Académique Documents
Professionnel Documents
Culture Documents
UNIVERSITY PRESS
Great Clarendon Street, Oxford OX2 6DP
Oxford University Press is a department of the University of Oxford.
It furthers the University's objective of excellence in research, scholarship,
and education by publishing worldwide in
Oxford New York
Auckland Bangkok Buenos Aires Cape Town Chennai
Dar es Salaam Delhi Hong Kong Istanbul Karachi Kolkata
Kuala Lumpur Madrid Melbourne Mexico City Mumbai Nairobi
Sao Paulo Shanghai Taipei Tokyo Toronto
Oxford is a registered trade mark of Oxford University Press
in the UK and in certain other countries
Published in the United States
by Oxford University Press Inc., New York
© the several contributors 2003
The moral rights of the author have been asserted
Database right Oxford University Press (maker)
First published 2003
All rights reserved. No part of this publication may be reproduced,
stored in a retrieval system, or transmitted, in any form or by any means,
without the prior permission in writing of Oxford University Press,
or as expressly permitted by law, or under terms agreed with the appropriate
reprographics rights organization. Enquiries concerning reproduction
outside the scope of the above should be sent to the Rights Department,
Oxford University Press, at the address above
You must not circulate this book in any other binding or cover
and you must impose this same condition on any acquirer
British Library Cataloguing in Publication Data
Data available
Library of Congress Cataloging in Publication Data
Data available
ISBN 0-19-924561-4 (hbk.)
ISBN 0-19-924562-2 (pbk.)
1 3 5 7 9 10 8 6 4 2
Typeset by Newgen Imaging Systems (P) Ltd., Chennai, India
Printed in Great Britan
on acid-free paper by
BiddIes Ltd, Guildford and King's Lynn
1
Agency and Self-Awareness:
Mechanisms and Epistell101ogy
CONTENTS
1. Introduction
2. Mechanisms of control
3. Epistemological issues
4. Action and perceptual consciousness: dissociations and connections
4.1 Awareness of bodily movements in intentional action
4.2 The role of perceptual consciousness in the control of intentional
action
5. The sense of ownership
1. INTRODUCTION
In the mid-1980s Libet conducted a series of experiments which, he argued,
showed that we are aware of intentions to act only after the brain area responsible
for initiating acting has already been activated. He, and others, took these experi-
ments to show that our brain (decides' on courses of action independently of
our consciousness, where this in turn threatens the idea of free will, the idea that
our actions are controlled by our conscious intentions. For immediate purposes
the main point to note here is the intuitive connection highlighted by the debate
surrounding Libet's claims, between the kind of control we think we have of our
actions, on the one hand, and our awareness of our intentions and actions, on the
other. For the suggestion underpinning Libet's claims is that unless the initiation
of the action is something we are aware of, the action itself is not under the
kind of control we think we have as agents, the kind of control in virtue of which
we speak of freedom of the will.
Though there is undoubtedly a powerful intuition about the existence of some
kind of link between agency and self-awareness, it turns out to be extremely diffi-
cult to articulate and defend. A pathological case that brings out both the power
of the intuition and the complexities of the problems involved in articulating and
explaining it is the neurological syndrome labelled 'Anarchic Hand'. Patients with
2 Naomi Eilan and Johannes Roessler
Anarchic Hand syndrome sometimes find one of their hands perforn1ing
complex, apparently goal-directed movements they are unable to suppress (except
by using their 'good' hand). Sometimes the anarchic hand interferes unhelpfully
with intentional actions performed with the use of the other hand (it may unbut-
ton a shirt the patient is trying to button up). Sometimes it performs movements
apparently unrelated to any of the agent's intentions, such as (in one notorious
example) a movement resulting in picking up some leftovers from somebody
else's plate in a restaurant. In most of these cases patients go on to claim that it
feels as if the actions performed by the anarchic hand are not theirs, and that the
hand is doing something they did not intend or want, and cannot control.
Perhaps the most immediate question that might come to mind about the anarchic
hand's movements is: are they actions? In the light of patients' reaction to the n10ve-
ments, the first impulse may be to deny this. On the face of it, the patients are not
in control of, or responsible for, the movements at all. Yet there does seem to be a
sense in which the activities of the anarchic hand are skilfully controlled: they are
not pure reflexes, but clearly devoted to a particular goal, and, relative to the goal,
well-executed. Then the question is: what is the nature of the control that is absent
in these cases, and whose absence fuels the intuition that they are not to be regarded
as actions? A similar question arises when we look at patients' awareness of the
movements. Just as there is evidently some kind of control in the anarchic hand
case, there is also some kind of awareness on the subject's part of what is going on.
It is not as if the subject does things unknowingly. Rather, there is a sense in which
she has knowledge of what is going on of the same kind one might have as an
observer of someone else's actions. The question is: how should we characterize
normal 'inside' knowledge relative to which characterization we can describe the
anarchic hand case as deviant? It is here that we should hope to find some help in
characterizing the sense of agency, and ownership, that accompanies most actions.
So there is a question about the causal explanation of actions-the mechanisms
of control-in the normal case. And there is a question about the source of our
knowledge of actions (again, in the normal case). And, of course, there is the ques-
tion of how these two sets of issues are related to each other. In his tutorial paper
'Control of Mental Processes', Stephen Mansell says there are two distinct Il1yster-
ies that come up when we attempt to explain the nature of control. One is the ques-
tion of how, in information-processing terms, we should explain the kind of
control that seems to break down in cases like Anarchic Hand syndrome. The other
is how we should explain consciousness. His strategy is to divide and conquer, leav-
ing with 'some relief the phenomenal commitments of control to others', while
focusing instead on the first question only, namely, what observations of behaviour
can tell us about how we control our minds and behaviour (Monsell, 1996: 108).
Methodologically, as the title of the volume suggests, we have taken the opposite
tack. The explicit idea here is that we may be able to make progress with each of the
'mysteries' only by examining the interconnections between them. However, to take
-------------------------------------
Mechanisms and Epistemology 3
this methodological line is not the same as delivering a positive substantive answer
to the question: is the correct account of control one on which there is a constitu-
tive connection between the kind of control we exert over our everyday normal
actions, and self-awareness? Implicit in many of the papers collected here we find
support for both negative and positive answers to this question.
In taking this methodological line we are also simultaneously inviting readers
to draw on both psychological and philosophical resources in addressing these
questions. The central challenge here, as we hope to bring out in these introduct-
0ry comments, is to get right the relation between common-sense psychology
and various branches of professional psychology (e.g. neuropsychology, cognitive
psychology, and developmental psychology). In what follows we will not attempt
a comprehensive summary of the concerns of each contribution-the papers
speak eloquently for themselves, each one raising fascinating issues that we will
not be touching Oll. Rather, our aims are twofold. First, in Sections 2 and 3 we
sketch a map of some of the background issues that inform the psychological
and philosophical discussions of (a) control mechanisms and ( b) epistemology,
where the main aim will be to facilitate cross-disciplinary reading of the papers.
Second, in Sections 4 and 5, we consider how the epistemological and mechanism
questions interact in specific cases, drawing on the papers in this volume.
2. MECHANISMS OF CONTROL
Failures of Control
What if anything can we learn about normal intentional action control by con-
sidering abnormal actions such as that exhibited in Anarchic Hand syndrome?
There are in fact two kinds of questions they raise. One kind is: what does our baffle-
ment about such cases show about our everyday understanding of the way actions
are controlled, relative to which anarchic hands count as deviant? What are the
everyday expectations that are thwarted in such cases? The other question is: what
do such cases tell us about the mechanisms involved in control of everyday
action? How can we learn from such failures about the proper functioning of
action control?
In this section we want to layout some of the basic ideas that inform the
psychological and philosophical papers n10st directly concerned with control. We
do so by putting the initial bafflement we feel about the deviant cases in terms
of a crude paradox, to which these basic ideas can be read as a response (though
they are not explicitly presented as such). To give a flavour of the kinds of cases
at issue here, it will be helpful to have before us a brief list of the kinds of failure
of control that are considered in the psychological literature concerned with
supplying an information-processing account of normal action control. There are
basically three kinds.
4 Naomi Eilan and Johannes Roessler
1. Pathological cases. Earlier we mentioned Anarchic Hand syndrome. This is
one of many pathological cases, which psychologists use to try and gain insights
into the mechanisms and anatomical basis for the normal kind of controL Other
pathological examples of failure of control, all of which tend to be classed as
'frontal lobe' problems include:
(a) utilization behaviour, where subjects apparently lose the ability to overcome
the power of a stimulus to invoke a habitual action. So for example, such a
patient may be found making a cup of tea whenever a teabag comes into
view; or start dealing cards whenever they come into view, and so forth;
(b) perseveration, where subjects seem to be unable to stop performing a par-
ticular sequence of actions, for example, a patient will go on sorting cards
according to one dimension (colour), even though he has explicitly been
given the instruction to switch to another (e.g. shape);
(c) dissociations between verbal announcement of intentions and actions, for
example, a patient may announce an intention to do one thing while doing the
opposite and simultaneously describe what he is doing as the opposite of what
he intends. And nlany others (for a good summary of these see Monsell, 1996).
2. 'Slips of action'. The second kind of case psychologists use for formulating
theories about normal control of actions are labelled 'normal' failures of control,
that is, non-pathological, of the kind that happen to all of us, nlost of which we
tend to classify as cases of absent-mindedness. The locus classicus for this is
Reason's categorization of such slips of action, which takes as its point of depar-
ture James's description of going upstairs to change for dinner and finding him-
self getting into his nightclothes and climbing into bed instead. Reason labels this
a 'double capture error', where attention is captured by some internal preoccupa-
tion, allowing the action to be captured by a stimulus associated with a strong
habit. A double capture of omission rather than commission is failing to stop off
for the paper on the way horne, despite the intention· to do so. Another kind of
case he labels 'lost intentions', as in the familiar cases of finding oneself in a room
and not remembering why on earth one wanted to be there in the first place. Yet
another case is labelled one of 'detached intentions'-as when you cross a room
in order to shut the window but shut the cupboard door instead. (For a compre-
hensive survey and description of these and many other types of slips of actions
see Reason, 1984; and Monsell, 1996.)
3. Infant failures of controL The third kind of deviation from the norm
is found in the developnlentalliterature. Here, the concern is not to investigate
the adult norm on the basis of deviations, but rather to gain insights into
the developmental stages involved in achieving the human mature norm. An
example of a failure of control here is the failure of children until the age of about
5 to be able to play Simon says garnes, in which players must listen to commands
to perform simple actions but are required to carry out only those commands that
Mechanisms and Epistemology 5
are preceded by the phrase 'Simon says'. They evince every sign of understanding
the instruction verbally, but as soon as play begins seem unable to switch from one
kind of instruction to another. This feature has been replicated in a number of
experiments, discussed in Frye and Zelazo's chapter, where children seem to be
unable to switch from one task to another, in some kinds of case but not others,
while at the same time being able to repeat the instructions and manifesting some
kind of understanding of them.
Now a background question is, of course: how are these different kinds of deviations
from the norm related to each other? The question of the connection between
developmental findings, on the one hand, and slips of action and pathological cases,
on the other, is raised explicitly by Frye and Zelazo, and we return soon to their
paper, and Jennifer Hornsby's comment on it. Let us begin with the first two kinds
of deviation, pathologies and slips of action, which are normally considered
together in the psychological literature, and are thus treated by Perner and by
Humphreys and Riddoch. Here too there are, of course, fascinating and difficult
questions about the connections between different kinds of slips and different
pathologies. But for the purposes of introducing the issues we are interested in, let
us focus on the n'\'o kinds of cases that figure most prominently in these two papers.
Double Capture Errors. Suppose you are driving along a familiar route (say, on
your way home from work), intending to take an unfamiliar turn (say, to go the
supermarket). It is a common experience that such intentions are sometimes
ineffective: when you reach the critical junction there is a chance that you will
absent-mindedly drive on along the familiar route.
The (cups' experiment. Humphreys and Riddoch report on their experimental
work with a patient, ES, who had suffered brain damage to the supplementary
motor area in both hemispheres. ES sometimes showed the kinds of symptoms
associated with Anarchic Hand syndrome. But under experimental conditions,
she also showed a more specific impairn1ent. In one experiment, ES was presented
with a cup placed on the table in front of her, aligned with either the right or the
left side of her body. She was asked to pick up the cup, using the left hand when
the cup was placed on the left, and the right when it was placed on the right. ES
had no difficulty complying with the instruction to pick up the cup, but she often
used the wrong hand. Strikingly, a key factor was the direction in which the cup's
handle was facing. If the cup was placed on the left, with its handle pointing to the
right, ES would tend to pick it up, conveniently, with her right hand. She would
usually claim to be unaware of not having complied with the instruction.
The Paradox
Our first question was: what is it about our everyday conception of intentional
action relative to which both cases are surprising? An in1mediate response to the
6 Naomi Eilan and Johannes Roessler
very question might be that we do not find the slips of action surprising in any-
thing like the way we do the pathological cases. That may be so, but the explana-
tion for the lack of surprise might be simple amused familiarity in the slips of
action case. If there is a substantive difference between the two, it has to be argued
for, and we will return to such an argument later on. In the meantime, there is at
least prima-facie value, when our interest is in our everyday expectations, in being
forced to consider these two kinds of cases together. And when we do, the initial
bafflement about all these cases can be stated as follows. In many of them
an intention appears to be thwarted, not, familiarly and unproblematically, by a
physical obstacle, or physical inability, nor, on the face of it, by any failure of
reasoning capacities, or by other agents. Rather n1any of these cases seem to be ones
in which the agent herself thwarts her own intentions by acting with a conflicting
intention. And the air of paradox here turns on the potential threat to the unity of
agency implicit in such a description.
To spell this prima-facie paradox out in more detail, it will help to have
before us the bare bones of current philosophical understanding of the nature of
intentional action. Very roughly, on this conception, it is a constitutive feature of
intentional actions (as opposed to a reflex, say) that they are explained in terms
of the agent's intentions. Such explanations appeal to a good reason, from the
agent's perspective, for performing the action, where the agent's perspective will
normally involve desires for particular outcomes and beliefs about how to achieve
them. Though on some accounts this precludes the explanation from being
causal, most philosophers today would accept Davidson's point that the only way
an intention can rationally explain an action is if it causes it. So in citing the
agent's reason for performing an action we are citing its cause. Note that actions
tend to be construed here as particulars with nun1erous properties or descrip-
tions. For example, 'reaching for a drink' can be a description of the same action
as 'spilling the drink'. The notable difference, of course, is that when it comes to
explaining the action in terms of your intention, only one of the descriptions will
be suitable and relevant. A common way of putting the point is that actions are
intentional only under descriptions; an action may be intentional under one
description, yet unintended under another.
Consider now our example of a slip of action. If the action of turning left at the
light is an intentional action, then it must be done for a reason. Such a reason con-
flicts with the reason I have for not turning left, namely, my intention to stop off at
the supermarket. Such conflicts often occur, of course, and we resolve them by decid-
ing on one course of action rather than another, which in this case we would describe
as discarding the previous intention in favour of the one that informs my turning
left. We change our mind about what to do. But on the face of it no such change of
mind has occurred, that is what makes us baffled, amused, and irritated. How can it
be that the very same agent who decides on one course of action performs an action
that conflicts with it without havin~_~~~~~cie3_~~e_2~eY!Q~~!I!t~!1tipn3 ------
Mechanisms and Epistemology 7
The bafflement here, and the problem, are, of course, very similar to those that
get discussed under the heading of 'weakness of the will' or 'akrasia'. How can it be
that I decide to stop smoking forever and a second later reach out for a cigarette
V\Tithout having discarded the intention to stop? We seem to have before us two
competing sources of agency within the same agent. But a feature that prima facie
distinguishes the cases we are considering from those that fall under the heading of
akrasia is that we do not, even on the face of it, have a case of desires competing
against reason. There appears to be no sense in which a reason-dodging
powerful desire is at play in driving me to turn left. We appear to have cases here
of unmotivated, cold irrationality.
The moves one might make in resolving our paradox can be represented as
falling into two camps according to how they develop a distinction drawn by James
between environmentally driven actions, which he called 'ideon10tor actions', and
what he called 'willed actions'. The former are defined as bodily movements caused
solely by the 'bare idea' of their perceivable effects. (James's examples include
spotting a pin on the floor and picking it up while simultaneously carrying on with
a conversation.) The latter are said to be preceded by 'an additional conscious
element, in the shape of a fiat, mandate, or express consent' (189Gb: 522). The first
camp says that while there is an important distinction between environmentally
driven actions and those preceded by deliberation, it is a distinction between types
of intentional action. Correlatively, our everyday notion of intentional action
can, if not given misconceived philosophical glosses, dissolve the paradox. The
second camp says that amongst ideomotor actions we must recognize a class of
non-intentional actions, actions that cannot be explained in terms of personal-
level intentions but only by an information-processing psychology. Correlatively, it
is this information-processing notion of action that we need for dissolving the
paradox. Both the chapters by Perner and by Humphrey and Riddoch fall into the
second camp, though not in explicit response to the paradox, and we begin with
them, before returning to the first kind of response.
1 This kind of view might also be motivated by philosophical concerns with the nature of rule-
following. Crudely, the thought would be that we must make room for the notion of habitual inten-
tional actions, where one finds oneself saddled with an intention, on pain of getting wrong various
clearly intentional aspects of our actions. Compare, for exan1ple, John McDowell's paraphrase of
Wittgenstein's account of rule-following: 'When one follows an ordinary sign-post, one is not acting
on an interpretation. This gives an overly cerebral cast to such routine behaviour. Ordinary cases of
following a sign-post involve simply acting in the way that comes naturally to one in such circumstances,
in consequence of some training that one underwent in one's upbringin~~(~~~o~~~ ~9~~ ~Ol.
Mechanisms and Epistemology 11
get executed by priming the activity of the lower-level mechanism. But how should
we conceive of the interaction between the (higher' and the (lower' level? On this issue
the chapters by Perner and Humphreys & Riddoch present somewhat different per-
spectives. Humphreys and Riddoch speak of an information-processing system that
is sensitive to environmental factors as well as intentions, without proposing an
analysis of the latter in information-processing terms. On a natural interpretation,
the (higher level' here simply is the level of common-sense psychology, or the
(personal level'. So the claim might be put by saying that when you form a prior
intention you thereby prime the appropriate stimulus-response associations
(though of course this is not something you are aware of doing). We might think of
this top-down influence on the lower-level system as a mechanism of (delegating
control to the environment' (Gollwitzer, 1996). It enables you to carry out your
intentions effortlessly, as it were: when the time comes, the appropriate stimuli will
select the right actions for you.
In Perner's account, in contrast, the distinction between two levels is explicitly
understood as a distinction between two information-processing systems. An
influential version of this idea is the distinction between a system of (contention
scheduling' and the (supervisory attentional system' (Norman and Shallice, 1986).
Briefly, these are two different kinds of mechanisms for resolving conflicts amongst
action schemas (action programmes activated by stimuli). Contention scheduling
consists in the mutual inhibition of competing schemas and the mutual activation
of supporting schemas, with the most strongly activated schema winning out.
Alternatively, a schema may be selected because its activation value has been
strengthened by the (supervisory attentional system', an information-processing
system devoted to the planning and execution of non-routine actions. Now, a
common cOlnplaint against the idea of a (supervisory attentional system' is that it
credits an information-processing system with the causal powers of a rational
agent. For example, Dennett ridicules the system as (an ominously wise overseer-
homunculus who handles the hard cases in the workshop of consciousness' (1998:
288). On one reading, the complaint here is that the theory lacks explanatory
force-it does not tell us how the operation of the supervisory attentional system
carries out the kinds of functions we associate with rational agency. Perner aims to
meet this objection by offering a substantive account of the (higher level'. He sug-
gests that the distinction between the two levels corresponds to a distinction
between two kinds of representations. Put very crudely, the lower level involves
representations that are strongly context-dependent and (connectedly, as Perner
argues) lack explicit semantic structure. At the higher level, in contrast, representa-
tions not only have explicit semantic structure; they are also (fact-explicit'-they
make explicit whether they aim to represent fact or fiction. In other words, they
involve a simple form of metarepresentation. Perner goes on to put this distinction
to work in presenting a detailed taxonomy of the phenomenology of action, as well
as in defending a version of a (higher-order thought' theory of consciousness.
12 Naomi Eilan and Johannes Roessler
Without considering these further developments here, and indeed without going
into the details of Perner's proposal, it is worth highlighting a methodological issue
raised by it. What should count as evidence for the existence of two distinct informa-
tion-processing systems underpinning 'automatic' and 'deliberate' behaviour?
Norman and Shallice (1986) put considerable weight on neurological
disorders which may be taken to show that the two systems can be selectively
impaired. For example, in their view, Utilization behaviour provides an
illustration of contention scheduling operating on its own, unmodulated by the
supervisory attentional system. By the same token, though, it may be argued that
the functions associated with the 'higher level' are subserved by distinct, dissoci-
able systems. Just this argument is made in Humphreys and Riddoch's chapter-
entitled 'Fractionating the Intentional Control of Behaviour'. Their point is
not just that the processes underpinning intentional control cannot be located in
a single anatomical area. Rather, what ES's performance in the 'cups' experiment,
as well as other data, suggest is that there is a functional distinction
between intentional control of the selection of targets (e.g. which cup to pick
up) and intentional control of responses (e.g. which hand to use). Humphreys
and Riddoch argue that there are two distinct information-processing systems
that underpin these functions. Damage to one of the systems can leave the other
intact.
Of course, one might insist that there is still an important sense in which these
systems are components of a single 'higher-level' system. As for evidence for the
existence of such a systen1, Perner would argue that there is compelling evidence
with which we are all familiar. As he puts it, 'dual control theory helps avoid the
conclusion that conscious will is an illusion'. We are all comn1itted to thinking of
intentions formed on the basis of deliberation as explanatory of many of our
actions. The two-level inforn1ation-processing theory, Perner argues, vindicates
that commitment. The dialectal situation here is far from clear, though. One
response to Perner might be that the empirical evidence, as discussed, for
example, by Humphreys and Riddoch, tells against the hypothesis of a single
higher-level system, and that, if the latter is required to make good our common-
sense picture, then so much the worse for common sense. Again, one. might
be sceptical on philosophical grounds about the possibility of vindicating the
common-sense conception of 'conscious will'. The worry might be that the price for
vindication is reductive explanation, and that the latter is not a feasible project,
given the prominent role of normative notions, and/or conscious experience, in
common-sense explanations of action. And of course there is the background
question of whether such vindication by a particular kind of information-
processing theory is required to avoid the conclusion that conscious will is an illu-
sion. (The issues arising from this last question are familiar from other areas, such
as debates as to what should count as evidence for the 'language of thought'
hypothesis. See, for example, Davies, 2000.)
Mechanisms and Epistemology 13
-------------~
Mechanisms and Epistemology 15
children's problem is one of adopting a new intention in situations where
the response to a given kind of stimulus that is correct relative to the new inten-
tion would have been wrong relative to the previous intention. A central claim of
the chapter is that the problem is a general, domain-unspecific limitation in 3-year-
aIds' practical reasoning abilities, namely, the inability to employ a higher-order
rule to determine which of two rules to apply. In this sense, if the knowledge that
should be guiding them is knowledge of the switching rule, they do not know
wl;1at to do.
Independently of whether or not the account works as a theory that supplements
in some sense our common-sense theory (for critical discussion, see Hornsby),
one important ingredient highlighted in their account is the intuitive connection
between intentional action and knowledge. Hornsby is raising the question of
whether our everyday notion of intentional action applies to the children's beha-
viour in these experiments, when certain kinds of connection with knowledge are
not met. More specifically, underpinning our puzzlement about the children is an
intuitive connection between (a) acting intentionally; (b) knowing what one ought
to be doing in thus acting (grasping the rule); and (c) knowing whether or not one
is doing what one should be doing. That is, we assume that if an action is inten-
tional then, in normal circumstances, and when all is open to view, the subject
should be able to tell whether or not the action she has performed complies with
what she ought to be doing, given her intention.
This connection with knowledge has, in fact, been implicit all along, from the
first formulation of the paradox generated by failures of control. There is puzzle-
raising con1petition between two reasons for action only if we assume that sub-
jects are aware of how they ought to be acting, given both intentions, and aware
of whether or not their actions conform to these intentions. (This intuitive con-
nection can be brought out by considering one way of resolving the paradox in
both cases, namely, driving the offending action underground. If we assume the
second action is unconscious, the paradox dissolves.)
If this is right then whatever story we tell about the mechanisms of action con-
trol, some account will have to be taken, in whatever way, of this intuitive con-
nection in our common-sense psychology. This is the issue to which we now turn.
3. EPISTEMOLOGICAL ISSUES
The relevant notion of control, for Anscombe, is that of rational control, where
this is described as a matter of practical reasoning 'leading to action' (p. 60). The
problem, then, is to understand how practical reasoning can constitute a source of
knowledge of agency. On the face of it, there is a kind of category mistake here.
Practical reasoning yields a grasp of practical reasons, an understanding of why it
is desirable to perform a particular action. But what we need when we are trying
to elucidate the source of agents' knowledge is an account of the epistemic, or
theoretical, reasons on which such knowledge is based-reasons for believing a
proposition to be true. Put crudely, what is not clear is why a decision to do
2 See Moran (2001) for an illuminating discussion of the relation between these kinds of questions.
Mechanisms and Epistemology 17
something should constitute any kind of knowledge. Anscombe's radical move
in response to this worry is to characterize agents' knowledge as 'practical' in the
following sense. We tend to think of propositional knowledge 'as something that
is judged as such by being in accordance with the facts. The facts, reality, are prior,
and dictate what is to be said, if it is knowledge' (p. 57). Anscombe calls this
'speculative' knowledge. In contrast, 'practical knowledge' aims for the facts to fit
with it. If a claim to practical knowledge turns out to be incorrect, this is because
"reality has refused to comply with the dictate of knowledge: 'the mistake is in the
performance, not in the judgement' (p. 82).3 In effect, Anscombe maintains that
intentions that have been correctly executed constitute knowledge of the intended
action. As intentions are based on practical reasons, so is agents' knowledge.
Many contributors to the present volume share Anscombe's view that there are
important connections between action control and knowledge of actions 'from
within', but it is comn10n ground between them that the notion of practical
knowledge provides little help in unpacking the connections. For one thing,
Anscombe's account does not hold in full generality. Intention plus success does
not in general equal knowledge of succeeding. You may intend to buy a nice
melon, and succeed in doing so (there is no 'mistake in the performance'),
while being, at best, hopeful of having succeeded. Of course, it is plausible that
there are descriptions of your action under which you do know what you are
doing, for example, 'I am buying "this" melon' (said while pointing at a particular
melon) or 'I am trying to buy a nice melon'. What is theoretically unsatisfactory
about Anscombe's account is that it does not tell us how it is that you know what
you are doing under these latter descriptions, but not under the first description.
A natural suggestion at this point is the following. What is missing from
Anscombe's account is any appeal to the experience of agency. To determine under
which descriptions agents have knowledge 'from within' of what they are doing,
we need to give an account of the content of that experience. As the papers in
this collection testify, there are numerous ways in which this general suggestion
can be developed; and numerous substantive issues, both epistemological and
phenomenological, which need to be resolved to make it good (some of which are
discussed in the next section). Here we confine ourselves to sketching some
fundamental theoretical choices to be made.
3 For an early manifesto of what Hintikka calls the tradition of 'maker's knowledge', compare the
following passage from Maimonides: 'There is a great difference between the knowledge which the pro-
ducer of a thing possesses concerning it, and the knowledge which other persons possess concerning
the same thing. Suppose a thing is produced in accordance with the knowledge of the producer, the
producer was then guided by his knowledge in the act of producing the thing. Other people, however,
who examine this work and acquire a knowledge of the whole of it, depend for that knowledge on the
work itself. For instance, an artisan makes a box in which weights move with the running of water, and
thus indicate how many hours have passed ... His knowledge is not the result of observing the move-
ments as they are actually going on; but, on the contrary, the movements are produced in accordance
with his knowledge' (Guide for the Perplexed, part iii, ch. xxi, quoted in Hintikka, 1974: 84).
18 Naomi Eilan and Johannes Roessler
One question is what to make of Anscombe's project of charting a middle
course between the two extremes. Suppose we adopt the general line that agents'
knowledge is made available, in a sense to be explained, by the experience
associated with agency. Then we have three options. We might take the view
that doing proper justice to experience requires adopting the first extreme (aban-
doning the traditional idea of a first-person/third-person asymmetry relative
to knowledge of bodily actions). Or again, we may favour the second extreme
(equating knowledge of actions 'from within' with introspective knowledge of
inner states or events). Finally, we might argue that getting right the role of
experience will enable us to find the middle course Anscon1be was contemplating.
The first two options are defended in the chapters by Joelle Proust and Brian
O'Shaughnessy, respectively.
Proust takes issue with the orthodox view that knowledge of one's own inten-
tional action is non-observational. At least one reason for the attraction of that
view, she argues, is a traditional, but misconceived, account of the third-person
case. On this account, observation merely furnishes knowledge of bodily move-
ments. To establish that a movement constitutes an intentional action we need to
rely on an inference, invoking a particular intention as the best explanation of the
observed movement. In opposition to this, Proust contends that intentions can be
directly manifest in the 'dynamic pattern' of observed goal-directed movements.
A second corrective to the orthodox view is her focus on unpremeditated, spon-
taneous actions-actions that have an intention-in-action but no 'prior intention',
in Searle's sense (Searle, 1983). Proust claims that, at least in these cases, we
acquire knowledge of the intentional content of our own actions by observing our
movements. This leads her to deny any difference in principle, relative to such
cases, between the bases of knowledge of one's own and knowledge of others'
actions, though with the following important qualification. Proust draws a dis-
tinction between the source of knowledge of the intentional content of an action
(what is being done?) and the source of knowledge of ownership of an action
(who is doing it?). Appeal to observation is intended only to address the first issue.
Knowledge of ownership, in her view, arises from a conscious 'sense of effort'. (We
will return to this in Section 5.)
The other (by Anscombe's lights) extreme position is occupied by Brian
O'Shaughnessy. O'Shaughnessy argues that (a) where there is physical action,
there is a mental event of trying, or willing, which is identical with the action, and
(b) events of the type willing are conscious experiences-part of the content of
the 'stream of consciousness'. In his view, we have immediate experiential know-
ledge only of tryings; put differently, we are introspectively aware of our actions
only under descriptions of the form 'I am trying to ... '. Knowledge of successful
bodily actions, even actions as humble as moving one's arm, is always based on
inference. The route by which O'Shaughnessy reaches these conclusions might
be called the argument from total failure (a close relative of the argument from
Mechanisms and Epistemology 19
illusion). Consider the case of total failure, where, due to sudden paralysis, an
attempt at bodily action does not even involve a bodily movement. In this situ-
ation' O'Shaughnessy reasons, the subject is still aware of trying to act (not just of
the intention of doing so). He concludes that a phenomenologically salient
mental event of trying also occurs in the case of attempts that are crowned with
a measure of success; here the trying constitutes the 'inner', introspectively access-
ible aspect of bodily action.
Suppose, finally, that we are sympathetic to Anscombe's idea of a middle
ground, while remaining firn1ly committed to the view that agents' knowledge is
grounded on the experience of acting. Then our project might be described as that
of transcending the divide between the inner and the outer: we are seeking an
account on which the perspective of conscious agency yields direct knowledge of
bodily actions. However, the basic issue between Proust and O'Shaughnessy (does
agents' knowledge arise from an 'outer', sensory awareness or from an 'inner',
introspectively accessible experience?) remains relevant. For there are two stra-
tegies we may adopt in pursuing the middle course. Roughly, one is to bring the
model of introspective knowledge to bear on bodily actions; the other is to argue
that agency involves a distinctive kind of perceptual experience, a kind of experi-
ence that differs from that of an outside observer and constitutes a first-person
perspective on bodily actions.
We find versions of the first strategy in the contributions by Christopher
Peacocke and Lucy O'Brien (Chapters 3 and 17). Peacocke defends the view that
awareness of trying is a sound basis for non-inferential knowledge of successful
bodily action. His argument turns on the causal role of trying. Not only is it the
case that trying, say, to move one's hand normally causes one's hand to move.
Rather, Peacocke suggests, it is partly constitutive of trying to move one's hand that
it is an event of a kind which normally generates the appropriate movement.
Peacocke concludes that an agent who is aware of trying to move her hand is enti-
tled to judge that she is in fact moving her hand. (One aspect of the disagreement
between Peacocke and O'Shaughnessy concerns the relation between trying and
moving: O'Shaughnessy rejects the idea that tryings cause movements; in his view,
successful tryings are identical with bodily actions, and the latter 'incorporate'
(rather than cause) bodily moven1ents.) Lucy O'Brien's acccount makes no appeal
to tryings at all. Part of her project is to n1ake a case for the view that bodily actions
are 'as primitive a psychological phenomenon as beliefs and perceptions'. In other
words, she rejects the assumption that actions can be analysed into a mental and a
physical element (of which only the former would qualify as a possible object of
introspective awareness). In her view, getting right the epistemology of action
requires getting right what it is for bodily actions (rather than mere tryings) to be
conscious. The central notion of her account is that of actions occupying attention.
Briefly, her suggestion is that conscious actions-those that occupy attention-are
actions performed on the basis of an assessment of possible options, hence with
20 Naomi Eilan and Johannes Roessler
a <sense of control'. Such actions are immediately accessible to the subject in
the sense that engaging in the action provides a reason for the belief that one
is doing so.
The second, perceptual strategy is explored in Chapters 15 (by Jerome Dokic)
and 18 (by Johannes Roessler). Dokic's central claim is that experiential knowledge
of agency arises from proprioceptive bodily experience. In acting intentionally, he
argues, we are proprioceptively presented with our movements as controlled by
ourselves. This is a kind of knowledge <from within', not just for the obvious reason
that only the agent herself can be proprioceptively aware of her n10vements, but
also in this sense: the perceptual experience of the movement involves a sense of
control, attendant upon being the agent controlling the movement. In Chapter 18,
Johannes Roessler pursues a similar suggestion, though he is not specifically con-
cerned with proprioceptive experience. Roessler draws on Brian O'Shaughnessy's
discussion of observation and agency in The Will (1980, ch. 8), where
O'Shaughnessy gave a distinctive phenomenological twist to Anscombe's slogan
that agents' knowledge is non-observational. O'Shaughnessy's claim was that the
project of observing one's own actions is incoherent: we cannot perceptually attend
to our actions from the perspective of an external observer (with a view to extract-
ing theoretical information about the world), given that we have to attend to it
<from within the act' (with a view to controlling the action). Roessler argues that
the distinction between the two modes of perceptual attention corresponds to a
distinction between two kinds of perceptual knowledge, each with a distinctive
sense of control: as observers we are aware of the mind-independent world around
us as causally responsible for our experience, as agents we experience intended
consequences of our actions as the upshot of our intentions.
As these sketchy remarks indicate, there is controversy amongst the philosoph-
ical contributions both about the kinds of constraints on an adequate account of
agents' knowledge that we should accept, and about the right way to meet them.
In particular, the debate revolves around three principal constraints. (a) An
account of agents' knowledge should do justice to the intuitive connection, in
the case of intentional action, between activity and knowledge, to the intuitive
contrast between the perspective of a passive observer of an action and that of the
participant actively engaged in it. ( b) It should explain in what sense such knowl-
edge is based on experience. (c) It should enable us to see how it is possible to have
non-inferential knowledge <from the inside' of actions involving bodily move-
ments, and thus how such knowledge bridges, or deconstructs, the divide between
the inner and the outer.
An important background question is whether (a) and (b) can be consistently
combined. It might be said that there is son1ething perplexing about this combina-
tion of views. The perplexity arises from the plausible assumption that in experi-
encing something the subject is precisely not an agent, but a passive receiver of
information. If this is right, how can agents' knowledge have its source both in
Mechanisms and Epistemology 21
agents' control over and in their experience of what they are doing? As Christine
Korsgaard puts it, (to experience something is (in part) to be passively receptive
to it, and therefore we cannot have experiences of activity as such' (Korsgaard,
1996: 204). The conception of experience underpinning this objection is
examined in detail by Tom Baldwin. Drawing on a little-known passage of
Wittgenstein's, Baldwin argues that a conception of perceptual experience accord-
ing to which perceptual content is subject to the will is incoherent-not just for
the more obvious reason that this would disable perceptual experience from
providing us with knowledge of a mind-independent world, but also because it
would undermine the possibility of a shared language for describing the content
of our experience. In this sense, passivity on the part of the sq.bject does seem to
be an essential feature of perceptual experience. On the other hand, Baldwin
claims that the subject's activity is no less essential. To acquire a (coherent and
ordered' representation of one's environment one needs to be able to attend
selectively (and actively) to particular aspects of it (see Chapter 8).
Of course one might retort that while passivity is indeed an essential feature of
our experience of the mind-independent world, there is no reason to suppose that
the same is true of our experience of our own (conspicuously mind-dependent)
activity. Indeed one might argue, as O'Shaughnessy does, that we have knowledge
of our own activity, not in virtue of having experiences ofthe activity, but in virtue
of the fact that the activity itself-in O'Shaughnessy's terms, trying-constitutes an
experience. An alternative line of response to Korsgaard's challenge would be to
question whether she is right to assume that activity and passivity are mutually
exclusive: perhaps agency, just like, on Baldwin's account, perceptual experience,
involves a mix of both. But of course there remains the substantive challenge of
explaining how the experience bound up with agency may constitute, as Korsgaard
puts it, an experience ofactivity as such, that is, a source of knowledge of activity.
Progress with resolving the tension between the three constraints depends in
part on how we conceive of the phenomenology of action. In the following two
sections we pursue two central aspects of this issue: the role of perceptual experi-
ence' and the nature of the experience of ownership.
Basic Actions
Suppose you are raising your arm to attract someone's attention. What is the
scope of your knowledge 'from the inside' of what you are doing? Arguably, it is
wider than the purely psychological antecedents of your movement, such as your
intention, or your trying, conceived as a mental event. On the other hand, it is
narrower than the remote consequences of your action; to know whether you suc-
ceeded in attracting the right person's attention you will need to rely on observa-
.tion and inference. An intuitively attractive thought is that one thing you will
always be immediately aware of is that you are raising your arm. Generalizing, we
might say: when we are aware 'from the inside' of our bodily actions, we are prim-
arilyaware of them as intentional bodily movements. A familiar philosophical gloss
on this, versions of which can be found in Lucy O'Brien's and Christopher
Peacocke's chapters, is that when we act intentionally we are aware of our 'basic
actions', that is, the things we do directly, rather than by doing something else.
(While you attract someone's attention by raising your arm, you can raise your
arm straightaway, without doing anything else in order to raise it.)
Appeal to some such principle is one way of articulating the surprise we feel
about the results of Marcel's and Jeannerod's experiments. Thus, focusing on
Marcel, we want to say that subjects moved their hand to the position of the light
by moving it to the left. This description of their action seems to be basic. Yet,
while they certainly intentionally moved their hand to the light (and were aware
of doing so), they were unaware of moving it to the left. So they were unaware of
their 'basic' action.
Another way of articulating our surprise is this. Owing to their illusory propri-
oceptive experience, subjects believed incorrectly that their hand was (say) to the
left of the light. Now, we think of intentional actions as actions informed by our
beliefs about means to achieve our ends. So the following looks like a piece of
sheer common sense: if you believe your hand is to the left of x, and you intend
to move it to the position of x, and execute that intention, then you will inten-
tionally move your arm to the right. Marcel's findings suggest, at the very least,
that the principle does not hold in full generality: subjects intentionally moved
their hand to the position of the light, but not in the direction that would have
accorded with their belief about how to reach it.
Now, conditions in Marcel's experiments are undoubtedly unusual in several
ways. So it is natural to think that while the data may show that some familiar
principles of common-sense psychology break down under special, unfamiliar
circumstances, they don't tell us anything significant about the normal case. In
fact, though, as Jeannerod points out, there is a growing body of evidence that
suggests that even under perfectly norn1al circumstances many aspects of inten-
tional actions are controlled on the basis of non-conscious, 'impl~glI£~~e~~d~
24 Naomi Eilan and Johannes Roessler
information. For example: suppose you grasp a cup that is placed in front of you.
There is evidence that the spatial parameters of your movement (e.g. how far, and
in which direction, you reach, how you pre-shape your hand as it approaches the
cup) are determined on the basis of non-conscious representations of the relevant
properties of both the cup and your motor system. Furthermore, the execution of
the movement is controlled by various comparator mechanisms that operate
non-consciously, checking for mismatches, for example, between the goal and the
perceived effect of a motor command and rapidly correcting for any deviations.
Jeannerod labels this level of operation of the information-processing system
'automatic', where the contrast is with conscious operations. Note that the 'auto-
matic level' in this sense is to be distinguished from the 'lower level' of control
invoked to explain habitual actions in theories such as Perner's or Humphreys and
Riddoch's, which is also often labelled 'automatic'. The utilization of non-
conscious information is constitutive of automaticity in Jeannerod's sense, but
not of automaticity as it applies to habitual actions. (As we shall see soon, this has
further implications for the role we think conscious experience has in directing
action.) So, what is automatically processed in this sense, cannot be the content of
awareness, cannot in particular be the content of an awareness of a basic action.
The very least this shows is that uncritical appeal to the notion of basic action is
of little help in articulating anything systematic about the relation between the
description under which an action is intentional, on the one hand, and the con-
tents of a subject's awareness, on the other. And that, recall, is the challenge faced
by anyone interested in articulating, let alone defending, the idea that there is a
kind of control essentially bound up with subjects being aware of their action.
her action in terms of bodily movements at all? And what is the right method for
answering that question-is it something we can tell simply on the basis of the
phenomenology of action?
In a suggestive passage, Tony Marcel argues that (a) the key to the explanation
of subjects' performance in the vibro-tactile illusion experin1ents may lie in the
focus of their perceptual attention (roughly, they were attending to the object
whose position they were trying to reach, rather than to their own body), and
(b) an explanation along these lines may build on Wolfgang Prinz's work on
common coding.
(a) might be taken to suggest that, quite generally, getting right the description
under which agents are aware of their intentional bodily movements requires tak-
ing account of the object of their perceptual attention. In Marcel's experiments,
subjects are aware that they are n10ving their hand to the position of,this' (per-
ceptually presented) light; they are not aware of moving 'this' (perceptually pres-
ented) hand-in acting intentionally the agent's body normally occupies the
background of attention. 4 Note that there are two possible ways in which this
point might be put to work in explaining why subjects were poor at determining
the direction of their movement. One would be that, as subjects focused their
4 An interesting exception is the case of deafferented subjects. As Marcel points out, IW, the deaf-
ferented patient described by Jonathan Cole, lacks proprioceptive experience beneath the neck, but this
does not disable him from performing intentional physical actions (see Cole and Paillard, 1995).
Perhaps the most striking theme in Cole's account of IW's rehabilitation is the constant need for him
to pay attention to any physical actions he performs: 'constant visual vigilance is required for any
purposeful movement' (p. 250). So it is not as if IW's experience is unusual merely in the sensory
modality he uses to attend to the relevant body parts. Rather, it seems that focal attention to oneself as
a physical object is precisely a symptom of the lack of ordinary bodily awareness.
Mechanisms and Epistemology 27
attention on the light (and perhaps had to do so in order to move their hand
there), it was difficult for them to observe proprioceptively the direction in which
their hand was travelling. A second, perhaps more interesting point would be that
subjects did not control the egocentric parameters of the movement by focusing
their attention on the hand and deciding to move it, say, to the left. That is, intu-
itively' the direction of the movement was not something to which the subjects
were attending in controlling their actions. As a consequence (it might be argued),
the direction was not part of the content of their non-observational knowledge of
what they were doing. (For more extensive discussions of the role of attention in
the control, and awareness, of action, see the chapters by O'Brien and Roessler.)
Marcel's emphasis on the focus of agents' attention is intuitively plausible, but
it reinforces the question of whether agents are aware of their bodily movements
at all-as the work by Prinz, to which Marcel refers, helps to bring out. Two core
claims of Prinz's theory of common coding are these: actions are controlled by
representations of their perceivable effects; and there is a level at which the repres-
entations of perceived stimuli and representations of intended effects use the
(same language', or (commensurate codes'. To see what this means, consider
the following simple exanlple of what Prinz calls action induction. Subjects are
asked to respond differentially to two different stimuli (a light flashed to the left
or to the right of a fixation mark), by pressing one of two keys (one on the left,
one on the right). There are two kinds of assignment: in the conlpatible assign-
ment, subjects are asked to press the key on the same side as the stimulus, in the
incompatible assignment they are asked to respond to a left-hand stimulus by
pressing the right-hand key (and vice versa). The result, replicated in a variety of
studies using numerous permutations of the basic paradigm, is that reaction time
is significantly shorter in the compatible condition than in the incompatible
condition. What this indicates, Prinz argues, is that the same system of spatial
representation is used in relation to stimulus and response: it is in virtue of this
(common coding' that, in the compatible condition, the representation of
the spatial location of the stimulus primes, or pre-activates, the representation of
the location of the response, thus improving performance.
Prinz does not explicitly consider the relation between common coding
and perceptual attention, but it is very natural to think that the level at which
spatial compatibility effects occur-at which perception and action use the same
language, as he puts it-corresponds to the focus of perceptual attention in action.
So, in line with Marcel's proposal, we should expect the level of description where
action induction occurs to coincide with the description under which agents are
immediately aware of what they are doing.
Prinz presents two sets of data that are directly relevant here. One set of
data suggests that spatial compatibility effects depend on the representations of the
distal object acted upon (a left-hand key) rather than on representations of
body anatomy (a left hand). Accordingly, common coding, Prinz argues, operates
28 Naomi Eilan and Johannes Roessler
at a level where actions are represented by reference to distal rather than proximal
effects. Perhaps more surprisingly, the second set of data indicates that in cases
where there are further intended perceivable effects (e.g. given the task to switch
on a light by pressing a key), compatibility effects depend on the location of the
further effect (e.g. a light on the left) rather than on the location of the object more
immediately acted upon (e.g. a key on the right). As Prinz puts it, action induc-
tion occurs at a level where actions are represented in relatively abstract terms, for
example, as <switching on the light', rather than <pressing the key'.
Should we allow that subjects in these experiments have direct knowledge <from
within' of switching on the light? Indeed, should we say that it is only, or at least
most immediately, under this description that the subjects are aware of what they
are doing? Or should we insist that the only thing they are aware of, without
observation or inference, is that they are pressing the key (this being their <basic
action', the action they carry out <directly', rather than by doing something else)?
We should distinguish two issues raised by these questions. One is whether the
content of knowledge of bodily actions <from within' always includes descriptions
of bodily movements (where, plausibly, these may involve reference to distal
objects, e.g. <I am pressing this key'). The other issue is whether these descriptions
enjoy a special epistemological status, such as the status of being the content of
agents' direct awareness of their actions.
A radical-perhaps too radical-reading of Anscombe's remark about the dis-
tance of the content of agents' knowledge from bodily movements, quoted earlier,
might suggest the possibility of being aware of acting without having the least
awareness of the bodily movements that constitute the opening links of the relevant
causal chain. Certainly that does not seem to be the normal case. At the other end
of the spectrum, we find the view with which we started this section. This is the view
that it is only basic actions we are immediately aware of <from the inside', and that a
more con1plex account has to be given of our knowledge of actions described in
terms of their further effects. Peacocke suggests an interesting parallel with the case
of perceptual knowledge. You may perceive an installation as a particle accelerator-
<immediately', in the phenomenological sense that no deliberation may precede
your judgen1ent. Yet, Peacocke argues, the correct account of your entitlement to the
judgement <this is a particle accelerator' will be far more complex than in the case of
judgements involving observational concepts, such as <this object is green'. Similarly,
while the judgement <I have switched on the light' may appear phenon1enologically
immediate, your justification will depend on a number of factors (including your
awareness of pressing the key). In contrast, in his view, your entitlement to a judge-
ment detailing your basic actions, such as <I've pressed this key', simply derives from
its being made on the basis of the experience of trying. 5
5 For an an alternative perspective on the parallel between the content of perceptual experience and
the range of an agent's basic actions, see Brewer (1999, ch. 8).
Mechanisms and Epistemology 29
A third option would be to agree that agents' awareness always includes an
awareness of bodily movements, but to insist that knowledge of actions under
descriptions referring to further effects can be equally direct. According to Ulric
Neisser's version of an ecological theory of perception, in perceiving intended
effects, effects that match the intention informing our action, we may directly per-
ceive what we are doing. Roughly, the idea is that perceived intended effects of our
actions are often experienced as being under our control. On a view like this, there
may be no difference in principle between the basis of your awareness that you are
pressing a key and that you are switching on a light; or between your awareness
that you are depressing the accelerator and your awareness that you are accelerat-
ing. We may perceive what we are doing just as directly in perceiving more remote
effects as in perceiving 'resident' effects. (For further discussion of Neisser's pro-
posal, see Ch. 18.)
Pooh knows that someone is shouting, but mistakenly takes the person in
question to be himself (the shout was in fact Piglet's). Intuitively, though, the
situation here is unusual, as Pooh himself is the first to acknowledge. Pooh's
claim to self-knowledge is based on observation and inference: he hears someone
shouting and hypothesizes, by way of inference to the best explanation, that
he himself is shouting. On the face of it, this is not the only way in which you can
know about your current actions: ordinarily you are aware of what you are
doing without inference, and without any room for the possibility of error
through misidentification. From a philosophical perspective this clain1 is
interesting for at least two reasons. First, it suggests that contrary to the Cartesian
tradition, self-awareness includes an awareness of one's physical as well as
mental properties; and it may provide an interesting perspective on the question
of what it is to be aware of a body as one's own body. (See Chapter 15 by Jerome
Dokic for extended discussion of this issue.) Second, the phenomenon of immun-
ity to error through misidentification is of course not restricted to our awareness
of our own action. It applies equally to knowledge of our own experiences,
thoughts, plans, memories, and so forth. A central question is: what explains such
36 Naomi Eilan and Johannes Roessler
immunity in the action case, and how does this explanation relate to explanations
in other cases?
The current line of thought implies that in the case of one's own actions,
knowledge of what is being done and knowledge of who is doing it call for a sin-
gle explanation. There is a way of being aware of one's actions such that it makes
no sense to wonder, with respect to an action of which one is thus aware, who is
the agent of that action. There is no gap between knowledge that an action is tak-
ing place, and knowledge that one is the agent of that action. However, there are
two kinds of pathologies of action that might be taken to call this view into ques-
tion. As Tony Marcel points out, patients with Anarchic Hand syndrome are pro-
prioceptively aware of an action they disown. This, he argues, conflicts with
Gareth Evans's claim that proprioception is a way of gaining self-knowledge that
leaves no room for error through misidentification. Actually, Evans's discussion of
proprioception was concerned with knowledge of the position of one's limbs, not
with knowledge of bodily actions (1982: 220-2). But Marcel's point speaks against
a natural extension of Evans's account to the case of agency. The point is that it
seems possible to be proprioceptively aware of a (manifestly goal-directed) move-
ment, without being aware of it as one's own action.
In his commentary, Christopher Peacocke offers a subtle (partial) defence of
Evans's thesis. He suggests that 'anarchic' movements are actions that are not
owned by anyone; they are actions that have no agent. If this is right, then pro-
prioceptive experience of anarchic movements provides no basis for knowledge of
anyone's actions, nor even for the judgement 'someone is acting, but is it I who is
acting?' In Peacocke's view, Evans's thesis is best interpreted as a thesis about
knowledge of the properties of persons. On this interpretation, the case of 'anar-
chic' movements does not constitute a counter-example. It remains true that in so
far as proprioceptive experience yields knowledge of the actions of an agent, it
yields knowledge that oneself is the agent of the action. Of course this is consist-
ent with Marcel's basic point, that being proprioceptively aware of an action does
not entail a sense of ownership. The upshot is that (perhaps surprisingly) unqual-
ified appeal to proprioceptive experience is not going to help in elucidating the
sense in which we are ordinarily aware of actions 'from the inside'-supposing
that such awareness 'from within' does entail a sense of ownership.
There is another kind of pathology that might be taken to render this supposi-
tion doubtful. In Chapter 14, Toelle Proust considers schizophrenic delusions of
alien control. In one kind of case, the content of the delusion is of the form: my
body is engaged in performing a particular action, but it is under the intentional
control of some powerful alien agent, rather than under my own control. Proust
suggests that we should think of this as an error through misidentification.
Patients are aware of the intentional content of their action. But they misattribute
the action to someone else. Proust's explanation of the mistake turns on her dis-
tinction between the basis of knowledge of the intentional content of one's
Mechanisms and Epistemology 37
actions, and the basis of knowledge of the agent of one's actions. Briefly, the idea
is that due to a disruption of certain information-processing mechanisms,
patients' experience of their actions lacks the characteristic sense of effort which
normally enables us to be aware of our actions as our own. In contrast, the basis
of their knowledge of the intentional content of their action is left intact. In other
words, Proust's claim is precisely that there is a gap between being directly aware
of the content of one's current actions and being aware of who is the agent of the
action. Thus, in her view, delusions of control help to illustrate a theoretical dis-
tinction that applies equally in the case of actions of non-schizophrenic agents.
This is also true in another sense. Recall the second issue raised earlier, the
question of whether there is a necessary connection between the fact and the
sense of ownership. Proust's reply to this would be in the negative. On her
account, there is no question but that the actions disowned by schizophrenic
patients are, unqualifiedly, the patients' own actions. Delusions of control do not,
in her view, reflect any in1pairment in action control. As she puts it, while a patient
with Anarchic Hand syndrome has a problem with control, a schizophrenic
patient has a problem with awareness of control.
Efferent Binding
Proust's approach draws on psychological theories of the information-processing
mechanisms involved in attributing actions to oneself. The basic thought inform-
ing much of this work is that to find out who is the agent of an action, it is neces-
sary to compare information regarding bodily movements (or their effects) with
information regarding the content of one's current intentions. To the extent that
the two pieces of information match, the movement is interpreted as one's own
action. Patrick Haggard refers to the process of 'associating intentions with the
actions they produce' as the process of efferent binding. Mechanisms of this kind
were originally invoked by theories of perception concerned with the question of
how the perceptual system distinguishes those movements of the retinal in1age
that are due to perceived movements in the world from those that are due to
moven1ents of the perceiver. According to such theories, the motor system issues
a copy of motor commands, which is then compared to perceptual feedback. Only
those movements of the retinal image that are not predicted on the basis of the
content of the motor con1mand are interpreted as movements in the perceived
world (Sperry, 1950; von Holst and Mittelstaedt, 1950). Haggard argues that etler-
ent binding is also important for action. Not only is it a crucial prerequisite for
successful motor learning-to identify and correct errors we need to be able to
detect mismatches between intention and performance. Furthermore, he suggests,
efferent binding is a central element in the construction of a sense of self.
One reason for being attracted to the idea that self-ascriptions of agency are
based on something like efferent binding is this. As noted earlier, ownership of
38 Naomi Eilan and Johannes Roessler
actions is a causal notion. For an action to be mine is for it to be controlled by
myself, more precisely, by my intentions. To be aware of an action as one's own,
then, is to have a piece of causal knowledge. To acquire such knowledge you need
to be aware both of the cause and its effect; and, of course, you need to establish
that the two things are related as cause and effect. That is what comparing the
content of your intentions with the course of your bodily movements (and their
effects) enables you to do. The process holds the solution to Hume's problem of
explaining our knowledge of the 'connexion betwixt an act of volition and a
motion of the body' (1978: 632) (which he regarded as even more mysterious than
causal connections in other domains).
However, it is far from clear how the idea is to be developed. One proposal might
be that efferent binding is a matter of making inferences, of the type 'My arm went
up at t. I was intending to raise my arm at t. So very probably, I raised my arm at t.'
This would be one possible reading of Haggard's remark that consciousness of
intention and consciousness of movement give rise to a further, 'derived' con-
sciousness of owernship of actions. Under this proposal, though, the intuitive
notion that we have direct knowledge 'from within' of ownership of our actions
stands exposed as an illusion. We may have a special way of acquiring knowledge
of our intentions, and of movements of our body, but there is nothing direct about
our awareness of ourselves as the agents of bodily actions. (One might defend this
counter-intuitive result on the grounds that in general, as Hume taught us, causal
knowledge is available only on the basis of inference. But of course it might be
replied that the correctness of this principle is precisely what is at issue between
defenders and detractors of the intuition of direct awareness of ownership.)
