to currently perceived objects. And we could think of epistemic causation not as
but rather as
(b) a relation between the object and a body of non‐conceptual contents, of the kind
appealed to in scientific analyses of vision.
Since these non‐conceptual, perceptual contents will be more basic than propositional
knowledge, this approach might seem to give us a way of sustaining the idea of a knowledge of
things more basic than knowledge of truths. The first and still the most sustained development
of this approach is Evans's Varieties of Reference (1982).
This approach deviates from Russell, however, in giving no role to consciousness in
providing acquaintance with objects. For Russell, awareness of the object is central to
Page 3 of 17
Consciousness and Reference
We shall say that we have acquaintance with anything of which we are directly
aware, without the intermediary of any process of inference or any knowledge of
truth.
(1912: 25)
This idea is lost by an explanation of acquaintance as epistemic causation, where the relata of
the causal relation are the thing referred to and a body of non‐conceptual information. There is
no reason in principle why non‐conceptual information should (p. 651) be conscious; in fact, the
information‐processing contents appealed to by vision scientists are typically remote from
consciousness. The point is quite explicit in Evans's development of this line of thought.
Consciousness is seen as an epiphenomenon that emerges once the whole apparatus of thought
and reference has been imposed on a more primitive information‐processing system:
[W]e arrive at conscious perceptual experience when sensory input is not only
connected to behavioural dispositions in the way I have been describing—perhaps
in some phylogenetically more ancient part of the brain—but also serves as the
input to a thinking, concept‐applying and reasoning system; so that the subject's
thoughts, plans and deliberations are also systematically dependent on the
informational properties of the input. When there is such a further link, we can
say that the person, rather than just some part of his brain, receives and
possesses the information.
(1982: 158)
Page 4 of 17
Consciousness and Reference
constituents. Second, it is hard to see how we can hold on to the idea that experience is
even relevant to providing knowledge of reference. One can grasp propositions in many
ways: by believing them, by desiring that they be true, and so on. Visual experience has
now simply been added to the list, as one among many ways in which one can take an
attitude towards a proposition. Why vision should be particularly central among those
ways seems now to defy explanation. And even if there is something special about vision,
it is hard to see why it is specifically experience that is important; presumably one could
in principle, on this approach, grasp a proposition as the content of non‐conscious vision.
Finally, these problems evidently arise because this approach abandons Russell's idea of
finding a kind of acquaintance with things that is more fundamental than knowledge of
truths.
(p. 652)
Page 5 of 17
Consciousness and Reference
the sensation of green have in common? And in what are they different? Moore said that
we have to recognize there are two components bundled together in the ordinary notion
of ∼‘sensation’:
[There are] in every sensation two distinct terms, (1) ‘consciousness,’ in respect of
which all sensations are alike; and (2) something else in respect of which one
sensation differs from another. It will be convenient if I may be allowed to call this
second term the ‘object’ of a sensation: this also without yet attempting to say
what I mean by the word.
We have then in every sensation two distinct elements, one which I call
consciousness, and another which I call the object of consciousness.
(1903: 444)
The analysis here does not appeal to the idea of ‘mental paint’ at all, whether representational
or sensational. The idea is that all experiences are the same so far as (p. 653) their intrinsic
properties go; they are all acts of consciousness, and it is this aspect of the experience that
eludes introspection. It is for this reason that it is easy to be a materialist and deny the very
existence of consciousness:
[When] we refer to introspection and try to discover what the sensation of blue is,
it is very easy to suppose that we have before us only a single term. The term
‘blue’ is easy enough to distinguish, but the other element which I have called
‘consciousness’—that which sensation of blue has in common with sensation of
green—is extremely difficult to fix. That many people fail to distinguish it at all is
sufficiently shown by the fact that there are materialists. And, in general, that
which makes the sensation of blue a mental fact seems to escape us: it seems, if I
may use a metaphor, to be transparent—we look through it and see nothing but
the blue
In an intriguing discussion of Moore, Van Cleve (in press) puts it very strongly when he says that
‘Moore denies that experiences have intrinsic features. Instead, they owe everything they are to
their relation to objects … Moore's radical view … could perhaps be cited as the explanation of
[transparency]: we are not aware of any intrinsic features of experience because there aren't
any’. As the above passage makes evident, Moore does think there is something elusive in
consciousness: the relation of consciousness itself, which is common to all sensation, that
materialists miss. But Moore emphatically does make the point that there is no reason to think
there are intrinsic features of experience that differentiate the experience of blue from the
experience of green. There is no need to appeal to either the notion of a representation of colour
differentiating the experiences or the notion of an intrinsic sensational feature of the experience
differentiating the two colour experiences. The objects, blue in one case, green in the other,
adequately differentiate the experiences.
Page 6 of 17
Consciousness and Reference
The concluding paragraph of Moore's ‘Refutation of Idealism’ is worth quoting quite fully,
as it brings out that his case against the idealist uses the point that the objects of
awareness are material objects:
When, therefore, Berkeley supposed that the only thing of which I am directly
aware is my own sensations and ideas, he supposed what was false; and when
Kant supposed that the objectivity of things in space consisted in the fact that they
were ‘Vorstellungen’ having to one another different relations from those which
the same ‘Vorstellungen’ have to one another in subjective experience, he
supposed what was equally false. I am as directly aware of the existence of
material things in space as of my own sensations, and what I am aware of with
regard to each is exactly the same—namely that in one case the material thing,
and in the other case my sensation does really exist. The question requiring to be
asked about material things is thus not: What reason have we for supposing that
anything exists corresponding to our sensations? but: What reason have we for
supposing that material things do not exist, since their existence has precisely the
same evidence as that of our sensations? That either exist may be false; but if it is
a reason for doubting the existence of matter, that it is an inseparable aspect of
our experience, the same reasoning will prove conclusively that our experience
does not exist either, since that must also be an inseparable aspect of our
experience of it.
(1903: 453)
This notion of Moore's, that we have non‐propositional awareness of objects, which may be
sensations or physical objects, provides exactly what we need to fill out the (p. 654) view that
awareness of objects is what makes thought about them possible. The reason it has to be
awareness is that awareness is a generic relation between the thinker and the object: there is no
such thing as a particular type of awareness without the object being there to differentiate that
exercise of awareness from any other. Since awareness has this relational character, there is no
question of being able to substitute for it some other relation that would be ‘just as good’, which
is the problem that comes up if we think of awareness as a kind of monadic glow with which the
mental life is enlivened.
Page 7 of 17
Consciousness and Reference
Arguably the same point applies to properties, or at any rate some properties. To
experience the shape of a solid object you must have some capacity to recognize (p. 655)
manifest sameness of shape across movements by you or by the object. Otherwise it is
hard to see how you could be said to be encountering the property of three‐dimensional
shape at all. But this capacity has its limits. Particularly if you consider a large object
with a complex shape, it seems entirely possible that you could encounter the shape in
one way from one angle, and then, coming upon the object from a quite different
direction, be unable to recognize the sameness of shape. You have some ability to keep
track of sameness of shape across variation in perceptual presentation, but it has its
limits. And this variation in your experience of the property matters, in that it will affect
the inferential behaviour of the shape concepts you use to report the situation observed
on the two encounters. It will affect, for example, whether you can immediately draw
Page 8 of 17
Consciousness and Reference
conclusions about whether the object has changed shape. Again, this is difficult to
reconcile with Moore's radical transparency. If the full characterization of your
experience of shape is given by saying that you bear the generic relation of consciousness
to a particular three‐dimensional shape, then it is not possible for there to be different
experiences of the same shape.
This is a fundamental problem. The response usually given in the analytic tradition is to
say that when an object is experienced in different ways this is a matter of different
representations being associated with the two ways of experiencing the object. It has
proved quite difficult to think of an alternative to this approach. Following hints from
Frege, the different representations are usually taken to be different descriptions, or
clusters of descriptions. So, for example, you might explain the difference between your
two encounters with Venus by saying that in the one case the associated description was
‘the brightest star in the morning sky’ and in the second case the associated description
was ‘the brightest star in the evening sky’. The problem being addressed here is to
characterize the difference between the consciousness of an object that provides your
knowledge of the reference of ‘the morning star’, and the consciousness of an object that
provides your knowledge of the reference of ‘the evening star’. The strategy is to appeal
to a difference in the descriptions associated with the consciousness of the object on
those two occasions. And, in particular, there is a difference in the general terms used in
those descriptions: the difference between ‘morning’ and ‘evening’.
Moreover, this approach takes grasp of predicates to be more primitive than the
awareness of objects that provides our knowledge of the references of singular terms. In
particular, it supposes that the use of predicates in quantified expressions such as
descriptions is more basic than the awareness of objects that provides knowledge of the
references of singular terms. The argument of the logical atomists was that a grasp of
quantified propositions depends on an understanding of atomic propositions: there must
Page 9 of 17
Consciousness and Reference
be singular terms as basic as predicates. Suppose we have two such basic singular terms:
‘logically proper names’. Suppose these two singular terms refer to the same concrete
object, and that you cannot in inference trade on their identity of reference. And suppose
that in both cases it is awareness of the object that provides knowledge of reference of
the term. There must then be a difference between the awareness that provided
knowledge of the reference of the one term and the awareness that provided knowledge
of the reference of the other term. But this difference cannot be explained by appealing
to different predicates that characterize the two modes of awareness of the object. For we
agreed that there are no predicates more primitive than those names. There must, then,
be some other difference between the two different ways of experiencing the object.
Finally, we have to bear in mind that there can be differences in one's awareness of an
object or property consistently with manifest sameness of the object or property: we do
have a capacity to keep track of objects or properties across phenomenological variation
in our experience of them. So some phenomenological variation is consistent with
manifest sameness of the object. On a descriptivist approach to characterizing the
awareness of the object that provides knowledge of the reference of a term, it is hard to
see how to draw the right distinctions here. There must be variation in the associated
descriptions, to acknowledge the phenomenological variation associated with turns of the
head or momentary occlusion of the object. There must be similarity or some systematic
correspondence in the associated descriptions, to underpin the manifest sameness of the
object across these variations. It is not easy to see how to specify which similarities of
description matter for manifest sameness of reference.
Russell himself developed an alternative approach to the problem. Rather than appealing
to a difference in type of representation associated with different ways of experiencing
one and the same object, he held that ultimately reference is to objects that we can
experience in only one way. So we keep the idea that we should characterize
consciousness in Moore's terms, as the holding of a generic relation between the self and
an object. But we should restrict the range of the relation. We should keep it restricted to
objects of which the subject automatically has comprehensive knowledge. This in effect
was the solution adopted by Russell and Moore, when they talked of direct awareness as
a relation between the subject and a sense‐datum. Even when Moore liberalized the
notion to material objects, he tended to talk of such items as ‘the front of a chair’, of
which the subject might be held to have comprehensive (p. 657) knowledge. This kind of
solution has not been widely endorsed in the analytical tradition, but I think that it does,
in effect, live on in the phenomenological tradition. When theorists talk about ‘the
ontology of the lived world’ of a subject, I think that what they have in mind is an
ontology of objects each of which is comprehensively given to the subject, so that issues
about partial knowledge or the informativeness of identities simply do not arise within the
ontology of the lived world.
The problems here have to do with the difficulty of explaining the relations between these
comprehensively known objects and the partially known objects in terms of which we
ordinarily think. The ‘lived chair’ of which we have comprehensive knowledge seems a
Page 10 of 17
Consciousness and Reference
quite different kind of object from the everyday chair, that has a history unknown to most
of its users, and all kinds of forgotten objects under the cushions. The problems of
explaining what the ‘lived chair’ and the ordinary chair have to do with one another are
so great that you really may wind up embracing a kind of idealism, and abandoning the
ordinary chair altogether.
Page 11 of 17
Consciousness and Reference
You and I might have an argument about the best view to be had of San Francisco. You
think the view of the whole peninsula from Twin Peaks can't be beaten, I prefer the view
from across the bay, of the town's tallest buildings floating above the fog. What notion of
a ‘view’ are we using? What is a view? To characterize a ‘view’ you say what things are in
the scene. And you say from where they are being viewed. This characterization does not
somehow miss out the crucial thing: the ‘mental paint’ that is induced in the spectator, or
the ‘representations’ the spectator is supposed to form. You have told the whole story
about the view when you have specified what is being seen and where it is being seen
from. We can think of this kind of characterization as using the three‐place relation of
experiencing an object from a standpoint that we need in addressing the problem of
partial awareness.
The notion of a standpoint must encompass more than merely the position of the
observer, but to make explicit the conditions on an account of what is included we have to
step back a little. Our aim is to characterize a notion of knowledge of things more
fundamental than our knowledge of truths. When you have this knowledge of a thing, that
constitutes your knowledge of the reference of a term referring to the (p. 658) thing. The
notion of a standpoint comes in because you can have knowledge of one and the same
thing from different standpoints. Earlier I remarked that Frege in effect proposed that
the issues here could be structured around the notion of informativeness. In the basic
cases we are considering, your understanding of two co‐referential terms t1 and t2 is
provided by your having knowledge of the thing referred to. Suppose your understanding
of t1 is provided by your experience of the thing from standpoint X, and your
understanding of t2 is also provided by your experience of the thing from standpoint X.
That is constitutive of your understanding the identity statement ‘t1 is identical to t2’ as
uninformative; as an instance of the logical law of identity. Understanding the terms in
this way, you have the right to trade on identity in inferring from ‘t1 is F’ and ‘t2 is G’ to
‘something is both F and G’. In contrast, suppose your experiencing the object from
standpoint X provides your understanding of t1, and your experiencing the object from
some quite different standpoint Y provides your understanding of the co‐referential term
t2. This constitutes your understanding the identity statement ‘t1 is identical to t2’ as
informative; it is not merely an instance of the logical law of identity. You do not have the
right to trade on identity in an inference from ‘t1 is F’ and ‘t2 is G’ to ‘something is both F
and G’. To reach that conclusion using those premises you have to add a further premise,
Page 12 of 17
Consciousness and Reference
‘t1 is t2’. The points in this paragraph are all laid down in advance of any substantive
description of the notion of a ‘standpoint’. These points set out the basic conditions that
have to be met when we do give a substantive characterization of what it takes to be
observing an object from one standpoint or another.
Suppose you want to characterize the view that someone currently has of the Taj Mahal.
We say which thing it is and which person is in question. Then to describe the standpoint
explicitly we have to say which sensory modality is involved; and that will determine what
further factors we have to fill in. For example, suppose the modality is vision. Then we
need, further, position, but also the relative orientations of the viewer and object, how
close the viewer is to the object, whether there is anything obstructing the light between
them, and so on. In the case of hearing, a rather different set of factors would be
relevant: not just which object was in question, but what sounds it was making, and the
obstruction of light would not be to the point, though the obstruction of sound would be.
We do not usually spell out all these conditions, though we are perfectly capable of
articulating them when they are important in particular cases.
We will have to keep in mind the dynamics of the experience; that experience is typically
temporally extended. We should not think of experience of an object over time, as it
moves, or as you move around it, as a matter of having a series of (p. 659) momentary
views of the thing from different standpoints. It may be that if you move from position S1
to position S2, keeping your eye on the thing, it is manifest in experience that this is one
and the same thing: the visual demonstrative t1 that you use at S1 is manifestly referring
to the same thing as the visual demonstrative t2 that you use at S2, and the identity ‘t1 is
t2’ is uninformative. Nonetheless, it may also be that if you were placed initially at S1 and
then a moment later at S2 you would not be able to formulate such an uninformative
identity, because you had not kept your eye on the object in the meantime and so could
reasonably wonder whether you were encountering the same object again. The dynamics
of experience—which things you are keeping track of over time—have to enter into the
characterization of your standpoint on a scene.
As I said, in practice we do not need to make all the relevant parameters here explicit,
because we are able to imagine how things are from standpoints other than our current
standpoint. In performing this exercise you can use the fact of the similarity of your visual
system to the other person's visual system, without having to make explicit what all the
relevant points of similarity are. So your knowledge of the other person's interests and of
Page 13 of 17
Consciousness and Reference
the scene may be enough to allow you to determine, by imagining the other person's
perception, the saliencies of the scene.
In carrying out this exercise of articulating the notion of a standpoint we show how we
can characterize a way of experiencing an object without appealing to either the idea that
‘ways’ are characterized by associated representations or the idea that ‘ways’ are
characterized by the idiosyncracies of the mental paint involved. Rather than either the
idea of an intervening level of mental representation or the idea of an intervening level of
mental paint, we can simply appeal to the notion of experience as a three‐place relation
between an object, an experiencer, and a standpoint. In this way we can do justice to
Russell's notion of acquaintance as a knowledge of things more basic than knowledge of
truths.
37.5 Revelation
Why should we pursue Russell's idea that consciousness, or experience of the world, has
a role to play in explaining our knowledge of reference? Intuitively, experience of things
provides us with grasp of what they are. Experience of the world reveals something to us,
provides us with some understanding of what is there.
It is not difficult to make sense of the idea of an animal that can be said to represent its
environment, even without ascribing consciousness to the animal. In the case of honey
bees, for example, you might feel quite sure that they are representing the locations of
targets to one another yet not confident about whether they are conscious. Such an
animal is certainly representing the affordances provided by the objects around it.
Without consciousness, though, it is hard to see how the animal could have any grasp of
the categorical objects and properties that ground those affordances (Campbell 2002).
This intuitive idea seems naturally applicable to demonstratives (p. 660) referring to
perceived concrete things, such as ‘that woman’ or ‘that tree’: experience of the things
provides knowledge of what is being talked about. The idea also seems appealing when
we consider the names of colours: knowledge of the references of colour terms seems to
be provided by experience of the colours.
Russell provided a canonical text that dominates current discussion of the idea that
acquaintance reveals some aspect of the world to us. However, he does not focus on the
notion of a categorical property. I want to close by remarking that despite the frequency
with which it is cited, this passage is currently quite misunderstood; and anyhow it does
not provide the most promising way of pursuing the idea that acquaintance reveals what
is out there. Russell said:
The particular shade of colour that I am seeing … may have many things to be said
about it … But such statements, though they make me know truths about the
colour, do not make me know the colour itself any better than I did before: so far
Page 14 of 17
Consciousness and Reference
(1912: 47)
Russell's view here is that one naturally does take and should take one's visual
experience as of, e.g. a canary yellow surface, as completely revealing the intrinsic
nature of canary yellow, so that canary yellow is counted as having just those
essential and intrinsic features which are evident in an experience as of canary
yellow.
(1997: 138–9)
The idea is now interpreted to be that experience reveals something of the nature or essence of
the world to us. It is in this strong sense that experience provides us with knowledge of what we
are referring to: it provides us with knowledge of the nature or essence of the thing. Most
writers have taken it that the idea here is not just that if X is the nature, or essence, of the
referent then experience provides you with knowledge of X. Rather, experience has to provide
you with knowledge that X is the nature, or essence, of the referent.
To review the situation. Russell argued that our knowledge of truths depends on
something more fundamental, our knowledge of things. Knowledge of things is provided
by acquaintance with—that is, direct awareness of—those things. Currently many writers
are trying to explain the intuitive notion of direct awareness here as a matter of the
subject having propositional knowledge of the nature, or essence, of the referent. This
way of explaining the idea immediately has far‐reaching implications. For example, in a
careful recent discussion Byrne and Hilbert (forthcoming) propose that the doctrine of
revelation, as applied to colour terms, should be regarded as the conjunction of two
theses. First, that if it is in the nature of the colours that p, then after careful reflection
on colour experience it seems to be in the nature of the colours that p. Secondly, if after
careful reflection on colour experience it seems to be in the nature of the colours that p,
then it is in the nature of the colours that p. As Byrne and Hilbert point out, this doctrine
immediately threatens physicalism about the colours. If it is in the nature of the colours
to (p. 661) be physical‐reflectance types, for instance, then by this doctrine of revelation it
should seem after careful reflection on colour experience that colours are physical‐
reflectance types. But no such thing is true; you could reflect on colour experience as
carefully as you liked for as long as you liked without it seeming that colours had any
such physical nature. Similarly, Lewis considers formulating revelation as a thesis about
colour experiences, or perhaps about the colours themselves. The idea here is that
revelation is the doctrine that each type of colour experience has its own essence, E, and
that having the experience provides you with propositional knowledge to the effect that
this type of experience has the essence E:
Page 15 of 17
Consciousness and Reference
Some philosophers think that each sort of colour experience has a simple,
ineffable unique essence that is instantly revealed to everyone who has that
experience. When I was shown the crayon mark and told that it was magenta (and
I believed what I was told, and it was true) straightway I knew all there is to know
about experience of magenta. I knew that it was the experience with the simple,
ineffable, unique essence E. And that is all there is to it. (Or perhaps it is the
colour magenta itself that has the simple, ineffable, unique essence that is
instantly revealed to each beholder, or anyway to each beholder with normal
visual capacities in normal light.)
And, as Lewis remarks, this view is inconsistent with materialism, at any rate when materialism
is formulated as the view that colour experiences and colours (and, presumably, everything else)
have physical essences (1997: 338).
The trouble with this whole exegetical line is that Russell's comment is being interpreted
as a remark about the relation between experience and propositional knowledge of the
essences or natures of colours. This misses the point that Russell's remark was about
knowledge of things, rather than knowledge of truths. Acquaintance with the colours is
not a matter of possessing propositional knowledge about them. It is a matter of having
knowledge of the thing, not knowledge of truths about natures or essences. To say that
knowledge of the thing is complete is not of itself to deny the possibility of there being
further propositional knowledge to be had to the effect that this thing has certain
essential features. These further essential features, of which we have propositional
knowledge, may indeed be physical, or of some sort quite unsuspected by the naive
observer.
Page 16 of 17
Consciousness and Reference
References
Byrne, A., and Hilbert, D. R. (forthcoming), ‘Color Primitivism’, in R. Schumacher (ed.),
Perception and the Status of Secondary Qualities (Dordrecht: Kluwer).
Johnston, M. (1997), ‘How to Speak of the Colors’, in A. Byrne and D. R. Hilbert (eds.),
Readings on Color, i (Cambridge, Mass.: MIT Press), 137–76.
Kaplan, D. (1971), ‘Quantifying In’, in L. Linsky (ed.), Reference and Modality (Oxford:
Oxford University Press), 112–44.
Lewis, D. (1997), ‘Naming the Colors’, Australasian Journal of Philosophy, 75: 325–42.
John Campbell
Page 17 of 17
Memory
The psychological study of memory has made exciting advances. In the last two decades
neurophysiology has given us insight into what happens in the brain when we remember.
First, the biology of nerve cells is much better understood; since signalling by nerve cells
is altered by experience, these cells might be the elementary devices of information
storage. At the higher level of brain systems, researchers have identified mechanisms
that make possible various memory functions. One task for neuropsychology is to unite
the results at the level of nerve cells and brain systems. One moral for the rest of us is
that while we often speak of a unified capacity — memory — responsible for much that is
distinctive in human life, we are really talking about a range of capacities, variously
realized in the brain and nervous system.
Keywords: memory, neurophysiology, information storage, brain systems, memory functions, neuropsychology,
nervous system
38.1 Introduction
THE psychological study of memory has made exciting advances. In the last two decades
neurophysiology has given us insight into what happens in the brain when we remember
(Squire and Kandel 1999). First, the biology of nerve cells is much better understood;
since signalling by nerve cells is altered by experience, these cells might be the
elementary devices of information storage. At the higher level of brain systems,
researchers have identified mechanisms that make possible various memory functions.
One task for neuropsychology is to unite the results at the level of nerve cells and brain
systems. One moral for the rest of us is that while we often speak of a unified capacity—
Page 1 of 17
Memory
memory—responsible for much that is distinctive in human life, we are really talking
about a range of capacities, variously realized in the brain and nervous system.
Likewise, in the last two decades episodic and working memory have continued to be
subjects of important and productive research (Tulving 1993; Baddeley 2001; Baddeley,
Conway, and Aggelton 2001). And cognitive psychology has given us insight into
reconstructive processes responsible for simple recall and autobiography. Emphasizing
the reconstructive nature of memory focuses attention on sources of memory distortion, a
topic with great social importance (Schacter 1995).
It is also an exciting time for the philosophy of memory. Old issues are enlivened by fresh
research, and new issues emerge as conceptual connections are forged at the heart of
philosophy of mind, language, metaphysics, and action theory.
There is no logical impossibility in the hypothesis that the world sprang into being
five minutes ago, exactly as it then was, with a population that ‘remembered’ a
wholly unreal past.
Russell goes on to claim that the hypothesis is, like all sceptical hypotheses, uninteresting.
Whether Russell is right or not about the interest of scepticism, recently realism about the past
has presented the more important philosophical problem.
Realism about the past implies that statements about the past are either true or false,
whether or not we can confront evidence for them. Either I was in Tanner Library on 17
November 1994 or I wasn't. Moreover, it seems we can understand such a claim despite
the fact that there is no evidence now for or against. So realism about the past presents a
problem for those who hold that understanding the meaning of a sentence requires being
able to confront the conditions under which the sentence is true (Dummett 1969). The
problem for such a broadly speaking verificationist theory of meaning is that the past is
unrevisitable. (In this way, the past is not a foreign country.)
One response for the verificationist (Green 2001) is to reconsider the kind of evidence
that memory can provide. Our memories can after all give us evidence that we had
evidence in the past. I remember the general fact of being a student at the University of
Michigan in 1994. So I have evidence now that on 17 November 1994 either I had
evidence that I was in Tanner Library or I had evidence I wasn't. It is only on such
memorial grounds, the verificationist might argue, that we meaningfully assert
statements of the form P or not‐P about the past.
Page 2 of 17
Memory
Dummett suggests that such a manoeuvre is still ‘repugnant’ (2002: 28), as it conflicts
with our intuition that what makes statements about the past true is independent of
whether we have now any grounds whatsoever for asserting them. Modifying his earlier
view, Dummett (2002) rejects radical verificationism, in favour of the view that a
statement about the past is true if a suitably placed observer could have confronted
evidence for it, and our knowing as much is central to our understanding such
statements.
Whether verificationists can meet the challenge that realism presents, the burden is not
solely theirs, for the point is that our common‐sense realist convictions about the past
need active defence. We must say something about how it is that we can even understand
past‐tense statements, such as ‘Yesterday, it was raining’, if the truth of such statements
transcends our ability to confront evidence for them. Christopher Peacocke (1999, 2001)
argues that memory plays a constitutive role here: what it is to understand past‐tense
statements is to have information of a kind that only a (p. 665) capacity for being causally
affected by past events—that is, memory—could provide. (Compare: our understanding of
spatial statements such as ‘Memorial Church is ten yards from here’ requires information
about the spatial layout of the world that only perception could provide.) On Peacocke's
view, a defence of realism about the past begins with the fact that memory is one of
several information systems (Evans 1982) and the fact that the content of memory beliefs
is externally individuated (Burge 1979). John Campbell (1994) argues on similar lines that
realism about the past is a presupposition of our ordinary conception of memory as an
informational system. For Campbell, however, this result about memory comes within a
larger story about the conceptual requirements for self‐consciousness. Campbell argues
that self‐consciousness requires one to grasp one's own causal and temporal structure:
one must be able to think of oneself as having internal states that cause other internal
states as well as change in external objects, and of events as occurring at particular
times. Thus, self‐consciousness demands one have more powerful timekeeping
sensitivities than some animals possess—one cannot simply be sensitive to time's passage
in cycles or phases. Self‐consciousness, on Campbell's view, requires all the capacities
needed for autobiographical memory.
We will see more about proposed links between autobiographical memory and self‐
consciousness below. For now, note that Campbell's broadly Strawsonian programme
shifts emphasis from the role of causation in an analysis of what memory is (Martin and
Deutscher 1966; Deutscher 1989) to the role the concept of causation plays in our
capacities for particular kinds of memory. New research in developmental psychology on
the role of causal concepts in childhood amnesia and in cognitive ethology on the role of
timekeeping in episodic memory (Hoerl and McCormack 2001) points to strong bonds
between memory and capacities for causal and temporal thought.
Page 3 of 17
Memory
The dominant view of memory in contemporary philosophy of mind and cognitive science
is indirect realism centring on the notion of a ‘memory trace’ or ‘engram’ (Schacter
1996). But the notion of a trace is more than simply that of a representation—a trace is a
permanent, discrete information store, providing a causal link (p. 666) to the past
perceptual experience, whose staying power explains or justifies one's remembering.
Opponents of the notion of a memory trace object to the independence of the trace from
processes of remembering (Christensen and Kornblith 1997) and point to the way
remembering relies on information left in the external world (Clark and Chalmers 1998;
Dennett 2000), as well as to the fact that what we call a memory is a joint product of
memory trace and cue. A memory is not simply an ‘activated engram’ (Schacter 1996).
Some have tried to reply to these concerns by revising the core notion, claiming a
memory trace is a ‘distributed representation’ in a connectionist network, where traces
are ‘dynamic’ and contribute to a ‘composite, superimposed memory
representation’ (McClelland and Rumelhart 1986: 193; see also McClelland 1995; Sutton
1998). The fact that ‘the units that participate in representing one episode or event also
participate in representing other episodes’ (McClelland 1995: 70) is supposed to be a
strength of connectionist models in accounting for memory distortion, but critical tests of
the superiority of connectionist models await development (Ratcliff and McKoon 2000).
Moreover, it is unclear just what a ‘superimposed’ representation is, and how, since both
memorial representation and trace are ex hypothesi changed with every new stimulus,
particular experiences can be preserved with any accuracy. (It is important not to lose
sight of how accurate memory can be, despite the potential for distortion.) Further
clarification is needed if we are to assess the connectionist alternative.
Does memory store a distinctive sort of ‘non‐conceptual content’ (Evans 1982)? The
question is vexed, in no small part because the debate can be cast in terms of the
existence of two kinds of content, informational versus conceptually articulated, or in
terms of whether it is appropriate to ascribe contentful states to creatures who lack the
conceptual resources to articulate or grasp the content. McDowell (1994) holds that the
ascription of content to creatures without relevant conceptual capacities is only
pragmatically appropriate; in reality computational processes in such creatures are
purely syntactic as opposed to semantic or contentful. Stalnaker (2003) takes a more
Page 4 of 17
Memory
liberal view, arguing that while certain states or activities require conceptual resources,
the content that characterizes all kinds of activities from perception and memory to
speech acts can be uniformly understood in information‐theoretic terms. Michael Martin's
argument (1992) for non‐conceptual content begins with the fact that one might fail, for
lack of the relevant concepts or through inattention, to form a belief at the time of
perception, but later one's memory of what one perceived makes new beliefs possible. So,
Martin claims, one's initial experience must be characterized by non‐conceptual content,
which is preserved by memory and provides the raw materials for later belief. It is not
clear whether Martin's argument is decisive, however, since it seems to rest on the claim
that the authority of memory derives from its being the ‘re‐presentation’ of an earlier
experience. McDowell might respond that memorial experience is not the mere re‐
presentation of earlier experience, but a new experience, with newly constructed content,
built out of more basic raw materials. The debate between conceptualists and non‐
conceptualists may not yet be settled, (p. 667) but Martin's insight is a good one: a more
careful understanding of memorial experience may hold the key to a resolution.
In his study of multiple‐personality disorder Ian Hacking (1995) details how nineteenth
century memory sciences became sources of treatment for psychological illness. One
upshot of Hacking's analysis is a limited but deeply important indeterminacy in the
personal past, not at the level of events—Hacking is a realist about past events—but at
the level of human actions. Briefly, his argument is this: all intentional action is action
‘under a description’ (Anscombe 1959); in remembering intentional action one's
description employs present concepts, which may have been unavailable to one as an
actor in the past. Since it is not determined in the past which concepts one will have in
future, one's agential past is likewise not fully determined, but is a story to be told in the
present. Hacking's conclusion is not just Freud's that past events may be given new
meaning as new conceptual resources make reconstrual in recollection possible, but the
more radical claim that one's actions themselves are sensitive to reconstrual. In a sense,
then, actions remembered may not have occurred, because the descriptions under which
they were chosen are not the descriptions under which they now make most sense. If
Hacking is right about this curious sensitivity to reconstrual of personal agential pasts,
the door is opened to important normative questions about what persons should be, and
what psychology should do for them: should persons have coherent pasts, and how should
the therapist help in constructing a narrative past for the patient, since this construction
helps to shape the facts about what the person did? (See also Engel 1998.)
Page 5 of 17
Memory
To grasp how these connections emerge we must start with episodic memory. Of all the
kinds of memory feat of which humans are capable, none is as gripping as (p. 668)
memory for past episodes of one's life. I remember the fall I took skating when I was
eight, and how many stitches were needed for the gash. Exactly what distinguishes my
episodic memory of the fall from other kinds of memory, such as my semantic or
declarative memory that ice is hard? (See Schacter and Tulving 1994; Schacter et al.
2000.)
As noted above, Campbell (1994) argues that episodic memory is bound up with a host of
conceptual capacities required for being self‐conscious. Dokic (2001) argues that episodic
memory is an experience whose content is that of a past experience, where the content is
preserved through a continuous causal chain or ‘information link’. Although one need not
be aware of this causal link, one must be aware of the memory as having its source in
one's own past. Thus, according to Dokic, one must be capable of thought about oneself,
or what we might call ‘reflexive’ thought. Moreover, according to Dokic, this sort of
reflexive thinking is only possible if one can think metarepresentational thoughts—
thoughts involving mental representations that have mental representations as their
objects. (Whether reflexivity requires metarepresentation is open to question; we shall
see a specific alternative account of reflexive thought presently.)
Others argue that the common‐sense conception of episodic memory involves ‘direct
acquaintance’ with the past through ‘mental time travel’ or ‘reliving’ the past. How are
we to make sense of such notions? Michael Martin (2001) makes an interesting stab,
arguing that in episodic memory what is retained is an apprehending of an event, where
the apprehending is itself an episode, and a conscious experience. We might find this
perplexing: how can an experience, as opposed to a representation of the experience, be
retained in memory? Martin replies that ‘memory can be experientially the same as
perception through being the representational recall of such an experiential
encounter’ (2001: 270). In other words, remembering counts as the retention of an
experience—as reliving the past—because memorial representations are of the original
encounter, and experiential sameness of memory and percept consists in linked
intentional content.
Page 6 of 17
Memory
Martin's focus on ‘mental time travel’ owes much to the work of the psychologist Endel
Tulving. Tulving's understanding of ‘mental time travel’, however, is less focused on the
notion of direct apprehension of the past, and more on an array of conceptual capacities
he takes to be required for genuine episodic memory.
In the early seventies Tulving's seminal research was framed by a distinction between
semantic and pure‐episodic memory, or memory for events. Soon enough focus on
memory for events was refined to a concern specifically for memory of events as they
occur in the course of one's own life history. Tulving's most recent work expands on this
theme:
(2001: 270)
(p. 669) Now conceived as fully autobiographical, episodic memory is characterized by: (a) a
capacity for thought about specific times and places of events; (b) a grasp of time as continuous,
forming a framework in which the specific events of one's life have a position; (c) a capacity for
self‐consciousness (i.e. ‘autonoetic awareness’); and, finally, since the relevant notion of the self
is the notion of a creature with a future as well as a past, who plans future events as much as
recalls past events, (d) a capacity for agency (see also Wheeler et al. 1997; Moscovitch 2000).
In (a) and (b) Tulving's claims agree with Campbell's about the role of our grasp of a
space–time framework and a capacity for temporal thought. It should be noted that
Tulving also claims, with many others, that because such advanced cognitive capacities
are required, only humans enjoy genuine episodic memory. The debate concerning animal
cognition, however, has only begun. It is now well known that animals remember specific
events: scrub jays, for instance, recall the where and when of food caching (McCormack
2001), and so seem capable of thought about particular past events. But does a jay have
episodic memory of the kind humans have, recollecting and planning specific events of
food caching as occurring in the course of its life? Tulving hypothesizes that animals lack
both self‐consciousness and the flexible cognition required for agency—that is, both (c)
and (d)—and so are not capable of genuine episodic memory.
Discerning just who is capable of episodic memory, then, requires greater clarity about
both self‐consciousness and agency.
38.5.2 Self‐consciousness
Page 7 of 17
Memory
There is broad agreement among philosophers and psychologists that episodic memory
requires self‐consciousness. But agreement here is only as deep as agreement about what
self‐consciousness in turn requires, and the topic of self‐consciousness has been the focus
of philosophical dispute for a long time. Minimally, the consensus is that self‐
consciousness involves the capacity to entertain thoughts about oneself, or what we have
called ‘reflexive’ thoughts. I have to be able to think about my fall on the ice, and my
stitches, to remember the episode. So the question becomes: What makes thought
reflexive? How can a representational system have thoughts about itself?
Over this question, too, there has been much dispute and confusion. John Perry's work
(1998) is a helpful tonic. Perry begins by distinguishing a broad set of agent‐relative
thoughts of which genuinely reflexive thoughts are a subset.1 An agent‐relative thought
about a thing is possible because the agent has specific means of finding out about and
affecting the thing, which means are keyed to the role the thing happens to occupy with
respect to the agent. For instance, I think an agent‐relative thought about the glass of
water to my left ‘There's a glass of water’ because the glass is to my left—it occupies the
agent‐relative role to the left—and I can both look left and reach (p. 670) left. To have a
genuinely reflexive thought, Perry claims, two conditions must be met. First, one must
have various normal means of finding out about and affecting the thing that occupies the
agent‐relative role of identity. For instance, I can look in the mirror and touch my nose,
glance down at my feet and brush the sand from them. These are normal routes of self‐
perception and self‐directed action. Second, one must be faced with information about
oneself deriving from non‐normal sources, which information must be integrated with
information derived from normal sources. For instance, if John Perry reads on a printed
schedule that John Perry is giving a talk at noon, then if he is to get himself to the talk on
time he must integrate that information with information he can gain in the normal ways.
This integration task is crucial: if one were never faced with the problem of integrating
objective information about oneself and information gained from normal sources, one
would never need a self‐notion. Normally self‐informative perception (looking down at
one's feet) and normally self‐affecting action (brushing them) could be linked without an
explicit representation of oneself. (One simply brushes what one is looking at.) The
representational store that integrates both kinds of information is what Perry terms a
‘self‐notion’. Possessing a self‐notion, one is in a position to think reflexive thoughts—
thoughts that are about oneself.
Page 8 of 17
Memory
It should be noted that some memory capacities must already come into play in order to
build an enduring self‐notion and thereby to engage in reflexive ‘I’‐thoughts. And here we
face a concern. If reflexive thought requires storage capacities on the scale required for
full‐blown episodic memory, then we're chasing our tails in a circle, at least if we hope to
define episodic memory in terms of a capacity for reflexive thought. Fortunately, if Perry
is correct, it would seem little memory is called for in order to possess a self‐notion.
Short‐term stores would seem enough to enable one to integrate information gained
through normal and non‐normal epistemic enquiries, and to execute actions based on that
information. The memory demands for reflexive thought are less than for full‐blown
memory of specific events of a subject's life. So no problem of circularity arises with
Tulving's requirement (c), on which episodic memory requires self‐consciousness, where
that is understood minimally as reflexive thought.
(p. 671)
With Tulving's requirement (d) we face a similar concern: memory capacities must also
already come into play in order to be an agent, a creature who directs her actions as
much as recalls them. Just how much memory is that?
How much memory does it take to be an agent? Is memory sufficient to ground reflexive
‘I’‐thoughts also sufficient to ground an autonomous self who directs her own actions?
It might seem that more memory is required to underwrite autonomous agency than it
takes to underwrite self‐consciousness or reflexive thought. After all, as we have seen,
only brief stretches of memory would seem necessary for thinking reflexive ‘I’‐thoughts,
and how could that be enough to produce autonomous agency? It is important to see that
this is a genuinely open question, the answer to which depends on one's theory of agency.