Alternatively, on Proust's account, efferent binding is the work of a subpersonal
information-processing mechanism, rather than of the subject's reasoning. Both
the operation and the input of the mechanism may be remote from conscious
awareness. You need not be aware of the content of the motor command that is
being compared to perceptual feedback. What is conscious, according to Proust, is
the sense of effort, which she suggests is the 'felt counterpart' of the subpersonal
efferent binding process (see also Frith, 1992). Tony Marcel presses a general ques-
tion facing this kind of account-the question of how, and why, the comparator
mechanism generates a (particular kind of) conscious experience. But there is also
the question of how the sense of effort contributes to our knowledge of owner-
ship. Perhaps the most natural interpretation of Proust's account is this. When
you automatically catch a ball thrown at you, your perception of the dynamic
structure of your movement provides you with knowledge that 'my body is
engaged in the action of catching a ball'. Now, a further ingredient in your overall
experience is a sense of effort. It is on that basis that you make the identity judge-
ment 'these bodily movements are under my intentional control'. You may then
conclude 'I am catching a ball'. On this account, the normal basis of self-knowledge
of (unpremeditated) actions certainly leaves room for the possibility of error
Mechanisms and Epistemology 39
through misidentification; and, as noted earlier, in Proust's view this is a good
thing, given that, in her view, the possibility is realized in the case of certain schizo-
phrenic delusions. Once again, though, the account seems to be in conflict with
the powerful intuition that we can be directly, non-inferentially aware of an action
as our own action.
We can make a rough division between two ways in which one may elaborate
that intuition. The first strategy is to accept the assumption that a sense of own-
ership of an action requires being aware of the action as something that is con-
trolled by one's intention, but to argue that such awareness is possible without
inference. The second strategy is to reject the assumption, and offer an account of
non-inferential awareness of ownership without appeal to any causal under-
standing. A clear example of the latter strategy is Marcel's suggestion that the sense
of ownership arises from the egocentric spatial content of motor specifications.
The self enters the content of agents' awareness here only as the point of origin of an
egocentric frame of reference, not as the subject of representations controlling the
action. (We will return to this account below.)
The first strategy is represented, in various versions, by the papers of O'Brien,
Roessler, and Dokic, and perhaps O'Shaughnessy. The point on which they agree,
notwithstanding their disagreement on much else, may be put like this. Consider
again the idea that the sense of ownership arises from a process of comparing
intentions and effects. Integral to that idea is the assumption that the representa-
tions of effects which provide part of the input to the process are neutral as to
whether the effects are self-generated or not. The effects may either be represented
as nlere movements, and perhaps their further consequences Cmy arnl is going
up', 'my arm is pushing the door'), or as actions of uncertain ownership ('my arm
is being raised'). In either case, the correctness of the representations depends in
no way on the movements being the subject's own action (rather than no action
at all, or the result of someone's else action). Put differently, the content of these
representations is provided by the 'highest common factor' of what is available in
the case of self-generated actions and in the case of passively witnessed actions or
movements alike. It is this 'common factor' conception that, at least in relation to
the personal level, proponents of the first strategy reject. For example, Roessler
argues that the perceptual experience of indended effects can intrinsically involve
a sense of control (and hence ownership), in so far as it is errlbedded in percep-
tual expectations arising from the agent's perceptually informed intentions.
Similarly, in spelling out the suggestion that a conscious action provides the agent
with a reason for self-ascribing the action, O'Brien argues that the consciousness
of an action consists in its being done with a sense of control, where this is a mat-
ter, roughly, of acting on the basis of an evaluation of the options. So the idea is
that there is a way of being aware of an action that is inseparable from an aware-
ness of the causal antecedents of the action-there is no gap between being aware
of the action and being aware of it as one's own, controlled by one's evaluations.
40 Naomi Eilan and Johannes Roessler
A final example is O'Shaughnessy's suggestion that trying is a conscious event of
which the agent is directly aware. Now, on the face of it, this may look like a ver-
sion of the second strategy, on which the sense of ownership does not constitu-
tively involve any causal understanding. However, O'Shaughnessy argues that
awareness of trying does not occur <in sheer isolation', but depends, in part, on the
agent's having committed herself (in other words, on her intending) to do what
she is trying to do, and on her belief, based on that commitment, that she will try
to do it. On a natural reading, then, part of the explanation why an introspectively
presented trying is experienced as one's own trying is that it is experienced as
controlled by the relevant commitments.
Immersion
A central concern of Tony Marcel's chapter is to articulate the sense in which our
awareness of actions as our own is immersed rather than detached. In general
terms, the idea is that while the self is in some sense present in the phenomeno-
logy of action, it does not figure as an object of awareness. More specifically, we can
distinguish three ingredients in Marcel's conception of immersed self-awareness.
First, in immersed self-awareness the self is not an object of perceptual attention,
it is 'perceptually recessive'. Second, imn1ersed self-awareness is to be contrasted
with reflective consciousness, that is, or with being introspectively aware of one's
mental properties. Third, Marcel concludes that in agents' self-awareness, the self
figures only implicitly. It is not hard to bring out the intuitive appeal of the first
two points. As noted earlier, there are strong grounds, both phenomenological
and experimental, for thinking that perceptual attention in action is typically
focused on the objects acted upon rather than on the actor's own body. Further,
consider the case of what Marcel calls immersed action, where (to add some fur-
ther metaphors) you are engrossed, or absorbed, by some temporally extended
activity. Intuitively, the self does not figure as an object of attention here, not just
in the sense that perceptual attention is not focused on the body, but also in the
sense that you are not entertaining any thoughts about your current mental states.
There is a prima-facie absence of introspective awareness in such cases.
The issue that these points help to bring to the fore is: in virtue of what does
action awareness involve an awareness of the self at all? How does the self enter
the scene? Part of the difficulty here stems from the intuitive connection (in the
case of intentional action) between, as we put it earlier, the fact and the sense of
ownership. If this connection holds, it is not as if awareness that I am the agent of
an action is an occurrence confined to those more or less infrequent occasions
when I choose to engage in self-reflection. Rather, it is a non-contingent feature of
actions that are controlled in such a way as to count as my own (intentional)
actions in the first place. If immersed actions are intentional, as they appear to be,
then it looks as if in explaining the sense of ownership we are barred from using
44 Naomi Eilan and Johannes Roessler
REFERENCES
ANSCOMBE, G. E. M. (1957), Intention. Oxford: Blackwell.
BREWER, B. (1999), Perception and Reason. Oxford: Oxford University Press.
CAMPBELL, 1. (1993), 'The role of physical objects in spatial thinking', In N. Eilan,
R. McCarthy, and B. Brewer (eds.), Spatial Representation. Oxford: Blackwell.
COLE, J., and PAILLARD, J. (1995), 'Living without touch and peripheral information about
body position and movement: studies with deafferented subjects', in J. Bermudez,
A. Marcel, and N. Eilan (eds.), The Body and the Self Cambridge, Mass.: MIT Press.
DAVIES, M. (2000), 'Persons and their underpinnings'. Philosophical Explorations, 3: 43-62.
DENNETT, D. (1998), 'Reflections on language and mind', in P. Carruthers and J. Boucher
(eds.), Language and Thought: Interdisciplinary Themes. Cambridge: Cambridge University
Press.
EVANS, G. (1982), The Varieties of Reference. Oxford: Oxford University Press.
FRITH, C. (1992), The Cognitive Neuropsychology of Schizophrenia. Hove: Erlbaum.
FRYE, D. (1991), 'The origins of intention in infancy', in D. Frye and C. Moore (eds.),
Children's Theories of Mind. Hillsdale, NJ: Lawrence Erlbaum Associates.
Mechanisms and Epistemology 47
Anthony Marcel
If it stinks, if it rots your little conscience, in the passive voice it goes! Nuclear
devices were dropped-shots were fired-feelings exist-No! Say it in your
person, I dropped, I fired, I feel!
Bela, No End ofBlame, II. iv, by Howard Barker (emphasis in original)
1. INTRODUCTION
1.1 Overview
All four quotations cited above connect action and self, or self consciousness, and
as such are in a long tradition. But the first and second pairs pull in opposite
I am the author of this chapter in that I wrote it. However, certain other people are part agents of it, in
that they played causal roles in its production and its content, and therefore must take partial respons-
ibility, whether they intended the result or not. Johannes Roessler and Naomi Eilan organized and
invited me to the workshop where this started. Johannes persisted in aiding and abetting the execution
and inclusion of this chapter. He has also failed to decline to converse with me over several years.
Naomi Eilan has failed to decline to converse with me for a much longer period, and has helped to cla-
rify the chapter. They are clearly accessories both before and after the fact. John Campbell has said
much that has affected the content. Christopher Peacocke has even more directly contributed, though
I have failed to profit from many of his comments. Others that have perhaps unwittingly played a
causal role in conversation (though ignorance is no excuse) are Jonathan Miller, Jim Russell, Chris
Frith, Nick Humphrey, Sergio Della Sala, and Jonathan Cole. For their part, I plead as Portia that mercy
may mitigate justice; for my part, I declare as Shylock responds to her, 'my deeds upon my head'.
The Sense ofAgency 49
directions, the first in focusing on action and the second on self. Put slightly
differently, the first two stress the non-detached immersion of self in action, the
second two insist that actions are someone's: they constitute that person, who is
responsible for them, and their sense of self. But if one's consciousness is focused
on action, how does this relate to consciousness of self, especially to a self that is
constituted by past deeds and can be reflected on? And do data from cognitive
neuroscience, which tends to deal with the subpersonal domain, throw light on
these personal-level issues?
This chapter is about the sense of agency as a form of self consciousness. More
precisely, it is about the relationship between the sense of oneself as an agent "/
and one's awareness of one's own actions. The very use of these terms requires
some preliminary ground-clearing and clarification. Empirical data will be
described which challenge implicit and opposite assumptions in philosophy and
psychology. A crude generalized description of these two (sets of) assumptions is
as follows.
In philosophy, it is often assumed that (a) it is essential to an event being an J
action that it is linked to a subject's sense of ownership of the action, (b) being
an action is linked to the subject's (conscious?) knowledge of what he is doing
in acting, (c) there is a constitutive link between the actor's awareness of an
action and a sense of agency.
In psychology, in contrast, it is often implicitly assumed, in the way that research
has been conducted, that we can give a psychological account of action which is '/
independent of the further (on this view) psychological questions of (a) whether
actions are accompanied by a sense of ownership, (b) whether actions are accom-
panied by awareness of what one is doing, (c) whether one can give separate
accounts of awareness of action and a sense of agency or ownership of the action.
Obviously, these assumptions turn on the relevant definitions of their terms. In
what follows my interest is principally in the third assumption in each set, namely,·
the dependence or independence of awareness of one's action and the sense of . j
ownership of that action. Necessarily this involves aspects of the other assump-
tions. This raises a large number of conceptual and empirical issues. I shall use
two sets of empirical observations to narrow these issues down to two main ones.
(a) What is the nature of the content of awareness of our current actions? This
decomposes into two questions. (i) To what extent are we aware of exactly ./'
what we do? (ii) In what form is the content of that awareness, in terms of
whether or not it includes the agent of the action; and, if so, in what way?
(b) Can a sense of self-ownership of action come apart from awareness of one's , (
action? If so, what do the relevant phenomena tell us about the kind or
kinds of awareness we have of our selves and of our actions?
I shall end by making a proposal about how a sense of agency comes into one's
awareness of action in the normal case and how it relates to pathology.
50 Anthony Marcel
Interestingly, while psychology has assumed that separate accounts can be given
of awareness of action and of a sense of agency or ownership of action (to the
extent that it has treated ownership at all), psychologists treat as pathological and
are puzzled by cases where someone performs a seemingly well-formed action but
says that they do not feel that they themselves have done it. Their puzzlement
implies that they assume that awareness and ownership of action normally go
together. Whether they go together by necesssity or contingently is another
matter, but one that will be raised here.
1 I use the term self consciousness here without a hyphen to refer to something general, as distinct
from self-consciousness with a hyphen referring to being aware of oneself from the outside as in
mirror reflection or through the eyes of others, connected with embarrassment.
The Sense ofAgency 51
implicit in the particular experience. For these and other cases, the question arises
to what extent and how the different selves are linked. Although the same word is
used and although we may think that these refer to a single entity, it is not clear
that they are mentally linked or how, other than that they are all self conscious-
nesses born or experienced by the same mind. Just like in the preceding examples,
there is a distinction between a long-term sense of agency and an occurrent sense
of agency. The former is the sense of one~elf as an ~gent apart fro~ any particu- , (
lar action, for example, as causally effective over tune; the latter IS the sense of
oneself as performing a particular action at or around the time it is performed.
What each of these consists in is the topic of this chapter. I am more especially
concerned here with the occurrent sense of agency, but I shall want to say
something about the long-term sense of agency, one reason being that the former
may be related to it or based on it in important ways.
The third distinction is between detached and immersed self consciousness.
This is akin to William James's (1890/1950, ch. 10) distinction between the Me
and the I, though not exactly the same. The former is the reflective, introspective
observation of something which is phenomenologically separate from the
observer and which is a perceptually distinct object. The latter is the kind of self
awareness one has when much involved in an activity that is not self-focused,
where the self is implicit and perceptually recessive. It applies not only to the experi-
ence of oneself as the doer of whatever activity, but also to what Gibsonian psy-
chology refers to as ecological self-awareness. Thus perceptual information
specifies egocentric self-location and spatial embodiment, and this along with
proprioceptive awareness gives one a sense of oneself as an experiencing organism
and a sense of one's current capabilities. Primarily I mean the distinction to be
between reflective and non-reflective consciousness, in this case of oneself. (See
Gallagher and Marcel, 1999, for further discussion.) How this distinction mani-
fests itself in accounts of one's sense of agency and the awareness of one's action
will be developed throughout this chapter. /
I am mainly concerned here with a minimal sense of agency, one that is occur- J
rent and immersed. It may be that a full sense of agency requires a third-person
detached and long-term conception of oneself. Whether one is built on the other
will be returned to at the end. It is important to bear in mind that the topic here is
phenomenology, though primarily in a non-technical sense. Although what it is
conceptually to be an agent may affect one's feeling of oneself as agentive, the
latter is the present focus. Therefore what is of concern is the nature of one's aware-
ness of one's actions. Furthermore, it is not merely the awareness of such actions, j
but awareness of them as actions and as one's own. There are two kinds of question
here: (a) 'what is it to be aware of one's action as one's own?' or 'what does one
mean by being aware of an action as one's own?' and (b) 'what makes it such that J
one experiences an action as one's own?' The first is more a matter of definition,
the second is more empirical and is that with which I shall be concerned here.
52 Anthony Marcel
To be more specific, when we are concurrently aware of relatively punctate bodily
actions, what is the content of that phenomenology? What is the relation between
the awareness of the action and the experience of its ownership or source?
intentionally to do anything then one has to be 'within' the action we are attempting to
observe, in which case we have an entirely empty and self-delusive experience of observa-
tion .... or else we remain 'without' in some less serious sense and genuinely seem to
observe the action. But remaining 'without', we lose the action as ours in gaining the observa-
tion: we lose any'withinness'. The action becomes for us a mere event in the world, and we
ourselves become dispersed and lost among the bric-a-brac of the world: we become the
world in our own eyes; we suffer the experience of loss of identity.
2. AWARENESS OF ACTION
To what extent are we aware of our intentions? One reason for asking this ques-
tion is that it will playa part later in assessing the extent to which awareness of
intention plays a part in our awareness and ownership of action. Another reason
is that being aware of intentions may playa part in having a long-term sense of
agency or a concept of agency in oneself or others or both. Oddly, many psycholo-
gists seem to assume that intentions are by their nature conscious. In fact, there
are several kinds of ways in which we are unaware of them.
1. We frequently lack explicit awareness of subgoals in achieving a goal. This
applies not only to how one is achieving the goal in the sense of the manner of an
action, but also to actions that are instrumental to the overall goal. When I pick
up something fronl a low table I may be quite unaware whether I intend to do it
by bending at the waist or at the knee. Even when I perform a subgoal first as a
discrete action, I may be unaware of intending it, for example, moving an object
to get to the one behind it. In most cases, one can often (though not always)
become aware of such intentions. But this does not contradict the fact that we
often happen to be unaware of such intentions.
2. In the course of temporally extended actions, we may forget our intention
or the reason for it. These cases often consist of awareness of the subgoal with
temporary unawareness of the goal. Such cases of absent-mindedness sometimes
result in our wondering why we are doing something (why did I come upstairs?)
and sometimes result in the action being smoothly transformed into another
more habitual action, for example, on entering the bedroom to retrieve a book,
instead one gets into bed (Reason, 1990).
3. There exist certain kinds of speech errors (and supposedly errors of other
kinds of action) which appear to be explicable only on the basis of a non-
conscious intention realized as a speech or action plan that is synchronous with
and competitive with our conscious intention and plan for speech or action, but
unrelated to it. It emerges into speech (or action) only where there is a phonolo-
gical (or nl0tor) equivalence at a structurally equivalent point in an utterance or
action. These kinds of errors were originally explored by Meringer and Mayer
(1895/1978); Butterworth's (1981) review suggests the reality of such errors and
their resistance to explanation in other ways. If so, then we do have intentions for
immediate action of which we are unaware.
4. Theorists often identify non-conscious long-term intentions whose presence is
implied by the effect of threat to their violation. These may be seen as dispositional
The Sense ofAgency 61
or as concerns rather than as intentions. However, it is easy to come up with ones that
are hard to reduce to biology, such as the goals to be liked or to maintain self-esteem.
We are unaware of such intentions until they are impinged on.
5. One kind of case is particularly relevant to this chapter. In immersed
ongoing action, where we are not in a detached, self-reflective state, we may not
only have a general sense of acting intentionally but be aware of each intention. But
we are often unaware of the specific content of each intention. When focused, in
sport or a craft, the expert acts intentionally but may be unaware of the specifics of
tactical intentions, for example, in tennis whether one intends to play a drop
voll~y or a drive volley even when the postural aspects of one's approach to the net
is a selective preparation for one rather than the other. Indeed, this is why even
when such experts sincerely claim unawareness of their intention, their opponent
can anticipate the shot, though the opponent himself may not know how he did so.
Unawareness of intention appears to be more common than supposed. In so far
as we are unaware of our intention, it is hard to see how intention, in those cases
where we are unaware of it, can be an ingredient of felt ownership of action or
playa significant part in the sense of our causality of an action.
One of the main questions here is the nature of the content of awareness of
actions. But along with this question must go another that is more usual in psy-
chology but has not been addressed adequately: 'To what extent are we aware of
exactly what we do?' There are several examples of quite dramatic unawareness of
our actions. This is important, because if we are wrong about our actions it may
mean that what feeds our awareness of actions is not the actions themselves. If
not, this not only raises the question of how we control our real behaviour, but
also the questions of what it is that actually gives rise to our sense of producing
actions and of the nature of our knowledge of them.
Unfortunately, the norn1al case provides little relevant evidence. Where action
accords with intention, which it normally does, accurate awareness of action may
be no more than accurate awareness of the intention. One might respond to this
suggestion by saying that the fact that we correct errors in actions that leave no
mark on the world to be perceived at leisure surely means that we are able to mon-
itor our actions accurately. However, we have to be careful. In studies of error
detection and correction in serial choice reaction times, corrections can occur as
fast as 64ms after emission of error responses (Rabbitt, 1968), implying that the
error was detected before the erroneous response itself was made. This may be due
to detection of perceptual rather than response error. But it does raise the possib-
ility that what we monitor is something prior to the action itself. This does not
mean that we are unable to be aware of feedback from movement itself, only that
62 Anthony Marcel
our initial awareness of action is pre-motor and that we may not always monitor
motor feedback. We need to examine cases where there is a discordance of action
with intention, to see if there are cases where, even when the actor is trying to
attend to what he does, his awareness of his actions is deficient. This amounts to
producing illusions of action. Jones (1988) has reviewed several motor illusions
that reveal a lot about proprioception. I shall focus on one of these on which I
have recently conducted research, since the results are dramatic in their relevance
to the present issue. This concerns vibro-tactile illusions of limb position and
movement.
110
-+- Reported
100 -0- Pointing
c:::: -0- Grasping
0 90
'';:;
'Vi
0
0- 80
~
~
70
E 60
~ 50
V'l
OJ
~ 40
.01
OJ
"0 30
OJ
01
C'O 20
OJ
~ 10
0
-10
-7 -6 -5 -4 -3 -2 -1 0 2 3 4 5 6 7 8 9 10
-seconds +seconds
Point of initial
maximal illusion
Time preceding/following initial maximal illusion
with the other hand, or (c) to point to it with a pointer held in the free hand. Since
the time taken to reach the maximal point of the illusion varied between indi-
viduals, we calibrated the results according to time of the initial maximal illusion.
The results are ShO¥lll in Figure 2.1, which displays the time course of each kind
of response. in terms of the average error in angular degrees (around the elbow)
from the true position of the wrist, as well as the range for different subjects.
There are four interesting effects. First, if you ask people to point to the location
of their wrist with a pointer in their free hand, they point to where they feel their
hand to be, as is indexed by oral report. My interpretation is that pointing as deixis
is a communicative speech act, just as matching by mimicking with the other hand
is a kind of 'showing' one's phenomenology. Second, if asked to grasp the wrist,
they grasp the correct location, and do so if asked within about 4 or 5 seconds after
the initial maximum illusion. But, third, after that the grasp seems to be captured
by the experienced location. This is very similar to other phenomena, observed by
Bridgeman (1989), where motor adjustments of the hand and eyes are initially
immune to visual illusions of location and movement of external objects, but get
captured by the illusion after a few seconds (the Roelofs effect). Like Bridgen1an,
we interpret such effects in terms of what is under the control of conscious
experience or intentions based on it at different times. A fourth finding in our own
situation was that people most often seemed not to notice any disparity between
64 Anthony Marcel
the experienced and grasped locations! This is what we followed up in a second
experiment.
The situation is depicted in Figure 2.2. The person's dominant arm is lowered
to the elbow on a rest and their forearm is held on a horizontal splint attached to
the rest that swivels in the horizontal plane from the elbow, but can be held
Surface
occludes arm
+ with target LEDs
\~ Surface paper
to draw intended and/or
performed movement
\ 11..--II....-./
Forearm fixed to
swivelling armrest
-fixed or free-moving
.
.
·tJl
..
, ,
,': '\
Real position Illusory experienced
position
2.4 Explanations
How should we think about these cases of unawareness or radically inaccurate
awareness of movement? Hypotheses that psychologists have drawn on in move-
ment control may be applicable. First, are any of these phenomena a matter of the
level at which we attend, for example in immersed versus non-immersed action? In
The Sense ofAgency 69
driving, I may consciously intend to overtake and to gain more acceleration, but be
unaware of what downshift of gear I make, from fourth or fifth to fourth, third or
second, or even of whether I have shifted gear. But in such cases I am able to be so
aware, by attending appropriately. While it may be that the anosognosic patient's
problem may be one of the target of or control of attention, this is unlikely to be
the case for the vibro-tactile experiments, because no matter how hard they try, a
number of those normal people who are una,vare of their true movement direc-
tion still get it wrong even when informed. The reality of our bodily disposition
and movements seems inaccessible to awareness for at least some people. The same
point has been made in research by Fourneret and Jeannerod (1998) in which
people tried to draw lines in the sagittal axis with a stylus but where their hand was
hidden and they saw the lines on a computer screen, Unknown to them, the
computer made the lines appear to deviate in either direction. Some subjects
experienced their movements as underestimates of their apparent deviation,
others as in the opposite direction to their actual adjusted movements.
Two possible reasons for such inaccessibility have been proposed. The first may
be seen as a variant on the level of attention hypothesis. Several people, notably
Wolfgang Prinz (1990; but see Brewer, 1993, for review), have suggested that in
action our control target or our awareness is focused on the distal object in
the world or on the goal rather than on the unfolding of our own action, and it
is extremely difficult to become aware of the latter. The second possible reason lies
in the notion that the implementation of our intentions in moven1ent is in tern1S
that are incommensurable with both our intentions and our phenomenology.
One example of this is procedural versus declarative knowledge, which is the heart
of Polanyi's (1966) discussion of tacit knowledge. Another example is Jacques
Paillard's (1991) suggestion that intentions for spatial action are in the form of
a static representational geometry of location which is implen1ented in terms of a
vectorial geometry of transformatory movement paths. Although these ideas may
be generally true, a problem with applying either of these kinds of account to the
present effects is that they seem to predict universal effects, whereas in the vibro-
tactile experiments there were individual differences. Indeed, skilled athletes,
dancers, and actors can become aware of the specifics of their actions. A nice
example of a universal dissociation is in catching balls travelling in a high para-
bolic trajectory towards oneself, as fielders in cricket do. Reed, McLeod, and
Dienes (2000) have recently documented the fact that people not only believe but
experience their angle of gaze as rising then falling as the ball falls to eye-level. But
in fact, if the ball is caught at eye-level, the gaze goes up continuously and never
falls. One can see this easily if one draws the parabola and a stick-figure. This is
probably paradigmatic of the 'Inner Game' technique in sports, where accuracy
and control of movement is aided by attending away from the movement itself.
Since one's awareness is acknowledged to be inaccurate, attempting to base learn-
ing and control on it is likely to be disruptive. William James's advice to avoid
70 Anthony Marcel
attending to your movement while descending the stairs is correct but possibly for
the wrong reason.
One should be careful in drawing general inferences from data such as these or
the vibro-tactile experiments. One may be tempted to infer that in the normal
case our actions accord with our intentions and our awareness of our actions is
accurate only fortuitously, because normally the physics and biology ensure agree-
ment between action and intention and awareness is restricted to our intentions.
But if this were the case, it is difficult to see how we could do what we intend
whenever the physics changes (e.g. in gravity change or in water) or how we could
learn to control our action finely (whether in sport or under prismatic distor-
tion). In addition, as already said, people can become aware of their movements.
However, although one can become aware of the specifics of one's movements, for
example, by learning to attend to proprioceptive feedback, it may be that there is
something that has to be overcome to do so, a kind of default or something that
predominates in awareness of action. A clue to this is provided by recent research
by Patrick Haggard (see Chapter 4, this volume). Indeed Haggard's results will
play an important part in a central hypothesis developed later in this chapter.
Haggard and Magno (1999) investigated awareness of one's voluntary action in
a simple reaction time task where subjects had to judge when they made their
response. On a proportion of trials they applied Transcranial Magnetic
Stimulation (TMS) to the cortex, which introduces a localized neural interruption
and lag, at a temporal point chosen to produce substantial latency delays. When
TMS was applied over the primary motor cortex it produced a long delay in reac-
tion time (RT) but a small one in awareness of the response. When TMS was
applied over the Supplementary Motor Area it produced a small delay in RT
but a long delay in awareness of the response. Haggard and Eimer (1999) tried to
distinguish whether awareness of intention and action in Libet et al.'s famous
(1983) work are related to specific or non-specific motor processes. They did so
by relating such awarenesses to the General Readiness Potential found before
movement and the Lateralized Readiness Potential (LRP, or Bereitschaftspotential)
associated with cortical selection of specific movements and the Supplementary
Motor Area. Essentially, subjects freely chose when to make a movement and had
to report by reference to a clock either when they moved or when they 'first began
to prepare the movement'. The more informative condition turned out to be
where subjects could also choose whether to make the movement with the left or
right index finger. Their finding was that not only the awareness of movement but
also that of initial preparation of the movement were unrelated to the General
Readiness Potential but both covaried with the Lateralized Readiness Potential:
early and late awareness were related to early and late LRPs respectively. This
strongly suggests that initial awareness of movement is pre-motor, in the sense of
deriving from the specifications for movement rather than the movement itself
(as recorded from electrical activity in muscles). However, it should be noted that
The Sense ofAgency 71
the LRP is associated with both movement specifications and energizing of effec-
tors. These data are very important. Initial awareness of a voluntary action
appears to derive from a stage later than intention but earlier than movement
itself. It appears to be underlain by the specification of the movement and this at
least partakes in the Supplementary Motor Area.
These points will playa part in the view which I will later develop in relation to
felt ownership of action. However, two points should be made about Haggard's
work. First, the data do not tell one about the total possible awareness of action or
intention; they only tell one about initial awareness and what may underlie it. Much
information may derive from later stages of action, especially in non-punctate
actions. Second, Haggard refers to the content of our awareness of action, propos-
ing that it consists in awareness of pre-motor processes that may playa role in the
organization of forthcoming n10vements (Haggard and Magno, 1999). However,
the data do not tell one directly about the content of the initial awareness. They only
give a clue to what non-consciously underlies that content. Although motor specifica-
tions n1ay underlie initial awareness of action, we are not necessarily aware of them
as such: that is, not under that description.
At this point we can review what are the possible constituents of awareness of
action, and of awareness of it as action rather than as movement or reflex. I am
concerned with the potential contributors to the phenomenal ingredients of such
awareness rather than with potential causes or substrates, but for the moment
without regard to the description under which such phenomenal ingredients are
experienced. I shall briefly review several candidates in terms of whether each is
(a) necessary or (b) sufficient for such awareness. The overall answers and the
main reasons are summarized in Table 2.2 below. I will also comment here on the
relevance of certain phenomena to felt ownership of action and movement,
although these inferences are not shown in Table 2.2, since I will deal separately
with ownership in its own right in the section following this.
Proprioceptive Awareness
Proprioceptive awareness cannot be necessary for awareness of action. This is
shown by the case of Ian Waterman (IW), who has been reported by Jonathan
Cole (Cole, 1993; Cole and Paillard, 1995). Through a peripheral neuropathy that
caused demyelinization of the large afferent fibres, IW was deprived of all proprio-
ceptive experience and bodily sensation beneath the neck, except for deep pain
and temperature at the surface. When he does something he may not be aware
without vision of the disposition of his limbs and body, but he does know that he
has acted. Incidentally, this also suggests that proprioceptive awareness is not
necessary for ownership of action.
72 Anthony Marcel
TABLE2.2. Necessity and sufficiency of possible constituents of awareness of action:
summary of the arguments and evidence
Awareness ofIntention
It is obvious from the large variety of unawareness of intention, summarized
earlier, that awareness of intention cannot be necessary for awareness of actions,
either for awareness of them as actions or as one's own. Trivially, there are many occa-
sions when we cannot remember why we are doing something (absent-mindedness)
where we neither deny its status as an action nor disown it. This is not to deny that
some sense of voluntariness may be necessary for one to be aware of actions as
actions, but awareness of the specific intention per se is not. A stronger reason for
excluding the necessity of awareness of intention as necessary for awareness
of action, since it does not involve forgetting, is a phenomenon that I shall return
to later in connection with ownership of action. Prior to certain surgical
The Sense ofAgency 73
operations, Hecaen et al. (1949) electrically stimulated the patients' central thala-
mic nucleus on one side, producing in the contralateral hand clenching and
unclenching of the patients' fist or 'pill-rolling' movements. The patients 'asserted
the wilful character of such induced movements' and described them as actions,
saying that the movements had a goal-directed and voluntary character, but had
no idea why they had made them. The question arises as to whether thalamic
stimulation merely induces movements and some aspect of will (e.g. effort) or
whether it produces an unconscious intention in some sense. Certainly the
patients never said that they felt an urge to do the action, but did say that the
.action had a wilful character. Nor did they disown them. Apparently, one does not
need to be aware of an intention to experience an action as an action.
Is awareness of intention sufficient for awareness of an action? The behaviour of
patients anosognosic for plegia may suggest that it is. As noted above, some such
patients when asked to raise each arm separately seem convinced that they have
raised their plegic arm in the same way that they are convinced that they have
raised their non-plegic arm (veridically). It may be that awareness of their inten-
tion to do so, if not contradicted (at least not consciously contradicted) by feed-
back or proprioception, suffices for experience of the action. However, what is
relevant may rather be something that is closer to action itself than intention,
which may be either effort or awareness deriving from n10tor specifications or
commands as suggested by Haggard's experiments. On the other hand, as Marcel
et ale (forthcoming) point out, since only some anosognosic patients assert they
have n10ved their plegic limbs, and only in response to questioning, it is possible
that they do not have the experience of action prior to the question.
Awareness ofEffort
It is difficult to rule out entirely the necessity of the experience of effort. However,
there are cases of virtually effortless actions of which we are not unaware. One
example is movement of body segments that have very little mass or inertia, such
as the index finger. An even better example is action in zero gravity conditions:
astronauts are not unaware of their actions when they are not visually monitoring
them. There remain the resting forces of muscle, skin, and so on that hold body
parts in their postures of relaxation. But normal people are hardly aware at all of
overcoming such forces with the forefinger or arm. The case of IW may also be
relevant here. He does have some preserved but greatly diminished awareness of
force production. However, Jonathan Cole suggests for several reasons that the
source of this awareness is peripheral intramuscular tension rather than central.
The normal sense of 'trying' is central. IW has some sense of peripheral muscular
tension which he then interprets as 'trying'. IW is aware in voluntary actions of
doing something, even if, without vision, he is unaware of precisely what he is
doing. If, as Cole suggests, he lacks the normal central sense of effort from volun-
tary movement, this cannot be contributing to his awareness of his actions.
74 Anthony Marcel
By contrast, experience of effort may well be sufficient for awareness of action,
especially where there is no prior intention. When we resist a sudden increase in
wind pressure to maintain posture, or adjust to resistance or force changes when
walking in water, these are not reflexes in the strong technical sense of the word;
they are voluntary in that we can decide not to do them. But they certainly require
no conscious intention. Apparently, the effort is sufficient for awareness of action.
3. OWt~ERSHIP OF ACTION
Can self-ownership come apart from awareness of one's action? This question can
take two forms. (a) Can one's action be experienced as another's? ( b) Can one's
action be experienced as unowned or of uncertain ownership? The question arises
of whether one can be uncertain or mistaken as to who performed someone else's
action but never be uncertain or mistaken about who performed one's own
action. If the answer is that one can never be mistaken or uncertain about owner-
ship of one's own actions that one is aware of, then we might look to the content
of awareness of action per se to find the determination of experienced ownership.
This is one of the variants of the content of awareness of action that I set up
earlier in Section 1.5. If awareness of action and experience of its ownership can
come apart, the way and circumstances in which they come apart should be
inforlnative. I shall give two kinds of example of disowned actions. I shall then
review cases where the urge, reason, or intention for an action is disowned but not
the action itself. In distinguishing what aspects of action can be owned or
disowned, this should help to sharpen where a sense of agency may lie.
The first, weak, case of disownership of action is semi-involuntary imitation of
uncharacteristic action. To explain what I mean, it will help to recount an anec-
dote told me by Jonathan Miller, the theatre director (personal communication).
As a young man, he came across a crowd who, he learned, were waiting for the
Queen to pass in a car. Although this was one of the last things Jonathan would
consider doing, he could not easily get past. When the car passed, the crowd
cheered and waved. (This happened in the 1950s when royalist adulation was still
widespread.) Jonathan says that to his surprise and horror, given his self-image,
(1 found myself raising my arm and starting to cheer', which he quickly suppressed.
One thing that embarrassed him was that his behaviour was something which
seemed under his voluntary control and could have been inhibited even earlier
had he been aware of it before. Yet he says that he did not feel that the actions
(raising his arm and cheering) were (his'. He supposes that they were induced by
76 Anthony Marcel
a kind of contagious imitation. Note the phrase 'I found myself doing x. Such
phrasing means either that one is not until a certain moment self-aware at all or
that one was not aware (specifically) of engendering the action. One is aware only
from the outside as an observer. The type of action in question and the circum-
stance of its induction is often one that we ascribe to fugue states or hypnosis
where one says 'it wasn't me that did it'. Usually in fact, the action itself is owned
but not the intention ('I didn't mean to do it.'), or even if the intention is owned
the responsibility for the action is disowned. One of the probable reasons for this
is that this kind of disowning usually relates to actions that are socially, personally,
or legally illegitimized. Neither self-image nor character witnesses are a reliable
guide to what we are actually capable of doing or thinking, but they are clearly
powerful influences on what we feel we are capable of. Incidentally, the kind of
imitation recounted here puts pressure on the neat dichotomies between volun-
tary and involuntary and between intended and unintended. Indeed, 'lynch mob'
induction is not automatic: being in a critical or detached attitude reduces its like-
lihood and so does the situational content and actors: I doubt that a Jew would
ever be susceptible to join a crowd of Jew-baiters, unless deliberately pretending.
One further point is worth making. Jonathan Miller concedes that he did not dis-
own doing the movements, but it was the action that he felt alien to him. When we
discuss agency, it is important to distinguish, and to realise that people distin-
guish, what it is of which one is the putative agent: the level of description and the
description under which it is known. This is important, since I want to consider a
second, stronger, case, where everything about an action, at all levels of descrip-
tion, is disowned, and more immediately.
The second, stronger and more pertinent, case of disownership of action is a
phenomenon found in neurological patients that has been called Anarchic Hand
(Della Sala et aI., 1991, 1994). This phenomenon first has to be distinguished from
two other clinical phenomena. After discussing it, I shall also distinguish it from
other kinds of disownership related to action.
Anarchic Hand has an experiential aspect and a component of action control;
the former is frequently confused with Alien Hand and the latter with Utilization
Behaviour. In Hemisomatoagnosia or Alien Hand, one hand does not feel to be
one's own, especially when held with the other hand. This is a sensory phenom-
enon and has little to do with movement, whereas in Anarchic Hand it is certain
actions performed by one hand that are disowned, not the hand itself. These dis-
owned actions are unintended or seem so to the patient, and it is this that gets con-
fused with Utilization Behaviour. (Indeed in Anarchic Hand there is no evidence to
discount the explanation of an unconscious intention.) It is worth describing
Utilization Behaviour since it is also relevant to agency. It is observed in patients
with a bilateral focal frontal lesion. If there is some object that can be used or
manipulated within the patient's vision and within reach, the patient will use it to
perform actions appropriate to the object, though they have been asked not to do
The Sense ofAgency 77
so. The actions can be unimanual or bimanual. Even if the patient is given some
other task unrelated to this object and even if they are asked to desist from their
behaviour, they cannot stop themselves performing actions with the irrelevant
object (Lhermitte, 1983). Such patients do not appear to understand that they are
doing anything inappropriate, even though they can repeat the instruction to
desist. Thus, if on entering the bedroom the bed is visible they will undress and go
to bed, or if a matchbox is within reach they will light matches and if a candle is
present light the candle. If asked what they are doing, they will calmly say that they
are going to bed or lighting the candle, while either being unable to say why or
fabricating a reason or saying that they felt like doing it. In such cases, the patient
does not disown the action. However, this and the patients' explanation or lack of
one could be due to a lack of concern ·which is characteristic (attributed to such
patients' large bilateral frontal lesions). In Utilization Behaviour the non-intended
actions are environment-driven. Most current explanations of this suggest a weak-
ening of whatever is responsible for ensuring the implementation of intended
actions such that they normally override habits or environmentally driven
behaviour. Environmentally driven behaviour is quite common in normal people
especially during low arousal or after waking, and is responsible for characteristic
slips of action noted by Freud (1914/1975) and reviewed by Reason (1990).
In Anarchic Hand the phenomenon is unimanual (bimanual cases have not so
far been observed) and can involve either the left or right hand, consistent for an
individual. (One case of Anarchic Foot has been reported.) The review by Della
Sala et ale (1994) implicates damage to the Supplementary Motor Area, though
cortico-basal degeneration is also often involved. The affected hand performs
unintended but complex, well-executed, goal-directed actions. Often when the
patient is trying to do something with the unaffected hand, the other hand
appears to do the opposite or compete with it. (The patient tries to answer a ques-
tion by pointing to (Yes', and the other hand points to (No'; the patient tries to
button their shirt, and the other hand tries to unbutton it.) But the (anarchic'
action can be quite unrelated to an intended action or can occur even in the
absence of any concurrent action. Although the action is often unwanted or even
socially unacceptable (taking left-overs off a neighbouring diner's plate was
observed by Della Sala), it is not always so. The phenomenology of this is differ-
ent from that in Utilization Behaviour, where there is lack of concern. Here the
patients describe the anarchic hand as having (a will of its own' and are often
terrified. Sometimes the patients talk to their hand asking it to desist and occa-
sionally talking to it in a high pitch, as if to a child, often only able to stop the
action by holding the hand down with the other one or by sitting on it (observed
by myself). Yet the hand per se is not felt as (other'; at most, it is sometimes felt
only as an alien agent when performing its unwanted actions.
Although this differs in a number of ways from Utilization Behaviour, most
examples are of responses to visually perceptible stimuli. However, not only is the
78 Anthony Marcel
case of unbuttoning the shirt unlikely to be visually driven (typically, this action
is not perceptually monitored); there has also been one case reported of the anar-
chic hand trying to strangle the patient while he was asleep, where there was no
external perception, and which also suggests that another, intended action is not
necessary to provoke the anarchic action. What seems to be true is that in such
cases the individual has no sense of their own intention or urge or effort in the
anarchic action, and indeed they may struggle with the other hand to suppress the
action. The wilfulness of the anarchic action is felt only by the unaffected hand as
a resistance to its attempts at suppression, a point I shall return to later. This phe-
nomenon is depicted in the film Dr Strangelove and is a long-standing theme in
the gothic tradition. Indeed, the long cultural and artistic history of the phenom-
enon might suggest that, apart from its neurological existence, it can occur as a
form of dissociative defence mechanism, much as splitting and fugue.
It has been suggested that Anarchic Hand is really a unilateral form of
Utilization Behaviour, the latter involving extra damage that leads to lack of
concern. If so, then disowning in Anarchic Hand is due to the non-suppressed
environmental provoking of action. However, there are a number of features
which makes this unlikely. First, this does not account for the difficulty the suf-
ferer has to stop the action, which seems to have a persisting wilfulness. Second,
cases exist where there is no obvious environmental provocation. Third, when
normal people produce environmentally driven actions (e.g. shortly after waking;
see Reason, 1990), they do not experience them as alien, and in Utilization
Behaviour the patient does not disown the actions. Fourth, I have seen a patient
who simultaneously exhibited Anarchic Behaviour in his left hand and Utilization
Behaviour in his right hand, being troubled by the former but unconcerned about
the latter. 2 This is important since it discredits both the notion that Anarchic
Hand is a unilateral form of Utilization Behaviour and the notion that the phe-
nomenology of being concerned is a matter of extent or bilaterality of the lesion.
In addition, some of this patient's Anarchic actions were objectless gestures to his
rear outside his field of vision. They could not have been environmentally driven,
except if one includes the possibility of an imagined person or object behind him.
As far as the control of action is concerned (as distinct from the phenomenology
of ownership), it is plausible that Anarchic Hand represents the failure of a
suppression mechanism parallel to that in Utilization Behaviour but one which
normally suppresses internally driven action, that is, our own non-conscious
intentions or urges that are unacceptable or inopportune. While suppression of
internally and externally driven actions plausibly overlap, their non-coextensiveness
in the two syndromes suggests some separation.
2 I would like to thank John Hodges for drawing this patient to my attention and arranging for me
to see him.
The Sense ofAgency 79
The patient I saw, an alert man in his seventies, said that he was not doing the
anarchic actions. He quickly followed this by adding that 'of course I know that I
am doing it. It just doesn't feel like me.' When I asked him if he meant that he did
not intend to do those actions he replied that of course he didn't, but that wasn't
the reason that they didn't feel like him doing them. When I asked him what was the
reason, he responded that he was unable to say ('I can't put my finger on it.'). I men-
tion this for two reasons. First, the disownership does not have the characteristic of
conviction seen in schizophrenia, for example, when the schizophrenic claims that
someone else is inserting thoughts. The person with Anarchic Hand is not neces-
sarily confused at the time and is often clear that their experience of the action as
disowned is a 'seeming'. Second, the crucial phenomenological feature was in this
case ineffable. One might be reluctant to rely on the phenon1enological reports of
disownership of action by patients with Anarchic Hand, since they have frontal
lesions. However, as Della Sala (2000) has recently pointed out, there is a consistency
between such patients and within each in disowning anarchic actions. Whatever
their other deficits, this seems to be a faithful report of what it's like for them.
What I wish to draw from Anarchic Hand at present is that it is possible for
what we would call an action to be experienced as another's (or not one's own)
and that, in the adequately recorded cases, when it occurs, awareness from the
inside of relevant intention, effort, and will are lacking. When actions are thus
experienced as another's, either some feature specifying alienness of ownership
overrides whatever specifies self-ownership or the latter is lacking. The fact that
there is a lesion certainly makes it plausible that something is deficient or lacking.
Further, the relevant common lesion site as far as we know is the Supplementary
Motor Area. Let me also say that although the phenomena of generation and con-
trol of anarchic actions are of clear interest, I am mainly concerned at present with
the phenomenology. Finally, these phenomena should not be confused with what
Glyn Humphreys and Jane Riddoch describe in this volume (Chapter 9). That is,
although the actions of the patient they describe deviate in detail from the appar-
ent intention, they are not unrelated to it and they are often performed with the
intended limb. These are both untypical of Anarchic Hand.
Is disownership of action, in Anarchic Hand or otherwise, due to a lack of aware-
ness or acknowledgement of a reason or of an urge for the action? rrhe answer must
be negative, because there are cases where such lack does not lead to disownership
of action. I return to the movements induced by Hecaen et al. (1949) by stimula-
tion of the central thalamic nucleus. Although their subjects had no idea why they
had made the movements, which were described by them as actions, nor felt an
urge to do them, they never con1n1ented that the actions were not their own and
there is no reason to suppose that the actions had an alien character.
In Tourette's Syndrome and in Obsessive Compulsive Disorder (OCD), people
feel an urge to carry out an action. In Tourette's Syndrome, the actions often seem
meaningless and are often referred to as tics but are certainly goal-directed. In
80 Anthony Marcel
OCD, the actions sometimes are or seem meaningful. In both cases, the sufferer
can withhold or suppress the actions for some time, and to that extent they are
under the person's control; and when they 'allow them to pour out' they feel tem-
porary relief. In such cases, the urge is felt as alien and coming from elsewhere. But
the action itself is owned. Indeed, both in Tourette's Syndrome and in OCD, the
individual feels that the action has to be performed to some criterion and if it fails
this (which it usually does) he or she feels compelled to repeat it adequately a cer-
tain pre-ordained number of times. The person feels responsible for the execution
of the action, and yet it is as though some external authority demands it. The indi-
vidual usually feels passive, not about the action itself, but about the motivation
for it. To some extent this is also true of schizophrenia. That is, when schizo-
phrenics assert that their thoughts are not their own, they do not mean that they
themselves are not having the thoughts, but rather that someone else has inserted
them and that they themselves are not responsible for generating them. (The
attribution in schizophrenics of heard voices to others who are telling them to do
things is another matter. The extent to which schizophrenics produce actions that
they deny that they themselves are producing, as opposed to asserting they are
doing so under another's commands, is unclear.)3 The point of these cases is that
they illustrate a double dissociation between disownership of action and disown-
ership of intention or of the cause of an action. Disownership of the latter does
not lead inevitably to disownership of the former. For this reason, the disowner-
ship of action in Anarchic Hand cannot simply be attributed to unawareness of
intention, and it may be treated as a 'pure' case of disownership of action.
I have gone to such lengths here because this is important in addressing the
phenomenology of the sense of agency, in order to clarify the prerequisites for dis-
ownership of action, and because to my knowledge no one has as yet put together
the phenomena discussed here to separate out the issues. The data on awareness
and felt ownership of action indicate that they can come apart. Does this mean
that accounts of each should be distinct? I can now proceed to examine theoretic-
ally the phenomenology of ownership in relation to the experience of agency.
Before proceeding, one point is worth making about the implication of
Anarchic Hand. This concerns one of the ways that self knowledge has been
thought to be privileged that I raised in Section 1.3. If one is aware through inter-
nal proprioceptive awareness of an action, of a posture, or of a sensation, one
might think that it is impossible to be mistaken about whose it is. By 'mistaken' I
mean mistaken about whose it seems phenomenally to be, rather than mistaken in
logical or rational reflection about whose it must be. Also, the contrast here is with
external visual or haptic perception of an action or posture. Indeed, this supposi-
tion has been raised to axiomatic status by some authors (Shoemaker, 1968;
Evans, 1982) and termed 'immunity to error through misidentification relative to
REFERENCES
ABRAMSON, L. Y., SELIGMAN, M. E. P., and TEASDALE, J. D. (1978), 'Learned helplessness in
humans: critique and reformulation', Journal ofAbnormal Psychology, 87: 49-74.
The Sense ofAgency 91
BARNARD, P. J., and MAY, J. (1999), 'Representing cognitive activity in complex tasks',
Human-Computer Interaction, 14: 93-158.
BETTELHEIM, B. (1961), The Informed Heart. London: Thames and Hudson.
BISIACH, E., and GEMINIANI, G. (1991), 'Anosognosia related to hemiplegia and
hemianopia', in G. P. Prigatano and D. L. Schacter (eds.), Awareness ofDeficit after Brain
Injury. New York: Oxford University Press.
BORGES, J. L. (1970), 'Borges and 1', in Labyrinths: Selected Stories and Other Writings.
London: Penguin Books.
BREWER, B. (1993), 'The Integration of spatial vision and action', in N. Eilan, R. McCarthy,
and B. Brewer (eds.), Spatial Representation. Oxford: Blackwell, 294-316.
BRIDGEMAN, B. (1989), 'Separate visual representations for perception and for visually
guided behavior', in S. Ellis, M. Kayser, and A. Grunwald (eds.), Spatial Instruments and
Spatial Displays. MountainView, Calif.: NASA, 40.1-40.16.
BUTTERWORTH, B. (1981), 'Speech errors: old data in search of new theories', Linguistics,
19: 627-62.
COLE, J. (1993), Pride and a Daily Marathon. Cambridge, Mass.: MIT Press.
--and PAILLARD, J. (1995), 'Living without touch and peripheral information about
body position and movement', in J. Bermudez, A. Marcel, and N. Eilan (eds.), The Body
and the Self. Cambridge, Mass.: MIT Press.
CSIKSZENTMIHALYI, M. (1978), 'Attention and the holistic approach to behavior', in
K. S. Pope and J. L. Singer (eds.), The Stream of Consciousness. New York: Plenum Press.
DELLA SALA, S. (2000), 'Clinical aspects of Anarchic Hand: disowned actions'. Paper
presented at 16th European Workshop on Cognitive Neuropsychology, Bressanone, Italy.
- - MARCHETTI, C., and SPINNLER, H. (1991), 'Right-sided anarchic (alien) hand:
a longitudinal study', Neuropsychologia, 29: 1113-27.
------(1994), 'The anarchic hand: a fronto-mesial sign', in F. Boller and
J. Grafman (eds.), Handbook ofNeuropsychology, ix. Amsterdam: Elsevier, North-Holland.
DENNETT, D. C. (1991), Consciousness Explained. London: Little, Brown and Co.
EILAN, N., MARCEL, A., and BERMUDEZ, J. (1995), 'Self-consciousness and the body', in
J. Bermudez, A. Marcel, and N. Eilan (eds.), The Body and the Self. Cambridge, Mass.:
MIT Press.
EVANS, G. (1982), The Varieties ofReference. Oxford: Oxford University Press.
FOURNERET, P., and JEANNEROD, M. (1998), 'Limited conscious monitoring of motor
performance in normal subjects', Neuropsychologia, 36: 1133-40.
FREUD, S. (1914/1975), The Psychopathology ofEveryday Life. Harmondsworth: Penguin.
FRIED, 1., KATZ, A., MCCARTHY, G., SASS, K. J., WILLIAMSON, P., SPENCER, S. S., and
SPENCER, D. D. (1991), 'Functional organization of human supplementary motor cortex
studied by electrical stimulation', Journal ofNeuroscience, 11: 3656-66.
FRITH, C. D. (1992), The Cognitive Neuropsychology ofSchizophrenia. Hove: Taylor and Francis.
GALLAGHER, S., and MARCEL, A. J. (1999), 'The self in contextualized action', Journal of
Consciousness Studies, 6 (4): 4-30.
GIBSON, J. J. (1979), The Ecological Approach to Visual Perception. Boston: Houghton Mifflin.
GOODWIN, G. M., MCCLOSKEY, D. 1., and MATTHEWS, P. B. C. (1972), 'The contribution of
muscle afferents to kinaesthesia shown by vibration induced illusions of movement and
by the effects of paralysing joint afferents', Brain, 95: 705-48.
92 Anthony Marcel
HAGBARTH, K. E., and EKLUND, G. (1966), 'Motor effects of vibratory muscle stimuli in
man', in R. Granit (ed.), Muscular Afferents and Motor Control. Stockholm: Almqvist &
Wiksell,177-86.
HAGGARD, P., and EIMER, M. (1999), 'On the relation between brain potentials and the
awareness of voluntary movements', Experimental Brain Research, 126: 128-33.
--and MAGNO, E. (1999), 'Localising awareness of action with transcranial magnetic
stimulation', Experimental Brain Research, 127: 102-7.
HECAEN, H., TALAIRACH, J., DAVID, M., and DELL, M. B. (1949), 'Coagulations limitees
du thalamus dans les algies du syndrome thalamique: resultats therapeutiques et
physiologiques', Revue de Neurologie, 81: 917-31.
HEIDER, E, and SIMMEL, M. (1944), 'An experimental study of apparent behavior',
American Journal of Psychology, 57: 243-59.
HERRIGLE, E. (1988), Zen in the Art ofArchery. London: Penguin Books.
HUMPHREY, N. (1983), Consciousness Regained. Oxford: Oxford University Press.
--(1992), A History of the Mind. New York: Simon and Schuster.
JAMES, W. (1890/1950), The Principles of Psychology, i. 1950 edn., New York: Dover.
JONES, 1. (1988), 'Motor illusions: what do they reveal about proprioception?', Psychological
Bulletin, 103: 72-86.
KELSO, J. A. S. (1995), Dynamic Patterns: the Self-Organization of Brain and Behavior.
London: MIT Press.
KIRSCH, 1., and LYNN, S. J. (1999), 'Automaticity in clinical psychology', American
Psychologist, 54: 504-15.
LACKNER, J. R, and TAUBLIEB, A. B. (1983), 'Reciprocal interactions between the position
sense representations of the two forearms', Journal of Neuroscience, 3: 2280-5.
LAMBIE, J. A., and MARCEL, A. J. (2002), 'Consciousness and Emotion Experience:
A Theoretical Framework', Psychological Review, 109: 219-59.
LEVINE, D. N. (1990), 'Unawareness of visual and sensorimotor defects: a hypothesis', Brain
and Cognition, 13: 233-81.
LHERMITTE, E (1983), '''Utilization Behaviour" and its relation to lesions of the frontal
cortex', Brain, 106: 237-55.
LIBET, B. (1993), 'The neural time factor in conscious and unconscious events', in
G. R Bock and J. Marsh (eds.), Ciba Symposium 174: Experimental and Theoretical
Studies of Consciousness. Chichester: John Wiley.
- - GLEASON, C. A., WRIGHT, E. W., and PEARL, D. K. (1983), 'Time of conscious
intention to act in relation to onset of cerebral activity (readiness potential): the uncon-
scious initiation of a freely voluntary act', Brain, 106: 623-42.
MCCORMACK, T., and HOERL, C. (1999), 'Memory and temporal perspective: the role of
temporal frameworks in memory development', Developmental Review, 19: 145-82.
MARCEL, A. J. (1993), 'Slippage in the unity of consciousness', in G. R. Bock and J. Marsh
(eds.), Ciba Symposium No 174: Experimental and Theoretical Studies of Consciousness.
Chichester: John Wiley, 168-86.
- - (1998), 'Blindsight and shape perception: deficit of visual consciousness or of visual
function?', Brain, 121: 1565-88.
--and TEGNER, R (1995), 'Anosognosia for plegia'. Paper presented at lith European
Workshop on Cognitive Neuropsychology, Bressanone, Italy.
The Sense ofAgency 93
_---and NIMMO-SMITH, 1. (forthcoming), 'Anosognosia for plegia: specificity, exten-
sion, partiality and disunity of bodily unawareness', Cortex.
MERINGER, R., and MAYER, C. (1895/1978), Versprechen und verlesen: eine psychologisch-
linguistische Studie. Reprinted with introduction by A. Cutler and D. Fay. Amsterdam:
John Benjamins.
MERLEAU-PONTY, M. (1962), Phenomenology of Perception, trans. Colin Smith. London:
Routledge.
O'SHAUGHNESSY, B. (1980), The Will, ii. Cambridge: Cambridge University Press.
PAILLARD, J. (1991), 'Motor and representational framing of space', in J. Paillard (ed.), Brain
and Space. Oxford: Oxford University Press.
POLANYI, M. (1966), The Tacit Dimension. Garden City, NY: Doubleday.
PRINZ, W. (1990), 'A common coding approach to perception and action', in O. Neumann
and W. Prinz (eds.), Relationships betvveen Perception and Action: Current Approaches.
Berlin: Springer, 167-20l.
RABBITT, P. M. A. (1968), 'Three kinds of error-signalling responses in a serial choice task',
Quarterly .Journal ofExperimental Psychology, 20: 179-88.
REASON, J. (1990), Human Error. Cambridge: Cambridge University Press.
REED, N., McLEOD, P. D., and DIENES, Z. (2000), 'In1plicit learning and motor skills: what
people who know how to catch a ball do not know'. Unpublished paper.
RUSSELL, J. (1996), Agency: Its Role in Mental Development. Hove, UK: Erlbaum.
SHALLICE. T. (1972), 'The Ulster depth interrogation techniques and their relation to
sensory deprivation research', Cognition, 1: 385-405.
SHOEMAKER, S. (1968), 'Self-reference and self-awareness', Journal ofPhilosophy, 65: 555-67.
TURVEY, M. T. (1977), 'Prelin1inaries to a theory of action with reference to vision', in
R. Shaw and J. Bransford (eds.), Perceiving, Acting and Knowing: Toward an Ecological
Psychology. Hillsdale, NJ: Erlbaum.
VON HOLST, E., and MITTELSTAEDT, If. (1950), 'Das Reafferenzprinzip. Wechselwirkungen
zwischen Zentralnervensystem und Peripherie', Naturwissenschaften, 37: 464-76.
WATSON, J. S., and RAMEY, C. T. (1987), 'Reactions to response-contingent stimulation in
early infancy', in J. Oates and S. Sheldon (eds.), Cognitive Development in Infancy.
Hillsdale, NJ: Lawrence Erlbaum.
3
Action: Awareness, Ownership,
and Knowledge
Christopher Peacocke
Tony Marcel's chapter 'The Sense of Agency: Ownership and Awareness of Action'
is a rich and original contribution to our philosophical and psychological under-
standing of awareness of action. Even experienced enthusiasts are likely to find
Tony's chapter to be very closely packed with ideas and proposals. So it will help
to structure the discussion if we extract some of its main theses here:
1. There is a way of being aware of one's own actions that is distinct from, and
can exist in the absence of, both one's own observation and proprioception
of those actions. This special way of being aware of one's own actions is
naturally described as being aware of them 'from the inside', and it involves
in normal cases a species of first-person awareness that one is doing some-
thing. (Special Awareness)
2. It can seem to an agent from the inside that he has made a bodily movement
in one direction, even though he has in fact moved in a different direction,
and even though his proprioceptive mechanisms are functioning. The
experience of the subjects in Tony's own vibro-tactile experiments exemplify
this possibility. (Independence of Movement)
3. The initial subjective awareness of a bodily action from the inside precedes
the occurrence of the bodily movement. It 'appears to be underlain by the
specification of the movement' (p. 71), and this specification involves the
states of the Supplementary Motor Area. (Movement Specification)
4. Neither intention nor effort is necessary for the distinctive phenomenology
of action.
5. In the phenomenon of Anarchic Hand, the subject experiences an action as
being not his own.
6. Motor specifications must be given in egocentric coordinates, and this use of
an egocentric frame, or set of interrelated frames, in action gives a 'minimal
(and physical) sense of ownership of action' (p. 85). Tony adds that he
supposes that 'the perspectival content of action specifications is reflected in
My thanks to Naomi Eilan and Johannes Roessler for valuable advice, both substantive and editorial,
and to Tony Marcel for stimulating discussions at the Warwick conference in 1998. The present com-
ments are more extensive than those I delivered at the meeting.
Awareness, Ownership, and Knowledge 95
the corresponding phenomenology that one experiences as action itself'
(p. 85). (Minimal Ownership as Egocentric Origin)
7. The occurrent sense of agency that is present in the phenomenology of
particular action relies on longer-term reliable causal effectiveness, practical
knowledge, and long-term sense of agency. (Longer-Term Dependence)
The Thesis of Special Awareness (1) and the Thesis of Independence of
Movement (2) seem to me true, important, and well-established by what Tony and
others have said. The Thesis of Movement Specification (3) is also made plausible
by his discussion and the literature cited. The description of Anarchic Hand in
(5) also appears to be a datum. My comments here will fall into three parts.
(i) I will argue that the significance of the phenomenon of Anarchic Hand
and of the striking results in Tony's own vibro-tactile experiments is both
wider than, and also rather different from, that given in Tony's chapter.
Reflection on the whole set of cases suggests a different explanation of the
phenomenology of agency. Such reflection also prompts doubts about
Tony's thesis of Minimal Ownership as Egocentric Origin (6).
(ii) I will pursue an issue about the relations between the phenomenology of
action and of ownership that are raised by, but not addressed in, Tony's
discussion. When a subject is aware of an action from the inside, must he
also be aware of it as his own, and if so, why?
(iii) In a final section, I turn to the very sharp epistemological question raised
by Tony's work, and to which he acutely alludes at one point: if Tony's
principal claims are right, what is the correct philosophical account of
how we know about our own actions?l
2 S. Della Sala, 'Clinical aspects of Anarchic Hand: disowned actions', paper presented at
3 N. Reed, P. McLeod, and Z. Dienes, 'Implicit learning and motor skills: what people who know
4 In the case of some non-human animals, the first-person notion may be some non-conceptual
analogue of the first-person concept. On non-conceptual forms of the first person, see S. Hurley,
Consciousness in Action (Cambridge, Mass.: Harvard University Press, 1998) and my Postscript 'The
relations between conceptual and nonconceptual content', in Essays on Non-Conceptual Content, ed.
Y. Gunther (Cambridge, Mass.: MIT Press, 2002).
104 Christopher Peacocke
from the inside is not merely one whose correctness condition implies that he
himself is the agent of the ¢-ing in question; further, the experience represents
this as being so, as part of its representational content.
In case this explanation seems ad hoc, it should be noted that there is a parallel
datum about perception, and a broadly corresponding explanation of the parallel
datum. The datum in the perceptual case also involves the first person. It too links
two features of awareness. Consider any perceptual experience as of a presented
object standing in a given spatial relation R to the body that provides the egocen-
tric frame of reference for the perception, in a case in which this experience is
enjoyed by someone who possesses the first-person notion. It is equally apparently
a datum that such an experience is one in which it seems to the subject that the
presented object stands in relation R to himself-a first-person content.
Again, three facts seem to explain this: a Constitutive Fact; a fact about percep-
tual experience; and a corresponding and parallel further hypothesis. The
Constitutive Fact is that for any given thinker's first-person judgement 'That F
(perceptual demonstrative) bears R to here' to be correct, the referent of the
perceptual-demonstrative must bear R to the egocentric origin fixed by the subject's
body (or relevant limb). The fact about perceptual experience is that it has repre-
sentational content, in the sense discussed above. These first two facts in the per-
ceptual case imply the representational content of a subject's experience in which
an object is represented as bearing R to the egocentric origin is correct (and for
constitutive reasons) only if that object bears R to the subject himself. The further
hypothesis is once again that this in1plied content is absorbed into the represen-
tational content of the experience itself: the experience itself represents the object
as bearing R to he himself (a first-person content).
Both in the case of the phenomenology of action from the inside, and the con-
tent of perceptual experience, the representational content can be incorrect. This
is possible even when the perception supplies genuine information about
the world somewhere or other, just not around oneself. Examples of the kind of
case in which a subject is perceiving the environment which is in fact around
someone else have been elaborated by writers as diverse as H. G. Wells in a short
story, Sydney Shoemaker in discussions of personal identity, and Daniel Dennett
in 'Where am I?'S Although in these examples, the world around the subject is
not as it is represented as being in the subject's perceptual experience, it remains
the case that the misleading experience represents objects and events as being
thus-and-so in relation to him. That is a necessary condition of the experiences'
being misleading about him.
5 H. G. Wells, 'The Remarkable Case of Davidson's Eyes', in The Complete Short Stories of H. G. Wells,
ed. J. Hammond (London: Orion, 1998); S. Shoemaker, 'Persons and their pasts', American
Philosophical Quarterly (1970), 7: 269-85; 'Where am I?', in D. Dennett, Brainstorms (Montgomery, Vt.:
Bradford Books, 1978).
Awareness, Ownership, and Knowledge 105
In the case of action, here is a way in which a subject can be aware of an action
from the inside, the action seems to be his own, and yet it is not in fact his own.
We can imagine Siamese twins joined at the head, sharing a common motor
specification area. This common area can receive information and instructions
from their two separate centres that sustain numerically and possibly type-
distinct beliefs, desires, intentions, emotions, and so forth. But when either of
them succeeds in trying to do something, each of them has the distinctive kind of
awareness from the inside, generated by activity in the common motor specifica-
tion area, that he himself is doing it. Aside from the case of overdetermination,
only one of these impressions will be correct.
6 For one statement, see 1. Bonjour, The Structure of Empirical Knowledge (Cambridge, Mass.:
some preliminary discussion in my piece 'The past, necessity, externalism and entitlement', Philosophical
Books (2001),42: 106-17, and in my 'Explaining Perceptual Entitlement', forthcoming in Philosophy and
Phenomenological Research (Proceedings of the Rutgers Epistemology Conference 2003).
Awareness, Ownership, and Knowledge 107
account-is that thinkers are entitled to presume, in the absence of explanation to
the contrary, that they are in the circumstances with respect to which the obser-
vational intentional contents of their perceptual states are individuated. What
makes it the case that a perceptual state has a certain spatial and temporal content,
for instance, is that in the basic case in which the thinker is properly embedded in
the world, the occurrence of perceptions with these spatial and temporal contents
will be explained by the very spatial and temporal conditions they represent as
holding. These observational intentional contents are externally individuated, in
ways familiar from discussions of externalism in the theory of content. S Their
external individuation, a constitutive fact about their nature, contributes, on this
view, to their role in entitlement and epistemology. There is evidently a huge
amount more to be said on these matters; but this is enough for an initial paral-
lelism with entitlement in the case of action.
The deeper parallel with the action case starts from the parallel constitutive
point that what, for instance, makes something a trying to move one's hand is that
it is an event of a kind which, when the subject's states are properly embedded in
his body and the world, causes his hand to move. This point holds for a range of
kinds of tryings which are a subset of those which are basic actions for the agent
in question. For non-basic action types, such as turning on the computer, what
makes something a trying to do that is that, for example, the person is trying to
push a certain button, and believes this will turn the computer on, this being part
of the subject's plan.
Equally in parallel with the perceptual case, subjects are entitled, in the absence
of evidence to the contrary, to presume that they are in the circumstances with
respect to which the intentional content of their states is individuated. But for
types of action which are basic for a given subject, these will be circumstances in
which trying to perform an action of one of these kinds will result in (or be part
of) an action of that kind. So when a thinker has a distinctive awareness from the
inside of trying to ¢, where ¢-ing is basic for him, he is entitled to judge that
he is ¢-ing. In this way, experiences of agency from the inside can, in suitable
circumstances, lead to knowledge. This parallel with the perceptual case holds,
despite differences in the direction of causation in the two cases.
Knowledge of action in non-basic cases must involve entitlement of a some-
what different kind. This seems to me to parallel the role of non-observational
concepts in the representational content of perceptual experience. There is such a
thing as seeing an object as a Macintosh computer, seeing a huge installation as a
particle accelerator, and perceiving an event as an H-bomb explosion. Such experi-
ences can lead to knowledge; but the account of the nature of the entitlement is, I
suggest, more complex and certainly different from that of the observational
8 T. Burge, 'Individualism and Psychology', Philosophical Review (1986), 85: 3-45; C. Peacocke,
'Externalist explanation', Proceedings of the Aristotelian Society (1993), 93: 203-30.
108 Christopher Peacocke
cases. It must involve the subject's possessing the knowledge that things that
fall under observational properties also have certain theoretical properties, or
properties not given in the relevant perception itself. There is no such role for
knowledge in the observational cases themselves. Corresponding points apply
pari passu to such non-based action cases as that of a subject's knowledge that he
has turned on the computer, on the basis of the awareness from the inside of
having done so.
The next major task on the agenda for future work would be to move to further
explanatory depth in the action case, to address the question of why we are
entitled to presume that we are in the circumstances with respect to which the
intentional content of perception and action is individuated. These comments
are not the place for that; but I conjecture that carrying out that task will reveal,
as reflection on Tony's paper has also revealed, that the theory of entitlement and
knowledge on the one hand, and the theory of the nature of the contents of
conscious states involved in action on the other, can illuminate one another.
APPENDIX
This appendix discusses some issues raised by Tony's paper that are ancillary to the argu-
ment above, but are still of independent interest.
9 H. Hecaen, J. Talairach, M. David, and M. Dell, 'Coagulations limitees du thalamus dans les algies
10 The following formulations are those given in G. Evans, The Varieties of Reference (Oxford:
Patrick Haggard
1. INTRODUCTION
'That's funny: thought Pooh. 'I said «Ow!" without really oo'ing:
'That's me again,' thought Pooh. 'I've had an Accident, and fallen down a well,
and my voice has gone all squeaky and works before I'm ready for it, because
I've done something to myself inside. Bother!'
'Help-help!'
'There you are! I say things when I'm not trying. So it must be a very bad
Accident.'
(from A. A. Milne, The House at Pooh Corner)
2. AWARENESS OF INTENTION
Libet et al. (1983) is the classical study in this field. Libet et al. developed an experi-
mental method which combined three important elements traditionally found
within separate scientific strands. First, they asked people to provide subjective
judgements about their own intentions and actions. As such, they were admitting
the phenomenology of conscious states as an area of valid scientific enquiry, in
contrast with many other researchers in this area. Second, they studied the timing
of these conscious phenomena. This allows a chronometric approach to intention.
This chronometric approach opens the way to the carefully designed experiments
of the experimental psychology tradition. Third, they understood the crucial
importance of the relation between phenomenology and underlying brain activity,
in that they combined subjective report with objective physiological recordings of
brain activation during intentional action. These bold innovations have made
Libet et al.'s work the starting-point for any discussion of the relationship between
conscious state and intentional action.
Libet et al.'s (1983) study was as follows. Subjects sat watching a rotating spot
on an oscilloscope, which resembled the hand of a clock. The clock hand rotated
once every 2,560 ms. The clock is intended to provide an independent external
metric against which subjects can judge the time of occurrence of the various
internal phenomenological states occurring in intentional action. At a time of
their own choosing, the subjects made freely willed voluntary movements of the
right hand. The clock continued to rotate for a random interval of a number of
seconds, and then stopped. Once the clock had stopped, the subject reported the
position of the clock at which they had experienced specific phenomenological
events. In Libet et al.'s original paper (1983), subjects made three subjective judge-
ments, according to condition, to report on three specific phenomenological
events. First, in a control condition, subjects reported the perceived time of an
electrical somatosensory stimulus delivered to the skin of the right hand. By com-
paring the clock times reported by subjects with the actual time at which the
114 Patrick Haggard
experimenter delivered the stimulus, Libet et al. calculated that subjects perceived
this somatic stimulus as occurring some 32 ms on average before it was actually \
delivered. Libet et al. proposed that this value could be used as a control condition 1
to indicate the extent to which the perceived time of events using their method \
differed from the time of their actual occurrence.
In a second condition, Libet et al. asked subjects to make freely willed inten-
tional actions at a time of their own choosing, and to report the position of the clock
hand when they actually initiated the action. The perceived time of this motor
event, which Libet et al. termed M (movement), was compared to the actual onset
of muscle activity, by recording the electromyogram (EMG) of the hand muscles.
Libet and colleagues found that subjects thought they moved on average 86 ms
before the onset of muscle activity. On the face of it, this result appears to suggest
that an awareness of action cannot arise from any peripheral sensations telling us
that movement is in fact occurring. Instead, the findings point towards a central
source for the awareness of action, within the brain structures that prepare and
execute intentional actions.
Finally, Libet et al. studied a third condition in which they asked subjects to use
the clock hand to report when they first felt the urge to move. Libet et al. termed
this W (will) judgement. They assumed that W reflected the moment of conscious
willing of the subsequent action. On average, Libet et al.'s subjects reported
awareness of the will to move some 200 ms before the first muscle activity. Since
Libet et al. concurrently recorded the brain activity involved in the neural pre-
paration of movement, using scalp electrodes positioned over the motor cortical
areas, they were able to relate the timing of the subjective experience of will to the
timing of neural preparation of movement.
The robust finding of previous studies on the neural preparation for inten-
tional action has been a gradual ramp-like increase in frontal cortical activity,
beginning one second before movement onset. This has been termed the readiness
potential. Libet et al. observed similar readiness potentials in their subjects, begin-
ning some 900 to 700ms before movement onset. Intriguingly, therefore, the neural
preparation of intentional action began in these subjects some 500 ms at least
before the subjects themselves reported awareness of the intention to move. Libet
et al. inferred that (1) intentional action begins with an unconscious set of neural
events in the brain, and (2) intentions only reach conscious awareness when these
neural processes have progressed to a later stage in the chain from intention into
action. Libet expressed his view of this apparent paradox well in the subtitle of his
paper: 'the unconscious initiation of a freely voluntary act'. A tabloid paraphrase
of his result might be 'your brain knows you are going to move before you do'.
Libet's work has had considerable influence. Unfortunately, this influence has
divided into two strands of controversy, neither of which, with the benefit of a few
decades' hindsight, seems to have been particularly productive. A good synopsis
of reaction to Libet's finding comes in the commentaries on a 1985 target article
by Libet (Libet, 1985).
Action Awareness 115
On the one hand, psychologists have tended to be sceptical about Libet's work
on technical grounds. Thus, for example, the method involves comparison of tim-
ings across two perceptual streams (the external clock and the subject's internal
motoric events). People's ability to make such cross-modal judgements is known
to be poor. Moreover, cross-modal judgements are highly dependent on which
particular stream they may be attending to at a given time (the so-called 'prior
entry' phenomenon). Further, empirical psychologists have doubted that subjects'
verbal reports of states such as W really reflect a phenomenology with a strong
and distinctive conscious content. In essence, these critics claim that subjects may
not really be judging any mental entity at all when they give a W judgement.
Instead, they may be making some inferences on the basis of the (admittedly
vague) instructions they were given, and their folk psychological beliefs about
'free will', with the aim of producing plausible judgements about events which
they do not normally judge in the course of their mental life.
Libet et al.'s paper also had substantial impact within the philosophical world.
There, the main interest has been in whether Libet's data are compatible with the
traditional Cartesian assumption of conscious free will. Libet himself believed
they were not: consciousness of willing in his experiment was a by-product of the
brain processes of movement initiation. Thus, the conscious will did not cause the
brain preparation, but the other way around.
Libet's data do not, however, clarify what brain events may underlie the con-
scious awareness of will on which the W judgement is based. His data only show
that some preparatory brain processes precede Wawareness. Temporal precedence
is a poor argument for causation, and is also a poor guide to the content of con-
scious states. Thus, it remains unclear what the W judgement is a judgement of
We have addressed this issue in a recent experiment extending Libet's work.
This experiment has been described in detail elsewhere (Haggard and Eimer,
1999) and will be summarized here. Subjects made voluntary button press move-
ments while watching a rotating clock similar to that used in Libet's original
experiment. In our study, subjects either freely chose on each trial whether to
depress a left-hand or a right-hand button, or moved with the same hand
throughout an experimental block, according to condition. Since no reliable
differences between fixed and free choice actions were found, either in the psy-
chophysical judgements, or in physiological measurements of brain activity, this
difference will not be discussed further. Cortical activity was recorded continu-
ously from electrodes positioned in standard locations, and was processed using
conventional techniques to calculate the average brain activity during the prepara-
-tion of voluntary action. Our analysis of the physiological signals concentrated on
two well-characterized potentials related to movement preparation. First, we cal-
culated the bereitschaftspotentia~ or readiness potential, using methods comparable
to Libet's. The bereitschaftspotential is maximal over the vertex of the scalp, and
probably reflects post-synaptic potentials in medial pre-motor areas of both
hemispheres, and in the motor cortices themselves. We also measured the lateralized
116 Patrick Haggard
readiness potential (LRP). This represents the specific activity in the motor cortex
actually responsible for the impending movement. For example, when the subject
moves their right hand, the specific signals which drive the muscles arise largely
from the left motor cortex, while the right motor cortex remains relatively inactive.
Subtracting the electrical activity recorded over the inactive right motor cortex ,
from the electrical activity recorded over the active left motor cortex gives a measure '
of the specific brain activity associated with the subject making that particular
movement, rather than a movement with the other hand.
The robust finding of bereitschaftspotential studies has been that the early com-
ponents of the bereitschaftspotential are symmetric. That is, the earliest stages of
preparation for intentional action involve both hemispheres of the brain equally.
This bilateral activity typically appears one to two seconds before the onset of
movement itself, with substantial variations from one task to another. Around
500 ms before movement onset, the bereitschaftspotential begins to lateralize, and
the activity in the cortex contralateral to the hand which will move begins to
exceed that in the cortex ipsilateral to the movement. This event corresponds to LRP
onset. LRP onsets are psychologically significant. They represent the point at
which the motor processing leading to movement shifts from the abstract stage
of intending an action (e.g. pressing a button) to the more specific concrete stage
of initiating one particular movement (pressing this button, as opposed to that
one) as the means of achieving the action goal.
We wanted to find a stronger argument for associating conscious awareness of
intention with a particular brain event than the reliance on a single numerical
estimate used in Libet's work. Therefore, we used a version of Mill's argument
from concomitant variation. Mill (1843) proposed this argument as one method
for identifying causal relations. It states that variations in a cause should produce
concomitant variations in an effect. We used random trial-to-trial variation in the
timing of brain activity and psychophysical judgement to investigate the relations
between them. The ideal approach, following Mill's method directly, would
involve correlating the time of bereitschaftspotential onset, for example, with
the time of Wawareness. Since Libet holds that the bereitschaftspotential causes W
awareness, this correlation should be positive. Unfortunately, this direct approach
is not practical because bereitschaftspotentials cannot easily be measured on single
trial, but need to be averaged over several trials to produce a reliable signal. W
judgements, however, are made on a single trial basis, and can be used in an
inverted form of Mill's argument. Thus, on Libet's view, trials with early Wawareness
should also have early bereitschaftspotential onsets, while trials with late Waware-
ness should have a later bereitschaftspotential onset. Because bereitschaftspotentials
must be studied as averages, rather than as single trials, we investigated this
covariation by splitting each subject's W judgements at their median value. This
classified the trials into those with Wawareness occurring long before movement
onset (early W) and those where W awareness preceded movement onset by a
Action Awareness 117
Readiness potential (RP) at Cz
Median split on W awareness values
- lO IlV
Q)
"0
.~
a.
E
'"
c..
c:
shorter time (late W). We separately averaged bereitschaftspotentials for these early
and late Wtrials. The results are shown in Figure 4.1.
On Libet's view, the bereitschaftspotential for early W trials should lead that for
late Wtrials during the critical pre-movement period. In fact, we found a difference
in the opposite direction, though this did not achieve statistical significance. The
failure to find covariation between the timing of bereitschaftspotential and of W
awareness appears to rule out Libet's argument for a causal relation between them.
We next repeated the same covariation analysis with lateralized readiness
potential (LRPs). The results are shown in Figure 4.2.
In the LRP traces, a downward deflection indicates an increased activity of the
cortex contralateral to the moving hand. Notice the relatively flat initial portion
of the trace, corresponding to the early phases of bereitschaftspotential in which
cortical activity is bilaterally symmetric. During the critical period of the deflec-
tion of the LRP, when cortical activity related to the specific movement is develop-
ing, there is a slight lead of the LRP for early W trials over that for late W trials
(hatched area in Fig. 4.2). To characterize this difference statistically, we fitted two
piecewise regressions to the pre-movement portion of the LRP, corresponding to
the initial flat portion and the downward deflection respectively. We iteratively
searched for the best joining-point of these two phases of the potential, by identi-
fying the joining-point between the two regressions which best fitted the data.
This joining-point can be taken as an estimate of LRP onset. The average LRP
118 Patrick Haggard
Lateralized readiness potential (LRP)
Median split on W awareness values
Wearly 1
-2000 -1500 -1000 -500
w
"0
.~
a.
E
'"
Cl..
""
--l
onset times and W judgement times are indicated in the figure. LRP onsets for
early W trials were significantly earlier than those for late W trials. This finding is
evidence for a concomitant variation between these events.
The significant covariation between Wawareness and LRP onset is consistent
with the hypothesis that the LRP causes W awareness, though of course it does not
prove this hypothesis. Because the LRP is a well-characterized potential, whose
physiological and behavioural features are known, the relation between LRP and
Wawareness can also clarify what Wawareness is an awareness of Our findings
suggest that awareness of intention is tied not to the general aspects of action
preparation, but to the development of a specific intention to execute an action
using a particular movement (in our case a left or right button press). That is, we
are conscious of the intention to act in a particular way. This view places con-
sciousness of intention much closer to the detailed pattern of motor execution
than some other accounts. Wawareness thus looks rather like intention in action,
and much less like prior intention (Searle, 1983).
One interesting feature of this result is the remaining gap of around 500 ms
between average LRP onset, and subjects' average W judgements. First, this gap
appears to rule out the possibility that subjects became aware of their intentions
when the motor system solves the problem of which specific movement to make
Action Awareness 119
to achieve a goal. That hypothesis would predict synchronous LRP onset and W
awareness. The gap could be explained by a threshold-type effect: lateralization of
the readiness potential, and thus the specificity of the movement to be imple-
mented, must reach a particular threshold level for the subject to have the
conscious experience of intending an action.
The covariation of awareness with LRP also suggests a relation between
conscious awareness of intention and movement planning. In the movement plan-
ning literature (see Wolpert et al., 1998 for a review), selecting a specific movement
to achieve a goal is understood to be a hard computational problem. Most of the
action goals I undertake can be achieved by anyone of an infinite set of possible
movements. For example, if I want to switch on the light I could use my left or right
hand to touch the light switch, I could rotate any combination of the shoulder,
elbow, and wrist joints, and so on. An important aspect of movement selection is
the inverse kinematics problem. This refers to the problem of selecting which of the
infinite set of possible joint rotations I should choose in order to move my hand to
a given location in extrapersonal space. There are typically many possible joint
movement patterns compatible with any desired hand movement pattern, so the
inverse kinematics problem is typically ill-posed. Experiments on human reaching
movements (Haggard and Richardson, 1996; Haggard et al., 1995, 1997) have
found evidence for the difficulty of inverse kinematic transformations in the move-
ment patterns of normal subjects. Specifically, movement patterns are more vari-
able under conditions where the inverse kinematics are harder to solve than under
conditions where they are easier. Similarly, robotics research has demonstrated the
difficulties of the inverse kinematics transformation, and often uses biologically
implausible solutions to solve it.
In our experiment above, the specification of which hand to use to press a but-
ton can be seen as a very simple example of a movement selection problem. The
subject's goal on each trial is to press a button, and they have two equally accept-
able motor alternatives with which to achieve it. It is interesting, therefore, that the
conscious awareness of intention appears to be associated with the computation-
ally difficult process of selecting which movement to make. Very speculatively, our
conscious access to the intention and preparation of our movement may guide
and/or verify processes of movement selection. For example, conscious awareness
of intention may bring the impending movement into focal attention, and allow
a movement to be evaluated prior to its actually being executed. On this view,
a conscious state associated with action may be intertwined with the internal
models thought to underlie movement control.
3. AWARENESS OF ACTION
In the quotation that began this chapter, the process of binding intentions to
actions was shown to be important for agency and the self, and was shown to
120 Patrick Haggard
depend on conscious representations of both intention and action. So far, we have
discussed the awareness of intention, and related it to movement specification.
Now, we review some experimental evidence regarding the awareness of action.
Several studies on the perceived time of action have found that people think
they move before they actually do. Thus, Libet et al. (1983) found that their sub-
jects' M judgements (the judgement of when intentional hand movements actu- ,
ally occurred) showed awareness of actions an average 86ms before movement:
onset. Similar results have been reported by McCloskey et al. (1983). This anticip-
atory awareness of action appears to point to a central, rather than peripheral
origin of M awareness. That is, we are aware not of our movements themselves,
but of some pre-motor process associated with them. If we were aware of Our
actual peripheral movements, then our awareness would depend on sensory
information about them. Since all sensory systems have delays in the conduction
of stimuli to the brain, a delayed awareness of movement would be expected on
this view. This anticipatory awareness argument suffers from reliance on a single
absolute numerical value, just as Libet's original argument about W awareness
did. However, some more robust recent evidence also agrees that awareness of
'action' is actually awareness of pre-motor processing.
In a behavioural experiment, Haggard et al. (1999) asked subjects to type
sequences of pre-rehearsed letters on a computer keyboard, beginning at a time of
their own choosing. Libet's method was used to provide an external metric for
subjects to report when they pressed the first key in the sequence. The sequences
could be either one, three, or five letters in length, but the subject's judgement
always related to the moment where the first key in the sequence was pressed.
Sequences of key presses were used in this experiment because of the robust find-
ing that longer sequences require more preparation. This additional preparation
takes time, so reaction time experiments reliably find greater reaction time for
longer sequences than for shorter sequences (Sternberg et al., 1978). Theoretical
models differ in the exact locus of this sequence length effect, but all agree that it
involves some aspect of preparation. The correlation between preparation time
and sequence length makes direct predictions regarding M awareness. If anticipa-
tory awareness indeed reflects preparation, then the anticipatory judgement error
should increase with sequence length.
Our data are summarized in Figure 4.3. Briefly, the perceived time of the first key
press in a sequence became more anticipatory as the sequence became longer. We
draw two conclusions from this finding. First, the awareness of movement does not
relate to the sensory feedback consequences of those movements. Studies on the
perceived time of sensory stimuli, particularly speech stimuli, reliably show an
effect called the P-centre effect (Morton et al., 1976). The original demonstration
of P-centres noted that the perceived time of a speech syllable does not coincide
with its physical onset, but is closer to the middle of the stimulus duration, at a
point presumed to reflect the stimulus's 'centre of perceptual gravity'. A P-centre
Action Awareness 121
,.....
100
g
Q)
E
'p
'"'"Q)
c -100
2:
'"~
...'"
Q)
..Cl
:.:::; -300
3 5
Number of letters in sequence
interpretation of our typing data would suggest that the perceived time of the
movement should occur near the middle of the typing sequence. Were this the case,
awareness of movement would be more delayed for longer movement sequences,
because the interval between movement onset and the hypothesized P-centre is
longer than for shorter sequences. In fact, the opposite result was found. The per-
ceived time of movements becomes more anticipatory as the sequence gets longer,
and as the presumed perceptual centre is delayed. Therefore, the awareness of
movement appears to be less related to the actual motor production than to
preparatory processes.
Second, this result shows that the awareness of movement is generated in the
motor processing system upstream of the locus of the sequence length effect. The
sequence length effect causes an increasing delay between a go signal and movement
onset, as the sequence gets longer. Our result suggests that subjects are not aware
of this increasing delay, since awareness of movement has been formed upstream
of this stage. As a result, the result that the interval between perceived time of
movement and actual onset becomes increasingly negative with sequence length.
This argument does not by itself clarify what processing generated M awareness
since there is no agreed account of the processes responsible for the sequence
length effect. However, on one influential model (Sternberg et al., 1978) the
sequence length effect is attributed to the time taken to retrieve from a holding
buffer the pre-loaded programme for the first movement. On this view, awareness
of movement would be generated after the appropriate movement 'programs' had
been selected for use, but before they were ordered into the appropriate pattern.
Again, the process of movement selection appears important in awareness of
action. In the Sternberg model, selection has two aspects: first, retrieving the
appropriate motor programmes, and, subsequently, accessing the order in which
to use them. Our data suggest movement awareness may be generated between
these two stages. We should add that the Sternberg model is purely functional, and
122 Patrick Haggard
that its stages are not well-localized in the neural sense. Therefore, this experiment
cannot directly clarify the brain processes responsible for the conscious awareness
of movement.
A second experiment, however, did address the issue of neural localization of ;
M awareness. Again, we reasoned that a clearer understanding of movement
awareness could flow from experimental paradigms which selectively influence
a single stage in motor processing, and assess the effects of movement awareness. I
Day et al. (1989) produced a robust and well-localized example of such selective
influence on motor processing stages. They delayed simple reaction time move-
ments by transcranial magnetic stimulation (TMS) over the contralateral primary
motor cortex. Briefly, they delivered brief magnetic pulses around 75 ms before
the time when the subject was expected to react to a stimulus. These pulses ,
delayed subjects' responses (measured by voluntary EMG in the appropriate
muscles) by periods of up to 70ms. Day et al. proposed that the motor cortex con-
tains a temporary representation of the movement which will be executed. TMS
during the reaction time period abolishes this stored representation, and requires
that it be re-established. The process of re-establishing the 'motor programme'
takes time, leading to delays in RT.
We wished to know if subjects were conscious of these TMS-induced delays.
Since we know from Day et al.'s work that this effect operates on execution
processes in the motor cortex, we can use Day et al.'s delaying effect to localize
M awareness. If subjects are conscious of a TMS-induced delay in their motor
responses, then their awareness of movement must be generated downstream of
. the locus of the TMS effect, namely, in the final execution of the motor command
by the cortex. If, on the other hand, subjects are not aware of the TMS-induced
delays, then motor awareness must be generated upstream of the locus of the TMS
effect, for example, in preparatory pre-motor processes which assemble the motor
programme for subsequent execution. This approach can be extended to identify
two sites, such that awareness is generated downstream of the first and upstream
of the second.
In our expedment (Haggard and Magno, 1999), seven subjects watched a Libet
clock. At a random time after the onset of each trial an auditory warningstimu-
Ius occurred, followed 900 ms later by an auditory go signal. On hearing the go
signal, subjects made speeded key press responses with the right index finger. The
EMG of the principal flexor muscle of the index finger was recorded to allow
accurate detection of the onset of the voluntary muscle activity associated with
the key press. On some trials, the warning signal was not followed by the go sig-
nal, to prevent subjects anticipating. On other control trials, no TMS was deliv-
ered. At the end of each trial, subjects judged the clock position at the moment gf
their,key-press response. TMS was delivered either over the contralateral motor
cortex, over the pre-motor areas (at electrode site FCz), or over occipital cortex
according to condition. FCz roughly overlies the supplementary motor area
Action Awareness 123
300
MI stimulation
-------
0--0 Fez stimulation
250 ~'-6 Occipital stimulation
r-.
I
V'l
200
~
l-
e...
u
OJ
150
~
OJ
I"",
O'l
C
';;"
til 100
Qj
Cl
50
8,- ------·---6
0
Reaction time judf:jed reaction time
(Goldberg, 1985). This pre-motor area has been strongly implicated in the gener-
ation of willed actions. We used a stronger TMS pulse than Day et aI., with the aim
of producing larger delays in reaction time. We did this to ensure that we did not
miss any effects due to delays falling below the resolution of subjects' averaged M
judgements.
To study the effects of TMS on reaction time and on motor awareness, we sub-
tracted subjects' actual reaction times in control (no-TMS) trials from their reac-
tion times in trials where TMS was delivered. Similarly, we subtracted the time of
their M awareness in control trials from the time of M awareness in TMS trials.
The results represent the delaying effects of TMS on actual reaction times, and on
judgements of reaction time (reflecting M awareness) respectively. These results
are shown in Figure 4.4.
First, TMS over the primary motor cortex produced big delays (average 220 ms)
in subject's reaction times. These delays are much larger than those reported by
Day et aI., presumably because of the higher TMS strengths used in our study.
Relatively little of this TMS-induced delay entered into awareness: subjects' judge-
ments of when they pressed the button revealed that less than 100 ms of the TMS-
induced delay entered into conscious awareness.
124 Patrick Haggard
Stimulation over pre-motor areas at electrode site FCz revealed a different pattern
of results. First, TMS-induced delays in reaction time were lower than over primary
motor cortex. Inspection of the EMG traces suggested that even these reduced
delays were not an intrinsic consequence of FCz stimulation, but an indirect con-
sequence due to the strong magnetic stimuli spreading back to the primary motor
cortex. Crucially, however, a greater proportion of the TMS-induced delays
entered into subjects' M awareness following FCz stimulation than following primary
motor cortex stimulation. That is, the subjects' judgements of their own reaction
times following FCz stimulation showed a greater awareness of TMS-induced
delay than following motor cortical stimulation. The interaction between these
data points was significant. Finally, these effects were specific to the neural structures
effected by TMS, rather than a non-specific consequence of the startle caused by
the relatively large magnetic pulses. In a control condition, stimulation over
occipital cortex produced only minimal delays in actual reaction times, and in
awareness of reactions.
In summary, the results of stimulation over motor cortical stimulation and over
pre-motor areas differ. Relatively little of the delay caused by motor cortical stimu-
lation entered into awareness. In contrast, more of the delays produced by FCz
stimulation entered into awarenss. This pattern of results lends support to the
view that M awareness is generated upstream of the primary motor cortex, but
downstream of the pre-motor structures, such as the supplementary motor area,
underlying site FCz. That is, the experiment suggests a pre-motor origin of M
awareness. Awareness of movement is generated not by the final execution of the
pre-assembled motor command, but by the preparatory pre-motor processes
which assemble that motor command.
4. CONCLUSION
This chapter has described some of the experimental data investigating subjects'
conscious awareness of the generation of movement and has related those con-
scious states to the neural processes generating movement. In the first experiment,
awareness of intention was related to the LRP: a brain potential associated with
the specification of which of two movements to make. In the second experiment,
behavioural evidence showed an association between awareness of movement,
and preparation or assembly of 'motor programme'. In the third experiment,
intervention on motor processing with TMS produced converging psychophysio-
logical evidence; again, awareness of movement was associated with brain
processes concerned with the assembly and preparation of movement, rather than
those concerned with execution.
What general speculations can we make about conscious awareness of action
based on these data? I will restrict myself to three comments in an ascending order
Action Awareness 125
of extravagance and distance from the data. First, both awareness of intention
(W) and awareness of action (M) appear to occur in a narrow window of pre-
motor processing between the abstract prior intention to do something, and the
completion of a specific programme or plan of how to do it. Awareness of intention
and awareness of movement are conceptually distinct, and have different absolute
numerical values in judgement task like Libet's. Nevertheless, I suggest they derive
from a single processing stage in the motor pathway. I have already suggested that
this locus may correspond to the computationally difficult process of movement
selection. Therefore, there appears to be a close correspondence between motor
awareness, at least in these tasks, and motor processing.
Second, the strong relation between pre-motor processing and consciousness
points to an important and qualitative difference between consciousness of action
and consciousness of external stimuli. In vision, our consciousness is tied to our
perception of objects. Several stages of processing ensure perceptual constancy:
our conscious percept of objects remains the same despite large changes in stimu-
lus luminosity, incident light, viewing angle, and so on. We do not have conscious
access to representations of the stimuli before those constancy processes are
applied, nor to operation of the constancy processes themselves. We have access
only to the output: namely, the reliable, informationally economical, functional
percept of the object in our world. That is, visual consciousness appears to be
anchored at the output end of visual processing. A good example from vision
comes from the work of Logothetis (1998) on stimuli exhibiting rivalry. Within
the higher levels of visual processing in the cortex, single cells appear to respond
to which of two rival stimuli the monkey sees at a given time, based on beha-
vioural evidence. The neural and behavioural data from monkeys correlate closely
with verbal reports of human subjects in the same visual settings. Again, both the
conscious percepts and their presumed neural correlates are localized at the
higher, output end of visual processing.
The psychology of action is often considered as the inverse of the psychology of
perception. However, I now wish to speculate that our consciousness of actions
differs from our consciousness of external visual events. First, consciousness of
action is not a conscious awareness of the output of the system, as it is in vision.
The output of the motor system is the physical movement of part of the body
across space and time. The experiments reported here have shown that the per-
ceived time of actions is more closely tied to movement preparation than to
movement execution. Similar results have been reported in the perception of spatial
parameters of movement. Fourneret and Jeannerod (1998) found that subjects
were remarkably unaware of the actual movement parameters, such as the path of
the hand through space, that they used when reaching with prismatically distorting
visual displays. It seems unlikely, then, that motor awareness relates to actual
motor outputs. On the other hand, neither do the data reviewed here localize
awareness of action at the very start of the motor processing hierarchy, as a strict
126 Patrick Haggard
inversion of visual models might suggest. Even when we asked subjects to report
awareness of the earliest intention leading to movement, conscious access to
motor processing was restricted to the narrow window of pre-motor activity
measured by the LRP. This seems well downstream from the very earliest neural and
functional processes which precede action. Certainly, our chronometric approach
does not suggest a strong conscious representation of the long-term antecedents
such as Searle's (1982) prior intention. Unlike vision, conscious awareness within
the motor system is localized to the computationally intensive processes in the
middle of the processing chain.
Third, and finally, I began this chapter by observing the importance of binding
consciousness of action to consciousness of intention, in order to generate an agen-
tic self. The experiments reported here do not address the self or self-consciousness
directly, and it is hard to see how the experimental chronometry approach taken
here could do that. Nevertheless, the coexistence of awareness of intention (W)
and awareness of action (M) within a single narrow window of motor processing
does suggest that binding these two conscious representations may be important.
We have access to awareness of both intention and of action: they appear to be gen-
erated by similar processing stages, at comparable times in the development of
action. The conjunction of these conscious representations suggests an important
function of action awareness in binding the two representations. The efferent bind-
ing process could have the dual function of bringing to consciousness mismatches
between intention and action, and of making possible a second, derived kind of
consciousness of the relations between my intentions and my actions. This could
be a part of the sense of self. I would like to speculate that a psychologically and
neurally coherent account of self and of self-consciousness must build on the foun-
dations of associating the awareness of intention and of action within voluntary
movement. In our future research, we plan to investigate judgements of agency, to
assess how tight the couplings in space and time between intention and action need
to be for subjects to attribute actions to themselves.
REFERENCES
BAUMEISTER, R. F. (1998), 'The Self', in D. T. Gilber and S. T. Fiske (eds.), The Handbook of
Social Psychology, ii. Boston: McGraw-Hill, 680-740.
BLAKEMORE, S. J., WOLPERT, D. M., and Frith, C. D. (1998), 'Central cancellation of
self-produced tickle sensation', Nature Neuroscience, 1: 635-40.
DAY, B. 1., ROTHWELL, J. c., THOMPSON, P. D., MAERTENS-DE-NoORDHOUT, A.,
NAKASHIMA, K., SHANNON, K., and MARSDEN, C. D. (1989), 'Delay in the execution of
voluntary movement by electrical or magnetic brain stimulation in intact man: evidence
for the storage of motor programs in the brain', Brain, 112: 649-63.
FOURNERET, P., and JEANNEROD, M. (1998), 'Limited conscious monitoring of motor
performance in normal subjects', Neuropsychologia, 36: 1133-40.
Action Awareness 127
GOLDBERG, G. (1985), 'Supplementary motor area structure and function: review and
hypotheses', Behavioral and Brain Sciences, 8: 567-616.
HAGGARD, P., and EIMER, M. (1999), 'On the relation between brain potentials and the
awareness of voluntary movements', Experimental Brain Research, 126: 128-33.
_-and MAGNO, E. (1999), 'Localising awareness of action with Transcranial Magnetic
Stimulation', Experimental Brain Research, 127: 102-7.
_-and RICHARDSON, J. (1996), 'Spatial patterns in the control of human movement',
Journal ofExperimental Psychology: Human Perception and Performance, 22: 42-62.
_-HUTCHINSON, K., and STEIN, J. F. (1995), 'Patterns of coordinated multijoint move-
ment', Experimental Brain Research, 107: 254-66.
--LESCHZINER, G., MIALL, R. c., and STEIN, J. F. (1997), 'Local learning of inverse kine-
matics in human reaching movement', Human Movement Science, 16: 133--47.
--NEWMAN, c., and MAGNO, E. (1999), 'On the perceived time of voluntary action',
British Journal ofPsychology, 90: 291-303.
LIBET, B. (1985), 'Unconscious cerebral initiative and the role of conscious will in volun-
tary action', Behavioural-and-Brain-Sciences, 8: 529-66.
--GLEASON, C. A., WRIGHT, E. W., and PEARL, D. K. (1983), 'Time of conscious intention
to act in relation to onset of cerebral activity (readiness potential)', Brain, 106: 623--42.
LOGOTHETIS, N. K. (1998), 'Single units and conscious vision', Philosophical Transactions of
the Royal Society (B), 353: 1801-18.
MCCLOSKEY, D. 1., COLEBATCH, J. G., POTTER, E. K., and BURKE, D. (1983), 'Judgements
about onset of rapid voluntary movements in man', Journal of Neurophysiology, 49:
851-63.
MILL, J. S. (1843), A System of Logic. London: Murray.
MORTON, J., MARCUS, S. M., and FRANKISH, C. (1976), 'Perceptual centres (P-centres)"
Psychological Review, 83: 405-8.
SEARLE, J. R. (1983), Intentionality. Cambridge: Cambridge University Press.
STERNBERG, S., MONSELL, S., KNOLL, R. 1., and WRIGHT, C. E. (1978), 'The latency and
duration of rapid movement sequences: comparisons of speech and typewriting', in
G. E. Stelmach (ed.), Information Processing in Motor Control and Learning. Amsterdam:
North-Holland, 117-52.
WOLPERT, D. M., MIALL. R. c., and KAWATO, M. (1998), 'Internal models in the cerebellum',
Trends in Cognitive Sciences, 2: 338--47.
5
Consciousness of Action and Self-Consciousness:
A Cognitive Neuroscience Approach
Marc Jeannerod
INTRODUCTION
The mutual relationships of action to consciousness are complex and far from
unequivocal. We execute many of our daily actions unconsciously; conversely, we
can consciously simulate or imagine actions we do not execute. This vast domain
of research, motor cognition, is central to the study, not only of action itself (how
it is planned, prepared, and finally executed), but also of how action contributes
to the representations we build from objects and from other selves. In this chap-
ter, I will use the broad term of representation to include the various internal states
in relation to action. There could be better terms, with the advantage of greater
precision, but with the disadvantage of losing continuity between different levels
of functioning. Indeed, the term representation of an action can be used in its
strong sense, to designate a mental state in relation to goals and desires, as well as
in its weak sense, to indicate the ensemble of mechanisms that precede execution
of a movement. Finally, it can also be accepted by biologists to designate the state
of the neural network during a mental state related to action.
The various levels of representation of an action will be described with regard
to the subjective experience felt by an agent during interactions with objects in
space or with other individuals. Three main levels will be identified depending on
whether consciousness of objects, consciousness of goals, or consciousness of the
self is concerned.
1. OBJECT-ORIENTED MOVEMENTS:
THE AUTOMATIC LEVEL
Most of our movements directed at objects are prepared and executed automatically.
Once started, they are performed accurately and rapidly (within less than one sec-
ond), which leaves only a little time for top-down control. Models accounting for
such characteristics have used the notion of optimization principles, that is, prin-
ciples that are represented within the motor system and operate during execution
itself. In the motor literature, such principles have been described in the kinematic
A Cognitive Neuroscience Approach 129
domain (e.g. the minimum jerk model, Hogan and Flash, 1987), in the domain of
biomechanics (optimized trajectories, Desmurget et al., 1997; end-point comfort,
Rosenbaum et al., 1990), and in the time domain (Hoff and Arbib, 1993). Besides
optimization of execution, other features of object-oriented movements strongly
suggest that they are organized, or represented, prior to execution. A widely studied
example is the action of grasping. Changes in finger position appropriate for a
stable grasp occur during the reaching component that transports the hand at the
object location, that is, far ahead of contact with the object. This process of grip
formation has been epitomized by the parameter of maximum grip amplitude: at
about 70 per cent of movement time, the distance between the fingers that com-
pose the grip (e.g. thumb and index finger) reaches a maximum value which
exceeds object size. The important point, however, is that maximum grip ampli-
tude has been shown to be exactly correlated to object size and is therefore likely
to reflect a representation of the object (see Jeannerod, 1986).
The representation that could account for such anticipatory adjustments must
therefore encode not only the properties of the central and peripheral motor sys-
tem used for optimization of the movement execution. It must also encode those
properties of the object that are relevant to potential interactions with the agent,
according to his intentions or needs: object's shape and size are relevant to grip
formation (maximum grip size, number of fingers involved), its texture and estim-
ated weight are relevant to anticipatory computation of grip and load forces, and
so on. In addition, the processing of object properties must take into account the
location and orientation of the object with respect to the body. In other words, the
representation of object properties interacts with motor factors, such as the bio-
mechanics of the arm, in defining the final pattern of the grasp. To demonstrate
this point, Paulignan et al. (1996) systematically studied the orientation of the
opposition axis (the axis defined by the positions of the thumb and the index fin-
ger on the object surface) during the process of grasping the same object placed
at different positions in the working space. The object they used (an upright cylin-
der) was visually 'simple' in that it afforded an infinite number of possible posi-
tions of the fingers on its surface. In this condition, the orientation of the
opposition axis was only guided by the biomechanics of the arm. They found that
the orientation of the opposition axis remained invariant with respect to the body
axis of the subject, which is a clear indication that computation of the grasp by the
visual system was effected within a body-centred (egocentric) frame ofreference
(Fig. 5.1 (a) ). In most everyday situations, however, the interaction between visual
and motor factors is dominated by visual constraints. Indeed, if the shape of the
object affords only a few positions for the opposition axis, the arm biomechanics
have to adapt to the object. As shown by Figure 5.1h, a slight change in the orien-
tation of such an object may cause a major reconfiguration of the grasping
movement directed at that object (Stelmach et al., 1994).
80
~ 70
~ 60 ~
_
~ .....:J
40
:;- 50
t 40
~
-i1 30
~ 20
o -0- Small
Y :$i 10 --e- Middle
i
x
o
8: 0 -0- Large
o -lOLl_,-,-_ _-,--_ _,-,--_ _-:-_ _--,-,,_ _----:_
I T
(a)
w (b) -10 0 10 20 30 40
Object position (degrees)
80
~ 70 --0- Small
~ 60
-0-
~Large
Middle
:;- 50
t 40 ~ ~ '-"
[]J'
~-
!~~
~ 10
o
• •~.
:' •
8: 0
0_ 10 ---:-:---::----:-;:----,:;-;0----,;;;-----:;",--
-10 0 10 20 30 40
(d)
(c) Object position (degrees)
The term of pragmatic representation has been proposed (see Jeannerod, 1994) to
qualify this mode of representing objects as goals for action. The most striking
characteristic of pragmatic representation is its implicit functioning and, correl-
atively, its unconscious nature. Although further qualification will be needed to
better define the term pragmatic (see below), this mode of representation opposes
another, not directly related to action (the semantic representation), whereby the
same objects can be processed for identification, naming, and so on. This distinction
stresses the fact that objects are multiply represented, according to the task in
which they are involved. For the purpose of the present chapter, which is to discuss
the degree of consciousness attached to different forms of action, we have to con-
centrate on two controversial and related issues: (i) whether pragmatic processing
is associated with a specific neural system and (ii) whether the unconscious nature
of pragmatic processing is a consequence of the structure of this system or a
consequence of constraints imposed by pragmatic processing itself.
The first issue, which will be discussed only briefly here (see Jeannerod, 1997),
relies on an anatomical theory for the processing of visual information. This theory,
which has been developed in the last twenty years or so, proposes a duality of
cortico-cortical visual pathways (Ungerleider and Mishkin, 1982): an occipito-
parietal route (the dorsal visual pathway) specialized for visuospatial processing
and, by extension, for processing object-oriented movements; and an occipito-
temporal route (the ventral pathway) for object recognition. Subjects with lesions
located in specific areas of the parietal lobe exhibit a typical impairment in object-
oriented behaviour with their contralesional arm. They misreach object location
and are unable to form the proper grip: grip size is improperly calibrated, orien-
tation of opposition axis is inappropriate, and so on (Jeannerod et al., 1994). This
impairment, however, appears to be limited to processing those visual properties
of an object that are relevant to interacting with it. Of course, in the same subjects,
semantic identification of the objects is preserved. These observations stress the
fact that pragmatic processing might take place in the parietal lobe (i.e. in the dorsal
pathway of visual cortico-cortical connections), whereas semantic processing
would take place somewhere else (indeed, there are examples of subjects with
large lesions in the ventral part of the visual system, who fail to recognize objects,
whereas they can correctly handle them; Goodale et al., 1991). The fact that, in
parietal patients, the grasping deficit is limited to the arm contralateral to the
lesion supports the notion that pragmatic processing is controlled by local mech-
anisms and, by way of consequence, should be little influenced by the more global
top-down mechanisms which are required for conscious access.
This raises the second issue: are automatic movements unconscious because
they pertain to a neural system (the dorsal pathway) that, by design, does not
operate on the conscious mode? This attractive possibility is not entirely sup-
ported by experimental data in normal subjects. In a recent study using position
132 Marc Jeannerod
emission tomography (PET) Faillenot et ai. (1997) have compared the patterns of
cortical activation during two different tasks, an action task (grasping objects of
different sizes and shapes by hand), and a perceptual task (perceptually matching
these objects with each other). In the first task, the main activation foci were in
the motor areas and the inferior parietal lobule contralateral to the hand used
for the task, but also in the right posterior part of the intraparietal sulcus. In
the second task, two foci were found, one in the left inferotemporal cortex, and one
in the right posterior parietal cortex: this latter focus clearly overlapped with the
homologous focus activated during grasping. This result means that perceptual
analysis, even when no action is to occur, uses resources that pertain to the dorsal
pathway and are also used during object-oriented action. Following up this result,
Faillenot et al. (1999) tested further the degree of involvement of parietal cortex
during perceptual discrimination. Subjects had to discriminate shapes which were
presented with different degrees of slant in the frontal plane (a 2-D orientation
task) or in the sagittal plane (a 3-D orientation task). These tasks, where no action
was ever required, did produce activation in areas located in the posterior part of
the intraparietal sulcus (the dorsal pathway), as well as at the occipito-temporal
junction and in the inferior temporal gyrus (the ventral pathway).
These results in normal subjects illustrating perceptual functions of the parietal
lobe are to be compared with the effects of lesions of the same areas. Patients with
such lesions, besides their problems in visuomotor transformation already
described, may also exhibit typical perceptual deficits. They may be unable to
copy objects by drawing (the so-called constructive apraxia); they may have great
difficulties recognizing on a photograph objects displayed in a non-canonical ori-
entation (Warrington and James, 1986). They cannot resolve, either motorically
or perceptually, the 3-D orientation of objects or their spatial relations with
respect to other objects. One could speculate that these perceptual difficulties
represent one and the same deficit as the visuomotor ones: only those aspects of
perception which relate to action would be affected by the lesion, whereas other
aspects-those related to identification and semantic processing-would be left
intact. Obviously, this would require a radical redefinition of what has been called
pragmatic processing, as it would now include representation of the same visual
stimuli both in an allocentric and an egocentric reference frame, where that part
of the processing coded egocentrically (the visuomotor transformation) would be
devoid of conscious counterpart, whereas the perceptual part would be conscious.
Another interpretation for automaticity of object-oriented movements would
be that these movements are unconscious because this is a prerequisite for accu-
racy. The main argument would be that the life span of the working memory used
for a goal-directed movement must not exceed the duration of the movement
itself: if corrections have to be generated, they must be effected on the basis of the
goal of each movement, and the representation of that goal must be erased before
another segment of the action starts. In other words, if one assumes that access to
A Cognitive Neuroscience Approach 133
conscious processing is a time-consuming process, the requirements for accuracy
would not leave enough time for consciousness to appear. Experiments where the
goal is modified during the movement support this view. If a target briskly
changes its location during the ocular saccade that precedes a pointing movement
towards that target, subjects usually remain unaware of the displacement (they see
only one, stationary, target). Yet, they correctly point at the final target location
(e.g. Bridgeman et ai., 1981). Goodale et al. (1986) reported a pointing experiment
where the target occasionally made jumps of several degrees, unnoticed by the
subjects. They found that the subjects were none the less able to adjust the trajec-
tory of their moving hand to the target position. Interestingly, no additional time
was needed for producing the correction, and no secondary movement was
observed, suggesting that the visual signals related to the target shift were used
without delay for adjusting the trajectory. According to this view, generating a
motor response to a stimulus and building a perceptual experience of that same
stimulus would activate different mechanisms with different time constants
(Castiello et al., 1991). If the motor response to a stimulus were delayed by a few
seconds, the fast mechanism would become inefficient and the movement would
be effected under the control of the slow mechanism. As observed by several
authors, this would result in a severe degradation of its accuracy (e.g. Jakobson
and Goodale, 1991).
lli
areir~:~gmirror
..-.J
Pen
stylus
Graphic
tablet
I _
.- Onset of right
perturbation (10·)
(n=9)
10 5 6 7
1: 9
C1
ii:
8 +
~
c:
0
~~CJ?T
pepc!'~
Q.