On Velleman's view of what makes action autonomous, for instance, memory for reflexive
thought is memory enough. Autonomous action, Velleman argues (2005), occurs when
behaviour is governed by the desire for narrative coherence. Narrative coherence need
only accrue over relatively short time frames—only long enough to say, ‘I'm reaching for
Page 9 of 17
Memory
the glass’ and then reach for it. Working memory sufficient to form intentions and get
them acted on (what psychologists call ‘prospective memory’) suffices. Narrative
coherence over longer time frames, of the sort required to ground the self as a person,
isn't required to ground the self as agent. If Velleman is right, the relevant memory
demands for autonomous action are less than for memory of specific events of one's life,
so no problem of circularity need arise for a view, like Tulving's, on which episodic
memory is defined in terms of a capacity for autonomous action.
One might object that merely prospective memory cannot be enough for autonomous
action. Imagine an amnesiac who makes perfect sense of his actions in intervals of
several minutes, and who decides in favour of one course over another, but who lacks the
larger frame of a life story (or even a significant stretch thereof) in which such local
narratives fit. The amnesiac can seem both less a self and less autonomous, despite his
capacities for short‐term narrative coherence. (Note that if this objection stands, a circle
again threatens Tulving's proposed definition of episodic memory: if agency requires
fully‐fledged autobiographical memory, then we cannot say, with Tulving, that episodic
memory is defined by a capacity for agency.)
How deep can memory failure go before we withhold our attribution of autonomous
agency? How much memory does it take to ground each of the selves—moral, (p. 672)
psychological, metaphysical—we have an interest in? We have only recently begun to
have the philosophical resources to address these questions with rigour. One thing is
clear: in light of recent philosophical advances into the nature of the self and self‐
consciousness we can no longer rest content with simple formulas and general truisms
about the importance of memory to selfhood.
Epistemologists routinely distinguish memory and inference, as each has a distinct role in
the production of knowledge. Perhaps for this reason, questions about the role of
inference in memory have not received much notice from philosophers. The recent
explosion of research in psychology on suggestibility and other sources of memory
distortion (Schacter 1995) promises to challenge philosophical complacency about such
questions. Although memory failure and distortion can hold a grim fascination for us,
what is most interesting is not the distortion itself, but the light it casts on the
mechanisms at work when memory is veridical. One important paradigm that seeks a
unified account of many such mechanisms is the ‘source‐monitoring framework’ (Mitchell
Page 10 of 17
Memory
and Johnson 2000), according to which memory for attitudes and events is driven by both
subperson‐level and person‐level inferences about the original sources of one's
information. If the SMF is correct, inference plays a central role, often unnoticed by the
subject, in the construction of veridical memory. Philosophers have only begun to take
account of the implications of the claims made by the SMF. For instance, that self‐
knowledge of one's past attitudes might involve inference contravenes at least one
important theory of diachronic self‐knowledge (see Shoemaker 1996; Lawlor 2005).
Clearly, memory plays a role in inference. No one could reason without the capacity to
hold two thoughts together or to draw new conclusions from knowledge held in long‐term
memory. Until recently, however, psychologists have tended to separate their study of
memory from their study of inference. (For some new work on the role of working
memory in cognition see Logie and Gilhooly 1998.) Philosophers have tended to focus on
two questions about memory's role in inference.
(p. 673)
(1) First, if all inference relies on memory, does that pose a problem for semantic
externalism? Semantic externalism is the view, roughly, that the semantic value of words,
thoughts, and other representations is determined by facts metaphysically external to the
individual thinker and thus potentially outside her epistemic ken (Burge 1979). Many
have held that because memory plays a central role in inference, if semantic externalism
is true then validity is not a property of an inference that is knowable a priori to the
inference maker (Boghossian 1992). Briefly, the argument is this. Suppose one makes an
inference about a thing one remembers and takes oneself to currently perceive. One
reasons: ice is slippery (memory‐based belief), and that is ice (perception), so that is
slippery. Now suppose that one has unknowingly been switched from earth to Twin Earth,
where the chemical composition of the stuff called ‘ice’ is not H2O, but XYZ. If semantic
externalism is true, the argument goes, the semantic values of one's ‘ice’ tokens are fixed
by two different environments, so one's memorial belief is framed with a token mental
representation, ‘ice’, that refers to frozen H2O, but one's current perception is framed
with a token mental representation, ‘ice’, that refers to frozen XZY. When the time comes
to reason, tokens of the same type are taken to co‐refer, but, as our case reveals,
purportedly co‐referential tokens may not in fact co‐refer, and one won't be sensitive to
the difference. Consequently, the argument continues, if semantic externalism is true,
then it is possible for one to reason equivocally without knowing as much.
Some have taken this argument to constitute a reductio against semantic externalism,
while others have argued, in a nettle‐grasping mood, that validity is thus shown to be an a
posteriori property of specific inferences—knowing that one is thinking of the same
things over time is required for ‘epistemically valid’ inference, and one simply cannot
Page 11 of 17
Memory
have such knowledge a priori (Campbell 1987; Millikan 1993). Still others, more
conservatively, resolve the problem by supposing that the semantic value of token mental
representations is sensitive to the context of inference (Burge 1998), so that whatever
semantic externalism dictates about how the semantic values of words, thoughts, and
other representations are determined, in the context of an inference the semantic value
of two tokens that purport to be the same will be the same. Thus, the role of memory in
inference does not disrupt our capacity to know a priori of a specific inference whether it
is valid or not. The debate here rages on, and can be expected to for some time (Ludlow
and Martin 1998; Nuccitelli 2003). Nonetheless, Burge's initial response retains its
appeal, as it promises the least perturbation of common sense. However, it does require
supposing that the reference of one's memories might shift with new environments.
(2) The second question about memory's role in inference is epistemological. In the
context of reasoning, one's memory must enable one to retrieve propositions (e.g. that ice
is slippery), as well as to have continued access to these as one reasons. Memory thus
performs a crucial service in inference, making thought content available for the duration
of one's reasoning. But epistemically speaking, are new burdens added when one trusts
one's memory to perform this service? Must one add to every inference a premise such as
‘my memory is working well enough to retain all the foregoing premises’? Aside from the
concern that such a premise would launch a (p. 674) regress of demands to the effect that
such premises are also vouchsafed, there is the additional concern that such a premise
would turn all inferential justification a posteriori. After all, how do I know that my
memory is working well enough unless I check? (There are further puzzles here of course
about what checking could amount to, once I doubt my memory.) But certainly some of
our inferences produce knowledge a priori. For example, suppose one reasons through a
proof of Pythagoras's theorem. The fact that one relies on one's memory to hold premises
long enough to draw the conclusion does not make one's justification for believing
a2+b2=c2 a posteriori. This is common sense. As Tyler Burge (1993) puts it, memory
functions in an inference to preserve content: it neither adds to justificational demands,
nor enters new content into an inference (in the form of a premise concerning memory's
reliability).
To say as much only raises the question, why should it be that memory adds no
justificational demands to inference? Burge's answer, briefly, is that one has a defeasible
entitlement to trust one's memory without making memory's reliability itself a matter of
further explicit judgement, because memory is a resource for reason, and reliance on
such a resource in the absence of special reasons for distrust is necessary to the
functioning of reason. Burge sees memory as one of several such ‘content‐preserving’
resources, along with testimony and perception.
This answer, however, raises the further question, how does memory do its job of
preserving content? Some have argued that because working memory involves active and
ongoing reconstruction of contents, just as episodic or autobiographical memory does, no
such preservation of content is possible (Christensen and Kornblith 1997). These
Page 12 of 17
Memory
philosophers argue that we must change our view of justification by inference: common
sense about the apriority of the conclusions of reason is overthrown in light of the
reconstructive nature of all memory.
To rebut this argument we must supply a specific model of how working memory might
function to preserve content, while acknowledging that memory involves reconstruction.
At this stage in our empirical investigations of working memory any model of how
memory preserves content will be abstract, but we can nonetheless make an initial
conjecture: suppose that memory preserves content in something like the way anaphoric
representations preserve reference in speech. For example, when John says ‘Ken will
know; he went to the meeting’, John's use of the pronoun ‘he’ refers to Ken, and in this
sense manages to preserve reference throughout the duration of the utterance. If
working memory involves stringing together representations in thought with anaphoric
relations to each other, then content might in this way be preserved. The conjecture that
working memory might involve anaphoric representation will perhaps trouble those who
reject language‐of‐thought models of cognition, but for those with such scruples
anaphoric thinking can itself be modelled on alternative grounds, without reliance on
‘mental pronouns’ (Lawlor 2002).
References
Anscombe, G. E. M. (1959), Intention (Oxford: Blackwell).
Baddeley, A. (2001), ‘Is Working Memory Still Working’, American Psychologist, 56: 849–
64.
Baddeley, C., Conway, M., and Aggelton, J. (2001) (eds.), Episodic Memory: New
Directions in Research (Oxford: Oxford University Press).
Burge, T. (1979), ‘Individualism and the Mental’, Midwest Studies in Philosophy, 4: 73–
121.
Page 13 of 17
Memory
Christensen, D., and Kornblith, H. (1997), ‘Testimony, Memory, and the Limits of the A
Priori’, Philosophical Studies, 86: 4–20.
Clark, A. (2002), ‘On Dennett: Minds, Brains, and Tools’, in H. Clapin (ed.), Philosophy of
Mental Representation (Oxford: Oxford University Press), 67–90.
Clark A., and Chalmers, D. (1998), ‘The Extended Mind’, Analysis, 58: 10–23.
Dokic, J. (2001), ‘Is Memory Purely Preservative?’, in C. Hoerl and T. McCormack (eds.),
Time and Memory (Oxford: Clarendon), 213–34.
Dummett, M. (1969), ‘The Reality of the Past’, Proceedings of the Aristotelian Society, 69:
239–58.
Engel, S. (1998), Context is Everything: The Nature of Memory (New York: Freeman).
Hacking, I. (1995), Rewriting the Soul: Multiple Personality and the Sciences of Memory
(Princeton, NJ: Princeton University Press).
Hoerl, C., and McCormack, T. (2001) (eds.), Time and Memory (Oxford: Oxford
(p. 676)
University Press).
Page 14 of 17
Memory
—— (2005), ‘Reason and the Past: The Role of Rationality in Diachronic Self‐knowledge’,
Synthese, 145/3: 467–95.
Logie, R. H., and Gilhooly, K. (1998) (eds.), Working Memory and Thinking (Hove:
Psychology).
Ludlow, P., and Martin, N. (1998) (eds.), Externalism and Self‐Knowledge (Stanford, Calif.:
CSLI).
Martin, C. B., and Deutscher, M. (1966), ‘Remembering’, Philosophical Review, 75: 161–
96.
—— (2001), ‘Out of the Past: Episodic Recall as Retained Acquaintance’, in C. Hoerl and T.
McCormack (eds.), Time and Memory (Oxford: Clarendon), 257–84.
Millikan, R. (1993), White Queen Psychology and Other Essays for Alice (Cambridge,
Mass.: MIT Press).
—— (2001), ‘Understanding the Past Tense’, in C. Hoerl and T. McCormack (eds.), Time
and Memory (Oxford: Clarendon), 339–74.
Perry, J. (1998), ‘Myself and I’, in M. Stamm (ed.), Philosophie in synthetischer Absicht
(Stuttgart: Klett‐Cotta), 83–103.
Page 15 of 17
Memory
Ratcliff, R., and McKoon, G. (2000), ‘Memory Models’, in E. Tulving and F. Craik (eds.),
The Oxford Handbook of Memory (Oxford: Oxford University Press), 571–81.
Rumelhart, D. E., McClelland, J. L., and the PDP research group (1986), Parallel
Distributed Processing: Explorations in the Microstructure of Cognition, i (Cambridge,
Mass.: MIT Press).
Schacter, D. (1995) (ed.), Memory Distortion: How Minds, Brains and Societies
Reconstruct the Past (Cambridge, Mass.: Harvard University Press).
—— (1996), Searching for Memory: The Brain, the Mind, and the Past (New York: Basic).
Schacter, D., and Tulving, E. (1994), ‘What are the Memory Systems of 1994?’, in D.
Schacter and E. Tulving (eds.), Memory Systems (Cambridge, Mass.: MIT Press), 1–38.
Schacter, D., Wagner, A., and Buckner, R. (2000), ‘Memory Systems of 1999’, in E. Tulving
and F. Craik (eds.), The Oxford Handbook of Memory (Oxford: Oxford University Press),
627–43.
Squire, R., and Kandel, E. (1999), Memory: From Mind to Molecules (New York: Scientific
American).
Stalnaker, R. (2003), ‘What Might Nonconceptual Content Be’, in Y. Gunther (ed.), Essays
on Nonconceptual Content (Cambridge, Mass.: MIT Press), 95–106.
—— (2001), ‘Episodic Memory and Common Sense: How Far Apart’, in A. Baddeley, M.
Conway, and J. Aggelton (eds.), Episodic Memory: New Directions in Research (Oxford:
Oxford University Press), 269–80.
Wheeler, M., Stuss, D., and Tulving, E. (1997), ‘Toward a Theory of Episodic Memory: The
Frontal Lobes and Autonoetic Consciousness’, Psychological Bulletin, 121: 331–54.
Page 16 of 17
Memory
Notes:
(1) Note that in his own work Perry uses the term ‘reflexive’ to characterize a special kind
of content, and not to mean, as I have here, thoughts about oneself. Perry prefers to call
thoughts about oneself ‘self‐beliefs’.
Krista Lawlor
Page 17 of 17
Emotions: Motivating Feelings
Keywords: emotions, thoughts or feelings, theories of emotions, decision making, action, motivation
39.1 Cognitivism
Page 1 of 15
Emotions: Motivating Feelings
There are almost as many theories of the emotions as there are emotion researchers.
Each contributor to this sizeable literature has a perspective on the nature of the
emotions, and these perspectives often vary considerably. But within this variety a
boundary can be drawn between two major classes of theories, and much of the debate
about emotions has been located at the battle line. On one side, there are those who think
emotions are essentially cognitive. Others deny that emotions are cognitive. In this
section I will briefly present these two approaches. Within philosophy (p. 679) cognitive
theories have been more popular, but I will argue that non‐cognitive theories currently
enjoy more support.
To say that emotions are cognitive is to say that emotions are constituted by or essentially
connected to cognitive states. A cognitive state is a mental state that contains concepts.
For example, being afraid may involve deployment of the concept of danger, being angry
may involve the concept of a slight, and being sad may involve the concept of loss. One
kind of cognitive state is a judgement. Judgements are constituted by concepts. The
judgement that zebras migrate contains the concept of zebras and the concept of
migration. Many cognitive theorists assume that emotions always involve judgements
(Aristotle c.350BC 1984; Spinoza 1677/1994; Solomon 1976; Nussbaum 2001). For
example, they say that I cannot be afraid without judging that I am in danger. Other
cognitive theorists distinguish between judgements and thoughts. To judge that
something is the case is tantamount to adopting the belief that it is true. If I judge that I
am in danger, I must believe that I am in danger. Thoughts are less demanding. I can
think about something, as when I ruminate, imagine, or reflect, without believing it. I can
entertain the thought that I am in danger, even if I believe that I am not. Greenspan
(1988) argues that emotions involve thoughts, rather than judgements, because we can
be afraid even if we know there is not danger at hand. To illustrate Greenspan gives the
example of a phobic person who knows that her phobias are irrational, but nevertheless
continues to feel afraid. Armon‐Jones (1989) suggests that emotions are interpretative
states. To be afraid of something is to see it as dangerous; that is, to place it under the
concept of danger.
Some cognitive theorists argue that emotions are nothing but cognitive states. To have an
emotion is simply to make a judgement, have a thought, or conceptualize something in a
certain way. This view was defended by some of the Stoic philosophers, such as
Chrysippus (Nussbaum 2001). Pure cognitive theories claim that emotions do not have
any other constituent parts. For example, they deny that emotions contain feelings. Pure
cognitive theorists admit that we often have feelings when we emote, but they claim that
those feelings are not components of our emotions. Rather, emotions can cause us to have
feelings, but they need not. Pure cognitive theorists also deny that emotions necessarily
involve behavioural dispositions or changes in the body. A disembodied being could have
emotions on their view. To be afraid is just to deploy the concept of danger in a particular
way.
Page 2 of 15
Emotions: Motivating Feelings
On the face of it pure cognitive theories may sound peculiar. Surely I can judge that
something is dangerous without being afraid. Many people who smoke regard cigarettes
as dangerous, but don't fear them. To use a common metaphor, emotions are not cool
judgements; they are hot. In response, pure cognitive theorists can try to capture the
heat of emotions by distinguishing between two different ways of thinking about danger.
When I think about danger in an emotional way, it captures my attention, and leads me to
focus, perhaps perseveratively, on certain features of my situation. If I am afraid of the
spider on my pillow I may focus all my attention on it, and I may be flooded by thoughts
about how it might harm me (see Öhman et al. 2001). A cool thought about spiders would
not occupy so many of my cognitive resources.
(p. 680)
This attempt to capture the heat of emotions is certainly right about one thing. Emotions
do tend to promote certain patterns of thinking and attention. (See de Sousa 1987 for a
theory that places emphasis on this fact.) But critics of pure cognitive theories should not
be satisfied. For it seems possible to have a perseverative, attention‐demanding series of
thoughts without having an emotion. Imagine a teenage boy who has a pet tarantula and
delights in the fact that tarantulas are dangerous creatures. Whenever he sees his
beloved pet he might attend to it and ruminate compulsively on the hideous agony its
venom would cause. Still, we might imagine that this teenager takes perverse pleasure in
these thoughts, and has no fear. Intuitively, fear requires more than thoughts; it requires
a distinctive feeling. Or consider another example. Suppose a woman says she loves her
husband, and, in keeping with the pure cognitive theory, she focuses her attention on him
when he's in the room, and thinks compulsively about all of his merits. But imagine that
she has no feelings, no flutter in her heart, no arousal when he nears. Imagine that
despite her being fully cognizant of his merits they leave her unmoved. It would be fair to
say she no longer loves him. By comparison, someone who is indifferent to jazz
composition might cognitively understand that Charlie Parker was a great soloist without
ever enjoying his work. Fear, love, and pleasure require more than mere thoughts; they
seem to require feelings. The heat of emotions seems to have a non‐cognitive character.
Recognition of this fact has led some cognitive theorists to adopt impure theories. They
concede that emotions are not wholly exhausted by cognitive states. In addition, they say
that emotions involve feelings. For example, Spinoza (1677/1994) argued that emotions
are judgements that are accompanied by a feeling of pleasure or pain. Descartes
(1649/1985) argued that emotions are judgements that follow felt movements of our
animal spirits—substances that convey information about changes in our body. Many
psychologists defend impure cognitive theories as well (Scherer 1984; Frijda 1986;
Lazarus 1991; Roseman 1994). According to a popular class of views, emotions are
feelings that are caused by appraisals. An appraisal is a judgement about a relationship
between an organism and its environment that bears on well‐being (Lazarus 1991). On
many psychological models we are constantly assessing our well‐being, by unconsciously
and automatically asking ourselves questions such as: Does the present situation concern
me? Is it consistent or inconsistent with my goals? What are my options for coping with
Page 3 of 15
Emotions: Motivating Feelings
it? Am I to blame for this situation or is someone else? How certain am I about the
outcome? Lazarus and Smith (1993) call these molecular appraisals, and Scherer (1984)
calls them stimulus evaluation checks. For these psychologists each emotion corresponds
to a different constellation of answers to questions like these. Fear arises when I judge
that a situation concerns me, that it's inconsistent with my goals, that it is caused by
something other than me, that the outcome is uncertain, and that my coping options
include finding a way to escape. When I make this collection of judgements a feeling
follows, and that feeling is defined as fear.
In response to examples of this kind cognitive theorists have two choices. They can either
argue that, despite appearances, there are cognitive states in each of these cases, or they
can deny that these cases qualify as emotions. The latter strategy is not especially
promising. These seem to be paradigm cases of emotions: anger at being physically
harmed, delight in play, surprise at an unexpected event, and so on. An adequate theory
of emotions must accommodate cases like these. Cognitive theorists can do that only by
saying that each case involves a covert thought or judgement. If cognitive theories are
right, then when you taste the rotten milk you form the thought that the milk is
contaminated and incongruent with your goals. I think that this is extremely implausible
for three reasons. First, it would bad engineering. If certain perceptual events, such as
rotting food, tickling, sounds of distress, and physical assaults, have consistent emotional
implications, then there is no need for an intervening judgement; we can move directly
from perception to emotional response without any intervening step. Second, in each of
the imagined cases there is no conscious experience of a judgement, but there is a
conscious experience of a feeling, and that feeling can be recognized as an emotion on its
own. There is a distinctive feeling of disgust which differs from the feeling of delight, and
both differ from anger and fear. If we discovered a person who had a delight experience
without an accompanying judgement, she would insist that she was delighted. What right
would a theoretician have to say ‘No, you only think you're delighted’? Third, there is
Page 4 of 15
Emotions: Motivating Feelings
evidence from neuroscience that emotions can be elicited subcortically. For example,
there is a pathway from low‐level perceptual centres (the pulvinar and superior
colliculus) to the amygdala which can trigger responses that people will experience as
emotions (LeDoux 1996). Seeing a snake, for example, can trigger a fear feeling before
the snake's appearance has been processed by the neocortex. Subcortical structures are
not believed to be the locus of judgement or thought. There is no evidence that structures
such as the colliculus or the amygdala harbour concepts, such as danger. The pulvinar
and the colliculus are responsive to simple perceptual features, and the amygdala
contains some nuclei that respond to perceptual features and others that send commands
to brain structures that orchestrate responses in the body (including the autonomic
nervous system, the endocrine system, and stereotype responses (p. 682) in the
musculuskeletal system). There is no place for judgement beneath the cortical mantle.
Thus, there is very strong evidence for emotion induction in the absence of cognitive
states.
39.2 Non‐cognitivism
This leads us to non‐cognitive theories of emotion. According to defenders of non‐
cognitive theories, emotions have no cognitive components, and they can occur in the
absence of any judgement, thought, or conceptualization. For most non‐cognitive
theorists emotions are identified with feelings, but there is some disagreement about
what those feelings are. According to some non‐cognitive theories emotional feelings are
sui generis: they cannot be reduced to any other kind of feeling. According to others
emotional feelings are reducible. William James (1884) and Carl Lange (1885) argued
that emotions are reducible to bodily feelings: each emotion is experienced as a felt
pattern of changes in the body. Lange emphasized vascular changes; emotions are felt
changes in blood flow. James had a more encompassing view, citing heart rate, muscle
tension, respiration, and facial expressions. James was a functionalist, in a Darwinian
sense, and he regarded these bodily changes as adaptive. Eyes widen in fear to take in
more information, and the heart races to facilitate flight. Darwin (1872/1998) pointed out
that our hair follicles stand on end when we are afraid, and this may be a vestige from
our earlier mammalian ancestors who were covered with hair; piloerection can increase
apparent girth in a hairy mammal and that can scare off a would‐be predator. According
to James, emotions arise when a thought or perception of a precipitating event triggers a
pattern of bodily changes, and those changes are felt; the feeling of the bodily changes is
the emotion. James argued that if you were to systematically subtract each bodily
symptom of an emotion, there would be nothing left. It seems impossible to imagine the
feeling of fear without imagining tensed muscles, a racing heart, and strained breathing.
James's theory enjoys considerable empirical support (Damasio 1994; Prinz 2004;
Robinson 2005). In study after study emotional feelings have been linked to the body. For
example, when emotions occur, there are measurable changes in autonomic response
(e.g. Levenson et al. 1990), and neuroimaging studies show that brain areas associated
Page 5 of 15
Emotions: Motivating Feelings
with the body are active when people experience emotions (e.g. Damasio et al. 2000).
There is also evidence that people who are given false information about their bodily
states can form false beliefs about their emotional attitudes (Valins 1966), and induction
of bodily states can lead people to report emotion‐like experiences (Marañon 1924;
Duclos et al. 1989). There is evidence too that people who are better at reporting their
bodily states (e.g. at perceiving their heart rate) are more aware of their emotions
(Schnall and Laird 2007). Collectively such findings suggest that emotional feelings are
not sui generis; they are somatic. The most plausible non‐cognitive theory identifies
emotions with felt changes in the body.
(p. 683)
If James's theory of emotions is the most plausible non‐cognitive theory, then cognitivists
must show that it is false. They must show that felt changes in the body are either not
necessary for emotions or not sufficient. Let's begin with necessity. Opponents of James
can try to establish that bodily changes are not necessary by identifying emotions that
occur without bodily changes. Most people grant that our simplest emotions, such as fear,
anger, and disgust, have homologues in non‐human animals, and may usually have
associated behavioural responses. We flee in fear, aggress in anger, and withdraw or
expel in disgust. The somatic states associated with these emotions may be an
evolutionary residue of these behaviour programmes. But some of our emotions are more
sophisticated, and less likely to have homologues in non‐human animals. Consider
evaluative emotions. These include moral emotions, such as guilt and shame, and
aesthetic appreciation, which Hume (1739–40/1978) called a calm passion, because it is
not associated with strong arousal. Emotions of social regard are also not traditionally
related to body states: consider pride, envy, and esteem for others.
I grant that these emotions sometimes seem to occur without patterned changes in the
body. There are, for example, no facial expressions associated with guilt or esteem, and
it's rare to feel your heart racing when looking at art or viewing another's possessions
with envy. But closer analysis tells another story. Each of these emotions can affect the
body. Guilt is associated with reparative behaviours, such as submissively approaching
the party who's been wronged. Shame is associated with gaze avoidance and blushing.
Aesthetic experiences can take our breath away. Pride is associated with smiling and
exhilaration. The envious person stares longingly and disdainfully at envied persons and
possessions. Esteem often blends bodily responses associated with surprise, submission,
and happiness. Ultimately it's an empirical issue whether these emotions have bodily
concomitants, and current research suggests that they do. For example, moral and
aesthetic emotions have been studied using neuroimaging, and in each case researchers
have found activation in brain areas associated with bodily responses, such as the
cingulate cortex and the insula (e.g. Moll et al. 2003; Vartanian and Goel 2004). Work in
‘neuroeconomics’ suggests that envy is associated with those brain structures as well;
they become more active when one player in a game receives a lower monetary award
than another player (Sanfey et al. 2003). Bodily concomitants can also be studied
behaviourally. For example, recent experiments have shown that pride has a recognizable
Page 6 of 15
Emotions: Motivating Feelings
expression, which includes a smile, a tilted back head, and expanded shoulders or chest
(Tracy and Robins 2004). Obviously, each alleged counter‐example to James's theory
would have to be carefully considered, but so far the prognosis looks good. Whenever an
emotion has been studied in the lab, evidence for a somatic component has been found.
So far, no one has identified an emotion that leaves the body unperturbed.
Defenders of impure cognitive theories could concede this point. They could admit that
every emotion involves a perceived change in the body, just as James claimed. But they
might deny that such bodily experiences are sufficient. Cognitive theorists claim that
emotions require judgements as well. Earlier I argued that this is implausible, but that
argument was based on cases in which emotions arise immediately after (p. 684) a
perceived event. Cognitive theorists might reply by arguing that some emotions cannot
be elicited in this way; some emotions require prior thoughts or judgements. The
emotions that I have just been considering are often regarded as examples. Evaluative
and social emotions are associated with sophisticated mental capacities. Cognitive
theorists insist that a person cannot feel guilty without judging that she has done
something wrong, and a person cannot feel pride without believing that her
accomplishments are praiseworthy. In light of these cases one might try to split the
differences between cognitivists and non‐cognitivists. One might say that some emotions
(e.g. fear in response to snakes) are non‐cognitive, while others (e.g. shame and envy)
have essential cognitive components. This would be inelegant, of course; it would force us
to conclude that emotions are not a natural kind (Griffiths 1997). Alternatively, the non‐
cognitivist can try to argue that evaluative and social emotions do not have necessary
cognitive components after all. This strikes me as a promising line of response. It is
plausible that the emotions under consideration can be triggered by perception, in the
absence of judgement. Consider guilt. Imagine that I accidentally step on your foot, and
you squeal in pain. I might feel a sharp pang of guilt, and begin making amends before I
have formed any judgements or thoughts. Notice how people immediately and
instinctively hold up their hands apologetically and defensively and raise their eyebrows
when they bump into each other. Or consider envy. It's easy to imagine that the mere
appearance of another person's being rewarded could elicit an aversive emotional state,
especially if we expected or desired the reward ourselves. This probably occurs in many
mammalian species. Imagine waiving a food treat in front of one mouse and then giving it
to another. My guess is that this perceptual experience will trigger a rodent homologue of
envy: an aversive response to something possessed by a conspecific. Judgement does not
seem necessary. Similar examples can be devised for any of the emotions that I have been
considering.
In response the cognitive theorist might insist that the eliciting perceptions in these cases
qualify as thoughts. Perhaps when the mouse perceives a conspecific getting a coveted
food reward that perception qualifies as the thought ‘He got the food that I wanted’. I
think this is a sensible response, but it does not undermine James's theory. At this point
the cognitive theory collapses into a notational variant of the non‐cognitive approach.
Traditionally cognitive theories have insisted that emotions require the deployment of
fairly abstract concepts, such as danger, insult, loss, and contamination. On appraisal
Page 7 of 15
Emotions: Motivating Feelings
theories, for example, emotion elicitation takes three steps: there is a perception of a
specific event, a construal of that event under one of these abstract concepts, and a
subsequent emotional response. On the present story there are only two steps: a
perception of a specific event and an emotion. If one wants to call perceptions of events
‘thoughts’, then the present story technically qualifies as a cognitive theory, but the label
is terribly misleading, because traditional non‐cognitivists have always claimed that
emotions are triggered by perceptions when they are not triggered by more abstract
judgements. In light of this we might redefine the divide between cognitivists and non‐
cognitivists as a debate about (p. 685) whether emotions can occur in the absence of
abstract concepts, as opposed to concrete representations of specific events. On this way
of framing the debate the cognitive theorist cannot undermine non‐cognitivism by
claiming that perceptions are thoughts. If all emotions can be perceptually triggered,
then no emotions require cognitive states in the sense that is relevant to this debate (i.e.
abstract cognitive states). If guilt and envy can be perceptually triggered, then James's
theory of the emotions extends to cases that have been regarded as counter‐examples. If
all emotions can be perceptually triggered and all emotions involve perceptions of the
body, then emotions are a natural kind, and James's account applies to the entire
category. This would be a victory for non‐cognitivists.
Let me consider one more standard objection to non‐cognitivism. In the examples I have
been considering, emotions are states triggered by precipitating events. But talk of
triggering gives the impression that emotions have a merely causal relation to their
elicitors, rather than a semantic relation. Pepper can trigger a sneeze and acidic food can
trigger indigestion, but we would not say that the sneeze is about pepper or that
indigestion refers to acidic food. In contrast, we do say that emotions are about their
elicitors. We get angry at those who insult us, and we are afraid of things that go bump in
the night. In a word, emotions are intentional. Indeed, they have two kinds of intentional
object. Every token experience of an emotion is about some ‘particular object’ (e.g. fear
of this snake in my path), and every emotion type has a ‘formal object’ which is the
feature of the world that it alerts us to (e.g. fear concerns danger and sadness concerns
loss) (Kenny 1963). Mere bodily sensations are usually thought to lack formal and
particular objects; a mere twinge isn't a representation of anything (except perhaps some
state of the body). Therefore, James's bodily theory of emotions does not explain how
emotions can refer to their objects (Pitcher 1965).
This objection is sometimes presented as if it were a principled and fatal blow to the
Jamesian theory: it's assumed that bodily feelings are just not the kinds of things that can
have intentional objects. But I think the objection is far from fatal; it just shows a lack of
imagination. Some bodily sensations probably lack intentional objects, or refer, at most,
to bodily states. Think of twinges and tingles. But there is no reason to assume that bodily
sensations could not refer to things outside the body. To see whether they can, we need to
ask more generally: How does a mental state come to refer? Philosophers of mind and
language typically distinguish between two different routes to references. On the one
hand, a mental state can refer by description. The concept expressed by ‘the fattest worm
in my garden’ refers by describing something. But mental states can also refer by
Page 8 of 15
Emotions: Motivating Feelings
causation. Philosophers of mind and language have argued that many of our mental
states refer to their objects by being causally related to them in some way. There are
three kinds of causal approach that have been developed in the literature (Stich and
Warfield 1994). On a causal‐history approach, a mental state refers to what actually
caused it to be generated; on a nomological approach, a mental state refers to whatever
would reliably cause it to be activated; and on a teleological approach, a mental state
refers to whatever has the function of causing it to be activated (i.e. those causes in
virtue (p. 686) of which it managed to be replicated in the past). Much of the work on
intentionality involves refinements of these three basic approaches. Obviously, if emotions
do not contain thoughts or judgements they cannot refer by description. An experience of
a racing heart does not describe something as dangerous or insulting. But bodily
experiences could refer causally. Suppose that seeing a snake triggers a pattern of bodily
responses in me. That bodily pattern was caused by the snake, so it could refer to the
snake by means of its causal history. In addition, bodily patterns of this kind would be
reliably caused by snakes and by a variety of other potentially harmful things, so it could
refer nomologically to the class of dangers. And, finally, the pattern of bodily responses,
which is an experience of the body preparing for flight, may have been passed down to us
in evolution in virtue of being caused by dangerous things; the bodily pattern is a
teleological adaptation that plays a role in escaping dangers.
In sum, the most popular causal theories of reference can be readily applied to emotions
even if emotions are feelings of bodily changes (Prinz 2004). Indeed, these theories seem
to entail that such feelings refer. It looks like a straightforward consequence of both
nomological and teleological theories that the bodily sensation associated with fear is
about danger, because that response is caused by dangerous things in a lawlike way, and
it has been passed on through the genome in virtue of its role in successfully reacting to
danger. Thus, nomological and teleological theories could account for the formal objects
of emotions. In addition, on any given occasion the bodily response associated with fear
will have a causal history relating it to some specific elicitor. A token instance of fear
might be caused by a snake, and it can be said to refer, by its causal induction, to that
snake. Thus, causal history can account for the particular object of an emotion. Other
accounts of the intentionality of emotions may be available to the Jamesian as well.
Notice that, in addition to being caused by precipitating events, our emotions also have
behavioural effects that are related to external objects. An angry person might aggress
against someone and the frightened person may flee from something. One could account
for the particular object of an emotion by identifying the person, event, or thing towards
which emotional behaviours are directed. This proposal is consistent with James's theory.
If a bodily sensation leads me to flee from a snake, then we can say that the sensation is
about the snake. The details of this proposal, like all the others, would need to be worked
out. I am not here offering a fully developed theory of how emotions refer. I am merely
pointing out that there are many resources available for explaining how emotions refer,
and these resources are fully available to those who identify emotions with bodily
Page 9 of 15
Emotions: Motivating Feelings
feelings. A feeling of a racing heart, tense muscles, and strained breathing can represent
danger.
I conclude that the bodily theory expounded by William James has the resources to
accommodate all emotions, and it can be defended against standard objections to non‐
cognitivism. If I am right, then emotions can occur in the absence of abstract thoughts,
and they do not have thoughts as components. Instead, emotions are feelings of changes
in the body, caused by thoughts or perceptions, and they represent formal and particular
objects by means of causal relations.
The Jamesian theory of emotions is well suited for explaining the link between emotions
and motivation. For James, fear is a feeling of a bodily change, and that bodily change is a
preparation for flight. Therefore, if a person experiences fear it is ordinarily because her
body has already initiated the early stages of escape behaviour. It's built into James's
theory that we have behavioural programmes that are activated by certain entities and
events, whether real or imagined. Dangerous things set the flight programme in motion,
and fear is the experience of that programme at work. We find such experiences
motivating. When we perceive our bodies in flight mode, we have an urge to flee. By
analogy, the feeling of being tired is an experience of our bodies trying to sleep. It takes
energy to resist what our bodies have already set out to do. Metaphorically, an emotion is
like a ball rolling down a hill; once the action tendency is set in motion, it is easier to go
with the flow than to resist it. If our bodies are trying to flee, we feel motivated to give in.
If this is right, then each emotion comes along with a disposition to act in a certain way.
Psychologists call such dispositions ‘action tendencies’ (Frijda 1986). Fear comes with a
flight tendency, anger comes with an aggression tendency, and delight comes with a
tendency to approach and explore. In some cases the same emotion label actually covers
a family of different felt action tendencies. Fear coincides with flight in many cases, but it
is also associated with freezing and fighting, depending on context. In each case we also
Page 10 of 15
Emotions: Motivating Feelings
need to use contextual information to figure out how to carry out the action for which
we're prepared. If anger comes with a motivation to aggress, we still need to figure out
how to channel the aggression—we must decide who to aggress against and how, or
whether we should resist the urge. Without emotions, we would not be motivated to act.
Emotion deficits have been investigated in the lab. Damasio (1994) has studied patients
with injuries in the ventromedial prefrontal cortex. These individuals are perfectly
intelligent, and they come out normal on all standard cognitive tests, but they make bad
decisions. They can't hold down their jobs, they go into business with charlatans, their
marriages fall apart, and they behave in socially inappropriate ways. In a gambling task
developed by one of Damasio's colleagues it is shown that they (p. 688) play losing
strategies (Bechara et al. 1997). Interestingly, these patients will often say that they knew
the winning strategy but they couldn't bring themselves to play it. Damasio and his
colleagues discovered that performance on the gambling task and, presumably, all the
bad decisions these people make in daily life are the result of an emotion disorder. These
patients have emotions, but their brain injuries disrupt their ability to reliably anticipate
the emotional consequences of their actions. Without this ability they make infelicitous
choices. Just as we need fear to be motivated to flee a burning house, we need to be able
to anticipate the sting of defeat to avoid losing strategies when gambling or choosing
business partners.
Page 11 of 15
Emotions: Motivating Feelings
dictates of morality. Human life would certainly be very different if we didn't emote, and
survival would be difficult if not impossible.
References
Adolphs, R., Gosselin, F., Buchanan, T. W., Tranel, D., Schyns, P., and Damasio, A. R.
(2005), ‘A Mechanism for Impaired Fear Recognition After Amygdala Damage’, Nature,
433: 68–72.
Aristotle (c.350 BC/1984), ‘Rhetoric’, in The Rhetoric and Poetics of Aristotle, ed. E. P. J.
Corbett, trans. W. R. Roberts (New York, NY: Modern Library).
(p. 689) Bechara, A., Damasio, H., Tranel, D., and Damasio, A. R. (1997), ‘Deciding
Advantageously Before Knowing the Advantageous Strategy’, Science, 275: 1293–5.
Chapman, L. J., Chapman, J. P., and Raulin, M. L. (1976), ‘Scales for Physical and Social
Anhedonia’, Journal of Abnormal Psychology, 85: 374–82.
Damasio, A. R. (1994), Descartes' Error: Emotion, Reason, and the Human Brain (New
York: Gossett/Putnam).
Damasio, A. R., Grabowski, T. J., Bechara, A., Damasio, H., Ponto, L. L. B., Parvizi, J., and
Hichwa, R. D. (2000), ‘Subcortical and Cortical Brain Activity during the Feeling of Self‐
generated Emotions’, Nature Neuroscience, 3: 1049–56.
Darwin, C. (1872/1998), The Expression of the Emotions in Man and Animals. 3rd edn.,
ed. P. Ekman (Oxford: Oxford University Press).
Duclos, S. E., Laird, J. D., Schneider, E., Sexter, M., Stern, L., and Van Lighten, O. (1989),
‘Emotion‐specific Effects of Facial Expressions and Postures on Emotional Experience’,
Journal of Personality and Social Psychology, 57: 100–8.
Griffiths, P. E. (1997), What Emotions Really Are (Chicago, Ill.: University of Chicago
Press).