~
""i;
6 +
-e
>'" 5
<l:' 4
.3
+
2
(c) -10 -7 -5 -2 2 5 7 10 (d)
the representation of the desired state would not be erased. It has been previously
suggested, on the basis of introspective evidence, that in such a situation one
should shift to a conscious strategy whereby one becomes aware of the goal of the
movement and of the cause ofits failure (Jeannerod, 1994). This suggestion seems
to be supported by the results of a recent experiment by Slachevsky et al. (2001),
using the same design as above. Although Fourneret and Jeannerod (1998) had
limited the amplitude of the bias to 10 degrees and randomized the trials with
136 Marc Jeannerod
different amplitudes and directions of bias, Slachevsky and her colleagues used a
bias in one direction only (e.g. right), with increasing amplitude from trial to trial,
up to 42 degrees. At an average amplitude between 14 and 24 degrees, normal sub-
jects became clearly aware of the bias and declared that the movement of their
hand erred in a direction different from that seen on the screen. Interestingly,
beyond this value of 24 degrees, their motor performance suddenly improved.
This result tends to confirm the introspective evidence that a movement becomes
conscious when it fails (e.g. Pacherie, 1997). When large discrepancies appear and
corrections are beyond the capacities of the automatic level, another level of the
action system is activated. This hypothesis will be further developed below.
Little is known about the neural basis of this mechanism. Slachevsky and her
colleagues examined further a group of patients with frontal lobe lesions. Unlike
normal subjects, frontal patients did not shift from the automatic to the conscious
mode when the bias increased. Instead, they kept using the same automatic strat-
egy, making larger and larger, unnoticed errors (Slachevsky et aI., 2001). Indeed,
patients with frontal lobe lesions exhibit typical impairments demonstrating their
inability to consciously monitor their performance and to override their automatic
responses. Some of these patients present with so-called 'imitation behaviour'
whereby they compulsively reproduce actions performed by other people in front
of them or compulsively use objects displayed in front of them (Lhermitte, 1983).
Let us consider here the situation where an action is executed, but its origin is
rendered ambiguous. The question will be to determine how an agent can disen-
tangle this ambiguity to produce an agency judgement. To this end, an experi-
mental situation was designed by Daprati et ale (1997), using a paradigm initially
designed by Nielsen (1963). In this experiment, the subject's hand and the experi-
menter's hand were filmed by two different cameras. By changing the position of
a switch, one or the other hand could be briefly displayed on the video screen seen
by the subject. The two hands looked alike as they were both covered with a sim-
ilar glove. In each trial, both the experimenter and the subject had to' perform a
given hand movement on command (e.g. stretch thumb, stretch fingers one and
two, etc.): on some trials, however, the experimenter's movement departed from
the instruction. As a result of this experimental arrangement, the subject was ran-
domly shown either his own hand, or the experimenter's hand performing the
same movement as him, or a different movement. At the end of each trial, a ver-
bal agency judgement was recorded: the subject had to say whether the hand he
had seen was his hand or another hand. Normal subjects were able to determine
unambiguously whether the moving hand seen on the screen was theirs or not, in
A Cognitive Neuroscience Approach 143
the two 'easy' conditions: first, when they saw their own hand, they correctly
attributed the movement to themselves; second, when they saw the experimenter's
hand performing a movement which departed from the instruction they had
received, they correctly denied seeing their own hand. By contrast, their perform-
ance degraded in the 'difficult' trials where they saw the experimenter's hand per-
forming the same movement as required by the instruction: in this case, they
misjudged the alien hand to be theirs in about 30 per cent of cases (Fig. 5.3(a) ).
One possibility is that the subjects in this experiment based their judgement on
internal signals arising from their own action, and on monitoring the degree of
correlation between these internal signals and other types of signals (e.g. visual).
However, if one considers that these action-related signals are not consciously
monitorable, the problem remains to understand how these could influence the
state of the conscious level where the agency judgement was generated. Another
possibility is that subjects were able to make a conscious distinction between the
representations, in their own brain, pertaining to a self-produced action or an
action produced by another agent, based on the pattern of cortical activation gen-
erated for each of these modalities. Internal, self-generated, signals contributed to
the correct agency judgement by reinforcing those representations pertaining to a
self-produced action (see Jeannerod, 1999).
E
---------~
Video camera 2
TV screen
(0)
24
20
16
c
'" 12
'6
OJ
2: 8
Non-delusional
4
o
Exp. Same
Exp. Ditt. Subject
(b) Hand presented
4. CONCLUSION
Several levels of representation of actions have been identified. The automatic level
essentially corresponds to non-conscious interaction with objects in the external
world. It relies on pragmatic processing of these objects, in order to extract those
146 Marc ]eannerod
properties (extrinsic and intrinsic) that are relevant to action. Brain areas respons-
ible for visuomotor transformation, like parietal areas surrounding the intraparietal
sulcus, and premotor areas, are likely to be involved in this mechanism.
The content of motor representations may become conscious under a variety of
conditions. One condition is the failure of the automatic level. Mental simulation
of an action, where the representation is voluntarily activated from within, or
intentional selection of an action from several possible alternatives, are other
examples. Brain areas located in the left dorsolateral prefrontal cortex and in the
anterior cingular region might be critical for this function, as suggested by neuro-
imaging and also by behavioural observations in patients with frontal lobe lesions:
these patients consistently cannot shift from the non-conscious to the conscious
mode in the event of failure of the automatic level. Whatever way it is accessed,
however, the conscious content remains remarkably limited.
Finally, attribution of an action to its proper agent represents the ultimate
aspect of consciousness of action. A possible hypothesis for the neural mechanism
of agency is that internal self-generated signals would exert an inhibitory influ-
ence on those cortical sites which are responsible for analysing the effects of an
action. Lack of inhibition and correlative hyperactivity in those sites (due to dis-
ruption of these self-generated signals, for example) would automatically refer the
origin of the action to an external agent. Conversely, lack of disinhibition would
lead to overattribution to the self. The specific impairment of this network in
schizophrenia would explain the variety of agency problems in these patients.
This could represent the core of the disease, if one thinks that the ability to pro-
duce agency judgements is one of the foundations of self-consciousness and, by
extension, represents the first step in communication between individuals and
social interactions. Those questions of how can the self become aware of its own
productions, how does it distinguish itself from other selves, in other words, how
can the 'Who?' of an action be determined, are critical questions inherent to the
social nature of human beings (Georgieff and Jeannerod, 1998).
REFERENCES
BRIDGEMAN, B., KIRCH, M., and SPERLING, A. (1981), 'Segregation of cognitive and motor
aspect of visual function using induced motion', Perception and Psychophysics, 29:
336-42.
CASTIELLO, u., PAULIGNAN, Y., and JEANNEROD, M. (1991), 'Temporal dissociation of
motor responses and subjective awareness: a study in normal subjects', Brain, 114:
2639-55.
DAPRATI, E., FRANCK, N., GEORGIEFF, N., PROUST, J., PACHERIE, E., DALERY, J., and
JEANNEROD, M. (1997), 'Looking for the agent: an investigation into consciousness of
action and self-consciousness in schizophrenic patients', Cognition, 65: 71-86.
DASSER, V., ULBAEK, I., and PREMACK, D. (1989), 'The perception of intention', Science,
243: 365-7.
A Cognitive Neuroscience Approach 147
DECETY, J., PERANI, D., JEANNEROD, M., BETTINARDI, v., TADARY, B., WOODS, R.,
MAZZIOTTA, J. c., and FAZIO, F. (1994), 'Mapping motor representations with PET',
Nature, 371: 600-2.
_GREZES, J., COSTES, N., PERANI, D., JEANNEROD, M., PROCYK, E., GRASSI, F., and
FAZIO, F. (1997), 'Brain activity during observation of action: influence of action content
and subject's strategy', Brain, 120: 1763-77.
DESMURGET, M., JORDAN, M. A., PRABLANC, c., and JEANNEROD, M. (1997), 'Constrained
and unconstrained movements involve different control strategies', Journal of
Neurophysiology, 20: 1644-50.
DIERKS, T., LINDEN, D. E. J., JANDL, M., FORMISANO, E., GOEBEL, R., LANFERMAN, H., and
SINGER, W. (1999), 'Activation of the Heschl's gyrus during auditory hallucinations',
Neuron, 22: 615-2I.
ENGELKAMP, J. (1995), 'Visual imagery and enactment of actions in memory', British
Journal of Psychology, 86: 227--40.
FAILLENOT, 1., TONI, 1., DECETY, J., GREGOIRE, M. c., and JEANNEROD, M. (1997), 'Visual
pathways for object-oriented action and object recognition: functional anatomy with
PET', Cerebral Cortex, 7: 77-85.
--DECETY, J., and JEANNEROD, M. (1999), 'Human brain activity related to the percep-
tion of spatial features of objects', Neuroimage, 10: 114-24.
FOURNERET, P., and JEANNEROD, M. (1998), 'Limited conscious monitoring of motor per-
formance in normal subjects', Neuropsychologia, 36: 1133--40.
FRAK, V. G., PAULIGNAN, Y., and JEANNEROD, M. (2001), 'Orientation of the opposition axis
in mentally simulated grasping', Experimental Brain Research, 136: 120-7.
FRITH, C. (1996), 'The role of the prefrontal cortex in self-consciousness: the case of audi-
tory hallucinations', Philosophical Transactions of the Royal Society, London B, 351:
1505-12.
--FRISTON, K., LIDDLE, P. F., and FRACKOWIAK, R. S. J. (1991), 'Willed action and the
prefrontal cortex in man: A study with PET', Proceedings of the Royal Society, 244: 241-6.
GALLESE, v., and GOLDMAN, A. (1998), 'Mirror neurons and the simulation theory of
mind-reading', Trends in Cognitive Science, 2: 493-50I.
GEORGIEFF, N., and JEANNEROD, M. (1998), 'Beyond consciousness of external reality: a
"Who" system for consciousness of action and self-consciousness', Consciousness and
Cognition, 7: 465-77.
GOODALE, M. A., PELISSON, D., and PRABLANC, C. (1986), 'Large adjustments in visually
guided reaching do not depend on vision of the hand or perception of target displace-
ment', Nature, 320: 748-50.
--MILNER, A. D., JAKOBSON, 1. S., and CAREY, D. P. (1991), 'Perceiving the world and
grasping it: a neurological dissociation', Nature, 349: 154-6.
GRAFTON, S. T., ARBIB, M. A., FADIGA, 1., and RIZZOLATTI, G. (1996), 'Localization of grasp
representations in humans by positron emission tomography. 2. Observation compared
with imagination', Experimental Brain Research, 112: 103-1 I.
GRAY, J. A., FELDON, J. N., RAWLINS, J. N. P., HEMSLEY, D. R., and SMITH, A. D. (1991), 'The
neuropsychology of schizophrenia', Behavioral and Brain Sciences, 14: 1-84.
HOFF, B., and ARBIB, M. A. (1993), 'Models of trajectory formation and temporal interac-
tion of reach and grasp', Journal of Motor Behavior, 25: 175-92.
148 Marc Jeannerod
HOGAN, N., and FLASH, T. (1987), 'Moving gracefully: quantitative theories of motor coor-
dination', Trends in Neuroscience, 10: 170-4.
HOLST, E. von (1954), 'Relations between the central nervous system and the peripheral
organs', British Journal ofAnimal Behavior, 2: 89-94.
JAKOBSON, 1. 5., and GOODALE, M. A. (1991), 'Factors affecting higher order movement
planning: a kinematic analyis of human prehension', Experimental Brain Research, 86: I
199-208.
JEANNEROD, M. (1986), The Neural and Behavioural Organisation of Goal-Directed
Movements. Oxford: Oxford University Press.
--(1994), 'The representing brain: neural correlates of motor intention and imagery',
Behavioral and Brain Sciences, 17: 187-245.
--(1997), The Cognitive Neuroscience ofAction. Oxford: Blackwell.
--(1999), 'To act or not to act: perspectives on the representation of actions', Quarterly
Journal of Experimental Psychology, 52A: 1-29.
--(2001), 'Neural simulation of action: a unifying mechanism for motor cognition',
Neuroimage, 14: 5103-5109.
--DECETY, J., and MICHEL, F. (1994), 'Impairment of grasping movements following a
bilateral posterior parietal lesion', Neuropsychologia, 32: 369-80.
JOHANSSON, G. (1977), 'Studies on visual perception oflocomotion', Perception, 6: 365-76.
LHERMITTE, F. (1983), "'Utilisation behaviour" and its relation to lesions of the frontal
lobes', Brain, 106: 237-55.
NIELSON, T. 1. (1963), 'Volition: a new experimental approach', Scandinavian Journal of
Psychology, 4: 225-30.
PACHERIE, E. (1997), 'Motor images, self-consciousness and autism', in J. Russell (ed.),
Autism as an Executive Disorder. Oxford: Oxford University Press, 215-55.
PAIVIO, A. (1986), Mental Representations: A Dual Coding Approach. Oxford: Clarendon
Press.
PARSONS, 1. M. (1994), 'Temporal and kinematic properties of motor behavior reflected in
mentally simulated action', Journal of Experimental Psychology: Human Perception and
Performance, 20: 709-30.
- - - - F o x , P. T., DOWNS, J. H., GLASS, T., HIRSCH, T. B., MARTIN, C. c.,
JERABEK, P. A., and LANCASTER, J. 1. (1995), 'Use of implicit motor imagery for visual
shape discrimination as revealed by PET', Nature, 375: 54-8.
PAULIGNAN, Y., FRAK, V. G., TONI, 1., and JEANNEROD, M. (1997), 'Influence of object posi-
tion and size on human prehension movements', Experimental Brain Research, 114:
226-34.
PERANI, D., CAPPA, S. F., BETTINARDI, V., BRESSI, 5., GORNO-TEMPINI, M., MATARRESE, M.,
and FAZIO, F. (1995), 'Different neural systems for the recognition of animals and man-
made tools', NeuroReport, 6: 1637-41.
REQUIN, J. (1992), 'From action representation to motor control', in G. E. Stelmach and
J. Requin (eds.), Tutorials in Motor Behavior. Amsterdam: Elsevier.
RIZZOLATTI, G., FADIGA, 1., GALLESE, v., and FOGASSI, 1. (1995), 'Premotor cortex and the
recognition of motor actions', Cognitive Brain Research, 3: 131-41.
ROSENBAUM, D. A., MARCHAK, F., BARNES, H. J., VAUGHAN, J., SLOTTA, J. D., and
JORGENSEN, M. J. (1990), 'Constraints for action selection: overhand versus underhand
A Cognitive Neuroscience Approach 149
grips', in M. Jeannerod (ed.), Motor Representation and Control, Attention and
Performance XIII. Hillsdale, NJ: Erlbaum, 321-42.
SCHNEIDER, K. (1959), Clinical Psychopathology. New York: Grunne and Stratton.
SHEPARD, R. N., and METZLER, J. (1971), 'Mental rotation of three dimensional objects',
Science, 171: 701-3.
SIRIGU, A., DUHAMEL, J.-R., COHEN, 1., PILLON, B., DUBOIS, B., and AGID, Y. (1996), 'The
mental representation of hand movements after parietal cortex damage', Science, 273:
1564-8.
SLACHEVSKY, A., PILLON, B., FOURNERET, P., PRADAT-DIEHL, P., JEANNEROD, M., and
DUBOIS, B. (2001), 'Preserved adjustment but impaired awareness in a sensory-motor
conflict following prefrontal lesions', Journal of Cognitive Neuroscience, 13: 332-40.
SPENCE, S. A., BROOKS, D. J., HIRSCH, S. R., LIDDLE, P. F., MEEHAN, J., and GRASBY, P. M.
(1997), 'A PET study of voluntary movement in schizophrenic patients experiencing
passivity phenomena (delusious of alien control)', Brain, 120: 1997-201l.
STELMACH, G. E., CASTIELLO, u., and JEANNEROD, M. (1994), 'Orienting the finger opposi-
tion space during prehension movements', Journal ofMotor Behavior, 26: 178-86.
STEPHAN, K. M., FINK, G. R., PASSINGHAM, R. E., SILBERSWEIG, D., CEBALLOS-BAUMANN,
A. 0., FRITH, C. D., and FRACKOWIAK, R. S. J. (1995), 'Functional anatomy of the mental
representation of upper extremity movements in healthy subjects', Journal of
Neurophysiology, 73: 373-86.
UNGERLEIDER, 1., and MISHKIN, M. (1982), 'Two cortical visual systems', in D. J. Ingle,
M. A. Goodale, and R. J. w. Mansfield (eds.), Analysis of Visual Behavior. Cambridge,
Mass.: MIT Press, 549-86.
WARRINGTON, E. K., and JAMES, M. (1986), 'Visual object recognition in patients with right
hemisphere lesions: axis or features', Perception, 15: 355-66.
WOLPERT, D. M., GHAHRAMANI, Z., and JORDAN, M. 1. (1995), 'An internal model for sens-
orimotor integration', Science, 269: 1880-2.
6
The Role of Demonstratives
in Action-Explanation
John Campbell
1. THE DATUM
It has often been remarked that perceptual demonstratives have a special role to play
in the explanation of action. In 'The Problem of the Essential Indexical', Perry dis-
cussed the case of a hiker standing in the wilderness, unsure as to which of the
mountains he sees is which, and not sure which trail he is looking at. The hiker has
a period of deliberation. 'Then he begins to move along the Mt. Tallac trail. If asked,
he would have explained the crucial change in his beliefs this way: "I came to believe
that this is the Mt. Tallac trail and that is Gilmore Lake'" (Perry 1993: 35).
The connection between perceptual demonstratives and action was given a
strong formulation by Peacocke, who said that 'no set of attitudes gives a satisfactory
psychological explanation of a person's acting on a given object unless the content
of those attitudes includes a demonstrative mode of presentation of that object'
(Peacocke, 1981: 205-6). This is quite a strong thesis. As it stands, it is open to the
point that I can act on the candidate standing in an election by voting against him,
or that I can act on you by sending you an e-mail, without a demonstrative mode
of presentation being involved at all. But that leaves it open that there may be
some more primitive types of action, such as choosing which path to take, or
reaching and grasping an object, whose explanation does typically require the use
of a demonstrative.
I think that another way to get at the connection between perceptual demon-
stratives and action, other than by simply insisting that demonstratives are needed
for the explanation of the action, is suggested by a distinction which Austin (1964)
drew between doing something by accident and doing something by mistake. His
example was this. You have a donkey, and so have 1. They graze in the same field.
One day, I decide to shoot mine. I go to the field where the two donkeys are graz-
ing side by side, and take careful aim at mine. At the last moment, the two don-
keys move, and I shoot yours by accident. In the second case, I take careful aim at
what is in fact your donkey, believing it to be mine, and shoot it through the heart.
Here I did not shoot your donkey by accident; I shot it by mistake. The content of
the mistake is easy enough to specify. It is something like the judgement, 'my don-
key is that donkey', where 'that donkey' is a demonstrative identifying a perceived
Demonstratives in Action-Explanation 151
donkey. In the case in which I shoot your donkey by accident, though, there is still
such an identity judgement-I am depending upon some such belief as 'my donkey
is that donkey', only in this case my identity belief is correct, and the consequence
of the donkeys n10ving at the last minute is that I do not shoot the donkey I iden-
tified demonstratively. What this example brings out is that there is a hierarchy of
ways of identifying objects for the purpose of acting upon them, and that demon-
stratives are at the base of this hierarchy. To act on 'my donkey' I typically need
some further identification of the object as 'that donkey'; but to act on 'that don-
.key' I do not need any further identification of the donkey. Putting it like this does
not mean that demonstrative contents are always needed in the explanation of
action on an object. What the relevant hierarchy is, depends on the specific action
in question: demonstrative identification is not to the point when it comes to vot-
ing for someone or ringing them up. And even in the case of an ordinary action,
such as shooting, or reaching and grasping, the fact that demonstratives are at
the base of the relevant hierarchy does not mean that action on the object is impos-
sible without the use of demonstrative identification; whatever the special work
done by demonstratives in this case, it is possible that other types of identification
could, on occasion, do that work.
But what is the special work done by demonstratives in action-explanation?
The natural appeal here is to the special connection between demonstrative iden-
tification of an object and experience of the object; demonstratives depend on
perceptual experience of the object as other modes of identification do not. In
fact, it is not just experience of the object that matters: it is consciously singling-
out the thing that is required by demonstrative identification. Suppose that you
and I are passing a tall building with an array of windows. You say to me, 'that
window is the window of my office'. Since I can take in the building at a glance, I
can be sure that window to which you are referring is somewhere in my field of
view, so I am consciously experiencing it. But that is not enough for me to inter-
pret your remark. To know which thing you are talking about, I need not just to
have the thing in my field of view, I have to single it out consciously; I have to
attend to it.
You might wonder whether experience of the object really is essential for an
understanding of the demonstrative. Suppose we consider someone blindsighted,
who also has the building in her blind field. Could this person not be in a position
to understand the demonstrative, 'that window', in just the wayan ordinary subject
does? After all, such a subject could, in principle, despite not being conscious of
the object, be capable of simple actions with respect to the thing, such as pointing,
and capable of verifying simple judgements about the object, by guessing reliably,
for example, whether the window is shut or open, round or rectangular, and so on.
Should we not say that such a subject has after all interpreted the demonstrative,
'that window'? There are two points to make here. One is that this blindsighted sub-
ject has not interpreted the demonstrative in the way we ordinarily do. A simple ~ay
152 John Campbell
to bring this out is to consider again the case in which you and I are looking at the
building, and you say to me, 'that window is the window of my office'. SUppose
that although I do have conscious experience of the whole side of the building in
which the relevant window figures, I have not yet singled out which window it is.
At this stage, I would say that I don't know which window you mean. If you ask
me to try to point to the window, I would say that I can't, because I don't yet know
which one you mean. And if you ask me to try to guess whether the window is
open or shut, I would again protest that I don't know which one you are talking
about. Suppose, though, that you insist that I should try to point, or to guess
whether the window is open. Suppose that at this point it appears that I have
exactly the same capacities as the blindsighted subject, and I am reliably success-
ful in my guesses or attempts. It still seems evident that, since I still cannot con-
sciously single out which window you mean, I do not know which window you
mean. I am not in a position to demonstratively identify the relevant window
myself. None the less, my movements, as when I try to point, are intentional, as
are the movements of the blindsighted subject who tries to point. It is just that in
saying which mode of identification of the object is the mode under which those
actions are intentional, we cannot appeal to a demonstrative mode of identifica-
tion. At most, I am using some such identifier as 'the window in which my inter-
locutor is interested', and somehow that identifier is engaging with my
visuomotor system to yield the pointing. This contrasts with the ordinary case, in
which it is my demonstrative identification of an object that interacts with my
visuomotor system to yield action on the object. And in the ordinary case, it is
demonstrative identification of the object, depending on my experience of the
object, that is at the base of the hierarchy; that provides the most basic identifica-
tion of the object under which my action is intentional.
Activation
of reach
,.,::::.-------...1 Distance ~'--- _ _ ----l
Activation
Object of grasp
Identification
Visual processing
What is the evidence for this distinction? If the distinction is real, it ought to
be possible to find an anatomical basis for it, and it ought to be possible to find
people who have had damage to one pathway but not the other. So are there, for
example, people who have the 'action' pathway intact but the 'perception' pathway
impaired? The patient DF, studied by Goodale and Milner (1995), seems to have
had an intact 'action' pathway and an impaired 'perception' pathway. After carbon-
monoxide induced anoxia, DF was unable to recognize everyday objects or faces,
and could not visually identify simple shapes. She could not even show how big
objects were or in what orientations. So she certainly has impairment of the
'perception' pathway. None the less, she has strikingly intact visuomotor capabil-
ities. When she had to pick up an irregular object, she angled her fingers optimally
for the grip, though she could not say which irregular objects were the same or
different shapes. When she had to post a card through a slit, her movements were
accurate: she oriented the card correctly and the size of her finger grip was correl-
ated with object size. If we ask about the nature of her conscious experience of the
156 John Campbell
objects she sees, it does not seem that conscious experience is what is setting the
parameters for her actions, since she seems to lack any basis for verbal reports of
sameness or difference of shape, size, and orientation, though in action she does
manage to set these parameters correctly. Similarly, Perenin and Rossetti (1996)
looked at a blindsight patient PJG, who was requested to post a card through a slot
in the blind field, or to grasp blocks presented in the blind field. Despite the lack
of awareness, the patient was accurate on those tasks: orientation of the hand
posting the card was correlated with the orientation of the slot, and finger-grip
size in reaching for the blocks was correlated with the size of the
block. So the 'action' pathway was preserved. But the patient was at chance on
the corresponding matching tasks: showing how big the blocks were or what the
orientation of the slit was.
There are many cases which show the opposite dissociation, intact perception
and an impaired action system. In cases of optic ataxia, patients can accurately dis-
play the size of an object by holding up their fingers to show how big it is. Or they
can show the orientation of a slot they can see by holding up a card in the same
orientation. So the 'perception' pathway seems to be preserved. But these patients
cannot accurately grasp the object, and they cannot post a card through the slot in
the right orientation. The 'action' pathway has been impaired (Goodale, 1996).
So much for the clinical evidence for there being two visual pathways; does the
distinction have any anatomical basis? Goodale and Milner propose that the dif-
ference between perception and action pathways maps on to the difference
between the two streams first identified in the macaque monkey by Ungerleider
and Mishkin (1982), between the dorsal stream projecting from the primary
visual cortex to the posterior parietal cortex, and the ventral stream projecting
from the primary visual cortex to the inferotemporal cortex. Ungerleider and
Mishkin proposed that the dorsal stream is for locating objects, whereas the
ventral stream is for the identification of objects. Goodale and Milner (1995)
propose instead that the distinction is between a dorsal stream carrying visuomotor
processing, and a ventral stream carrying 'perceptual' processing. So on this
diagnosis, the deficits in DF are due to impairments in the ventral stream, whereas
the problems of a patient with optic ataxia would be due to impairments in the
dorsal stream.
At this point, we can already state some problems about how to think of our
ordinary understanding of demonstratives. Suppose we consider the blindsighted
patient PJG, who seems to have a preserved 'action' pathway, and who can cor-
rectly set the parameters for action on a thing. PJG is none the less not aware of
the object, so PJG cannot identify the object demonstratively. In contrast, an optic
ataxic who is aware of the object, who has an intact 'perception' pathway, seems
capable of demonstratively identifying the object, but cannot accurately set the
parameters for action on that thing. At this point, the capacity for demonstrative
identification of the object seems to have come apart from the ability to set the
Demonstratives in Action-Explanation 157
parameters for action on that thing. So what connection does demonstrative
identification have with the explanation of action?
The Grounding Thesis, which was intended to explain the role of demonstrative
identification in the explanation of action, has at this point been lost.
Consciousness of the object is one thing, and setting the parameters for your
action on the object is another. The idea of the Grounding Thesis was that the
properties of the object which show up in conscious experience of the object are
also those which simultaneously (a) ground the meaning of the perceptual
demonstrative, and (b) set the parameters for the subject's actions. But the picture
we have reached so far already threatens that idea. So far, it looks as though the
nature of your conscious experience of the object-the 'picture' that you con-
sciously have of it-depends on what is happening in the 'perception' pathway,
whereas it is what happens in the 'action' pathway that sets the parameters for
your action on the thing.
You might say that these dissociations in clinical patients do not touch the
question whether the Grounding Thesis is correct for ordinary people. Can we
find any evidence against the Grounding Thesis for ordinary subjects? Can we
find any evidence for a dissociation between the 'action' and 'perception' pathways
in ordinary subjects? That is, can we find cases in which the parameters for
the subject's action on an object are clearly not being set by the content of experi-
ence of the thing? In fact, there are a number of illustrations of this possibility.
One example, quoted by Jeannerod (1997), is the so-called 'Roeloff effect'. Here a
large frame surrounds a target. The frame moves, while the target remains fixed.
In that case, subjects experience the conscious illusion that the frame is fixed and
the target moving (in a direction opposite to that in which the frame is actually
moving). None the less, when subjects are asked to point to the location of the tar-
get, they are reliably successful (Bridgeman et al., 1981). This implies that the content
of the conscious experience is being determined by the 'perception' system which
is computing the illusion, rather than the low-level 'action' system which is man-
aging to keep track of the object so that pointing to it remains accurate. It also
implies that the parameters for action in an ordinary subject are not being set by
the content of your conscious experience; rather, the parameters for action are
being set by the low-level 'action' system, remote from consciousness.
Again, we might consider the so-called 'Ebbinghaus' illusion, in which you
present the subject with two target circles of the same size. One is surrounded by
a ring of circles each of which is larger than the target circles. The other target
circle is surrounded by a ring of circles each of which is smaller than the target
circles. The target surrounded by the ring of larger circles is typically experienced
as being smaller than the target surrounded by a ring of smaller circles. The illusion
is preserved when the circles are not simply drawn on a page, but realized by thin
solid discs such as poker chips. And the illusion can be manipulated so that two
target circles of different sizes are experienced as being the same size. This is done
158 John Campbell
by surrounding the smaller target with a ring of smaller circles and surrounding ,
the larger target with a ring of larger circles. Yet although the perceptual illusion
is robust, people asked to pick up the discs set the parameters for their actions
correctly. People scale their grip according to the actual sizes of the objects. So the
grip size is the same when the two targets are really of the same size, and different
when the two targets really are of different sizes, perceptual illusions notwith-
standing (Goodale, 1996; Agliota et al., 1995). Once again, it seems that for ordin-
ary subjects, the parameters for action on a demonstratively identified object are
not being set by the content of the subject's experience of the object.
Finally, the same point can be made by considering experiments in which
a subject is asked to point towards a target which is moved during a saccadic I
eye-movement of the subject's (that is, the rapid jump in position of the eye that
ordinarily occurs several times a minute). The subject adjusts his or her pointing
to keep track of the object, though there is no conscious awareness of any change
in the position of the target (Bridgeman et al., 1975). Similarly, subjects asked
to reach for a target, which was then moved during a saccade, adjusted their reach
to compensate for this movement, but were unaware of the change in position of
the target (Goodale et al., 1986; Jeannerod, 1997). Again, it seems that for ordinary
subjects, (a) the content of conscious experience of the target is not being deter-
mined by the 'action' system, since the target is experienced as stationary but the
'action' system registers its movement, and (b) the parameters of your movement
are being set by the 'action' system rather than by the content of your experience.
In this case as in the previous cases, it seems that the Grounding Thesis has to be
abandoned. This, though, means that we do not have any explanation for why
perceptual demonstratives should have a basic place in the explanation of action.
If the conscious perception that you use to interpret a perceptual demonstrative
is not what sets the parameters for your action, why should there be anything par-
ticularly basic about perceptual demonstratives in the explanation of intentional
action?
REFERENCES
AGLIOTI, S., DE SOUZA, J. F. X., and GOODALE, M. A. (1995), 'Size-contrast illusions deceive
the eye but not the hand', Current Biology, 5: 679-85.
ALLPORT, A. (1989), 'Visual attention', in M. 1. Posner (ed.), Foundations of Cognitive
Science. Cambridge, Mass.: MIT Press.
AUSTIN, J. 1. (1964), 'A plea for excuses', in Donald F. Gustafson (ed.), Essays in
Philosophical Psychology. London: Macmillan.
164 John Campbell
BRIDGEMAN, B., HENDRY, D., and STARK, L. (1975), 'Failure to detect displacement of the
visual world during saccadic eye movements', Vision Research, 15: 719-22.
- - KIRSCH, M., and SPERLING, A. (1981), 'Segregation of cognitive and motor aspects of
visual function using induced motion', Perception and Psychophysics, 29: 336-42.
GIBSON, J. J. (1979), The Ecological Approach to Visual Perception. Hillsdale, NT: Erlbaum.
GOODALE, M. A. (1996), 'One visual experience, n1any visual systems', in T. Inui and
J. L. McClelland (eds.), Attention and Performance XVI. Cambridge, Mass.: MIT Press,
369-93.
- - and MILNER, A. D. (1995), The Visual Brain in Action. Oxford: Oxford University
Press.
- - PELISSON, D., and PRABLANC, C. (1986), 'Large adjustments in visually guided reach-
ing do not depend on vision of the hand or perception of target displacement', Nature,
320: 748-50.
JEANNEROD, M. (1994), 'The hand and the object: the role of posterior parietal cortex in
forming motor representations', Canadian Journal of Physiology and Pharmacology, 72:
525-34.
- - (1997), The Cognitive Neuroscience ofAction. Oxford: Blackwell.
PEACOCKE, C. (1981), 'Demonstrative thought and psychological explanation', Synthese, 49:
187-217.
PERENIN, M. T., and ROSSETTI, Y. (1996), 'Residual grasping in a hemianopic field: a fur-
ther instance of dissociation between perception and action', Neuroreport, 7: 793-7.
PERRY, J. (1993), 'The problem of the essential indexical', in his The Problem ofthe Essential
Indexical and Other Essays. Oxford: Oxford University Press.
SHOEMAKER, S. (1984), 'Self-reference and self-awareness', in his Identity, Cause and Mind.
Cambridge: Cambridge University Press.
UNGERLEIDER, L. G., and MISHKIN, M. (1982), 'Two cortical visual systems', in D. J. Ingle,
M. A. Goodale, and R. J. W. Mansfield (eds.), Analysis of Visual Behaviour. Cambridge,
Mass.: MIT Press.
7
Experimental Approaches to Action
Wolfgang Prinz
This chapter has two sections, theoretical and experimental. In the first section I
will discuss the relative merits of two broad views on human action, sensorimotor
and ideomotor. Sensorimotor approaches treat action as re-action to stimulation.
They date back to Cartesian physiology and are still predominant in both modern
physiology and cognitive sciences. Ideomotor approaches emphasize the inten-
tional and volitional basis of action. They date back to Lotze's and James's discus-
sions of the anatomy of voluntary action and have since then fallen into oblivion.
I will argue for reviving ideomotor theory and outline ways of combining it with
sensorimotor theory within a novel framework. In the second section I will then
discuss some recent experiments from our lab in an attempt to illustrate how
human action can be studied under experimentally controlled conditions and
how the interaction between the perceptual and the intentional basis of action can
be analysed in both experimental and theoretical terms.
1. BASIC ISSUES
1.1 Two Views on Action
The sensorimotor framework regards actions as re-actions, that is, as responses
triggered by stimuli. Strict versions of the approach (like classic behaviourism)
claim that such reduction of responses to stimuli is a necessary and sufficient con-
dition for a full account of action. More liberal versions may also consider a role
for additional factors that cannot be traced back to the actual stimulus situation.
Historically, the sensorimotor strand of theorizing has been dominant in
action theory for decades, if not centuries. One of the early influential sources to
which its origin can be traced back is Descartes's analysis of the relationship
between perception and action. According to Descartes, activity on the motor side
must be understood as response to activity on the sensory side of the mind/brain
(Descartes, 1664). This doctrine has laid the groundwork for the notion of the
sensorimotor arc which has since played an important role in shaping physiolo-
gical theorizing about action. At the same time, this notion has deeply influenced
psychological theorizing as well. This can be shown in a number of independent
historical and theoretical contexts like, for example, the foundation of reaction
166 Wolfgang Prinz
time measurement (Helmholtz, 1852; Donders, 1862), the formation of the
behaviourist programme (Watson, 1913; Hull, 1943) or, more recently, the devel-
opment of the linear stage theory of human performance (Sternberg, 1969;
Sanders, 1980).
The core assumption shared by all brands of sensorimotor theories of action
may be called the stimulus-trigger hypothesis. It holds that, at least in experimental
tasks with well-structured sets of stimuli and responses,l the presentation of the
stimulus is both a necessary and sufficient condition for triggering the appro-
priate response. Accordingly, action is always re-action: it comes into being as a
causal consequence of stimulation.
Unlike sensorimotor views, ideomotor views stress the role of internal (voli-
tional) causes of action and often disregard the role of external (sensory) causes.
Correspondingly, the evidence on which ideomotor theories are grounded does
not come from stimulus-controlled reaction tasks but rather from more open sit-
uations where individuals pursue certain goals and, from time to time, perform
certain actions in an attempt to approach, or achieve them. On this view, actions
are considered creations of the will-events that come into being by virtue of the
fact that people pursue goals and entertain intentions to realize them.
Historically, ideomotor approaches to action have entered the scientific discus-
sion much later than sensorimotor approaches. This may be due to the fact that
they deal with internal (latent) rather than external (manifest) causes of action.
Moreover, their causal role appears to be different. Unlike a stimulus which acts as
a causa efficiens for the responses to follow, a goal seems to come closer to the mys-
terious role of a causa finalis, that is, a causal determinant that, at first glance,
seems to work backward in time. Lotze (1852) and James (1890) were the first to
solve the puzzle, at least in principle. Their solution was based on the distinction
between the physical goal state itself (as achieved through the action-and, hence,
following it in time) and its mental representation (as formed before the action-
and, hence, potentially involved in its causation).2
Though this move solved the puzzle of backward causation it did not help
much in bringing goals to the forefront of action theory and assigning to them the
same status and dignity that everybody (and every theory from the sensorimotor
1 Typical experimental tasks exhibit well-defined sets of stimuli and responses, individuated on the
basis of instructions. Instructions typically specify (i) which stimuli can occur, (ii) which responses
can be selected, and (iii) which rules govern the mapping of responses to stimuli. The minimum
requirement for avoiding ambiguities in individuating stimuli is that the structure of the task is trans-
parent to the experimenter. In many cases it may be transparent to the participant as well.
2 What still remains is the problem of backward specification, that is, of the specification of an
action that causes a desired effect: given a certain intended event, or goal-how can the body move-
ments suited to effectuate that event be determined? This problem is closely related to the issue of
inverse kinematics and inverse dynamics in motor control. However, in the motor literature it applies
to the physics of action, that is, the dynamics and kinematics of body movements (cf. e.g. Rosenbaum,
1991, ch. 6). In ideomotor theory the issue is generalized to also include the 'semantics of action.
Experimental Approaches to Action 167
domain) would find natural to assign to stimuli. More important was-and still
is-their difference in methodological status: stimuli are observable physical
entities, and therefore stimulus information is easy to record and/or manipulate.
Goals, however, are unobservable mental entities that cannot be recorded and
manipulated that way. As long as there is no solution to this problem, the study of
the role of goals is much more delicate than the study of the role of stimuli. As a
consequence, it is not too surprising that we do not dispose of a comprehensive
conceptual framework for understanding the role of goals in action. 3
The core assumption shared by various brands of ideomotor theory may be
called the goal-trigger hypothesis. It holds that goal representations, which are func-
tional anticipations of action effects, play a crucial role in action control. For
instance, Lotze speaks of 'Vorstellungen des Gewollten' (= image of the intended
happenings; Lotze, 1852: 301) and James of the 'bare idea of a movement's sensible
effects' which serves the function of a 'sufficient mental cue' to the movement itself
(James, 1890: 522). Given this crucial role for goal representations it is natural that,
according to ideomotor theory, the proper way to individuate actions is neither in
terms of stimuli nor responses but in terms of goals and their representations. 4
In summary, it seems that goals have not lost much of the elusive character they
used to have in the times of Lotze and James-at least as far as action control is con-
cerned. Though goals pose a major challenge for the sensorimotor framework, this
framework still plays a dominant role in guiding research on human performance.
3 This is not to say that goals have been completely ignored in research on action. First, in the motor
control literature, a substantial body of knowledge has been accumulated on target-related perform-
ance (as in reaching, grasping, hitting targets etc.; cf. e.g. jeannerod, 1997). Targets may be viewed as
goals that are specified in terms of spatial and temporal coordinates. Yet in most of these paradigms
the targets (or some aspects thereof) can be perceived in the same way as the triggering stimuli.
Therefore this literature is only of limited use for ideomotor theories of action whose main point is to
clarify the role of goals that are hidden. Second, the study of goal selection and goal implementation
has ever been one of the key issues in theories of motivation and volition (ef. e.g. Freese and Sabini,
1985; Hershberger, 1989; Gollwitzer and Bargh, 1996). These theories have much to say about the
internal dynamics of processes that precede the action. Yet most of them are virtually silent when it
comes to the details of the operations by which goal representations contribute to the control of the
physical realization of actions. Third, more general theories are available on the nature of the mecha-
nisms for goal-directed control of cognition and action. As concerns cognition, production-rule-based
computational models like ACT' or EPIC provide a central role for goals and goal representations in
their architecture (Anderson, 1983; Newell, 1990; Meyer and Kieras, 1999). As concerns action, there
is a long-standing debate on the role of goal representations in learning and performance, that is, on
how animals learn to build up representations of future goals from consequences of previous actions
and how these goal representations then become involved in action control (Hull, 1943; Greenwald,
1970; Miller et al., 1960). However, these approaches, too, have provided us with broad principles
rather than detailed mechanisms.
4. It should be noted that the goal-trigger hypothesis implies a stronger role for goal representations
than is usually implied in theories about the role of feedback in motor learning and motor control.
These theories also hold that representations of intended action effects are maintained ('Plans'; Miller
et aI., 1960; 'Image of Achievement'; Pribram, 1971). However, their main function is evaluation, that
is, testing observed against expected outcome after the action has been performed. They do not playa
major role in action planning and control.
168 Wolfgang Prinz
1.2 Action Induction
Another challenge for the sensorimotor framework comes from action induction.
By this term I refer to a number of observations that suggest much closer functional
links between perception and action than the standard framework provides. These
links appear to be based on structural resemblances, that is, on (non-arbitrary)
relations based on similarity rather than (arbitrary) relations based on contiguity,
suggesting a functional role for similarity in the mediation between perception and
action (d. Prinz, 1990, 1997b; Hommel et al., 2001).
Consider first some simple examples from everyday activities like, for example,
action imitation, movement synchronization, or sympathetic movements.
Imitating another person's action appears to be mere child's play. Everybody can
do it, but virtually nobody is aware of the delicate functional problem of how the
motor system can generate a movement pattern that resembles the pattern of
movements seen in the other person. There is, of course, a large variety of imit-
ative actions, ranging from instances of imitation of simple bodily gestures to
cases of observational learning of habits, attitudes, or even traits. Yet, in all of these
cases, the imitator's action resembles the model's action in one or the other
respect. In social learning this resemblance may refer to the action's eventual out-
come (like wearing the same clothes as the model), whereas in skill acquisition it
may refer to the kinematics of the movement pattern itself. In a similar vein, spon-
taneous synchronization of body movements with environmental events can be
observed in a number of situations, for example, when people are exposed to
music or other kinds of rhythmic sound patterns. In this case too, it seems that a
certain stimulus pattern induces, on the part of the listener, a movement pattern
that exhibits the same temporal structure, suggesting that the latter is called forth
by the former by virtue of similarity. Sympathetic action arises when a particular
motor activity is induced (or maybe even seduced) by a certain stimulus event in
which the observer has an invested interest. For instance, a person who has just
pushed a bowling ball and is now following its course can often hardly prevent
him or herself from moving his or her hand or twisting his or her b()dy while
watching-at least as long as it is still open whether or not the ball reaches its
intended position or not.
Another example of action induction, which can be considered the spatial
counterpart of synchronization, comes from experimental observations on the
effects of spatial compatibility between stimuli and responses. Though everybody
finds these effects natural, too, they cannot be directly observed in everyday life.
However, they can easily be demonstrated in simple experimental set-ups. As an
example one may think of a disjunctive reaction task with two stimuli and two
responses where on each trial a stimulus light is flashed, either to the left or to
the right of a fixation mark (5/ and 5r> respectively). In response to this stimulus
one of two response keys in front of the Subject has to be operated, either the left
Experimental Approaches to Action 169
one or the right one (R 1 and Rr> respectively). A set-up like this allows for two
tasks, differing in how responses are assigned to stimuli:
SI Sr
compatible: incompatible: X
RI Rr
5 For instance, according to motor theories of speech and motion perception, the perception of
speech and motion is based on mandatory entries in motor modules for the generation of speech ges-
tures and body movements, respectively (ef. e.g. Liberman et aI., 1967; Viviani et aI., 1997). Note that
the notion of a perceptual basis for action representation is much broader in scope than the notion of
an action basis for perceptual representation. An action basis for perceptual representation can only be
claimed for the limited range of perceptual events that can be directly generated through correspond-
ing action (as in speech, movements, etc.). Conversely, since all action goes along with perceptual
consequences, a perceptual basis for action representation can be claimed for any action.
Experimental Approaches to Action 173
The principle of common coding, as I conceive it, adopts the first of these two
views. It implies the notion that actions are represented at the common meeting
place in terms of their sensory or perceptual consequences. This has important
implications for both learning (i.e. the formation of linkages between actions and
their consequences) and performance (i.e. the control of goal-directed action).
As to the second dimension, the information contained in the supposed com-
mensurate codes can be represented at various levels of coding. At the one
extreme, one can think of low-level codes of elementary sensory features of stimuli
and responses. At the other extreme, one can think of high-level codes of complex
semantic features of environmental events and actions. Since we must assume that
coding at all of these levels is involved in all kinds of sensory and motor activity,
the question of interest is not what codes and what levels of coding exist,
but rather at which of these levels the meeting place for commensurate codes is
established.
Empirical evidence and theoretical considerations both argue for high-level
representation and abstract semantic codes. As was indicated above, evidence
from the Simon task suggests that action induction relies on correspondence
between events and actions in distal, extracorporal space and not on correspond-
ence between stimulus and response patterns defined in coordinates of proximal
body anatomy. This suggests a strong role for abstract semantic codes. Theoretical
considerations argue for abstract coding and distal reference, too. In a way,
abstraction must be considered a prerequisite for creating commensurability
among otherwise incommensurate entities. For instance, the sensory representa-
tion of an action that a person performs him or herself will be entirely incom-
mensurate with a perceptual representation of the same action he or she observes
in another person. Since one representation is in terms of kinesthetic features
and the other in terms of visual features, there is no obvious way to match them.
Yet, at a more abstract level of representation, the same two events may be com-
mensurate, for instance, with respect to their kinematic structure (spatio-temporal
pattern) or their semantic content (meaning or goal). It is only at this level where
one could induce the other by virtue of similarity.6
Thus, we can now narrow down the functional locus of the meeting place for
commensurate codes for perception and action as follows. It needs to be a repres-
entational domain where stimulus-related and response-related information are
both coded as environmental events. The difference between the two is that events
are actor-independent whereas actions are actor-dependent. Otherwise, they are
6 One could argue that codes at low and high levels of representation differ from each other in two
independent respects, both referring to the information contained in them, namely, degree ofabstraction
and reference. Degree of abstraction concerns the difference between concrete (sensory) versus abstract
(semantic) features. Reference concerns the difference between proximal versus distal representation.
The two factors may be logically independent. Yet, in functional terms, they are correlated in the sense
that abstract codes with distal reference offer optimal conditions for commensurate coding.
174 Wolfgang Prinz
completely commensurate. They share a common space for semantic, or concep_
tual representation and a common reference system for spatial and temporal
localization. As a result, they are represented as two interrelated components of a
single, coherent stream of meaningful happenings.
7 The structure of action codes may be functional not only in action control but in ordinary percep-
tion as well. Action codes may account for the enactive aspects of perceptual functioning as claimed by
motor theories of perception (cf. n. 5). As goal codes refer to certain environmental events (those that
can Occur as effects of one's own movements), they will not only be activated when these movements
176 Wolfgang Prinz
Since action codes are made up of two basic constituents, action induction can
take two different forms, namely, goal induction and movement induction. As
indicated before, there is often no way to distinguish them. This is because most
of the evidence relies on examples where the goal is resident in the movement, so
that the theoretical distinction cannot be mapped to a corresponding difference
in empirical observations (except for Hommel's dissociation experiment which
strongly supports goal induction).
At this point, we may leave it as an empirical question to which extent action
induction relies on goal or on movement induction. The sole constraint derived
from theoretical considerations is that, as a rule, environmental events cannot
directly induce the execution of movements unless corresponding goals are activ-
ated. This assumption is required in recognition of the trivial fact that-at least in
humans-action does not come about as a mandatory consequence of percep-
tion. Even induced action does not follow from perception without intention, that
is, without goal codes switched in.
Matching Codes
The theory of common coding suggests that action induction relies on matching. s
I use the term of matching in order to refer to computational operations suited to
link codes to each other within the boundaries of a common representational
domain. This view implies that any two codes that belong to the same representa-
tional system can get matched to each other, irrespective of their contents. Still,
the fact that the matching is always performed within the boundaries of a closed
representational system has important functional implications for both performance
and learning.
are performed but also when pertinent events are perceived independent of one's own movements.
The perception of such events will then imply two activations: first, activation of corresponding event
codes, and second, by virtue of the links inherent in action codes, activation of codes for movements
suited to effectuate these events. Interestingly, this notion can also be phrased in an entirely different
theoretical jargon. Ecological approaches claim that there can be no perception of events that does not
at the same time specify (more or less explicitly) certain movements afforded by these events. Action
codes could thus be used for building functional architectures for motor theories as well as Gibsonian
(affordance-based) theories of perception.
8 Note that common coding theory does not suggest that matching replaces mapping throughout.
Quite on the contrary, mapping operations must be ubiquitous in the system. First, as indicated above,
there is no reason to believe that the supposed meeting place for commensurate codes is the sole route
that leads from perception to action (and vice versa). There must be a number of other routes as well,
most of them presumably located at lower levels of coding where mapping is the only way to link codes
to each other. Second, we must not forget that, in order to allow for matching at one place, the system
needs to invest in additional mappings in a number of other places. For instance, the claim that events
and actions can only be commensurate when movements are represented through their effects requires
efficient mapping relationships between movements and effects-and vice versa. In the same vein, the
claim that common coding applies to abstract codes implies complex transformations from low- to
high-level coding, which all require the mapping of codes across differently structured and, hence,
incommensurate domains.
Experimental Approaches to Action 177
First, codes that overlap will prime each other. For instance, when a red stimulus
which requires a right-hand response is presented on the right-hand side of the
stimulus screen, the codes representing the event on the screen and the required
action in response to it will overlap with respect to the feature of spatial location
(right-hand side). In a case like this, the event code will partially pre-activate, or
prime the action code, with the strength of priming depending on the degree of
overlap. This priming accounts for action induction, with obvious implications
for performance.
Second, when the codes do not overlap, the matching operation will still help
to specify their mutual relationships within the common representational
domain. This specification is derived from the functional architecture of the com-
mon representational system. On the basis of its inherent representational dimen-
sionality this architecture allows us to compute how far the two codes are away
from each other (functional distance) and on which particular dimensions they
differ (functional difference). This has important implications for learning and
automatization. 9
Consider first the implications for performance. The notion of overlap-based
priming implies that, upon activation of a particular event code, action codes may
get primed or partially primed, dependent on code overlap. This priming, since it
is due to the fact that the codes share strictly identical components in the same
representational domain, will be mandatory and automatic, and the system has no
way to suppress or escape from it.
Obviously, such mandatory action priming can either support or act against
the required mapping, depending on the structure of the task. Consider, for the
sake of illustration, three different trials from the above-mentioned task where the
onset of a coloured light is the imperative stimulus for pressing one of two keys-
a left-hand key in the case of a green light and a right-hand key in the case of a
red light. In the first trial, a red light appears at the centre of the screen. This trial
is neutral in the sense that no obvious overlap and, hence, no priming between
the stimulus code and the response code can be claimed. In the second trial, a red
light appears on the right-hand side. In this case, where stimulus and response
share the same (relative) location, the stimulus code will partially prime the
response code for the correct response, thereby supporting the mapping required
by the instruction. Conversely, in the third trial where the red light is flashed
on the left-hand side of the screen, it will prime the action code for the
incorrect response, thereby acting against the required mapping by creating conflict
9 Both implications do not apply to the mapping of codes (i.e. linking codes across representational
domains). First, if there is no common representational system shared by the codes to be linked, there
can be no overlap and, hence, no priming. Second, mapping across domains has no way to compute
any meaningful relationships between the codes involved and is therefore blind to their functional
difference and distance.
178 Wolfgang Prinz
between the response required by instructions (right-hand) and the response
induced by the stimulus (left-hand).
As a result, in any task, action induction will emerge to the extent a given stimu- :
Ius primes certain responses that compete within the task. lO The neutral status of I I
the first trial derives from the fact that no such priming occurs at all. The differ-
ence between the second and the third trial is that the priming either pertains to
the required response or to one of its competitors, with beneficial and detrimental
effects on performance, respectively.
Consider now the implications for learning. On the one hand, event codes and
action codes are built on the basis of sensory codes representing physical charac-
teristics of stimulus events and action effects. On the other hand, they represent
objects and events with distal reference and in terms of their semantic properties. ,
As a result, the information contained in these codes goes far beyond the sensory'
basis it is built on. In fact, we must assume that the system learns to represent
(physical) events like stimuli and responses through appropriate (semantic) event
codes, both actor-dependent and independent.
This view implies that the way in which stimuli and responses are represented
in this domain is not fixed at all. Instead, it is subject to learning, that is, operations
resulting in structural alterations of the event codes involved. For instance, it is
reasonable to assume that, in the course of practising a specific task, the structure of
the codes involved gets altered so as to optimize the conditions for matching, that is,
for both exploiting the benefits and avoiding the detrimental effects of priming as
much as possible. This goal can be achieved by two means: by creating new overlap
(where priming helps) or by erasing old overlap (where priming hurts).ll Task per-
formance will then become automatic to the degree that the required mappings
become supported by priming and, hence, take the form of matches.
There will certainly be limitations to the structural alterations that can be
achieved through learning. For example, the fact that the Stroop effect does not
10 A different picture may emerge with code overlap across tasks, for instance, in a situation where
two tasks (51 -> R\; 52 -> R2 ) are arranged such that the stimulus for the second task (52) is presented
at the time where the response for the first task (R,) is being selected and prepared. In a situation like
this, where two independent tasks address the same codes at the same time, there should be a way to
protect and isolate these activations from each other. Therefore, we need to distinguish between two
basic cases: we may expect to see induction when stimuli codes overlap with reponse codes within
tasks. However, when the same overlap arises across tasks we may expect to see interference arising from
that overlap (ef. e.g. Mtisseler, 1999; Mtisseler and Hommel, 1997; Stoet and Hommel, 1999; SchuM
et al., 2001)
11 New overlap may be generated in various ways. For instance, one of the two codes may expand
to overlap with (part of) the other, or both may expand and create a new zone of overlap. Conversely,
overlap may become deleted by erasing overlapping features from one of the two codes or from both.
There is a noteworthy parallel between the distinction between creation and deletion of overlap and
the distinction between enrichment and differentiation in perceptual learning (cf. Gibson and Gibson,
1955; Postman, 1955). However, unlike Gibson and Postman who each defended one of them, we hold
that both can be operating at the same time: creating code overlap requires associative enrichment,
whereas selective differentiation is required for deleting code overlap.
Experimental Approaches to Action 179
go away after extensive specific practice seems to suggest that code components
that have been established in life-long learning cannot be ignored, or deleted, for
the sake of a particular task. Whether the same kind of limitation holds for the
creation of new overlap remains to be explored.
2. EXPERIMENTAL EVIDENCE
In the following, I will go through some recent experiments from our lab in order
to indicate ways of studying action induction, that is, similarity-based relation-
ships in the interaction between stimuli, goals, and actions as well as the putative
operations in the underlying codes.
Response 5election
In one of our tasks we wanted to study the relative contributions of symbolic and
iconic response specification in a response selection paradigm (Brass, 1999; Brass
et al., 2000). In this task, participants were always required to select among two
responses, namely to lift, as fast as possible, either the index finger or the middle
finger of their right hands in response to the stimulus presented. In other words,
the choice was between two fingers that could be used to perform the same man-
ual gesture, that is, lifting.
The stimulus was always provided by a hand on the screen which was the mir-
ror image of the participant's right hand. We used two different instructions. One
was iconic, or imitative, requiring the participant to lift the same finger as was
being lifted by the hand on the display. The other instruction was symbolic,
requiring the participant to lift the same finger as was marked by a cross on the
display. With each of these two instructions, three different classes of stimuli
could be presented: baseline, congruent, and incongruent. For baseline stimuli,
only one of the two features was shown (under iconic instructions one finger
was lifted and no cross was shown; under symbolic instructions one finger was
180 Wolfgang Prinz
marked by a cross, and no lifting was shown). For congruent stimuli, the same
finger that was lifted was also marked by a cross, whereas for incongruent stimuli
one finger was lifted and the other one was marked by a cross.
The major results from this experiment can be summarized as follows. First,
when the finger to be lifted was cued by a stimulus finger performing the same
gesture (= baseline/iconic), response times were much shorter than when it was
cued by a stationary stimulus finger marked by a cross (= baseline/symbolic).
This seems to reflect a difference in the degree to which the selection of a given
response can be supported by iconic versus symbolic information, suggesting a
pronounced advantage of iconic matching over symbolic mapping. Second, there
was strong iconic interference with symbolic instructions, and it was observed in
both directions: iconic congruency helped and iconic incongruency hurt (relative
to baseline). Third, there was also an (albeit weaker) symbolic interference effect
with iconic instructions, this time only in the sense that symbolic incongruency
hurt (relative to baseline). In sum, the results suggest that iconic cueing of
response gestures is much more powerful than symbolic cueing.
In a further experiment we decided to weaken the iconic similarity between
stimulus and response gestures, with everything else completely unchanged. In
this experiment the same two instructions and the same three types of stimuli
were combined, but a different response gesture was used throughout: instead of
an upward lift, the task this time required a downward tap of the finger indicated
by the stimulus. Under these conditions, the baseline difference virtually dis-
appeared. Under symbolic instructions iconic incongruency was still effective but
under iconic instructions only a slight effect of symbolic incongruency was pre-
served. In sum, though the strong advantage of iconic over symbolic response
specification had now gone, a substantial impact of iconic incongruency was still
preserved. Clearly, this finding suggests that weakening gesture similarity also
weakens the impact of iconic response specification-without, however, deleting
it completely.
Response Initiation
In another set of experiments, we addressed the issue to which extent iconic cue-
ing can even be effective under conditions of full response certainty, that is, when
the response to be generated is kept constant over a number of trials (Brass, 1999;
Brass et al., 2001). The issue of whether or not compatibility effects are obtained
in simple reaction tasks is controversial in the literature. If any such effects are
observed they tend to be weak and not very robust. This has often been taken to
support the claim that stimulus-response compatibility effects arise in response
selection-an operation that, by definition, is involved in response selection tasks
but not in simple initiation tasks (see Hommel, 1997, for a discussion and
overview). Therefore, if one could show that substantial compatibility effects arise
Experimental Approaches to Action 181
in a task involving no choices and, hence, no response selection at all, this would
challenge the notion that this particular operation is the functional locus where
compatibility effects emerge.
In the experiments participants were presented with a randomized sequence of
two different stimulus gestures, namely, an index finger that would move either up
or down at an unpredictable point within a time window of a few seconds.
Participants had to respond with one of the same two gestures with their index
finger. This time, however, response gestures were always kept constant within
blocks, requiring the same, pre-known response gesture over and over again (say,
moving up), irrespective of the stimulus gestures shown (moving up or down).
Therefore, since stimulus gestures and stimulus onset times were randomized
within blocks, the task exhibited (1) some degree of stimulus uncertainty,
(2) some degree of temporal uncertainty, but at the same time (3) full response
certainty.
Over a number of experiments we observed huge compatibility effects for both
response gestures (somewhat more pronounced for downward than for upward
movements): response gestures were much faster when prompted by correspond-
ing stimulus gestures as compared to non-corresponding gestures. Further, in one
of our experiments we made an attempt at separating two factors that were con-
founded in the first experiment, namely, direction compatibility (upward vs.
downward movement of the finger) and movement compatibility (flexion vs.
extension of the finger). In this experiment, participants ran through four blocks.
Two blocks were an exact replication of the previous experiment. For the two
remaining blocks, we turned the display upside down and mirrored the resulting
image. Under these conditions, two new stimulus gestures emerge: tapping
(= flexion) goes up and lifting (= extension) goes down. Using the same two
response gestures throughout (= tapping down and lifting up), these four blocks
should allow us to decouple the effects of direction and movement compatibility.
Results showed that both of these factors were effective and added to each other:
response gestures were particularly fast when prompted by stimulus gestures with
the same direction and the same movement, and they were particularly slow when
stimulus gestures were different on both dimensions. Performance was inter-
mediate with similarity on one and dissimilarity on the other dimension.
These observations appear to rule out the classical view that similarity-based
compatibility effects can only arise in the process of response selection proper.
Instead, they support the notion that iconic, or imitative, response specification
may be higWy effective even under conditions in which the reaction to be per-
formed is completely prespecified and predetermined. Moreover, they suggest a
functional role for similarity on (at least) two independent dimensions, one refer-
ring to the direction of movements in spatial terms and the other to the pattern
of movements in bodily terms.
182 Wolfgang Prinz
2.3 Spontaneous Action
3. TO CONCLUDE
What can we learn from our experiments about the functional underpinnings of
action control? Two major lessons appear to emerge, supporting the basic claims
made in the first section. The first one is that watching certain events may auto-
matically induce a tendency to produce actions that resemble these events.
Interestingly, this does not only apply to observing other people's actions but may
also extend to physical events that follow from one's own or other people's
actions. Second, and perhaps even more importantly, when people watch actions
or their effects they seem to represent them not only in physical terms (i.e. their
spatio-temporal pattern) but in semantic terms as well (i.e. their underlying goals
and the extent to which they are being achieved). In summary, theoretical consid-
erations and experimental observations both suggest a novel conceptual frame-
work for action control that provides roles for similarity and for goals in the
underlying functional machinery. In order to develop such a novel framework, we
need to combine sensorimotor with ideomotor strands of thought. This is what
common coding theory is about.
REFERENCES
ACH, N. (1905), Ober die Willenstatigkeit und das Denken. Gottingen: Vandenhoeck &
Ruprecht.
Experimental Approaches to Action 185
ANDERSON, J. R. (1983), 'The architecture of cognition', Cambridge, Mass.: Harvard
University Press.
BRASS, M. (1999), 'Imitation and ideomotor compatibility'. Dissertation, University of
Munich.
_ _ BEKKERING, H., WOHLSCHLAGER, A., and PRINZ, W. (2000), 'Compatibility between
observed and executed: comparing symbolic, spatial, and imitative cues', Brain and
Cognirion,44: 124-43.
_ - - - - - (2001), 'Movement observation affects movement execution in a simple
response task', Acta PsychoLogica, 106: 3-22.
DESCARTES, R. (1664), Traitee de I'Homme. Paris: Girard.
DONDERS, F. C. (1862), 'aber die Schnelligkeit psychischer Processe', Archiv fur Anatomie
und PhysioLogie, 657-81.
FITTS, P. M., and BIEDERMAN, 1. (1965), 'S-R compatibility and information reduction',
JournaL ofExperimentaL PsychoLogy, 69: 408-12.
FREESE, M., and SABINI, J. (eds.) (1985), GoaL-Directed Behavior: The Concept ofAction in
PsychoLogy. Hillsdale, NJ: Erlbaum.
GIBSON, J. J., and GIBSON, E. J. (1955), 'What is learned in perceptual learning? A reply to
Professor Postman', PsychoLogicaL Review, 62: 447-50.
GOLLWITZER, P. M., and BARGH, J. A. (eds.) (1996), The PsychoLogy of Action. Linking
Cognition and Morivation to Behavior. New York: Guilford Press.
GREENWALD, A. (1970), 'Sensory feedback mechanisms in performance control: with spe-
cial reference to the ideomotor mechanism', PsychologicaL Review, 77: 73-99.
HELMHOLTZ, H. (1852), 'Messungen tiber Fortpflanzungsgeschwindigkeit der Reizung in
den Nerven', Archiv fur Anatomie, Physiologie und wissenschaftLiche Medizin, 199-216.
HERSHBERGER, W. A. (ed.) (1989), VolitionaL Acrion: Conation and Control. Amsterdam:
North-Holland.
HOMMEL, B. (1993), 'Inverting the Simon effect by intention: determinants of direction and
extent of effects of irrelevant spatial information', Psychological ResearchlPsychoLogische
Forschung, 55: 270-9.
--(1997), 'Toward an action concept model of stimulus-response compatibility', in
B. Hommel and W. Prinz (eds.), Theoretical Issues in Stimulus-Response Compatibility.
Amsterdam: North-Holland, 281-320.
- - MOSSELER, J., ASCHERSLEBEN, G., and PRINZ, W. (2001), 'The theory of event coding
(TEe): a framework for perception and action', BehavioraL and Brain Sciences, 24:
849-937.
HULL, C. L. (1943), PrincipLes of Behavior. New York: Appleton-Century-Crofts.
JAMES, W. (1890), The Principles ofPsychoLogy. New York: Macmillan.
JEANNEROD, M. (1997), The Cognitive Neuroscience ofAcrion. Oxford, UK: Blackwell.
KNUF, L. (1998), Ideomotorische Phiinomene: Neue Fakten [iir ein altes Problem. Aachen: Shaker.
- - ASCHERSLEBEN, G., and PRINZ, W. (2001), 'An analysis of ideomotor action'. Journal
of ExperimentaL Psychology: Genera~ 130 (4): 779-98.
LIBERMAN, A. M., COOPER, F. S., SHANKWEILER, D. P., and STUDDERT-KENNEDY, M. (1967),
'Perception of the speech code', Psychological Review, 74: 431-61.
LOTZE, R. H. (1852), Medicinische Psychologie oder die Physiologie der Seele. Leipzig:
Weidmann.
186 Wolfgang Prinz
MASSARO, D. W. (1990), 'An information-processing analysis of perception and action', in
O. Neumann and W. Prinz (eds.), Relationships between Perception and Action: Current
Approaches. Berlin: Springer-Verlag, 133-66.
MEYER, D. E., and KIERAS, D. E. (1999), 'Precis to a practical unified theory of cognition
and action: some lessons from EPIC computational models of human multiple-task per-
formance', in D. Gopher and A. Koriat (eds.), Attention and Performance, XVII.
Cambridge, Mass.: MIT Press, 17-88.
MILLER, G. A., GALANTER, E., and PRIBRAM, K. H. (1960), Plans and the Structure of
Behavior. New York: Holt, Rinehart & Winston.
MOSSELER, J. (1999), 'How independent from action control is perception?', in
G. Aschersleben, T. Bachmann, and J. Musseler (eds.), Cognitive Contributions to the
Perception of Spatial and Temporal Events. Amsterdam: Elsevier, 121-47.
--and HOMMEL, B. (1997), 'Blindness to response-compatible stimuli', Journal of
Experimental Psychology: Human Perception and Performance, 23: 861-72.
NEWELL, R. (1990), Unified Theories of Cognition. Cambridge, Mass.: Harvard University
Press.
POSTMAN,1. (1955), 'Association theory and perceptual learning', Psychological Review, 62:
438-46.
PRIBRAM, K. H. (1971), Languages of the Brain: Experimental Paradoxes and Principles in
Neuropsychology. Englewood Cliffs, NJ: Prentice Hall.
PRINZ, W. (1984), 'Modes of linkage between perception and action', In W. Prinz and
A. F. Sanders (eds.), Cognition and Motor Processes. Berlin, Heidelberg: Springer-Verlag,
185-93.
--(1987), 'Ideomotor action', in H. Heuer and A. F. Sanders (eds.), Perspectives on
Perception and Action. Hillsdale, NJ: Erlbaum, 47-76.
--(1990), 'A common coding approach to perception and action', in O. Neumann and
W. Prinz (eds.), Relationships between Perception and Action: Current Approaches. Berlin,
New York: Springer, 167-201.
--(1992), 'Why don't we perceive our brain states?', European Journal of Cognitive
Psychology, 4: 1-20.
--(1997a), 'Perception and action planning', European Journal ofCognitive Psychology, 9
(2): 129-54.
--(1997b), 'Why Donders has led us astray', in B. Hommel and W. Prinz (eds.),
Theoretical Issues in Stimulus-Response Compatibility. Amsterdam: North-Holland,
247-67.
ROSENBAUM, D. A. (1991), Human Motor Control. San Diego: Academic Press.
SANDERS, A. F. (1980), 'Stage analysis of reaction processes', in G. E. Stelmach and J. Requin
(eds.), Tutorials in Motor Behavior. Amsterdam: Elsevier, 331-54.
SCHUBO, A., ASCHERSLEBEN, G., and PRINZ, W. (2001), 'Interactions between perception
and action in a reaction task with overlapping S-R assignments', Psychological Research,
65 (3): 145-57.
SIMON, J. R. (1969), 'Reactions towards the source of stimulation', Journal of Experimental
Psychology, 81: 174-6.
--and RUDELL, A. P. (1967), 'Auditory S-R compatibility: the effect of an irrelevant cue
on information processing', Journal ofApplied Psychology, 51: 300-4.
Experimental Approaches to Action 187
_HINRICHS, J. V., and CRAFT, J. L. (1970), 'Auditory S-R compatibility: reaction time as
a function of ear-hand correspondence and ear-response-location correspondence',
Journal of Experimental Psychology, 86: 97-102.
STERNBERG, S. (1969), 'The discovery of processing stages: extensions of Donders' method',
Acta Psychologica, 30: 276-315.
STOET, G., and HOMMEL, B. (1999), 'Action planning and the temporal binding of response
codes', Journal of Experimental Psychology: Human Perception and Performance, 25:
1625-40.
VIVIANI, P., BAUD-Bovy, G., and REDOLFI, M. (1997), 'Perceiving and tracking kinesthetic
stimuli: further evidence of motor-perceptual interactions', Journal of Experimental
Psychology: Human Perception and Performance, 23: 1232-52.
WALLACE, R. A. (1971), 'S-R compatibility and the idea of a response code', Journal of
Experimental Psychology, 88: 354-60.
WATSON, R. I. (1913), 'Psychology as the behaviorist views it', Psychological Review, 20:
158-78.
WELFORD, A. T. (1968), Fundamentals ofSkill. London: Methuen.
--(ed.) (1980), Reaction Times. London: Academic Press.
8
Perception and Agency
Thomas Baldwin
I want to start, not with perception itself, but with belief, and in particular with
Moore's paradox, since it illustrates well the ambivalent relationship between
belief and the will which is itself one aspect of the complex relationship between
perception and the will. Moore's paradox is that I cannot coherently affirm such
things as 'The sun is shining, but I don't believe that it is' or 'I believe that the sun
is not shining, but it is' despite the fact that I know that there are many truths
which I do not believe and that many of my beliefs are false. At one level, the
explanation for this paradox is that a rational thinker who affirms that the sun is
shining is committed to accepting that he believes that the sun is shining, so that
he cannot coherently combine the simple affirmation that the sun is shining with
the denial that he has that very belief or with the affirmation that he has the con-
tradictory belief. In his discussion of Moore's paradox, however, Wittgenstein
comes at the matter from a slightly different direction. 3 He suggests that Moore's
paradox shows that for each of us the self-ascription of present belief is redundant
in the sense that there is no difference from a first-person perspective between
affirming 'I believe that the sun is shining' and just affirming 'The sun is shining'.
An earlier version of this chapter was presented at Oxford and I am much indebted to the discussion
on that occasion, in particular to Bill Brewer, Naomi Eilan, and Martin Davies.
1 J. Locke, An Essay concerning Human Understanding, ed. P. Nidditch (Oxford: Clarendon Press,
So far I have used Moore's paradox to discuss the ambivalent relationship between
belief and the will. In turning to perception, there can be no immediate extension of
exactly the same line of argument: for there is no incoherence in thinking, while I
gaze at a waterfall 'it appears as though the rocks are moving, but they are not'. As
Wittgenstein observed when discussing Moore's paradox: 'One can mistrust one's
senses but not one's own belief'.4 None the less, the intimate connections between
perception and belief lead one to expect that if belief is not subject to the will, then
the content of perception must be similarly independent of it. For although percep-
tions are not beliefs, their content provides us with evidence and thereby reasons for
beliefs: the fact that it appears to me that p is a reason for me to believe that p. Hence
if we were to suppose that the content of perception was subject to the will, we
would have to suppose that such reasons for belief can be created at will: for in
deciding how things appear to us, we would be creating evidence for ourselves. But
since reasons for belief are reasons for the truth of what is believed, evidence cannot
be in this way subject to the will. Evidence which is created at will can provide no
rational support for hypotheses whose truth is independent of the will.
The argument so far is structural: perception could not provide us with evidence
if we regarded it as entirely subject to the will. If one now adds the familiar empiri-
cist thesis that meaning is dependent upon perceptual evidence, it will follow that
a conception of perception as subject to the will be destructive of meaning and
mental content in general. This conclusion is confirmed by an argument of
Wittgenstein's which is not as well known as it deserves to be. The argument occurs
in his Remarks on the Philosophy ofPsychology, ii and begins with a series of questions:
79 Isn't it conceivable that there should be a man for whom ordinary seeing was subject to
the will? Would seeing then teach him about the external world? Would things have colours
if we could see them as we wished?
Wittgenstein does not answer these questions at once; initially he discusses the
sense in which imagery is subject to the will. But then he returns to his questions:
91 Is it conceivable that visual impressions could be banished or called back? What is
more, isn't it really possible? If I look at my hand and then move it out of my visual field,
This long passage is, in effect, an application of the famous Private Language
Argument. The basic thought is that the hypothesis that the content of perception
is subject to the will turns out to be incoherent-like the thought that the time is 5
o'clock on the sun-because it undermines the possibility of perceptual-, or image-,
content. The argument has two stages. In the first (§92), Wittgenstein points out
that the hypothesis undermines the possibility of any shared language for describ-
ing the appearances of things (e.g. as red and round). For pointing at red round
things in order to check that there is agreement in judgement concerning their
colour and shape provides, under this hypothesis, no way of grounding a common
understanding of the language: for who is to know how such things look to another
192 Thomas Baldwin
person if their appearance to him is subject to his will. But without any basis for
agreement in judgement on such obvious matters as shape and colour there can be
no shared language concerning them. This then undermines my confidence that my
own experience is experience of a real world; indeed, once I recognize that my own
experience is subject to my will, I should acknowledge that I have no reason to take
it as experience of a real world at all. Instead the world as I experience it will be just
my own 'fiction' and it will, for example, be for me an open question whether things
themselves are coloured in anything like the ways in which I have decided they are
to appear to me to be (as Wittgenstein suggested in the last sentence of §79).
The second stage of the argument (§93) then takes the point further by criticizing
the assumption that there is any conceptual content at all in the experiences and
images of this subject isolated within their own fictional world. For without any per-
ceptual contents independent of the subject's will and thus justifiable by reference to
observed features of the world itself, the thinker's judgements as to what its images
are images of (e.g. a red cross) are vacuous-not for simple verificationist reasons,
but because there is no basis for a distinction between 'is right' and 'seems right' in
the thinker's putative application of these concepts. We are straightforwardly back on
the territory of the Private Language Argument here; and in this case, as ever, appeal
to a God who 'knows' what images I have simply begs the question as to whether
there is anything for such a 'higher being' to know in the first place.
Wittgenstein's argument shows nicely how the hypothesis that perception is
subject to the will undermines not only serious epistemology (the possibility of
learning about the external world), but also, and thereby, the possibility of any
conceptualization of perceptual content. Furthermore, although the argument
does draw on the thought that voluntary perceptions cannot be regarded as pro-
viding evidence concerning a real world, the argument does not depend on a sim-
ple-minded empiricist verificationism concerning meaning and content; instead,
as with the original Private Language argument, it draws only on the need for
some basis for the 'is right'l'seems right' distinction, and the absence of any such
basis within the purely voluntary experience of the subject envisaged in §§92-3.
Hence, recoiling from Wittgenstein's reductio we have to accept that, in this fun-
damental sense in which imaging is subject to the will, perception is not. Thus any
account of perception which seeks to show how agency, and thus the will, is in
some respects implicated in perception has to respect this conclusion. There are
limits to the involvement of the will in perception.
Having explored these limits, however, I now want to see how much involvement
of the will remains possible in the case of perception. In the case of belief, it will
be recalled, I argued that despite the fact that belief is not subject to the will, the
involvement of beliefs in practical reasonings, and thus indirectly in agency, is an
Perception and Agency 193
essential condition of the fact that beliefs incorporate a commitment to truth. So
there is here a precedent for supposing that some involvement of perception in
agency is not only possible, but essential.
An initial point to make is that it certainly does not follow from the considera-
tions advanced so far that our epistemology has to be based on passive observa-
tion. As Collingwood famously maintained,s an important part of scientific
inquiry is working out the right questions to ask, and this is not a matter of sim-
ply waiting for them to emerge from the data. So there is plenty of space here for
mental and practical activity in constructing hypotheses and experimental meth-
ods to test them. All that follows from the conclusion that perception is not sub-
ject to the will is that at the heart of empirical inquiry there has to be a moment
when the investigator has to stand back in order to wait and see what the results
are-whether the light bends, what proportion of peas are smooth or wrinkled,
or whatever. Empirical inquiry is in this respect a curious enterprise: preparations,
often lengthy and massively expensive, are undertaken-but always and
inescapably in such a way as to construct a space (Heidegger's 'clearing') within
which the results of inquiry become apparent to investigators who cannot dictate
what they are to be. Duhem's famous thesis was that the significance of the results
of empirical inquiries is always open to question; but it does not follow that we
are free to decide what these results were.
This conception of the practical question and answer dialectic ('Man proposes;
nature disposes') suffices, I think, to show that empiricism is not tied to a merely
passive 'spectator theory of knowledge'. But the activity involved in the arrangement
of empirical inquiries is always prior to, and essentially separate from, observation
itself. The issue I now want to pursue is whether there are any more intimate rela-
tionships between perception and agency whereby, somehow, agency contributes to
the structure of perception despite the limits on this involvement explored earlier.
One kind of involvement is in fact discussed by Wittgenstein himself, namely
that which occurs in cases of 'seeing as' where a figure can be seen in two ways-
for example, as a drawing of a duck or as of a rabbit (Jastrow's duck-rabbit).6 For
in a case of this kind, once we are familiar with it, we can switch from one way of
seeing it to the other more or less at will. But the role of the will in this context is
restricted to one of selection between alternatives that perception informed
by past experience makes available to us. It is not that we can see things however
we choose-as if we could see the duck-rabbit as a drawing of a cow if we wanted.
Rather, as with an ambiguous sentence, the figure admits of more than one per-
ceptual interpretation; and once we are familiar with two or more of them, we can
choose between them. These interpretations are not, however, created at will: even
with advice from others one needs to find for oneself the requisite perceptual
7 N. Eilan, 'Consciousness and the Self', in The Body and the Self, ed. J. Bermudez, A. Marcel, and
N. Eilan (Cambridge, Mass.: MIT Press, 1995),337-57.
Perception and Agency 195
If we think of sight and touch, perceptual attention is closely related to voluntary
control of the relevant sense-organs, and this then provides another way in which
the will is involved in perception. But what remains debatable is how deep this
involvement runs-how far, first, attention does involve a capacity for adjustment
of the relevant sense-organs; and second, how far, as a consequence, the existence
of these bodily sense-organs is implicated in perceptual consciousness itself. One
way to think about these questions is to connect them with Quasim Cassam's the-
sis that, as self-conscious subjects, we possess an 'intuitive' awareness of ourselves
as embodied. 8 For if perceptual consciousness brings with it perceptual attention,
and if attention involves a capacity to direct one's sense-organs, then Cassam's
conclusion that we possess an intuitive awareness of ourselves as embodied fol-
lows rather more straightforwardly from the hypothesis of self-consciousness
than he supposes.
A case which shows the need for caution here, however, is hearing; for, intu-
itively, our auditory system is one over which we exercise little voluntary control
and which, for that very reason, is of all our senses the most apparently'disem-
bodied'. It is no accident that Strawson chose to write about a 'sound' world in
order to explore his Kantian fantasies about objectivity and space9-for in this
sense-modality we are, phenomenologically, as unlocated in space as we can be.
Though of course we often turn our head in the direction of a sound, shifts in
auditory attention do not require movement of the ears-when listening to an
orchestra we do not have to physically 'focus' our ears on different sound sources
in order to attend to one type of instrument rather than another; we can do the
selection at a later stage in the auditory processing. It is less clear that we can alto-
gether detach our conception of ourselves as hearing sounds from an awareness
of our ears: Russell was at one time very scornful of Dewey's suggestion that we
can only speak of sounds as 'heard' where we imply that our ears were involved, 10
though he later changed sides on this issue and was, characteristically, equally
scornful about his earlier view. II On balance I think a version of Russell's early
view was right: we can imagine some primitive people thinking that we use our
noses to hear things. Admittedly, this belief might lead to some odd habits, and
it would not take much to correct it; but there does not seem to me anything
incoherent in the supposition.
The case of hearing shows, then, that the argument suggested earlier, that
perceptual attention involves voluntary control of a physical sense-organ and thus
brings with it an intuitive awareness of ourselves as embodied subjects, is too
hasty. Except in the case of touch, perceptual consciousness does not need to be
informed by an awareness of the role of the sense-organs involved. This does not
So far I have suggested that, despite my earlier conclusion that the content
of perception cannot be subject to the will, the will can have an important part to
play in the development and control of perception, particularly in the direction of
perceptual attention. What I now want to discuss is how far a connection with
the will is a condition for the possibility of perceptual content, even though the will
cannot determine this content. This is of course an issue which connects with
many disputed questions concerning the determination of mental content in gen-
eral, but I hope to be able to discuss it without too many contentious commitments.
A good place to start is at the subpersonallevel, at the level of the sensory repres-
entations to be found in simple organisms. Whatever one's favoured theory of
content, it seems plausible to hold that it is only where properties of these repres-
entations are such as to enable them to contribute to the determination of beha-
viour which contributes to meeting an organism's needs or fulfilling its goals
that these representations have any content at all. For without a connection of this
kind with behaviour, a putative sensory representation is no more than a state of
the organism with a structure systematically determined by significant features
of its environment; but a state which is in this wayan effect of the environment is
not thereby a representation of it. Exposure to radiation may well have systematic
effects within an organism; but in the absence of any connection between these
effects and a capacity for appropriate behaviour, these effects are not states which
represent for the organism the level of radioactivity to which it has been exposed.
The content of a sensory representation in a case of this kind is, therefore, a
function both of its distinctive environmental causes and of the way in which it is
apt to contribute to the determination of appropriate behaviour. Different
accounts of 'naturalized intentionality' will fill out in different ways details of the
requisite causation, the appropriate behaviour, and the relationship between the
two. But these need not concern us. What is important here is just that there is an
essential role for the determination of appropriate behaviour. For this sets the stage
for the thesis with which I am primarily concerned, that the possibility of percep-
tual content is dependent upon a role for agency. This thesis is not established
12 Cf. T. Baldwin, 'Objectivity, Causality and Agency', in The Body and the Self, ed. J, Bermudez,
A. Marcel, and N. Eilan (Cambridge, Mass.: MIT Press, 1995), 107-26.
Perception and Agency 197
simply by the conclusion already reached; for that concerns subpersonal systems,
whereas agency requires an agent, and thus at least a subject of action, if not a
person. But it will be clear that in thinking about cases of this latter kind we can
build upon the simpler one.
The first case to think about is that in which perceptual consciousness is 'sub-
jective' in the sense that the perceiver is present within it as its subject. In this case
perceptual consciousness involves the organization of sensory fields in such a way
that the perceiver's 'point of view' emerges as a point of origin within these fields.
This point of origin need not be conceptualized as such; indeed conceptual
thought need not be presupposed at all. What is primarily implied is an egocentric
organization of perceptual content which gives it coherence and order from the
perceiver's point of view, including sufficient integration of short-term memory
with present experience to enable the perceiver to keep track of their relationship
to objects within the perceived environment. But is it also implied that this sub-
ject be an agent, something which we can think of as capable of setting itself to do
something, and thus of forming intentions or giving itself goals which it is also
capable of abandoning when it fails to achieve them?
We certainly connect the subjective structure of what we perceive with
its potential for purposive activity: for example, what is near is what it would be
easy to reach. It may well be argued, however, that this connection just reflects the
truth of the converse of the thesis that is under consideration here, that is,
the truth of the thesis that agency requires subjectivity. The plausibility of this
thesis derives from the fact that agency is essentially first-person (as agents, we set
ourselves to do things) and therefore requires beliefs, and thus perceptions, which
represent the world from a first-person point of view. One cannot set oneself to
chase a cat unless one can locate oneself in relation to the cat; and this latter
capacity involves perceptions with an egocentric structure.
But what of the original thesis, that the subjectivity of perception depends
upon agency? One line of thought starts from the assumption that a subject-
ive perceptual consciousness is one in which we are able to keep track of our
location within a changing environment in which we are ourselves altering our
location as we move around. This ability requires the ability to distinguish
changes in the perceived world which are consequences of one's own movements
from other changes in the environment, and, the suggestion is, our ability to make
this distinction draws on the fact that it is where our movements are purpos-
ive that we are immediately aware of them as our own and are therefore able to
distinguish their consequences from other changes in the perceived world. The
objection to this suggestion, however, is that one can just as readily envisage this
distinction being accomplished through the operation of a subpersonal system
which monitors bodily movements and feeds information about them into the
processing of perceptual input. The appeal to agency here achieves nothing that
cannot be more straightforwardly achieved without it.
198 Thomas Baldwin
This objection seems right to me, and to be of a type which can be readily
replicated to shoot down other suggestions which invoke agency as a condition of
other distinctions inherent in subjective perceptual consciousness. But there
remains one more general consideration in favour of the thesis that subjectivity
requires agency, namely by treating it as an extension of the earlier thesis that sens-
ory states have representational content only in so far as they are capable of con-
tributing systematically to the determination of appropriate behaviour. For if we
apply this thesis specifically to the hypothesis that sensory states have subjective
representational content, then, the suggestion is, this supposed subjective content
should be such that it makes a distinctive contribution to the determination of
appropriate behaviour, and the only obvious way in which such a distinctive con-
tribution can be made is precisely where the behaviour has the first-person structure
that cornes with agency.
The difficulty with this argument is that it appears to involve a verificationist
assumption: put crudely, the thought seems to be that the way to verify that some-
thing has a subjective perceptual consciousness is to find that this consciousness
is ll1anifested in the first-person structure of its behaviour as an agent. But I think
the argument's rationale can be set at a deeper level than that of straightforward
verificationisn1: instead, it rests on the Kantian thought that the first-person struc-
ture of subjective perceptual consciousness is such a fundamental feature of it (an
'a priori' feature of it, in Kant's terminology) that it is not something that can be
grounded within the empirical structure of consciousness alone. Hence it requires
vindication through considerations that look beyond consciousness to its place in
human life. For if this is granted then a connection with agency does look to
be the only plausible 'transcendental condition' for the possibility of subjective
perceptual consciousness, since agency is so deeply implicated in first-person
modes of thought.
This line of argument remains somewhat speculative, as well as bringing with it
the contentious presumptions of Kantian transcendental arguments, which I shall
not seek to discharge here. No doubt, therefore, the argument needs refinement, but
rather than attempt that task here I want to consider, finally, what difference it n1akes
if one assumes that the subject of perception is also a person capable of rational
thought. Clearly, if the preceding argument is a good one, it applies to persons in
particular: personal subjects of perception must be agents. But there is a separate
question as to how far their rationality also brings with it a commitment to agency.
One general argument for this thesis rests on the claim made earlier in this
chapter concerning the fundamental role of practical reasoning in explaining
the commitment to truth inherent in belief. For in taking the subject of percep-
tion to be a person, we take it that their perceptual consciousness provides them
with evidence which gives them reasons for belief. Hence if it is correct to suppose
that the role of such reasons rests on the involvement of beliefs in practical
reasoning, it will follow that a personal subject of perception must also be a person
capable of rational action.
Perception and Agency 199
This general line of thought is rather abstract, but it can be filled out by con-
sidering what is involved in the mastery of some of the most familiar concepts
which an ordinary person employs on the basis of their experience, namely prac-
tical concepts such as house, pen, bus, and so on. Since these are concepts which
characterize things by the use that is made of them, it is obvious that someone
who grasps these concepts must have some understanding of these uses; the sub-
stantive thesis is then that this latter understanding requires the ability to use
things of this kind, which is a form of rational action. At its simplest, the thesis is
that to recognize something as a pen one must be able to write with a pen. The
argument for this is that without the ability to use a pen, one will not be able to
grasp the basis for distinguishing between pens and other apparently similar
things; where the basis for a classification is essentially practical, the ability to clas-
sify correctly requires the practical ability in question. Admittedly, in the case of
some sophisticated but distinctive pieces of equipment, such as an electron micro-
scope, one might have the ability to identify the equipment without the ability to
use it; but one's understanding in such cases is limited and derivative: without the
ability to use it oneself, one cannot tell when it is working, and one can only iden-
tify it because those who can use it have taught one its characteristic appearance.
Thus in general, to use Ryle's idioms, in the case of these practical concepts,
'knowing-that' presupposes 'knowing-how'.
It may be objected that this case is too easy to be properly exemplary, since a
connection with practice is all too obviously built in to a grasp of practical con-
cepts. But even if the argument in this case is straightforward, the case is none the
less important because these concepts are characteristic both of our everyday 'com-
mon sense' understanding of the world and of a scientific understanding which
acknowledges the role of experimental equipment in its justifications. In both cases
practical concepts make an ineliminable contribution to our epistemology, and
with them comes a necessary connection between rationality and agency.
Anyway, the general thesis here can also be supported by considering what is
involved in mastery of natural kind concepts. These are certainly not practical
concepts; instead they are concepts which purport to identify fundamental kinds
within a theoretical framework that unifies and explains a domain of inquiry such
as chemistry. None the less there is considerable intuitive plausibility in the prag-
matist hypothesis that mastery of these concepts requires the ability to undertake
the kinds of empirical inquiry which justify their application. As it stands, this is
clearly too strong a demand, since one can have a reasonable understanding of the
basic concepts of chemistry, and in particular of the table of elements, while lacking
any serious experimental experience and expertise. But the way to reinstate the
pragmatist thesis is by distinguishing a first-hand from a derivative understanding
of the concepts in question. 13 The derivative understanding of chemistry which
13 Cf. H. Putnam, 'The Meaning of"Meaning" " reprinted in his Philosophical Papers, ii (Cambridge:
Cambridge University Press, 1975), 215-71, esp. 227-8.
200 Thomas Baldwin
most of us possess is based upon familiarity with textbooks and so on;
but those who write these textbooks rely on the authority of investigators with a
first-hand understanding to which the pragmatist thesis does plausibly apply.
A simple argument for this pragmatist thesis would be that mastery of the
theoretical concepts involved requires a capacity for knowledge of the relevant
features of the world which can be obtained, first hand, only by empirical inquiry.
The difficulty with this argument, however, is that the demand that knowledge be
possible is, on the face of it, verificationist. But this demand can, I think, be cir-
cumvented by substituting the demand that the concepts involved be employed in
such a way that there can be agreement in judgement concerning their application.
For this will be possible only where first-hand investigators apply these concepts
in situations where others can exercise their own judgment concerning the case
in hand-and these will necessarily be situations in which empirical inquiries
are conducted. Thus even where one is dealing with theoretical concepts such as
natural kind concepts, there is good reason to hold that the ability to apply these
concepts when reasoning on the basis of experience requires rational action, albeit,
for most of us, that exemplified by the investigations of those to whom we defer.
In this final argument I have substituted the requirement that agreement in
judgement be possible for the requirement that knowledge be possible. Such a
requirement obviously belongs within a Wittgensteinian account of concepts and
language-games, and in setting out a brief review of the position put forward in
this chapter this final line of thought can be placed alongside the considerations
which I took earlier from Wittgenstein himself. The earlier claim was that if per-
ception is imagined to be entirely subject to the will, its content will be entirely
idiosyncratic and there will then be no possibility of agreement in judgement even
concerning apparently obvious features of a shared situation; and where there is
no such possibility of agreement in judgement, Wittgenstein argued, there is no
basis for supposing that experience has any conceptual content at all. The final
argument can be reconstructed as starting from the opposite hypothesis: that
the will, so far from being almost omnipotent in our lives, is largely impotent, so
that we are unable to conduct rational activities such as empirical inquiry. It is
then argued that without this capacity we cannot construct situations within
which agreement in judgement is possible; but without such agreement, again,
conceptual thought is not possible. Hence the two arguments frame the con-
straints which govern the relationship between the will and perception: on the one
hand, the contents of perception cannot be dictated by the will; on the other
hand, a person who is a subject of perceptual consciousness needs to be someone
who is capable of rational action.
9
Fractionating the Intentional Control of
Behaviour: A Neuropsychological Analysis
1. INTRODUCTION
This work was supported by grants from the Medical Research Council and the Wellcome Trust (UK).
202 Glyn W Humphreys and M. Jane Riddoch
We begin by discussing frameworks for the intentional control of action, derived
from studies of normal subjects. Such frameworks are then useful for reviewing
neuropsychological evidence on the fractionation of intentional behaviour.
In everyday life, our actions usually correspond to aspects of our general inten-
tional will. The intentional control of action occurs even in instances where
psychologists have shown that the behaviour is targeted at a stimulus that the sub-
ject appears to have no awareness of-an example being patients with 'blindsight',
who may reach with accuracy to a stimulus that they self-report as not being able
to 'see' (e.g. Weiskrantz, 1986). Although targets in such instances appear nQt to be
coded consciously, we would hold that the reaching action itself is intentional and
based on the instructions given to the patients by the experimenters ('just guess
where the object is'). One might even suggest that actions such as reaching and
grasping need to be elicited intentionally, by instruction, in order to be directed to
targets that are not themselves available for intentional report. Initiation of the
motor response itself is contingent on intentional control, even if stimulus pro-
cessing, and the linkage of the stimulus to the response, is not. Thus, once recruited
by intention, the action can be made to a stimulus that the actor seems unaware of.
But, one might counter, this is surely a matter of degree, for some responses do
appear to be generated unintentionally-an example being 'leakages' in non-verbal
behaviour when an actor is telling a falsehood. There is certainly good evidence
that electrophysiological measures, such as evoked potentials or galvanic skin
responses, can be generated unintentionally. In neuropsychology, such measures
have been used to demonstrate unconscious coding of stimuli in a variety of dis-
orders, from visual neglect (Vallar et al., 1991), to achromatopsia (Humphreys
et al., 1992) and prosopagnosia (Bauer, 1984), where patients cannot make inten-
tional discrimination responses to the stimuli involved. Further examples are bal-
ance reactions and eye movements. Balance reactions can be based in part on
variations in patterns of 'optical flow' as we move around the environment (or as
objects move around us), even though people are typically unaware of the informa-
tion that they are responding to (or even that some form of balance reaction
has taken place; see Lee and Aaronson, 1974). Similarly eye movements may be
elicited and directed to the locations where stimuli appear even when we are
directed to delay the movement and to make it in the direction opposite to where
the stimulus appears (as in so-called 'anti-saccade' tasks; see Rafal et al., 2000).
Here eye movements are made contrary to the intended behaviour.
Is it the case then that some behaviours may only be initiated intentionally (e.g.
reaching and grasping), whilst other-perhaps more 'primitive' actions-can be
Fractionating Intentional Control 203
evoked even without intention? Galvanic skin responses, balance reactions, and
stimulus-elicited eye movements, for example, may operate through special-purpose,
hard-wired processes, perhaps represented sub-cortically. Other behaviours, not
hard-wired in, are based on activation within the cortex, and subject to inten-
tional control. This point, on whether some behaviours are critically dependent
on intentional control, is one that we will return to at the close when we discuss
dissociations between actions such as pointing and grasping. For now, we wish to
contrast this dichotomous view (some behaviours are dependent on intention,
others are not), with a more continuous approach. The continuous approach
states that all motor behaviours can be generated unintentionally under some cir-
cumstances (including reaching and grasping, our starting examples); what is
critical is to define the circumstances where intentional control does and does not
operate, and how intentional control of behaviour is effected (when it is).
Support for this continuous approach comes from the study of 'slips of action',
where we make errors that to some degree transgress an intentional goal (see
Reason, 1984, for documentation of such slips, based on diary studies). An example
might be taking a familiar turning on a route instead of following a less familiar road
to a new destination. Such errors can involve actions that are not only learned (e.g.
turning a wheel and changing gear in a car), but that also involve a whole sequence
of motor behaviours (avoiding pedestrians, responding to traffic lights). It seems
unlikely that these actions are generated by special-purpose (sub-cortical) mech-
anisms, and, indeed on the surface, these action errors seem little different in kind
from intended actions. What seems more critical here than the nature of the actions
is the overlap between the stimuli being processed and the specification of the
intention. In particular, action slips often arise when there is some degree of partial
match between stimuli and the intention. In our example of driving, the intention
may involve components such as 'starting along a familiar route', 'taking a new turn
at a particular junction', 'taking particular care when changing gear in this car', and
so forth. The general intention, then, can specify a set of component actions, any
of which can be released by appropriate stimuli. We can speak of the intention
providing a response set that is activated by stimuli (cf. Broadbent, 1971). The con-
trol of action itself, though, is not directly based on the intention but on whether
the excitation of representations in the intended response set is greater than that of
other actions that are activated by the stimuli. If the intended actions are activated
more weakly than actions not in the intended response set, but which are never-
theless strongly linked to the stimulus, then these other actions may be elicited and
an action error results. Figure 9.1 illustrates this idea.
The continuous approach we have outlined makes two points: (i) that there are
not classes of behaviour that are intentional and others that are not; all behaviours
may be generated unintentionally, and (ii) that intention does not directly generate
behaviour, rather it modulates response activation within a system that is sensitive
to environmental (bottom-up) factors as well as intention (top-down control).
204 Glyn W Humphreys and M. Jane Riddoch
Intention
Stimuli:
Start on familiar
Familiar landmarks - - - - - - . . route
~
~~ :r-----------------,
Turn left at :
Ll~~_:~?_~ j
FIGURE 9.1. Outline framework for how intention may modulate our taking
a novel turn on a familiar road
Supervisory attentional
system
Hierarchy
of
Stimulus - - - - e t Response
action
The approach is incorporated into models such as that proposed by Norman and
Shallice (1986). Norman and Shallice distinguish between a 'contention schedul-
ing systen1' (CSS) and a supervisory attentional system (SAS). The CSS contains
response programs that are activated partly by stimuli, in a bottom-up manner,
and partly by the SAS, which modulates activity in the CSS. Within the Norman
and Shallice framework, an action slip would arise when a response is strongly
activated by a stimulus and there is insufficient modulation of other responses
from the SASe The framework is presented in Figure 9.2.
Fractionating Intentional Control 205
This approach is also similar to what have traditionally been labelled 'late-
selection' theories of selection (cf. Deutsch and Deutsch, 1963). Late-selection
theories propose that all stimuli are processed to a high level, with stored repres-
entations being accessed without any limits apart from those imposed by sensory
constraints (e.g. stimuli falling on the periphery of the retina will be less likely to
activate stored representations than those falling on the fovea, due to differences
in acuity). On this view, all possible response associations are accessed by stimuli,
and behaviours are selected from the competition between the activated
responses. Response activation will be modulated both by associated stimulus-
response contingencies and by top-down intention.
In experimental psychology, arguments about whether selection operates 'late'
(after stimuli access stored representations) or 'early' (prior to access to stored
representations) have been waged for the past forty years (see Broadbent, 1958;
Treisman, 1960, for some initial views; see Mack and Rock, 1999, for a later exposi-
tion). Recently, behavioural experiments have been added to by studies using
techniques from cognitive neuroscience (including physiological recordings of
single-cell responses, neuropsychological and electrophysiological studies of
evoked potentials, and studies of functional brain imaging), and, we suggest, some
movement has been made towards resolution of the long-standing question.
There is clear evidence that selection does not only operate at a late stage of pro-
cessing but also on early stages concerned with perceptual analysis of stimuli. To
give but one example, Rees, Frith, and Lavie (1997) used functional brain imaging
to measure activation in a brain region (MT) known to be responsible for coding
motion in the stimulus (see Zeki, 1993). They presented a central word against a
background of moving dots and found that activation of MT by dots was affected
by the degree to which subjects attended to the central word. When subjects had
to perform a 'low-level' task on the word (detect a particular letter) there was
more activation of MT by the background dots than when subjects had to iden-
tify the word. This demonstrates that perceptual processing of stimuli is affected
by ongoing task activity. When we are engaged in a more difficult task there is less
perceptual analysis of irrelevant stimuli than when we undertake an easier task on
a relevant stimulus. It seems that selection of the relevant over the irrelevant
stimulus can be 'early' (at a perceptual stage) or 'late' (even at a response stage),
depending on factors such as the difficulty of the task at hand. Lavie (1995) pro-
poses that task difficulty influences the attentional resources available to process
irrelevant stimuli. When the primary task is difficult there are fewer resources
available to process irrelevant stimuli to high levels than when the primary task is
easy. There is 'early selection' in the former case and 'late selection' in the latter.
We return to this point concerning task difficulty when we consider the factors
governing the control of behaviour in neuropsychological patients.
This empirical work suggests that intentional processes not only influence
selection between competing responses but also selection at earlier, perceptual
206 Glyn W Humphreys and M. Jane Riddoch
stages of stimulus processing. Theories need to be elaborated to account for how
intentions are implemented at these earlier stages of processing, as well as at the
level of response selection. The neuropsychological evidence we present below
supports a view in which processes of stimulus selection and response selection
are functionally separable and even dependent on different neural systems.
Intentional control of behaviour operates through separable stimulus and
response-based mechanisms of selection.
3. NEUROPSYCHOLOGICAL DISORDERS OF
INTENTIONAL CONTROL OF ACTION
As we have noted, slips of action by normal subjects can provide important
insights into the nature of intentional (and unintentional) behaviour. However
(and probably fortunately, as far as our survival is concerned!), these slips of
action occur infrequently; also their occurrence is opportunistic rather than being
under experimental control. For both of these reasons, slips of action do not pro-
vide an ideal database for theory development. In contrast, following brain injury,
patients can have profound deficits in the intentional control of action. The
deficits may even occur with sufficient frequency and systematicity to make them
amenable to experimental investigation.
One form of impairment in intentional action can be found in patients with
what has been termed 'anarchic hand syndrome' (see e.g. Della Sala et ai., 1991).
In this syndrome, patients make hand movements that are not under their voli-
tional control. As one hand goes to unlock a door, the anarchic hand may move
to lock it again-though the patient intends that the door be opened. In one early
case, Goldstein (1908) described a patient making spontaneous movements of the
left hand that were unwilled and that could not be inhibited, and stated that 'the
hand does what it likes'-as if the hand had one form of intention that was dis-
tinct from the intention consciously expressed by the patient. Della Sala et ai.
(1991) argued that a distinction be drawn between anarchic hand actions and
behaviourally similar responses labelled as 'alien hand' symptoms (e.g. Brion and
Jedynak, 1972). In alien hand behaviours, patients fail to report ownership of the
wayward limb and they may not acknowledge when an inappropriate action has
been made. Anarchic hand behaviours, however, are acknowledged by the patient
as being their own, even though not under the control of intention. Our own view
is that whether a behaviour is acknowledged as being a patient's own may be
dependent on several factors, including the perceived mismatch between the
inappropriate action and the goal of the patient, as we discuss below. The distinc-
tion may not necessarily reflect some qualitative difference between patients. We
will use the term anarchic hand simply to describe actions that are not under
willed control by the patient.
Fractionating Intentional Control 207
For the most part, descriptions of anarchic hand behaviours have been acecdotal
and have concentrated on providing an anatomical account of the syndrome. Della
Sala et al. (1994), for example, discuss these behaviours by one hand in terms of
patients having damage within one hemisphere to the neural region involved in the
internal control of action (the supplementary motor area, or SMA), along with
lesions of the corpus callosum that disrupt communication from the unaffected
hemisphere. As a consequence, hand actions made by the damaged hemisphere are
driven by environmental factors rather than the patient's intention. In our own
work, we have attempted to study the nature of these environmental
factors experimentally, to provide a more articulated functional account of how
unintended actions arise. We have asked: are the unintended actions based on
learned stimulus-response associations? Do different factors determine the actions
by each hand? And, can there be partial rather than complete loss of intentional
control? We consider the first two questions to begin with.
Riddoch et al. (1998) conducted an experimental analysis of one patient, ES,
who had cortico-basal degeneration, which could have prevented the SMA within
each hemisphere from modulating action. As a consequence, ES showed aspects
of anarchic behaviour with both hands in everyday life. For example, ES described
how her left hand once struck her aunt at a dinner party, though ES was morti-
fied when this happened! Indeed, in many circumstances, ES sat on her hands to
prevent inappropriate actions from occurring. Hence we conclude that she was
aware when gross errors of action arose.
Our experimental analysis used a very simple task. ES was presented with a cup
on a table in front of her, with the cup positioned in line with either the left or right
side of her body. The position of the handle of the cup varied orthogonally with the
position of the cup with respect to ES, so that a left-side cup could have its handle
on the left or right (across different trials). ES was required to pick up the cup using
the hand aligned with the position of the cup with respect to her body, and to ignore
the position of the handle. However, despite the simplicity of the task, ES made
many errors. These errors typically involved her making cross-body hand move-
ments to pick up the cup, particularly when the handle of the cup was aligned with
the opposite (and task-inappropriate) hand. For instance, her right hand might pick
up a cup on her left side when the handle of the cup faced right. Interestingly, ES did
not seem to be aware that she made errors on these trials-when asked whether she
was doing the task correctly she replied, 'I think so!' Was there then a failure to grasp
the task rule? We think not. ES was able to repeat back the instructions, she showed
generally good comprehension, and she could discriminate her left from her right
side. Moreover, as we elaborate in the next paragraph, her performance could be
altered systematically by varying the task and the nature and orientation of the
stimulus. There were circumstances when the task rule could be followed.
In subsequent experiments, Riddoch et al. demonstrated that the frequency of
inappropriate hand actions made by ES varied with the task and the stimulus.
208 Glyn W Humphreys and M. Jane Riddoch
Incorrect responses were reduced when she had to point to the position of the
handle of the cup on each trial. They were also reduced when we replaced the cup
with a cup-like non-object, formed by glueing together two cylinders (a small one
on the side of a larger one, to act as the handle),1 and when we turned the real
cups upside-down so that they were no longer in a familiar orientation. All of
these manipulations affected inappropriate responses by ES's right hand more
than those made by her left hand. In contrast, her inappropriate left-hand errors
were reduced when there was less spatial uncertainty in the task (e.g. when
responses were made to a constant position, but with the hand of response being
cued randomly by the word left or right). The fact that ES could make appropri-
ate responses when the task or the orientation of the stimulus changed (to become
less familiar) indicates that ES could comprehend the task rules-simply she
found it difficult to implement the rules when the stimulus was strongly associ-
ated with the inappropriate response (e.g. a cup with its handle on the right being
strongly associated with a right-hand grasp response).
ES's right-hand responses were modulated by the task, and by both the famili-
arity of the object (the cup vs. the cup-like non-object) and the familiarity of its
orientation (the upright vs. the inverted cup). The effects of object familiarity
indicate that responses were activated on the basis of learned stimulus-response
associations. The effects of object orientation, though, suggest that associations
were probably not formed between actions and the semantic representations of
the stimuli (that would mediate recognition of the cup as a familiar drinking vessel),
since semantic representations are likely to be indifferent to object orientation (so
the cup is recognized as the same object across different orientations; see Biederman
and Cooper, 1991, for evidence of the lack of an effect of left-right orientation on
object recognition). Rather, responses were activated from learned associations
between visual representations and actions. These learned associations may be
particularly strong for the right hand, given that ES was pre-morbidly right-hand
dominant. The effect of the task is also of interest, since it demonstrates that
intentional processes could be implemented under some circumstances. For
example, ES may have been able to specify a response set from the task instructions
(e.g. for either pointing or grasping), but she had difficulty in modulating the
excitation of actions within the response set. Due to a lack of top-down, inten-
tional modulation, she was liable to make errors by selecting the overlearned (and
highly activated) response rather than the task-based response.
ES's left hand was affected less by variations in the strength of stimulus-
response associations, and more by manipulations of positional uncertainty. This
I Since ES tended to make more correct responses to the handle of the cup-like non-object than to
a real cup, it cannot be argued that her errors with the cup were due to the difficulty of picking up the
cup when its handle was incongruent with the hand required by the task rule (e.g. the difficulty of
picking up a left-side cup with her left hand when its handle faced right). This difficulty would have
occurred with both the real cup and the cup-like non-object.
Fractionating Intentional Control 209
can be accounted for if the right and left hemispheres are influenced by contrast-
ing factors. For instance, the left hemisphere (controlling the right hand) seems
affected by learned stimulus-response associations (see above). The right hemi-
sphere (controlling left-hand responses) may be dominant under conditions of
spatial uncertainty; hence inappropriate left-hand responses were made fre-
quently unless the spatial uncertainty of the target reduced.
In summary, the evidence from Riddoch et al. indicated that unintentional
grasp responses in anarchic hand syndrome could be studied systematically. These
responses were determined by learned relations between actions and visual repre-
sentations of stimuli (for the right hand) and by the spatial uncertainty of the
target (for the left hand). Some effects of intention remained apparent, though,
since changing the response reduced the frequency of unintentional grasp actions,
but such intentional effects were not sufficient to overcome the influence of over-
learned stimulus-response associations on behaviour.
Riddoch et al. also tested whether the inappropriate responses generated in their
experimental set-up were indeed unintentional. In this test, ES was asked to make
grasp responses across her mid-line to blocks aligned to the left and right sides of
her body: make a right-hand response to a block on the left and a left-hand response
to a block on the left-exactly the errors she made frequently in the experiment with
cups. Despite ES being physically able to make these reaching responses (shown by
the errors with the cups), we were in fact unable to teach her the cross mid-line task
rule-she always made grasp responses using the hand on the side where the target
object fell. Now she was unable to make cross mid-line responses intentionally! This
result is not too surprising. The cross mid-line responses to blocks had to compete
with overlearned tendencies to respond using the hand nearer to the stimulus, and
she then had difficulty in preventing this overlearned response from being made.
This again demonstrates lack of intentional control of action.
The results were also not confined to hand responses. Riddoch et al. (2001)
found that the pattern of performance generalized to foot responses (ES in fact
had anarchic feet!). We performed the equivalent experiment to the study with
cups, but this time ES had to place either her left or right foot into a shoe placed
on her left or right side, irrespective of whether the shoe was for her left or right
foot. She again had great difficulty in preventing an overlearned response to the
stimulus. For example, she often moved her right foot into a right shoe placed on
her left side.
though the work on utilization behaviours has been advanced by showing some
of the circumstances under which such behaviours arise. In addition, and unlike
some patients manifesting utilization behaviour, ES demonstrated some effects of
task instructions on performance. This is evidenced by the contrasting errors
found when pointing rather than grasping responses were required. It was also
clear in other studies that examined ES's ability to select which of two objects to
make an action to. Instead of presenting just one object, Riddoch et ai. (2000a, b)
used two objects and required grasping responses to just one. As before, the task
was to respond to this target using the hand aligned with the side where the object
appeared. The target was cued by its colour or its shape. To test the effects of
response activation, Riddoch et al. sometimes used a target that was less associ-
ated with the response than the distractor-for example, the target might be an
inverted cup and the distractor an upright cup. Even under these circumstances,
ES made very few errors in which a response was made to the distractor. However,
she did make many errors in which the wrong response was made to the target. As
before, these responses reflected learned stimulus-response associations (e.g. to an
inverted cup on the left, with its handle facing right, ES was liable to make a right-
hand grasping response).
These results reveal that ES was relatively good at selecting the object to which
she had to respond, even though she was subsequently impaired at selecting the
correct response. Also relative response activation, from the distractor compared
with the target, had little influence on selection of the target stimulus (though it
clearly influenced response selection). This suggests a dissociation between the
intentional selection of the appropriate stimulus and that of the appropriate
response; intentional response selection was more impaired than intentional
stimulus selection. This contrast, between stimulus and response selection, is con-
sistent with a neuroanatomical distinction drawn by Posner and Petersen (1990).
They propose that there exist different neural networks supporting stimulus- and
response-based selection. Stimulus selection is contingent on a 'posterior' net-
work centred in the parietal lobes, that prioritizes processing of a target object
over other, distractor objects present in the environment. Prioritization is based
on perceptual properties that characterize the target rather than any distractors.
Target features may be primed top-down, by intention (see Chelazzi et aI., 1993,
for physiological evidence for top-down priming). Response selection, however,
is contingent on an 'anterior' network involving the frontal lobes and the anterior
cingulate. Here we propose that selection is based on the relative activation of
items in a response set specified by intention, with this activation also modulated
by intention (biasing selection towards task-appropriate actions). A framework
for these ideas is presented in Figure 9.3.
Let us apply this framework to account for our behaviour in situations with
multiple objects-for example, reaching to pick up a cup of coffee when there are
other breakfast objects on a table. According to the framework, there would first
need to be selection of a target based on specification of its perceptual attributes.
Fractionating Intentional Control 211
(Resources)
(Resources)
---1 Response 31
o Primed by intention i_-:~ Not primed by intention
Following this, responses associated with the selected object are activated, with the
reaching response (hopefully!) being selected for action. The other objects present
would not necessarily activate their associated responses, so that effects of relative
response strength do not have a major impact on stimulus selection. 2
Other neuropsychological studies are informative about the kinds of templates
that can be established for target objects. We conducted a case study with a patient
with unilateral neglect who was impaired at finding a target specified by its name
or even by a salient perceptual property, such as its colour ('find the red object')
(Humphreys and Riddoch, 2001). However, the patient could find a target if we
indicated the action that could be performed with it (e.g. if the examiner gestured
an action before target and distractor stimuli were presented). This intriguing
result suggests that templates can be relatively abstract, even being specified terms
of an intended action which could be matched by perceptual properties that
'afford' that behaviour. These action-based templates may normally be repres-
ented along with templates denoting particular perceptual properties of targets
(colour, size, etc.). In the patient we examined, the effect of the brain lesion was
to disrupt use of specific perceptual templates for search.