Hare, R. D. (1993), Without Conscience: The Disturbing World of the Psychopaths Among
Us (New York: Simon & Schuster).
Page 12 of 15
Emotions: Motivating Feelings
repr. in The Emotions, ed. C. G. Lange and W. James, trans. I. A. Haupt (Baltimore, Md.:
Williams and Wilkins, 1922), 188–205.
Lazarus, R. S., and Smith, C. (1993), ‘Appraisal Components, Core Relational Themes, and
the Emotions’, in N. Frijda (ed.), Appraisal and Beyond (Hillsdale, NJ: Erlbaum), 233–70.
LeDoux, J. E. (1996), The Emotional Brain (New York: Simon & Schuster).
Levenson, R. W., Ekman, P., and Friesen, W. V. (1990), ‘Voluntary Facial Action Generates
Emotion‐specific Autonomic Nervous System Activity’, Psychophysiology, 27: 363–84.
Moll, J., de Oliveirra‐Souza, R., and Eslinger, P. J. (2003), ‘Morals and the Human Brain: A
Working Model’, Neuroreport, 14: 299–305.
Öhman, A., Flykt, A., and Esteves, F. (2001), ‘Emotion Drives Attention: Detecting the
Snake in the Grass’, Journal of Experimental Psychology General, 130: 466–78.
Phillips, M. L., Young, A. W., Senior, C., Brammer, M., Andrew, C., Calder, A. J., Bullmore,
E. T., Perrett, D. I., Rowland, D., Williams, S. C. R., Gray, J. A., and David, A. S. (1997), ‘A
Specific Neural Substrate for Perceiving Facial Expressions of Disgust’, Nature, 389: 495–
8.
Robinson, J. (2005), Deeper than Reason: Emotion and its Role in Literature,
(p. 690)
Page 13 of 15
Emotions: Motivating Feelings
Sanfey, A. G., Rilling, J. A., Aronson, J. K., Nystrom, L., and Cohen, J. D. (2003), ‘The
Neural Basis of Economic Decision Making in the Ultimatum Game’, Science, 300: 1755–
7.
Scherer, K. R. (1984), ‘On the Nature and Function of Emotion: A Component Process
Approach’, in K. R. Scherer and P. Ekman (eds.), Approaches to Emotion (Hillsdale, NJ:
Erlbaum), 293–317.
Schnall, S., and Laird, J. D. (2007), ‘Facing Fear: Expression of Fear Facilitates Processing
of Emotional Information’, Social Behavior and Personality, 35: 513–24.
Stich, S., and Warfield, T. A. (1994) (eds.), Mental Representation: A Reader (Oxford:
Blackwell).
Tracy, J. L., and Robins, R. W. (2004), ‘Show Your Pride: Evidence for a Discrete Emotion
Expression’, Psychological Science, 15: 194–7.
Jesse Prinz
Page 14 of 15
Intention and Intentional Action
Intention, intentional action, and the connections between them are central topics of the
philosophy of action, a branch of the philosophy of mind. One who regards the subject
matter of the philosophy of mind as having at its core some aspect of what lies between
environmental input to beings with minds and behavioural output may be inclined to see
the philosophy of action as concerned only with the output end of things. That would be a
mistake. Many intentional actions depend for their development on the processing of
input — both from the environment and from the body.
Keywords: intention, intentional action, philosophy of action, philosophy of mind, behavioural output, processing
of input
INTENTION, intentional action, and the connections between them are central topics of the
philosophy of action, a branch of the philosophy of mind. One who regards the subject
matter of the philosophy of mind as having at its core some aspect of what lies between
environmental input to beings with minds and behavioural output may be inclined to see
the philosophy of action as concerned only with the output end of things. That would be a
mistake. Many intentional actions depend for their development on the processing of
input—both from the environment and from the body. Try to imagine yourself cooking a
meal or driving to work without processing such input. Conceptualizing and explaining
intentional actions of intelligent agents—or some central subset of such actions—is the
primary business of the philosophy of action, and properly conducting that business
requires a firm understanding of most of the ins and outs of the philosophy of mind.
40.1 Intentions
Page 1 of 24
Intention and Intentional Action
It is generally agreed that intentions are closely linked to desires and beliefs. Intention
has a motivational dimension, and ‘desire’ (like ‘want’) is often used in the literature as
an umbrella term for motivation.2 Having an intention also is widely regarded as
requiring the satisfaction of a belief condition of some sort. Few philosophers of action
are inclined to assert that a person who believes that her chances of winning today's
lottery are about one in a million intends to win the lottery, no matter how strongly
motivated she is to win.3 However, theorists are divided on how tight the connection is
between intentions, on the one hand, and desires and beliefs, on the other. Some—
attracted, perhaps, by the popular idea that desire and belief are the most fundamental
representational states of mind4—hold that intentions are reducible to combinations of
desires and beliefs (Audi 1973, 1997; Beardsley 1978; Davis 1997; Ridge 1998). Others
have argued that attempts at such reduction are doomed to failure (Davidson 1980: ch. 5;
Searle 1983; Brand 1984; McCann 1986a; Bratman 1987; Harman 1997).
I discussed this dispute in Mele (1992a: ch. 9), where I weighed in on the side of the anti‐
reductionists. Here I am content to observe that the central issue is whether the
settledness that intention encompasses can be captured in terms of beliefs and desires.
Functions plausibly attributed to intentions include (1) initiating and sustaining
intentional actions, (2) guiding intentional behaviour, (3) helping to coordinate agents'
behaviour over time and their interaction with other agents, and (4) prompting and
appropriately terminating practical reasoning.5 Some philosophers have advanced non‐
reductive accounts of intention designed to accommodate many or all of these functions
(Searle 1983; Brand 1984; McCann 1986a; Bratman 1987; Mele 1992a; Harman 1997).
According to a representative account of this kind (Mele 1992a), intentions are executive
attitudes toward plans. Plans—which range from simple representations of ‘basic’ actions
to complex strategies for achieving remote goals—constitute the representational
contents (p. 693) of intentions.6 What distinguishes intentions from other practical
attitudes—for example, desires to act—is their distinctive practical nature. Although one
can desire to do something without being at all settled on doing it, to intend to do
something is, in part, to be settled on doing it (but not necessarily irrevocably). Such
settledness on a course of action constitutes a psychological commitment to executing the
pertinent plan of action, a commitment of a kind arguably constituted exclusively by
intentions.7
Page 2 of 24
Intention and Intentional Action
to do (Mele 2003a: ch. 9). In saying that deciding to A is momentary, I mean to distinguish
it from, for example, a combination of deliberating and deciding. A student who is
speaking loosely may say, ‘I was up all night deciding to major in English’ when what he
means is that he was up all night deliberating or fretting about what major to declare and
eventually decided to major in English. Not all intentions are actively formed, or so I have
argued elsewhere. For example, when I intentionally unlocked my office door this
morning, I intended to unlock it. But since I am in the habit of unlocking my door in the
morning and conditions were normal, nothing called for a decision to unlock it (Mele
1992a: 231). If I had heard a fight in my office, I might have paused to consider whether
to unlock the door or walk away, and I might have decided to unlock it. But given the
routine nature of my conduct, there is no need to posit an act of intention formation in
this case. My intention to unlock the door may have been acquired without having been
actively formed.
Some of our intentions are for the non‐immediate future and others are not. I might
decide on Tuesday to attend a meeting on Friday, and I might decide now to phone my
father now. The intention formed in the former decision is aimed at action three days in
the future. (Of course, if I need to prepare for the meeting—or need to write a note on my
calendar to remind myself of it—the intention may motivate relevant overt conduct sooner
than that.) The intention I form when I decide to phone my father now is about what to do
now. I call intentions of these kinds, respectively, distal and proximal intentions (Mele
1992a: 143–4, 158).8 Proximal intentions also include intentions to continue doing
something that one is doing and intentions to start A‐ing (e.g. start running a mile)
straight away.
Problems
Next on the agenda are some problems faced by anyone intent on analysing intentional
action. As I mentioned, it is natural to suppose that intention and intentional action are
importantly related. The same is often said of reasons and intentional action. For
expository purposes it will be useful to formulate a pair of proto‐analyses of doing
something intentionally, one framed in terms of intentions and the other in terms of
reasons. With the stipulation that ‘A’ is an action variable, the following will do:9
A1. Necessarily, S intentionally A‐ed if and only if S A‐ed in the way that S intended to
A.
A2. Necessarily, S intentionally A‐ed if and only if S A‐ed for a reason.
Page 3 of 24
Intention and Intentional Action
Both proto‐analyses enjoy intuitive support. Again, there presumably is some important
connection between what we intend to do and what we do intentionally. Similarly, it seems,
intentional action is a species of behaviour intimately bound up with agents' desires and beliefs;
and desires, perhaps typically in conjunction with beliefs linking desired goals to prospective
instrumental behaviour, arguably constitute reasons for action.10 In this section I consider four
general problems for the proto‐analyses.
(p. 695)
Consider the following from Gilbert Harman (1997: 151): ‘In firing his gun’, a sniper who
is trying to kill a soldier ‘knowingly alerts the enemy to his presence’. Harman claims that
although the sniper ‘does not intend to alert the enemy’, he intentionally alerts the enemy,
‘thinking that the gain is worth the possible cost’. If Harman is right, A1 and A2 are both
false. Not only does the sniper not intend to alert the enemy, he does not alert them for a
reason either. He has no reason for alerting the enemy, even if he does have a reason to
fire his weapon despite the cost of alerting the enemy. Michael Bratman makes the same
general claim as Harman, illustrated by a scenario featuring a runner who reluctantly
wears down some heirloom shoes (Bratman 1987: 123; see Ginet 1990: 75–6).
There is evidence that, on the topic of side‐effect actions, Harman and Bratman are more
sensitive to ‘the folk concept’ of intentional action than their opponents are. Joshua
Knobe presented subjects with the following story:
S1. 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 … 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.
(2003b: 191)
He reports that 82 per cent of his 38 respondents said that the chairman intentionally harmed
the environment (2003b: 192). Plainly, the chairman is not aiming at harming the environment.
Page 4 of 24
Intention and Intentional Action
(p. 696)
There also is evidence that in responding as they do to this story lay folk are especially
sensitive to the ‘badness’ of the action at issue. Knobe presented another group of
subjects with a similar story that substitutes a good side effect for the bad one:
S2. 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, and it
will also help the environment.’ The chairman … answered, ‘I don't care at all
about 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 helped.
(2003b: 191)
He reports that 77 per cent of his 40 respondents said that the chairman did not intentionally
help the environment (2003b: 192). Evidently, respondents' attributions of intentionality vary
with their moral assessment of the side‐effect actions at issue. Knobe also asked respondents to
S1 to indicate how blameworthy the chairman is on a scale from 0 to 6 and asked respondents to
S2 to indicate how praiseworthy he is on the same scale. The mean rating for blame was much
higher than that for praise (Knobe 2003b: 193). Knobe suggests that this asymmetry is a source
of the asymmetry in intentionality judgements.
How concerned should philosophers of mind and action be with majority judgements of
lay folk about what is or is not done intentionally in relatively straightforward cases such
as these? There is no single correct answer. Philosophers with the project of analysing
‘the folk concept’ of intentional action should definitely be concerned with such
judgements. What about philosophers whose primary business is building an adequate
theory about how agents produce their actions? Such philosophers do not aim to
construct a theory that will handle all actions, including such actions as Ann's
unknowingly frightening some ants—or harming the environment—when she starts her
lawnmower and Bob's accidentally driving his car into a ditch. Most such philosophers
are concerned with a species of goal‐directed action, and ‘intentional action’ may be the
ordinary English expression that most nearly names their target. Of course, an action
theorist's predictions about what lay folk would deem to be intentional or unintentional in
particular cases may be biased by her or his theoretical interests. This undoubtedly
happens. But if it is discovered that most lay folk are happy to count some non‐goal‐
directed actions as intentional and if it is decided that ‘the folk concept’ of intentional
action embraces such actions, philosophers with the project of developing a theory of
action‐production can retain their target and accommodate ordinary usage by modifying
their name for it; they can call it ‘intentional goal‐directed action’. If that sounds
redundant, bear in mind that agents of goal‐directed actions may be as simple as
mosquitos, and mosquitos presumably do not act intentionally. Differences in
philosophical aims are, I suspect, a source of some disagreements in the literature about
the precise nature of intentional action.11
Page 5 of 24
Intention and Intentional Action
Such anticipated side‐effect actions as the chairman's harming the environment are
apparent counter‐examples to the thesis—dubbed by Bratman the ‘simple view’—that
intentionally A‐ing entails intending to A. Some putative belief constraints on intentions,
or rational intentions, also pose apparent problems for that thesis and for proto‐analysis
A1. Bratman argues that intention has a normative side that includes requirements that
an agent's intentions be internally consistent (individually and collectively), consistent
with the agent's beliefs, and means–end coherent (Bratman 1987; see Harman 1997).
Rational intentions, he maintains, satisfy those requirements, and he contends that S
rationally intends to A only if, ‘other things being equal’, S does ‘not have beliefs
inconsistent with the belief that [he] will A’ (1987: 116).
If Amber hit T1 intentionally, proponents of the simple view must say that she intended to
hit it. Since Amber's attitude toward hitting that target is not relevantly different from
her attitude toward hitting T2, they apparently must also hold that she intended to hit T2.
Bratman claims that having both intentions, given what Amber knows (namely, that she
cannot hit both targets), would be irrational. Yet it seems perfectly rational of Amber to
have proceeded as she did. So given the point about the symmetry of Amber's attitudes
toward the targets, Bratman concludes that she did not have either intention. And if
Amber hit T1 intentionally in the absence of an intention to hit it, the simple view and A1
are false.
Some critics of the simple view are also critical of the idea that intentions are reducible to
complexes of beliefs and desires (e.g. Bratman, Harman, and Mele), and Hugh McCann
(1997) argues that they are in danger of having to settle for an unwanted reductive
analysis of intention. Bratman's suggestion that a ‘guiding desire’ to hit T1 can play the
role of an intention (1987: 137) is the primary target of McCann's objection. McCann
observes that once it is conceded that desires can stand in for intentions, reductionists
will justifiably ask what functional need there is for a notion of intention that is
irreducible to desire and belief. However, opponents (p. 698) of the simple view need not
follow Bratman in appealing to guiding desires. On my own view, for example, intentions
Page 6 of 24
Intention and Intentional Action
to try to A can stand in for intentions to A, but intentions to try to A are intentions (Mele
1992a: ch. 8).13 The agent's attitude toward A‐ing is not one of intending to A, but neither
is it merely one of desiring to A. It is, rather, an intending‐to‐try attitude toward A‐ing,
and intending to try is a species of intending. There is nothing irrational about Amber's
intending to try to hit T1 while also intending to try to hit T2.14
McCann (1986b, 1997) and Frederick Adams (1986) are leading proponents of the simple
view. McCann claims that ‘we get at least the beginning of a correct distinction’ between
intentional and unintentional action if we endorse the simple view, and that distinguishing
between non‐intended intentional actions and non‐intended non‐intentional ones
‘promises to be a most intractable problem’ (1986b: 191–2). Adams asserts that ‘we rely
(at least implicitly) on the Simple View to distinguish’ intentional from unintentional
actions (1986: 285). However, Knobe's subjects seem not to be relying on the simple view
when they judge that the chairman intentionally harms the environment. Now, according
to McCann, ‘the Simple View … pertains to the everyday concept of intending, not a
stipulated one’ (1997: 219). Presumably, he would have said the same, if asked, about the
concept of intentional action to which the view pertains. Some of Knobe's data give
McCann reason to doubt that his concept of intentional action is a folk concept of such
action. In a follow‐up study Knobe asked some people whether the chairman in S1
intentionally harmed the environment and asked others whether he intended to harm it:
87 per cent of the 15 respondents to the former question said that he intentionally did
this whereas only 29 per cent of the 17 respondents to the latter question said that he
intended to do it (Knobe 2004: 185). This result is troublesome for the simple view
understood as a view about ‘the folk concepts’ of intention and intentional action.15
I turn to reasons and to A2. Again, on a popular account, the reasons for which we act—
effective reasons—are complexes of beliefs and desires (Davidson 1980). For example, the
reason for which I crossed the road might be constituted by a desire to get to the other
side and a belief that doing so requires a crossing. This account of effective reasons
seems not to do justice to intrinsically motivated actions—actions done for their own
sakes, from ‘intrinsic’ desires. When something, A, is done for its own sake alone, it is not
done from a desire for something further, F, and a belief that identifies A‐ing as suitably
related to F. However, if the reasons for which we A are sometimes constituted wholly by
intrinsic desires to A, such cases can easily be (p. 699) handled (Mele 1992a: ch. 6).
Intrinsically motivated actions arguably are a problem not for A2 itself but for a particular
conception of effective reasons.
Page 7 of 24
Intention and Intentional Action
furry when ‘seized by a sudden desire’ to do so (1991: 62–3). Such actions, as Hursthouse
observes, often are not done for the sake of some further goal, and they typically seem
unreasonable. But it would be a mistake to infer that they are done for no reason at all. If
our reasons can be every bit as bizarre as our actions, proponents of A2 have no special
cause for worry. A man with an irrational urge to drink a can of paint (Davidson 1980: 4)
and the knowledge that drinking the paint requires removing the lid may pry off the lid
for a reason, and I have not encountered compelling grounds for thinking that he cannot
drink the paint for a reason, too—a reason constituted by an intrinsic desire to drink it.
40.2.4 Luck
A more interesting problem for A2 has attracted little attention (as a problem for that
thesis). The popular thesis that all actions done for a reason are intentional (Audi 1997:
104) is challenged by some cases of lucky success. Ann, who has never fired a gun, is
offered a large cash prize for hitting the bull's‐eye on a distant target that even experts
normally miss. She carefully aims and fires, hitting the target dead centre in just the
(direct) way she hoped she would. Many people, I think, would happily (but perhaps
mistakenly) say that Ann's hitting the bull's‐eye—that action—was done for a reason.16
After all, she wanted the money and believed that to get it she must hit the bull's‐eye, and
this helps to explain her carefully aiming and firing at the target. But was Ann's hitting
the bull's‐eye an intentional action? Suppose that Ann has no natural talent with firearms:
she tries equally hard to win even larger prizes for duplicating the feat, fires hundreds of
additional rounds at the target, and does not even come close. Here philosophers'
intuitions differ. According to Christopher Peacocke, an agent who makes a successful
attempt ‘to hit a croquet ball through a distant hoop’ intentionally hits the ball through
the hoop (1985: 69). But Brian O'Shaughnessy maintains that a novice who similarly
succeeds in hitting the bull's‐eye on a dart board does not intentionally hit the bull's‐eye
(O'Shaughnessy 1980: ii. 325; see Harman 1997: 151–2).
Luck is also a problem for A1. Just suppose that Ann, who mistakenly thinks that modern
weaponry makes target shooting extremely easy, intends to hit the bull's‐eye by aiming
and firing at it. She luckily hits it in just the way intended, but was (p. 700) her hitting it
an intentional action? Readers inclined to answer affirmatively should consider a similarly
benighted person who intends to disarm a bomb. She thinks that all she need do to
disarm it is to punch in any ten‐digit sequence of numbers, whereas, in fact, only one ten‐
digit code will work; and wanting to disarm the bomb, she intends to disarm it by
entering ten digits. If she luckily punches in the right code, thereby disarming the bomb,
does she disarm it intentionally? (Does she disarm it for a reason—perhaps one
constituted by a desire to save the townsfolk and a belief that she can ensure her doing
that by disarming the bomb?)
The stories I have spun in this subsection feature either morally neutral or morally
positive lucky actions. Morally negative properties of actions apparently have a
Page 8 of 24
Intention and Intentional Action
significant influence on folk judgements about what is and is not done intentionally in
cases of lucky action. Knobe (2003a: 313) finds that people respond very differently to the
following two stories:
S3. Jake desperately wants to win the rifle contest. He knows that he will only
win the contest if he hits the bull's‐eye. He raises the rifle, gets the bull's‐eye
in his sights, and presses the trigger. But Jake isn't very good at using his
rifle. His hand slips on the barrel of the gun, and the shot goes wild.
Nonetheless, the bullet lands directly on the bull's‐eye. Jake wins the contest.
S4. Jake desperately wants to have more money. He knows that he will inherit
a lot of money when his aunt dies. One day he sees his aunt walking by the
window. He raises his rifle, gets her in his sights, and presses the trigger. But
Jake isn't very good at using his rifle. His hand slips on the barrel of the gun,
and the shot goes wild. Nonetheless, the bullet hits her directly in the heart.
She dies instantly.
Knobe reports that 28 per cent of the 18 respondents to S3 said that Jake intentionally hit the
bull's‐eye whereas 76 per cent of the 21 respondents to S4 said that he intentionally killed his
aunt (2003a: 313–14). Obviously, the main difference between the cases is a moral one.17
Knobe's result is confirmation for my conjecture elsewhere that a story like S4 would get
‘a significantly higher intentionality rating’ from untutored respondents than a story like
S3 (2001: 40). I also conjectured, however, that with a group of subjects to whom it had
recently been made salient that people can be blameworthy for things they do
unintentionally (e.g. that a drunk driver who loses control of his car and unintentionally
kills a family of five can be blameworthy for killing them) the difference in intentionality
ratings of stories like S3 and S4 would shrink significantly (2001: 41). Knobe tests the
latter conjecture too. He reports that the result of an experiment with people primed in
this way is very similar to the original (p. 701) result: 84 per cent of the 32 respondents
said that Jake intentionally killed his aunt in S4 and 40 per cent of the 25 respondents
said that he intentionally hit the bull's‐eye in S3 (Knobe 2003a: 318). As in the case of lay
judgements about side‐effect actions, how concerned philosophers of mind and action
should be with results such as these depends partly on what their aims are.18
Page 9 of 24
Intention and Intentional Action
Familiar causal theories of action feature as causes such mental items as beliefs, desires,
intentions, and related events (e.g. acquiring a proximal intention to A). On an attractive
theory of action, actions are analogous to money in a noteworthy respect. The piece of
paper with which I just purchased a drink is a genuine US dollar bill partly in virtue of its
having been produced (in the right way) by the US Treasury Department. A duplicate bill
produced with plates and paper stolen from the Treasury Department is a counterfeit
dollar bill, not a genuine one. Similarly, according to one kind of causal theory of action, a
certain event is my buying a drink—an action—partly in virtue of its having been
produced ‘in the right way’ by certain mental items (Davidson 1980; Brand 1984).20 An
event someone else covertly produces by remote control—one including visually
indistinguishable bodily motions not appropriately produced by mental states of mine—is
not a purchasing of a drink by me, even if it feels to me as though I am in charge.21
The idea that actions are to be explained, causally, in terms of mental states or events is
at least as old as Aristotle: ‘the origin of action—its efficient, not its final cause—is choice,
and that of choice is desire and reasoning with a view to an end’ (Nicomachean Ethics,
1139a31–2). It continues to have a considerable following (including Goldman 1970;
Davidson 1980; Brand 1984; Bishop 1989; Velleman 1989, 2000; Mele 1992a, 2003a).
Owing partly to the influence of Wittgenstein (1953) and Ryle (1949), causalism fell into
disfavour for a time. The first major source of its revival is Donald Davidson's ‘Actions,
Reasons, and Causes’ (1963; 1980: ch. 1).
the challenge is to provide an account of the reasons for which we in fact act that does
not treat (our having) those reasons as figuring in the causation of the relevant behaviour
(or, one might add, as realized in physical causes of the behaviour). The challenge is
particularly acute when an agent has two or more reasons for A‐ing but A‐s only for only
one of them. Here is an illustration:
Al has a pair of reasons for mowing his lawn this morning. First, he wants to mow
it this week and he believes that this morning is the most convenient time.
Second, Al has an urge 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
his mowing his lawn this morning would constitute suitable repayment. As it
happens, Al mows his lawn this morning only for one of these reasons. In virtue of
what is it true that he mowed his lawn for this reason, and not the other, if not
that this reason (or his having it), and not the other, played a suitable causal role
in his mowing his lawn?
In Mele (2003a: ch. 2) I review detailed attempts to answer this challenge (Wilson 1989; Ginet
1990; Sehon 1994; Wallace 1999) and argue that they fail. Space constraints preclude pursuing
the issue here.
(p. 703)
Deviant causal chains raise difficulties for causal analyses of action and of doing
something intentionally. The alleged problem is that whatever causes are deemed both
necessary and sufficient for a resultant event's being an action or for an action's being
intentional, cases can be described in which, owing to a deviant causal connection
between the favoured causes (e.g. events of intention acquisition) and a resultant event,
that event is not an action, or a pertinent resultant action is not done intentionally.
The most common examples of deviance divide into two types.22 Cases of primary
deviance raise a problem about a relatively direct causal connection between mental
items (or their neural realizers) and resultant bodily motion. Cases of secondary deviance
focus on behavioural consequences of intentional actions and on the connection between
Page 11 of 24
Intention and Intentional Action
these actions and their consequences. The following are, respectively, representative
instances of the two types of case:
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 [or, one might suppose,
his intention to let go of the rope] might so unnerve him as to cause him to loosen
his hold [unintentionally].
A man may try to kill someone by shooting at him. Suppose the killer misses his
victim by a mile, but the shot stampedes a herd of wild pigs that trample the
intended victim to death.
Instructive attempts to resolve the problems such cases pose highlight four claims (Searle 1983:
chs. 3–4; Brand 1984: ch. 1; Thalberg 1984; Audi 1997; Harman 1997; Mele and Moser 1997;
Mele 2003a: ch. 2). (1) Obviously, A is an intentional action only if it is an action; and in many
cases of deviance the pertinent event seems not to be an action. For example, the climber's
‘loosening his hold’ is more aptly described as the rope's slipping from his fingers. A non‐action
cannot falsify a claim of the form ‘an action is intentional only if it is caused in way w’ (although
a non‐action might falsify a claim about actions, an issue discussed shortly). (2) A proper
analysis of intentional action may preclude there being a gap between an intentional action's
psychological causal initiator (or what realizes it) and the beginning of the action. If, for
example, every intentional action necessarily has the acquisition of a proximal intention (again,
in the basic case, an intention to A straight away) as a proximate cause, there is no room
between cause and the beginning of action for primary deviance.23 Perhaps the climber's
loosening his hold also is not an intentional action because it lacks a proximate cause of the
right sort. (3) Intention has a continuous guiding function in the development of intentional
action. Arguably, because the climber's loosening his hold is not guided by a pertinent intention,
it is not an (p. 704) intentional action. (4) An action's being intentional depends on its fitting the
agent's conception or representation of the manner in which it will be performed—a condition
violated in Davidson's shooting scenario, standardly interpreted. How close the fit must be is, of
course, an issue that requires attention (see Mele and Moser 1997). An analysis of intentional
action that incorporates these four claims would yield the correct judgement about Davidson's
two cases.
George Wilson challenges claim (2). Sometimes, Wilson observes, ‘intentions cause states
of nervous agitation that positively enable the agent to perform the type of action
intended’ (1989: 252). He offers the example of a weightlifter whose ‘intention to lift the
weight then caused a rush of nervous excitement … necessary for him to [raise] the great
weight even slightly’. However, this point and example arguably leave the requirement of
proximate causation untouched. What is required is not that intention‐inspired
nervousness, agitation, or the like play no role in the production of intentional actions,
but rather that they not fill a gap between the acquisition of a pertinent proximal
intention and action in such a way that intention acquisition figures only mediately in the
production of the corresponding action. In Wilson's example, one may contend, there is
Page 12 of 24
Intention and Intentional Action
no gap between intention acquisition and the beginning of the lifting that is filled by
nervousness. Rather, one may argue, intention acquisition proximately initiates the lifting
—which action, according to some causalists, begins before the weight's rising (Brand
1984: ch. 1; see Davis 1979; Davidson 1980: ch. 3; Hornsby 1980; McGinn 1982: ch. 5;
and Adams and Mele 1992)—while also producing nervousness required for the agent's
even budging the weight.
Proximal intentions typically are not momentary states. Presumably, the intention to lift
the weight in the case at hand is at work as long as the lifting continues. Even if
nervousness were required for the occurrence of the agent's muscular movements
themselves, a nervousness‐producing proximal intention to lift the weight whose
acquisition is a cause of a corresponding intentional lift would, in conjunction with the
nervousness, figure in the proximate initiation of those movements.24 If, instead, the
causal role of an intention to lift the weight were exhausted by the intention's issuing in
nervousness and the nervousness were somehow to result in the upward movement of
limbs and weight independently of any pertinent intention present at the time, the
‘weightlifting’ would not be intentional. We would have then a case that, aside from its
failure to proffer an intuitively appealing mechanistic explanation of the focal occurrence,
is of the same kind as familiar instances in the literature of non‐intentional occurrences
caused by intention‐inspired nervousness; for example, the case of the climber.
The point about the continued functioning of proximal intentions blunts an objection John
Bishop raises to Myles Brand's position on primary deviance. Bishop observes that
deviance can break in after intention acquisition has (properly) initiated (p. 705) a causal
chain but before bodily movement occurs and deprive the agent of control over his
motions (1989: 139). In such a case, although an agent's motions may accord with his
intention, he does not act intentionally. On Brand's view, however, the proximal intentions
to A that initiate intentional A‐ings also sustain and guide them: ‘Given that intention is in
part guidance … of activity, the intention continues as long as guidance …
continues’ (Brand 1984: 175; see Searle 1983; Thalberg 1984; Alston 1986; Mele 1992a,
2003a: ch. 2; Mele and Moser 1997). In a case of the kind Bishop imagines, guidance is
absent. (For Bishop's own response to primary deviance see his 1989: ch. 5.)
Some causal theorists who assess cases of primary deviance as attempted counter‐
examples to a causal account of what it is for an action to be intentional dismiss them on
the grounds that they are not cases of action at all (Brand 1984: 18; Thalberg 1984). If
this diagnosis is correct, primary deviance poses an apparent problem for the project of
constructing a causal analysis of action. Can causalists identify something of a causal
nature in virtue of which it is false that the climber performed the action of loosening his
grip on the rope?
Page 13 of 24
Intention and Intentional Action
characteristic of intentional action is a problem mainly for the special sciences’ (1970: 62;
see also 166–9). Goldman's remark strikes some readers as evasive (McCann 1974: 462–
3: Bishop 1989: 143–4); but he does have a point. A deviant causal connection between an
x and a y is deviant relative to normal causal routes from xs to ys. Moreover, what counts
as normal here is perspective‐relative. From the point of view of physics, for example,
there is nothing abnormal about Davidson's examples of deviance. And, for beings of a
particular kind, the normal route from intention to action may be best articulated partly
in neurophysiological terms.
One way around the problem posed by our neuroscientific ignorance is to design (in
imagination, of course) an agent's motor‐control system. Knowing the biological being's
design, we have a partial basis for distinguishing causal chains associated with overt
action (that is, action essentially involving peripheral bodily motion) from deviant motion‐
producing chains. If we can distinguish deviant from non‐deviant causal chains in agents
we design—that is, chains not appropriate to action from action‐producing chains—then
perhaps we would be able to do the same for normal human beings if we were to know a
lot more than we do about the human body. I pursue this line of thought in Mele (2003a:
ch. 2), where I develop an account of the place of intentions (or their physical realizers) in
the causal initiating, sustaining, and guiding of overt actions performed by agents I
design. Space constraints oblige me to set this topic aside here.
Some philosophers claim that causalism is inconsistent with there being any actions at
all, that it makes agents vanish. A. I. Melden writes: ‘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. … There is no place in this picture … even for the
conduct that was to have been explained’ (1961: 128–9). Thomas Nagel expresses a
similar worry: ‘The essential source of the problem is a view of persons and their actions
as part of the order of nature, causally determined or not. That conception, if pressed,
leads to the feeling that we are not agents at all. … my doing of an act—or the doing of an
act by someone else—seems to disappear when we think of the world objectively. There
seems no room for agency in [such] a world … there is only what happens’ (1986: 110–
11).
Straightforwardly interpreted, Nagel's worry is not very worrisome. Tigers and bears are
part of the natural order. Setting aside the mind–body problem and radical sceptical
hypotheses (it is all a dream, the only biological entities are brains in vats, and the like), it
seems clear that such animals act. Tigers and bears fight, run, eat, and so on. When they
do these things they are acting. The same is true of us, even if we are part of the natural
order. And what is it not to be part of the natural order? Supernatural beings—for
example, gods and ghosts—are outside the natural order. That a being needs to be
supernatural in order to act is an interesting thought, but I will not try to undermine it
Page 14 of 24
Intention and Intentional Action
until I see an argument for it that commands respect. Another way not to be part of the
natural order is to be abstract in the way numbers are often said to be. Human beings are
not abstract in that way.
David Velleman voices a variant of Nagel's worry. He contends that standard causal
accounts of action and its explanation do not capture what ‘distinguishes human action
from other animal behavior’ and do not accommodate ‘human action par
excellence’ (1992: 462; see 2000: ch. 1). He also reports that his objection to what he
calls ‘the standard story of human action’ (1992: 461), a causal story, ‘is not that it
mentions mental occurrences in the agent instead of the agent himself [but] that the
occurrences it mentions in the agent are no more than occurrences in him, because their
involvement in an action does not add up to the agent's being involved’ (1992: 463).
Velleman says that this problem would remain even if the mind–body problem were to be
solved (1992: 468–9), and, like Nagel (1986: 110–11), he regards the problem as ‘distinct
from the problem of free‐will’ (Velleman 1992: 465, n. 13).
Velleman here runs together two separate issues. Human agents may be involved in some
of their intentional actions in ways that tigers and bears are involved in many of their
intentional actions. Human agents do not vanish in such actions. Scenarios in which
human agents vanish are one thing; scenarios is which actions of human agents do not
come up to the level of human action par excellence, whatever that may be, are another.
Typical causalists are entitled to complain that Velleman has been unfair to them. In his
description of ‘the standard story’ (1992: 461) he apparently has in mind the sort of thing
found in the work of causalists searching for what is common to (p. 707) all (overt)
intentional actions, or all (overt) actions done for reasons, and for what distinguishes
actions of these broad kinds from everything else. If some non‐human animals act
intentionally and for reasons, a story with this topic should apply to them. Also, human
action par excellence may be intentional action, or action done for a reason, in virtue of
its having the properties identified in a standard causal analysis of these things. That the
analysis does not provide sufficient conditions for, or a story about, human action par
excellence is not a flaw in the analysis, given its target. If Velleman were to believe that
causalism lacks the resources for accommodating human action par excellence, he might
attack ‘the standard story’ on that front, arguing that it cannot be extended to handle
such action. But Velleman himself is a causalist. Moreover, causalists have offered
accounts of kinds of action—for example, free or autonomous action and action exhibiting
self‐control (the contrary of weakness of will)—that exceed minimal requirements for
intentional action or action done for a reason.25 Their story about minimally sufficient
conditions for action of the latter kinds is not their entire story about human actions.
Locating interesting differences between some intentional human actions and the
intentional actions of non‐human animals is an important project. Good work on it will
continue to emerge, including work exploring potential roles for higher‐order attitudes
and other relatively sophisticated attitudes in understanding philosophically interesting
differences (see Frankfurt 1988, 1999; Bratman 1999; Velleman 2000; Mele 2003a: ch.
Page 15 of 24
Intention and Intentional Action
10). That work will be more useful and less likely to mislead if its claims are consistent
with the theses that to be a human agent is to be a human being who acts and that many
non‐human animals act.26
40.5 Conclusion
I am not sure where the philosophy of action and its work on intention and intentional
action are headed. Owing to its location at the intersection of the philosophy of mind,
ethics, and metaphysics, the philosophy of action can be pulled in several directions at
once. I have heard it suggested that philosophical work on intention and intentional
action will be rendered otiose by discoveries in neuroscience about how intentional
actions are produced. I do not see that happening. What it is for something to be an
intentional action is a conceptual matter that cannot be settled wholly empirically. The
same is true of intention. Thus, whether specific neuroscientific discoveries are
discoveries about intentional actions and intentions or something else is also a partly
conceptual question that cannot be settled wholly empirically. Elsewhere I have argued
that scientists and philosophers can benefit from one another in their (p. 708) efforts to
understand and explain intentional behaviour (Mele 2003a, 2006: ch. 2, in press). That,
however, is much too long a story to be told here.27
References
Adams, F. (1986), ‘Intention and Intentional Action: The Simple View’, Mind and
Language, 1: 281–301.
Adams, F., and Mele, A. (1992), ‘The Intention/Volition Debate’, Canadian Journal of
Philosophy, 22: 323–38.
Aristotle (1925), Nicomachean Ethics, in The Works of Aristotle, ix, ed. W. Ross (Oxford:
Oxford University Press).
—— (1991), ‘Autonomy, Reason, and Desire’, Pacific Philosophical Quarterly, 11: 1–14.
—— (1997), ‘Acting for Reasons’, in A. Mele (ed.), The Philosophy of Action (Oxford:
Oxford University Press), 75–105.
Page 16 of 24
Intention and Intentional Action
Beardsley, M. (1978), ‘Intending, in A. Goldman and J. Kim (eds.), Values and Morals
(Dordrecht: Reidel), 163–84.
Bratman, M. (1987), Intention, Plans, and Practical Reason (Cambridge, Mass.: Harvard
University Press).
Davidson, D. (1963), ‘Actions, Reasons, and Causes’, Journal of Philosophy, 60: 685–700;
repr. in Davidson, Essays on Actions and Events (Oxford: Clarendon, 1980) 3–20.
Davis, W. (1997), ‘A Causal Theory of Intending’, in A. Mele (ed.), The Philosophy of Action
(Oxford: Oxford University Press).
Page 17 of 24
Intention and Intentional Action
Jackson, F., and Pettit, P. (1988), ‘Functionalism and Broad Content’, Mind, 97:
(p. 709)
381–400.
Kane, R. (1996), The Significance of Free Will (Oxford: Oxford University Press).
—— (2003b), ‘Intentional Action and Side Effects in Ordinary Language’, Analysis, 63:
190–4.
—— (2004), ‘Intention, Intentional Action and Moral Considerations’, Analysis, 64: 181–7.
McCann, H. (1974), ‘Volition and Basic Action’, Philosophical Review, 83: 451–73.
—— (1986b), ‘Rationality and the Range of Intention’, Midwest Studies in Philosophy, 10:
191–211.
—— (1997), ‘Settled Objectives and Rational Constraints', in A. Mele (ed.), The Philosophy
of Action (Oxford: Oxford University Press), 204–22.
—— (1992c), ‘Acting for Reasons and Acting Intentionally’, Pacific Philosophical Quarterly,
73: 355–74.
Page 18 of 24
Intention and Intentional Action
—— (in press), Effective Intentions: The Power of Conscious Will (Oxford: Oxford
University Press).
Mele, A., and Moser, P. (1997), ‘Intentional Action’, in A. Mele (ed.), The Philosophy of
Action (Oxford: Oxford University Press), 223–55.
Mele, A., and Sverdlik, S. (1996), ‘Intention, Intentional Action, and Moral Responsibility’,
Philosophical Studies, 82: 265–87.
Nagel, T. (1986), The View from Nowhere (Oxford: Oxford University Press).
Scanlon, T. (1998), What We Owe to Each Other (Cambridge, Mass.: Harvard University
Press).
Page 19 of 24
Intention and Intentional Action
Sehon, S. (1994), ‘Teleology and the Nature of Mental States’, American Philosophical
Quarterly, 31: 63–72.
—— (1984), ‘Do Our Intentions Cause Our Intentional Actions?’, American Philosophical
Quarterly, 21: 249–60.
Thomson, J. (1977), Acts and Other Events (Ithaca, NY: Cornell University Press).
Wallace, R. J. (1999), ‘Three Conceptions of Rational Agency’, Ethical Theory and Moral
Practice, 2: 217–42.
Notes:
(1) See Wilson (1989) for a third approach, according to which intentions are not
attitudes.
(3) A relatively popular claim among philosophers is that agents intend at a time t1 to A at
a time t2—where t1 and t2 may or may not be identical—only if they believe at t1 that they
(probably) will A at t2. (Proponents—some of whom omit the parenthetical qualifier—
include Audi 1973, 1991, 1997, Beardsley 1978, Harman 1986: ch. 8, 1997, Velleman
1989, and Davis 1997). The proposal is designed to capture, among other things, the
confidence in one's success that intending allegedly involves. In Mele (1992a: ch. 8) I
defend the less demanding claim that intending at t1 to A at t2 requires that, at t1, one
lack the belief that one (probably) will not do A at t2. (The person might have no belief on
the matter.) Other alternatives include the requirement that S believe to a ‘degree’ (even
Page 20 of 24
Intention and Intentional Action
a degree associated with a subjective probability significantly less than 0.5) that he will A
(Pears 1984: 124; also see Pears 1985) and the requirement that S believe that ‘there is
some chance that he can’ A (Davidson 1985: 215).
(4) For resistance to this idea see Searle (1983) and Brand (1984).
(5) On the first pair of functions see Brand (1984), Thalberg (1984), Bishop (1989), and
Mele (2003a: ch. 2). On the second pair see Harman (1986, 1997) and Bratman (1987,
1999). For discussion of all four functions see Mele (1992a: ch. 8).
(6) Roughly speaking, basic actions differ from non‐basic actions in not being performed
by way of performing another action.
(7) The commitment aspect of intention receives detailed treatment in Bratman (1987).
See also Audi (1991) and Mele (1992a: chs. 9–10). For a more detailed summary of the
dispute between belief/desire reductionists about intention and their opponents see Mele
(1997b: 17–19).
(8) Searle makes a case for the importance of what he calls ‘intentions in actions’ (1983:
ch. 3). As far as I can tell, they are tryings. On this matter see Mele (1992a: 183–6).
(9) By the end of the 1970s a lively debate over the question of action‐individuation had
produced a collection of relatively precise alternatives. Imagine that Don flips the switch,
turns on the light, illuminates the room, and—unbeknownst to him—alerts a prowler to
his presence (Davidson 1980: 4). How many actions does he perform? Davidson's coarse‐
grained answer is one action ‘of which four descriptions have been given’ (1980: 4; also
see Anscombe 1957). 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, since the act‐properties at issue are distinct. An
agent may exemplify any of these act‐properties without exemplifying any of the others.
One may even turn on a light in a room without illuminating the room: the light may be
painted black. Another alternative, a componential view, represents Don's illuminating
the room as an action having various components, including (but not limited to) his
moving his arm, his flipping the switch, and the light's going on (Thalberg 1977; Thomson
1977; 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. In this chapter I proceed in a neutral way
regarding the leading contending theories of individuation. Readers may read my action
variable ‘A’ in accordance with their preferred theory of action‐individuation. The same
goes for the term ‘action’.
(10) This broadly Davidsonian conception of reasons for action is controversial. Some
philosophers contend that Davidsonian reasons for action are not reasons at all. T. M.
Scanlon, for example, argues that ‘desires almost never provide reasons for action in the
way described by [this] model’ (1998: 43). Philosophical work on reasons for action tends
Page 21 of 24
Intention and Intentional Action
(12) For a more comprehensive survey of the literature on this example see Mele (1992b:
202–5).
(15) Incidentally, Knobe asked parallel questions about the chairman in S2 who benefitted
the environment without caring whether he did. 20 per cent said that he intentionally
benefitted the environment and none said that he intended to benefit it (Knobe 2004:
185).
(16) For resistance see Mele (1992c). Incidentally, intuitions tend to shift when the agent
succeeds in a way that diverges significantly from her plan, or from her (perhaps tacit)
assumptions about what a successful A‐ing would involve. If the bullet had ricocheted off
several rocks into the bull's‐eye, would Ann's hitting the bull's‐eye have been done for a
reason?
Page 22 of 24
Intention and Intentional Action
(17) If hitting the bull's‐eye seems to be a more difficult task than killing the woman, that
may account for part of the difference in the respondents' judgements. Knobe also asked
about variants of these cases in which ‘Jake is an expert marksman. His hands are steady.
The gun is aimed perfectly’. Here 95 per cent of the respondents said that he
intentionally killed his aunt and 79 per cent said that he intentionally hit the bull's‐eye
(Knobe 2003a: 314). This difference in responses might be partially accounted for by
differences in respondents' assumptions about difficulty. In an ideal test, it would be clear
that the moral and non‐moral tasks are equally difficult. For a pair of cases of this kind
see Mele and Sverdlik (1996: 281–2).
(18) For a review of problems for A1 and A2 posed by sudden and impulsive actions and
by subsidiary actions see Mele (1992b).
(19) I borrow the term ‘causalism’ from Wilson (1989). (Wilson is a non‐causalist.)
(21) This view does not identify actions with non‐actional events caused in the right way.
That would be analogous to identifying genuine US dollar bills with pieces of printed
paper that are not genuine US dollar bills and are produced in the right way by the US
Treasury Department, and so identifying genuine US dollar bills would be absurd.
(23) ‘Proximate cause’ may be defined as follows: x is a proximate cause of y if and only if
x is a cause of y and there is nothing z such that x is a cause of z and z is a cause of y.
(24) Here I set aside cases in which a nervousness‐producing intention to lift the weight
issues in a distinct intention to lift the weight that has a more direct causal role in the
lifting. Such cases pose no special problems.
(25) Many compatibilists are causalists about action. For a causalist account of self‐
control see Mele (1995).
(26) For discussion of a third alleged problem for causalism—the problem of so‐called
negative actions (e.g. intentionally not voting in an election)—see Mele (2003a: 146–54).
(27) Parts of this chapter derive from Mele (1992a, 1992b, 1997b, 2003a, 2003b, and
2005).
Page 23 of 24
Intention and Intentional Action
Alfred R. Mele
Page 24 of 24
Folk Psychology
We describe people in terms of their beliefs, desires, emotions, and personalities, and we
attempt to explain their actions in terms of these and other such states. These are folk-
psychological concepts; they are the ones we use when we think of people in terms of
their minds. By ‘folk psychology’ it is meant whatever serves this linking role for folk-
psychological concepts. Much philosophical writing about concepts would suggest that
the linking must be done primarily by beliefs or theories; a central question is whether
this is so for folk-psychological concepts.
Keywords: folk psychology, belief and desire, concepts of emotion, human action, physical world, state of mind
WE describe people in terms of their beliefs, desires, emotions, and personalities, and we
attempt to explain their actions in terms of these and other such states. These are folk‐
psychological concepts; they are the ones we use when we think of people in terms of
their minds. (Strawson 1959 called them ‘M‐concepts’.) Let us leave it fairly vague what is
to count as a folk‐psychological concept, taking as core examples the concepts of belief
and desire and concepts of emotion, and then including as folk psychological concepts
any concepts whose primary function is to enter into combination with the core examples
to give explanations and predictions of human action. (It is important to remember,
though, that explaining and predicting action is not the only purpose of ascribing states
of mind. Often, for example, the ascription is part of an inference to a conclusion about
the physical world, as when we say ‘She seems to be lying, so there is probably
something hidden in the basement’.) Usually when there are concepts there are beliefs or
theories, whose structure allows us to understand the concepts by linking them to other
concepts and to experience. By ‘folk psychology’ let us mean whatever serves this linking
role for folk‐psychological concepts. Much philosophical writing about concepts would
Page 1 of 18
Folk Psychology
suggest that the linking must be done primarily by beliefs or theories; a central question
is whether this is so for folk‐psychological concepts.
Page 2 of 18
Folk Psychology
41.1 Theory
The idea now labelled ‘folk psychology’ arose around the 1970s when a number of
philosophers, often influenced by Sellars's description (1956) of ‘the myth of Jones’ and
by Quine's assimilation (1953) of all of our thinking to explanatory (p. 714) hypotheses,
began to explore the consequences of assuming that individual people acquire from their
cultures an explicit theory which postulates that individuals have beliefs, desires, and
other states, which interact in specific ways to produce actions. (Typical examples are
Putnam 1960, Dennett 1978, Churchland 1979.) This theory would be the common‐sense
precursor of psychological theories developed as part of the scientific approach to human
nature.
The position had many attractive features. First, it fitted very neatly with the dominant
account at the time of the nature of mental states, functionalism, according to which,
roughly, mental states are any states of an organism which interact in the right ways. The
right ways can then be taken as those described by folk psychology. Second, the position
also provided a new take on the old epistemological problem of other minds. The
traditional argument from analogy was intellectually discredited, but now it became
possible to see each person's reason for attributing states of mind to others as an
inference to the best explanation, in which the ‘hypothesis’ that another has states
interacting in the way required by folk psychology emerges as the best explanation of the
person's actions. And, third, it became possible to formulate a new range of philosophical
questions analogous to old‐fashioned scepticism about the existence of other minds but
much more varied and plausible. Is folk psychology true? Is it a first approximation to an
eventual scientific psychology? Might it eventually be replaced by a theory based on
experimental evidence? These questions are much saner than other‐minds scepticism:
one can quite naturally imagine the answers to any of them going either way.
Page 3 of 18
Folk Psychology
information‐poor accounts, on the other hand, the fundamental features of our thinking
about mind are determined by routines and capacities which do not embody any
assumptions about how one state of mind leads to another.
Page 4 of 18
Folk Psychology
The other pioneering simulation account, Gordon's, applies best to cases in which a
person is predicting another person's decision. Suppose that A is driving behind B and
sees a deer run into the road in front of B's car. A can anticipate B's braking heavily and
swerving to the left. A will then also brake, before seeing B's brake lights, and may decide
not to swerve left in the hope of missing both B and the deer. One way in which A could
form this expectation about B's behaviour is by deciding what she would do in B's place,
where this means not thinking about B's decision but making it, feeding into her own
decision‐making processes the facts about B's situation and then taking the result not as
an intention for her own action but as a prediction about B's. A crucial part of the
capacity to think of others as minds, on such an account, is being able to take one's own
decision processes ‘off line’ and use them to form expectations about other people's
actions. To do this one has to be able to take the other person's point of view, at least to
the extent of understanding what aspects of the situation are and are not known to the
other person, and one has to be able to insulate it from the processes by which one turns
decisions into actions. As Currie (1995) and others have pointed out, these capacities are
also required for conditional thinking, in which one feeds into one's decision processes
some hypothetical facts and then stores the outcome as a conditional decision, not to be
acted on unless the facts turn out to be actual.
The contrast between simulation and theory should not be made too stark. In most cases
a Gordon‐type simulation will have to be guided by some information about how it is to be
performed. (‘The driver ahead can see the deer because it is nearer to him than it is to
me, and so I can take that information into the simulation; but he may not know that I am
close behind, so I can leave that information out.’) In many cases, too, a simulation will
also require information about the target person's (p. 716) desires, which may not be the
same as those of the ascriber. (Suppose the driver ahead is known to hate deer more than
he loves his car. Then in order to predict his actions one has to replace one's own
aversion to a collision with an aim towards one.) So even an information‐poor simulation
Page 5 of 18
Folk Psychology
account will have to make use of information about when and how to simulate, and this
information is potentially very rich. Or consider how a Heal‐type account can apply to
situations in which the target person's problem‐solving capacities are greater or less than
those of the ascriber. (Their grasp of English syntax is feeble; their arithmetic is brilliant
or eccentric.) Then in addition to solving the problem oneself one will have to apply some
correction or transformation, to come up with the other person's solution. Most real cases
will have some element of this, and will as a result require guidance from some beliefs
about the other's problem‐solving capacities and how to take them into account (Heal
2000; Morton 2002: ch. 2).
The line between applying a theory and reproducing the other person's thinking in your
own mind is also vague. Suppose for example one is anticipating a person's actions by
using a theory of belief/desire/action. Ascriptions of beliefs and desires will be needed,
and there will usually be unmanageably many possibilities consistent with the person's
prior behaviour. A natural approach to the problem is to use one's own desires and beliefs
as a first approximation, adjusting the ascription as needed later on. In fact, we do
unreflectively assume that others believe that the sky is above the earth and that death is
to be avoided, in the absence of very strong evidence to the contrary. As Goldman (1989)
has pointed out, this amounts to a kind of simulation: one uses one's own assessment of
the situation as a guide to the thinking of the other. (The default ascriptions are
information‐rich in that a lot of information about the environment is taken into account,
and information‐poor in that one entertains fewer thoughts about the other person's
thinking.)
Information‐poor procedures have generally been found attractive when the task of
anticipating another person's actions by tracing their beliefs and desires and reasoning is
too daunting. In fact, even in the most favourable cases it is hard to see how one could
anticipate actions purely by use of a theory of motivation. The reason is the combinatorial
explosion of possible lines of reasoning. Given a desire for a small cup of coffee and
enough coins to operate a machine that vends a large cup of decent coffee, a person
might of course use the coins to buy the coffee and drink half of it, but she might also
wait till someone else buys a cup and then buy half of it from them, or buy a cup and then
sell half of it to another, or any of indefinitely many other actions, all of them easily
rationalizable in terms of her beliefs and desires. But in many real situations we feel sure
—somehow—what lines of thought people are likely to follow. It might then seem
miraculous that this confidence succeeds, that people do quite often do what we expect
them to. The reason, though, is that we do not blindly follow out all possible lines of
reasoning available to a person. Instead, we expect people to follow reasoning that seems
natural to us, the reasoning we ourselves would follow. And, of course, we adjust this
expectation given what we learn about particular people's peculiarities.
(p. 717)
Page 6 of 18
Folk Psychology
The question of the exact mix of cognitively demanding information‐rich theory and less
demanding information‐poor procedures remains very puzzling. It seems clear that we
very rarely get to a prediction of what someone will do by combining explicit information
about the person's beliefs and desires with explicit information about how human beings
combine beliefs and desires to form intentions. On the other hand, it seems clear that we
normally deploy a rich variety of information, about particular people and about people in
general, in forming our expectations of what they are likely to do. Here is a suggestion
about how the pieces fit together, a reasonable suggestion given the current state of the
debate, but definitely a speculation.
There are three basic resources, on this suggestion. The first is a core folk‐psychological
theory, which specifies the basic characteristics of the basic psychological states. Call it
the ‘category theory’, as it outlines the general categories of states that can be ascribed
to people. It postulates informational states that represent the world as it is taken to be,
target states that represent the world as the person aims to make it, emotions and other
dispositions to types of behaviour, and processes of reasoning that lead from beliefs and
desires to intentions or actions. Most of this theory is in place in four‐year‐old children,
and parts of it emerge very early in life. But the theory does not specify what
informational and target states there are. (It is neutral on the relation between what
someone ‘thinks’, what she ‘supposes’, and what she ‘believes’, or between what she
‘wants’, ‘needs’, and ‘would like’.) It does not specify what forms reasoning takes. (It
simply says that people think and form intentions, and consider evidence and change
their beliefs.) And it does not embody any ideas about rationality. (Wishful thinking is as
likely as syllogisms, according to the core theory.)
The category theory becomes more powerful if it is scaffolded with some other skills. One
obvious skill is linguistic. In learning to speak one acquires one's culture's vocabulary of
mental‐state terms and learns how to ascribe them. Belief‐ascription—in particular,
knowing what state is ascribed to a person when someone says ‘she believes that p’—
must be especially hard to learn, as the rules governing it are extremely subtle (Braun
1998). Given a command of the language, one can attribute to people the states ascribed
to them by other people or themselves, as a basis for thinking about their minds. (How
much of this ascription is literally true, and how much is part of a social web of mutual
ascriptions which domesticated adults cannot easily act contrary to? That is not the kind
of question that the philosophers or psychologists thinking about folk psychology pay a
lot of attention to. For an exception see Kusch 1999.) And, given language, a great stock
of platitudes, logical principles, folk generalizations, and old wives' tales becomes
available, some of which will apply in almost any situation.
No amount of linguistic competence will allow one more than a primitive level of
prediction and explanation without the second basic resource. That is simulation: the
capacity to use one's own information and thinking as a guide to the thought of other
people. One has to be able to say where a person's beliefs and desires might take her, for
which the core theory gives no help. There are two related and possibly (p. 718) prior
capacities from which capacities of a generally simulational kind can develop. The first is
Page 7 of 18
Folk Psychology
the capacity, which develops early in normal children, to track the gaze of another
person, to know where the other is looking. This allows a child to know which of the
situations she is observing are also part of the information available to another. The
second is the capacity for conditional thinking, which allows you to see what conclusions
you would have come to if you had had counterfactual information and aims. Mastering
counterfactual thinking requires that one learn how to separate beliefs and desires into
consistent strands: wondering what would happen if the chair were put on the table
involves separating off one's belief that the chair is not on the table, just as one has to
when thinking what someone will do who thinks that the chair is on the table.
The capacity to follow another person's gaze is an essential prerequisite to using one's
own mind to model the thinking of another. Conditional thinking on the other hand
provides increasingly subtle ways of using information about another. For example, it
allows one to explore the consequences of the fact that someone has not noticed
something. They are both part of a third resource, neither category theory nor simulation.
Let me call it the ‘how‐to manual’. This is a body of information about what kinds of
simulation work under what conditions, and about what information about another person
can be used as input to one's own thinking in modelling another's. It must also have an
element of knowing how: how to use what one knows about others to fine‐tune one's
simulation of them. I call it a manual rather than a theory as it is likely to consist of a
large number of rough generalizations and unconnected facts, and even of little tricks
that work only for anticipating the thinking of particular individuals. Still, the manual is a
body of rich if disparate information, largely second‐order information about how to use
the first‐order information one has about other's minds to guide one's information‐poor
explanatory capacities.
Page 8 of 18
Folk Psychology
as a social network in which a delicate cooperation is needed so that each person gets the
information they need. Non‐human animals do not have these capacities in anything like
human form, and human infants have to go through a considerable process of
development or learning before they can exercise adult competence. If therefore we could
understand how individual humans come to acquire mind‐ascribing capacities and how
humanity as a species came to possess them, we would understand something very basic
about what it is to be human.
To psychologists, as to philosophers, the idea that our capacity to think in mental terms is
based on a theory was immediately very appealing. But psychologists immediately
supposed that the ‘theory of mind’ must be an implicit theory that develops during a
child's first few years of life. Early writers in the theory‐of‐mind movement, such as
Wellman (1990) and Perner (1991), supposed that a child moves towards a theory the
core of which postulates that human beings have representations of the environment,
some in the form of beliefs and some in the form of desires. A key theoretical aim was to
understand false‐belief problems, the surprising difficulty children before the age of three
have in attributing false beliefs. The experimental data, originally due to Wimmer and
Perner (1983), concerns situations in which unambiguous evidence of a fact is available
to a child but not to a ‘target’ person. For example candy is moved from one container to
another, in the sight of the child but not of the target person. An adult would
unhesitatingly ascribe to the target person a false belief, the belief that makes sense in
the absence of the evidence. But small children instead ascribe to the target person a
true belief, the belief that they themselves have formed in response to the evidence. They
say that the target person will look for the candy in the place to which it has been moved,
although she has not seen it moved.
The phenomenon is robust under a number of variations. It certainly shows that small
children have difficulty with knowing when others will and will not make inferences. The
difficult thought is: Evidence E is not available to X, so X will not have made the inference
‘E therefore not p’. It does suggest that the concepts of knowledge and ignorance are
easier to grasp than the concept of belief, for small children have little problem
understanding that people can lack particular true beliefs. (That is why hide‐and‐seek is
not a conceptual problem: the child can think ‘X does not believe I am here’ even though
she may have difficulty with ‘X believes that I am there when I am here’.) It is not clear
how much trouble small children have with false beliefs when understanding inference is
Page 9 of 18
Folk Psychology
not required: for example when an authority simply announces that X thinks the candy is
in location L (when in fact the child knows that it is in location L′). There does not seem to
be a robust phenomenon of children then expecting X to look in location L′. Nor are false‐
belief difficulties insuperable; suitable coaching can improve the performance of children
who are under the threshold. But the threshold is real and does mark a point in the
development of normal human children. Some older children with developmental
difficulties, notably autistic children, have the same difficulties with false‐belief tasks that
young children have, even though their general intelligence is on a level with (p. 720)
children who are well past the threshold. Great apes, such as chimpanzees, bonobos, and
gorillas, do not solve analogues of classic false‐belief tasks successfully, even though
deliberate deception and distraction is not uncommon in their social life. It is as if they
too can understand ‘p, but X does not believe p’ but cannot understand ‘p, but X believes
not p’.
The rival accounts of folk psychology that we have already seen can be used to explain
what is going on in false‐belief tasks. Simulation accounts fit very naturally with the pre‐
threshold child's behaviour: she models the target person's beliefs on her own and
predicts that the target person will do what she would do. What the child struggles to
learn, on these accounts, is how to keep facts available to herself but not to the target
person out of her modelling of the other's thinking. Theory theory accounts, on the other
hand, can explain very simply what the child past the threshold has learned. She has
improved her theory of thinking so that it includes a clause ‘if X cannot observe that P
then, other things being equal, X will not believe that p’. (This combined with the fact that
X initially believes that not p, will entail that X will continue to hold the false belief that
not p.)
Each account's strength as applied to the false‐belief situation links to a weakness. When
a simulation account suggests that in passing the crucial threshold children learn how to
be selective about which of their own beliefs to use in modelling another person's
thinking, it draws attention to the absence of any account of how one makes this
selection. Since a selection is suitable or not depending on the situation of the target
person, including that person's beliefs and desires, it must be guided by some reflection
on the person's mind (at a minimum, reflection on whether the person was paying
attention to the crucial inference‐triggering evidence). On the other hand, when a theory
theory suggests that the crucial transition consists in improving the child's theory of
inference, it draws attention to its ascription to the child of a theory of inference. But, as
we have seen, explicit theories of inference are unmanageable monsters, quite useless for
predicting what conclusions a person will arrive at.
The speculative account in the previous section, according to which a core theory
specifies simply that people have beliefs and desires, and avoids any commitments about
inference, can be used to get around these problems. Suppose that at the earlier stage of
children's development the core theory makes available to them the thought that others
have beliefs, but that children rely for the content of these beliefs largely on simple rules
such as: When a person can perceive something they have true beliefs about it; when
Page 10 of 18
Folk Psychology
something is obvious and a person has a belief about it then they believe the obvious. For
updating the beliefs ascribed to a person in the face of evidence children would simply
use their own belief changes as a guide to the thinking of the other person. Then, as they
pass the threshold they would learn which of their own inferences can be used to model
other people's thinking under which circumstances. They would do this by forming a
theory—not a theory of inference but rather a theory of how to track someone else's
reasoning with their own. Some of the applications of this theory are very subtle and
complicated, and in principle the theory could continually improve throughout a person's
life, but at the threshold stage (p. 721) it can consist of some fairly simple principles such
as ‘Model X's belief changes only using information available to X’.
If this suggestion is correct, a child emerges past the false‐belief threshold with a
conceptual structure characteristic of adult thinking about minds, though the range and
power of her thinking about personality and motive may be much less than that of most
adults. It consists of a collection of simulational capacities sandwiched between two
theories, the category theory and the how‐to manual mentioned above.
The suggestion here does not fit easily into a simple evolutionary story of how human
mental‐state ascription developed from the capacities of other apes. It neither points to a
single crucial conceptual development which characterizes the human situation, nor
allows us to see a clear series of steps leading from primate sociality to human thinking.
For the basic elements of the category theory can be expected to be found in other apes,
and the crucial false‐belief threshold consists not in a development of this theory but in a
series of small accretions to the how‐to manual, which is much more dependent on
language and on social interaction. And though the suggestion may of course be wrong,
recent comparative primate work has marked a retreat from the once‐fashionable view
(Byrne and Whiten 1988; Byrne 1995) that the cooperation and deception in primate life
makes it likely that many primates, and in particular chimpanzees, bonobos, and gorillas,
possess many of the capacities essential to human thinking about mind, which require
just a little tweaking and reconnecting to come together in their human form. Now, in
contrast, the consensus (Povinelli and Eddy 1996; Sterelney 2004: ch. 4) is that ape
capacities to track the attention of other apes, and to take account of information in the
possession of others, are much more limited than we had for a while supposed, and that
there are far more discontinuities than continuities with human cognition. That is not to
say that we should see human folk psychology as emerging in a discontinuous
evolutionary leap. In fact, it is much easier to give plausible accounts of its origins than
for the origins of, say, human language.
It is not deeply puzzling how folk psychology could have evolved, though we do not know
nearly enough to choose from among the possible routes. Our ancestors, like other apes,
had the capacity to recognize dozens of individual conspecifics and remember their
important characteristics; they could judge the mood of individual conspecifics by
attention to their gaits and faces; they could form long‐term and temporary coalitions for
purposes as diverse as hunting and raising offspring; they could commit themselves to
cooperative behaviour in situations in which it was in the interests of each individual to
Page 11 of 18
Folk Psychology
defect from such behaviour, and estimate the likelihood that another individual would
defect; they could participate in routines for distribution of food and other resources and
detect individuals who were abusing the routines. Many of these capacities were based
on special‐purpose cognitive processes aimed at specific types of situation, which, if
evolutionary psychologists are correct (Barkow et al. 1992), still operate in human
thought. But when we consider the effect of the slow accretion of capacities such as these
we can see many ways in which they can scaffold the development of more general‐
purpose capacities, and in (p. 722) particular the attribution of information‐bearing states
like those the philosophers and developmental psychologists have focused on.
Part of one story might go as follows. Proto‐humans engage in cooperative hunting and
foraging which requires them to keep track of what situations are perceivable to which
others. This has a conceptual requirement, that they be able to represent the relation ‘X
sees that p’, and a ‘pragmatic’ requirement, that they be able to figure out what one
might be able to see from position p under current conditions. (These might be taken as
among the precursors of the two folk‐psychological theories mentioned above.) As these
capacities improve, over thousands of years, language develops and with it the capacity
to register reports of what people have seen, or take themselves to have seen, from
places outside one's field of view. Then the eventual combination of these three things—
the concept of perception, the ability to take a perspective, and the understanding of
verbal reports—results in the roots of an understanding of knowledge or belief. (The co‐
evolution of folk psychology and language is something we have to understand in order to
get much further with these questions. The topic is very promising, but also very hard to
handle responsibly.)
A similar story could be told for the evolution of the capacity to use attributions of belief
and desire to anticipate behaviour. It would be a separate story, though, requiring a
different, if overlapping, set of social skills as its starting point. (I would conjecture that
crucial to this second story might be the capacity to manipulate others with misleading
information, and to resist such manipulation.) Even when we combine the origin‐of‐belief
story and this practical‐reasoning story we do not have a full account, though. We need to
explain how the conceptual and pragmatic components of these and other metacognitive
skills come together to form a relatively integrated body of cognitive routines for
connecting information, desire, emotion, and all the other states we attribute. It is
plausible that this integration could not happen without language and its specific
vocabulary for states of mind and more particularly its resources for using attributions in
explanations of actions.
There are alternative stories, and the present evidence is not going to decide between
them. The important point, though, is that we can see in principle how the evolution of
folk psychology is possible. The crucial element is ‘niche construction’ (Sterelny 2004: ch.
8): the way in which a set of attributes can permit an animal, particularly a social animal,
to change its environment in ways that allow other attributes to evolve. In the case at
hand the first attributes are those that allow specific forms of cooperative activity, the
Page 12 of 18
Folk Psychology
niche is stable social life in which more effective cooperation and resistance to
exploitation pays off, and the other attributes are the components of folk‐psychological
competence.
Page 13 of 18
Folk Psychology
justified, or worth holding on to? It is not immediately obvious that these questions get
answered with Yes. Compare crude folk meteorology. People look at the sky and smell the
wind and consider the behaviour of plants and animals, and by instinct and traditional
lore come up with opinions about whether it is a good time for the harvest. Sometimes
they are right and sometimes they are wrong. They are wrong significantly often. The
explanations they give have little to do with cold fronts or wind patterns in the
stratosphere. Often supernatural or simply superstitious elements play an important role.
It is far from obvious that the old‐time farmer has a reasonable belief that the crops
should be taken in early this year. Are our beliefs about one another any different?
Two basic considerations pull in opposite directions here. On the one hand, our attempts
to understand one another are largely successful, in that human social life and human
science flourish, and without an effective folk psychology they would be impossible. Isn't
success it's own justification? On the other hand, there is a now enormous body of social‐
psychological experiment which shows how wrong we often are about one another and
ourselves. Among our failings, we think that people are more likely to repeat patterns of
behaviour than in fact they are (Nisbett and Ross 1980); we underestimate how much
people's opinions are affected by their social situations (Festinger 1964); we are often
completely wrong about the reasons for which we have made choices (Nisbett and Wilson
1977; Kornblith 1989; Gopnik 1993); and we overestimate the accuracy of people's
memory for details of events they have experienced (Loftus 1979; Conway 1997). We
seem to make many, and systematic, mistakes.
Focus first on true beliefs. Human social competence entails that very often we anticipate
one another's actions correctly. And it is not pushing this fact very much further to
conclude that we are often accurate about other people's emotions (e.g. whether they are
enraged or conciliatory) and about the information on which they are acting (e.g. whether
someone has seen the letter which would bring about rage and confrontation). Accuracy
about these things is consistent with error in other attributions that we make along the
way to them. In particular, we may be right about what someone will do and wrong about
why they will do it. The reasoning we attribute to someone in the course of predicting
their action may contain thoughts the person will not think, and more profoundly may
represent as causes of the person's action thoughts that are merely incidental. In fact, the
psychological data on errors of attribution suggests that we are much less accurate about
why we (and others) think what we do than about what we think. In the domain of mind,
our explanations are even more suspect than our beliefs. (A potentially sceptical
possibility is that our social competence relies less than we think on our beliefs, also, and
Page 14 of 18
Folk Psychology
instead on specific social skills for whose success folk psychology narrowly construed
takes undeserved credit, see Morton 2002: ch. 1; Bermúdez 2004.)
lie somewhere between these extremes. And anywhere between these extremes we are
going to be faced with a delicate mixture of true and false beliefs, very often with true
beliefs derived from false beliefs, and very often with roughly reliable routes to
conclusions about other people producing numbers of false beliefs as by‐products. Is this
an acceptable situation? Is it one that we should or can (p. 725) try to change? If one is
committed to an internalist epistemology one will find the situation uncomfortable, even
if perhaps inevitable. For an externalist, though, it is a fairly typical example of the price
we have to pay to get beliefs that serve reliably for particular purposes.
References
Barkow, J., Cosmides, L., and Tooby, J. (1992) (eds.), The Adapted Mind: Evolutionary
Psychology and the Development of Culture (Oxford: Oxford University Press).
Bermúdez, J. (2004), ‘The Domain of Folk Psychology’, in A. O'Hear (ed.), Mind and
Persons (Cambridge: Cambridge University Press), 25–48.
Byrne, R. (1995), The Thinking Ape: The Evolutionary Origins of Intelligence (Oxford:
Oxford University Press).
Byrne, R., and Whiten, A. (1988) (eds.), Machiavellian Intelligence: Social Expertise and
the Evolution of Intelligence in Monkeys, Apes, and Humans (Oxford: Oxford University
Press).
Conway, M. (1997), Recovered Memories and False Memories (Oxford: Oxford University
Press).
Festinger, L. (1964), Conflict, Decision and Dissonance (Palo Alto, Calif.: Stanford
University Press).
Page 16 of 18
Folk Psychology
Gopnik, A. (1993), ‘How We Know Our Minds: The Illusion of First‐person Knowledge of
Intentionality’, Brain and Behavioral Sciences, 16: 1–14.
Gopnik, A.A., and Meltzoff, A. (1997), Words, Thoughts, and Theories (Cambridge, Mass.:
MIT Press).
Heal, J. (1995), ‘How to Think About Thinking’, in M. Davies and T. Stone (eds.), Mental
Simulation: Evaluations and Applications (Oxford: Blackwell), 33–52.
Nichols, S., and Stich, S. P. (2003), Mindreading (Oxford: Oxford University Press).
Nisbett, R., and Ross, L. (1980), Human Inference: Strategies and Shortcomings of Social
Judgment (Englewood Cliffs, NJ: Prentice‐Hall).
Nisbett, R. E., and Wilson, T. (1977), ‘Telling More Than We Can Know: Verbal Reports on
Mental Processes’, Psychological Review, 84: 231–59.
Povinelli, D., and Eddy, T. (1996), ‘What Young Chimpanzees Know About Seeing’,
Monographs of the Society for the Study of Child Development, 61: 1–152.
Putnam, H. (1960), ‘Minds and Machines’, in S. Hook (ed.), Dimensions of Mind (New
York: New York University Press), 148–79.
Quine, W. V. (1953), ‘Two Dogmas of Empiricism’, in Quine, From a Logical Point of View
(Cambridge, Mass.: Harvard University Press), 20–46.
Page 17 of 18
Folk Psychology
Sellars, W. (1956), ‘Empiricism and the Philosophy of Mind’, in H. Feigl and M. Scriven
(eds.), The Foundations of Science and the Concepts of Psychology and Psychoanalysis
(Minneapolis, Minn.: University of Minnesota Press), 253–329.
Wellman, H. (1990), The Child's Theory of Mind (Cambridge, Mass.: MIT Press).
Wimmer, H., and Perner, J. (1983), ‘Beliefs About Beliefs: Representation and
Constraining Function of Wrong Beliefs in Young Children's Understanding of Deception’,
Cognition, 13: 103–28.
Adam Morton
Adam Morton has a Canada Research Chair in Epistemology and Decision Theory at
the University of Alberta. He is currently working on intellectual virtues of
adaptation to one's own and human limitations, while writing on imagination and
morality in a scattered way that he hopes eventually to bring together. He is the
author of seven books, most recently On Evil (Routledge 2004) and The Importance
of Being Understood (Routledge 2002).
Page 18 of 18
Other Minds
Print Publication Date: Jan 2009 Subject: Philosophy, Metaphysics, Philosophy of Mind
Online Publication Date: Sep 2009 DOI: 10.1093/oxfordhb/9780199262618.003.0043
It is a serious question both in philosophy and the history of ideas just how our
conception of mind has evolved. There is something of a consensus that Descartes
introduced a particular conception and one, interestingly, that can be seen to take shape
against the backdrop of an evolving scientific conception of the world. This scientific
account of the world had no room for mind, and Descartes helped things along by hiving
off the mind, so to speak, into its own realm. In Descartes's world-view mind and body are
distinct, and distinct substances.
Keywords: conception of mind, scientific conception, conception of the world, Descartes, mind and body,
behaviour, mental antecedents
THE topic is other minds. Shall we say, the problem is that of other minds? Yes and no. Yes,
if what is meant is that there is an issue here that raises philosophical perplexity and
must be addressed. No, if what is meant is that there is a specific problem that can be
formulated thus: How can I know that there are other minds? Recent study in the
philosophy of mind has tended either to overlook or to attempt to downplay issues here.
And where there is engagement with the issues, there is little agreement even about what
the question is that ought to be occupying philosophers. This is reflected in two recent
books devoted to the topic: Hyslop (1995) (see also Hyslop and Jackson 1972) and
Avramides (2001).1 The author of the first, Alec Hyslop, claims that there is a problem
about how we justify our belief that there are other minds, and argues that that
justification is to be achieved through analogical reasoning. In proposing this solution
Hyslop here is following a traditional line—albeit one that has fallen out of favour in
recent decades. In the second book, I take the line that there is no epistemological
problem concerning the existence of other minds, although there is a philosophical
perplexity that must be addressed and there is much that needs to be said about why so
many philosophers think that there is such a problem here. The philosophical perplexity
Page 1 of 17
Other Minds
that arises concerns how we are to understand the concept of mind as one general
enough to cover others as well as oneself.2
Questions about others have rarely been to the fore in philosophical discussion. Rather,
comments on the topic of others, appended to discussions of other issues, (p. 728) are to
be found from time to time—especially since the work of René Descartes. Discussions in
more recent times have tended to take their cue from some remarks by John Stuart Mill.
One could identify a Cartesian tradition within which problems about our knowledge of
the mind of another arise. Mill is located in this tradition, as is Alec Hyslop. Pace Hyslop,
the analogogical argument has been largely rejected in recent times in favour of a more
contemporary scientific approach and an argument from best explanation (see, for
example, Pargetter 1984). The argument from best explanation can be placed alongside
the analogical argument as a reply to a question concerning our knowledge of the mind of
another, which question arises within the Cartesian tradition with its very particular
conception of mind. There have, however, been those who deny that the question to which
these forms of argument purport to be a reply is a real question. These philosophers
argue that this question only arises because a very particular conception of mind has
been adopted. Furthermore, reasoning of the sort employed by these forms of argument
can only issue in conjectures or hypotheses—albeit highly confirmed conjectures or very
probable hypotheses. What it cannot do is issue in the sort of certainty that some think is
requisite when it comes to our relations with others. Philosophers of this school would
argue that the existence of others is foundational to all our reasoning and so cannot be
the result of any reasoning.
It may be thought overly tendentious to claim that there is no question about our
knowledge of another mind. Surely questions about others arise all the time—
philosophical discussions aside. In my daily life questions arise about whether another
feels as much pain as I do, and in some cases I can question whether another feels any
pain at all. We have all questioned from time to time whether the colour one sees is the
same as that which others experience. And the medical profession is bedevilled by
questions concerning the experiences of patients who are paralysed or in a coma.
Children—and infants in particular—can leave us pondering about how their mental lives
square with ours. Can questions here really be dismissed? Not at all. But we must be
careful what question is being asked. Philosophers start with these questions but soon
find themselves asking how I can know that there is any mind other than my own. Indeed,
it is often argued that the very possibility of the former sorts of question brings in its
wake the more radical question and, thereby, raises the possibility of solipsism—of my
aloneness in the universe. If we want to forestall the question of solipsism, we need to do
so in such a way that allows for questions of the less radical sort to arise.
It is a serious question both in philosophy and the history of ideas just how our
conception of mind has evolved. There is something of a consensus that Descartes
introduced a particular conception and one, interestingly, that can be seen to take shape
against the backdrop of an evolving scientific conception of the world. This scientific
Page 2 of 17
Other Minds
account of the world had no room for mind, and Descartes helped things along by hiving
off the mind, so to speak, into its own realm. In Descartes's world‐view mind and body are
distinct, and distinct substances. As Gilbert Ryle was to comment four centuries later:
When Galileo showed that his methods of scientific discovery were competent to
provide a mechanical theory which should cover every occupant of space,
Descartes found in himself (p. 729) two conflicting motives. As a man of scientific
genius he could not but endorse the claims of mechanics, yet as a religious and
moral man he could not accept, as Hobbes accepted, the discouraging rider to
those claims, namely that human nature differs only in degree of complexity from
clockwork.
(1949: 20)
Descartes's solution was to introduce, alongside physical sorts of things, mental sorts of things.
Just as there is physical ‘stuff’, and physical events, states, processes, and the like, so Descartes
posited that there is mental ‘stuff’, and mental processes, events, states, and the like. The latter
consists of, inter alia, beliefs, desires, sensations, and they are taken to work just like physical
processes, states, and events: they cause other processes, states, and events. Mental causation
may seem in evidence; however, the understanding of this process has eluded intellectual grasp,
and one of the outstanding problems of this Cartesian legacy is to understand this process of
mental causation. In particular, we need to understand how ‘stuff’ so different as the mental and
the physical can interact in so intimate a way that, as Descartes remarked, ‘I am not merely
present in my body as a sailor is present in a ship’ (1641/1984: 56).3 There is a particular
closeness here, and this needs to be reflected in our understanding of mind in relation to body.