5. DISTRACTOR CAPTURE
The overwhelming effects of learned stimulus-response relations on behaviour
cannot only be observed in a patient such as ES, with cortico-basal degeneration,
2 Although we argue that stimulus selection precedes response selection, it may be that there is
some partial activation of the response set before stimulus selection is completed. Nevertheless effects
of response competition on stimulus selection should be small compared with effects of competition
between stimuli having similar stimulus attributes. .
212 Glyn W Humphreys and M. Jane Riddoch
but also in patients with frontal lobe damage and impaired intentional control of
response selection. Riddoch et al. (2000a), for example, replicated the 'cups' experi-
ments with a patient, FK, with bilateral damage to medial regions of the frontal
lobes. Like ES, patient FK made errors by selecting the overlearned rather than the
task-based response when a target cup had its handle oriented towards the usual
but task-inappropriate hand (e.g. reaching with his right hand to a cup on his left
that had its handle facing to the right). FK was also good at rejecting a distractor
and at selecting a target specified by its colour, even when the distractor had a
stronger overlearned response than the target (again, with an inverted target cup
and an upright distractor cup). This supports our distinction between stimulus
and response selection (Fig. 9.3).
There was also one circumstance in which FK made errors by responding to the
distractor rather than the target. This was when the distractor fell in the line of the
trajectory that would be followed by FK's hand as he reached to the target. Under
this circumstance, FK made errors by picking up the distractor rather than the
target, even when the stimuli had different colours (see also Humphreys and
Riddoch, 2000). Does this mean that distractors activated associated responses
along with targets? We think not. This is because the hand that FK used to pick up
the distractor was determined by the orientation of the target not the distractor.
Consider a circumstance in which the target was a right-facing cup and the dis-
tractor, falling in the path of FK's right hand, had its handle facing to the left
(Fig. 9.4). FK made errors by picking up the left-facing distractor with his right
hand. This type of mistake was striking because FK never made errors in which he
responded to a target using a hand incongruent with its handle. It seems that these
distractor errors occurred because distractor objects in the reach trajectory were
Target:
Distractor:
When FK made errors in the 'cups' experiments, he, like ES, failed to acknowledge
that a mistake was made-this held even when he picked up a distractor that lay
in the reach-path to the target. ES, however, did acknowledge other anarchic hand
actions, such as when her hand struck her aunt! She did not have a general prob-
lem in detecting errors in action. We speculate that this discrepancy between her
report of errors in the 'cups' experiment and in everyday life is due to the size of
the disparity between the action made and ES's intentional goals. Patients such as
ES and FK are impaired at implementing intentional control of responses. At best
the patients are able to determine which actions are part of a response set (e.g.
grasping or pointing), but they are poor at modulating activation of the response
set. This may reflect generally poor specification of intentional goals for action.
These intentional goals may be important not only for modulating the activation
of responses, but also for matching against behaviour so that errors are detected.
We suggest that error monitoring involves matching behaviour against specified
intentions. Impoverished representations of intended actions, then, will tend to be
associated with impaired error monitoring. For instance, patients such as ES and
FK may be able to match their action (pick up a cup) with a general intentional
goal, but the precise details of the action (which hand was used) cannot be
assessed because the more detailed intentional information is degraded. ES, how-
ever, remains able to detect a flagrant anarchic hand movement because this
transgresses even a degraded intentional goal.
We have argued, and presented neuropsychological evidence, that many actions can
be effected unintentionally-including, for example, grasp responses to target stim-
uli. Indeed, under some circumstances, actions can be elicited unintentionally even
when a patient cannot generate the same action intentionally (ES's cross-mid-line
reaches being a case in point). In presenting a framework for understanding
Fractionating Intentional Control 215
intentional behaviour, though, we initially discussed the possibility that some
actions are necessarily contingent on intention-they cannot be performed unin-
tentionally. Now, when discussing evidence from the 'cups' experiments, we noted
that task-inappropriate responses made by ES's right hand reduced when point-
ing rather than grasping actions were required (for similar data with FK, see
Riddoch et al., 2000a). Is it possible that right-hand responses differ qualitatively
from other forms of responses, in that they may only be generated intentionally?
ES may have generated few cross-mid-line pointing errors because pointing was
under intentional control. Pointing responses made with the right hand are based
on activation of the left hemisphere and they may serve a unique communicative
purpose, linked to language (see Edwards and Humphreys, 1999; Robertson et aI.,
1995, for some discussion of this). To the extent that language production is under
intentional control, so the same may hold for these pointing responses.
Of course, these last proposals are highly speculative. Nevertheless, there are
neuropsychological data indicating that pointing and grasping dissociate in a
number of ways besides the differences we have highlighted here. For example,
patients can show unilateral neglect in pointing tasks-when asked to point to the
centre of an object they may be biased towards their non-neglected side. However,
when asked to grasp the same object, their hand may go to the true centre (e.g.
Edwards and Humphreys, 1999; Robertson et al., 1995). In such patients, point-
ing may be based on a conscious representation of the world, which is spatially
distorted. Grasping, in contrast, may be dependent on different representations,
that are not necessarily available for conscious report, and that remain unim-
paired in neglect. These speculations on the differences between pointing and
grasping require further explorations of the relations between the two behaviours.
9. SOME CONCLUSIONS
We have discussed findings from neuropsychological patients that indicate that
intentional control of behaviour can be effected through dissociable neural systems:
a posterior system for intentional selection of objects, and an anterior system for
intentional selection of actions. Patients can have poor intentional selection of
action but relatively preserved intentional selection of target objects. The data show
that unintentional actions can be complex (including reaching and grasping), and
that they can be based on learned associations between stimuli and actions. In
addition, responses can be biased towards perceptual or higher-level properties of
stimuli (e.g. associated word meaning), by varying the task load. However, these
effects of task load appear to occur automatically and do not reflect intentional
processes. We have also proposed that an error-monitoring operation can be effected,
involving the matching of action against a specified intentional goal. If there is poor
specification of the intentional goal, error monitoring will be inaccurate.
216 Glyn W Humphreys and M. Jane Riddoch
The general tenet of our argument has been that intentional behaviours are not
unitary. Self-agency does not depend on a single, indivisible process; rather there are
distinct ways in which different forms of intention are implemented. The imple-
mentation of different intentions takes place through separable neural systems.
REFERENCES
BAUER, R. M. (1984), 'Autonomic recognition of names and faces in prosopagnosia: a neu-
ropsychological application of the Guilty Knowledge Test', Neuropsychologia, 22: 457-69.
BIEDERMAN, I., and COOPER, E. E. (1991), 'Object recognition and laterality: null effects',
Neuropsychologia, 29: 685-94.
BROADBENT, D. E. (1958), Perception and Communication. London: Pergamon Press.
--(1971), Decision and Stress. Oxford: Oxford University Press.
BRION,S., and JEDYNAK, C. P. (1972), 'Trouble du transfert interhemispherique it propos de
trois observations de tumeurs du corps calleux: Ie signe de la main etrangere', Revue
Neurologique, 126: 257--66.
CHELAZZI, L., MILER, E. K, DUNCAN, J., and DESIMONE, R. (1993), 'A neural basis for visual
search in inferior temporal cortex', Nature, 363: 345-47.
COLTHEART, M., PATTERSON, K. E., and MARSHALL, J. C. (1980), Deep Dyslexia. London:
Routledge and Kegan Paul.
DELLA SALA, 5., MARCHETTI, c., and SPINNLER, H. (1991), 'Right-sided anarchic (alien)
hand: a longitudinal study', Neuropsychologia, 29: 1113-27.
--'The anarchic hand: a fronto-mesial sign', in F. Boller and J. Grafman (eds.), Handbook
ofNeuropsychology, ix. Amsterdam: Elsevier.
DEUTSCH, J. A., and DEUTSCH, D. (1963), 'Attention, some theoretical considerations',
Psychological Review, 70: 80-90.
EDWARDS, M. G., and HUMPHREYS, G. W. (1999), 'Pointing and grasping in unilateral visual
neglect: effect of on-line visual feedback in grasping', Neuropsychologia, 37: 959-73.
GOLDSTEIN, K. (1908), 'Zur lehre der motischen apraxie', Journal fUr Psychologie und
Neurologie, 11: 169-87.
HALL, D. A., and RIDDOCH, M. J. (1997), 'Word meaning deafness: spelling words that are
not understood', Cognitive Neuropsychology, 14: 1131-64.
HUMPHREYS, G. W. (1999), 'Integrative agnosia', in G. W. Humphreys (ed.), Case Studies in
the Cognitive Neuropsychology of Vision. London: Psychology Press.
--and RIDDOCH, M. J. (2000), 'One more cup of coffee for the road: object-action
assemblies, response blocking and response capture after frontal lobe damage',
Experimental Brain Research, 133: 81-93.
--(2001), 'Detection by action: neuropsychological evidence for action-defined tem-
plates in visual search', Nature Neuroscience, 4: 84-8.
--TROSCIANKO, T., RIDDOCH, M. J., BOUCART, M., DONNELLY, N., and HARDING, G.
(1992), 'Covert processing in different visual recognition systems', in D. Milner and
M. Rugg (eds.), The Neuropsychology of Consciousness. London: Academic Press.
KUMADA, T., and HUMPHREYS, G. W. (2002), 'Internal vs. external control of visual selec-
tion following frontal lobe damage', Cognitive Neuropsychology, 19: 49-65.
Fractionating Intentional Control 217
LAVIE, N. (1995), 'Perceptual load as a necessary condition for selective attention', Journal
of Experimental Psychology: Human Perception and Performance, 21: 451--68.
LEE, D. N., and AARONSON, E. (1974), 'Visual proprioceptive control of standing in infants',
Perception & Psychophysics, 15: 529-32.
LHERMITTE, F. (1983), 'Utilisation behaviour and its relation to lesions of the frontal lobes',
Brain, 106: 237-55.
MACK, A., and ROCK, I. (1999), Inattentional Blindness. Cambridge, Mass.: MIT Press.
NORMAN, D., and SHALLICE, T. (1986), 'Attention to action: willed and automatic control
of behavior', in R. Davidson, R. Schwartz, and D. Shapiro (eds.), Consciousness and Self-
Regulation: Advances in Research and Theory, iv. New York: Plenum Press, 1-18.
. POSNER, M. I., and PETERSON, S. E. (1990), 'The attention system of the human brain',
Annual Review ofNeuroscience, 13: 25-42.
RAFAL, R. D., MACHADO, 1., Ro, T., and INGLE, H. (2000), 'Looking forward to looking: sac-
cade preparation and control of the visual grasp reflex', in S. Monsell and J. Driver (eds.),
Attention and Performance XVIII. Cambridge, Mass.: MIT Press, 155-74.
REASON, J. T. (1984), 'Lapses of attention in everyday life', in W. Parasuraman and R. Davies
(eds.), Varieties ofAttention. Orlando, Fla.: Academic Press, ch. 14: 515-49.
REES, G., FRITH, c., and LAVIE, N. (1997), 'Modulating irrelevant motion perception by
varying attentionalload in an unrelated task', Science, 278: 1616-19.
RIDDOCH, M. J" EDWARDS, M. G., HUMPHREYS, G. W., WEST, R., and HEAFIELD, T. (1998),
'Visual affordances direct action: neuropsychological evidence from manual interfer-
ence', Cognitive Neuropsychology, 15: 645-84.
--HUMPHREYS, G. W., and EDWARDS, M. G. (2000a), 'Visual affordance and object selec-
tion', in S. Monsell and J. Driver (eds.), Attention and Performance XVIII. Cambridge,
Mass.: MIT Press, 603-26.
--(2000b), 'Neuropsychological evidence distinguishing object selection from action
(effector) selection', Cognitive Neuropsychology, 17: 547--62.
- - (200 1), 'An experimental analysis of anarchic lower limb action', Neuropsychologia, 39:
574-9.
ROBERTSON, I. H., NICO, D., and HOOD, B. M. (1995), 'The intention to act improves uni-
lateral neglect: two demonstrations', NeuroReport, 7: 246-8.
TREISMAN, A. (1960), 'Contextual cues in selective listening', Quarterly Journal of
Experimental Psychology, 12: 242-8.
VALLAR, G., SANDRONI, P., RUSCONI, M. 1., and BARBIERI, S. (1991), 'Hemianopia, hemi-
anesthesia, and spatial attention', Neurology, 41: 1918-22.
WEISKRANTZ,L. (1986), Blindsight. Oxford: Oxford University Press.
ZEKI, S. (1993), A Vision of the Brain. Oxford: Blackwell Scientific Publications.
10
Dual Control and the Causal Theory of Action:
The Case of Non-intentional Action
Josef Perner
1. INTRODUCTION
This chapter is concerned with how the causal theory of action (Davidson, 1963)
in its extended version (Searle, 1983) can be combined with a model of dual con-
trol (Norman and Shallice, 1986) in order to distinguish intended actions from
non-intended actions and those from mere movements or happenings. My spe-
cific focus is on accounting for the fact that one variety of intentional action,
willed (controlled or voluntary) action, is part of a cluster of empirical phenom-
ena while automatic actions belong to a different cluster. The former cluster
comprises conscious awareness and attention, verbal justification, moral respons-
ibility, executive control (tasks on which frontal lobe patients seem specifically
impaired) with the self in charge, and so on, while the latter, surrounding
automatic action, includes potential lack of conscious awareness, lack of verbal
expressibility, reduced responsibility, and bypasses the control through the self.
I would like to thank Naomi Eilan, Johannes Roessler, Elisabeth Pacherie, and Zoltan Dienes for their
most perspicacious comments on earlier drafts of this chapter.
The Case ofNon-intentional Action 219
that can vary in its temporal relationship to the bodily movement as O'Shaughnessy
(1991) suggested.!
With this move two of the named problems can be solved. The intentional
component as integral part of the action (Searle's intention in action) enables dif-
ferentiation of automatic actions without apparent causally prior intention (which
the classic proposal seemed to assume) from mere movements. It also explains the
phenomenology of action. Since there is a mental component inherent in actions
but missing in mere movements, actions feel differently than mere movements. The
third problem of causal deviancy can be illustrated with the young man who intends
to kill his uncle. When driving on his way home in foggy conditions the preoccupa-
tion with his intention to kill his uncle causes him to fatally hit a pedestrian who
happens to be his uncle. Although his intention to kill his uncle was (part of) the
cause of the action that led to the uncle's death (and so by the traditional theory
would count as an intentional act of killing his uncle), he did not kill his uncle inten-
tionally, that is, it was manslaughter not murder. 2 This example can be accounted
for by requiring that the intention stipulates in its content not just the goal to kill
the uncle but also that this particular instance of an action is to achieve the goal. The
example then falls short of being intentional because the young man did not have
the intention to kill his uncle with this particular act of running him over.
A problem that remains even with the refined causal theory, and which Wakefield
and Dreyfus (1991) commented on in relation to Searle's version of it, is that it
distinguishes intentional actions from mere movement, but it leaves no room for
non-intentional action. Under the similar heading of'unintentionaP action' Searle
(1983: 101-3) does discuss the problem of alternative descriptions of one and
the same act. Oedipus intentionally married Jocasta and thereby ended up
marrying his mother. He did not intend to marry his mother. He married his
mother unintentionally.
I The intentional part can start well ahead of the movement and even be aborted before the move-
ment begins, giving the impression of a pure 'prior intention'. Or the intentional part can commence
simultaneously with the movement giving the impression of a pure 'intention in action'. In many cases,
though, the intention begins prior to the movement and then lasts and exerts its influence through to
the completion of the action.
2 Since the preoccupation with the intention to kill his uncle was also an integral part of the chaotic
action that led to the pedestrian's death, the intention to kill could-without further criteria-not be
differentiated from an intention in action. I emphasize this to show that Searle's distinction between
prior intention and intentions in action on its own is not sufficient for cleanly solving the causal
deviancy problems.
3 On Naomi Eilan's suggestion I adopted the expression 'non-intended' for what I have in mind in
4 It is natural to assume that 'non-intentional' actions are as much governed by some motor repres-
entation including goal representations in the sense of Jeannerod in order to assure the coherent
sequence of such complicated action. That is, the changing driving situation must implant local goals
that direct the next sequence of driving actions. It is unlikely that a whole drive home can be
done purely by forward modelling based on the goal specification of 'drive home' at the beginning
of the journey. An interesting question though is whether these local goals would pass the test for goal
The Case ofNon-intentional Action 221
helps distinguish the one case as an intended action and the other as non-intended.
This is one reason for why a dual control model is needed.
representation in animal learning: sensitivity to goal devaluation (Dickinson, 1994). I think it would.
If you have learned since the last drive home to be afraid of a certain traffic situation then this will
make a difference, but at the same time will jolt you out of being absent-minded as this newly devalued
situation will break the old routine.
S Thanks to Zoltan Dienes who reminded me to include the term 'voluntary'.
222 Josef Perner
;
The basic idea is that there are two levels of control which are distinguished by how
they represent action schemata using a distinction developed by Dienes and Perner
(1999; Perner, 1998). At the lower-level action schemata, environmental conditions
and the actions to be performed under these conditions, are represented predication
implicitly (a more precise definition of the kind of representation typically associated
with procedural representation) and the control within and between schemata is lim-
ited to control guided by properties of the representational vehicle (vehicle control).
At the higher level action schemata are represented prediction and fact explicitly
(related to the notion of declarative representation) and control can be guided in
terms of the representational content (content control) of action schemata.
Interplay ofLevels
The relationship between the two levels needs to be specified in more detail. I have
mentioned so far that the higher level exerts control over the lower level by chang-
ing connections at the lower level. The actual production of action, however,
remains the duty of the lower level. In other words, the higher level represents
what should be done and modifies the lower level accordingly. The lower level
then executes the wanted action when the occasion arises. Moreover, the higher
level cannot only represent what ought to be done but can also provide the trig-
gering stimuli for the relevant action schema and monitor what is being done. Out
of this interplay of levels several interesting possibilities can be distinguished that
help meet the different objectives.
7 In animal psychology (e.g. Dickinson, 1994) the term 'habit' is used for a stimulus-response asso-
ciation that is not dependent on a representation of the reinforcing result (goal) of the response as
illustrated by the habit to scratch oneself where it itches. Elisabeth Pacherie (personal communication)
pointed out that this usage of the term does not easily generalize to the case of human habits like
smoking. There needs to be some goal representation (as part of a motor representation in the sense
of Jeannerod) without which no coordinated cigarette scrounging and lighting behaviour would be
possible. A possible way to make these usages consistent would be to suggest that smoking is called a
habit because it is only the motivational state of nicotine deprivation and perception of the social
situation without any goal of wanting to smoke a cigarette that trigger the motor representation (including
motor goal) of finding and lighting a cigarette.
228 Josef Perner
4.2 Conscious Awareness: A 11/z-order-thought theory
Voluntary control is often practically identified with conscious awareness. So
much so, that Jacoby's (1991) process dissociation procedure, a prevalent method
of estimating the relative contributions of conscious versus unconscious knowledge
to task completion, is based on the assumption that voluntary control requires
conscious awareness (e.g. refrain from using a subliminally perceived word to
complete a word stem). In contrast, knowledge whose influence cannot be con-
trolled in this way is assumed to be unconscious. My prime objective here is to
explain to what degree this quasi-identification is justified. To address this issue I
first layout some of my basic intuitions about conscious awareness.
Ned Block (1994, 1995) distinguished different types or aspects of consciousness
two of which are now widely used: phenomenal consciousness and access con-
sciousness. I first want to address access consciousness, how it (and the related
aspects of monitoring and self-consciousness) captures (my) natural intuitions
about conscious awareness, and how it relates to higher-level control. Then I
will address the issue of how phenomenal consciousness might be related to access
consciousness and higher-level control.
Access Consciousness
Access consciousness captures the intuition that conscious contents are promis-
cuously accessible to a variety of inferential processes, in particular to the rational
control of action and speech (Block, 1994: 214). One should add accessibility to
higher-order thoughts (often distinguished as monitoring consciousness) and self-
ascription of these thoughts (self-consciousness). To me, only some of these aspects
seem critical, namely some form of accessibility to higher-order thoughts.
Inferential promiscuity and access to rational action and speech are not defining
of conscious awareness. They are features whose typical linkage with conscious
awareness is in need of explanation.
Accessibility to higher-order thought, one might think, necessitates self-ascription.
However, as we are exclusively concerned with how subjects see the world from
their own perspective there is no need to differentiate one's own thoughts from
those of others. There is no need for a concept of self in order to ascribe the
experienced thoughts to oneself, that is, the self as the focal point of conscious
experiences can remain representationally tacit (Eilan, 1995: 63).
Here are some of my intuitions of why access to higher-order thoughts, a position
currently defended by Rosenthal (1986, 2000) and Carruthers (1996, 2000), is
critical for consciousness. To me it seems incoherent to claim that I am con-
sciously aware of, say, the monitor in front of me without being able to specify the
mental attitude with which I behold the fact of the monitor being in front of me,
that is, whether I am seeing it, or just thinking about it, and so on. This intuition
is also reflected in most experimental investigations of the distinction between
The Case ofNon-intentional Action 229
implicit/unconscious as opposed to explicit/conscious perception, memory, or
learning in healthy people and in patients with neurological deficits such as blind-
sight. Unconscious knowledge is typically inferred by demonstrating existing
knowledge about some state of affairs (e.g. the location of a spot of light) to the
person's (blindsight patient's) surprise. On this intuition, a higher-order thought
that one knows where the spot of light is seems to distinguish conscious knowledge
from the blindsight patient's knowledge.
It is often objected that it remains a puzzle how simply making something second
order could produce consciousness. This objection, however, ignores two critical
points; One point is that not any first-order representation may be accessible to
higher-order thought formation as, perhaps, Carruthers's (1996) position suggests.
We (Dienes and Perner, 1999), for instance, suggest that higher-order thoughts
can only be formed of knowledge with predication- and fact-explicit content. I
will give a brief justification of this claim in my examples below. The second,
related point is that going from having a first-order state to forming a higher-
order thought (HOT) about a first-order state requires representing the first-order
state. This is far from being a trivial step whose magnitude has, perhaps, not been
sufficiently emphasized by higher-order thought theorists. Concentrating on this
point brings into focus the question of how precisely the HOT has to represent the
first-order attitude to create conscious awareness. In an attempt to answer critics
who suggest that consciousness is not dependent on HOTs but consciousness
makes them possible, let me develop a minimalist higher-order account that
satisfies the intuition that conscious knowledge implies knowing that one knows.
The intuition that conscious awareness implies that we know what we know
minimally entails (even for an organism lacking a concept of knowledge) that one
should not be surprised (at least not gain any new knowledge) from renewed infor-
mation about a known fact. That is lacking in blindsight. When a blindsight patient
having 'seen' the spot in his blind field turns his head so that he now sees the spot
in his healthy field of vision he will subjectively gain 'new' information. s In con-
trast, a person with unimpaired vision will gain no new information. What makes
normal sight different from blindsight is not (so much) the ability to attribute
knowledge to oneself, it is the ability to identify the fact in question (location of the
spot) across two different instances of informational events, so that these two
information events can be related to the fact in question and can be appropriately
integrated in a coherent body of knowledge. The integration is only possible if the
fact in question is represented predication explicitly. Only then can the information
coming from the two visual fields be taken as pertaining to a single fact. 9
8 With this example I want to avoid the task demands that arise in the typical case where the blind-
sight patient is told that he had pointed correctly to the spot of light in his blind field. To be surprised,
the patient needs the higher-order understanding that his correct behaviour must be mediated by
'knowledge' about the spot.
9 Campbell (l997a) developed a similar position about the nature of selective attention, a close
associate of conscious awareness.
230 Josef Perner
On this minimal intuition about consciously known facts, predication-explicitness
is prerequisite for conscious awareness. Moreover, if we plausibly require that
conscious awareness should keep factual states of affairs distinct from fiction then
conscious awareness also requires fact-explicitness as a necessary condition.
Are predication- and fact-explicitness sufficient for conscious awareness?
Higher-order thought theorists will argue that it falls short of accounting for our
intuitions. However, we may be closer to higher-order thoughts than it seems.
This hinges on the question of ,,,,hat constitutes our concept of knowledge. As
Gordon (1995) pointed out, when we stick exclusively to our own present per-
spective then there is an isomorphism between what we know and what is a fact
for us. It is only when we want to understand what others know or don't know in
contrast to ourselves (or what we didn't know but know now) that we have to
expand by simulation in terms of perspective switching (as Gordon would argue;
see also Perner, 1999). OUf concerns about consciousness, however, only pertain
to the subject's own perspective. Hence, fact-explicit representation minimally
constitutes some meta-awareness of what one knows, which satisfies the basic
intuition behind the higher-order thought theory that being consciously aware of
a state of affairs entails knowing (assertively thinking) that one knows.
So, to claim that predication- and fact-explicit representation is sufficient for
consciousness, adn1ittedly, falls short of being a typical HOT theory, since there
is no first-order mental state represented by a HOT independent of the explicit
representation of factuality. Looked at in this way, we are dealing with a first-order
theory of consciousness. Yet, the reason why fact-explicitness accounts for con-
sciousness is that it provides the distinction between what I know and what I don't
know-although this distinction remains limited to one's own present perspec-
tive. So, knowing/assertively thinking what is a fact and what isn't amounts to
(second-order) knowing/assertively thinking what one knows and what one doesn't.
In this sense, it is a higher-order thought theory. Caught between the canlps, let it
be known as the llj2-order-thought theory of consciousness.
More needs to be said about the limitations of this kind of consciousness. Fact-
explicitness only provides meta-knowledge within one's present perspective. It
works for the here and now. It might work for the past (as seen from the present)
depending on whether a proper understanding of the past isn't itself dependent
on an understanding of perspective and its causal relations to the remen1bering
mind (Campbell, 1997b; McCormack and Hoerl, 1999; Perner and Ruffman,
1995). What it doesn't allow is to distinguish what one knew then from what one
knows now about then. One's rudimentary HOTs are tied to present knowledge
about present or past. To know about past knowledge (or thoughts) one needs to
separate one's present perspective from one's past perspective and for this one
needs a notion of knowledge independent of factuality.
Perspective problems of this kind occur not only with the past but also with
fiction where a possible world is created in distinction to the real world. One
The Case ofNon-intentional Action 231
could say that 1liz-order consciousness is world specific as in the case of dreaming.
On the 1liz-order model, when I operate in the real world then I am consciously
aware of what I represent fact-explicitly about the real world. When I am dream-
ing I am completely immersed in my dream world. I can be consciously aware of
the fact-explicitly represented dream facts, but I would not be able to represent
that I am only dreaming 10 and not be able to remember my dreams when back in the
real world of waking life. To remember dreams, a full second-order HOT of
knowing that one dreamt these scenes is required. This, it is said, is only possible
for a few of the many dreams we have (those close to being woken up). But those
we do remember, we remember as 'conscious experiences' of the dream world, but
are reluctant to say that we are really conscious while dreaming. This ambivalence
can be accounted for by the 1liz-order theory: we are conscious during dreaming
in the basic sense of fact-explicitness, but not 'fully' conscious as an integral part
of our waking consciousness, because the rudimentary dream consciousness is
tied to the dream world and cannot be integrated with our perspective from waking
life. It can only be integrated if an explicit mental state concept of, for example,
dreaming, becomes available so that the experienced dream facts can be put in
place when seen from the waking perspective outside the dream.
In sum, on the 11Iz-order-thought theory, conscious awareness becomes possible
by representing not just features of the world but by representing that there is a
world that has these features. Predication- and fact-explicitness allows representation
of what features the world has and with that one has an incipient meta-knowledge
of what one knows, as the higher-order thought theories of consciousness require.
It implies that predication- and fact-explicit representation is not only necessary
but also sufficient for, at least, this rudimentary form of access consciousness. Since
predication- and fact-explicitness is-as I have argued-necessary for higher-
order control, we have gained a justification for the intuition that voluntary con-
trol implies consciousness as, for instance, Jacoby (1991) has formalized in his
procedure to separate unconscious from conscious influences. The question
remains what role phenomenal consciousness plays in this picture.
Phenomenal Consciousness
Phenomenal consciousness (Block, 1994) means the 'subjective feel' of our con-
scious experiences or 'What it is like to be a bat' to follow Nagel's (1974) famous
title. A first question for us concerns the relation between subjective feel and
access consciousness. Block has argued that they are independent. However, I find
it difficult to intuit how one can have feel without access. Carruthers (2000)
and Rosenthal (2000b) have argued that access (in particular to higher-order
thoughts) is necessary for phenomenal consciousness and even sufficient.
10 Potential exceptions to this view are reports of lucid dreaming (LaBerge, 1985) where one is
conscious of dreaming during one's dream.
232 Josef Perner
Carruthers, for instance, starts by pointing out that the specification of phenomenal
consciousness as <what it is like to be us' is in need of clarification by distinguish-
ing <what the world is like for us' (worldly subjectivity) from <what our experiences
of the world are like for us' (experiential subjectivity). First-order theories
(Dretske, 1995; Tye, 1995) can account for worldly subjectivity due to the fact that
mental representation presents the world under a certain mode of presentation.
However, this only explains what the world is like for the organism (that the world
takes on a subjective aspect by being presented). It does not account for-what
we really need-experiential subjectivity, namely, what an experience is like (that
the organism's experience takes on a subjective aspect). This only follows from
a higher-order representation of the experience of the world, because the higher-
order representation presents the experience under a certain mode of presentation
and thereby confers a subjective aspect upon the experience.
Here again, we can ask the question of how the HOT has to represent the first-
order state; how that state has to be conceived, so that it provides experiential
subjectivity. Rosenthal and Carruthers provide diametrically opposed answers.
Rosenthal (2000a: 207) says that it needs not to be conceived as mental, just as a
state. This strikes me as too loose a requirement because if the HOT represents the
mental state not as mental but just as a state, then why is the HOT higher order
and not just a first-order mental state. In contrast, Carruthers (2000: 195) requires
that the first-order state be minimally conceived of as giving a perspective, an
appearance of what it is about. This is an extremely strong requirement with
extremely counter-intuitive developmental consequences that Dretske (1995)
has noted before. Existing developmental data suggest that children do not
acquire the notion of appearance much before 4 years (Flavell et al., 1983) and
visual perspective (Masangkay et al., 1974; Flavell et al., 1981) and false belief
as a mistaken perspective (Wimmer and Perner, 1983). Hence, according to
Carruther's theory, children younger than 4 years would not be phenomenally
conscious in our adult sense.
The counter-intuitive nature of this consequence is underlined when we consider
that a host of abilities that are usually closely associated with conscious awareness
develop much earlier at around 9 to 18 months. At this age children start to COlll-
municate verbally (Fenson et al., 1994), follow instructions (Shatz, 1978; Babelot and
Marcos, 1999), plan novel action sequences (Piaget, 1937; Haake and Somerville,
1985; Willatts, 1997), and show persistency in focal attention (Gardner et al.,
2000). They also engage in delayed imitation (Meltzoff, 1988; Bauer, 1996), which
amnesic patients lacking explicit/conscious memory do not do (McDonough et
aI., 1995). This is also the age at which children show various signs of fact-explicit
representation (for review, see Perner, 1991), in particular pretend play (typically
also requiring conscious awareness). So, if the 1I/2-order theory, where conscious
awareness comes with predication- and fact-explicitness, can also serve as a
basis for phenomenal consciousness then we arrive at a much more coherent
The Case ofNon-intentional Action 233
developmental picture and much simpler general theory (see Perner and Dienes,
2003).
To see whether this is possible, let me trace the levels of subjectivity again and
see how they relate to levels of explicitness. Worldly subjectivity is shown even by
predication-implicit representations. If an object's colour registers as 'red' the
world is different for the subject than if this colour does not register or registers
as 'rose'. However, there is no possibility of an internal appreciation that that is
what the world is like. This appreciation only becomes possible, according to
Carruthers, with a second-order representation of how red seems to me. The subject
can think of an experience of green that is distinct from a concurrent experience of
red (Carruthers, 2000: 195).
However, a very similar contrastive appreciation also becomes possible with
predication- and fact-explicit representation, i.e., that the world is red rather than
green. The difference to Carruther's proposal is the following. His proposal
requires the ability to freely contrast an experience of red with an experience of
green pertaining to a single state of affairs, whereas fact-explicitness leaves each
experience tied to a particular state of affairs. However, that still leaves the ability
to contrast a concurrent experience of red (pertaining to the real world) with a
hypothetical experience of green (by considering a counterfactual world).
Predication- and fact-explicitness enables an understanding that the world can be
this way or that way and thereby creates different experiences providing a pract-
ical understanding of different experiences of the world. So we see that the 1'/z-
order-thought theory suggests that with predication- and fact-explicitness one
can enjoy not only a minimalist kind of (albeit perspective bound) access to
higher-order thoughts but also a minimalist kind of phenomenal consciousness
in terms of a practical appreciation of experiential subjectivity.
Summary
By arguing for a 1'/z-order theory of consciousness we get less demanding require-
ments for conscious awareness than from the full-blown higher (second) order
theories (Carruthers, 2000; Rosenthal, 1986). Nevertheless, we still capture the
intuitions behind these theories, in particular the notion that conscious knowledge
of a fact implies some knowledge that one knows that fact. Following on from
Carruther's argument about experiential subjectivity, it can also account for a
rudimentary, practical (but not conceptual) appreciation of the subjectivity of
our experiences. It also enables the more coherent developmental picture that
infants become capable of conscious awareness at around 9 to 18 months (Perner
and Dienes, 2003) when they show signs of predication- and fact-explicit repre-
sentation (e.g. pretend play) and correspondingly start to engage in activities that
are typically not possible without conscious awareness in adults (e.g. verbal
communication) .
234 Josef Perner
4.3 Verbal Report
There is the entrenched intuition that verbal communication requires conscious
awareness. Dennett (1978) even made verbal reportability the hallmark of con-
sciousness. Why should there be this intricate link? If I can drive a car through
dense traffic absent-mindedly while talking to my passenger, why can't I talk
absent-mindedly to my passenger while concentrating on the driving? Perhaps
sometimes this can happen, but then the conversation tends to consist of empty
phrases, in a similar way to my experience of singing bedtime songs to my chil-
dren. I kept singing while thinking about more important matters. When inter-
rupted I had no idea where I was in the song and I had to start again from the
beginning. My impression is that this mindless singing is possible only because '
the song is known by heart and the control is purely at the level of the represen-
tational vehicle, one word follows another regardless of its content. In contrast, in
'intelligent' conversation one needs to control one's flow of words according to the
meaning that one wants to convey and cannot rely on an overlearned standard
sequence. l1 These observations suggest that for 'intelligent conversation' I need
content control. For content control I need predication- and fact-explicitness, that
is, conscious awareness on the 1Vz-order-thought theory of consciousness.
11 Perhaps I should not be so sanguine about my intuitions. It could be that experienced sports
reporters are able to work absent-mindedly even though they do not rehearse a set story but make their
words contingent on what they see. This could be possible without content control if sports reporters-
due to their extensive practice-have developed pathways where the visually observed automatically
activates verbal descriptions of the same content. Similar to reading where the visual input of a word
automatically activates the phonetic realization of that word as demonstrated by the Stroop effect.
The Case ofNon-intentional Action 235
5. NOTES ON SEARLE
5.1 Causal Self-Referentiality
Searle (1983) has made the widely known claim that the content of intentions (of
prior intentions as well as intentions-in-action) needs to specify self-referentially
that the intention be the cause of the action. This claim also made for percep-
tion has come under strong criticism (Armstrong, 1991; Burge, 1991) as over-
intellectualizing basic functions. The aim here is to see to what degree Searle's
claim can be partly substantiated without excessive over-intellectualization.
The Case ofNon-intentional Action 237
Searle resists the suggestion that causal reference need only exist in the objective
satisfaction conditions: the intention of an action must specify the action
which it causes, or else it won't be an intentional action. He insists that causal self-
referentiality is internal to content. Pacherie (2000) took up the challenge to ask
about the psychological implementation of causal self-referentiality within
Jeannerod's theory of motor representations but her answer to the question of
how causal self-referentiality is internal to content remained rather hedged.
My task differs from Pacherie's because I do not look for causal self-reference
within the (intention in actions of the) lower action producing system but in the
interplay of the higher-level system with the lower system, as briefly outlined in
Pemer (1998). The higher level controls the lower level by modifying it or sup-
porting particular schemata. The higher level thus implements (or helps imple-
ment) concrete action tendencies. The mechanism (an implementation schema)12
that implements the prescriptions given by the higher level in the lower level as
action schemata, procedurally represents the causal responsibility of producing
the desired action, just as an ordinary action schema procedurally represents
stimulus-action-outcome sequences.
Let me work out these issues in more detail by focusing on the analogy with the
perception case. When looking at a cat causes the tokening of 'cat' (for instance,
by a CAT node being activated), then that token represents predication-implicitly
the fact that the individual, which caused the tokening, is a cat. It does so because
it represents the fact that the individual in front of me is a cat, not just cat-ness. It
does so according to consumer semantics (Millikan, 1984; Dretske, 1988) because
the token 'cat' has the function of covarying with the presence of a cat and of
directing behaviour in relation to the presence of a cat. It does not represent the
fact that the 'cat' token was caused by the presence of a cat (although it covaries
with being a cause, it does not direct behaviour in relation to being a cause).
In case of an 5-A-O action schema, stimulus 5 elicits the action A in order to
reach the outcome (goal) O. 50 when 5 occurs and the schema gets activated we
can say that the's' token explicitly represents the type of situation 5 and predication-
implicitly represents the presence of a concrete occurrence of 5 (just like in the
case of the cat). Corresponding tokens 'a' and '0' and the 's-a-o' connection explicitly
represent the action types, outcome (goal) types, and the transformation of 5 into
o by applying A. Predication-implicitly they represent the particular action A
produced aqd the particular change thereby effected of transforming 5 into O.
But, again, there is no implicit representation of the fact that the token 's-a-o'
causally leads to the emission of action A.
Now consider the implementation schema. It is like an action schema except
that it takes as stimulus condition a declarative representation at the higher level
6. SUMMARY
REFERENCES
ARMSTRONG, D. M. (1991), 'Intentionality, perception, and causality: reflections on John
Searle's Intentionality', in E. Lepore and R. Van Gulick (eds.), John Searle and his Critics.
Oxford: Blackwell, 149-58.
BABELOT, G., and MARCOS, H. (1999), 'Comprehension of directives in young children:
influences of social situation and linguistic form', First Language, 19: 165-86.
BADDELEY, A. (1996), 'Exploring the central executive', The Quarterly Journal of
Experimental Psychology, 49A: 5-28.
--and DELLA SALA, S. (1998), 'Working memory and executive control', in A. C. Roberts,
T. W. Robbins, and L. Weiskrantz (eds.), The Prefrontal Cortex: Executive and Cognitive
Functions. Oxford: Oxford University Press, 9-2l.
--and HITCH, G. J. (1974), 'Working memory', in G. H. Bower (ed.), The Psychology of
Learning and Motivation. New York: Academic Press, 47-89.
BARGH, J. A., and CHARTRAND, T. L. (1999), 'The unbearable automaticity of being',
American Psychologist, 54: 462-79.
BAUER, P. J. (1996), 'What do infants recall of their lives?', American Psychologist, 51: 29-41.
BIANCHI, L. (1922), The Mechanism of the Brain and the Function of the Frontal Lobes.
Edinburgh: Livingstone.
BLOCK, N. (1994), 'Consciousness', in S. Guttenplan (ed.), A Companion to the Philosophy
ofMind. Oxford: Blackwell, 210-19.
240 Josef Perner
BLOCK, N. (1995), 'On a confusion about a function of consciousness', Behavioral and Brain
Sciences, 18 (2): 227-87.
BRAND, M. (1984), Intending and Acting. Cambridge, Mass.: MIT Press. A Bradford book.
BURGE, T. (1991), 'Vision and intentional content', in E. Lepore and R. Van Gulick
(eds.), John Searle and his Critics. Oxford: Blackwell, 195-213.
BURGESS, C. A., KIRSCH, I., SHANE, H., NIEDERAUER, K. 1., GRAHAM, S. M., and BACON, A.
(1998), 'Facilitated communication as an ideomotor response', Psychological Science,
9: 71-4.
CAMPBELL, J. (l997a), 'Sense, reference and selective attention', Aristotelian Society, Suppl.,
71: 55-74.
--(1997b), 'The structure of time in autobiographical memory', European Journal of
Philosophy,S (2): 105-18.
CARRUTHERS, P. (1996), Language, Thought and Consciousness. An essay in Philosophical
Psychology. Cambridge: Cambrige University Press.
--(2000), Phenomenal Consciousness: A Naturalistic Theory. Cambridge: Cambridge
University Press.
DAVIDSON, D. (1963), 'Actions, reasons, and causes', Journal ofPhilosophy, 60: 685-700.
DELLA SALA, S., MARCHETTI, c., and SPINNLER, H. (1994), 'The anarchic hand: a franto-
mesial sign', in F. Boller and J. Grafman (eds.), Handbook ofNeuropsychology, 9: 233-55.
DENNETT, D. (1978), 'Toward a cognitive theory of consciousness', in C. Savage (ed.),
Minnesota Studies in the Philosophy of Science, 9.
DICKINSON, A. (1994), 'Instrumental conditioning', in N. J. Mackintosh (ed.), Animal
Cognition and Learning. London: Academic Press, 45-79.
DIENES, Z., and PERNER, J. (1999), 'A theory of implicit and explicit knowledge (target article)',
Behavioral and Brain Sciences, 22: 735-55.
DRETSKE, F. (1988), Explaining Behavior: Reasons in a World of Causes. Cambridge, Mass.:
MIT Press. A Bradford book.
--(1995), Naturalizing the Mind. Cambridge, Mass. and London: MIT Press.
EILAN, N. (1995), 'IV*-The first person perspective', Proceedings of the Aristotelian
Society, 95: 51--66.
FENSON, 1., DALE, P. S., REZNICK, J. S., BATES, E., THAL, D. J., and PETHICK, S. J, (1994),
Variability in Early Communicative Development: Monographs of the Society for Research
in Child Development, serial no. 242, vol. 59, no. 5. Chicago: Society for Research in Child
Development.
FLAVELL, J. H., EVERETT, B. A., CROFT, K., and FLAVELL, E. R. (1981), 'Young children's
knowledge about visual perception: further evidence for the Level I-Level 2 distinction',
Developmental Psychology, 17: 99-103.
--FLAVELL, E. R., and GREEN, F. 1. (1983), 'Development of the appearance-reality
distinction', Cognitive Psychology, 15: 95-120.
FRITH, C. D. (1992), The Cognitive Neuropsychology ofSchizophrenia. Hillsdale, NJ: Erlbaum.
GARDNER, J. M., FELDMAN, I. J., KARMEL, B. Z., and FREEDLAND, R. 1. (2000), 'Development
of focused attention from 10 to 16 months: effects of CNS pathology and intra-uterine
cocaine exposure'. Poster presented at the European research conference on brain devel-
opment and cognition in human infants: Development and functional specialization of
the cortex, Agelonde, France, September 2000.
The Case ofNon-intentional Action 241
During the first half-dozen years of life, there are regular-and often remark-
able-changes in children's ability to control their own actions (see Zelazo et al., '
1997, for a review). For example, consider the familiar game of 'Simon Says', in
which players must listen to commands to perform simple actions but are '
required to carry out only those commands that are preceded by the phrase ;
'Simon says'. Young children have a particularly difficult time complying with
these instructions even though they seem to understand them. Usually, they per-
form all of the actions, including those they should ignore. Indeed, 4-year-olds
may err on almost every occasion, and consistent success is typically only achieved
by children several years older (Strommen, 1973).
What might account for these age-related changes in action control? One pos-
sibility is that young children know perfectly well what they are trying to do, and
simply have trouble stopping themselves from acting. This view suggests that the
problem is one of response inhibition, and development in this instance reflects
an increase in inhibitory control. Another possibility is that with age, children
become increasingly aware of the constituents of their actions. According to this
view, developments in awareness permit changes in the way that action plans or
intentions are formulated, which in turn permits children to act selectively in a
wider range of situations.
The proposal that there are developmental changes in the formulation of inten-
tions has to be distinguished from at least one other proposal-namely, that the
changes are simply in children's skills. Changes in skill clearly alter the intentional
actions we can carry out, and perhaps even the intentions we can have. For
example, without the requisite training, none of us can perform a splenectomy, no
matter how much a medical emergency may demand one. Furthermore, when
there is a total lack of knowledge of the skills involved, and hence no chance of
accomplishing the act, it seems wrong for someone to say, 'I intend to do a
splenectomy today'. There must be innumerable changes in skills, from riding
bicycles to tying shoes, that expand the range of actions that children can perform,
and consequently, the range of intentions that they are able to entertain.
Developmental changes in intention will have more extensive implications for
action than changes in specific skills. The Tower of Hanoi, a well-studied measure
Children's Action Control and Awareness 245
of planning, illustrates one such change. At least up until the age of 12 years, the
Tower of Hanoi reveals an increase in the number of subgoals that children can
plan in advance, and this increase then leads to new success on more complicated,
multi-step problems (Welsh, 1991). Increases in the number of subgoals children
can plan and execute enable them to perform more complex actions, extended
over longer periods of time. Indeed, being able to conjoin subgoals has the poten-
tial of affecting any complex action. Thus, changes in planning alter what children
can intend to do, and unlike the acquisition of specific skills, they will have implica-
tions for children's intentional action across a wide variety of situations.
/\
'" ,
, ,
IF 51 52
/\ /\ /\ /\
IF a1 a2 a1 a2 IF a1 a2 a1 a2
(A) I I
(B) I(C) I(D) (A) I (B) I I(C) I(D)
THEN c1 c2 c2 c1 THEN c1 c2 c2 c1
FIGURE 11.1. Tree structures depicting the unintegrated rules of 3-year-olds (left)
and integrated rules of 5-year-olds (right) in the dimensional change card sort
Notes. In the current example. al and a2 refer to aspects of the test cards (blue triangle, red circle).
cl and c2 refer to aspects of the target cards (red triangle, blue circle), and sl and s2 are the setting
conditions of shape and colour.
both involve getting to work, and yet they are also conflicting in that the satisfac-
tion of one goal precludes the satisfaction of the other. The DCCS is similar. Both
games involve matching the same test cards, but because of the way the task is
structured, matching a card in one game mismatches it in the other.
The left panel in Figure ILl depicts the CCC characterization of the 3-year-olds'
goal formulations in the DCCS. The two games are shown with the two choices
each. Any given choice on one dimension (match red) results in a mis-sorting on
the other (mismatch triangle). It has been established that 3-year-olds have
knowledge of how to play both games but are unable to select the new rules flexibly
when the instructions change. The account proposes that 5-year-olds can govern
their actions in the task because they are able to add setting conditions, as shown
on the right side of the figure, that allow them to select between the sets of rules.
It is assumed that the setting conditions further specify the goal. Thus, the satis-
faction conditions of the action in the DCCS become something like 'match the
cards for the colour game' and 'match the cards for the shape game'. The further
specification explicitly differentiates the goals, and it allows the correct selection
of rules because when reasoning about what action to take, children can choose
on the basis of an adequately specified goal. In other words, 5-year-olds are able
to change their actions in these situations not because they gain more control over
their responses per se, but because they are better able to specify the goal that they
are attempting to reach.
Recent results by Munakata and Yerys (2001) lend support to this account.
These authors found that when the knowledge questions are made more complex
(where do the blue flowers go in the shape game?), so that they, like the post-
switch sorting questions, require recognition of the more closely specified goal,
3-year-olds often have difficulty on the knowledge questions too. This new finding
252 Douglas Frye and Philip David Zelazo
shows that the complexity of the inferences required is what is important, and:
again reinforces the point that it is not the modality of questions versus the sorting'
response that matters (Zelazo et ai., 1996, expo 4).
Given the explanation that characteristic problems in intentional action may
occur because of the underspecification of goals, it is worth asking whether there
are other cases in which this difficulty might arise. One possibility is the 'anarchic
hand' behaviours exhibited by patients who have suffered various forms of cortical
damage (see Humphreys and Riddoch, this volume). One of the patients dis-
cussed by Humphreys and Riddoch, referred to as ES, made characteristic errors
when asked to use whichever hand was aligned with a cup to pick it up. If the cup
was on the left, but its handle was pointed to the right, then ES tended to violate
the instructions and reach with the right hand. As with 3-year-olds in the DCCS,
this pattern might be the result of an habitual action that is difficult to control.
Alternatively, it might be that ES's difficulty lies in the specification of the goal.
The instruction that the cup must be picked up with the aligned hand has to be
maintained as a part of the goal, but there were indications that it was not. For
example, when the hand to be used was explicitly cued by the word left or right,
errors were reduced. And, as with 3-year-olds in the error-detection version of the
DCCS, ES seemed to judge the cross-reach to be correct, so the failure did not
seem to be a case of doing the wrong thing in spite of recognizing it was wrong.
Another important case that might be understood in terms of CCC theory is
the development of children's theory of mind (see Wellman et ai., 2001, for a
recent meta-analysis). Theory of mind has been taken to mean children's under-
standing of their own and others' mental states. One primary change is the under-
standing of false belief, or the appreciation that someone else can have a view of
the world that is out of accord with reality or the child's own view. Wimmer and
Perner (1983) first demonstrated the effect with a story of a character who placed
a desirable object in one location, and then was absent when the object happened to
be moved. Young pre-schoolers, when asked, predicted that the character would
look for the object where it was currently located.
This outcome seems similar to the ones that have previously been discussed in
that, the child might be underspecifying the intent of the question just as they
underspecify the instructions in the DCCS. That is, instead of attempting to deter-
mine 'where the character should look from the character's perspective to find the
object' they attempt to determine 'where the character should look to find the
object' (Frye, 1992). The CCC theory can be generalized to provide an explanation
of the various theory of mind findings (Frye, 1999,2000). As such, it can account
for the fact that the development of theory of mind shows the same improvement
between the ages of 3 and 5, and for the findings that tests of theory of mind
performance correlate with measures of executive function, including the DCCS
(e.g. Carlson and Moses, 2001; Frye et al., 1995; Perner and Lang, 1999; Zelazo,
Burack, Benedetto, and Frye, 1996; Zelazo et ai., 2002). Together, theory of mind
Children's Action Control and Awareness 253
and executive function span a substantial portion of cognitive development, so
the generality of the CCC theory is apparent.
smiles when you hit it. What is Sally going to do?' This problem is analogous to
the DCCS or the ramp task because there were two setting conditions of causal
system (normal or non-canonical) and within each setting condition, children
had to consider the agent's disposition (nice or mean) in order to predict the
agent's action (pet or hit).
Although the 5-year-olds did well, the 3-year-olds performed poorly on behav-
ioural predictions with the non-canonical sequences. In the normal situation,
children could perhaps take the causal system for granted and ignore it, focusing
only on a pair of relations between disposition and action (if nice then pet, if
mean then hit). In the non-canonical situation, however, the causal system could
not be ignored. In order to make accurate behavioural predictions in this situation,
children needed first to consider the appropriate action-outcome relation (e.g.
hitting produces smiling), and then to consider the agent's disposition in order to
predict the agent's act. Even though 3-year-olds demonstrated that they under-
stood the non-canonical causal system (namely, that hitting the mugwump made
the mugwump happy), they did not select this information when determining what
the actor would do; instead they reasoned from the usual perspective of a normal
causal relation, as if hitting caused harm. This pattern mirrors what has been found
in other studies with other tasks, such as the DCCS and the ramp task, and it indic-
ates that the constraint on 3-year-oids' ability to specify actions limits their ability
to predict other people's behaviour. By 5 years of age, however, children are able to
make accurate predictions even in unusual situations.
Similar age differences emerged for children's moral judgements. Mer the
agent acted, either hitting the animal or petting it, children were asked about the act's
acceptability, 'Was it okay to X?', and about possible punishment, 'Should [the agent]
get in trouble?' Although act-acceptability judgements at all ages (and among
Children's Action Control and Awareness 259
college students) tended to be based on outcome, there was an age-related
increase in the complexity of children's punishment judgements. Three-year-olds'
punishment judgements tended to focus either on the act's consequences or the
agent's disposition, whereas older pre-schoolers and college students tended to
use integrative rules, such as a conjunction rule, in which punishment is given
only if both disposition and consequences are negative. As shown in previous
studies, the use of integrative rules requiring the simultaneous consideration of
two dimensions of judgement first emerges at about 5 years of age or later.
Thus, for both a non-social domain, physical causality, and a social one, moral
understanding, children's level of awareness and specification of action plans and
goals can be seen to make a difference, both for the intentional actions that chil-
dren themselves can carry out and for the actions they understand. These results
demonstrate that the changes in children's understanding of action control will
have effects across a broad range of content domains.
5. CONCLUSION
It has been argued that developmental changes in children's awareness of actions
expand the range of actions that children can perform intentionally. Beginning in
infancy, developmental changes in action control depend on shifts in children's
awareness of the constituents of action (means and goals and their relations) and
their specification within increasingly complex action plans. Young pre-schoolers'
characteristic pattern of perseverative responding when they must follow altered
instructions does not appear to be the result of being unable to stop themselves
from acting (i.e., a lack of inhibition). In these situations, 3-year-olds display
knowledge-action dissociations in which they show an understanding of what
should be done, even though they do not perform the correct actions. At the same
time, however, their performance on tests of error detection indicates that their
difficulty consists in a failure to select the appropriate actions for execution. That
is, when they see the correct actions carried out by another person, 3-year-olds
judge that person to be mistaken.
Our explanation of this pattern is that 3-year-olds underspecify the goals or
satisfaction conditions of the actions indicated in the instructions. Although they
have knowledge of the correct actions, they do not select them because they do
not adequately characterize the goal they are trying to reach. Characterization of
the appropriate goal in this situation requires a degree of reflection that is gener-
ally beyond them. In contrast, 5-year-olds are able to specify the satisfaction con-
ditions sufficiently to contend with the altered instructions. This improvement
allows 5-year-olds to perform the correct intentional actions in more complex
situations in which there has been a change or an exception to the way things
typically occur. Importantly, this improvement has consequences for children's
260 Douglas Frye and Philip David Zelazo
intentional actions across a wide variety of domains, including instances in which
children must act to manipulate atypical causal sequences in the physical world
and those in which children must understand that at times it can be praiseworthy
to perform social actions that in other circumstances have harmful consequences.
REFERENCES
BALDWIN, J. M. (1968), Mental Development in the Child and the Race, 3rd edn. New York:
Augustus M. Kelley. Original work published in 1894.
CARLSON, S. M., and MOSES, L. J. (2001), 'Individual differences in inhibitory control and
children's theory of mind', Child Development, 72: 1032-53.
COHEN, L. B. (1998), 'An information processing approach to infant perception and cogni-
tion', in F. Simion and G. Butterworth (eds.), Development of Sensory, Motor, and
Cognitive Capabilities in Early Infancy: From Sensation to Cognition. East Sussex, UK:
Psychology Press.
DAS GUPTA, P., and BRYANT, P. E. (1989), 'Young children's causal inferences', Child
Development, 60: 1138-46.
DELoACHE, J. (1993), 'Distancing and dual representation', in R. R. Cocking and
K. A. Renning~r (eds.), The Development and Meaning of Psychological Distance.
Hillsdale, NJ: Erlbaum, 91-107.
DEMPSTER, F. N. (1992), 'The rise and fall of the inhibitory mechanism: toward a unified
theory of cognitive development and aging', Developmental Review, 12: 45-75.
DEWEY, J. (1985), 'Context and thought', in J. A. Boydston (ed.) and A. Sharpe (textual ed.),
John Dewey: The Later Works, 1925-1953, vol. vi, 1931-2. Carbondale, Ill.: Southern
Illinois University Press, 3-21. Original work published in 1931.
DIAMOND, A., and TAYLOR, C. (1996), 'Development of an aspect of executive control:
development of the abilities to remember what I said and to "Do as I say, not as I do"',
Developmental Psychobiology, 29: 315-34.
EDELMAN, G. (1989), The Remembered Present. New York: Basic Books.
ELMAN, J. (1990), 'Finding structure in time', Cognitive Science, 14: 179-211.
FRYE, D. (1991), 'The origins of intention in infancy', in D. Frye and C. Moore (eds.),
Children's Theories of Mind: The Development of Social Understanding of Others.