Once the mind and body have been separated in this Cartesian way, questions arise not
just concerning the mind's relation to body, but also concerning the mind's relation to
other minds. Descartes's conception of mind and world is such as to raise the radical
epistemological questions: ‘How do I know that the world of bodies external to my mind
exists?’ and ‘How do I know that there are any minds other than my own?’.4 The second
of these questions is sometimes masked by the fact that Descartes—like many other
philosophers—is prone to talk of our knowledge of the external world. Cursory inspection
of his writings reveals, however, that he is not entitled to the plural pronoun here. What is
most significant is the fact that throughout his Meditations, for example, Descartes writes
entirely from the first‐person perspective. And the Archimedean point from which he
attempts to draw the whole world is captured in the proposition I am; I exist. In so far as
there is a world of bodies or minds external to Descartes's own, this is something that
requires argument. Unlike his observation about the relation of mind to body, what
Descartes failed to note is that there is also a particular sort of closeness involved in the
relation of one mind to another.
This Cartesian conception and its repercussions for our knowledge of another mind are
nicely captured by Augustine, a philosopher who, in his thinking about mind, is often said
to have ‘anticipated’ Descartes. Augustine writes:
Page 3 of 17
Other Minds
‘Know thyself’ is not said to the mind as ‘Know the cherubim and seraphim’; for
they are absent, and we believe concerning them, according to that belief they are
declared to be certain celestial powers. Nor again as it is said, ‘Know the will of
that man’, for it is not within (p. 730) our reach to perceive at all, either by sense
or understanding, unless by corporeal signs set forth
The problem, as Augustine sees it, is that we cannot perceive another's will. And Augustine
explains that the case of another is very different from one's own. He writes:
But when it is said to the mind ‘Know thyself’; then it knows itself by the very act
by which it understands the word ‘thyself’; and this for no reason than that it is
present to itself.
This division between what is possible in one's own case and what is not possible in the case of
another is repeated throughout the history of philosophy.5
The problem seems clear. All these philosophers, like Descartes, begin with themselves,
with the first person. They find that there is one thing that is known intuitively and, thus,
with certainty. This is either myself, or my existence. And what is then observed, quite
correctly, is that there cannot be a similar relation to another. I cannot know another
immediately and intuitively; I cannot know another in this way. Nor can I perceive the
mind of another. When it comes to knowing about the mind of another there is a question
to be answered, and it is not at all clear how to go about answering it. And the question is
a radical sceptical one. A positive reply to this question is required as a bulwark against
solipsism. Sometimes it is acknowledged that a reply to this radical question will still
leave us with questions concerning the particular contours of this other mind: just what
the other is thinking and feeling (call this ‘weak scepticism’). Sometimes, however, it is
allowed that the argument that justifies our belief in the existence of another mind
justifies, as well, our belief that that mind is (roughly) like our own. Very occasionally in
the history of the topic one finds the weak sceptical question raised but not the more
radical one. An example of this can be found in the work of one of Descartes's immediate
successors, Nicolas Malebranche. Unlike Descartes, Malebranche clearly separates out
three kinds of knowledge: knowledge of body, of one's own mind, and of other minds. And
concerning our knowledge of another mind—which he takes it that we know about ‘by
conjecture’—Malebranche makes an important distinction between our knowledge of
another's sensations and of his thoughts (see esp. 1674/1997: bk. III, pt. ii, ch. 7, sect. V,
p. 239). In connection with both sensations and thoughts, however, the conjecture that
Malebranche has in mind seems clearly to do with what the other is feeling and thinking.
Malebranche is not asking the radical sceptical, ‘How do I know that another mind
exists?’. This is especially notable as the parallel question concerning the existence of
bodies is clearly identified by Malebranche (see esp. 1674/1997: (p. 731) bk. III, pt. ii).
According to Charles McCracken, Malebranche acknowledges that there is a crucial
difference between our knowledge of the existence of bodies and our knowledge of the
Page 4 of 17
Other Minds
existence of minds: while it may be possible to deny the former, it is not possible to deny
this possibility. McCracken suggests that the latter follows from the fact that
Malebranche holds that in the case of minds or souls I am assured of their existence from
consideration of the ‘whole moral order’ (1983: 81). Malebranche is astute in many of his
observations here, but I take his greatest insight to be an understanding of the very depth
and significance of our relation to others. Despite it being possible—indeed, necessary—
to raise questions about our knowledge of the nature of another mind, it is difficult to
imagine questioning the existence of other minds. With this observation Malebranche
acknowledges something that (it is arguable that) Descartes missed: my relations to
others are such that to question them—to question the existence of others—may not be a
real possibility.
The argument proceeds as follows: (i) I know that I have a mind; (ii) I observe that my
mind is the antecedent condition for my behaviour; (iii) I observe behaviour in the bodies
of others similar to that which I know in my own case to be the result of mental
antecedents; I conclude, by analogy, that this behaviour in others is the result of the
existence of other minds.7 I mentioned the argument from analogy earlier, as both a
traditional response and one lately adopted by at least one prominent contemporary
writer on the subject. What is not clear is whether this argument can address the
philosophical question of solipsism or deal with radical sceptical issues.
Arguments from analogy are not only proposed in connection with the mind; they are
used in the sciences. In an early paper Hyslop gives the following example of the use of
analogy in science: Helium has a characteristic spectrum, the nature of which was
determined by experiments carried out on earth. It was then observed that light from the
sun displayed this spectrum, and it was inferred that helium was present on the sun
(Hyslop and Jackson 1972: 169). Thomas Reid, in the mid‐eighteenth century, discussed
the use of analogy in the study of the planets—as well as medicine (p. 732) and politics.
This use he took to be unobjectionable, but Reid insisted on a contrast between our
thinking about the planets and our thinking about the mind. Reid's point is that in the
case of mind reasoning by analogy (along with what he called hypothesis and conjecture)
yields unstable and insecure foundations for knowledge (1785/1969: vol. I, ch. iii, p. 41).
Page 5 of 17
Other Minds
According to Reid, stability and security are the result of first principles, on which more
below. Where Hyslop and others see a parallel between our thinking concerning scientific
matters and our thinking concerning others, Reid notes an important difference.
Reid's objection to employing analogy when the subject is mind is insightful. It is echoed
in a comment by F. H. Bradley, who is reported to have held that we don't want inferred
friends who are mere hypotheses to explain physical phenomena (see Urmson 1982: 61).8
Although this objection has sometimes been raised in more recent times to the argument
from analogy, philosophers tend to emphasize other objections. One of these other
objections also draws on a disanalogy with the use of the argument in the sciences: it is
pointed out that in the case of another mind but not in the sciences the conclusion drawn
is logically uncheckable. A second objection is that in the case of mind but not in the
sciences the argument proceeds from one case alone. Hyslop discusses, and attempts a
defence against, these two objections in some depth. There is also a third objection, that
arises out of the work of Ludwig Wittgenstein. According to this objection the argument
from analogy goes wrong not in the way it derives its conclusion but in the formulation of
its very first premise: I know that I have a mind. The problems here are various and
complex. One way to formulate them is due to Norman Malcolm (1966) and draws on
considerations from Wittgenstein's private‐language argument. Another is due to P. F.
Strawson and draws on considerations concerning our concept of mind (see esp.
Strawson 1959). Around the mid‐twentieth century there was a heightened interest in the
problem of other minds, and a considerable number of those writing on the topic thought
that the Wittgensteinian considerations of the third objection pretty much demolished the
argument from analogy.9 In any case, for those who did not favour this objection, the
other two objections were considered sufficient for the argument to lose favour.10
Those who object to the argument from analogy on the grounds that our model should not
be that of the physical sciences will be even less impressed by the argument that has
come to replace it in many contemporary circles: the argument from best explanation. It
is fair to say that at the start of the twenty‐first century this argument is either explicitly
or implicitly held by the majority of philosophers. The argument was championed in an
early paper by C. S. Chihara and J. A. Fodor (1966). Chihara and Fodor's paper was
designed to diminish the impact that Wittgenstein's writing was taken to have had on the
problem of other minds, while at the same time to promote another—more scientific—
approach to this problem. Chihara and (p. 733) Fodor claim that Wittgenstein was right to
point out that we learn conceptual connections when we learn a language. However,
where they saw Wittgenstein as arguing that those connections are between words like
‘pain’ and characteristic patterns of behaviour, they proposed that the real conceptual
connection is between our concept of one mental state and that of a wide variety of other
mental states. Chihara and Fodor argue against (their understanding of) Wittgenstein's
post‐Tractatus writings on philosophical psychology—against, that is, what they label
Wittgenstein's ‘logical behaviourism’ and his ‘operationalist’ approach to meaning.11
Chihara and Fodor take as their starting point the Wittgensteinian idea that the
connection between, say, pain and pain behaviour is ‘criterial’. The notion of a criterion is
contrasted by Wittgenstein with that of a symptom (see esp. 1958: 24–5); the introduction
Page 6 of 17
Other Minds
(1966: 413)
Page 7 of 17
Other Minds
Instead of taking behaviour as criterial of pain, Chihara and Fodor hold that the behaviour on
the basis of which mental predicates are applied might properly be thought of as supplying
evidence for the existence of the mental processes we postulate. In this respect behaviour
should be thought of as analogous to the Wilson cloud‐chamber track on the basis of which we
detect the presence and motion of charged (p. 734) particles, and our interrelated mental
concepts should be thought of as analogous to the physical theory in which the properties of
these particles are formulated. On this way of looking at things, the belief that other people feel
pain ‘provides the only plausible explanation of the facts I know about the way that they behave
in and vis à vis the sorts of situations I find painful’ (1966: 414). Chihara and Fodor insist that
this argument is not committed to the idea that the child learns what pain is from his or her own
case. They suggest that it would be possible for a child to learn the concept of pain before—or
even in the absence of—experiencing pain. What the child would have learned is the complex
web of conceptual connections including the concept of pain that goes to understanding and
explaining another's behaviour. Furthermore, Chihara and Fodor follow Chomsky and suggest
that the system of mental concepts that the child comes to have may not require explicit
teaching or training in the use of language but may develop quite naturally in the child in
response to very little training. In other words, it may be innate. They point out that this way of
looking at things has the interesting result that there is no asymmetry in the use of mental
predicates. Whether I use the predicate to apply to myself or to another, I am applying a
predicate that picks out a state or process that interacts with other mental states and explains a
certain way of behaving.
Chihara and Fodor offer a powerful and radical challenge to an equally powerful and
radical idea that can be found in Wittgenstein's post‐Tractatus writing on the philosophy
of psychology. As arguments from best explanation have come in the sciences to replace
arguments by analogy, it is not surprising that this is the argument that has come to find
favour in the philosophy of mind. It is thought that this new argument avoids some of the
objections that were considered to have sunk the earlier argument from analogy. In
particular, it is claimed that this argument does not rely on extension from one's own
case. (Although this in itself raises the equally difficult issue of whether such an
argument can be thought to be true to the first‐person perspective.12) What is not clear is
that the argument from best explanation best addresses the issues that Wittgenstein was
concerned to address.
The main part of my complaint against Chihara and Fodor (as well as Fodor in some of his
subsequent, solo, discussions of these issues) is that the approach is wholly scientific.
This is bound to yield difficulties when the approach is taken with a text that professes a
fundamental distinction between science and philosophy. The difficulties affect the
interpretation of Wittgenstein that one finds in Chihara and Fodor's work, as well as the
alternative proposal for our understanding of mind that we find there. Of course, Chihara
and Fodor hope to break down this distinction.
I am not sure the problem doesn't begin with Chihara and Fodor's discussion of a
criterion. Like so many others, Chihara and Fodor cannot comprehend what Wittgenstein
is looking for here, and they are mesmerized by the undeniable fact that one can have, for
example, pain behaviour in the absence of pain and vice versa. They (p. 735) cannot see
any way to acknowledge this fact without running away with the point and they end up
having to hold that one can make sense of the idea of pain in the absence of any
Page 8 of 17
Other Minds
behavioural manifestation. The problem with saying this is that it completely divorces the
idea of pain—or, more generally, a mental phenomenon—from activity, and it allows for
the possibility of our aloneness in the universe. And the two go together. Chihara and
Fodor, however, divorce the two: their account of mental activity—of the mind as the
activating cause of behaviour—leaves in place the question of our aloneness in the
universe. The only answer they can give to this question is that it is highly improbable
that we are alone.
Sceptical questions are not to everyone's taste, but they do lie at the heart of many
important philosophical issues. They are clearly not to Fodor's taste (see, for example,
Fodor 1994: 292), but they did concern Wittgenstein. In one of the very few places where
Fodor does address the sceptic he writes that what he is committed to is ‘minimal
scepticism’. Fodor says that we can distinguish logical from moral certainty, and that it is
the latter—not the former—that can be achieved concerning another's (state of) mind
(1968: 60–3).13 What Fodor is mainly concerned to distance himself from is the idea that
we can achieve logical certainty concerning another. In response to the sceptical question
of whether we can know that there are other minds or whether we have a reason to
suppose that another has a mind, Fodor writes: ‘it would be pertinent to inquire into his
use of know’, or ‘to inquire into what his concept is of “having a reason” ’ (1968: 61).
What Fodor does not do here is explain what is his use of ‘know’. One thing that is clear is
that Fodor's use of ‘know’ is such that one can be said to know even where there is the
possibility of radical scepticism concerning the mind of another. Solipsism may be
‘absurd’, but it is possible given Fodor's concept of mind (1968: 64–6). Fodor and Chihara
emphasize that aspect of Wittgenstein's notion of a criterion involved in telling
(recognizing, seeing) that another is in pain on the basis of the presence of a certain
behaviour. And Fodor rightly understands that Wittgenstein is concerned with a kind of
certainty concerning other minds—a kind of certainty that does not allow for any (logical)
distance between one mind and another. This is the certainty—or better, the assurance—
one person has vis‐à‐vis another when he runs to help when there has been a serious
accident. There are circumstances in which doubt is out of place. And where doubt is out
of place so is the radical sceptical question. This is not just a matter of pragmatics. (We
should recall that Fodor was an early champion of Grice against Wittgenstein here.) It is
not just what we would say, but what this tells us about what we are talking about. Fodor
—and so many philosophers working in philosophy of mind today—take it that when we
talk about mind we are talking about something that acts as the characteristic cause of
the behaviour it is invoked to explain. On this conception of mind, all the observable
behaviour could be present while the mind is absent. Of course, this is highly unlikely (or
God most probably did not arrange things in this way), but it is not impossible—not
impossible given the sort of thing the mind is. If, however, we take behaviour in relation
to the world and (p. 736) to others as our starting point, this could not be the case. Where
Fodor emphasizes what he takes to be Wittgenstein's insistence on a logical connection
between mind and behaviour, I would emphasize a conception of mind which is such that
it makes no sense to speak of mind in the absence of such behaviour. There is nothing to
tie mind to behaviour, and hence one mind to another, on Fodor's way of thinking about
Page 9 of 17
Other Minds
the mind. It may seem absurd to deny that there are other minds, but Fodor offers no
understanding of why this is the case (1968).14 What we do know is that it is possible—
logically possible for it to be the case that God made a world that consists of one mind
only. But appeal to God is unnecessary. The starting point is not God, but one mind in
relation to another and the world.15
The certainty this alternative starting point gives us is not best expressed as a certainty
that allows for no error. (After all, no one can deny that there may be instances of
behaviour that are mere behaviour, lacking mind.) The certainty at issue here is best
thought of as a surefootedness vis‐à‐vis others. It is my behaviour towards another in
certain circumstances that leaves no room for doubt—no room for the sceptic to ask his
questions. It is not enough to say I have very good reason to believe that another is in
pain. It is not enough to argue that my theory that he is in pain is the simplest, most
plausible, and predictively adequate explanation of his behaviour. This leaves us at best
with a well‐confirmed hypothesis that the other is in pain. On this picture of things my
surefootedness towards others is merely a matter of the way I am—it tells us nothing
about the deep absurdity of the radical sceptical question that purports to show that this
surefootedness could be radically in error. It is not a matter of knowing that I am not
alone in the universe, it is not a matter of having a good reason to think that I am not
alone. Reid recognized this. Reid writes about the child even before its first birthday who
‘clings to its nurse in danger, enters into her grief and joy, is happy in her soothing
caresses, and unhappy in her displeasure’ (1785/1969: essay I, ch. iii, sect. 41, pp. 633–
4). There is no room for doubt. And this is not to say that the child knows that its nurse is
another mind. As Reid insists, that there are minds in our fellow creatures is a first
principle, and one that must be in place before we even can begin to talk about knowing.
First principles leave no room for the sceptic. Like Wittgenstein after him, Reid directed
much of his intellectual energy to showing the error of scepticism. If one understands the
operation of first principles, one will understand the closeness of our relations to others.
The existence of another mind is not the result of arguing by analogy and hypothesis; it is
a matter of first principles. Reid is clear that it is possible to be wrong about first
principles, but he writes that when first principles are brought into dispute ‘they require
to be handled in a way peculiar to themselves’ (1785/1969: essay I, ch. ii, p. 33). Reid is
here (p. 737) recognizing the important and unique part played by such principles as that
‘there is life and intelligence in our fellow men with whom we converse’. And he
understands that there is something fundamentally misguided about reasoning by
analogy or hypothesis to the existence of this fellow intelligence.
Despite his insights into the place of first principles, however, Reid's work does not go far
enough in its understanding of the sceptic's mistake. What Reid fully understood was the
kind of certainty—the kind of surefootedness—that one intelligence has vis‐à‐vis another.
Furthermore, he understood that one's conviction of the existence of other intelligent
beings is ‘absolutely necessary’ and fundamental to all reasoning and all knowledge
(1785/1969: essay I, ch. iii, 41, p. 635; see Malebranche 1674/1997). But Reid did not link
up this necessity with our concept of mind. It was Wittgenstein who tried to show how
these observations about our relationship to others have repercussions for how we should
Page 10 of 17
Other Minds
understand this concept. Until this conceptual work is done, it is arguable that the sceptic
cannot be fully silenced. The insistence on first principles can look like a dogmatic
assertion of common sense in the face of philosophical objection. It is only when we take
seriously the place of behaviour in our concept of mind that we can begin to understand
that which we find so hard to deny (the importance of our relations to others).
Chihara and Fodor aim to expose the errors of what they called ‘logical behaviourism’.
They insist that mind is only contingently related to behaviour. Theirs is a concept of mind
that leads to insecurity in our relations to others. It is the fundamental nature of our
relations to others that Chihara and Fodor fail to appreciate. And their failure to
appreciate this has led them to believe that the methods of science and theory are
appropriate when it comes to our knowledge of other minds. This is where one ends up if
one naturalizes an essentially Cartesian conception of mind. Perhaps the integration of
mind and world requires that we revise our very conception of mind. When we do this,
the question that lies at the heart of so much thinking about other minds—‘How do I
know that other minds exist?’—fades away. It is not a question of knowledge. It is a
matter of understanding aright our relationship to others. It is a matter of understanding
that others must be taken to be in place before questions of knowledge can arise.
Collectively we can ask questions about the far‐off planets, about subatomic particles,
and the like. But without this community of minds these questions cannot arise. The real
question concerning others is not how I know that they exist; it is, ‘What is my conception
of mind such that I can understand how it can include myself as well as others?’.
If one has to imagine someone else's pain on the model of one's own, this is none
too easy a thing to do: for I have to imagine pain which I do not feel on the model
of pain which I do feel.
Page 11 of 17
Other Minds
Colin McGinn (1984) has suggested that things are ‘none too easy’ here because the concept
gained through experience of things in one's own case bears the ‘inexpungible stamp’ of the
first‐person mode of presentation by which one acquires that concept. The problem is that it is
not at all clear how we are to move from such a concept to one that includes others. This is the
problem of generality. The problem of unity serves to exacerbate the problem of generality, and
is inherent in that problem. The problem here is that we need to understand the generality of
the concept in such a way as to incorporate others while at the same time maintaining that
inexpungible stamp of the first person. It is ‘none too easy’ to see how this can be done. Part of
the work that needs to be done here is to explore how best to understand what McGinn has
characterized as the inexpungible stamp of the first person.
The conception of mind that I am urging be rejected can be glimpsed in the thought—
expressed so well in the writings of St Augustine (see above)—that the mind of another
cannot be perceived. The naturalness of this thought masks a difficulty: what would it be
to perceive another mind? It is not just that another's mind is over there, so to speak,
inside another's body. It is not just that the behaviour I observe blocks my view of the
other's mind. (If only I could peer around it and perceive the other's mind.)17 The problem
lies in trying to comprehend what it is we would be perceiving. As is so often observed,
we do not perceive our own mind. How we are to understand this is not easy to
comprehend, but philosophers have been prone to talk of intuition, immediate knowledge
(again, see above). Yet when it comes to another, we gesture at perception. Perhaps we
should employ the light hand of analogy here and conclude that as we do not perceive our
own mind we should not think of perception in connection with the mind of another. But if
not perception, what? This question bedevils us. Whatever the thinking concerning the
mind of another, many philosophers have understood that in our own case mind is a
matter of agency. Here we may remind ourselves of some things Berkeley has written.
Berkeley distinguishes between ideas and spirits: ‘things so wholly different, that when
we say, they exist, they are known, or the like, these words must not be thought to signify
anything common to both natures’ (1710/1975a: 142). And Berkeley cites as the reason
for this difference the fact that ideas are passive while (p. 739) spirit is active. This
activity that is so evident in connection with mind is the key to our understanding its
nature. It is a real question how we are to understand something whose nature is
essentially active. I am not sure it is sufficient to say that it is the sort of thing that can
serve as the cause of behaviour.
These thoughts, in their turn, put pressure on the idea of causation as the correct way to
understand the relation of mind to behaviour. This is a topic too large to engage in now. It
is enough that we begin to appreciate the pressures here, and the reasons for them.
Another large topic that I do want briefly to touch on is the asymmetry that has been
taken to exist in connection with the way mental predicates apply to oneself and to
others, which several philosophers have highlighted. Alex Hyslop, for example, begins his
book with the statement that the problem of other minds is generated by an asymmetry in
respect of knowledge (1995: 6). Davidson notes that self‐knowledge is unmediated while
knowledge of others is mediated by behaviour. Chihara and Fodor have suggested that it
may be thought to be a virtue of their approach that there is no asymmetry here: mental
predicates apply in both cases as the cause of behaviour. It is arguable, however, that
Page 12 of 17
Other Minds
Chihara and Fodor seek to abolish the wrong problem. It seems hard to deny that there is
some sort of asymmetry here. But it is an interesting question just what the asymmetry
here amounts to. Pace Hyslop (and many others) I do not think the asymmetry is one of
knowledge. (This would follow from my views about the foundational nature of my
relations to others.) There is, however, an asymmetry in the use of mental predicates, in
connection with myself and with others. Not only do I not observe my behaviour when I
speak of my pain, but I don't even think of it. My pain is, in some sense, more immediate
than that, and in connection with this we insist that the questions that can arise here are
severely circumscribed. And no matter how close I am to another, questions continue to
arise over just what she is thinking and feeling. (This is the weak scepticism I referred to
above. It now arises divorced from the question of a more radical scepticism.) It is these
sorts of observations that are reflected in the claim that there is an asymmetry of
knowledge here. But we may, perhaps, reflect on these observations in a different way.
Concerning the asymmetry he observed in connection with the mind, Wittgenstein once
wrote:
(1998: 10e)
Wittgenstein seems here to be offering us another way of thinking about the asymmetry we
observe, and one in keeping with the idea that our attitude here is not cognitive. An asymmetry
lies at the heart of our use of mental terms, and it is something we need to understand. At the
same time we also need to understand how our concept of mind is such that it can apply to
others as well as oneself. As I see it, understanding these things is part and parcel of responding
to Descartes's challenge: how to understand the peculiar closeness of mind and body—a
closeness that is echoed in a closeness that Descartes missed: the closeness of myself to others.
References
Augustine (1974), ‘De Trinitate’, in P. Schaff (ed.), Nicene and Post‐Nicene Fathers of the
Christian Church, iii (Grand Rapids, Mich.: Erdmans).
Page 13 of 17
Other Minds
Hyslop, A., and Jackson, F. C. (1972), ‘The Analogical Inference to Other Minds’, American
Philosophical Quarterly, 9: 168–76.
McGinn, C. (1984), ‘What is the Problem of Other Minds?’, Proceedings of the Aristotelian
Society, suppl. vol. 58: 119–37.
Malebranche, N. (1674/1997), The Search after Truth, trans. and ed. T. M. Lennon and P.
J. Olscamp (Cambridge: Cambridge University Press).
Mill, J. S. (1872), An Examination of Sir William Hamilton's Philosophy, 4th edn. (London:
Longman, Green, Reader, & Dyer).
Nagel, T. (1986), The View From Nowhere (Oxford: Oxford University Press).
Page 14 of 17
Other Minds
Wittgenstein, L. (1958), The Blue and Brown Books (New York: Harper).
—— (1998), The Last Writings on the Philosophy of Psychology: The Inner and the Outer,
ii, trans. C. G. Luckhardt and M. A. E. Ane, ed. G. H. von Wright and H. Nyman (Oxford:
Blackwell).
Notes:
(1) These are the only books to be devoted to this topic since the 1960s.
(2) See Bilgrami (1992), who introduces the problem as a conceptual, and not an
epistemological, one. In The View from Nowhere Thomas Nagel remarked: ‘The
interesting problem of other minds is not the epistemological problem, how I can know
that other people are not zombies. It is the conceptual problem, how I can understand the
attribution of mental states to others’ (1986: 19–20). For a detailed discussion of Nagel
here see Avramides (2001: ch. 9).
(3) Note that simply adopting materialism is no help here—not until such a move can
offer an acceptable explanation of the distinctive character of mind.
(4) Donald Davidson has written: ‘If there is a logical or epistemic barrier between the
mind and nature, it not only prevents us from seeing out; it also prevents a view from the
outside in’ (1991: 154).
(5) See, for example, George Berkeley, who writes concerning others: ‘A human spirit or
person is not perceived by sense, as not being an idea; … Hence it is plain, we do not see
a man—if by man is meant that which lives, moves, perceives, and thinks as we
do’ (1710/1975a: 148). Concerning oneself Berkeley writes: ‘I know what I mean by the
terms I and myself … and I know this immediately, or intuitively, though I do not perceive
it as I perceive a triangle, a colour, or a sound’ (1713/1975b: 232).
(7) I base this formulation of the argument on a passage from the writings of John Stuart
Mill (1872: 243–4).
(8) Urmson claims that Bradley made this claim in response to the argument from
analogy, although he does not cite the reference.
(9) Bilgrami (1992) ignores altogether the first two of these further objections to the
argument from analogy and concentrates on variations of the third, Wittgensteinian,
objection.
(10) Hence the need for a book like Hyslop's by the 1990s.
Page 15 of 17
Other Minds
(11) Thus, they write: ‘To hold that the sceptical premiss is false is ipso facto to commit
oneself to some version of logical behaviourism where by ‘logical behaviourism’ we mean
the doctrine that there are logical or conceptual relations of the sort denied by the
sceptical premiss'; and, ‘It is clear that Wittgenstein thought that analysing the meaning
of a word involves exhibiting the role or use of the word in the various language‐games in
which it occurs … This notion of analysis leads rather naturally to an operationalistic view
of the meaning of certain sorts of predicates’ (1966: 387, 388).
(12) A difficulty, curiously enough, that many philosophers are inclined to raise as a fairly
decisive objection in connection with what they see as Wittgenstein's ‘behaviourism’. It is
interesting to note that this difficulty may also arise in connection with Chihara and
Fodor's work.
(13) Descartes, too, writes of ‘moral certainty’, which he describes as ‘sufficient … for
application to ordinary life’ (1641/1984: 289–90).
(14) To be fair, Fodor correctly insists that there is no direct line from ‘S is certain’ or
‘The denial of S is absurd’ to ‘S is necessary’ or ‘S is true a priori’. My complaint is that
Fodor's examples, which allow for alternative understandings of ‘the denial of S is absurd’
don't cut very deep. They certainly do not help us to understand the sort of absurdity
involved in denying the existence of others.
(15) The work of Descartes and Davidson illustrates these two positions: where Descartes
relies on God to guarantee our knowledge of the external world, Davidson posits a
triangle consisting of one's own mind, another mind and the world (see esp. Descartes
1641/1984; Davidson 1991).
(16) For a detailed of discussion of these issues see Avramides (2001: pt. 3).
Anita Avramides
Anita Avramides is Fellow and Tutor, St. Hilda's College, University of Oxford.
Cowley Place, Oxford, OX4 1DY.
Page 16 of 17
Introspection
Keywords: introspection, mental states, conscious states, self-knowledge, conscious thinkings, sensations
Page 1 of 30
Introspection
IT is a bleak winter day. I am gazing out of the window as I sit at my desk. A friend looks
over and asks me what I am thinking. I reply: ‘I'm thinking that a holiday in the sun would
be nice’. This avowal seems to be a report on a current mental state of mine, of which I
am aware. I am not only thinking that a holiday in the sun would be nice, but know what I
am thinking, and I seem to know it in a way that my friend does not. How do I know what
I am currently thinking?
There is considerable disagreement, however, amongst those working in the area of self‐
knowledge—even amongst introspectionists, as we shall see—whether such knowledge
possesses the features just mentioned, and if so, to what extent. Partly by way of clearing
the ground for discussion of the introspectionist position, therefore, I begin, in Section
43.1 below, by characterizing two classical but radically opposed positions on self‐
knowledge, one broadly Cartesian (Descartes 1641/1984) and the other Rylean (Ryle
1949). Whereas the Cartesian position takes certain self‐knowledge to be distinctive in
the ways described above, the Rylean one denies that self‐knowledge is distinctive in any
way at all. I set out some reasons that might be adduced against the Rylean position,
delaying discussion of the Cartesian one until Section 43.3. In Section 43.2 I discuss some
examples of ‘deflationary’ positions on self‐knowledge, ones that take some such
knowledge to be distinctive in at least some of the ways mentioned above (specifically, in
being authoritative) but reject the view that there is a special method or way by which
one obtains such knowledge. I give some reasons why these positions might be thought to
be unsatisfactory. Finally, in Section 43.3 I revisit the Cartesian position along with other
introspectionist positions distinguishable from it, and conclude by recommending my own
position over the others.
Since the case for an introspectionist position is intuitively stronger for knowledge of
one's own sensations than for knowledge of one's own propositional attitudes, even when
Page 2 of 30
Introspection
these are currently consciously undergone, my focus in what follows will be, by and large,
on how introspectionist accounts of self‐knowledge of current conscious propositional
attitudes fare.
Descartes held that we know some of our intentional states, namely those that we are
consciously undergoing while we are thinking about them, in an epistemically direct and
authoritative way.3 His paradigm for this kind of knowledge was the cogito, which
includes not just thoughts like I am now thinking, but ones like I am thinking that water is
transparent. Descartes thought that he had this special kind of knowledge because of the
special epistemic relation he bore to his thoughts while he was thinking them. Many have
construed this epistemic relation as being at the very least immediate, in the sense of
being non‐evidence‐based (see Alston 1971; Davidson 1984, 1987, 1988; Burge 1985,
1988; Heil 1988, 1992; Wright 1989).4 This immediacy is thought to extend beyond
knowing that one is in a given intentional state—say, a state of thinking—to knowing what
the content of that state is—say, a state of thinking that water is transparent.
Page 3 of 30
Introspection
(p. 744)
The mere fact that I think my thoughts whereas others do not cannot be the answer, true
though this might be, since it does not by itself explain my favoured position with regard
to knowledge of the contents of these thoughts. It may be that one cannot think a thought
without thinking its content. But thinking a content and knowing that a thought has that
content are distinct matters. Further, since not all knowledge is authoritative, knowing
authoritatively that a thought has the content it does is another matter still.
The Cartesian view is that it is the epistemically direct or immediate nature of the
relation between the subject of the second‐order thought and the first‐order thought
reviewed in thinking that second‐order thought which explains the authoritative nature of
cogito‐type thoughts. This view is associated with a particular conception of the mind,
according to which it is a kind of inner theatre, viewable by a kind of ‘inner
eye’ (McDowell 1986, 1998; Wright 1998). The immediate objects of one's thoughts are
‘inner’: mental phenomena such as sensations, perceptual experiences, and current,
conscious thinkings. By attending to these so‐called inner objects, one can know both
one's own mind and what seems to be the case in the world beyond one's mind. Further,
the existence and the nature not only of one's sensations but also of one's contentful
thoughts is independent of what may or may not exist beyond one's mind.
The Cartesian conception of the mind is associated with a commitment to the view that
subjects have privileged access to their own current, consciously entertained thoughts.
This is so not just in the sense that they are in a better position than others to ‘view’
them, and so to know them as the thoughts they are, but also in the stronger sense that
their knowledge of such thoughts is either incorrigible, in that they cannot be shown to
be mistaken about them, or infallible, in that they simply cannot be mistaken about them,
or both.
For the Cartesian, the special status of such thoughts that derives from the immediate,
non‐evidence‐based relation between subject and thought makes for a kind of
transparency of the thought reviewed to the reviewing subject, but not to others, where a
transparent thought is one that is both infallible and certain, or indubitable, in that it is
Page 4 of 30
Introspection
impossible for there to be any grounds for doubting that one has it and what its content is
(Alston 1971). This transparency is due to the special method by which a subject has
access to the contents of her own intentional states, which differs from others' access to
those contents, and this is what confers an epistemic advantage on the subject in the
sense of better placing her to know what the contents of those states are. The Cartesian
account is thus a sort of introspectionist account of self‐knowledge, what we might call a
direct observational one (which we will revisit in Section 43.3).
This view contrasts sharply with the Rylean one. For the Cartesian, the special status of
self‐knowledge is a datum that serves as a secure basis on which to construct a theory of
knowledge in general. For the Rylean, however, self‐knowledge does not have any
distinctive status. Unlike the Cartesian, the Rylean denies that there are features peculiar
to self‐knowledge that need accounting for by any theory (p. 745) of self‐knowledge (as
distinct from a theory of knowledge in general). As Ryle himself puts it:
The sorts of things that I can find out about myself are the same as the sorts of
things that I can find out about other people, and the methods of finding them out
are much the same. A residual difference in the supplies of residual data makes
some differences in degree between what I can know about myself and what I can
know about you, but these differences are not all in favour of self‐knowledge. In
certain quite important respects it is easier for me to find out what I want to know
about you than it is for me to find out the same sorts of things about myself. In
certain other important respects it is harder. But in principle, as distinct from
practice, John Doe's ways of finding out about John Doe are the same as John
Doe's ways of finding about Richard Roe.
(1949: 155–6)
What ‘privileges’ self‐knowledge over knowledge of others, when it does, is just the fact that we
are more often able to observe our own behaviour than we are able to observe the behaviour of
others, and so we have more data of the same kind (rather than data of a different kind) on
which to base such knowledge.
Ryle's main objection to the view that self‐knowledge is distinctive is to the idea of
privileged access, or a special way of knowing one's own inner states, of the kind
presumed by the Cartesian; namely, a sort of ‘inner’ perception of one's mental episodes.
His objection is that if self‐knowledge involved a second‐order perception‐like awareness,
or consciousness, of a first‐order state, this second‐order awareness would itself have to
be conscious, in which case it would itself require the subject to be aware of it by means
of a ‘third‐order’ awareness of the second‐order state, and so on into infinity. However,
Sosa (2003) claims that this argument rests on a confusion between a conscious episode
(such as my consciously thinking that water is transparent) and being conscious of that
episode (as I am when I am reflecting on my conscious thinking that water is
transparent). I can be in a state of consciously thinking that water is transparent without
being in a state of being conscious of that state. So I can be in a state of consciously
Page 5 of 30
Introspection
thinking that I am thinking that water is transparent without being in a state of being
conscious of that state.5
A neo‐Rylean version of the position (McGeer 1996), which goes further in attempting to
account for the authoritative status of self‐knowledge, construes self‐knowledge as a
matter not just of there being a kind of fit between what one avows when one says ‘I
believe that p’ and what one goes on to do or say, but of a subject's actually ensuring that
there is that fit, by making her future actions and verbal behaviour fit her avowals. In that
case, the truth‐makers for one's avowals are not one's present ‘inner perceptions’ of ones
present inner states, but the future deeds that follow the words, which one brings about.7
The Rylean position is undermined by the fact that it seems so obviously true that, in at
least some cases, there is an epistemic asymmetry between self‐knowledge and
knowledge of others and that this is due to the fact that subjects often know what they
are thinking when they make avowals without appeal to evidence from their own
behaviour as well before engaging in any such behaviour (Boghossian 1989), and by the
fact that the arguments attacking the Cartesian alternative, which give support to this
apparent truth, are not compelling.8 The Rylean account may fit some cases of self‐
knowledge, specifically, cases of knowledge of one's own non‐episodic mental states, such
as one's standing beliefs, particularly ones about which one is self‐deceived, but it is the
episodic ones, like the cogito‐type cases, knowledge of which seems peculiarly
authoritative. The neo‐Rylean attempt to rescue authority while maintaining the central
claim that avowals are not reports on inner states that underlie and cause behaviour by
claiming that one's ability to ensure that one's future behaviour fits with and makes true
one's avowals suffers from these problems as well as from problems of its own. It has the
consequence, for example, that one cannot explain why people make the avowals they do
when they do, as well as allowing almost any first‐order belief which one can make true
by undertaking a course of action, such (p. 747) as my belief now that the balloon will pop
(which I can make true by undertaking a course of action involving a dart), to count as
authoritative (Brueckner 2001).
Page 6 of 30
Introspection
There are a number of different positions that are of this general kind, and here I can
only canvass a few. One stems from considerations about the nature of interpretation
(Davidson 1984, 1987, 1989). According to this, there is an asymmetry between self‐
knowledge and knowledge of others because the grounds for self‐ascriptions of beliefs
differ from the grounds for other‐ascriptions of beliefs. One argument for this stems from
considerations concerning the nature of interpretation. When another interprets my
speech, she must work out, from the external conditions in which my utterances occur,
both what my utterances mean and what the contents of my beliefs expressed by those
utterances are. But I do not need to work out what my utterances mean; when I say ‘The
balloon is red’ I know what I mean by this. This knowledge is not based on evidence,
observation of my own behaviour, or inference. Further, when another sets out to
interpret my utterance, she needs to work out whether I hold the sentence uttered true;
that is, whether I believe that it is true. My utterance itself along with other factors
external to me provide defeasible empirical evidence for whether I do hold that sentence
true. Unlike my interpreter, however, I do not need to rely on this evidence to determine
whether I hold that sentence true. So, just as I do not have to work out what I mean by
my utterance, I do not need to work out whether I hold the sentence uttered true. But if I
know what my utterance of that sentence means, and I know that I hold that sentence
true, I know what I believe.
One response to this argument is that it does not explain how I know that I hold a
sentence true, that is, how I know that I believe that it is true, and so does not explain
how I know what I believe. Without an explanation of this, the argument seems to
presume rather than explain authoritative self‐knowledge (Gallois 1996). Davidson (1991)
gives a different argument for the position, however, which doesn't depend on the nature
of interpretation. According to this, although an interpreter must work out, from the
external conditions in which my utterances occur, both the meanings of my utterances
and the contents of my beliefs, I do not have to do this. But since whatever determines
the contents of my beliefs also determines the contents of my beliefs about my beliefs, I
cannot be mistaken about what I believe. My authority with (p. 748) regard to my beliefs
about my own beliefs and other mental states consists in the fact that such beliefs are
infallible: whatever I think I believe I know I believe.9
Page 7 of 30
Introspection
In a similar vein, Burge (1988, 1996, 1998) argues that the authoritative status of certain
self‐knowledge—the kind of knowledge involved in the cogito‐type cases—is due to the
fact that they are contextually self‐verifying. In his opinion, the proper way to view the
cogito‐type cases is as ones in which one's first‐order thought is literally contained within,
or is a constituent of, one's second‐order thought, which is contextually self‐verifying for
the reason that in thinking the reflective thought one makes that thought true. One makes
it true because in thinking the reflective thought one brings into being the first‐order
thought upon which one is reflecting. The relation between first‐ and second‐order
thought is non‐causal, at least in part because there are not here two separate acts of
thinking: there is no first‐order thought, considered as a state distinct from the second‐
order, reflective one, to serve as one of the relata of the causal relation.