Hillsdale, NJ: Erlbaum, 15-38.
--(1992), 'Causes and precursors of children's theories of mind', in D. Hay and
A. Angold (eds.), Precursors, Causes, and Psychopathy. West Sussex, England: Wiley.
--(1999), 'Development of intention: the relation of executive function to theory of
mind', in P. D. Zelazo, J. W. Astington, and D. R. Olson (eds.), Developing Theories of
Intention: Social Understanding and Self-Control. Mahwah, NJ: Erlbaum, 119-32.
--(2000), 'Theory of mind, domain specificity, and reasoning', in P. Mitchell and
K. Riggs (eds.), Children's Reasoning and the Mind. Hove, UK: Psychology Press, 149-67.
--ZELAZO, P. D., BROOKS, P. J., and SAMUELS, M. C. (1996), 'Inference and action in early
causal reasoning', Developmental Psychology, 32: 120-31.
- - - - a n d BURACK, J. (1998), 'Cognitive complexity and control: I. Theory of mind in
typical and atypical development', Current Directions in Psychological Science, 7: 116-21.
Children's Action Control and Awareness 261
_ _- and PALFAI, T. (1995), 'Theory of mind and rule-based reasoning', Cognitive
Development, 10: 483-527.
JACQUES, S., ZELAZO, P. D., KIRKHAM, N. Z., and SEMCESEN, T. K. (1999), 'Rule selection
versuS rule execution in preschoolers: an error-detection approach', Developmental
Psychology, 35: 770-80.
LURIA, A. R. (1961), The Role of Speech in the Regulation of Normal and Abnormal
Behaviour, ed. J. Tizard. New York: Pergamon Press.
MELTZOFF, A. N. (1995), 'Understanding the intentions of others: re-enactment of intended
acts by 18-month-old children', Developmental Psychology, 31: 838-50.
MUNAKATA, Y., and YERYS, B. E. (2001), 'All together now: when dissociations between
knowledge and action disappear', Psychological Science, 12: 335-7.
PERNER, J., and LANG, B. (1999), 'Development of theory of mind and executive control',
Trends in Cognitive Sciences, 3: 337-44.
PIAGET, J. (1952), The Origins ofIntelligence in Children, trans. M. Cook. New York: Vintage.
Original work published in 1936.
--(1954), The Construction of Reality in the Child, trans. M. Cook. New York: Basic
Books. Original work published in 1937.
SEARLE, J. R. (1983), Intentionality: An Essay in Philosophy ofMind. Cambridge: Cambridge
University Press.
SIGEL, 1. (1993), 'The centrality of a distancing model for the development of representa-
tional competence', in R. R. Cocking and K. A. Renninger (eds.), The Development and
Meaning ofPsychological Distance. Hillsdale, NJ: Erlbaum, 91-107.
STROMMEN, E. A. (1973), 'Verbal self-regulation in a children's game: impulsive errors on
"Simon Says"', Child Development, 44: 849-53.
WELSH, M. C. (1991), 'Rule-guided behavior and self-monitoring on the Tower of Hanoi
disk-transfer task', Cognitive Development, 6: 59-76.
WELLMAN, H., CROSS, D., and WATSON, 1. (2001), 'Meta-analysis of theory of mind devel-
opment: the truth about false-belief', Child Development, 655-84.
WHITE, S. H. (1965), 'Evidence for a hierarchical arrangement of learning processes',
in 1. P. Lipsitt and C. C. Spiker (eds.), Advances in Child Development and Behavior.
New York: Academic Press, 187-220.
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.
ZELAZO, P. D. (1996), 'Towards a characterization of minimal consciousness', New Ideas in
Psychology, 14: 63-80.
- - (1999), 'Language, levels of consciousness, and the development of intentional action',
in P. D. Zelazo, J. W. Astington, and D. R. Olson (eds.), Developing Theories of Intention:
Social Understanding and Self-Control. Mahwah, NJ: Erlbaum, 95-1170.
--(2000), 'Self-reflection and the development of consciously controlled processing', in
P. Mitchell and K. Riggs (eds.), Children's Reasoning and the Mind. Hove, UK: Psychology
Press, 169-89.
--BURACK, J. A., BENEDETTO, E., and FRYE, D. (1996), 'Theory of mind and rule use in
people with Down Syndrome', Journal of Child Psychology and Psychiatry, 37: 479-84.
--CARTER, A., REZNICK, J. S., and FRYE, D. (1997), 'Early development of executive func-
tion: a problem-solving framework', Review of General Psychology, 1: 1-29.
262 Douglas Frye and Philip David Zelazo
ZELAZO, P. D., and FRYE, D. (1997), 'Cognitive complexity and control: a theory of the
development of deliberate reasoning and intentional action', in M. Stamenov (ed.),
Language Structure, Discourse and the Access to Consciousness. Amsterdam and
Philadelphia: John Benjamins, 113-53.
- - - - (1998), 'Cognitive complexity and control: II. The development of executive
function in childhood', Current Directions in Psychological Science, 7: 121-6.
- - - - a n d RApus, T. (1996), 'An age-related dissociation between knowing rules and '
using them; Cognitive Development, 11: 37-63.
- - HELWIG, C. c., and LAU, A. (1996), 'Intention, act, and outcome in behavioral pre-
diction and moral judgement; Child Development, 67: 2478-92.
- - JACQUES, S., BURACK, J. A., and FRYE, D. (2002), 'The relation between theory of mind
and rule use: evidence from persons with autism-spectrum disorders', Infant and Child
Development (Special Issue: Executive Function and its Development), 11: 171-95.
--and REZNICK, J. S. (1991), 'Age-related aynchrony of knowledge and action', Child
Development, 62: 719-35.
- - - - and PINON, D. E. (1995), 'Response control and the execution of verbal rules',
Developmental Psychology, 31: 508-17.
ZELAZO, P. R., and ZELAZO, P. D. (1998), 'The emergence of consciousness', in H. H. Jasper,
1. Descarries, V. F. Catellucci, and S. Rossignol (eds.), Consciousness: At the Frontiers of
Neuroscience: Advances in Neurology, !xxvii. New York: Lippincott-Raven Press, 149-65.
12
Jennifer Hornsby
Frye and Zelazo present findings that uncover a particular developmental change
in young children's capacities for intentional action. They bring these findings
within the purview of two theoretical models-the Cognitive Complexity and
Control account (CCC), and the Levels of Consciousness account (LOC). The two
models have been claimed to help to explain not only the change discussed in
detail in their chapter but a wide range of other findings besides. The special inter-
est of the authors' present contribution, signalled in its conjunctive title 'Control'
and 'Awareness', resides in its claim to harmonize the two models. LOC postulates
four major developmental changes, occurring at around 9 to 12, 21, 30, and 48
months; CCC focuses on changes between 3 and 5 years of age. Frye and Zelazo
suggest now that the LOC's account of a child's development through infancy
meshes with the CCC's account of development at the pre-school stage.
r want to approach the theoretical models by way of a more everyday perspective
on the findings. There is a very general question about how our ordinary, so-called
common-sense psychological perspective relates to that of experimentalists and theo-
reticians of psychology. An instance of this general question seems bound to arise in
evaluating the claims that Frye and Zelazo make in developmental psychology.
1. COMMON-SENSE PSYCHOLOGY
1 The term 'theory of mind' has also come to be used sometimes to stand for a field of research
within developmental psychology, rather than for anything that anyone, child or adult, putatively pos-
sesses or knows. But this evidently is not what Frye and Zelazo mean by it.
Whether, as Frye and Zelazo along with many others assume, psychological knowledge is always
knowledge 'of mental states' is something I put in question below.
2 I allude to the debate between simulationists and theory theorists. For a useful discussion, see
Davies (1994) and Heal (1994). Use of the term 'theory of mind' seems to me to reinforce a picture that
sometimes lies behind the theory theory. But I agree with Davies that the debate has many strands, and
is much less clear-cut than is often supposed.
3 See further Hornsby (1997b), where I suggest that, developmentally speaking, not only does
understanding of 'mind' go hand in hand with understanding of 'mind-world' relations, but under-
standings of both of these go hand in hand with understanding of 'world'.
Comment on Frye and Zelazo 265
So I prefer the label 'common-sense psychology' to 'theory of mind' for what we
(human adults) are governed by and all use all the time, and what children
gradually come to be governed by and to use. Prejudices would be introduced if
the term 'theory of mind' was substituted for 'common-sense psychology', because
the answers to all three of the questions I have just raised would then seem to have
been settled. In my own view, all three questions would be settled wrongly if
one followed through the implications of speaking of 'theory of mind'. For I want
to affirm all of the following: (1) Someone's ability to apply psychological con-
cepts to others is not a matter of their possessing a theory (not even a tacit theory).
(2) The appropriateness of applying psychological concepts to someone is insepa-
rable from that someone's being able to apply those concepts appropriately to
others. 4 (3) Our everyday common-sense psychological understanding involves
not 'mind' (mental states) specifically but interactions between 'world' (surroundings)
and 'mind'.
Now common-sense psychology, on this view of it, is a massive and unwieldy
subject matter. (1) It is not theoretically tractable; (2) it encompasses individual
psychological subjects only in so far as they are mutually intelligible; and (3) it
encompasses mind and world. Very evidently, empirical psychologists have to narrow
their sights and to study something much more delimited. Perhaps the term 'theory
of mind' stands for a field of studies whose concern is specifically understanding
of others' mental states. And perhaps the term 'executive function' has grown up in
developmental psychology to stand for a field of studies whose concern is specif-
ically action-related capacities-capacities to do things at will and for reasons. If
that is so, then common-sense psychology will subsume what is brought under
the heads of theory of mind and of executive function, and very much else
besides.
4 This is a philosophical claim about the conditions for common-sense psychological predications
to be in place. It is not that we can never understand someone by making use of a concept that she
doesn't apply to others. When it comes to children, we obviously do use concepts that they haven't yet
acquired, as I acknowledge below. The claim may be put in terms that Daniel Dennett has used: some-
one who takes the intentional stance towards people is someone towards whom that stance can appro-
priately be taken. In Hornsby (l997a), I attempt to make it seem reasonable that a third-personal point
of view (i.e. Dennett's 'intentional stance') and first-personal points of view go hand in hand. The
philosophical claim provokes high-level metaphysical questions, so that my treatment here is
inevitably cursory.
266 Jennifer Hornsby
mental states and that person's action-related capacities might be supposed to be
connected. A connection is made as soon as both are located within common-sense
psychology. It is made then by way of two other connections-(a) between common-
sense psychological concepts being applicable to X and X's having a capacity to
apply those concepts to others (cf. (2) above); and (b) between understanding
mental states and understanding of mind-world relations (cf. (3) above).
These connections, which link theory of mind with executive function, are not
developmental of course. From an everyday perspective, the process by which an
infant comes to be a common-sense psychological subject is surely a gradual one
to which the acquisition oflanguage (which of course is itself gradual) is crucial.
The question, or at least one large question, for developmental psychology is
whether one can distinguish steps in the process, and discern definite develop-
mental stages that children pass through. What seems certain, before any theoriz-
ing is undertaken, is that such steps will not map neatly onto distinctions that we
make commonsensically. It is not as if the child first became intelligible using a
batch of concepts and then in a separate step became able to apply those concepts
to others. Nor is it as if the various psychological concepts deployed commonsen-
sically might be mastered one at a time. The thesis of the holism of the mental
counts against this. And even before philosophical theses are on the scene, it seems
right to say that coming to have a grasp of any particular concept is itself a grad-
ual matter, often caught up with the acquisition of other, related concepts.
Children engage in the kind of human interaction for which common-sense
psychology provides long before they become mature common-sense psycholo-
gical subjects themselves. Consider the phenomena that psychologists study under
the various heads of imitation, affective exchange, pointing behaviours, pretend
play, make-believe. Many of us have first-hand experience of all of these. In early
joint visual attention, for instance, an adult turns to look at an object, and a baby
responds by looking in the same sort of direction; when the baby comes to focus
on the object, there is, as it were, a meeting of minds. It is natural to think of such
sharing of mental states as part of a process of initiation into a community of per-
sons. Joining in a range of activities partaking of intersubjectivity precedes the
achievement that has been labelled acquisition of a theory of mind. Thus
although the child's achievement is a gradual one, it cannot seem right to think
that children gradually come to 'have minds'.
Here we start to see some of the reasons for resisting the picture that can lie
behind tall< of acquisition of a 'theory of mind'. In the first place, we see why one
might resist assimilating the capacity to deploy psychological concepts to the pos-
session of a theory. The child's accomplishment in early development is a sort of
socialization into full personhood, and that fits ill with thinking of it as the learn-
ing of any theory. Second, suppose that it is correct to connect being a full-fledged
user of common-sense psychology with being a full-fledged subject of common-
sense psychology. Then if 'theory of mind' is supposed to stand for what is
Comment on Frye and Zelazo 267
acquired in becoming a user of common-sense psychology, it might seem as if
becoming a subject of common-sense psychology had to be equated with coming
to be a subject of mentality. But the case of young children shows that our idea of
a participant in phenomena of intersubjectivity needs to be distinguished from
our idea of being a full-fledged common-sense psychological subject. (In this con-
nection, it is worth noticing that, in most people's view, non-human animals, like
very young children, are much stronger candidates for being subjects of experi-
ence than they are for being common-sense psychological subjects. s But the pres-
ent point is only that psychological terms-in a sense of 'psychological' that
relates to 'having a mind' as opposed to 'being an automaton'-apply to infants,
but that infants are not yet subjects of common-sense psychology.)
5 Evidently much here is controversial; and no doubt my own view that only creatures with lan-
guage are common-sense psychological subjects is very controversial.
6 One might wish to add, for example, 'and if one is not forgetful of what one should be doing'. But,
as Frye and Zelazo say, it can't be made plausible that the 3-year-olds suffer memory lapses.
268 Jennifer Hornsby
by the generalization, then we have to say she does not know what she should be
doing. (The fact that she has the wrong view of what a puppet who is supposed to
be following the same instructions should do is more evidence that the child does
not know what she should be doing.) It seems in short that at the post-switch
stage the child, if she were treated as properly common-sense psychologically
intelligible, would both know and not know what she should be doing.
A similarly surd description could be offered of 3-year-old performance on the
ramp task, and again on the 'morality and action' tasks where the children's judge-
ments are not informed by what they seem really to know. Three-year-olds are
simply not the sort of rational beings that common-sense psychology takes for
granted when it is used in understanding adults. This is not to say that we cannot
apply any of common-sense psychology's concepts to 3-year-olds. Indeed, in
formulating any view of what is going on with 3-year-olds, we might find common-
sense psychological concepts very useful: they might help us to understand our
children better. In the light of the findings that Frye and Zelazo present, for example,
we might decide that we had sometimes wrongly seen young children as obstinate,
having mistaken what is actually a maturational deficit for a defect of the will.
There need not then be any difficulty with Frye and Zelazo's claim that a distinction
between what the child does intentionally and what she does unintentionally can
perfectly well be made during infancy: it is not as if common-sense psychology had
no purchase at all on children. It is only that such a distinction cannot have exactly
the same significance in young children's case as it does for adults. If one describes
the child as doing something intentionally, then one describes her by reference to
a network of interconnected concepts, not all the implications of which are yet in
place in application to her. We can use common-sense psychological descriptions,
but we may be led into absurdities if we follow through on what they would entail
as applied to normal adults.
The reason, I think, why common-sense psychology can lead to inconsistencies
if young children are taken to fall in its domain is that it takes so much for
granted: common-sense presupposes an enormous amount about what common-
sense psychological subjects are capable of. In the normal case, of normal adults,
its presuppositions are warranted. But not all of common-sense psychology's
presuppositions are in place where young children are concerned. Nor are they in
place in the case of patients with a range of neurological deficits, including the
example of the anarchic hand phenomenon to which Frye and Zelazo refer. And
its presuppositions let us down again when, for instance, people have extraordin-
arily outlandish beliefsJ Perhaps part of the explanation of why we can none the
7 See Stich (1983). Stich is among those who view common-sense psychology's breakdowns as part
of a case for its elimination: examples of outlandish beliefs feature in an argument which concludes 'so
much the worse for common-sense psychology'. It puts a new slant on the debate about eliminativism,
I think, to consider child development. This helps one to see some of the problems with the supposi-
tion, made by Stich and many others, that common-sense psychology gives a protoscientific account
and uses categories that might fix onto well-defined states in some science.
Comment on Frye and Zelazo 269
less rely on common-sense psychology is that we normally encounter mature
psychological beings whose brains and nervous systems subserve the range of
capacities which it presupposes; and when we encounter others-whether children,
or patients with neuropathologies, or people suffering from one or another kind
of insanity-we still find beings whose brains and nervous systems are such as to
subserve some large range of these capacities. At the points at which common-
sense psychology breaks down, we find ourselves with very little to say. We can
only try to get at what capacity is lacking. About our 3-year-olds, we might say
. that their knowledge does not impinge as it should upon their actions because
they are not yet able to keep track of what should be done. Such an account
evidently does not cast much light. It avoids paradox simply by acknowledging the
absence of a capacity present in mature psychological subjects, generalizations
about whom can then be supposed to be inapplicable. (Similarly unenlighteningly,
one could say of children who fail the 'false belief test' that they do not yet have
a workable notion of 'what he (another) will think'.)
Thus I suggest that there are features of young children that common-sense
psychology, so far from helping us to understand, finds paradoxical in its own
terms. It is here evidently that theory may come in. And it seems now that two sorts
of theory might be given. First, theory might attempt exactly to pinpoint specific
capacities that young children lack. Then childhood deficits could be described in
a much less rough and ready way than they are when it is allowed simply that some
relevant capacity or other is not developed. (Theory of this sort begins to encroach
upon common-sense if we decide, as I suggested above that we might, that 3-year-
olds are not as straightforwardly obstinate as we might have supposed.) Second,
theory might postulate capacities attributable specifically to children, treating
them as having a psychology proper to them. Theory of this second sort is more
ambitious: it attempts to do more than to isolate respects in which children are
deficient as compared with adults. It might introduce a stock of concepts to tell
a developmental story, which provides, at any stage, a psychological description of
children that achieves a consistency missing from a common-sense psychological
description of them. For example, it might be postulated that children at a certain
developmental state have 'proto-intentions'-states of mind which do not
conform to those of common-sense psychology's generalizations about intentions
that land us in inconsistency, but which none the less playa role related to that of
intentions proper.
The division between these two sorts of theory is evidently not a very clear-cut
one. Indeed, since the two kinds of theoretical intervention I am imagining would
ideally mesh with one another, the separation cannot be sharp. Theories of the
first sort work back from common-sense psychology, as it were. Theories of the
270 Jennifer Hornsby
second sort work up towards common-sense psychology. If theories of the two
sorts were available, and could be made to dovetail, then we should have a story
to tell about how a child comes to be a common-sense psychological subject.
According to this division, it seems that CCC attempts a theory of the first sort,
and LOC a theory of the second sort. The interest of looking at CCC and LOC
together, then, is not merely that CCC deals with a stage of development that is
recognized in the more developmentally comprehensive LOC. It seems as if the
two accounts might together yield an informative account of how a child comes
to be an intentional agent. But despite the potential promise that Frye's and
Zelazo's accounts hold out, I fear that I have difficulties about reconstructing from
them the kind of overall account I think we want. I shall now take the CCC and
LOC accounts in turn, and explain some of my difficulties.
Frye and Zelazo use the idea of underspecification also in explaining failures in
'the false belief' test. They say that the 3-year-old 'attempts to determine where the
character should look to find the object' but fails to 'attempt to determine where
the character should look from the character's perspective'.
CCC initially seems hard to fault. Certainly, in some sense or other the 3-year-old
acts on an inadequately specified goal. But we need to know more exactly what it
is for a goal to be 'underspecified'. One suggestion might be this: a complete spec-
ification of a goal is the content of an intention of someone who performs the task
adequately; and there is underspecification of the goal where an intention with a
similar content but with something omitted is present and explains the deficient
performance on the task. According to this suggestion, what is lacking in com-
pleteness, where there is underspecification, is the content of an intention. This
suggestion fits with the most natural way of understanding what Frye and
Zelazo say about the false belief test: they think that the concept of 'the character's
perspective' goes missing from the thoughts of 3-year-olds who fail the test.
But there is a difficulty with the view we reach if we construe underspecification
in line with this suggestion. We have to say that the difference between the 3-year-old
and 5-year-old, is a difference of thought content. The 5-year-old who passes the
Comment on Frye and Zelazo 271
false belief test and is capable of realizing that another person will get something
wrong has a thought whose content, unlike that of the 3-year-old, includes a concept
of 'the perspective [of the other]'. And this is questionable. For we can distinguish
between having the ability to appreciate another's perspective, and exercising the
concept of another's perspective in one's thoughts. And while it may be very
plausible that in a range of circumstances, 5-year-olds can and 3-year-olds cannot
appreciate another's perspective, it is surely not so plausible that 5-year-olds actually
exercise the concept of another's perspective in their thoughts. (To me it does not
seem plausible that adults exercise the concept of 'another's perspective' in
thought very often, although of course they often manifest an ability to know
what someone else will think about something.)
There is not necessarily a criticism of Frye and Zelazo here, because underspeci-
fication might be understood differently. But questions are bound to arise about
what exactly is added to an account of development when claims are made about
specifications of goals. We need to be told exactly which states of mind suffer from
underspecification in 3-year-olds, and exactly what is absent from the underspeci-
fied states. (If we are indeed meant to understand underspecification as a matter of
something's being missing from a state's content, then we must supply an analogue of
'another's perspective' to do the work in accounting for the deficit in the DCCS case
that 'another's perspective' does in accounting for the deficit when there are failures
on the false belief test.) It may be that the states that CCC postulates (whose objects
Frye and Zelazo call goals) are not states of intention as these are ascribed in com-
mon-sense psychology, so that we should not think of them as having the kinds of
content that adults' thoughts and intentions have. But if this is so, then CCC will
emerge as a more ambitious theory (with the pretensions of theory of the second sort
I distinguished above). Then we should need to know more about the postulated
states, and about how those of 3-year-olds differ from those of 5-year-olds.
I have raised these questions about what an underspecification of goals
amounts to in order to draw attention to how little in the way of actual explana-
tion CCC appears to provide. It is not clear that, as it stands, it offers us much
more than the unenlightening account we looked at earlier, which recognizes that
young children cannot keep track of what they should do. The interest of CCC
would seem to lie principally in its bringing together a range of findings, and thus
in its suggestion of a developmental step which makes a crucial difference across
a variety of task domains. Its chief implication seems to be that what is lacking in
the 3-year-old in the DCCS task is not some specialized capacity.
LOC proposes that the developmental change addressed by CCC is the final
change in a series of changes in 'recursive awareness'. LOC, moreover, promises to
be more specific about underspecification: it appears to introduce a tangible claim
272 Jennifer Hornsby
about what a greater degree of specification involves-namely, a specification at a
higher level of consciousness. Perhaps it was wrong to turn to common-sense psy-
chological concepts in seeking to discover what underspecification amounts to:
perhaps we needed further theory, of a more ambitious kind, to put the relevant
notion of underspecification into service. Prima facie, then, LaC might not only
add to CCC by speaking also to earlier developmental history, it might also add
shape and definition to CCC itself.
But LaC is hard to judge. How are we meant to understand 'level of conscious-
ness'? And how is the idea of such levels related to an information-processing story?
With consciousness in the picture, one expects phenomenology to be some sort
of touchstone. But thinking of what one can be aware of oneself doesn't appear to
reveal what is meant by a different 'level' of consciousness in the LaC model.
Certainly, there can be more to being a conscious subject than enjoying the 'minimal
consciousness' that corresponds to the initial level of consciousness. But why
should a subject who enjoys more than minimal consciousness be thought to be
conscious at a new and different level? Perhaps we secure an idea of a second level
of consciousness, above the minimal, by thinking of consciously reflecting on our
own conscious states. It doesn't seem credible, however, that such thinking about
our own states (consciously reflecting on consciously reflecting on ...) can provide
us with as many levels as five. Yet, according to LaC, a fifth level of consciousness
is what 5-year-olds reach, having passed from minimal consciousness through
four developmental steps.
Despite the name that he gives to his account, Zelazo is perhaps not really
committed to as many levels of consciousness as there are stages in the account. It
might be suggested that what increases is actually the number of the model's layers
that are relevant to what can be present to consciousness. On this construal,
consciousness itself does not assume new levels, but only the recursions in a
process by which information feeds into the contents of consciousness. But if one
assumes that the different outputs of greater levels of processing are what really
constitute differences in 'levels of consciousness', then one wants to know how the
model's different layers connect with the child's awareness, or connect with the
contents of states that the child's performance warrants the ascription of. The idea
would have to be that each developmental stage brings a different layer of informa-
tion processing, and at any stage the whole processing system existing at that
stage somehow determines what is available to the conscious subject at that stage,
or determines states which explain the child's performance. If this is indeed the
idea, then something will need to be said about the nature of the determination
of the contents of consciousness, or of mental states, by information processors.
Does Zelazo want us to think of a special centre in the brain where the deliver-
ances of the various recursions (their number depending upon the child's stage of
development) somehow arrive? If so, his model seems to be a version of the
Cartesian Theatre, which Dennett criticized in Consciousness Explained (1991).
Comment on Frye and Zelazo 273
But unless we are told more about the relationship between the outputs of the
model and capacities of the child,s the model's claims about the inner workings
of the child's brain will not seem even to start to explain the different degrees of
specification of goals which CCC tells us that the child goes in for.
7. OVERVIEW
In the ordinary way we make use of common-sense psychological concepts in
understanding mind and agency. Developmental psychology, like any other
branch of human psychology, seems bound to take off from, and ultimately be
answerable to, a common-sense psychological conception of mind. It begins with
one because it treats children who are experimental subjects as we all do; and we
treat children as, as it were, en route to full personhood. It is answerable to common-
sense psychology, because common-sense psychology provides the categories that
we have eventually to rely upon to assess the value of any theories that are put
forward. I have made a suggestion about how developmental theory (coming in
two roughly distinguishable sorts) might playa role such that it could both
take off from and be answerable to common-sense psychological conceptions of
children. But I have also suggested that CCC fails to take us much beyond what
common-sense already suggests we might say about 3-year-olds' performance on
the DCCS, and that LOC departs from common-sense in ways that make it hard
to assess.
REFERENCES
DAVIES, M. (1994), 'The mental simulation debate', in C. Peacocke (ed.), Objectivity
Simulation and the Unity of Consciousness. New York: Oxford University Press, 99-127.
Proceedings of the British Academy 83.
DENNETT, D. (1991), Consciousness Explained. London: Little, Brown and Co.
FRYE, D. (1999), 'Development of intention: the relation of executive function to theory of
mind', in P. D. Zelazo, J. W. Astington, and D. R. Olson (eds.), Developing Theories of
Intention: Social Understanding and Self-Control. Mahwah, NJ: Erlbaum, 119-32.
HEAL, J. (1994), 'Simulation vs Theory Theory: What is at Issue?', in C. Peacocke (ed.),
Objectivity Simulation and the Unity of Consciousness. New York: Oxford University
Press, 129-44. Proceedings of the British Academy 83.
8 No doubt one needs to look at the data on younger children that are adduced in support of LOC
in order to learn more about the relationship between the outputs of the model and capacities of the
child. But in the present chapter, I confine myself to the question of what light has been cast on the
particular findings described in Frye and Zelazo's contribution.
274 Jennifer Hornsby
Michael Lewis
The concept of agency implies an active organism, one who desires, one who
makes plans, carries out actions, and compares their actions to their desires. Such
a concept of agency requires consciousness. There is little question that in adults,
such conscious processes exist. The question is, from a developmental perspective
when do they arise? In our work on the development of self-awareness, we have
developed measures which are related to this concept. They include self-recognition,
personal pronoun usage (me and mine), and pretend play.
In this chapter, we will first develop the idea that the mental state of the idea of
me needs to be distinguished from the system aspects of the self, that a theory of
mind-that is, the child's knowledge of these internal states, such as thinking and
intending-requires self-consciousness. Second, we will show the developmental
course of self-recognition and personal pronoun usage, and how they are related
to the onset of pretend play; and third, we will show how the onset of these capa-
cities affects the onset of self-conscious emotions, such as embarrassment, shame,
guilt, and pride.
f ! ~C_~_~ 1
B
_ _2_m_o_nt_h_s
L1 EX L2
(0) Phase of learning
2 months
I
f~j - -=~= = = =~_: . . . .:. . 2_C~: : .P:.:N_
2.0
l...-_ _
B L1 EX L2
(b) Phase of learning
.... 2 months
<Ii V>
O)c
Co 1.5 I
2.0 EXP
"'.-
o~
C .... 1.0
",c.. ~
2
<Ii ><
<Ii
0.5~
o L- --_...:::::::::::=-...- ===-----="=====-
----'=_
B L1 EX L2
(c) Phase of learning
phase. Anger increased markedly once the association between action and outcome
was broken and declined as rapidly once the second learning phase began.
The data for those 2-month-olds was identical to that shown by 4-, 6-, and
8-month-olds (Lewis et ai., 1990); thus 2-month-olds behaved equally like the
8-month-olds. These findings force us to conclude that the 8-week-old infant has
intentions; that is, they intend to cause the pictures to appear and are happy when
they do so. When in extinction, the arm pull does not result in the expected out-
come, they pull harder as if they intend to reinstate the event. Their emotional
behaviour parallels and gives support to this conclusion.
Is it reasonable then to say that the 8-week-old intends in the same way that the
40-year-old intends? I believe that it is a mistake to do so; first, because we then
must conclude that there is no development of intention, and, second, we fail to
consider the role of consciousness in intentions. Elsewhere, I have tried to consider
this problem by proposing three types of intentional processes (Lewis, 1990b, c).
Here I wish to draw a distinction between intentions of systems that do not have
consciousness and those that do. The two features of the self that we need to
The Development of Self-Consciousness 279
consider are what I call the system properties of the self or the machinery of the self
and the idea ofme or the mental state of knowing about one's self from an epistemo-
logical point of view. It is the distinction between I know and I know I know.
The failure to make the distinction between these two features of self can be
blamed for much confusion when studying the issue of development. The dis-
tinction between self and me or between knowing and knowing I know involves
two aspects of me. If we do not confuse knowing with knowing I know, then the
argument around the issue of the developmental sequence in self becomes clearer.
As I have already suggested, many features of the self exist early, and exist as part
of the system from birth or soon after; the idea of me-the knower who knows-
is not developed until somewhere in the middle of the second year oflife (Darwin,
1965; Lewis, 1992).
We have already given some attention to two early features of the self. These are
the self-other differentiation and self-regulation. Both features are likely to be
part of the machinery ofmyselfand not related to the idea of me. Certainly, by two
months, and most likely from birth, an infant can differentiate itself from other.
Self- and other-differentiation also has associated with it a type of self-awareness.
It is the self-awareness of elements of a system in communication with one
another. This type of recognition and the self-other differentiation is part of the
hardware of any complex system. T-cells recognize and differentiate themselves
from foreign protein. A rat does not run into the wall but runs around it. A new-
born infant recognizes and responds appropriately to intersensory information.
Therefore, we should not expect that these aspects of self are the differentiating
features when we compare widely different organisms. All organisms, as systems,
should have these capacities.
What may distinguish organisms in regard to their system organization is the
complexity of the machinery of these systems. What may differentiate humans from
most other living organisms are not the functions of the system, but the ability to
have mental states and, more specifically, the mental states related to the idea of me.
In any consideration of the concept of self, especially in regard to adult humans,
it is important to keep in mind that both features exist. There is, unbeknownst to
us most of the time, an elaborate complex of machinery that controls much of our
behaviour, learns from experience, has states and affects, and affects our bodies,
probably including what and how we think. The processes are, for the most part,
unavailable to us. What is available is the idea of me, a mental state.
What is particularly impressive is the research on brain function and the findings
which point to the possibility that different areas of the brain may be associated
with different functioning. Thus, both the machinery of the self and the mental
state involving the idea of me appear to be the consequences of different biological
processes and locations. For example, the work by LeDoux (1989) points to specific
brain regions that may be responsible for different kinds of self-processes. Working
with rats, LeDoux found that even after the removal of the auditory cortex, the
280 Michael Lewis
animals were able to learn to associate an auditory signal and shock. After a few
trials, the rats showed a negative emotional response to the sound, even though its
auditory cortex had been removed. These findings indicate that the production of
a fear state is likely to be mediated by subcortical, probably the thalamic-amygdala
sensory, pathways. Similar findings have been reported in humans, suggesting that
states can exist without one part of the self experiencing them. Weiskrantz (1986),
among others, has reported on a phenomenon called 'blindsighted'. Patients have
been found who lack the visual cortex, at least in one hemisphere. When they are
asked if they can see an object placed in their blind spot, they report that they can-
not see it; that is, that they do not have the experience of the visual event. The self
reflecting on itself, my recognition of what I know, the me-the mental state-in
fact does not see. When, however, they are asked to reach for it, they show that
they have the ability to reach, at least some of the time, for the object. Thus, they
can 'see' the event, but cannot experience their sight. These findings, as well as
Gazzaniga's work (1988) on split brain, suggest that separate brain regions are
responsible for the production and maintenance of both the machinery of self-
processes and the mental state of the idea of me. Tulving (1985) has suggested that
memory itself can be conceptualized as memories involving the self (episodic)
from other memories not involving the self. Lewis (1990c, 1992; Lewis and
Michalson, 1983) has made this argument in the emotional domain arguing that
emotional states such as fear can occur without the experience of these states or
for that matter that emotional experiences can exist without the states themselves.
I have tried to summarize the distinction between these two features of the
self (Lewis, 1999) arguing for six basic principles:
1. All living systems self-regulate. By this we mean that within any living system
there needs to be communication between parts of that system. This we can call
awareness, but not the mental state of awareness. This can include a unit as small
as a cell, a plant, or animal, or even more complex organisms. As I write, my systems
are regulating my temperature or regulating my blood sugar level. Regulation is a
property of living matter. Regulation makes no assumptions about objective self-
awareness or intentions, although there is intentionality in the process.
2. In order to act, it is necessary for organisms to be able to distinguish between
self and other. Whether this ability is learned as Watson suggests (1994) or, as
others have suggested, part of the process of action-including perceiving, feeling,
and thinking-is unknown (Butterworth, 1990). What appears to be is that no
organism can act without being able to distinguish between self and other. The
ability to regulate or to distinguish self from other is part of the machinery of all
living systems (von Bartalanffy, 1967).
3. Even higher-order functions such as perception and complex actions
(i.e. driving a car) can be carried out by adult humans without objective self-
awareness; that is, without their being able to look in and observe the processes
which allow these behaviours to be carried out. I cannot watch myself think, I can
only look at the product of my thinking.
The Development of Self-Consciousness 281
4. A unique aspect of some self-systems is objective self-awareness. By objective
self-awareness, I mean the capacity of a self to know it knows or to remember it
remembers. It is this mental state that we refer to when we say self-awareness. The
capacity of objective self-awareness may be uniquely human (although the great
apes, porpoises, and whales appear capable of this as well; Lewis, 1994, 1995a).
5. Once the emergence of consciousness occurs, processes of agency, originally
part of the machinery of the self come under the control of the conscious desires
of the organism. Thus, although agencies and intentions exist from birth, the
processes that support them undergo change.
6. Specific developmental processes of the self follow the general principles of
development. Earlier capacities, such as agency controlled by the machinery of the
self may give rise to later capacities, like mental states (e.g. the idea of 'me'), but
these capacities are not transformed. Thus, agency controlled by the machinery of
self exists once agency controlled by consciousness emerges. Thus, unlike a more
classical genetic epistemological approach, I see the retention of earlier functions
as not only possible, but a necessary aspect of development. In some sense, then,
old structures in interaction with the environment and/or as a function of mat-
uration give rise to new structures. These new structures do not replace the old
ones, but coexist with them. Under certain conditions, individuals will utilize the
most mature aspect they have achieved. However, this does not mean other
aspects are not utilized. In some sense, then, mature adults possess within their
repertoire all aspects of agency, whereas younger children or infants possess only
those earlier aspects (Lewis, 1997).
While others have focused on agency and intentions in early infancy, I have
preferred to study agency and intentions as they emerge as a function of con-
sciousness. Hence, here we turn our attention to the emergence of consciousness
and of adult-like intentions.
While certain aspects of self-development are likely to occur across the first year
and a half of life (Butterworth, 1992; Hobson, 1990; Lewis, 1995a; Lewis and
Brooks-Gunn, 1979a, b; Meltzoff, 1990; Meltzoff and Gopnik, 1993; Rochat,
1995), there is clear evidence for the onset of the idea of me or self-representation
by approximately the middle of the second year of life. Visual self-recognition has
been found to emerge in children at this point in development. Self-recognition
has been assessed by surreptiously applying a spot of rouge to children's faces and
then observing whether they touch the mark on their faces when in front of a mirror
(Lewis and Brooks-Gunn, 1979a; Lewis and Ramsay, 1997). In normally developing
children, this mark-directed behaviour has been found to emerge reliably between
15 and 24 months of age and not before. Moreover, self-recognition appears to be
282 Michael Lewis
a prerequisite for children's expression of various self-conscious emotions. For
example, Lewis et al. (1989) found a relation between self-recognition and embar-
rassment. Similarly, Bischof-Kohler (1988, 1994) found that empathic behaviour
was related to self-recognition, and Zahn-Waxler et al. (1992) found that altruism
was related to self-recognition. Based on these findings, it has been argued that
self-recognition onset marks more broadly the emergence of self-awareness
(Darwin, 1965; Duval and Wicklund, 1972), reflecting children's understanding
that they are an object in their own mental representations of the world (Lewis,
1992, 1995a, 1997). It is to be noted as well that children also begin to use personal
pronouns including 'me' and 'mine' by the latter part of the second year of life
which also provides a demonstration of the emergence of the idea of me (Harter,
1983; Hobson, 1990). In fact, one can observe two children playing together and
see that one child will take a toy from another, bring it toward herself, away from
the other child, and say 'mine'. This action of bringing the toy towards the space
occupied by the self and away from the space of the other child, and at the same
time saying 'mine,' strongly suggests that the language usage reflects a linguistic
marker of the idea of me. The use of one's own name might be another example
of the linguistic capacity if it were not the case that parents often use the child's
name and that this usage may give rise to a learning capacity not necessarily
reflecting a representation of the self (see Lewis and Brooks-Gunn, 1979a, for a
fuller description of this problem).
Thus there is evidence of the emergence of this meta-representation or idea of
me by the middle of the second year of life. This emergence is further supported
by studies indicating that children who are mentally retarded and do not have a
mental age of 18 months are not able to recognize themselves in the mirror task.
The emergence of self-recognition as a function of maturation can be found in
the data on brain myelinization. Research on myelinization of the frontal lobes
suggests that while myelinization is not complete by the middle of the second
year, it is on its way (Grodd, 1993; Staudt et al., 1993, 1994). As we shall see in the
next section of the chapter, the emergence of this meta-representation of me leads
to emotions which have as their focus intentionality and are related to the agency
of self (Lewis and Michalson, 1983; Lewis, 1992; Lewis and Ramsay, forthcoming).
There is good reason to suppose that this meta-representation is necessary for
pretence. A self-representation is necessary for there to be the dissociable relation
between the literal and pretend situation. Piaget (1962) understood that pretence
requires an awareness of self in order for children to distinguish between what is
reality and what is fantasy. The child feeding a doll imaginary food, or the child
drinking an imaginary liquid from a cup must know that what he or she is doing
is not real. If children did not know that their actions on objects were not real,
their play would not be pretence but a hallucination. Pretence involves, in effect,
a negation by the self that 'this is what I pretend it to be' rather than what it actually
is. Similarly, Leslie (1987) included a self-representation in his theoretical model
The Development of Self-Consciousness 283
on pretend play and theory of mind (ef. the 'I pretend--' function of the
'manipulator' in the model). Thus, the origins of a theory of mind involve a self-
pretending, that is, the appearance of the self that knows that it knows and knows
that its play is not real (Lewis and Ramsay, 1999).
Thus, from a variety of theoretical perspectives (e.g. Huttenlocher and Higgins,
1978; Leslie, 1987; McCune-Nicolich, 1981; Piaget, 1962), pretence is an early
manifestation of the ability to understand mental states including one's own and
others'. The dissociable relation (the double knowledge or dual representation of
the literal and pretend situation) allows the child to distinguish between appearance
and reality and provides the cognitive basis for a theory of mind. Research by
Piaget (1962) and subsequent investigators (e.g. Fein, 1975; Lowe, 1975; McCune,
1995; Nicolich, 1977) indicates that pretence emerges in children by the middle to
latter part of the second year oflife, thus suggesting that it too emerges at the same
time as self-recognition and personal pronoun use.
Lewis and Ramsay (1999) have recently observed children's capacity at sponta-
neous pretend play, mirror self-recognition, and personal pronoun use in a longi-
tudinal study of children at 15, 18, 21, and 24 months of age. Spontaneous pretend
play was observed in a play-room situation as was their ability to recognize them-
selves using the mirror recognition task. Maternal reports of personal pronoun
use were also obtained in order to see whether there was a sequence to these
emerging capacities. We hypothesized that visual self-recognition should be
preceded by personal pronoun usage or at least emerge at the same time and that
pretend play would be exhibited only after visual self-recognition occurred. Use of
latent transition analysis (Collins and Wugalter, 1992) confirmed that the sequential
emergence of the three behaviours followed in the order of self-recognition,
personal pronoun usage, and pretend play. Table 13.1 shows the perspective of
(Lewis and Michalson, 1983; Lewis, 1992) that there is relatively little dependency
between emotional states and emotional experiences and that a consideration of
the differences between the two is one way to bridge the gap between those who
would and those who would not hold that emotions need to involve cognition. It
is clear that emotional experiences are by their nature cognitive. Emotional states,
on the other hand, mayor may not have cognitive elicitors depending on whether
we want to consider sensory processing as cognitive and depending on what
emotions we consider.
experiences may affect the likelihood of children making the one or the other
attribution in such cases. In the achievement literature, there are data to suggest
that, in situations of failure, females are more likely to make an internal, global
attribution and males an external evaluation: men tend not to blame themselves
and, when they do, are more likely to make a specific attribution. Alessandri and
Lewis (1993) attempted to explore this issue by looking at parents' behaviour in
the context of an achievement test where they were asked to get their children to
solve both easy and difficult tasks. We observed that parents made very few global
attributions, but those that they did make were much more likely to be negative
for daughters than for sons, suggesting that parents may have an effect on both the
internal/external and the global/specific-attribution dimensions. Indeed, in the
same study, parents were more likely to make specific attributions to sons than
to daughters, and their specific attributions to girls were more negative than those
to boys.
Thus, it seems that direct teaching experiences may playa significant role in
affecting the child's self-evaluative behaviour. However, there are probably other
dimensions of socialization that playa significant role in regard to the acquisition
of self-evaluative strategies. In particular, the use of love withdrawal as well as the
use of contempt and disgust appear to be related to individual differences in
proneness to shame. It seems fairly clear, from a variety of observations, that con-
tempt and disgust, directed to another-in this case, from parent to child-are
likely to lead the latter to shame attributions around failure. Looking at videotapes
of middle-class parents teaching their children, we were amazed at the number of
instances in which disgust appeared as a component of the direct socialization of
an action. Thus, for example, a mother might say to her 2- or 3-year-old, 'Look,
don't do that. It's dirty', while concurrently presenting a disgust face to the child.
The sparse literature on shame (for a review, see Lewis, 1992) suggests that shame,
humiliation, and disgust directed towards another are likely to produce those
same emotions in the recipient. Thus, socialization of individual differences in
self-evaluative style not only may be a function of the explicit message that parents
give to children but may be located in the non-verbal, implicit messages as well.
In this regard, one is reminded of Miller's (1981) work on parental use of shame
and humiliation in the socialization of their children.
Finally, love withdrawal seems to be a particularly potent elicitor of shame.
As I have argued elsewhere, there are likely to be few stimulus events that have
direct consequences in the production of shame; rather, these are mediated by
self-evaluative cognitive processes. However, if I were to nominate any single event
as prototypical in this regard, that is, one likely to elicit shame with a minimal
amount of self-evaluation, it is withdrawal oflove by an important figure. It seems
to me that love withdrawal as a socialization technique is highly effective because
of its shame-inducing capacity. Thus, the withdrawal of love by the significant
other may be an important mechanism through which self-evaluation takes place,
The Development of Self-Consciousness 291
and, at the same time, it is likely to account for individual differences. Its effect on
the incorporation of the parent's value may be due to the potency of this elicitor.
So, for example, in order to avoid the withdrawal of love of the significant other,
it becomes increasingly important for children to remember to evaluate their
actions vis-a-vis those behaviours that are likely to cause love withdrawal. At the
same time, since love withdrawal is interpreted as a global, internal failure, it is,
consequently, likely to lead to shame.
4.4 Summary
The development of intentions is complicated by the fact that even very young
infants behave as if they intend to act in a particular fashion. Recognizing this
difficulty, Lewis (l990a) proposed that intentions vary in terms of their use of a
self-referent. Infants, prior to the development of a self-representation, show
intentional behaviour, but these behaviours are due to the inborn machinery of
the organism; they can be considered to be primitive intentions. Once a self
schema emerges, the processes underlying intentions change. Intentions now are
related to how the self views itself, the world, and its actions in the world. These
can be considered to be self schema intentions. Once these emerge, both types of
intentions exist, one does not replace the other. Nowhere is the change in the type
of intentions more obvious than in the study of emotional development. The
primitive intentions are associated with the primary emotions. Not until the self
schema intentions emerge do we also see the emergence of emotions which can
only exist as a function of the child's intentions about itself and its goals.
In terms of development, children acquire a meta-representation or 'the idea of
me' in the middle of the second year of life. These self-cognitions play an important
role in both generating a theory of mind as well as being directly related to the
development of a complex emotional life. While the critical events that lead to this
meta-representation have not been fully worked out, the advent of this capacity is a
milestone in the development of human cognition, action, and emotional life.
REFERENCES
ALESSANDRI, S. M., and LEWIS, M. (1993), 'Parental evaluation and its relation to shame
and pride in young children', Sex Roles, 29: 335-43.
ASENDORPF, J. B., and BAUDONNIERE, P. M. (1993), 'Self-awareness and other-awareness:
mirror self-recognition and synchronic imitation among unfamiliar peers',
Developmental Psychology, 29: 88-95.
--WARKENTIN, v., and BAUDONNIERE, P. M. (1996), 'Self-awareness and other-awareness
II: mirror self-recognition, social contingency awareness, and synchronic imitation',
Developmental Psychology, 32: 313-21.
292 Michael Lewis
BERTENTHAL, B. 1., and FISCHER, K. W. (1978), 'Development of self-recognition in the
infant', Developmental Psychology, 14: 44-50.
BISCHOF-KoHLER, D. (1988), 'On the connection between empathy and the ability to
recognize oneself in the mirror', Schweizerische Zeitschrift fUr Psychologie, 47: 147-59.
--(1994), 'Self-objectification and other-oriented emotions: self-recognition, empathy,
and prosocial behavior in the second year', Zeitschrift fUr Psychologie, 202: 349-77.
BUTTERWORTH, G. (1990), 'Origins of self-perception in infancy', in D. Cicchetti and
M. Beeghly (eds.), The Self in Transition: Infancy to Childhood. Chicago: University of
Chicago Press, 119-37.
- - (1992), 'Origins of self-perception in infancy', Psychological Inquiry, 3: 103-11.
COLLINS, L. M., and WUGALTER, S. E. (1992), 'Latent class models for stage-sequential
dynamic latent variables', Multivariate Behavioral Research, 27: 131-57.
DARWIN, C. (1872/1965), The Expression of the Emotions in Man and Animals. Chicago:
University of Chicago Press, 1965. Original work published in 1872.
DUVAL, S., and WICKLUND, R. A. (1972), A Theory of Objective Self-Awareness. New York:
Academic Press.
DWECK, C. S., and LEGGETT, E. L. (1988), 'A social-cognitive approach to motivation and
personality', Psychological Review, 95: 256-73.
FEIN, G. G. (1975), 'A transformational analysis of pretending', Developmental Psychology,
11: 291-6.
FREUD, S. (1959), 'The ego and the id', in J. Strachey (ed. and trans.), The Complete
Psychological Works ofSigmund Freud, xix. London: Hogarth Press, 3-66. Original work
published in 1923.
GALLUP, Jr., G. G. (1991), 'Toward a comparative psychology of self-awareness: species
limitations and cognitive consequences', in G. R. Goethals and H. Strauss (eds.), The Self
An Interdisciplinary Approach. New York: Springer-Verlag, 121-35.
GAZZANIGA, M. S. (1988), 'Brain modularity: towards a philosophy of consciousness experi-
ence', in A. J. Marcel and E. Beseach (eds.), Consciousness in Contemporary Science.
Oxford: Clarendon Press, 218-56.
GOPNIK, A., and MELTZOFF, A. N. (1994), 'Minds, bodies and persons: young children's
understanding of the self and others as reflected in imitation and theory of mind research',
in S. T. Parker, R. W. Mitchell, and M. L. Bocchia (eds.), Self-Awareness in Animals and
Humans: Developmental Perspectives. Cambridge: Cambridge University Press, 166-:-86.
GRODD, W. (1993), 'Normal and abnormal patterns of myelin development of the fetal and
infantile human brain using magnetic resonance imaging', Current Opinion in Neurology
and Neurosurgery, 6: 393-7.
GROSS, C. J., ROCHA-MIRANDA, C. E., and BENDER, D. B. (1972), 'Visual properties of neurons
in the inferotemporal cortex of the macaque', Journal ofNeuropsychology, 35: 96-:-111.
HARTER, S. (1983), 'Developmental perspectives on the self-system', in E. M. Hetherington
(ed.), Handbook of Child Psychology, iv: Socialization, Personality, and Social
Development, 4th edn. New York: Wiley, 275-385.
HOBSON, R. P. (1990), 'On the origins of self and the case of autism', Development and
Psychopathology, 2: 163-81.
HOCHSCHILD, A. R. (1979), 'Emotion work, feeling rules, and social structure', American
Journal ofSociology, 85: 551-75.
The Development of Self-Consciousness 293
HUTTENLOCHER, J., and HIGGINS, E. T. (1978), 'Issues in the study of symbolic develop-
ment', in W. Collins (ed.), Minnesota Symposia on Child Psychology, xi. Hillsdale, NJ:
Erlbaum, 98-140.
IZARD, C. E. (1979), The Maximally Discriminative Facial Movement Coding System (MAX).
Newark, Del.: Instructional Resources Center, University of Delaware.
JAMES, W. (1890), The Principles of Psychology. New York: Holt.
LEDoux, J. (1989), 'Cognitive and emotional interactions in the brain', Cognition and
Emotion, 3: 265-89.
LESLIE, A. M. (1987), 'Pretense and representation: the origins of "Theory of Mind"',
Psychological Review, 94: 412-26.
LEWIS, M. (1990a), 'Intention, consciousness, desires and development', Psychological
Inquiry, 1 (3): 278-83.
--(1990b), 'Social knowledge and social development', Merrill-Palmer Quarterly, 36:
93-116.
--(1990c), 'Thinking and feeling-the elephant's tail', in C. A. Maher, M. Schwebel,
and N. S. Fagley (eds.), Thinking and Problem Solving in Developmental Process:
International Perspectives (the WORK). Hillsdale, NJ: Erlbaum, 89-110.
--(1992), Shame: The Exposed Self. New York: Free Press.
--(1993a), 'Self-conscious emotions: embarrassment, pride, shame, and guilt', in
M. Lewis and J. Haviland (eds.), Handbook of Emotions. New York: Guilford Press,
563-73.
--(1993b), 'The emergence of human emotions', in M. Lewis and J. Haviland (eds.),
Handbook ofEmotions. New York: Guilford Press, 223-35.
--(1994), 'Myself and me', in S. T. Parker, R. W. Mitchell, and M. 1. Boccia (eds.), Self-
Awareness in Animals and Humans: Developmental Perspectives. New York: Cambridge
University Press, 20-34.
--(1995a), 'Aspects of self: from systems to ideas', in P. Rochat (ed.), The Self in Early
Infancy: Theory and Research, Advances in Psychology Series. Amsterdam: North-
Holland, Elsevier Science, 95-115.
- - (1995b), Shame: The Exposed Self, paperback edn. New York: Free Press.
- - (1997), Altering Fate: Why the Past Does Not Predict the Future. New York: Guilford.
--(1999), 'Social cognition and the self', in P. Rochat (ed.), Early Social Cognition:
Understanding Others in the First Months of Life. Mahwah, NJ: Erlbaum, 81-98.
--ALESSANDRI, S. M., and SULLIVAN, M. W. (1990), 'Violation of expectancy, loss of
control, and anger in young infants', Developmental Psychology, 26 (5): 745-51.
------(1992), 'Differences in shame and pride as a function of children's gender
and task difficulty', Child Development, 63: 630-8.
--and BRooKs-GuNN, J. (1979a), Social Cognition and the Acquisition ofSelf New York:
Plenum Press.
----(1979b), 'Toward a theory of social cognition: the development of self', New
Directions in Child Development, 4: 1-20.
--and MICHALSON, 1. (1983), Children's Emotions and Moods: Developmental Theory
and Measurement. New York: Plenum Press.
--and RAMSAY, D. S. (1997), 'Stress reactivity and self-recognition', Child Development,
68: 621-9.
294 Michael Lewis
LEWIS, M., and RAMSAY, D. S. (1999), 'Intentions, consciousness, and pretend play', in
P. D. Zelazo, J. W. Astington, and D. R. Olson (eds.), Developing Theories of Intention:
Social Understanding and Self-Control. Mahwah, NJ: Erlbaum, 77-94.
- - - - (forthcoming), 'Self-awareness and pretend play in the second year of life'.
--SULLIVAN, M. W., STANGER, c., and WEISS, M. (1989), 'Self-development and self-
conscious emotions', Child Development, 60: 146-56.
LOWE, M. (1975), 'Trends in the development of representational play in infants from one
to three years-an observational study', Journal of Child Psychology and Psychiatry, 16:
33--47.
MCCUNE, 1. (1995), 'A normative study of representational play at the transition to lan-
guage', Developmental Psychology, 31: 198-206.
MCCUNE-NICOLICH,1. (1981), 'Toward symbolic functioning: structure of early pretend
games and potential parallels with language', Child Development, 52: 785-97.
MELTZOFF, A. N. (1990), 'Foundations for developing a concept of self: the role of imita-
tion in relating self to other and the value of social mirroring, social modeling, and self
practice in infancy', in D. Cicchetti and M. BeegWy (eds.), The Self in Transition: Infancy
to Childhood. Chicago: University of Chicago Press, 139-64.
--(1995), 'Understanding the intentions of others: re-enactment of intended acts by
18-month-old children', Developmental Psychology, 31: 838-50.
--and GOPNIK, A. (1993), 'The role of imitation in understanding persons and devel-
oping a theory of mind', in S. Baron-Cohen, H. Tager-Flusberg, and D. J. Cohen (eds.),
Understanding Other Minds. Oxford: Oxford University Press, 335-66.
MILLER, A. (1981), The Drama of the Gifted Child, trans. R. Ward. New York: Basic Books.
MOUNOUD, P., and VINTER, A. (eds.) (1981), La reconnaissance de son image chez l'enfant
et l'animal. Paris: Delachaux et Niestle.
NICOLICH,1. (1977), 'Beyond sensorimotor intelligence: assessment of symbolic maturity
through analysis of pretend play', Merrill-Palmer Quarterly, 23: 89-102.
PIAGET, J. (1952), The Origins of Intelligence in Children, trans. M. Cook. New York:
International Universities Press. Original French edn., 1936.
--(1954), The Construction of Reality in the Child, trans. M. Cook. New York: Basic
Books. Original French edn., 1937.
- -(1962), Play, Dreams, and Imitation in Childhood, trans. C. Gattegno and F. M. Hodgson.
New York: Norton. Original French edn., 1951.
POVINELLI, D. J., and EDDY, T. J. (1996), 'What young chimpanzees know about-seeing',
Monographs of the Society for Research in Child Development, 61 (3), serial no. 247.
ROCHAT, P. (ed.) (1995), The Self in Early Infancy: Theory and Research. Amsterdam:
Elsevier.
STAUDT, M., SCHROPP, c., STAUDT, F., OBLETTER, N., BISE, K., and BREIT, A. (1993),
'Myelination of the brain in MRI: a staging system', Pediatric Radiology, 23: 169-76.
- - - - - - - - - - - - a n d WEINMANN, H. M. (1994), 'MRI assessment of
myelination: an age standardization', Pediatric Radiology, 24: 122-7.