Although there are differences between these two versions of the deflationary position,
both are committed to the view that certain self‐knowledge is authoritative just because
one and the same content is the content of both the first‐order thought or belief and the
content of the so‐called second‐order thought or belief about the first‐order one. This
being so, it is impossible that one should be in error about what that content is. One
cannot have a thought of a certain contentful type and misidentify it. Since to think a
thought is to think its content, to think of it as a thought with a different content would be
to think a different thought altogether.
Some may object to this position on the grounds that it relies on a particular conception
of what is involved in thinking second‐order thoughts about one's first‐order thoughts,
but this conception is false because all second‐order thoughts or beliefs about first‐order
ones involve distinct thoughts/beliefs. However, this would not by itself show that, in the
cogito‐type cases, such thoughts are not contextually self‐verifying, since this feature of
such cases might be explained differently. For example, it might be maintained that a
necessary condition of thinking a reflective thought with a given content is that that
content refers to a first‐order thought content, and that a necessary condition for this is
that the contents of the two thoughts are of the same type. If so, then although it would
be true that thinking such a second‐order thought suffices for its being true, and that
thinking it makes it true, this would not be because there is no first‐order thought content
distinct from the reflective one to which it is causally related.
(p. 749)
There is, however, another, deeper problem with this type of position. Even if we agree
that certain cases of self‐knowledge are special because their contents are infallibly
known, and even if we agree that this is due to the fact that there is just one content
thought and thought about, the authority awarded to their subjects by this means alone
does not seem to have much to do with epistemology. In one clear sense of ‘authority’,
someone whose knowledge is authoritative is in a better epistemic position, or is
epistemically better placed, or better justified, than another to claim to have that
knowledge. But this deflationary position does not award a subject's self‐knowledge
authoritative status for this reason. The authority here is simply to do with the fact that a
Page 8 of 30
Introspection
Davidson and Burge develop their positions on self‐knowledge against the background of
a common presumption. Both are externalists with regard to the contents of propositional
attitudes. Briefly, externalism is the view that certain intentional states of persons have
contents that are ‘world‐involving’ in that they depend on the existence of objects and/or
other factors beyond the bodies of their subjects (Putnam 1975; Burge 1979, 1985, 1988;
Davidson 1984, 1987, 1989—though Burge's view specifically concerns anti‐
individualism). Externalism has been thought to compromise the presumption of
authoritative self‐knowledge in virtue of its commitment to the view that the empirically
discoverable determinants of a subject's thoughts might be better known to be what they
are by another who has better knowledge of those determinants. Suppose, to use a well‐
worn example, that in the distant galaxies there is a planet, Twin Earth. Twin Earth is a
duplicate of earth in every way, physically and phenomenologically described, apart from
this: the substance that fills lakes, rivers, and streams and has all of the
phenomenological qualities of water is constituted not by H2O but by a different chemical
structure, XYZ (Putnam 1975). Now consider a situation on earth in which a subject, who
knows nothing about chemistry and Twin Earth, believes that she is currently thinking
that water is transparent. Her justification for this seems weaker than that of another
individual who, knowing about chemistry and Twin Earth, knows that the water content in
question relates to H2O, and that the subject has not (unbeknownst to her) been
transported to Twin Earth. This seems to be a case where externalism conflicts with
authoritative self‐knowledge, intuitively understood.
The deflationary position taken up by Davidson and Burge responds to this threat by
maintaining that since whatever determines the (water) content of the subject's first‐
order state to be what it is also determines the content of the second‐order state, (p. 750)
and since the content of the first‐order one just is the content of the second‐order one,
error with respect to the content of the first‐order thought is not possible. While this may
explain why the subject's thoughts about her water thoughts on earth are infallible, and
also why her Twin's thoughts about her (water*) thoughts on Twin Earth are infallible, it
does not really respond to the worry expressed by the above situation. That concern has
to do with the fact that another, who has knowledge that the subject does not about both
earth and Twin Earth, and water and water*, is better justified in her knowledge that the
subject is indeed thinking that water is transparent. And she is better justified because
she has, whereas the subject does not, discriminative knowledge—that the subject is
thinking a water, rather than a water*, thought. Switching cases, where the subject has
Page 9 of 30
Introspection
unwittingly been transported from earth to Twin Earth, are intended to illustrate the
force of this threat. In one such case a subject is told that she has been transported to
Twin Earth during sleep but is not told when the switch occurred, so that when asked
which thought she was thinking a year ago she has no idea, even though she suffers no
memory impairment (Boghossian 1989).10
Another version of the deflationary position that presumes the truth of externalism is held
by McDowell (1986, 1998). Unlike the version just discussed, this one does not attempt to
explain what makes for authoritative self‐knowledge. On the contrary, it takes a sceptical
stance on the view that there is something problematic about the claim that subjects have
authoritative self‐knowledge—indeed, infallible self‐knowledge—that needs explaining.
According to it, this is an illusion, whose source is the ‘fully Cartesian’ conception of the
inner, which construes the mental domain as autonomous with respect to the world
beyond the mind to the extent that a subject could think exactly the thoughts that she
thinks in the world as it is (p. 751) in a world in which there is nothing beyond her mind.
Once the source of the so‐called problem is exposed and the Cartesian grip is dislodged,
one is free to view the inner and outer domains as ‘interpenetrating’ by construing the
infallibly knowable appearances expressed by avowals of the form ‘It seems to me that I
am thinking that that F is G’ disjunctively, as constituted either by the fact that I am
indeed manifestly thinking that that F is G or by the fact that that merely seems to me to
be the case.11
A question that arises is whether this view can account for authoritative self‐knowledge
any better than does the Davidson/Burge view. Suppose that I am in a situation in which I
am thinking a singular tiger thought—say, I am thinking that that tiger has stripes—but
am ignorant of the facts of biology and Twin Earth. On the present view I can have
infallible knowledge that it seems to me that I am thinking that that tiger has stripes
consistently with externalism because what I infallibly know is either that I am indeed
manifestly thinking that that tiger has stripes or that that merely seems to me to be the
case, and this is compatible with another person's better knowledge of which of the two
disjuncts is in question. But in this situation another has this knowledge because she has
discriminative knowledge that I do not—knowledge about the facts of biology and Twin
Earth, where there are no tigers but only pligers (creatures that have all the
phenomenological properties of tigers but differ in their biological constitution),
knowledge that the tiger content in question relates to tigers rather than pligers, and
knowledge that I have not (unbeknownst to me) been transported to Twin Earth. Because
she knows this, she not only knows which disjunct is in question but knows the fact,
disjunctively construed, that I know. Further, and more importantly, she seems better
justified than I am to know, when it seems to me that I am thinking that that tiger has
stripes, that what it seems to me that I am thinking is either this: I am indeed manifestly
thinking that that tiger has stripes, or this: that that merely seems to me to be the case.12
Page 10 of 30
Introspection
conditionally interchangeable with other unproblematic statements, and (2) that they
exhibit epistemic asymmetries with these other unproblematic, truth‐conditionally
equivalent statements. Her view is that avowals (p. 752) are self‐ascriptions of, but not
judgements about, one's mental states, events, etc., and are expressions of one's mental
states, but not reports on them. This combination of views allows her to maintain that
avowals are both true and epistemically more secure than other self‐ascriptions and
other‐ascriptions.
Bar‐On recognizes that there is an inherent tension in this combination of views. On the
one hand, the view that avowals are expressions of, rather than reports on, one's mental
states encourages the view that they are not truth‐evaluable. On the other hand, the view
that avowals are self‐ascriptions of mental states and are truth‐evaluable encourages the
view that they are not expressions of, but are rather reports on, one's mental states. Bar‐
On's attempt to show how these views can be held in combination involves arguing that
the special epistemic security that avowals enjoy does not have an epistemic basis at all.
This is the core of her account, and what makes it deflationary.
According to it, the special epistemic security that avowals enjoy has its source in the fact
that subjects are immune to error in self‐ascribing present mental states and (in the case
of intentional avowals) in assignments of intentional objects to such states. This immunity
to error is in turn due to the expressive nature of avowals, rather than to any recognition‐
based introspective self‐knowledge. Thus, the asymmetry between certain self‐
ascriptions, on the one hand, and other self‐ascriptions and other‐ascriptions, on the
other, is due to the fact that the former, but not the latter, are immune from error because
they are expressions of rather than reports on one's mental states. This puts the account
in the same category as those, mentioned above, which take some self‐knowledge to be
distinctive in being infallible but deny that there is any special method or means by which
it is obtained.
Evidently, the main problem for all of the deflationary accounts considered here is that,
although all concede that certain self‐knowledge is authoritative in being infallible,
infallibility does not account for authority in the sense that seems to matter for a genuine
epistemic asymmetry between self‐knowledge and knowledge of others. Further, there is
a question whether, on deflationary accounts, subjects' infallible ‘knowledge’ of the
contents of their own mental states deserves to be called knowledge at all.
Page 11 of 30
Introspection
it. The Cartesian position is just one of these. In this final section I will explore this and a
few other such positions and conclude by recommending my own.
As noted earlier, introspectionist views attempt to account for some or all of the features
associated with certain self‐knowledge by appeal to a special method or way in which one
comes to know one's own mental states. Like deflationary accounts, (p. 753) they
recognize an asymmetry between self‐knowledge and knowledge of others. Unlike
deflationary ones, however, they attempt to explain how it is that at least some self‐
knowledge possesses some or all of the features associated with the asymmetry. It is
important to the introspectionist view that the relevant knowledge concerns one's own
current, conscious mental states, whether these are propositional attitudes or sensations.
While all introspectionist accounts appeal to a special method or way in which knowledge
of one's own mental states is obtained, they differ in their accounts of what this method
is, how it works, and whether the knowledge delivered by it has all of the features of self‐
knowledge listed at the outset of Section 43.1 associated with the asymmetry between
self‐knowledge and knowledge of others. This section will discuss three such accounts. All
are in agreement that one clear feature marks the asymmetry between self‐knowledge
and knowledge of others, namely the epistemic directness or immediacy of such
knowledge, and that appeal to introspection is what explains this feature.
The first is what I earlier called the direct‐observation account (see Gertler 2003). Its
most famous expositor was Descartes, but a version of it was also held by Russell (1910),
who maintained that subjects have direct epistemic access to their own mental states in
virtue of being directly acquainted with them. More recently, versions have been
endorsed by Chisholm (1981), Langsam (2002), and Chalmers (2003). While Chisholm's
version is decidedly Cartesian, others, such as Chalmers's version, restrict the account to
knowledge of one's sensation states. The reason why one might think it particularly
plausible in the case of sensations, including visual experiences such as being appeared
to redly, is that it is plausible to hold that no appearance/reality distinction applies to
them.13
What distinguishes the direct‐observation account from other introspectionist ones is not,
or not merely, that it is observational, since, as we shall see, the inner‐sense account to be
discussed shortly is also an observational one. It is that the relation between the knowing
state and the mental state known is held to be unmediated by any further mental state or
mechanism in the subject. Chalmers (1996), for example, maintains that the knowing
state is partly constituted by the phenomenal state known. This is thought to distinguish
the kind of method involved in knowing one's own mental states from the kind of method
involved in ordinary observation, or perception, of things in the world around one in at
least two important respects. First, in ordinary perception the objects or phenomena
perceived are distinct from, and can exist independently of, their being perceived.
Second, and more importantly for present purposes, in ordinary perception there is a
state of perceptual experience that mediates between objects of perception and beliefs/
judgements about them, and a perceptual mechanism whose (p. 754) operation is causally
Page 12 of 30
Introspection
This being so, one question that arises with this account, no less than with the
deflationary ones discussed in Section 43.2, is whether on it self‐knowledge has the status
of being genuine knowledge. Wright (1989, 1998), for example, argues that knowledge
involves the application of concepts, with respect to which error must be possible. If this
is right, then in order to be able to know, say, that I am thinking that the apple on my
desk is round, I need not just to think that the apple on my desk is round but also to have
the concept of round and the concept of thinking and apply these correctly, where
correctness implies the possibility of error. However, on the direct‐observation account,
where self‐knowledge of mental states is construed in the way that Chalmers construes it,
knowledge of one's own current conscious mental states is infallible. If I think that I am
thinking that the apple in front of me is round, then I know that I am thinking that the
apple in front of me is round. Wright's view is that on this kind of account self‐knowledge
is not genuine knowledge.
Even if Wright is mistaken about this and the account does not compromise the status of
self‐knowledge as knowledge, it does jettison the epistemic asymmetry between self‐
knowledge and knowledge of others, since this is connected with the notion of one's being
epistemically better placed, or being better justified, than another to know one's own
current conscious mental states. That notion seems to require that the reviewing thought
and the thought reviewed be distinct existences, contingently and causally related. But
this is just what the direct‐observation account, specifically the constitution view, denies.
Somewhat surprisingly, then, this account suffers from the same sorts of problems that
bedevil deflationary accounts.
(p. 755)
Page 13 of 30
Introspection
Later, at the end of this section, I will suggest a way of developing the direct‐observation
account that seems capable of avoiding these problems. But first let us consider a couple
of other introspectionist accounts. According to one of these, the inner‐sense account,
introspection is viewed as a kind of inner perception. Versions of it have been held by
Locke and Kant. Locke claimed:
This Source of Ideas, every Man has wholly in himself … And though it be not
Sense; as having nothing to do with external Objects; yet it is very like it, and
might properly enough be call'd internal Sense
(1690/1975: II.i.4)
Kant described introspection as ‘inner sense, by means of which the mind intuits itself or its
inner state’ (1787/1998: A23/B37; p. 157). More recently, the inner‐sense account has been
championed by philosophers such as Armstrong (1968, 1981) and Lycan (1987, 1996, 2004),
both of whom also hold a particular view about what it is to be a conscious state (although
commitment to this is not required by commitment to the inner‐sense account of self‐
knowledge). According to this, to be a conscious state is to be the object of higher‐order
awareness by a perception‐like faculty. This view is known as the higher‐order‐perception theory
of consciousness (one of a family of such theories, another of which is the higher‐order‐thought
theory of consciousness—see Rosenthal 1997).
The inner‐sense account differs from the direct‐observation one in important ways. First,
and most obviously, it takes knowledge of one's own current, conscious states to be
mediated by a causal, perception‐like process involving a perception‐like mechanism.
Armstrong's position is that introspection involves a kind of self‐scanning process by
which subjects are made aware of their own mental states (1981: 61), involving an
internal scanner or monitor that takes first‐order states as inputs and outputs second‐
order ones, where these are distinct and contingently, causally related to one another.
Lycan (1996: 14) describes it as ‘the functioning of internal attention mechanisms
directed at lower‐order psychological states and events’. Armstrong's position takes the
internal scanner to be in principle capable of scanning the lower‐order mental states of
others, so that the asymmetry between self‐knowledge and knowledge of others is merely
a contingent matter. Lycan's does not take the analogy with perception to be so complete
as this, since it takes introspection to involve the use of primitive lexemes in a language
of thought. This makes introspective self‐knowledge ineffable.
Second, the inner‐sense account attributes only one feature of those associated with the
asymmetry between self‐knowledge and knowledge of others described at the outset of
our discussion to self‐knowledge; namely, that it is direct and immediate, and this only in
the sense that it is non‐inferential. Self‐knowledge has this feature because a subject
need not know the causal relations that hold between her introspecting state and the
state introspected when she knows her own mental state.
One objection to the account is that self‐knowledge cannot be like perception because in
perception there are three items (perceived object, perceptual experience, (p. 756) and
belief/judgement based on it) whereas in self‐knowledge there are only two (first‐order
Page 14 of 30
Introspection
mental state and belief/judgement based on it) (Shoemaker 1994; Rosenthal 1997). To this
the response has been made that the inner‐sense view is not committed to construing
introspection as like perception in every respect (Lycan 1996: 28).15 Since, however, the
presence of an experiential state seems to be essential to perception, the absence of any
such state in introspective self‐knowledge is a crucial difference.
In response, one might agree that the inner‐sense view is incompatible with Shoemaker's
strong sense of ‘rational’ but claim that there is a weaker sense with which it is
compatible. This is the sense that is relevant when one considers very young children,
who are not yet capable of rational belief revision but are capable of adjusting their
beliefs about the world in response to changes in their environments. 16
But a more general worry about the account remains. Shoemaker's principal objection to
the inner‐sense account is that the relation that holds between the first‐order and second‐
order states involved in self‐knowledge is non‐contingent and constitutive, and it is
essential to human rationality that it be so. It is significant that the account marks an
asymmetry between self‐knowledge and knowledge of (p. 757) others but takes it to
consist merely in contingent facts about the method of knowing and the way in which the
method operates. It fails to account for nearly all of the features associated with the
asymmetry between self‐knowledge and knowledge of others because it rejects the view
that self‐knowledge possesses them. But this robs the account of a satisfactory
explanation of why at least some self‐knowledge is epistemically special; a fact that the
classically Cartesian account hoped to explain by appeal to introspection. Some will claim
that this cannot be accounted for by any model that construes the relation between
knowledge of one's own mental states and those states themselves as contingent and
causal (Burge 1996, 1998; Gertler 2000).
Page 15 of 30
Introspection
Further, and relatedly, the inner‐sense view suffers from a problem mentioned earlier in
connection with deflationary positions, that of reconciling authoritative self‐knowledge
with externalism about mental content. According to it, the authoritative status of certain
self‐knowledge is a purely contingent matter, arising from the reliability of the
introspective perception‐like faculty. As it happens, this faculty is in general reliable. But
despite this, if externalism is true, another may be better placed than I am to know the
contents of my intentional states because she may have better knowledge of the beyond‐
the‐body factors that individuate the contents of those states. This is because the internal
scanner or monitor can only detect what is internal to the subject, not what is beyond the
body.
A third type of introspectionist account agrees with the inner‐sense view that knowledge
of one's own mental states is perception‐like, but denies that it involves the operation of
any inner process or mechanism in addition to the process of perception itself. This view
is known as the displaced‐perception account (Dretske 1994, 1995, 1999). Whereas the
inner‐sense account takes introspection to be a matter of ‘looking inwards’, and attending
to one's inner states, the displaced‐perception account takes introspection to be a matter
of ‘looking outwards’, and attending to the objects of normal perception, such items as
trees, oranges, and human beings. Versions of it have been held by Evans (1982) and
Dretske (1994, 1995, 1999, 2006). As Evans describes it:
(1982: 225)17
This is displaced perception, since in perception one directs one's attention to the object
perceived, whereas in self‐knowledge one directs one's attention not to one's (p. 758) inner
state, but outward, to a state or feature of the world beyond one. According to this account, one
‘perceives’ one's own mental state by perceiving something else.
Evans's version of the account speaks of self‐ascriptions of beliefs, and his use of ‘think’
in the above example is most plausibly construed as functioning like ‘believe’. The
displaced‐perception account seems ideally suited to explain cases like these. This is
because it is plausible to maintain that it is in the nature of belief that beliefs aim at the
truth of what is believed. So it is not surprising that when asked what I think about a
possible or actual state of the world beyond me my attention is drawn not to the inner
workings of my mind but to factors in that world. But things seem otherwise when the
issue is not about what one believes, but about what one is currently consciously thinking.
When asked what I am thinking, rather than what I think is or will be the case in the
Page 16 of 30
Introspection
world beyond me, the more plausible view is that my attention turns not outward but
inward. This is all the more apparent in cases of self‐knowledge of one's sensation states.
To know what the experience is like, what properties it has, it is enough for the
experiencer to ‘look at’ what the experience is an experience of (something that,
as the experiencer, he cannot help but be doing). That will tell him what the
relevant properties of the experience are.
Dretske's view is that the knowledge that one has of one's own mental states is special in that it
is not only authoritative, but infallibly so.
(p. 759)
But how is this possible? On the representational view of mental content the factors that
determine the properties, or content, of an experience or other mental state are extrinsic,
or relational. That is to say, the representational view is externalist about mental content.
How, then, can one know authoritatively, let alone infallibly, what the properties of one's
own propositional or sensory states are?
This is explained by appeal to the distinction between ‘thing’ awareness and ‘fact’
awareness, a distinction that is brought explicitly into play in order to help explain why
the possession of these features by self‐knowledge is compatible with externalism. 18
What I know authoritatively when I know that I am thinking that water is transparent, or
that I am having an experience as of a red round apple, is what I am thinking or having,
rather than that I am thinking or having it. That is, I am authoritative about the content of
my thought or experience, but not about the fact that I am having a thought or
experience with that content. As Dretske puts the point:
Introspection is not how we know that we think and feel. It is how we know what
we think and feel. Introspection is no more a way of knowing that we think and
feel than is perception, our primary way of knowing what else is in the world, a
Page 17 of 30
Introspection
way of knowing that there is something else in the world … We cannot see that
there is an external world, although the things we come to know by seeing (that
there is beer in the fridge, keys in my pocket) imply that there are things (namely,
beer and keys) outside my mind.
(1999: 137)
The idea is that when I am ϕ‐ing that p, for some attitude ϕ and some propositional content p, I
can infallibly know that it is p that I am ϕ‐ing, but know only fallibly, if at all, and by some other
means, that it is ϕ‐ing (that p) that I am doing.
In order to see why Dretske thinks that self‐knowledge of content is infallible, we need to
understand an analogy he uses to demonstrate this. Consider an instrument, like a petrol
gauge or a thermometer, that represents the value of a quantity, Q, say, temperature.
When it isn't malfunctioning, it correctly represents the quantity of a source to which it is
connected. So, for example, when athermometer is functioning properly, when it registers
(or ‘says that’) the source has Q of 37 °C, the source does have a Q of 37 °C. It carries
this information, but does not say that it carries this information. But suppose we were to
rewire it so that that one state, registering the source as having a Q of 37 °C, could ‘say
that’ the source to which it is connected has a Q of 37 °C, but could also say something
about itself—that it is representing (p. 760) the source as having a Q of 37 °C. We might
do this simply by affixing to it the label ‘Value that Q is representing’ (Dretske 1999: 134).
This instrument now does two things by being in one state: it both represents the state of
its source, and it represents itself as representing the state of its source. Although the
instrument is fallible about the first thing it does (since it can malfunction), it is infallible
about the second thing it does (even when it is malfunctioning).
Because the account is externalist, it is also subject to the same sorts of objections based
on switching examples discussed in Section 43.1. It is true that simple switching will not
be a problem for the account, since the semantics associated with it is teleological, taking
the determinants of mental content to be causal‐historical, and a simple switch will not
suffice for a change in the content of, say, a water thought. But the type of switching
example discussed in Section 43.1 remains problematic for the account (Boghossian
1989). This is because, in the envisaged scenario, the subject has not only been
Page 18 of 30
Introspection
unwittingly transported from earth to Twin Earth, but she also remains there for a long
enough period for the causal‐historical factors that determine content to determine her
thoughts about the watery stuff to be water* thoughts. Evidently, apart from the inner‐
sense view, all of the accounts, both deflationary and introspectionist, thus far considered
suffer from the same sorts of objections: one stemming from commitments to the
infallibility of knowledge of one's own current conscious state, and the other stemming
from considerations having to do with externalism about mental content. The inner‐sense
view does not suffer from objections of the former sort, but then again the asymmetry
between self‐knowledge and knowledge of others doesn't amount to much on this view
either, since self‐knowledge doesn't have the special epistemic status it is typically
thought to have.
This is not true of other properties of objects of perception. Water, for example, is an
instance of the chemical structural property H2O, but it is not manifested to me as an
instance of that property through one of my senses. So certain properties seem to be ones
of which we have direct or immediate awareness in perception because they are
observable: whether objects are instances of them can be determined just by unaided
observation (but not, perhaps, by unconceptualized experience) of those objects. 19
Page 19 of 30
Introspection
is that such properties are the ones by which objects that instance them are typically
known in the first place. Knowing an object through instances of certain properties and
not others favours certain ones epistemically. Another is that they are in general as they
appear to normal perceivers in normal circumstances. This is true both for the primary
qualities, such as that of being rectangular, where the connection between an object's
being an instance of the property and how things look to normal observers in optimal
conditions is thought to be a posteriori and contingent, and for the secondary ones, like
that of being brown, where the connection between these and the best opinion of normal
observers under optimal conditions is thought to be a priori, and, further, thought by
some to determine the nature of the property itself. 20
(p. 762)
The rationale for focusing on the example of observable properties in perception is not to
argue that self‐knowledge is just like perceptual knowledge. There are clearly important
and fundamental differences between these two sorts of knowledge. Crucially, there is a
certain kind of phenomenology to perception of observable properties (for example, the
experience of something's looking wet is distinctive and very different from the
experience of something's looking red) that attaches to the contents of one's perceptual
experiences. This seems to be lacking in the case of self‐knowledge of one's thoughts,
since, as noted earlier, there seems to be no analogue of a perceptual experience in the
knowledge one has of one's own thoughts at all.
Despite this difference, there are important affinities that can help us to understand
better the nature (and extent) of authoritative self‐knowledge. In particular, the two
above‐mentioned features of observable properties of perception apply to contentful
properties in these special cases of authoritative self‐knowledge in a way that can help us
to see why subjects have immediate awareness of them and so authoritative knowledge of
their own thoughts. When I think about my own current, conscious intentional states
while undergoing them, I think about them first and foremost as states of certain
contentful types. Further, when I think about my states in this way, they are in general of
the contentful types I take them to be. What makes for authority, when I have it, is that
only I can be the subject of my intentional states, so that when I think about my
intentional states as states of particular contentful types I am the only one to whom those
contents appear to me in this epistemically basic and favoured way. So normally (i.e.
barring special cognitive failures) when I think about my current conscious intentional
states while undergoing them I am authoritative about the contents of those states.
Because the two features that characterize observable properties of perception are
abstract and general, they are not tied to cases of observation alone. Those who appeal to
such phenomena as ‘intellectual experience’ or ‘intellectual intuition’ in their accounts of
authoritative self‐knowledge may well appeal to such features (see, for instance, Burge
1996 and Bealer's theory of the a priori— 1999). This distinguishes the position from
other introspectionist ones. Further, the account is not tied to any particular conception
of the cogito‐type cases; specifically, to the conception that takes second‐order thoughts
Page 20 of 30
Introspection
about first‐order ones to contain the first‐order ones as constitutents. It can construe
such cases as involving both a thinking of a first‐order thought and a reflective thinking
about that thought by the same subject, where the truth of the reflective one requires
that it match the one reflected upon both with respect to content and with respect to the
attitude. When successful, in reflection, the subject thinks a content of the same type as
that tokened in the first‐order thought (that is, she thinks the content again) in the same
attitudinal mode (p. 763) and presents it as the content of the first‐order thought. 21 This
differentiates the view from the type of position held by Dretske (1999) discussed above,
with its attendant difficulties. In thinking again the same content as that contained in
one's first‐order thought, a subject is brought into direct epistemic contact with the
content of that first‐order thought itself. But because there are here not one but two
thoughts with their own contents, error is possible. This makes self‐knowledge, even in
the cogito‐type cases, a genuine cognitive achievement.
Finally, the account seems to be compatible with externalism. The kind of epistemic
access subjects have to the contents of their thoughts in such cases, in being direct and
immediate, contrasts with the epistemic access others have to those contents. When I am
currently consciously thinking and thinking about a thought with a given externalistically
individuated content, another may be in a better position than I am to know what content
is available for me to think, and so to think about. But that other is not in a better position
to grasp, and so to know, the particular content that constitutes the subject matter of my
thought, about which I am thinking. When I am both thinking it and thinking about it, I
have, whereas another does not, a special kind of epistemic access to the content of that
thought. Being in that position gives me an epistemic purchase on it that no other has.
While it is true that this does not give me better discriminative knowledge of the contents
of my thoughts, it is plausible to hold that, if externalism is indeed true, I could not be in a
better position than another who has better knowledge of my environment to have such
knowledge. But that does not compromise my authoritative knowledge of what I am
currently consciously thinking (Burge 1988, 1998).22
References
Alston, W. (1971), ‘Varieties of Privileged Access’, American Philosophical Quarterly, 8:
223–41.
Armstrong, D. (1968), A Materialist Theory of the Mind (London: Routledge & Kegan
Paul).
Page 21 of 30
Introspection
Burge, T. (1979), ‘Individualism and the Mental’, Midwest Studies in Philosophy, 4: 73–
121.
repr. in Davidson, Inquiries Into Truth and Interpretation: Philosophical Essays, 2nd edn.
(Oxford: Clarendon, 2001), 93–108.
—— (1987), ‘Knowing One's Own Mind’, Proceedings and Addresses of the American
Philosophical Association, 60: 441–58;
Page 22 of 30
Introspection
repr. in M. Krausz (ed.), Relativism: Interpretation and Confrontation (Notre Dame, Ill.:
University of Notre Dame Press, 1989), 159–72,
—— (1989), ‘What is Present to the Mind?’, Grazer Philosophische Studien, 36: 3–18;
Gallois, A. (1996), The World Without, The Mind Within: An Essay on First‐Person
Authority (Cambridge: Cambridge University Press).
Kant, I. (1787/1998), The Cambridge Edition of the Works of Immanuel Kant: Critique of
Pure Reason, trans. and ed. P. Guyer and A. Wood (Cambridge: Cambridge University
Press).
Page 23 of 30
Introspection
McDowell, J. (1986), ‘Singular Thought and the Extent of Inner Space’, in P. Pettit and J.
McDowell (eds.), Subject, Thought, and Context (Oxford: Clarendon), 137–68;
Page 24 of 30
Introspection
Putnam, H. (1975), ‘The Meaning of “meaning”’, in Putnam, Mind, Language, and Reality:
Philosophical Papers, II (Cambridge: Cambridge University Press), 215–71.
repr. in Russell, Mysticism and Logic (London: Allen & Unwin, 1963), 152–67.
Ryle, G. (1949), The Concept of Mind (New York: Barnes & Noble).
Wright, C. (1988), ‘Moral Values, Projection and Secondary Qualities’, Proceedings of the
Aristotelian Society, suppl. vol. 62: 1–26.
Notes:
(2) This contrasts with another use of the term ‘self‐knowledge’, to apply to knowledge of
the subject of mental states, or the self, and its nature (see e.g. Shoemaker 1968; Evans
1982; Bermúdez 1998). This use will not form part of our discussion.
Page 25 of 30
Introspection
(3) See his Meditations, especially Meditation 2: ‘Lastly, it is also the same “I” who has
sensory perceptions, or is aware of bodily things as it were through the senses. For
example, I am now seeing light, hearing a noise, feeling heat. But I am asleep, so all this
is false. Yet I certainly seem to see, to hear, and to be warmed. This cannot be false; what
is called “having a sensory perception” is strictly just this, and in this restricted sense of
the term it is simply thinking’ (1641/1984: 19).
(4) Some, like Wright, emphasize the non‐evidence‐based character of such knowledge,
whereas others, like Heil, emphasize its non‐empirical‐èvidence‐based character. Alston
gives an illuminating account of the different senses that might attach to the notion of
direct access. He argues that the notion of directness that is relevant to self‐knowledge is
epistemic, not causal, and is explicable in terms of being non‐evidence‐based, where this
is distinct from being non‐inferential. Heil (1992) endorses the view that the notion of
directness is epistemic, not causal. Gertler (2003) distinguishes two senses of ‘direct’ in
‘direct access’, one epistemic (what is here characterized as ‘non‐inferential’) and the
other metaphysical (here characterized as ‘unmediated’).
(6) As he puts it: ‘Our knowledge of other people and ourselves depends upon our
noticing how they and we behave’ (1949: 181).
The view I propose involves putting special emphasis on our own agency by
recognizing that we are actors as well as observers and so can be good, even
excellent, ‘predictors’ of our future behavior because we have the power to make
these ‘predictions’ come true. Put simply, we are able to ensure a fit between the
psychological profile we create of ourselves in first‐person utterances and the acts
our self‐attributed intentional states are meant to predict and explain simply by
adjusting our actions in appropriate ways.
(1996: 507)
Page 26 of 30
Introspection
(8) As Davidson puts it, ‘Ryle was wrong. It is seldom the case that I need or appeal to
evidence or observation in order to find out what I believe; normally I know what I think
before I speak or act. Even when I have evidence, I seldom make use of it’ (1987: 441).
(9) Davidson typically couches this argument in terms that speak of presumptions of
knowledge of meaning and the like, but, as the second of his arguments clearly indicates,
there is more to the position than the presumption of authority. Here is a passage where
the argument makes no use of it:
(1991: 196)
(10) It is not clear that this case shows that Burge's account is defective, unlike the
example given in the paragraph immediately preceding this one. Burge (1998) has
responded to it by maintaining that authoritative self‐knowledge does not extend to cases
of discriminative knowledge, and this is particularly so where memory is involved:
I have maintained that the individual may not know whether yesterday he had an
aluminum or twaluminum thought. He does not have discriminative knowledge of
this form. But memory need not work by discrimination; it can work through
preservation. The memory need not set out to identify or pick out an aluminum
rather than a twaluminum thought, trying to find one by working through the
obstacles set by the switches. Preservative memory normally retains the content
and attitude commitments of earlier thinkings, through causal connections to the
past thinkings. That is one of its functions—maintaining and preserving a point of
view over time. It need not take a past thought as an object of investigation, in
need of discrimination from other thoughts. Memory need not use the form
‘Yesterday I was thinking a—type of thought’, where the memory attempts to
identify the thought content as an object. Again, if it did, the individual might
perhaps err by using a thought appropriate to the second environment in making
an attribution to a thought event in the first environment. … The memory need not
be about a past event or content at all. It can simply link the past thought to the
present, by preserving it.
(1998: 357)
Page 27 of 30
Introspection
It might be thought that the example in the paragraph immediately preceding this in the text
makes essential use of the notion of discriminative knowledge, since the authoritative status of
the subject's self‐knowledge seems to be compromised by another's better knowledge, not only
of the facts of chemistry, but also of Twin Earth and XYZ. But the point can be made equally
effectively in terms of another's better knowledge of the facts of chemistry, without appeal to
knowledge of Twin Earth.
Short of the fully Cartesian picture, the infallibly knowable fact—its seeming to
one that things are thus and so—can be taken disjunctively, as constituted either
by the fact that things are manifestly thus and so or by the fact that that merely
seems to be the case. On this account, the idea of things being thus and so figures
straightforwardly in our understanding of the infallibly knowable appearance;
there is no problem about how experience can be understood to have a
representational directedness towards external reality.
(1986: 150)
(12) Again, it might be claimed that authoritative self‐knowledge does not extend to cases
of discriminative knowledge and that this counts as one such case (see n. 10), but then
the point can be put in terms of another's better knowledge of the facts of biology,
without appeal to knowledge of Twin Earth and pligers.
(13) Chisholm gives as examples of ‘self‐presenting’ properties not just feeling sad, but
thinking about a golden mountain, being appeared redly to, and believing oneself to be
wise, where a self‐presenting property is defined as ‘a property which is such that, if
while having it, you consider your having it, then you will believe yourself to have
it’ (1981: 81). Further, ‘the foundation of our knowledge consists of certain subjective—or
‘Cartesian’—apprehensions'. (1981: 92).
(14) Those who hold, with McDowell (1994), that perception itself is unmediated by an
unconceptualized experiential state, but who hold that one's own perceptual states can be
introspectively known in the observational way, will deny that there is this contrast but
claim to be introspectionists with respect to self‐knowledge (see Langsam 2002).
(15) Thus, Lycan says: ‘The inner‐sense theorist does not contend (at least neither
Armstrong nor I contend) that internal monitoring is like external perception in every
single respect. And in particular, we should not expect internal monitoring to share the
property of involving some presented sensory quality at its own level of operation’ (1996:
28).
(16) In a similar vein, Lycan (1996) responds to Shoemaker's analogous argument with
regard to pain by distinguishing between a strong and a weak sense of ‘pain’ and
insisting, against Shoemaker's claim that:
Page 28 of 30
Introspection
at least certain kinds of pain behaviour … are intelligible as pain behaviour only
on the assumption that the subject is aware of pain, for to see them as pain
behaviour is to see them as motivated by such states of the creature as the belief
that it is in pain, the desire to be rid of the pain, and the belief that such and such
a course of behaviour will achieve that result
(1982: 225)
This indicates that he takes at least some self‐knowledge to be infallible. But in fact he goes on
to say that more is involved in knowing whether I believe that p than just following whatever
procedure it takes to determine whether p. At the very least it involves possessing the concept of
belief, as it applies not just to oneself but also to others. Still, he claims, this further requirement
need not involve looking inward. Note that Evans does not endorse this model for self‐knowledge
of experiences (1982: 226 ff.). Note also that, as in the case of at least some inner‐sense models,
the appeal to perception and ‘looking inwards’ in the displaced‐perception account may be
metaphorical (see n. 1). As the Evans quote in the text indicates, the key idea in the account
seems to be that one comes to know what one thinks, believes, etc. not by appealing to any of
one's inner states but by appealing to an outer state of the world, and this leaves it open
whether the relevant state is one that is available to perception. Thanks here to John Williams.
(19) The distinction I am after can be captured by means of the distinction between
seeing x and seeing that p. I can both see that the table is brown and see the table's
Page 29 of 30
Introspection
brownness; but although I can see that the liquid before me is H20, I cannot see its ‘H20‐
ness’. Only the former counts as a case of direct epistemic access.
(20) However, unlike Wright (1988, 1992), who holds that one's authority with respect to
one's own states consists in the fact that subjects’ best opinions concerning those states
fix the extensions of content types, the view I favour is that one's authority consists in the
fact that one does not normally in reflection misidentify the object of one's reflection.
Reflection is, in one respect at least, an appropriate characterization of the special
relation that subjects’ second‐order thoughts bear to their first‐order ones. In physical
reflection—say, in a mirror—under certain ideal conditions, the object is not normally
misrepresented. So the object is as it appears to be. But the reflection does not determine
the object to be what it is. Similarly, in mental reflection the reflecting thought does not
determine the thought reflected upon to be what it is.
(21) How might ‘aboutness’ or reference to the first‐order content be secured? Roughly,
in the way suggested by Davidson's paratactic analysis of sentences involving indirect
discourse (1969/2001). Suppose I think to myself I am currently thinking that water is
transparent. This second‐order thought can be understood as a compression or
abbreviation of something like: I am currently thinking a thought whose content is the
same as the following: water is transparent. When I think this thought, I present my
thought as having the same content as that of another current thought of mine, and I do
this correctly if I am a samethinker with regard to these two thoughts (rather than a
‘samesayer’, as in Davidson's original suggestion).
(22) I would like to thank Bill Lycan, Graham Macdonald, and participants at the 2005
Australian Association of Philosophy Conference, Dunedin, New Zealand, for comments
on this chapter.
Cynthia Macdonald
Page 30 of 30
Semantic Externalism and Self‐Knowledge
The distinctive claim made by semantic externalism is that a subject's thought contents
are partly individuated by her environment, and do not supervene on her ‘inner states’,
such as her brain states. One of the main objections to this position is the claim that it is
incompatible with self-knowledge. A subject's knowledge of her own thoughts seems quite
different from her knowledge of what others think. A subject uses behavioural evidence to
know what others think. However, typically, a subject can know what she herself thinks
without inferring this from her own behaviour, and even prior to manifesting any
behaviour which could constitute grounds for such an inference.
Keywords: self-knowledge, semantic externalism, inner states, brain states, behavioural evidence, second-order
thoughts
THE distinctive claim made by semantic externalism is that a subject's thought contents
are partly individuated by her environment, and do not supervene on her ‘inner states’,
such as her brain states. One of the main objections to this position is the claim that it is
incompatible with self‐knowledge. A subject's knowledge of her own thoughts seems
quite different from her knowledge of what others think. A subject uses behavioural
evidence to know what others think. However, typically, a subject can know what she
herself thinks without inferring this from her own behaviour, and even prior to
manifesting any behaviour which could constitute grounds for such an inference.
Exceptions may be provided by cases of self‐deception and the unconscious. However, for
the most part a subject can have a priori knowledge of what she thinks (call this
‘privileged access’ for short).
Since semantic externalism is a thesis about thought content, the threat it poses to
privileged access specifically concerns a subject's knowledge of her thought contents,
rather than her attitudes to those contents. There are two main ways to fill out that
Page 1 of 16
Semantic Externalism and Self‐Knowledge
threat: what Davies (1998) usefully distinguishes as the ‘achievement’ and ‘consequence’
problems. According to the achievement problem, semantic externalism makes it difficult
for a subject to achieve privileged access. For instance, it has been argued that since
semantic externalism claims that a subject's thought contents are partly individuated by
her environment, a subject would need to use empirical information about her
environment in order to know her thought contents. According to the consequence
problem, the combination of semantic externalism (p. 768) and privileged access has
absurd consequences. Both supporters and opponents of semantic externalism accept
that it would be a serious objection to semantic externalism if it were incompatible with
privileged access. Thus, semantic externalists have argued that semantic externalism is
compatible with privileged access, whereas their opponents have argued for
incompatibility.
Since it's part of the case that S is switched unwittingly between the planets, it may seem
that, as a result of the switch, S may make a mistake about her thoughts. Perhaps, after
the switch, she would take herself to believe that water is wet when she actually believes
that twater is wet. Thus, semantic externalism may seem to threaten privileged access by
undermining a subject's reliability about her thought contents. Alternatively, the
incompatibilist may formulate her challenge using the notion of a discriminative capacity.
Consider S on earth when she thinks that water is wet. Plausibly, without empirical
investigation S cannot distinguish this situation from the alternative situation in which
she is instead on Twin Earth thinking that twater is wet. For Twin Earth seems just like
earth and water seems just like twater. But, then, it may be said, on earth S cannot know
a priori that she thinks that water is wet for, without empirical information, she cannot
distinguish the actual situation in which she thinks that water is wet from the alternative
situation in which she is instead on Twin Earth thinking that twater is wet. These two
Page 2 of 16
Semantic Externalism and Self‐Knowledge
(p. 769)
Two concerns arise about whether the standard response fully answers the achievement
problem. The first questions whether the extent of one's reliability about one's thoughts is
sufficient to defend privileged access. The second questions whether reliability, however
extensive, can answer the incompatibilist's concerns about discriminative abilities. First,
consider the extent of a subject's reliability about her thoughts. Notice that a second‐
order thought is self‐verifying only if it ascribes a current thought. By contrast, a second‐
order belief that one thought that p in the past is not self‐verifying. Believing that I
thought that p entails that I now think that p in the sense of entertaining the thought that
p. However, it leaves it open whether I thought that p at the relevant earlier time. The
problem of knowledge of past thoughts is exacerbated on certain externalist views which
hold that, after a switch a subject loses certain concepts (e.g. Ludlow 1995b; Heal 1998;
Tye 1998; though see Burge 1993 for a different view). For instance, on this view, once S
has been switched to Twin Earth she gains the concept twater and loses the concept
water. As a result, after the switch she no longer has the conceptual resources to think
Page 3 of 16
Semantic Externalism and Self‐Knowledge
that water is wet, let alone know that she used to think this. On this view, a victim of a
slow switch may be unable to have a priori knowledge of some of her past thought
contents.
Initially, it may seem that even if the standard response cannot be applied to show how
semantic externalism is compatible with privileged access to past thoughts, this need not
undermine its application to current thoughts. However, Boghossian (1989) argues that if
semantic externalism undermines a subject's a priori knowledge of her past thoughts it
also undermines her a priori knowledge of her current thoughts. If he were right, then
the standard response would not merely have (p. 770) a restricted application, but would
fail altogether. Boghossian's argument exploits what he calls a ‘platitude’ about memory:
if S knows that p at t1, and if at (some later time) t2 S remembers everything S knew at
t1, then S knows that p at t2. Let S be the subject of a slow switch so that t1 is a time
before the switch at which S thinks that water is wet, and t2 is a time sufficiently after
the switch for the switch to have affected S's thought contents. Boghossian says that we
can stipulate that S forgets nothing between t1 and t2. Boghossian appeals to the
stipulation and his ‘platitude’ to argue that if at t2 S does not know a priori that at t1 she
thought that water is wet, then at t1 she did not know a priori that she thought that water
is wet. In this way Boghossian seeks to extend a lack of privileged access to past thoughts
to current thoughts also. However, even if we accept that, post switch, S cannot know a
priori what she thought before the switch, we need not accept with Boghossian that this
undermines S's ability to know a priori her current thoughts. If between t1 and t2 S loses
the concepts required to think what she knew at t1, then at t2 she cannot know what she
earlier knew. We may either see this as undermining Boghossian's stipulation that she
forgets nothing between t1 and t2 (Brueckner 1997) or, leaving the stipulation in place, as
undermining Boghossian's ‘platitude’ (Ludlow 1995b).
Some argue that a subject's reliability about her thoughts is restricted in a different way,
to her thought contents, rather than to her attitudes to those contents. Bernecker (1996)
and Gibbons (2001) argue that Burgean semantic externalism makes it possible that a
subject may suppose that she has an attitude to a content when she lacks that attitude.
Burge accepts that a subject may have a concept despite incompletely understanding it.
Applying this idea to attitude concepts, Gibbons and Bernecker suppose that a subject
has the concept of belief while taking it to have a wider application than it in fact has. As
a result, a subject may believe that she believes that p although, in fact, she does not
believe that p.
It seems, then, that the standard response to the achievement problem needs
qualification: although a subject cannot falsely judge that she thinks that p, on some
versions of semantic externalism a subject may falsely judge that she has a certain
attitude to the content p and may be unable to recall some of her earlier thoughts.
However, these caveats need not worry the compatibilist unduly. As long as the conditions
which give rise to these mistakes are rare then, on a fallibilist conception of knowledge,
Page 4 of 16
Semantic Externalism and Self‐Knowledge
these errors do not show that in general subjects lack privileged access to their present
and past propositional attitudes.
However, the standard response faces a second, and more serious, worry; namely, that it
fails to fully answer that strand in the achievement problem which concerns
discriminative abilities rather than reliability (Vahid 2003; Brown 2004: 33–73). The
incompatibilist argued that S cannot know a priori that she thinks that water is wet since
she cannot non‐empirically distinguish the actual situation in which she has this thought
from the alternative in which she is on Twin Earth thinking that twater is wet. Stressing
that a subject is reliable about her thought contents does not provide any obvious answer
to this worry about discrimination. To answer this worry, the compatibilist needs to show
either that S can non‐empirically distinguish the actual and twin situation, or that such an
ability is not required for a priori knowledge (p. 771) of her thoughts. (Here I'm focusing
on the impact of discriminative abilities on self‐knowledge; a distinct worry concerns
their impact on reasoning; see Boghossian 1992; Goldberg 1997, 1999.)
The first of these options seems unpromising and has been rejected by many prominent
anti‐individualists (e.g. Burge 1988; Falvey and Owens 1994). It is part of the case that S
is unaware of the switch in her environment and her thoughts. After the switch, on being
asked, she would claim that she now expresses the same thought by ‘Water is wet’ as she
used to many years ago. Even if told that she has been switched, but not when, she could
not identify which thoughts involve the concept water and which involve the concept
twater (Burge 1988; Goldberg 1997). It initially seems, then, that S cannot non‐
empirically distinguish water and twater thoughts.
In response the compatibilist may suggest that there is a further test for having a
discriminative ability which S can pass; namely, being able to act differentially with
respect to the two kinds of thoughts. For instance, she may argue that although S reports
her water and twater thoughts using the same word ‘water’, since ‘water’ has a different
meaning on earth and Twin Earth S reports the relevant thoughts as having different
contents. Alternatively, she may argue that S acts differentially with respect to water and
twater thoughts when her actions are typed intentionally. For instance, a reaching which
is caused by the desire for water would be typed as a reaching for water whereas a
reaching which is caused by the desire for twater would be typed as a reaching for
twater. However, on further consideration this compatibilist response fails. If a subject
can distinguish two types of thing, then she can usually act differentially with respect to
those types for a wide range of actions within her behavioural repertoire. However, there
is a very large range of actions within S's behavioural repertoire that she cannot perform
differentially with respect to water and twater thoughts. For example, suppose that
before the switch we convince S of the truth of semantic externalism and forewarn her
that she will be switched in such a way as to change the concept she expresses with
‘water’. However, we do not tell her when the switch will take place. We ask her to press
a button when her thought contents change. It seems plausible that S will fail to pass this
test even if we offer her a large reward, say one million pounds, for doing so. Similar
arguments show that S cannot act differentially with respect to water and twater
Page 5 of 16
Semantic Externalism and Self‐Knowledge
thoughts by stamping her feet, clasping her hands, wiggling her ears, etc. In summary, S
does not have the action abilities we would expect if she genuinely had the ability to
distinguish the two types of thought (Brown 2004: 50–8).
It seems, then, that the compatibilist should accept that S cannot non‐empirically
distinguish water and twater thoughts. Instead, many hold that this ability is not required
for S to have a priori knowledge of her thought contents (e.g. Burge 1988; Falvey and
Owens 1994). However, there is little detailed argument for this view. (For an exception
see Falvey and Owens 1994.) Further, certain difficulties face any attempt to show that
knowledge requires only reliability and not discriminative abilities. A first problem is that
these two phenomena typically go hand in hand so that it's hard to construct cases which
show that knowledge requires one rather than the other. A second problem arises from
the motivations for the reliabilist approach to (p. 772) knowledge, which holds that a true
belief is knowledge if produced by a reliable process (Vahid 2003; Brown 2004: 45–8). Key
proponents of this view motivate reliabilism by the idea that knowledge requires
discriminative abilities (Goldman 1976, 1986; McGinn 1984).
It seems hard, then, for a compatibilist to argue that S can non‐empirically distinguish
water and twater thoughts, or that knowledge does not require discriminative abilities.
An alternative potential defence of compatibilism attempts to finesse the issue of whether
knowledge requires discriminative abilities or merely reliability by using the well‐known
notion of a relevant alternative (see e.g. Goldman 1976). It's widely acknowledged that,
on pain of scepticism, not all alternatives are knowledge‐undermining, but only those
which are ‘relevant’ or ‘nearby’. Perhaps the compatibilist could argue that the possibility
that one is in a twin scenario thinking a twin thought is not a relevant alternative? Since
there is no Twin Earth, the possibility that one is on Twin Earth thinking twater thoughts
does not seem relevant. However, there are more realistic examples of slow switches in
which a subject is slowly switched between two linguistic groups in which a word has
different meanings (Ludlow 1995a; Butler 1997). (For instance, ‘professor’ and ‘chip’ have
different meanings in American and British English.) However, even if such examples
succeed in showing that slow switches are sometimes relevant, this leaves space for a
compatibilist to argue that they are not generally relevant (Warfield 1992; Sawyer 1999;
Brown 2004). If that were so, then semantic externalism would be compatible with the
claim that, typically, a subject can have a priori knowledge of her thoughts.
In summary, there are a number of avenues for the compatibilist to explore to provide a
full answer to the achievement problem. She may provide new arguments to show that
knowledge requires only reliability and not discriminative abilities; alternatively she may
argue that slow switches are not usually relevant. However, even if the compatibilist can
solve the achievement problem, she still faces the consequence problem, according to
which the joint assumptions of semantic externalism and privileged access have an
absurd consequence and are thus incompatible.
Page 6 of 16
Semantic Externalism and Self‐Knowledge
(p. 773)
There are three main ways for the compatibilist to reply to the consequence problem.
First, she could accept that a subject can have a priori knowledge of her thought contents
and of entailments from her thoughts to her environment but deny that she can thereby
acquire a priori knowledge of her environment (the denial of ‘transmission’). Second, she
could accept that a subject's a priori knowledge of her thoughts and the relevant
entailments provide a priori knowledge of her environment but deny that such knowledge
is problematic. Third, she could deny that it follows from semantic externalism that there
are a priori knowable entailments from thought to the environment. I will examine each
option in turn. (Since privileged access is an assumption for reductio, one cannot respond
to the consequence problem by denying that a subject can have a priori knowledge of her
thought contents.)
According to the first response, even if a subject knows a priori that she thinks that water
is wet and that if she thinks that water is wet then her environment contains water, she
cannot combine these pieces of knowledge to gain by inference a priori knowledge that
her environment contains water. In other words, the relevant inference is a counter‐
example to the principle of transmission for a priori knowledge, according to which if a
subject knows a priori that p and she also knows a priori that p entails q, then she can
know a priori that q by that inference (Davies 1998, 2000, 2003; McLaughlin 2000, 2003;
Wright 2000, 2002, 2003).
There are a number of problems with this first response. Some reject the claim that
transmission fails across the relevant inference (Brewer 2000; Beebee 2001; Pritchard
2002; McLaughlin 2003; Brown 2003, 2004; Pryor 2004; Silins 2005). However, a more
serious objection is that even if transmission does fail, this would not answer the
consequence problem. Notice that the claim that transmission fails across the inference
merely amounts to the claim that one cannot acquire a priori knowledge of the
environment in one way (namely via the target inference); it leaves it open whether one
has that a priori knowledge on some other basis. But plausibly the alleged absurdity
consists merely in the idea that one could have a priori knowledge that one's environment
Page 7 of 16
Semantic Externalism and Self‐Knowledge
contains water, however this knowledge is supposedly acquired (Brown 2003; McKinsey
2003). In that case, denying that one could acquire the a priori knowledge in one way,
namely via the inference, would not answer the consequence problem. For it would leave
it open that one has the a priori knowledge in some other way. We may put the point more
formally using the principle of closure. According to closure for a priori knowledge, if S
knows a priori that p, and she knows a priori that p entails q, then she can know a priori
that q, where it is left unspecified whether she knows that q via the inference, or
independently of that inference. Closure is weaker than transmission and may hold even
if transmission fails. Even if one cannot know a priori that q via inference from one's a
priori knowledge that p and that p entails q, one may have a priori knowledge that q on
some other basis. Indeed, on one prominent account of the failure of transmission,
whenever transmission fails, closure holds (Wright 2002). Assuming closure for a priori
knowledge, if one knows a priori that one thinks that water is wet, and that one's thinking
that water is wet entails that one's environment contains water, then (p. 774) one can
know a priori that one's environment contains water. On the assumption that it's absurd
to suppose that one has a priori knowledge that one's environment contains water
however acquired, then semantic externalism and privileged access are incompatible. It
seems, then, that the consequence problem relies only on closure for a priori knowledge.
So the problem is not resolved by arguing that the relevant inference is a counter‐
example to transmission.
Given the failure of the first response to the consequence problem, let us turn to the
second. The second response defends the compatibility of semantic externalism and
privileged access by arguing that it's not absurd that a subject could have a priori
knowledge of her environment. One way to support this claim would be by performing a
modus ponens on the incompatibilist's modus tollens. The incompatibilist argues that
since it is absurd that one could have a priori knowledge of one's environment, semantic
externalism and privileged access are incompatible. Some compatibilists counter that
since semantic externalism and privileged access are both true, it follows that one can
have a priori knowledge of one's environment. That one can have such knowledge is just
an epistemic consequence of the metaphysical theory of semantic externalism (Sawyer
1998; Warfield 1998). However, the compatibilist modus ponens is unlikely to convince
the incompatibilist. Compatibilists would be better served by a defence which goes
beyond such a modus ponens.
Some have tried to undermine the alleged absurdity of a priori knowledge of one's
environment by drawing attention to the conditions for having the concept water. Some
externalists suggest that a subject who meets the conditions for having the concept water
would thereby already know her environment contains water (Sawyer 1998; Brewer
2000). For instance, if she acquires the concept by interacting with water, say by drinking
it or washing in it, she would thereby acquire the empirical knowledge that her
environment contains water. If that's right, then there's no danger that a subject could
combine her a priori knowledge of her thoughts and externalism to gain new knowledge
about her environment. Rather, in order to think with the concept water and know that
she thinks that water is wet, she would already have empirical knowledge that her
Page 8 of 16
Semantic Externalism and Self‐Knowledge
environment contains water. However, this line of thought does not resolve the
consequence problem. The alleged absurdity in supposing that a subject has one kind of
knowledge, a priori knowledge, of some environmental fact seems in no way lessened by
supposing that she has some other kind of knowledge, empirical knowledge, of that same
fact (Davies 2000).
A different way of casting doubt on the alleged absurdity of the claim that a subject can
have a priori knowledge of her environment is to suggest that the relevant knowledge is,
in a certain sense, weak. Say that knowledge is ‘weakly’ a priori if that knowledge is
warranted otherwise than on the basis of empirical evidence, and ‘strongly’ a priori if, in
addition, it is empirically indefeasible (McLaughlin 2000). If one's knowledge of one's
thoughts and the entailments from thought to the world is strongly a priori, then one's
knowledge of the relevant environmental conditions would also be strongly a priori. It
does indeed seem implausible that one could have strongly a priori knowledge that one's
environment contains water, that is knowledge indefeasible by empirical evidence.
However, if one's a priori knowledge of either (p. 775) one's thoughts or the entailments
from thought to the world is weak, then one's a priori knowledge that the relevant
environmental condition obtains would also be weak. Understood in this way, would
knowledge of one's environment be any less absurd? It seems not. For it remains
problematic that one could know, say, that one's environment contains water without
basing that knowledge on any empirical evidence.
Page 9 of 16
Semantic Externalism and Self‐Knowledge
It may be, then, that this last way of filling out the second response could be developed to
show that it is not absurd that one have a priori knowledge of one's environment.
However, one would not need to defend the possibility of such a priori knowledge if the
consequence problem could be stopped at an earlier stage. This is precisely what the
third response attempts to do. The third response denies that it follows from semantic
externalism that there are a priori knowable entailments from thought to the world. A
first concern is whether semantic externalism establishes any entailments from thought
to the world, whether a priori knowable or not (Brueckner 1992; Gallois and O'Leary‐
Hawthorne 1996: 11–12). The standard Twin Earth thought‐experiments most obviously
establish a counterfactual dependence of thought on the world. For instance, Putnam's
thought‐experiment most obviously shows that Oscar who is actually on earth with the
concept of water would instead have had the concept of twater if he had counterfactually
been brought up on Twin Earth. This counterfactual claim does not entail that it is a
necessary condition for having the concept of water that one's environment now contains
water, or used to contain water (I leave the second disjunct implicit in what follows).
(p. 776)
In fact, further consideration shows that semantic externalism does establish entailments
from thought to the world. However, the difficulty for incompatibilists has been to show
that a subject could have the a priori knowledge of these entailments required by the
consequence problem. To illustrate this consider Putnamian semantic externalism. Let us
set aside chemically knowledgeable subjects who, as semantic externalists accept, can
have the concept of water in a waterless world. Instead, let us consider a subject who has
no idea what the chemical composition of water is. Semantic externalists accept that such
a subject may have the concept of water in an environment which contains water and/or if
she is part of a linguistic community with the concept of water. However, it's hard to see
how such a subject could have the concept of water if she were instead in an environment
without water or other speakers. What would determine that she has a concept which
applies to all and only H2O? So semantic externalism establishes the following principle:
(A) If S has the concept of a natural kind and is ignorant of its fundamental
properties, then either she is in an environment which contains instances of that
kind, or she is part of a linguistic community with the concept of that kind (Brown
1995).
Page 10 of 16
Semantic Externalism and Self‐Knowledge
To use (A) to gain a priori knowledge of a specific entailment between her thought and empirical
facts about the world, a subject needs a priori knowledge that she has the concept of a natural
kind. However, such knowledge is not available. It is an empirical matter whether a term which
is intended to name a natural kind actually does so. A term intended to name a natural kind may
fail to do so because the exemplars form a heterogeneous motley of several natural kinds (e.g.
‘jade’) or, more radically, because we were mistaken in thinking that there are such exemplars
(Gallois and O'Leary‐Hawthorne 1996: 6–7; Boghossian 1997: 168; McLaughlin and Tye 1998).
Nor does it help if, as Boghossian (1997) suggests, principle (A) is reformulated to concern the
notion of a term which is intended to name a natural kind. A subject can infer no fact about her
environment from the fact that she has a term which is intended to name a natural kind unless
she also knows that it is not the case that the relevant term fails to name a natural kind. But a
subject cannot know a priori that that's so, regardless of whether, in such a situation, the term
would express no concept, or express a concept different from that which it expresses when it
does name a natural kind. (For discussion of these options see Boghossian 1997; Brown 2004.) If
a subject intends a term to name a natural kind but it fails to do so then she would still believe
that it's not the case that it fails to name a natural kind. So even if she believes correctly that a
term does not fail to name a natural kind, this belief does not constitute a priori knowledge.
Incompatibilists may hope to avoid these problems by targeting the consequence problem
not just at natural‐kind concepts, but, rather, a broader range of concepts (Brown 2001,
2004: 300–1). To set up this broadening, notice that any concept whatsoever is partly
individuated by its application conditions. For instance, the natural‐kind concept water is
partly individuated by the fact that it applies to all and only (p. 777) H2O. The non‐
natural‐kind concept sofa is partly individuated by the fact that it does not apply to large
armchairs. Semantic externalists accept that a subject can have a concept even if she is
ignorant of its application conditions or, as Burge puts it (1979), if she ‘incompletely
understands’ it. One kind of incomplete understanding is misunderstanding where a
subject has a mistaken view about the kinds of thing to which a concept applies. For
instance, Burge's arthritis patient mistakenly thinks that arthritis applies to problems of
the thighs. A different kind of incomplete understanding is agnosticism where a subject is
unsure whether a concept applies to a certain kind of thing where there is a determinate
fact about whether the concept does, or does not, apply to that kind of thing (Burge 1979:
77). Note that merely being unsure about whether a concept applies to certain types of
thing need not amount to being agnostic about it in the sense defined, since there may be
no fact of the matter whether the concept applies to the relevant type of thing or not; in
that case, the subject would not incompletely understand the concept. Semantic
externalists accept that a subject who incompletely understands a concept in either of
these senses may possess that concept if she is part of a linguistic community with that
concept and/or, in the case of natural‐kind concepts, if her environment contains
instances of the relevant kind. However, it seems hard to understand how a subject could
have a concept despite incompletely understanding it unless her environment aids her in
one of these two ways. Thus,
(B) If S has a concept despite incompletely understanding it, then either she is part
of a linguistic community with that concept or, in the case of a natural kind concept,
she is in an environment which contains instances of the relevant kind (Brown 2001).
Page 11 of 16
Semantic Externalism and Self‐Knowledge
A subject could use (B) to gain a priori knowledge of a specific entailment between her thoughts
and the world only if she knows a priori that she has a concept despite incompletely
understanding it. However, a subject cannot know a priori that she misunderstands a concept.
For she would need empirical information about linguistic usage to know that she has an
incorrect view of how the concept is applied. Further, a subject cannot know a priori that she is
agnostic about a concept. A subject is agnostic about a concept when she is unsure whether it
applies to things of a certain kind when there is a determinate fact about whether or not it
applies to things of that kind. While a subject may know a priori that she is unsure whether the
concept applies to things of a certain kind, she cannot know a priori that there is a determinate
fact about whether it applies to things of that kind as opposed, say, to her mistakenly thinking
that there is such a determinate fact (Brueckner 2002; Brown 2004). On the semantic‐externalist
view defended by Burge, incomplete understanding of concepts is rife, where such incomplete
understanding may include thinking that there is a determinate fact about whether a concept
applies to things of a certain kind, even though there is not.
In conclusion, the consequence problem fails to show that semantic externalism is
incompatible with privileged access. For it does not follow from semantic externalism that
there are a priori knowable entailments from thought to the environment which (p. 778) a
subject could use to gain a priori knowledge of her environment. Earlier we saw that
there are several potential ways for a compatibilist to answer the achievement problem,
either by showing that knowledge requires only reliability and not discriminative abilities,
or by showing that twin thoughts are not relevant. If such lines of thought can be
fruitfully developed, then semantic externalism would not, after all, be incompatible with
privileged access.
References
Beebee, H. (2001), ‘Transfer of Warrant, Begging the Question and Semantic
Externalism’, Philosophical Quarterly, 51: 356–74.
—— (1997), ‘What the Externalist Can Know A Priori’, Proceedings of the Aristotelian
Society, 97: 161–75.
Page 12 of 16
Semantic Externalism and Self‐Knowledge
Burge, T. (1979), ‘Individualism and the Mental’, Midwest Studies in Philosophy, 4: 73–
121.
DeRose, K. (2000), ‘How Can We Know that We're Not Brains in Vats?’ Southern Journal
of Philosophy, suppl. vol. 37: 121–38.
Page 13 of 16
Semantic Externalism and Self‐Knowledge
Heal, J. (1998), ‘Externalism and memory’, Proceedings of the Aristotelian Society, 77:
95–110.
McLaughlin, B., and Tye, M. (1998), ‘Content Externalism and Privileged Access’,
Philosophical Review, 107: 349–80.
Page 14 of 16
Semantic Externalism and Self‐Knowledge
Tye, M. (1998), ‘Externalism and Memory’, Proceedings of the Aristotelian Society, 72:
77–94.
—— (1998), ‘A Priori Knowledge of the World: Knowing the World by Knowing our Minds’,
Philosophical Studies, 92: 127–47.
—— (2002), ‘Anti‐sceptics Simple and Subtle: Moore and McDowell’, Philosophy and
Phenomenological Research, 65: 330–48.
Jessica Brown
Page 15 of 16
Self‐Deception
Print Publication Date: Jan 2009 Subject: Philosophy, Moral Philosophy, Philosophy of Mind
Online Publication Date: Sep 2009 DOI: 10.1093/oxfordhb/9780199262618.003.0046
Keywords: self-deception, contemporary philosophy, irrationality, lying model, epistemic support, intentional
action
SOMETIMES we have to face facts we'd just as soon ignore, give up on hopes that once seemed
realistic, and accept unpleasant surprises. We all experience rude awakenings. When something
disappointing or even shocking transpires, sooner or later we adjust—but not always. There are
times when we respond differently and put our heads in the sand. In the language of pop
psychology, we are unable to ‘face up’ to things that are ‘too painful to deal with’ and ‘go into
denial’. Sometimes just the opposite occurs: we can't stop dwelling on some terrible possibility
and, as a result, we're convinced it's for real. Naturally it's easier to see these tendencies in
others than in ourselves. Their thinking on some matter strikes us as biased, distorted, or
clouded by how they would like things to be or how they fear things are. They seem irrational,
and their irrationality seems motivated.
Irrationality can occur without being motivated, as shown by the widespread biases,
flaws, and errors in human reasoning identified by cognitive psychologists, but motivation
or emotion can surely add to it. Your rational abilities can be affected by the sudden death
of a loved one or the unexpected news that you have cancer. So it should not seem
puzzling that grief and dread, or other powerful emotions like rage and panic, can disrupt
Page 1 of 19
Self‐Deception
normal thinking. Our concern, though, is with something less extreme. Self‐deception is
of special interest, and puzzling, partly because it does not disrupt one's overall rational
faculties and, if anything, exploits them. When we spot instances of it, we wonder how
even intelligent people can come to believe things for reasons that do not begin to
support these beliefs, and how can they fail to believe things for which the evidence is
obvious and compelling. How can normal people's thinking, though usually fairly
responsive to rational and evidential considerations, (p. 782) be distorted by motivational
factors in regard to particular subjects? The self‐deceiver seems aware of the relevant
evidence and of its import, but manages to resist it somehow. We think they ‘should know
better’.
The main philosophical question asked about this phenomenon is prompted by its very
name. Is self‐deception really a matter of deliberately getting oneself to believe
something contrary to something else one already believes and, if so, how can one
succeed at it? Colloquial phrases for it, like ‘fooling yourself’ and ‘lying to yourself’,
evoke paradox. Although our discussion will stress the more recent philosophical work on
the subject (see Mele 1987 for a survey of work up to that time), most work on self‐
deception has focused on this paradox and sought ways to clear the air of it. Space
limitations preclude taking up the scattered remarks made about self‐deception (or
relevant to it) by such diverse thinkers as Plato, Aristotle, Butler, Hume, Kierkegaard,
Marx, Nietzsche, and Sartre, and possible connections between self‐deception and the
various ‘defence mechanisms’ proposed by Freud, such as denial, repression, projection,
and reaction formation. We will focus on the puzzles and questions of contemporary
philosophical interest.
Page 2 of 19
Self‐Deception
Like other words ending in ‘‐tion’, the word ‘self‐deception’ exhibits a process– product
ambiguity. It can refer to either the process of deceiving oneself or to the state of being
self‐deceived. Both seem paradoxical on the lying model. It represents (p. 783) the
product of self‐deception, being self‐deceived, as holding contradictory beliefs at the
same time (at least temporarily) and the process that leads to it, deceiving oneself, as
driven by an intention to form a belief that conflicts with a belief one already has.
However, if one were aware that one had such an intention, one could not carry it out. No
wonder paradox looms, never mind psychological implausibility. It was no surprise that
shortly after being endorsed by Demos the lying model was forcefully challenged
(Canfield and McNally 1961; Siegler 1962, 1963).
The lying model is paradoxical precisely because it views self‐deception on the model of
deceiving someone else. When you lie to someone, you tell them something that you
yourself do not believe, with the aim of getting them to believe it. (It is convenient to
assume that a lie—or a self‐deception—must be false, but what really matters is not its
truth‐value but that the liar disbelieve or at least doubt it.) Assuming you don't provide
any tell‐tale evidence of your deceitfulness, your deception can succeed if they find you
trustworthy and find what you tell them credible. There's nothing puzzling about what
happens in deceiving another, because the deceiver and the deceived are two different
Page 3 of 19
Self‐Deception
people. But how can one deceive oneself in this way? How can you lie to yourself and
succeed? Presumably the self‐deceiver's intention must be conscious if it is to be carried
out. But how, in so far as the self‐deceiver is aware of what he's doing, could he keep
from thwarting it in the course of trying to carry it out?
It is important to appreciate that the trouble with the lying model isn't merely that it has
the self‐deceiver being both agent and patient. After all, there's nothing puzzling about
feeding oneself or washing oneself. Deceiving oneself seems puzzling more in the way
that fighting oneself or carrying oneself might seem puzzling, at least until we notice that
what we call ‘fighting oneself’ and ‘carrying yourself’ are not the same sorts of things as
we do when we fight or carry someone else. More apt is the case of informing oneself.
When you inform someone else of something, you provide them with information you
already have. However, informing yourself can't be like that. You seek out and acquire
information that you don't already have. Similarly, perhaps, deceiving yourself is not like
deceiving someone else.
Page 4 of 19
Self‐Deception
One idea is that the self‐deceiver's intention is unconscious. This would explain why he
disavows what he's doing, but that explanation seems ad hoc at best. Or one might adopt
a compartmental or even homunculus model. Here the idea is to avoid paradox by
claiming that the two beliefs are segregated or partitioned, perhaps one confined to the
unconscious or each belonging to a different subperson. The strategy, of course, is to
dissociate the victim from the intending agent, so as to avoid representing either as
inconsistent or irrational. It is true that people sometimes say that ‘part of me’ thinks this
and ‘part of me’ thinks that, but it is hard to take such language literally. However, such
ideas generate mysteries of their own. Besides, as Johnston points out, if self‐deception
really were a matter of one subperson deceiving another, as on the homunculus model, its
motivation would be a mystery. For even if there were identifiable subpersons of the
required sorts—a culprit and a victim—why would one's deceiving the other even be
relevant to the person's being deceived? We can grant that ‘as a result of his own activity
[the self‐deceiver] gets into a state in which he is misled, at least at the level of conscious
belief’ without accepting the presupposition, which generates the paradox of self‐
deception, that this is the ‘reflexive case of lying’ (Johnston 1988: 65).
Page 5 of 19
Self‐Deception
term and seeking an account that rings true psychologically and does not give rise to
paradox. That would be preferable to an account on which the questions ‘What is self‐
deception?’ and ‘How is it possible?’ do not have compatible answers.
The same ideas can be put without the p's and q's. Whereas a belief about something
normally causes one to think the very thing one believes (when the subject comes up),
this is not what happens in self‐deception. Rather than adopt a new, contrary belief, what
the self‐deceiver does is keep himself, at least on a sustained and recurrent basis, from
thinking what he believes (or would otherwise believe). No contrary belief is needed to
suppress or inhibit the effect that the unpleasant belief normally has on his thinking. It is
enough for the self‐deceiver to clutter his mind with contrary thoughts. In that way the
truth of the warranted belief is kept from being settled for him.
Page 6 of 19
Self‐Deception
(p. 786)
The obvious objection to Rey's view is that avowed belief is not really a kind of belief, just
as counterfeit money is not really money. People don't always believe what they sincerely
avow; a belief does not have to be possessed to be sincerely expressed. It is enough that
the self‐deceiver seem to have it, and from his own point of view avowing it is the way of
seeming to have it. If the proposition that p is what he is deceiving himself about, it is
enough that he think that p whenever the thought of p occurs to him. The self‐deceiver's
‘avowed belief’ that p is his thought that p and what he would sincerely avow to others as
his belief, but not what he actually believes. After all, this is what we mean when we
describe a self‐deceiver as refusing to admit or to face up to something; we are accusing
him of disavowing a belief and avowing something else.
Avowal plays a role in other accounts of self‐deception as well. For example, Robert Audi,
who distinguishes the state of being self‐deceived from the act of deceiving oneself,
claims that someone who is self‐deceived that p must sincerely avow that p, or at least be
disposed to sincerely avow it (1988: 94). Audi requires further that the self‐deceiver
unconsciously know (or at least believe) that not‐p. Brian McLaughlin agrees with Audi
that self‐deception involves an unconscious or, as he calls it, an ‘inaccessible’ belief
(1988: 48–53). However, it seems that the suppressed belief can't be ‘too’ unconscious or
inaccessible for, as Allen Wood has observed, in cases of self‐deception (as opposed to
traumatic repression, for example), ‘the psychically upsetting awareness is dangerously
close at hand’ (1988: 359). It is tempting to indulge in Freudian metaphors here and say
that the suppressed belief, assuming that being self‐deceived involves having one, must
be relatively ‘near the surface’ rather than ‘deep in one's unconscious’, and threaten to
become conscious. This language can be avoided by capturing Wood's observation
dispositionally: being self‐deceived involves a disposition both to resist consciously
thinking something that one believes (or at least takes there to be strong evidence for
and normally would believe) and to affirmatively think something contrary to that
proposition. Indeed, in so far as having this self‐deceptive thought keeps one from
thinking the suppressed one, this is really just one disposition.
Page 7 of 19
Self‐Deception
What does it take to exercise that disposition? That is, what keeps one from affirmatively
thinking the nasty thought about the touchy subject? One obvious technique is
rationalization, the process of fabricating plausible but phoney reasons for rejecting the
warranted proposition and for buying into the preferred alternative. Rationalization can
both get one into the state of being self‐deceived and, when the touchy subject comes up,
keep one in that state.
Rationalization is not the only technique for keeping a nasty thought from occurring or
for suppressing it when it does occur. Johnston (1988), McLaughlin (1988), and I (Bach
1981) have each identified others. One technique is evasion; that is, diverting one's
attention from the troublesome subject and focusing it elsewhere, with one's imagination
supplying alluring possibilities. Even though one appreciates the weight of the evidence,
one avoids the implications of that evidence by thinking about other things. Evasion, in
the form of changing the subject, is especially effective in interpersonal contexts.
(p. 787)
Another technique falls somewhere between evasion and rationalization: with the help of
a good imagination, one conjures up fanciful alternatives to the warranted proposition
and implicitly uses these mere possibilities to keep the case for that proposition
unsettled. I have called this technique ‘jamming’, because of the radio/radar analogy
(Bach 1981: 361–2). One sticks to the subject but clutters one's mind with contrary
thoughts. To the extent that one's mind is redirected to appealing though epistemically
idle possibilities, one avoids thinking as if one believed that proposition. The persistent
occurrence of such thoughts provides a kind of perverse evidential support for the self‐
deceptive proposition.
Evasion and jamming play supporting roles in the process of deceiving oneself and
keeping oneself self‐deceived, by diverting the self‐deceiver from his occasional doubts.
Rationalization provides the foundation. It plays for him the role that the party line or
official doctrine plays in a political, religious, or other social group (family, business,
charity), providing not only canonical support for what he thinks but also the terms for
spinning it to others.
Page 8 of 19
Self‐Deception
deception the twisted kind involves motivated belief and mishandling of evidence, it poses
a special puzzle, since its motivation is not the result of desire. Why, after all, would
anyone deceive himself into thinking something he would rather not be so?
It might be objected that the case of twisted self‐deception does not really undermine the
claim that self‐deception is invariably motivated by desire. Of course, one could construe
desire so broadly as to include motivation of any kind, but that would be to trivialize the
objection. Rather, the idea is that the negative emotions that can give rise to self‐
deception, such as fear, anger, envy, and jealousy, have desires associated with them and
that these are what motivate the self‐deception (see Mele 2003). However, it seems to me
that to understand the motivation behind twisted self‐deception we need to look deeper.
Compare and contrast twisted and straight self‐deception. In the straight case the self‐
deceiver refuses to confront a painful truth and draw the obvious conclusion from the
strong evidence available to him. He thinks up reasons to the contrary, however flimsy,
conjures up alternative possibilities, however idle, or just gets his mind off the unpleasant
subject. In the case of twisted self‐deception, on the other hand, the self‐deceiver is
preoccupied with the subject. Whether it's his wife's (p. 788) infidelity, his boss's lack of
appreciation for his accomplishments, or his neighbour's vendetta against him, he lacks
solid evidence for the unwelcome state of affairs and yet is disposed to treat isolated bits
of flimsy evidence as showing that it obtains. Yet presumably he doesn't want, as the case
may be, his wife to be unfaithful, his boss not to appreciate him, or his neighbour to have
it in for him. So what motivates his thinking?
It seems to me that in such cases the source of the motivation for the twisted self‐
deceiver's thinking is his inability to get his mind off the subject. (That in turn requires
further explanation, of course.) The mere possibility of what he fears is enough to keep it
in mind, and its staying there is enough to make it seem far more realistic than it actually
is. The more realistic it seems, the more significant seem the stray bits of flimsy evidence
for it and the more easily the evidence against it can be dismissed. The twisted self‐
deceiver weighs the evidence in accordance with the degree to which it occupies his
attention. He dismisses some evidence not because it seems weak; rather, it seems weak
because he dismisses it, because he can so easily put it out of mind. Meanwhile, the
threads of evidence that support his fear seem strong because thoughts of them keep
recurring. Of course, he doesn't treat the degree to which a piece of evidence occupies
his attention as reason to weigh it accordingly; rather, its degree of occupying his
attention causes him to weigh it accordingly—and he doesn't realize that this is why he so
weighs it. What motivates the twisted self‐deceiver's thinking, then, is the fear itself (or
whatever the anxious emotion).
The twisted self‐deceiver can't get his mind off some undesired possibility, whereas the
straight self‐deceiver can't face up to some undesired actuality. Because he is consumed
with that possibility, the twisted self‐deceiver keeps thinking of evidence in its favour to
the exclusion of evidence against it. In contrast, the straight self‐deceiver, because he
can't face up to a certain actuality, dismisses its very possibility. He downplays, dismisses,
Page 9 of 19
Self‐Deception
or even disregards pieces of evidence for it. Evidence favouring the alternative to that
actuality keeps grabbing his attention and makes that alternative seem not only plausible
but convincing. Despite these differences, the two types of self‐deceivers have something
in common: their thinking is driven by possibilities and considerations they attend to and
is resistant to possibilities and considerations they downplay or disregard. Self‐deception
is a kind of self‐distraction. This is not an intentional act, but a motivated process.
Page 10 of 19
Self‐Deception
jump to conclusions but also need to know when to think twice (see Bach 1984). Whatever
the ultimate verdict about the rationality of normal human cognitive processes, it is
plausible that the phenomenon of self‐deception, though motivated, is continuous with
normal, unmotivated cognitive phenomena.
This is the basis for Alfred Mele's version (1997, 2001) of the deflationary approach.
According to his account, which focuses on entering self‐deception rather than
maintaining it (and assumes that the proposition in question must be false), a person
enters self‐deception in acquiring a belief that p if although his data favour not‐p over p
the person treats these data in a motivationally biased way and is thereby caused, in a
non‐deviant way, to come to believe that p. However, as Dion Scott‐Kakures (2002) has
argued, these conditions are not strong enough to exclude certain motivated beliefs that
do not count as self‐deceptive. He argues that to count as deceiving oneself, rather than
merely be engaged in wishful or anxious thinking, a person must exercise a sophisticated
reflective capacity, however deficiently. Genuine self‐deception requires misapprehending
the effect of desire (p. 790) or anxiety on one's thinking and misjudging the degree to
which one's belief is warranted. The self‐deceiver ‘engages in various cognitive
strategies, encounters recalcitrant data that must be explained away, generates friendly
theories, [whereas] in wishful thinking a proposition is precipitously and directly
embraced’ (2002: 600).
What distinguishes both Mele's and Scott‐Kakures's accounts from other deflationary
accounts is their suggestion that the processes that subserve self‐deception are
continuous with those underlying ‘unmotivated or accuracy directed hypothesis
testing’ (Scott‐Kakures 2002: 599). Mele thinks that the route to error in self‐deception is
no different in kind from motivationally neutral routes. He specifically mentions
overweighting positive evidence, downplaying negative evidence, selective evidence‐
gathering, and selective attention. To be sure, motivation can amplify these tendencies,
but they don't have to be motivated to occur. Indeed, in moderation these tendencies
support a reasonable degree of epistemic conservatism. As some epistemologists and
philosophers of science advise, there is much to be said for not being too quick to
challenge one's core beliefs or entrenched theories. So perhaps there is no clear
boundary between reasonable conservatism and excessive caution or even outright
dogmatism or, for that matter, between motivated and unmotivated irrationality, and
perhaps the same processes are at work.
To illustrate this idea, Mele and Scott‐Kakures both invoke a particular cognitive model,
according to which people implicitly adopt different ‘confidence thresholds’ for accepting
and for rejecting a given hypothesis (see Friedrich 1993; Trope and Lieberman 1996).
People do this because they differentially assess the cost of wrongly believing something
and of wrongly believing its negation. Costs and confidence thresholds affect both how
hypotheses are evaluated and how far one is willing to go in testing them before making
up one's mind. How does this bear on self‐deception? The self‐deceiver may cling to a
certain proposition, say that his child is innocent of a horrible crime, when the cost of
wrongly believing it is much less than the cost of wrongly believing its negation, that his
Page 11 of 19
Self‐Deception
son is guilty. This makes the proposition much easier to believe than to disbelieve.
Recognizing that self‐deception is not a one‐shot affair but requires ongoing resistance to
the emergence of the truth, Scott‐Kakures suggests that ‘in self‐deception motivation
plays a continuing or recurring role—mediated by the asymmetry in error costs—in the
generation of belief’, adding that ‘in wishful thinking it plays a merely triggering
role’ (2002: 600). However, the process thus motivated is of a piece with normal
cognition.
Page 12 of 19
Self‐Deception
Considerations of this sort have led some to suppose that self‐deception is inherently
about oneself, that being self‐deceived about anything includes being self‐deceived about
oneself. Indeed, Richard Holton (2001) contends that the role of the word ‘self’ in ‘self‐
deception’ is to indicate not that self‐deception is action on oneself but, like self‐
promotion or self‐effacement, that it is essentially about oneself. People do, after all,
deceive themselves about their motives, as when they invent explanations for their
shameful deeds, and about their abilities, as when they make up excuses for their abject
failures. And, no doubt, people who are nasty, obnoxious, phoney, incompetent, or boring
often seem to make an active effort to deceive themselves into thinking otherwise. Of
course, examples like this do not show that self‐deception is always about oneself, but
perhaps that is true in an indirect way.
Consider the case of the mother who ‘can't believe’ that her son, the prime suspect in a
murder case, is a killer. Like people whose self‐deception is about themselves, she shows
little self‐reflection and self‐criticism in the area of her self‐deception. David Sanford
gives another example, of a person who makes cruel jokes and, without appreciating their
impact on their victims, always thinks of himself as ‘just kidding’ (1988: 164). In general,
Sanford suggests, self‐deceivers ‘misapprehend [their] attitude structures’, and, as
Annette Barnes adds, they fail to appreciate how their desires, wishes, fears, or motives
have guided their reasoning on the subject and shaped their beliefs about it (1998: 102).
However, it does not seem that this lack of self‐reflection necessarily amounts to further
self‐deception. In particular, failure to recognize one's attitudes or to appreciate their
impact on one's thinking doesn't entail being positively mistaken, much less self‐deceived,
about them as well as the matter one is primarily self‐deceived about. Besides, in so far
as most people are not generally reflective or self‐critical about their reasoning anyway, it
seems excessive to accuse them of being self‐deceived about themselves whenever they
are self‐deceived about anything else.
Page 13 of 19
Self‐Deception
Then there is the matter of social role‐playing and the question of authenticity, of ‘being
true to oneself’, as opposed to what Sartre (1958) called mauvaise foi (bad faith). We play
along with other people's expectations of what we are. We get caught up in our social
roles, such as being a waiter, a gourmet, or a doctor, and do not spell out to ourselves
what we are really doing (Fingarette 1969/2000). Instead, like politicians we feign
sincerity and feign acceptance of others' sincerity. If we all do this, who's fooling who?
How do we distinguish hypocrisy from self‐deception (see Statman 1997)?
Finally, there is an epistemological question about self‐deception and other people: How
do we know when we've correctly identified instances of it? Perhaps we attribute self‐
deception to people much more readily than we should. Maybe we can't think of a better
explanation for why they believe something that strikes us as completely absurd. We fail
to recognize, even after questioning and challenging them, that they have a set of core
beliefs about the subject matter from the perspective of which the suspect belief makes
perfectly good sense (for example, that everything happens for a good reason or that
one's child is incapable of violence). So we think they must be deceiving themselves.
Similarly, we may suppose that people are more thoughtful than they really are, hence
that they must be making the special effort characteristic of self‐deception in order to
think as they do. For example, we suppose that they must be unconsciously aware of the
obvious character flaws or annoying habits that we associate with them and must
therefore do something to hide these traits from themselves.
Page 14 of 19
Self‐Deception
Self‐deception raises various other questions. For example, what is the relationship
between self‐deception and weakness of will? Arguably, self‐deception is itself a case of
weakness of will: just as weakness of will involves acting against one's better judgement,
so self‐deception involves thinking against one's better judgement. And certainly self‐
deception can facilitate weak‐willed action by leading to error about what one's better
judgement is. Then there are moral questions pertaining to self‐deception. Is deceiving
oneself something one can't help, or is one responsible for it? (p. 793) If the latter, is it
always blameworthy? When we attribute self‐deception to someone, we tend to regard
them as guilty of it, but perhaps there is something to be said in its favour, at least
sometimes. Shelley Taylor (1989) documents cases in which people benefit from believing
that they are fully recovered from cancer even if they are not, have more control over
their surroundings than they actually do, will succeed at tasks that they are likely to fail
at, and so on. Indeed, there is plenty of psychological evidence that most people, or at
least most college students (the usual subjects of studies), think of themselves as above
average (this is the so‐called Lake Wobegon effect), or at least not below average, as to
intelligence, attractiveness, etc.
References
Audi, R. (1988), ‘Self‐deception, Rationalization, and Reasons for Acting’, in B.
McLaughlin and A. Rorty (eds.), Perspectives on Self‐deception (Berkeley, Calif.:
University of California Press), 92–120.
Canfield, J., and McNally, P. (1961), ‘Paradoxes of Self‐deception’, Analysis, 21: 140–4.
Page 15 of 19
Self‐Deception
Gigerenzer, G., and Selten, R. (2001) (eds.), Bounded Rationality: The Adaptive Toolbox
(Cambridge, Mass.: MIT Press).
Doubleday).
Holton, R. (2001), ‘What is the Role of the Self in Self‐deception?’, Proceedings of the
Aristotelian Society, 101: 53–69.
Kahneman, D., Slovic, P., and Tversky, A. (1982) (eds.), Judgement Under Uncertainty:
Heuristics and Biases (Cambridge: Cambridge University Press).
—— (1997), ‘Real Self‐deception’, with open peer commentary and author's response,
Behavioral and Brain Sciences, 20: 91–136.
Nisbett, R., and Ross, L. (1980), Human Inference: Strategies and Shortcomings of Social
Judgment (Englewood Cliffs, NJ: Prentice‐Hall).
Page 16 of 19
Self‐Deception
Sartre, J. P. (1958), Being and Nothingness, trans. Hazel E. Barnes (New York: Washington
Square).
Scott‐Kakures, D. (2000), ‘Motivated Believing: Wishful and Unwelcome’, Noûs, 34: 348–
75.
Taylor, S. E. (1989), Positive Illusions: Creative Self‐deception and the Healthy Mind (New
York: Basic).
Trope, Y., and Lieberman, A. (1996), ‘Social Hypothesis Testing: Cognitive and
Motivational Mechanisms’, in E. Higgins and A. Kruglanski (eds.), Social Psychology:
Handbook of Basic Principles (New York: Guilford), 239–70.
Further Reading
Ainslie, G. (2001), Breakdown of Will (Cambridge: Cambridge University Press).
Dupuy, J.‐P. (1998) (ed.), Self‐deception and Paradoxes of Rationality (Stanford, Calif.:
CSLI).
Page 17 of 19
Self‐Deception
Goleman, D. P. (1985), Vital Lies, Simple Truths: The Psychology of Self‐deception (New
York: Simon & Schuster).
—— (1990), ‘The Case for Motivated Reasoning’, Psychological Bulletin, 108: 480–98.
Quattrone, G., and Tversky, A. (1984), ‘Causal Versus Diagnostic Contingencies: On Self‐
deception and on the Voter's Illusion’, Journal of Personality and Social Psychology, 46:
237–48.
Page 18 of 19
Self‐Deception
Sackheim, H. A., and Gur, R. C. (1985), ‘Voice Recognition and the Ontological Status of
Self‐deception’, Journal of Personality and Social Psychology, 48: 1365–8.
Silver, M., Sabini, J., and Miceli, M. (1989), ‘On Knowing Self‐deception’, Journal for the
Theory of Social Behaviour, 19: 213–47.
Whisner, W. N. (1998), ‘A Further Explanation and Defence of the New Model of Self‐
deception: A Reply to Martin’, Philosophia, 26: 195–206.
Wilkes, K. V. (1994), ‘Psychology and Politics: Lies, Damned Lies and Self‐deception’,
Philosophy, 37: 115–29.
Kent Bach
Page 19 of 19
Index
Index
The Oxford Handbook of Philosophy of Mind
Edited by Ansgar Beckermann, Brian P. McLaughlin, and Sven Walter
Print Publication Date: Jan 2009 Subject: Philosophy Online Publication Date: Sep 2009
non‐human 67, 185, 229–32, 244, 342, 382, 402, 405, 448, 604–6, 665, 669, 683
anomalism, see also anomalous monism
psychological 96–9
psychophysical 99–102
anomalous monism 35–8, 95–108, 111–13
and supervenience 105–7
Anscombe, G. E. M. 98, 494 n., 667, 694 n. 9
anthropomorphism 340
anti‐individualism, see individualism, anti‐
Antony, L. 97, 111, 119, 607, 622 n.
a posteriority 316–19, 325–8
appearances 515–20
and self‐knowledge 751, 753, 761, 762
a priority 315, 317, 319, 324–32, 673, 674
transmission of 773, 774
Ariew, A. 387 n. 1
Aristotle, 152, 193, 208, 241, 275, 500–2, 504, 678, 679, 702, 782
Armon‐Jones, C. 679
Armstrong, D. 131, 181, 200, 201 n., 241, 488, 755, 756 n. 15
artificial intelligence 132
ascription, see attribution
Ashwell, L. 324
associationism 626–8
attention
focal 499, 501, 502, 568, 573
and self‐deception 785–93
attitudes, propositional 49, 95, 100, 101, 113, 117–20, 346, 355, 361–3, 407–19, 459–69, 478–81
and attitude predicates 408–18
as causally efficacious 407, 413
as semantically evaluable 407, 413
attributions of 100, 101, 355, 413–6, 633–5, 640, see also folk psychology
core vs. derivative cases of 414
existence and reality of 418
individuation of types of 412–7, 425, 426, 432–5
measurement‐theoretic account of, see measurement‐theory and propositional attitudes
attribution
of belief 356
of content 356–64
of mental concepts 296, 298, 305
of propositional attitudes, see attitudes, propositional
Audi, R. 665, 692, 693 n. 7, 699, 703, 786
Augustine, 729–31, 738
Aunt Bubbles machine 618–24
autism 719
autonomy, see free will
Avramides, A. 88, 727, 731 n. 6, 737 n.
Aydede, M. 304
awareness 260, 261, 270–4, 278, 498–505
Page 2 of 37
Index
Baddeley, A. 663
Bain, D. 475 n. 3, 479–81, 488
Baker, L. 45, 49, 68, 72, 73, 93 n. 7, 109, 111, 116 n. 7, 120, 121 n. 15 & 18, 131, 588 n.
Balog, K. 292, 301, 306, 308, 324
Bar‐On, D. 751, 752
Barkow, J. 721
Barnes, A. 791
Barnes, J. 208
Baron‐Cohen, S. 344
Bayne, T. 566, 567, 570, 573, 574 n.
Bealer, G. 300, 762
Beardsley, M. 692
Bechara, A. 688
Beckermann, A. 152, 155, 156, 169 n. 24, 408
Beebee, H. 773
behaviour 731–9
explaining 284–9, 339–49, 351, 356, 360–5, 713–25
predicting 339–49, 351, 356, 360–5, 713–25
behaviourism
in psychology 569
logical 129–31, 152, 153, 211, 302, 733, 737
belief 226–8, 233, 239–49, 352–6, 361–5, 478–80, 631, 633–5, 638–40, 692–709, see also attitudes,
propositional
and rationalization 784–7
and self‐deception, see self‐deception
avowal 782–6, 791
as reliable indicator 364
as source of information 364
de dicto vs. de re 228
dispositional vs. occurrent 785, 786
belief box 346
Benecke, E. 88
Bennett, J. 82, 131, 142, 271, 570, 581 n. 8, 633
Bennett, K. 46, 63
Bennett, M. R. 344
Berkeley, G. 190, 194, 198, 199, 202, 210, 546, 626, 643, 653, 730 n., 731, 738
Bermúdez, J. L. 457, 458, 459, 461 n., 463 n., 467, 469, 636–9, 641, 723, 741 n. 2
Bernecker, S. 770
Bickle, J. 43 n.
Bilgrami, A. 727 n. 2, 732 n. 9
binding problem 556
biosemantics 394–406
Bird, A. 430
Bishop, J. 692 n. 5, 701 n. 20, 702–5
Page 3 of 37
Index
Page 4 of 37
Index
Casey, E. 599 n. 9
Cassam, Q. 569 n. 3
(p. 799) Caston, V. 500 n. 12, 502 n. 16
Page 5 of 37
Index
315–20, 326–9, 332, 363 n. 20, 376–8, 475 n. 3, 477 n. 5, 479, 480 n., 481, 486 n., 487, 497 n. 7,
503, 504 n. 23, 514, 534, 566, 567, 570, 573, 574 n., 666, 753, 754
Chapman, L. J. 688
Cheney, D. 344, 611
Chihara, C. S. 732–5, 737, 739
Child, T. W. 97, 99–102, 107, 434 n. 20
Chinese Room 620–4
Chisholm, R. M. 75, 753
choice 80–2, 339 see also free will
Chomsky, N. 184, 356, 612, 734
Christensen, D. 666, 674
Church, A. 614
Church, J. 524
Churchland, P. M. 42 n., 148, 207 n., 391, 408, 714
Churchland, P. S. 391
Clapp, L. 119, 120 n. 12, 122 n. 20
Clark, Andy, 633, 636, 637, 645, 666
Clark, Austen, 268–70, 271–3
Clarke, D. 553, 554 n.
Clifford, W. 214
closure, causal 38–45, 53–65, see also causal closure principle
evidence for 55–60
of the physical 38–42, 53–65, 78–83, 174, 185, 217
cognitive science 46, 48, 344, 361, 383, 458, 532, 665
coincidence 583–7
commitment 693–5
communication 233, 234
community, linguistic 351–5, 364, 365, 424, 425, 450
computationalism, 48, 625
Cartesian, 614
computational theory of mind, see computationalism
conceivability
and possibility 78, 79, 173, 284, 314–333
argument against materialism 314
argument for dualism 69–74
strong vs. weak versions 69
clear and distinct 331
ideal 315, 316, 318, 319, 324, 332
positive vs. negative 301 n. 28, 315–19, 324–6, 329
of materialism 332
of disembodiment 69
of dualism 314
of zombies, see zombies, conceivability of
prima facie 315–18, 324, 331, 332
primary vs. secondary (1‐ vs. 2‐) 317–19, 324
varieties of 315–19
concepts 227–9, 400, 404, 405, 437–55, 515, 518, 520, 527, 530, 531
acquisition of 227, 441
Page 6 of 37
Index
Page 7 of 37
Index
subjective 292
connectionism 626–8, 665, 666
consciousness 30, 35, 36, 41, 42 n. 15, 132, 139–49, 173, 313, 314, 320–4, 328, 332
and experiences 243–5, 281–90, 494–509, 513–33
and introspection, see introspection and consciousness
and non‐conscious thought, see thought, non‐conscious
and personal identity, see identity, personal
and reference 648–62
and self 669, 670
and what it’s like 211, 243, 247–50, 269, 270, 276, 283–90, 292–4, 300, 495, 496, 503
as a three‐place relation 657–9
as fiction 207 n. 2
as glow 648–54
co‐ 566
creature vs. state 239, 244 n. 6
degrees of 144
disorders of 570–3
hard problem of 281, 534
higher‐order theories of, see HOT (higher‐order thought)
in matter 191–8
inner sense view of 241–3, 246, 250, 753–60
locus of 554–61
phenomenal 77, 225, 236, see also qualia
relational vs. intrinsic 287
representationalist theories of, see representationalism
sameness of 581–3
stream of 227, 573
unity of 565–76
pathologies of the 570–2
content 244–50, 262–6, 351–65, 457–72
and epistemic norms 450, 451
and language 458, 459, 468–70
and self‐knowledge 743–54, 757–62, see also self‐knowledge
and the perspectival constraint 461–3
attribution of, see attribution of content
broad, see content, wide
conceptual vs. non‐conceptual 257–9, 296, 438, 448, 449, 457–72, 478–88, 650, 665, 666
definiteness of, see content, determinacy of
determinacy of 515, 526, 530, 531, 632, 635, 640, 643
disjunctivism about 506
externalism vs. internalism about 351–6, 476, 477, 504–6, 672–4, 767–78, see also content,
narrow; content, wide
fixing of 141
grain of, see grain of representational content
grasping of 460
inferential role 100
(p. 801)
Page 8 of 37
Index
Page 9 of 37
Index
Dawson, M. 572
Debellis, M. 458
decision, see also choice; free will
making of 678, 687, 688
Democritus, 152, 208
Demos, R. 782, 783
demonstratives
in concepts 444, 445
in experiences 231
in judgments 352
denial, see self‐deception and denial
Dennett, D. C. 42 n., 88, 131, 133, 140–5, 148, 174 n. 2, 207 n., 295, 305 n., 339, 344–8, 377 n. 10,
391, 408, 409, 475, 483, 515 n., 527–35, 560 n. 49, 562, 568, 574, 666, 714
dependence 111, 177, 178, 181, 184–6
and emergence 212, 213
and supervenience, see supervenience and dependence
causal 86, 89–93
constitutional account of 123
counterfactual 90 n., 111, 775, see also causation, counterfactual accounts of
of the mental on the physical, 40, 105, 106, 110, 123
DeRose, K. 775
Descartes, R. 29–37, 55, 56, 68, 69, 72, 85, 177, 206, 208, 240, 283, 293 n. 4, 316, 333 n., 525,
528, 529, 541, 550–6, 569, 607–16, 619, 620, 625, 626, 678, 680, 728–31, 735 n., 736 n. 15, 739,
742, 743, 753
designator
non‐rigid 321
rigid 162, 166, 170
design stance 340–4
desire 67, 68, 85, 95, 96, 100, 117, 118, 129–33, 227, 242, 243, 339–49, 375, 376, 382, 386, 401–5,
428, 469, 475–7, 521, 523, 572, 688, 692–4, 698–702, 713–22, 741
determinism 39, 217
and quantum mechanics 54, 59
deviance, causal 703–5
Deutsch, H. 579 n. 1
Deutscher, M. 665
Dewey, J. 499 n. 10
Díez, J. 410 n.
Diogenes 208
directly reductive narrow functionalism (DRNF) 375, 376
discrimination, perceptual, see perception and discrimination
disjunction problem 387–9
disjunctivism, see content, disjunctivism about
dispositionalism
normative vs. non‐normative 429–35
pure 195
(p. 802) dispositions 194–7, 203, 215, 218, 293 n. 5, 322, 347, 622, 646
Page 10 of 37
Index
Page 11 of 37
Index
Page 12 of 37
Index
levels of 110–20
physical 113
psychological 335–63, 374–6, 441
reductive 155–60, 300
and causal role 154, 155, 160, 169
in science 154, 159, 166–70
vs. identity 152–5, 157–60, 206 n. 1, 210
semantic 391, 392
explanatory gap 77, 82, 88, 169, 231, 270, 281–90, 300–5, 313, 529, 534, see also mind‐body
problem
expressivism about norms 423
extension 353, 354
determination of 429–31
externalism, see also content, externalism vs. internalism about
and self‐knowledge, see self‐knowledge and externalism
facts
abstract 201
natural 64
epistemology of non‐physical 64
Fain, H. 32 n.
Fair, D. 44
Falvey, K. 771
Fara, M. 430
Farkas, K. 477 n. 6
Feigl, H. 225, 237
Festinger, L. 723
Fichte, J. G. 546, 555
Field, H. 364, 408, 446
Fine, K. 330, 422 n. 4
Fingarette, H. 792
first‐person access 288, 298
first‐person perspective 289, 294, 295, see also subjectivity
fission, see identity, personal
Flanagan, O. 514 n. 5, 516, 518, 550
Fodor, J. 43 n., 49, 99, 111, 112, 143, 185, 288, 296 n. 14, 297, 357, 360 n. 16, 361 n. 19, 369–73,
377 n. 10, 381, 387, 388, 392, 408, 409, 414, 416, 431, 458, 467 n. 6, 611–16, 625–8, 631, 637,
640, 732–9
folk psychology 96, 99, 132, 133, 142, 339, 342, 349, 361, 362, 373, 713–25
and categories of mental states 717–21
and simulation 715–8
and theory 713, 714, 720
Foster, J. 32 n., 68, 198 n., 505 n., 506, 507
four‐dimensionalism, see identity, personal
Frankfurt, H. 693, 707
Frankish, K. 332
free will 85, 140, 208, 214, 223, 224
Frege, G. 153 n. 1, 163, 257, 259, 439–41, 444, 448, 452, 506, 614 n., 627, 654, 655, 658
Frege dictionary 163 n. 16
Page 13 of 37
Index
Page 14 of 37
Index
Page 15 of 37
Index
Page 16 of 37
Index
ideas 223–31
identity 163, see also individuation
conditions for 73–6, 158
of causes 61–3
(p. 805)
Page 17 of 37
Index
indexicals 378
indicator semantics, see semantics, information‐theoretic
individualism, see also content, narrow
anti‐ 227, 749, 771
individuation
of concepts, see concepts, individuation of
of objects 34
of properties 47
of propositional attitude types, see attitudes, propositional
sortal‐relative 578–86
inexistence, intentional 521, 522, see also intentionality and non‐existent objects
inference 397–405, see also reasoning
and memory, see memory and inference
information 382–4, 528, 533
and quantum mechanics 214
processing 285–9, 466, 467, 614–26, 651
‐theoretic semantics, see semantics, information‐theoretic
instrumentalism 346
intelligence 132, 140, 348, 395, 608–10, 617–20
intension 379
epistemic 376–8
primary vs. secondary 318, 320–26, 328, 331
intention 109 n. 1, 124, 691–708
‐dependent properties 117–19, 124
of a speaker 353, 354
intentionalism 474–91, see also representationalism
pure vs. impure 479–90
intentionality 30, 48, 67, 139–42, 144, 149, 190, 199–202, 207 n. 2, 214, 394–406, 414, 417, 418,
512–35, 609–13, 621–5
and content 244–50, 477, 478, 497, 498
and mode 475–8
and non‐existent objects 476
and normativity, see norms
and phenomenology, 512–35, see also separatism
as self‐transcendence 476
as the mark of the mental 474, 490
intrinsic 632, 642, see also intentionality, original vs. derived
objects of 474–90
of conscious states 482, 486–90
of moods 487–90
of perceptual and emotional states 486–90
of sensations 487–90
original vs. derived 343–6, 621, 625, see also meaning, original vs. derived
phenomenal 512–35
realism about 531
intentional stance 339–49
in evolutionary biology 347–9
objections to the 346–8
Page 18 of 37
Index
Page 19 of 37
Index
Page 20 of 37
Index
Page 21 of 37
Index
Page 22 of 37
Index
original vs. derived 382, see also intentionality, original vs. derived
and synonymy 153, 154
and use 424–32
measurement‐theory
numerical 410, 411
and propositional attitudes 407–19
measure predicate 408–15
mechanisms
cognitive 537–63
computational 537–63
semantic interpretation of 538–61
Meehl, P. E. 212 n.
Meixner, U. 69, 79
Melden, A. 706
Mele, A. R. 113, 625 n. 17, 691, 692–708, 782, 787–90
Mellor, D. H. 46, 57, 67, 184
Melnyk, A. 175 n. 4, 178 n. 10, 179–81, 184 n. 23, 284
Meltzoff, M. 471 n. 9, 724
memory 48, 141, 142, 254, 435, 572, 596, 598, 605, 663–75, 723, 750
and agency, see agency and memory
and inference 672–4
and realism about the past 664, 665
and representation 665–7
and the self 555, 562, 582, 627, 667–71
episodic 663, 665, 667–74
short term 557, n. 40
Mentalese 297, 346, 616, see also language of thought
Menzies, P. 46
Merricks, T. 69, 584, 588 n.
Mill, J. S. 257, 626, 701 n. 20, 728, 731 n. 7
Miller, A. 467 n. 7
Millikan, R. G. 133, 134, 373, 382, 394–6, 401–6, 625 n. 16, 673
Mills, E. 46
mind
as a fundamental natural kind 177, 180–6
disembodiment of 69, 74, 333 n. 2, 679
immaterial 38
mind‐body problem 29, 41, 109, 130, 138, 139, 142, 160, 180–6, 207, 293 n. 5, 303 n. 33
and reductive explanation 160
mind‐reading 305, see also folk psychology
misrepresentation 249, 387, 388, 487
Mitchell, K. J. 672
mode
of existing 549, 554
intentional 488–91, 498, 503
vs. content 477, 488, 489
Page 23 of 37
Index
Page 24 of 37
Index
Nuccitelli, S. 673
Nussbaum, M. C. 679
objects 31–5, 43
abstract 201, 439, 440, 443, 444
and propertiedness 546, 548 n. 17, 549, 551
as bearers of properties 66–8, 72
individuation of, see individuation of objects
intentional 476–8, 488–90, see also intentionality, objects of
mind‐independent 497, 503, 504
problem of coincident 34
of intentionality, see intentionality, objects of
tracking properties of 470
O’Connor, T. 41, 701 n. 20
O’Dea, J. 475 n. 3, 480 n. 8
Öhman, A. 679
Olson, E. T. 72, 543, 582 n. 10, 584 n. 16 & 17, 586 n. 22, 588, 589
ordinary‐language philosophy 201
O’Shaughnessy, B. 597 n. 4, 699
other minds 89, 714, 727–39
argument from 88, 89
existence of vs. contents of 727–9
scepticism about 733–5
overdetermination 46, 79, 125, 174
acceptable vs. unacceptable 62, 113
and causal exclusion 39
argument 114
problem of 38
strong vs. weak 63, 64
Owens, J. 771
pain 270, 276–8, 286–9, 480, 487–9, see also experience; experiences; qualia
and behaviour 728–39
pairing problem, see cause‐effect pairing
panpsychism 177, 180, 206–18
history of, 207–11
arguments for 211–16
arguments against 216–18
Papineau, D. 45, 46, 53, 57, 77, 157, 158, 162, 163, 184 n. 23, 185, 294 n. 8 &10, 296 n. 13, 298
n., 305–9, 333, 382
Parfit, D. 566, 569 n. 4, 570, 587 n. 25
Pargetter, R. 728
Parsons, T. 476
particulars
abstract 408, 409, 412
Page 25 of 37
Index
Page 26 of 37
Index
Page 27 of 37
Index
Page 28 of 37
Index
Page 29 of 37
Index
defeasible 433
practical 339, 399, 401, 404, 433, 562, 692, 722, 784
reasons
acting for 694, 695, 699–702, 707
recognition 224, 232–5
reducibility, see reduction
reduction 39, 42–4, 86, 93, 206, 210–12, 217, see also supervenience
and identity 506
of folk psychology to physics 96
reductionism, see monism; naturalism; physicalism
reference 352–6, 368, 369, 377, 446–8, 685, 686
and causal contact 648, 649
and consciousness, see consciousness and reference
by causation 685
by description, 685, 686
fixing of 356, 650–3
knowledge of, see knowledge of reference
of phenomenal concepts, see concepts, phenomenal
reflex action 609, 610
Reid, T. 273 n., 276, 501 n. 14, 504, 569, 581 n. 6, 731, 732, 736, 737
relations 190, 200–4
reportability of mental states 243, 247, see also introspection
representation 48, 381–92, 394–406, 457–71, 475–90, 494–509, 609–26
analogue 296, 305, 308, 644
and content, see content
descriptive vs. iconic 396–406
explicit 345
false, see misrepresentation
higher‐order, see HOT (higher‐order thought)
in language vs. pictures, see imagism vs. sententialism
pictorial, 632, 636–8, see also imagery, pictorial
pushmi‐pullyu 396–8, 401, 402, 405, 406
vehicles of 294 n. 8, 297, 306, 637–43
representationalism 502–4, 524, 525, see also intentionalism
and phenomenal consciousness 144–9, 253–67, 287, 288
and the representational theory of mind (RTM) 296, 297, 416
arguments for 260–5
externalist vs. internalist 257
Fregean 257–9
reductive vs. non‐reductive 257–9
strong vs. weak 256, 257
revelation 659–61
Revonsuo, A. 567
Rey, G. 294 n. 6, 785, 786
Richard, M. 408, 413
Ridge, M. 692
Rips, L. 639
Rives, B. 43
Page 30 of 37
Index
Robins, R. W. 683
(p. 812) Robinson, J. 682
Robinson, H. 189, 191 n. 1, 192 n., 194, 198 n., 302, 505 n.
Robinson, W. S. 78, 87
robots 132, 146, 160, 339, 343, 622
intelligent, 285, 286
non‐conscious 495
role
conceptual 141, 623–5
functional 131, 132, 137, 145–7
and content 375, 376
of perception vs. beliefs 460
Rollins, M. 642, 646 n.
Roseman, I. 680
Rosen, G. 422 n. 5, 423 n. 6 & 7
Rosenberg, G. 215
Rosenthal, D. 144, 147, 239, 242–6, 249 n. 10, 250, 273 n., 287, 293 n. 5, 482 n. 12, 502 n. 17,
568, 755, 756
Ross, D. 344
Ross, G. H. T. 29 n., 177
Ross, L. 723, 789
Rovane, C. 588, 589
Ruddick, W. 792
Rumelhart, D. E. 628, 666
Rumfitt, I. 445
Russell, B. 176, 183 n., 197, 215, 275, 290, 292 n. 3, 322, 643, 648–51, 656, 659–61, 664, 753
Ryle, G. 129, 194, 702, 728, 742, 745, 746
Sabates, M. 46
Sainsbury, M. 650
Salmon, N. 369, 478
Sanfey, A. G. 683
Sanford, D.H. 791
Sartre, J. P. 513, 598 n. 6, 782, 792
Sawyer, S. 772, 774
Scanlon, T. 694 n. 10, 699
Schacter, D. 663–8, 672
Schaffer, J. 125
Scherer, K. R. 680
Schiffer, S. 327, 364, 440, 446, 451, 452, 478, 497 n. 8
schizophrenia 571–3
Schnall, S. 682
Schütte, M. 167 n. 22
sciences, special, see laws, of special sciences
Scott‐Kakures, D. 787, 789, 790
Scruton, R. 73, 600 n. 13
Seager, W. 206, 208, 217, 520, 524
Searle, J. R. 48, 148, 226 n. 2, 243, 264 n. 12, 421, 467 n. 6, 477 n. 5, 480, 487, 488 n., 499, 514,
524, 526, 530, 575, 620–2, 625, 692, 693, 701 n. 20, 703, 705
Page 31 of 37
Index
Sedivy, S. 460 n.
Segal, G. 49, 125 n. 25, 166 n., 353–8, 367, 372, 373 n., 374–6, 379, 413, 415 n., 504 n.23
Sehon, S. 702
selection, natural, see evolution
self 67–76, 81, 82, 223–8, 236, 541–63
and memory, see memory and the self
and personality 545, 547, 562
as subject of experience, see experience, subject of
as transient 550, 551
awareness of 144, 243
Buddhist conception of 555, 560
ecological 545
identity and persistence 67–82, 560–2, see also identity, personal
illusion of 223
narrative 545, 560 n. 49, 562
reality and existence of 541, 545–9, 554, 559–60
social 545
transparency of 741, 744
self‐deception
and denial 781–3
and dispositional belief, see belief, dispositional vs. occurrent
and emotions, see emotions and self‐deception
and rationalization, see belief and rationalization
deflationary approaches to 784–7
twisted 787, 788
self‐knowledge 729, 730, 739, see also introspection
and externalism 767–78
and transparency, see self, transparency of
as incorrigible 741, 744
as privileged 741, 744, 745
Sellars, W. 85, 212 n., 246, 713
Selten, R. 789
semantics 48, 611, 625 n. 17, 626, see also biosemantics; teleosemantics
conceptual role 381
consumer 382
formal 407, 413, 422
information‐theoretic 381–92
naturalized 389
possible world 462
procedural 381
two‐dimensional 313–33
objections to 324–7
sense
data 263, 268, 274, 275
Fregean, see Fregean sense
impressions 269, 272
sententialism, see imagism vs. sententialism
separatism and inseparatism 514–36
Page 32 of 37
Index
Shoemaker, S. 92 n., 119, 120, 145–7, 246, 256 n., 259 n. 5, 322 n., 329, 330, 566, 574, 582 n. 11,
583 n. 12 & 13, 584 n. 16, 586 n. 21 & 24, 672, 741 n. 2, 754, 756
Sidelle, A. 330
Sider, T. 543 n. 7, 584, 589
Siegel, S. 499
Siegler, F. A. 783
Siewert, C. 475 n. 3, 482, 486 n., 496 n., 514, 522
sign, natural 275–9, 401–4
Silins, N. 773
simultagnosia 572, 573
skepticism
about other minds, see other minds, skepticism about
about phenomenal content 525
Skrbina, D. 208, 211
Slater, A. 471 n. 9
Smart, J. J. C. 129, 154, 174, 175, 180, 211, 274
Smith, B. 570
Smith, C. 680
Smith, D. 496 n., 502 n.19
Smith, M. 692 n. 2
Smith, Q. 194 n.
Smolensky, P. 628
Snowdon, P. 177, 495 n. 2
Soames, S. 324, 369, 478
Sober, E. 49, 125 n. 25
solipsism 625, 728, 730, 731, 735–7
Solomon, R. 679
Sosa, E. 37 n. 12, 587 n. 26, 745
soul 40, 72, 110, 152, 174, 283, 347, 544–8, 553, 559, 608, 731, see also dualism, substance
and spatio‐temporal location 31–4
Sousa, R. de, 680
Spelke, E. S. 471 n. 9, 561 n. 51, 616
Sperry, R. W. 36, 41
Spinoza, B. 29, 208–10, 678–80
Sprigge, T. 200, 204 n. 7
Squire, R. 663
Stalnaker, R. 156 n., 158, 159, 162, 163, 165, 324, 328, 408, 462, 666
Stampe, D. 381
state
abstract, higher‐order 285–8
cognitive 67, 227–9, 235
computational 458
information‐processing 466
inherently conscious 253, 254
of affairs 199–203
psychological 352, 362–4, 367–75
Page 33 of 37
Index
Page 34 of 37
Index
Page 35 of 37
Index
Tye, M. 97, 144, 149, 253, 256 n., 259, 261 n. 8, 275 n., 287, 288, 293 n. 5, 296, 299 n., 302 n. 31,
304, 458, 461, 475 n. 3, 479–83, 487, 488, 504, 514 n. 2, 524, 530, 566–8, 574 n., 575, 642, 769,
776
Unger, P. 176 n. 7, 183 n., 582 n. 11, 584 n. 15, 585 n. 19, 586 n. 20
understanding
functional 131, 134, 137, 139, 140
of concepts 522, 770, 777
of words 516, 620–4
unity
argument for dualism 74, 76
of consciousness, see consciousness, unity of
of the mind 512–35, see also separatism
of the world, 144, 168, 169
uniqueness theorem 411, 412, 415, 417
universals 201
Urmson, J. 732
use vs. mention, 623, 624
Vahind, H. 770, 772
Valins, S. 682
Vartanian, O. 683
Van Cleve, J. 653
Van Fraassen, B. 182
Van Gulick, R. 43, 111, 116 n. 7, 128, 133, 143, 144, 148, 249 n. 11
Van Inwagen, P. 76, 551 n. 21, 584 n. 17
Velleman, J. D. 264 n. 12, 671, 692 n. 3, 702, 706, 707
verificationism 153, 298, 640, 664
Von Eckart, B. 357
Vonnegut, K. 640 n. 8
Wallace, R.J. 702
Walter, S. 85, 91, 93 n. 7, 290
Warfield, T. 114, 685, 772, 774
water and H2O 129, 154, 158, 159, 162–7, 173, 286–8, 300, 316–20, 325, 326, 352, 353, 367–9,
373, 377, 389, 673, 749, 761, 776, 777
vs. twater and XYZ 368, 371, 372, 375, 768–72, 775
Watson, R. A. 30
Webb, S. 348
Wedgwood, R. 421, 422 n. 2
Weizenbaum, J. 618, 619
Wellman, H. 719
Wheeler, M. 669
Whisner, W. N. 785
White, P. A. 243
White, S. 300, 301 n. 27, 307, 377 n. 10, 570, 760
Whitehead, A. 210, 214, 217, 546
(p. 815) Whiten, A. 344, 721
Wimmer, H. 719
Wiggins, D. 579 n. 1, 584 n. 17, 585 n. 19, 586
Wigner, E. 214
Page 36 of 37
Index
Page 37 of 37