STIPEK, D., RECCHIA, S., and MCCLINTIC, S. (1992), 'Self-evaluation in young children',
Monographs of the Society for Research in Child Development, 57, serial no. 226.
TULVING, E. (1985), 'How many memory systems are there?', American Psychologist, 40:
385-98.
The Development of Self-Consciousness 295
VON BARTALANFFY, L. (1967), Robots, Men, and Mind. New York: Brazilles.
WATSON, J. S. (1994), 'Detection of self: the perfect algorithm', in S. T. Parker, R. W. Mitchell,
and M. 1. Bocchia (eds.), Self-Awareness in Animals and Humans: Developmental
Perspectives. New York: Cambridge University Press, 131-48.
WEINER, B. (1986), An Attributional Theory of Motivation and Emotion. New York:
Springer.
WEISKRANTZ, L. (1986), Blindsight: A Case Study and Implications. Oxford: Oxford
University Press.
WYLIE, R. C. (1961), The Self Concept. Lincoln, Nebr.: University of Nebraska Press.
ZAHN-WAXLER, c., RADKE-YARROW, M., WAGNER, E., and CHAPMAN, M. (1992),
'Development of concern for others', Developmental Psychology, 28: 126-36.
ZAjONe, R. B. (1980), 'Feeling and thinking: preferences need no inferences', American
Psychologist, 35: 151-75.
14
Perceiving Intentions
Joelle Proust
Among the many issues raised as part of the 'other minds' problem is the discov-
ery of the ways which justify inferring from public behaviour the existence of
mental predicates in others. How does one 'go beyond' physical moves, dynamical
patterns of bodily segments, to interpret these as buying a ticket, throwing a ball to
a dog, expressing joy? A traditional theory suggests that understanding others pre-
supposes understanding oneself. If a subject enjoys access to her own inner states
while she acts, she may be able to use analogical reasoning to attribute to others
intentions, beliefs, and desires corresponding to her own (Armstrong, 1968).
This way of framing and answering the problem draws on the intuition that self-
knowledge is the secure ground on which knowledge of others might be built. This
classical way of putting the problem may raise various objections. One may ques-
tion the Cartesian assumption according to which we learn about ourselves through
introspection, rather than by collecting information about the physical and the
social world. One may be worried by the verificationist flavour of choosing feelings
and sensations as the basis for our knowledge of others. It is furthermore contro-
versial that analogy might help bridge the gap between self and other minds. As Ryle
(1949) insisted, even granting privileged access to one's own intentions, analogy
does not solve the problem, for others differ from me, both in their 'observed
appearances' and in their actions (p. 53). Analogical reasoning does not seem to
offer a promising route to understanding others because at best it would only allow
for an understanding of actions and mental events already experienced by the
attributor. More radically, Wittgenstein showed that analogy falls short of attribut-
ing the relevant mental state to another mind: understanding someone else's pain
may simply consist in grasping how that might hurt oneself(in another body). The
analogy fails to bring in the notion of a different person having her own internal
states. How can a wincing behaviour be associated by an observer with pain felt by
someone else? How might an egocentric relation to her own action space help an
agent reconstruct alien movements as belonging to another agent's action space?
Many thanks to Naomi Eilan, Johannes Roessler, Paul Bernier, Jerome Dokic, and the members of
the APIC seminar at the former CREA, now Institut Jean-Nicod, as well as to the members of the
Philosophy of Action Seminar in Leipzig and to the audience of the Canadian Philosophy Association
2001 Congress for helpful comments on earlier versions of this chapter. Special thanks to Sliman
Bensmaia for his linguistic help.
Perceiving Intentions 297
The traditional view on the problem of other minds would be clarified, if not
quite solved, if it could be shown that the apparent gap between introspective and
social knowledge results from two mistakes: ignoring what can be directly
perceived about others and failing to acknowledge that information about oneself
can be gained by looking at the external world. In this chapter, we will concentrate
on the reasons we may have to claim that there is a common source of knowledge
for self and others' intentional behaviour. Our approach will, however, not
directly bear on mental understanding, but rather on one of its preconditions:
intentions, at least those of the most ordinary, physical kind, that give rise to
bodily movements and to changes in the external world.
3 For a full defence of this view, see Martin (1995). 4 See Roll et al. (1991).
5 Cf. Roll et al. (1990: 549-65).
300 Joe/Ie Proust
The cause of perceiving that one's legs are bent is that some specific muscles, ten-
dons, and joints are in the particular condition which causes the brain to be in
certain corresponding states underlying the perceiver's distinctive experience.
Proprioceptive states thus have a perceptual structure very similar to the visual
system's.
When it comes to intentional action, a perceptual theory seems at first blush
natural and adequate only in the case of a detached observation of another's
behaviour. It is a fact of experience that when a perceiver observes someone else's
bodily movements, she directly perceives these movements as goal-directed and
intentional. Moreover, what is consciously perceived and stored in memory is not
the pure sensorimotor aspect of the movement, but rather part of its teleological
content, that is, a specific dynamic interaction between behaviour and environ-
ment, as involving this or that part of the body, with this kind of timing and that
portion of space being a target of the action. These various demonstratives refer
to perceptual dynamical sequences which can be experienced without a conceptual
capacity coming into play. Now it is plausible to refer to what is immediately per-
ceived, in this kind of case, as the agent's intentional action, rather than as her
bodily motion.
Several arguments substantiate the claim that specialized perceptual mech-
anisms deal with the dynamical pattern of an observed movement. Humans can
easily discriminate a biological motion from a non-biological, mechanical move-
ment, even when the cues are limited to a few luminous dots located on the mov-
ing joints, with no background information available (Johansson, 1977). Such an
ability has also been shown present in 3-month-old babies (Bertenthal, 1993).
Newborns have furthermore been shown to imitate facial movements presented
in their visual field, such as tongue protruding, which suggests that they perceive
the stimulus in the format of a possible action (Meltzoff and Moore, 1995). I
Among the perceptual features that an infant might be innately disposed to use as
markers for intentional behaviour, searchers have identified the properties of
being a self-propelled movement (Premack, 1990), of having an irregular path
(Mandler, 1992), and of exerting non-rigid transformations on objects (Gibson
et al., 1978). Facial and vocal expressions are also cited as determining the value
of the goal for the agent (Haviland and Lelwica, 1987).
It will be objected here that this perceptual capacity to discriminate animate
movements does not amount to perceiving an action with a definite intentional
content. The recognition of an early perceptual capacity for discriminating inten-
tional agency in others should not lead one to jump to the conclusion that an
infant attributes intentions to others, for the latter capacity involves a perceptual
judgement using intentional concepts as well as the recognition of others as inde-
pendent selves. Just as an infant can manipulate a toy and extract the shape and
texture information concerning that object without forming the judgement
(that a toy is in her hands), she may look at the intentional action A of an agent 5
Perceiving Intentions 301
without recogmzmg (that 5 does A). Perceiving a biological movement as an
action, the objection goes, would require applying the whole explanatory struc-
ture of human agency in terms of intentions and reasons to act.
This argument thus rejects the transition from 'purely seeing' a movement that
is in fact an action to 'seeing a movement as an action', that is, it rejects the
transition from discriminating an intentional action to perceiving it as such, on
the basis of the additional cognitive resources that are needed for such a transi-
tion. A second argument would further object that such a transition is often
unwarranted. A well-coordinated and clearly target-oriented movement may fail
to be an action, for example, when the agent did not form any intention to act, as
in post-hypnotic compulsive behaviour, in imitation behaviour, where a subject
is involuntarily mirroring the actions she observes, or in utilization behaviour,
where she responds automatically to a functional stimulus. 6 So even leaving aside
the fact that many actions do not involve a physical movement-but rather
a refusal to move or speak-we must recognize that an animate movement is a
rather poor perceptual guide for action. There is more to action than biological,
target-oriented movement.
A first way of answering these objections consists in claiming that, without such
an innate sensitivity to this kind of movement in perception, a subject would be
unable to apply the concept of an action to specific kinds of dynamical patterns.
What is gained through perception is thus a sensory basis for later judgements,
when the concept of an action is acquired (independently from perception:
through language, learning, etc.). Now the question when a concept of intention
is grasped depends in part on a theoretical decision about what mastering a con-
cept amounts to. Several kinds of abilities have been suggested, such as an implicit
ability to attribute goals in pre-verbal communication (around 9 months),7
interpreting the goal-directed spatial behaviour as a marker of rational agency
(around 12 months),8 attributing intentions to dolls during play (around
20 months)9 or verbally attributing motives to agents (around 21 months),Io
This view thus grants that perceiving an action at a sensory level does not
amount to perceiving an action as such, but emphasizes that the perception of
actions plays a major role in the capacity to apply the concept of an action when
it is acquired. Even if the 'animate agency' feature may be unreliable in some
circumstances, it certainly is a precondition for reacting differently to physical
objects subject to natural regularities and those governed by motives.
Let us note here a difficulty of this position when it is associated with the view
that intentions are conceptually grasped only when a theory of mind is developed
(at around 3 to 4 years). Surely, non-human animals are able directly to see an
II We will presuppose here that non-human animals can use concepts to predict regularities in their
Let us first remark that, in order to draw this parallel, we need to exclude a large
family of cases of accessing one's own intentions that are irrelevant for it. A basic
distinction made by Searle (1983) will help us make this point. There are two
classes of intentions, named 'prior intentions' and 'intentions in action'. Prior
intentions are formed prior to the corresponding actions; intentions-in-action are
formed while the action occurs. There is a fundamental difference in the kind
of knowledge one can obtain when having one or the other type of intention.
For example, an agent may form the project to visit her uncle, and store this prior
intention in memory while her action develops. When an agent forms a prior
intention to perform action A, she uses her concepts and inferential capacities to
represent her goal as well as a relevant instrumental action. The question of how
12 SAM refers to a Shared Attention Module, and ToMM to a Theory of Mind Module. See Baron-
Cohen (1995).
304 Jodie Proust
the agent knows about her own intentions, in that case, can be solved by way
of postulating that the agent entertains the belief that she formed such and such
an intention with such a content. The agent may thus know that such and such an
intention will control her behaviour because she remembers having explicitly
formed this prior intention, and expects to find herself in the relevant context for
acting.
This kind of case diverges from the case of third-person intention-reading to the
extent that there seems to be, in the prior intention case, no public or perceivable
counterpart of the first-person, private event of intending. If we want to draw a fair
comparison between first- and third-person perception of intentional action, we
must therefore restrict our analysis to a situation in which no belief about a prior
intention is available to the agent while she acts, or at least in which such a belief is
not sufficient for the agent to dynamically keep track of her own doings.
Before we consider these restricted cases, let us first stress the importance of
perceiving one's own intentions (that will be called 'perception in action') in the
ordinary case. An agent who has stored conceptual knowledge about her own goal
in her working memory (and therefore believes she intends to A) has to check
what her progress towards her goal is. She must in particular appreciate whether
her current movement token converges to, or departs from, her pre-established
target. Not only a concept, but a perception of the action (or element of the
action) currently performed, is necessary to make the necessary adjustments and
corrections.
Therefore, even in the case where an agent formed a conscious prior intention
and carries it out at some later point in time, the agent must still have perceptual
access to her own developing intention. Important as it is, that case will not be con-
sidered in our argument because it mixes conceptual and perceptual expectations.
Three kinds of cases can be used to illustrate the necessity of recognizing a
perceptual level where the relevant evidence might be accessed:
(1) An agent may act without having formed any prior intention: she just acts
on the basis of an occurrent intention-in-action; scratching a bodily part,
pacing about a room, opening a door, shifting a gear, hitting a piece o~_
furniture out of anger, may all be performed in a purely automatic, non-
deliberate way. In that case, to know what she is doing, the agent must gain
access 'online', so to speak, to the kind of intentional action she is engaged in.
(2) An agent may perform several actions at the same time. In this situation of
divided attention, she must be able to reassess whether or not her occur-
rent behaviour conforms to her prior intentions.
(3) An agent subject to task interference may forget what the goal of her
current movement is.
The view defended here will be that, in these kinds of cases, an agent must rely
on a perceptual source of evidence to come to know what she is doing. This will
Perceiving Intentions 305
be shown by checking whether the conditions listed above for perceiving apply to
these kinds of situations. Let us start with condition (a). Having no belief
currently available on her prior intention, an agent is left to rely on specialized
captors to extract the relevant kind of information. Those specialized captors-in
particular an efferent copy mechanism,13 that allows a subject to keep track of
her own executive commands in the course of an action-exploit a subset of
the information available to the senses, namely, the stimuli having to do with the
anticipated feedback of the intended action. She perceives her own intention as
the particular way in which the world is given to her, that is, as a context afford-
ing some particular move or presenting some specific motivation-dependent
saliences. In other words, an intention can be recognized by the specific structure
and dynamics of the perceptual/motor field arising as a consequence of the atten-
tional features associated with the intended action.
There is again a much richer reading of an action context, in which the agent
masters a conceptual knowledge structure relative to that context, and hence can
use the inferential connections inherent in this structure to deploy her behaviour
rationally. An adult agent normally perceives her own intentions in a world
already conceptually categorized, just as she perceives the things in the world as
being trees, bushes, tables, and chairs. In a more basic sense, however, a context of
action can be described in terms of vectors and trajectories, salient affordances,
things with shapes, weights, colours, and fragrances, attractive and repulsive fea-
tures, and specific transformations. An infant presumably first senses her own
intentions in such a sensory way; she recognizes and identifies them through their
effects on her environment and her own body (including vision, audition, pro-
prioception, and verbal labelling by parents).
It is a commonplace of phenomenological inquiry that an observer directly
perceives the world as offering potential courses of action. The present approach
claims in addition that an action transforms the world in a way internally related
to the intention driving the action. 14 The agent can thus directly perceive her own
intention while her action develops. The meaning of the action lies, so to speak,
in the open. The specific organization of her visual field, resulting inter alia from
her moving intentionally through some particular action space, is part of what is
perceived, a material indication of what she means to do. Among the relevant fea-
tures for perceiving the intention driving one's behaviour are: the information
on the trajectory, on some of the relevant steps of the action, on the temporal
sequence of the various components of the action, on the effort involved in each,
on the modifications induced in the environment, and so on. Visual perception is
But here too various ordinary or pathological conditions may prevent someone
from perceiving what he or she is currently doing. Deafferented patients, for
example, may fail to identify what they do in case they cannot see their limbs.
Subjects with apraxia-who substitute incorrect actions into their own gestures-
also have difficulty in recognizing substitution errors in the gestures performed
by others and by themselves. In time pressure, or under toxic influence,
normal subjects may also fail to consciously perceive their own intention, and act
'without knowing what they do'.
15 This information can later be used to provide the content of a visual image for the corresponding
intention. Entertaining a prior intention before acting might thus consist in engaging in goal-directed
behaviour either on the basis of a conceptual representation of that goal, or on the basis of a perceptual
image of a prior successful token of the corresponding intention. Cf. James (1890, vol. 2: 487).
16 They also have problems with forming intentions, but these perturbations will not be discussed here.
Perceiving Intentions 307
In sum: intentions can be perceived or misperceived prior to their being
conceptually identified. Conditions (b) and (c) above are fulfilled: even when no
conscious belief is available for having formed an intention, there is some
distinctly phenomenological way associated with perceiving intentions through
the dynamic properties in the world contingent on one's agency.
Condition (c), that is, the ability to categorize perceived objects and properties,
will be accommodated in a different way according to the kind of strategy used to
address the objections raised above. In a nutshell, the two strategies differ on the
moment when categorization occurs: at the conceptual level, in the first stategy, and
ata sensory, proto-conceptual level in the second. In both cases, however, we use sen-
sory perception to judge that this (type of) intention is currently being carried out.
Our approach might, however, be put in difficulty by the two last conditions
in the list, namely (d) and (e) According to clause (d), what causes in veridical
instances the relevant perception of an intention should be nothing other than the
intention itself, appropriately connected to an intention detection mechanism.
According to clause (e), the perceived intentions should be shown to exist
independently from the fact that they are perceived.
17 The expression 'intentional content' in all its occurrences in this chapter means 'semantic
(2) 0 in L2.
The present account does not exclude that I may have a conceptual access to
(1) the fact that I intend to do P, (2) the object with the property on which real-
izing P depends (the target), (3) what is to be done to reach the target. As was
suggested above, a conceptual access allows a subject to memorize an intention as
part of a plan in a propositional format. The possibility of knowing an intention
conceptually should not, however, obliterate the possibility that an intention
could and even should be identified perceptually, when the agent has no conceptual
way of identifying her own current attempt.
Perceptual access to an intention-in-action has to do with the dynamics of an
action relevant for its intentional content, that is, with (3). Just as object-related
visual perception generally allows for the extraction of a sensory property or event
(seeing property F in an object, seeing an object undergo some change) and
forming a judgement (that grass is green, seeing the breaking of the vase), perceiving
an intentional behaviour allows for the extraction of agent-related information
about the dynamic properties of the world related to a target to be reached by her:
I go to the window and open it; I duck and avoid the stone thrown at me.
Now the puzzle about external versus internal evidence surfaces again in the
following way: is an agent's 'perception in action' distinct from the corresponding
intention-in-action or is it one and the same event? Confronting this question
faces a serious dilemma, for if it is recognized that perception-in-action and
intention-in-action are two ways of describing the same event, then the attempted
theory cannot qualify as a perceptual one. As we saw, one major clause for per-
ception (condition (d) above) is that the world having a property should cause the
corresponding perception in the observer. If, on the other hand, it is denied that
they describe the same event, then the task becomes one of finding out what kind
of relation connects perception-in-action and intention-in-action. What might
seem problematic is that, although an intention has a world-to-mind direction of
fit, it can nevertheless be perceived, and hence form the content of a mind-to-
world thought. How might the same thought-the intention perceived-be
simultaneously world-to-mind (in its executive dimension) and mind-to-world
(in its perceptual dimension)? Furthermore, given the time lag between when the
intention is formed and when the world is changed, how might an intention be
perceived if the world, so to speak, is only about to change?
Perceiving Intentions 309
Answering these questions adequately requires the clarification of the meta-
physics of the will. Let us start with the independence clause: how independent is
perception-in-action from intention-in-action? Forming an intention is the event
through which an action is initiated: when an intention is efficient, the world
undergoes a change as a causal consequence of the event of forming an intention.
Hence an efficient intention will generally fail to coincide temporally with the
dynamics realizing it, and with the corresponding perceptual event. Even in the
case of an intention-in-action, where muscle contractions are triggered without
any conscious prior intention, the dynamics of the action depend on the cerebral
process carrying it out. There must thus exist two events, one in which the pre-
motor cortex is activated (an activation instantiating a given intention), and the
other in which the muscular activity with the attached external changes takes place
(a corresponding action). It is during execution that the intention is controlling the
execution, which involves in part perceiving the relevant feedback. This gives us
licence to say that there are two events, one in which an intention is formed,
another in which an intention is carried out. It is only then that it can be perceived.
Such an independence is also manifest in the various misfirings that might
occur. An intention token may be formed but fail to trigger and/or control the
corresponding movement (e.g. because some new intention interfered with the
former), and therefore fail to offer visual or proprioceptive reafferences. Or it may
guide the bodily movements, but fail to be attended to: an agent engaged in a lively
conversation, for example, usually fails to perceive her own conversational ges-
tures, for lack of attention (her attention is engaged elsewhere, and, moreover,
they have in general a low saliency score for the speaker). If this is granted, then it
is correct to claim that the process through which an intention is perceived is dis-
tinct from the intention itself, that is, the event-token of a type which normally
generates and controls actions of that very kind.
Once the independence clause is secured, we are left with the first question
raised above: how is perceiving an intention connected to the intention itself?
We saw above that an intention causes a bodily movement and various changes in
the world, which can in turn be perceived. Still the representational relation
between efficient intention and action should be distinguished from the causal
connection between two events (forming vs. perceiving an intention). There is an
internal relationship between the representational content activated in the inten-
tion and the succession of bodily movements and interaction with the world in
the corresponding action. Where an intention is efficient, its content is preserved
in the causal process (from generating an intention to producing a bodily move-
ment). What is preserved is the instruction (3) deployed over time by the action.
310 Joe lie Proust
This explains why a physical action immediately presents its intentional content
to a perceiver.
Therefore we do have an answer to the first of our difficulties. There are two
distinct events in a causal relationship: the intention being read and the changes
it triggers in the perceived environment. The intention and the perception of that
intention have the same representational content (see (3) above), but with a dif-
ferent kind of direction of fit. The very difference of fit is a precondition for the
intention to control the action. What is at stake here is not only that the move-
ment that is executed satisfies what was intended, nor just that the properties in
the world are changed as intended, but also that this execution occurred and these
properties have been changed by virtue of that intention. The relatively fuzzy
expression 'by virtue of' corresponds to the experience referred to above as a
'sense of effort'. More precisely stated, this experience reflects the capacity to pre-
dict the reafferences, that is, to shift directions of fit as often as is needed for a
complete execution of the intention.
This kind of case again is not isolated. Consider the case when an agent A orders
B to leave the room. If the order is obeyed, A can observe that B performs the action
requested. What she observes in that case is the kind of feedback that she expects
when she utters the order. She perceives (in a mind-to-world direction of fit) a situ-
ation in the world that responds to her own intention (with a world-to-mind direc-
tion of fit). She expects the event to take place because she wants this event to
happen, and uses her own ability and social position to have it happen. Here, too,
there is a shift of directions of fit; one cannot act successfully without perceiving
(and judging) how the world responds to one's attempt at changing it.
Let us note in passing that the ability to shift directions of fit allows for an
explanation in economical, 'pre-subjective' terms, as to how a subject can get a very
basic sense of agency even if she does not grasp the concept of subject (and does
not use the first-person pronoun). The sense of being able in a given circumstance
to predict and modify the world by intending and carrying out meaningful actions,
that is, what is traditionally called the 'sense of effort', involves a kind of reflexivity
that does not involve any kind of metarepresentation. The process of holdillg the
conditions of satisfaction unchanged across the shift (from intending to perceiving
the intention developing through reafferences) generates an implicit non-conceptual
precursor of what could be made conceptually explicit through a metarepresentation
such as 'I perceive that I have this intention'.18
Now let us come to our final problem, concerning the temporal aspect of perceiv-
ing an intention in the dynamics of a movement. The time-lag separating
the event of forming an intention from the intention being made manifest in the
agent's behaviour should no more be an obstacle to perceiving intentions, as it is
to perceiving distant auditory or, in astronomy, visual phenomena. The temporal
18 On the indexical aspect of such a thought, see Burge (l991) and Pr9ust (2002).
Perceiving Intentions 311
puzzle specific to intention does not lie in the fact that it is perceived in a deferred
way-that is, not the common fact that the event in which it is identified is
posterior to the event in which it is generated-but rather in the fact that it can be
perceived even before the goal is reached. To answer this puzzle, we need to general-
ize a property of any kind of perception, that is, its aspectuality. A perceiver may
perceive a tree as a tree even though she only perceives the outline of its branches
in darkness; one can perceive a bird flyaway although what was perceived was a
brief flickering of a wing; similarly one can perceive an intentional process by being
exposed to some fragment of a bodily movement.l 9 Perceiving another agent's
intention may occur either in the attentional posture of someone, or in watching
her move and displace objects, or by seeing her reach her goal. Similarly for one-
self: one's own intention (besides the usual conceptual means for coding and
retrieving it) can be perceived and identified at some point in the dynamics of (3),
either very early in the saliences and dynamic projections in the perceived context,
or later in the succession of reafferences generated by the bodily movements. 20
19 On this, see Carey et al. (1997: 116). 20 See Roland (1978) and Pribram (1978).
312 Joe lIe Proust
a 'genuine' from an interfering, context-driven, action (and corresponding
intentional content)?
It is interesting to note that, in our examples, an intention supplanted another'
on the basis of the visual perception of the context in which the current sequence \
of the action was developing. Sight of a bedroom is associated with a definite
bedtime routine, that is, specific perceptual saliences and a sequence of actions.
James's case exemplifies the general fact that representations of action are .
normally embedded in canonical context representations. The latter seem to play'
a major role in both triggering and recognizing varieties of actions. 21 What hap-
pens in the two cases described is not that the agent misperceived her interfering
intention, for she correctly saw herself doing what she did (undressing, opening
the fridge). And it would be equally unfair to claim that the agent misperceived her
initial intention Il, because her action no longer reflected II. One should rather
say that the mishap occurred not at the perceptual level, but rather at the
executive level: the agent failed to maintain across time in working memory the
appropriate intention. Context overlap allowed an intention overlap to occur. This
overlap was beyond the agent's awareness, and as such was also in some sense
beyond her cognitive control.
It is all the more relevant to note that, on the basis of what she perceives as her
current intention, the agent may finally judge that what she does is not what she
meant to do. Perceiving her intention in her current action is a condition for exer-
cising her rational capacity at evaluating it, that is, understanding its irrelevance
with respect to her present goal, and finally reviving her initial intention.
Obviously, a mismatch can be detected only if what is observed can be compared
with what was anticipated: perceiving one's present intentions is a necessary pre-
condition for evaluating one's course of action. As was suggested above, the con-
ceptual content of an intention, as determined when forming it, normally
provides a template against which the current perceived intention is evaluated.
The agent who conceptually formed the intention to do R by way of F-ing gets
from perception some specific feedback that mayor may not be subsumed under
the relevant concepts. As was stressed by William James,22 activating a perceptual
image of her goal certainly plays a major role in this process. Visualizing in anti-
cipation what is to be done in some context allows her to evaluate how far the self-
induced transformations of the environment are the desired ones, and also helps
her guide her attention towards the relevant parts of the context. Where the agent
formed an intention which turned out to be non-efficient, there is no dynamical
scene developing as a consequence of that intention. Therefore the subject should
detect a mismatch by finding no correspondence between anticipated imagery
and observed scene.
6. SELF-GENERATED/EXTERNALLY
GENERATED PREDICTABLE STIMULI:
THE SEPARABILITY PROBLEM
28 Some patients seem on the contrary to misperceive intentions where there are none, i.e. to look
at the world as permeated with cues for a possible action. We will not address here the details of the
pathology. See Proust (2000a) for a discussion of the current theories.
29 On this question, cf. Feinberg (1978), Campbell (1999), Proust (2000a).
30 This case has been contrasted in Frith et al. (2000) with the symmetrical case of the anarchic
hand sign. In the latter case, the hand contralateral to a lesion in the supplementary motor area (SMA)
performs goal-directed movements which are, so to speak, unintended and intrusive. Although the
patient perceives the intentional content of these movements, she does not perceive them as caused by
alien forces. Such a patient has a problem of control, while the schizophrenic patient has a problem of
awareness of control.
31 One might speculate that this problem might result from an over-activity of the parietal area,
leading a subject to reinterpret other people's intentions in egocentric terms. Cf. Jeannerod (1999).
32 See among others Evans (1982), Shoemaker (1996).
316 Joe lie Proust
no such thing as introspection, if by 'introspection' is meant an inspection directed
to internal states. Still a world-directed kind of perception may be a source of
knowledge about oneself, as Evans already appreciated. 33 A paradigmatic 'internal'
state like intention can be directly perceived in the world, whether by the agent or
by an observer, in the dynamic pattern of the subject's interaction with objects and
people. Moreover, a specific marker in the developing pattern allows an agent to dis-
criminate among various features in the scene those changes contingent on her
action, giving her simultaneously a distinctive feeling of agency. Such a feeling is a
constitutive part of the experience of self-directed perception in action. When the
complex informational processing mechanism underlying the identification of self-
generated changes is defective, the agent is no longer able to detect her own agency
in the outer world. She becomes subjectively 'acted through'.
Shoemaker (1996) discusses a similar case in an argument to the effect that self-
blindness about intentions is incoherent. According to him, the potential inco-
herence of the case inheres in the fact that a subject has to possess an integrated
self to disavow an intention; otherwise the case might be described as involving
two subjects in a single body, the 'agent' and the 'agnostic' (i.e. the person dis-
avowing the intentions). These two roles only qualify as facets of one person if the
agnostic's beliefs and desires do rationalize the agent's actions and if she (the
agnostic) has the corresponding knowledge. Shoemaker concludes that it is neces-
sary to have introspective knowledge to one's beliefs and desires 'if one is to have
any access at all to one's intentions, volitions, and actions'.34
Such a conclusion is much too strong, however. If the view defended here is
correct, then one can have perceptual access to one's intentions, at least in the
restricted case of simple physical actions, even though no access to the (concep-
tual) content of one's beliefs or desires is currently available, just as one can have
access to others' intentions in a direct perceptual way, without further informa-
tion concerning the observed agents. Concepts of types of action, as well as infer-
ences from objects to affordances, will allow for the categorization of the
perceived intentional contents and for the prediction of behaviour in others in a
richer way, as was acknowledged earlier. The point is that informational access to
intentions is not constrained by the possession of concepts.
.When a schizophrenic patient denies having acted with a particular intention,
and even having acted at all, he or she mayor may not have beliefs and desires
rationalizing the corresponding type of action. Where she does not, we might be
tempted to say that she has grounds for supposing that some other agent (with the
relevant rationalizing beliefs and desires) made her act. But this temptation should
33 'Any informational state in which the subject has information about the world is ipso facto a state
in which he has information about himself, of the kind we are discussing, available to him. It is of the
utmost importance to appreciate that in order to understand the self-ascription of experience we need
to postulate no special faculty of inner sense or internal self-scanning' (Evans, 1982: 230).
34 Shoemaker (1996: 235).
Perceiving Intentions 317
be resisted, because the impression of being or not being the agent in one's actions
is not inferred from what one believes and desires. It is a genuine, direct feeling,
experienced dynamically in the changing world, a feeling functionally independ-
ent of the specific content of the corresponding intention. A patient may in fact
feel 'acted through' in actions she values as well as in actions she loathes.
Another difficulty plaguing discussion of self-blindness about intentions lies in
the failure to distinguish several levels in self-knowledge. Although it is tempting,
and in many cases right, to say that the mechanism for perceiving one's intentions
is a source of self-knowledge, it is not entirely correct. An agent may, properly
speaking, only have self-knowledge if she can form first-person thoughts, namely,
thoughts referring to the person having that thought according to a token-reflexive
rule, and if she is able to apply this rule in a general way, that is, as available to
other subjects. 35 An agent may, however, have thoughts that are in fact 'about him-
self' without realizing that he is having them; not only in the sense in which
Oedipus thinks that whoever slew Laius should die, that is, while utterly failing to
capture any reflexivity in the thought content, which, unknown to the thinker, is
in fact reflexive; but in the sense in which an agent may build up self-directed
thoughts while failing to have a concept of self. These latter thoughts can be said
to be self-directed either because they allow practical reflexivity to develop-as
when an agent identifies the target of a threat as herself, and flees; or because their
content is such that it involves the agent essentially (like proprioceptive informa-
tional states). The important point is that full-blown self-knowledge would not be
possible if there were no elementary steps through which an agent comes to grasp
her functional integrity.36
REFERENCES
ANSCOMBE, G. E. M. (1957), Intentions. Oxford: Blackwell.
ARMSTRONG, D. (1968), A Materialist Theory ofthe Mind. London: Routledge and Kegan Paul.
BARON-COHEN, S. (1995), Mindblindness, an Essay on Autism and Theory of Mind.
Cambridge, Mass.: MIT Press.
BERMUDEZ, J. L. (1995), 'Nonconceptual content: from Perceptual Experience to
Subpersonal Computational States', Mind and Language, 10: 333-69.
--(1998), The Paradox ofSelf-Knowledge. Cambridge, Mass.: MIT Press.
BERTHENTHAL, B. (1993), 'Infant's perception of biomechanical motions: intrinsic image
and knowledge-based constraints', in C. Granrud (ed.), Visual Perception and Cognition
in Infancy, Carnegie-Mellon Symposia on Cognition. Hillsdale, NJ: Erlbaum, 175-214.
BLAKEMORE, S.-J., and DECETY, J. (2001), 'From the perception of action to the under-
standing of intention', Nature Reviews/Neuroscience, 2 (Aug.): 561-7.
Jerome Dokic
1. INTRODUCTION
In his ground-breaking work on the will, O'Shaughnessy observes that 'we relate
epistemologically to bodily tryings as we do generally to (bodily) sensations'
(1980, ii: 81). In this chapter, I am interested in a close analogy, between action
and sensation. More precisely, I want to compare the intimate knowledge we have
of our own actions, and the distinctive way we know of our bodily sensations, like
pains, itches, and sensations of contact. Two aspects of the analogy will be particu-
1arly enlphasized.
First, both cases of knowledge have central properties in common with perceptual
knowledge. In particular, the states of affairs they are about are not significantly
different in kind from the states of affairs, accessible to external perception, of
other people being involved in feeling and acting. I shall go further and argue that
bodily experience, either of sensation or of action, has what I call the transparency
property, in the sense that it is often transparent or obvious to the subject that
the states of affairs disclosed by bodily experience are or are not of the same type
as the states of affairs she can observe through external perception. There is a fun-
danlentallevel of conscious experience at which others' sensations and actions are
presented to me in much the same way as I am aware of my own sensations and
actions. Arguably, the transparency of bodily experience explains how empathy,
imitation, and coordination are possible, in so far as these skills depend on the
perceptual ability to compare one's sensations and actions with others'.
The second aspect of the analogy is more familiar, but seems to point in
a different direction. Both cases of knowledge exhibit an asymmetry between the
first-person and the third-person perspectives. At best, I am aware of others'
sensations and actions 'from the outside', by observing their behaviour, whereas
I do not need observation in my own case. I typically know 'from the inside' that
I thank the participants of the 1999-2000 APIC seminar in CREA, Paris, where a first draft of this
chapter was read. I have been much helped by written comments from Ingar Brinck, Roberto Casati,
Julien Deonna, JodIe Proust, Anne Tuescher, and especially the editors, Naomi Eilan and Johannes
Roessler, who made stimulating remarks about several ancestors of the present chapter. What follows
owes much to them.
322 Jerome Dokic
I am in pain (I feel it), just as I typically know 'from the inside' that I am stretching
my arm. This asymmetry gives rise in both cases to a sense of ownership which
makes it difficult to imagine that I could feel someone else's pain, or that I could
know 'from the inside' what the other is doing.
How could bodily experience of sensation and action be transparent with
respect to external perception and yet so crucially different from it in being essen-
tially restricted to one's own case? In this chapter, I try to answer this question and l'
reconcile these two aspects. What we need is a proper explanation of the sense of
ownership in feeling and in acting. As we shall see, such an explanation has a com-
mon form, which does not presuppose that internal and external perception
introduce different modes of perception, or even that they involve altogether dif-
ferent ways of experiencing the world. Rather, what grounds the sense of owner-
ship is a constitutive relation between bodily experience and its intentional object
(what it is about), which makes such experience implicitly reflexive.
The chapter is divided into two parts. In the first part (Sects. 2-5), I raise the
question of the nature of bodily sensations, taking pain as a paradigm. In the
second part (Sects. 6-9), I ask whether and to what extent bodily experience is
necessarily involved in our awareness of our own actions. In the conclusion
(Sect. 10), I expand on a distinction which looms large in the rest of the chapter,
between psychophysical phenomena which are accessible through bodily experi-
ence, and non-spatial, fully psychological phenomena which are accessible
through non-perceptual introspection. Hopefully, this distinction can lead to a
better understanding of the distinctive form of self-awareness involved in bodily
experience.
bodily experience immediately presents to the subject is her body as a system of possible movements
(compare Merleau-Ponty, 1962). Of course, she can then ground on such experience a knowledge of
non-psychological (mainly spatial) properties of her body.
324 Jerome Dokic
Of course, it immediately follows that (1) is insufficient to capture what is spe-
cial about perceiving F 'from the inside', since the specified experience might as
well be a case of external perception. More generally, any intentional object of I
bodily experience can be the intentional object of external perception, in particu-
lar by others. Whatever pain is, what I am perceiving in my hand when I feel pain
is in principle open to observation by anybody else (or by me in the mirror).
Indeed, a stronger claim can be made. Not only another person's sensations can
be directly perceived, but the fact that the observer herself has similar (or differ-
ent) sensations can sometimes be fully transparent to her. This claim I shall call
the transparency of bodily experience. It seems to be a necessary condition of at
least some cases of empathy. There are empathetic situations in which types of
bodily sensations are immediately (i.e. non-inferentially) recognized as shared. 2
The transparency of bodily experience sets a constraint on any adequate
account of the difference between observing the instantiation of property F and
experiencing it 'from the inside'. Suppose that one tries to explain the relevant dif-
ference, not at the level of reference, but at the level of the perceptual modes under
which the 'scene' is presented to the subject, as in the following description:
(2) Experience (a particular body part is F [where Fis presented under mode M])
Here M is a mode of presentation of a property. When a bodily condition is
experienced 'from the inside', it is presented under mode M, whereas when it is
externally perceived, say visually, it is presented under the different mode M'.
However, if perceptual modes of presentation are conceived on a model analogous
to that of Fregean modes of presentation, it might be impossible to recognize
immediately that, for instance, the painful condition I am now in is exactly sim-
ilar to the one I simultaneously observe in you, just as it might be impossible to
recognize immediately (without an inference involving an identity thought) that
one and the same planet is presented both as Hesperus and as Phosphorus. The
discovery that Hesperus is Phosphorus was a substantial one, and the same may
hold for the discovery that a property presented under mode M is the same as a
property presented under mode M'. This is not to deny that I sometimes infer that
you and I are both suffering from pain. The claim is only that it can be otherwise.
Often, it is just transparent to me that we are in similar bodily states, although
only one of them is presented to me 'from the inside'.3
2 Interestingly, Hutchison et al. (1999) have recently discovered cortical neurons which respond to
painful stimuli in humans. Some of these neurons are activated by pinpricks to the subject's hand, and
also respond to the observation of pinpricks being applied to someone else's hand. These neurons can
be compared to the 'mirror neurons' in the action case; cf. n. 9 below. I should say that the relevance of
this discovery to the claim that bodily experience is transparent at the personal level is not watertight.
3 An alternative would be to explain the difference between external perception and bodily experi-
ence at the level of the perceptual mode under which the body part is presented to the subject.
Properties F themselves would be perceived in the same way in both cases. This alternative is
implicitly excluded at the end of Section 5, where it is shown that the body part is not identified
independently of the instantiation of properties F.
The Sense of Ownership 325
An analogous point holds for any conception according to which internal and
external perception pertain to different kinds of cognitive faculties. On such a
conception, if internal and external perception both belong to the genus 'percep-
tual experience', they are distinct species of that genus. One can thus modify
(1) in order to make explicit the species of the relevant experience:
(3) Internal experience (a particular body part is F)
The transparency of bodily experience is not easier to explain or even acknow-
ledge on the assumption that perceiving one's body 'from the inside' involves an
altogether different kind of intentional relation from that involved in external
perception. These relations seemed so different to Merleau-Ponty (1962) that he
concluded, from the correct observation that bodily experience is intentional (it
is about conditions of the body), that the 'space' in which the objects of such
experience are presented is not objective space, but a special kind of space.
(For instance, he talks of a 'pain space'.) More generally, the more we stress the
difference between internal and external perception, the less we are entitled to
conceive them as being about the same kinds of states of affairs.
4. IMPLICIT REFLEXIVITY
At this stage, I would like to explore the possibility that the sense of ownership
peculiar to bodily experience results from facts about the perceived properties
F--more precisely, from a constitutive relation between the instantiation of these
properties and the experience itself. 4
4 Of course, there are in principle other options. For instance, one might try to explain the sense
of ownership by reference to a further quality present in bodily experience, over and above the
property F and the spatial content of the experience.
The Sense of Ownership 327
It is not implausible to suppose that human pain has the essential property of
forcing itself into its subject's field of consciousness. (Strong pains have the
further property of clamouring for attention.) No one can have pain, in this sense,
who is not aware of it; for pain, esse est sentiri. So when one feels pain, as opposed
to observing it in others, one's experience is both about the pain and required by
it. Accordingly, I propose the following as a roughly correct description of the
experience 'from the inside' of one's bodily properties:
(4) Experience (a particular body part is F), where the perceived instantiation
of F is constitutively dependent on this particular experience
In this description, the term 'this particular experience' refers to the bodily
experience itself. Feeling pain is an experience with a spatial content; it presents a
body part as being both located in space and hurting. Now, hurting is a psycho-
logical property in so far as its instantiation entails the existence of a particular
experience. So bodily experience presents states of affairs which are both spatial
and psychological. Brewer writes that 'in bodily awareness, one is aware of
determinately spatially located properties of the body that are also necessarily
properties of the basic subject of that very awareness' (1995: 300). Indeed, (4) can
be seen as a development of Brewer's general claim, adding the notion of implicit
self-reference to be introduced right away. This is a radically anti-Cartesian inter-
pretation of body ownership. What is presented in (veridical) bodily experience
does not merely belong to me; it is a part of myself As Brewer puts it, 'the subject
of experience extends physically to encompass the bodily location of sensation'
(1995: 297).
On the present account, bodily experience is reflexive or self-referential, in the
sense that it is about instantiated properties which entail the experience itself.
However, it is implicitly rather than explicitly reflexive. What I mean is that (5) is
not an obviously valid consequence of (4):
(5) Experience (a particular body part has the property Fwhose instantiation
is constitutively dependent on this particular experience)
In (5), the fact that the instantiation of F depends on a particular experience is
explicitly presented in the subject's experience (its content is about that fact). In
general, bodily experience might be fine-grained or intensional enough to
block the inference from (4) to (5). It might lock onto the property F without
locking onto the different property of being constitutively dependent on a
particular experience, even if necessarily, the instantiation of the former involves
the instantiation of the latter. Analogously, I perceive the presence of water
without perceiving the presence of H 20, even if the former necessarily entails
the latter.
There is a deeper reason to be sceptical about (5), though. According to (4), a
particular bodily experience is about a psychophysical state of affairs which is
necessarily associated with this very experience. It does not follow from (4) that
328 jerome Dokic
when I feel pain in my arm, I locate my experience of pain in my arm, or anywhere
else. There is no perceptual faculty capable oflocating experiences as such. More gen-
erally, we should distinguish between the scarcely intelligible idea that we directly per-
ceive non-spatial experiences, which seems to be an unfortunate consequence of (5),
and the much less extravagant claim that we perceive something-for example, a
pain-which does not fall short of the fact that there is someone who experiences it.
In discussing knowledge of other minds, McDowell (1986: 212) distinguishes
between a 'model of direct observation', appropriate for instance to facial expres-
sions of emotional states, and a less straightforward application of the perceptual
model, more appropriate to 'avowals'. In the latter case, one does not perceive the
other's 'inner' state itself, but rather something which (does not fall short of it', and
can be described as 'her giving expression to her being in that "inner" sense'. Here
I am drawing a somewhat similar distinction with respect to the first-person case.
I can directly perceive, (from the inside', psychophysical conditions of my own
body. These conditions entail conscious experiences which I cannot directly per-
ceive, even 'from the inside'. Contrary to McDowell, perhaps, I would allow that it
is possible to experience some of these conditions without representing them as
being constitutively linked to the entailed experiences and attitudes, especially if
the subject does not possess the relevant mental concepts (see Sect. 10).
In a nutshell, bodily experience is indeed a variant of external perception. The
states of affairs experienced 'from the inside' are not different in kind from those
that can be observed 'from the outside'. Properties F are accessible to both internal
and external perception, although in neither case can the underlying non-spatial
attitudes be directly perceived. So external observation of another person's being
F has the following form:
(6) Experience (a particular body part is F), where the perceived instantiation
of F is constitutively dependent on the other's experience
Here, the difference between internal and external perception of someone's
being F is traced to the fact that in the former case, bodily experience is precisely
the conscious experience required by the instantiation of the property F. In
contrast, observing pain in someone else (or in oneself in the mirror) introduces
two experiences: the observation itself, and the conscious experience required by
the pain.
5 In exceptional cases, such as that of Sianlese twins, it seems that the same body part can be per-
ceived 'from the inside' by more than one person. When they both feel pain at a conlmon limb, Siamese
twins do not perceive the same states of affairs, since there are two instantiations of the property F,
which are implicitly linked to distinct experiences. But clearly they extend physically to encompass the
same bodily region. Does it follow that they share a part of themselves? Not necessarily. Depending on
the specific nature of the property F in question, the body part will be perceived as solid and impene-
trable, like a material limb, or as more ethereal and diaphanous, like a cloud. Kinsbourne (1995:
214-15) reports experiences in which two body parts are perceived as occupying the same location.
(Kinsbourne's own explanation is that 'the representations of separate body parts are separately
referred to their spatial locations and then related to each other'.) Obviously, these experiences are illu-
sory, but it does not follow that they are about impossible states of affairs. Some bodily experiences are
more capable than others of giving the subject a sense of a solid and impenetrable body. Perhaps we
can imagine Siamese twins who perceive exactly similar pains in what we (somewhat misleadingly)
describe as their common limb. It does not follow that they share a part of themselves, since we can
also imagine two clouds occupying the same position in space, for instance if they are passing through
each other. It is more difficult to imagine Siamese twins who simultaneously perceive 'from the inside'
bodily activity at the same place. Here it seems that if one of the twins takes control of the material
lin1b, it ceases (at least momentarily) to be a part of the other's body.
330 Jerome Dokic
This is an important objection. It certainly shows that there is still work to be ,
done, but it does not undermine the project of securing the ontological fact of
ownership so as to warrant the epistemological sense of ownership. What makes
bodily experience special, in comparison to vision and audition, is a fact external
to it, namely, that it happens to be the experience entailed by the perceived
(instantiated) properties. It does not follow that this fact is external to the subject's
cognitive system, that it has no cognitive significance. The general idea that the
subject can manifest a sensitivity to that fact not by representing it, but implicitly,
by differentiating on a functional level internal perception from external percep-
tion, has not been shown to be incoherent.
One way (perhaps not the only one) of implementing the general idea would
be to claim that the sense of ownership is at least partly determined by the sub-
ject's finding the transition from a particular case of bodily experience to a judge-
ment like 'My arm is hurting' primitively compelling in Peacocke (1992)'s sense. Of
course, such a transition is not to be construed as an ordinary inference, since
nothing in the content of experience itself warrants the conceptual self-ascription.
Although the bodily experience and the judgement are distinct mental states with
different, overlapping contents, the transition from the former to the latter is war-
ranted in part because the judgement makes explicit something that was already
implicit in the bodily experience. 6
It is a well-known observation that bodily experience is 'immune to error
through misidentification relative to the first person' (IEM for short).? If! experience
something 'from the inside', I can be wrong about what it is exactly, or where it is
located in body-space, but I cannot be wrong about whose body I am presented
with. The standard explanation of IEM is in terms of normal ways of gaining
knowledge of one's bodily properties (Evans, 1982; Perry, 1990). If I acquire
information that a property F is instantiated in the appropriate way (e.g. propri-
oceptively), I have the information that I am F. Up to this point, the standard
explanation mal<es no appeal to the nature of the property F, which might well be
purely physical (like the property of having one's legs crossed). However, we now
have to face the following dilemma. On the one hand, if the normal ways of gaining
information about oneself are too much like the ways one gains information
about others, it will be a mere accident of our cognitive architecture that they
yield information about oneself, rather than about someone else. The difficulty in
this case is to explain how bodily experience can ever be a non-inferential ground
for the judgement 'I am F' and convert it to knowledge. On the other hand, if justice
is done to the relevant contrast between internal and external perception, the normal
ways of gaining information about oneself will be very unlike the ways one gains
6 On rational transitions that manifest the subject's sensitivity to properties of mental states over
and above their explicit contents, cf. also Peacocke (1999). Much of what Peacocke says there about
such transitions and their epistemological significance can be applied to the present case.
7 The phrase is Shoemaker's; cf. his (1994), and Evans (1982).
The Sense of Ownership 331
information about others, and bodily experience won't be transparent. As we have
seen in Section 2, if internal and external perception are pictured as different kinds
of experience, a theory will almost certainly be needed to link what I feel 'from the
inside' to what I observe in others.
On the present reflexive account, this dilemma is escaped in the following way.
Of course, what it is like to acquire and use information about one's body by feel-
ing pain is not the same as what it is like to acquire and use the same information
by observing someone in pain. However, the relevant difference, which results in
.the fact that bodily experience is IEM, is explained at a more fundamental level
which takes into account the nature of the property F. Given the constitutive
dependence of the relevant instantiation of property F on bodily experience itself,
the latter simply cannot ground a judgement of the form 'Someone else's arm
hurts'. If such a judgement is made by a subject on the basis of her own experi-
ence, it can never amount to knowledge.
To sum up, bodily experience gives us a sense of ownership at least partly
because it is necessarily about a particular condition of my own body. This is
explained by the fact that the condition itself essentially depends on the very same
experience which discloses it to me. I get a sense that I physically extend to places
where I feel something in part because what I feel depends on my being aware of
it. Bodily experience and vision are both cases of spatial perception. What seems
to make bodily experience short-sighted is not that it places its object in a queer
space; it is just that what is called 'bodily experience' is perceptual experience as of
implicitly reflexive conditions located in objective space. Moreover, when I have a
bodily sensation, no one but me can be aware of the corresponding state of affairs
'from the inside', at least to the extent that no one else can have the very same
(numerically determined) experience as me while I am having it.
In the first part, I argued that the intimate knowledge we have of our own sensa-
tions is based on conscious bodily experience of these sensations. Now, we also
have a distinctive knowledge of our own physical actions, and it is now time to
enquire whether and to what extent such knowledge is grounded on conscious
experience.
A case can be made for the view that experiencing one's body in action is what
corresponds in the action case to feeling pain in the sensation case. We have
rejected the Cartesian picture according to which our sensations can be known
independently of their spatial locations (even if the latter can be rather indeter-
minate). Similarly, we should discard the equally Cartesian view that knowledge
of our actions does not involve any sense of how our bodies are actually moving.
Intuitively, bodily experience plays a central role in awareness of our own actions.
332 Jerome Dokic
The analogy between the sensation case and the action case can be pushed in
two further directions. First, awareness of our own actions seems to share the
transparency property with consciousness of sensations. Just as we are aware of
what we are doing 'from the inside', through bodily experience, we can directly
observe other people acting intentionally. Now, it seems that we have the ability to
recognize immediately that our action shares or does not share crucial properties
with the other's. For instance, I do not need to draw any inference to see that you
and I are trying to catch the same ball. Without such an immediate recognition,
action coordination would be much more difficult than it actually is. As in the
sensation case, the transparency of our experience of actions is best explained on
the hypothesis that there is a level of awareness at which my action is presented to
me in much the same way as the other's is presented to me. 8
The second point of analogy is related to the first. If our awareness of actions is
transparent, the difference between observing actions in others and experiencing
oneself in action must be explained otherwise than at the level of the contents of
the relevant experiences. In the sensation case, I have suggested that consciousness
of pain is implicitly reflexive: typically, pain is such that it forces itself into its sub-
ject's field of consciousness. Analogously, it can be argued that many actions are
such that their existence entails some consciousness of them on the agent's part.
Perhaps action performed automatically or absent-mindedly is not consciously
experienced by the agent, at least 'from the inside', but it is difficult to imagine
that, for instance, a dangerous or difficult action can be done without the agent
having any sense, however dim, of what she is doing. 9 In the case of controlled
actions at least, the agent is fully involved in her doing what she does, and her
experience at that moment is incompatible with the intentional activity in her
body being controlled by someone else, or not controlled at all.
So awareness of action is typically reflexive, in the sense that action has
the essential property of grabbing the agent's consciousness. However, just as in
the sensation case, there is reason to suppose that such reflexivity remains implicit
in the functional properties of the state of awareness. Intentional bodily activity
is experienced as a spatially located phenomenon, but the experience itself, being
8 As is now well known, there is empirical evidence that observing someone else's action and
performing the same action are underlied by overlapping subpersonal mechanisms. Rizzolatti and his
colleagues discovered so-called 'mirror neurons' in the pre-motor cortex of macaque monkeys (ef. e.g.
Rizzolatti et aI., 1996 and Jeannerod, 1997). These neurons fire both when the monkey actually
performs an action, and when it merely watches another monkey performing the same action. Similar
neurons are thought to exist in human beings as well. As in the case of pain, it is a delicate question
whether and in what ways these mechanisms ground the transparency of bodily experience at the
personal level.
9 My remarks here are meant to echo the famous distinction drawn by Norman and Shallice (1986)
between two mechanisms of action control, which they call 'Contention Scheduling' and 'Supervisory
Attentional System'. I take it that it is partly an empirical question what kinds of action require
implicitly reflexive consciousness in the sense explained here.
The Sense of Ownership 333
a non-spatial mental state, cannot be perceived as such, either 'from the inside' or
externally. All that is perceived 'from the inside' is that someone is doing
something in a certain way. So when specific properties F such as cases of bodily
activity are taken into consideration, (4) remains a more perspicuous description
of bodily experience than (5).
How is it possible, then, to ground a first-person judgement like 'I'm grasping
this glass' on an experience which is both transparent and only implicitly reflex-
ive? Answering this question amounts to explaining, in the action case, how the
sense of ownership can flow from the fact of ownership. We have seen that bodily
experience of sensations participates in our sense of ownership; the very idea of
feeling someone else's pain does not seem to make sense. It is plausible that
bodily experience of our own activity plays a similar role. When I experience bod-
ily activity 'from the inside', I necessarily feel some of the boundaries of my
bodily self. The very idea of feeling someone else's body intentionally moving is
prima-facie incoherent.
As in the case of bodily sensations, the sense of ownership can be explained by
reference to the more fundamental fact of ownership. When I experience bodily
activity 'from the inside', what I experience cannot be a particular condition of
someone else's body, because it is ontologically dependent on my own awareness.
This is something to which the subject can be rationally sensitive. For instance, the
transition from bodily experience of my intentional activity to the judgement that
I am grasping a particular glass will be both primitively compelling (in the nor-
mal case) and warranted.
10 It does not follow that whenever I intentionally act, I have to attend to every aspect of my own
bodily activity, for instance, in singling out a body part demonstratively, as 'that arm'. This would be
incompatible with the spontaneity and smoothness of many of our intentional actions. Attending to
the movements of one's fingers while at the piano almost always leads to playing wrong notes.
Nevertheless, these movements can be experienced 'from the inside' by the agent, if only peripherally.
The Sense of Ownership 335
action, but knowing it is not enough to know what one is doing. Suppose that I
want to drink from a glass before me. There is an indefinite number of specific
bodily movements which will get me what I want, but I do not have a detailed rep-
resentation of anyone of them at the beginning of the action. I cannot anticipate
the precise course of the movement. When my action involves controlled bodily
activity, I fully know what I am doing only while acting. I know that my action has
been successful not merely because I am aware of the goal being attained; I also
experience my hand as being continuously guided towards the glass.!!
In fact, we can be more precise about the relationship between control and con-
sciousness. Our first-person experience of controlled bodily activity has two
related aspects. First, it supervenes on subpersonal processes of comparison
between behavioural outputs and proprioceptive and exteroceptive inputs. At that
non-conscious level, the motor system confronts its predictions with behavioural
results as it records them, and makes the necessary adjustments. These processes
are essential to action-monitoring, as they allow for a smooth development of the
action towards the goal (Jeannerod, 1997; Russell, 1996, 1998; Pacherie, 2000).
Moreover, as James Russell pointed out, it is crucial that they are subpersonal; if
this aspect of action monitoring was done at the personal level, our tryings would
be related to bodily activity only externally, as in the picture we have just rejected.
However, subpersonal processes of comparison and adjustment are not enough
to give rise to a first-person experience of control. What makes a bodily action
controlled is that it mobilizes the agent's consciousness in a specific way, requir-
ing not only a representation of the goal but also an ongoing representation of the
best way of attaining the goal, given the constantly changing circumstances of
action. Now, the second aspect of our first-person experience of control
arises from the fact that processes of comparison and adjustment themselves
are governed, on the whole, by these conscious representations. The 'decisions' of
the motor system are made 'in the light of' the agent's understanding of what
she is currently doing. There need be no circularity in this causal influence: the
agent's consciousness of what is to be done partly determines what adjustments
are to be made by the motor system, and the actual adjustments modify in turn
. the agent's consciousness of what she is doing.!2 Moreover, the agent herself is not
11 What I say here is compatible with the view that success is an essential property of the (success-
ful) action. This leads to a disjunctive account of tryings. When someone tries to do something,
her trying is either a mere trying, or a successful action, these being essentially different species of
psychological states. O'Shaughnessy (Ch. 16, this volume) seems to defend an opposing view.
12 This sketchy account allows for different levels of control, depending on the specificity of one's
conscious representation of the goal and of the best way of getting at it. Absent-minded behaviour can
count as a 'weakly controlled' action if it is internally governed by appropriate conscious representa-
tions. For instance, if my general goal is to drive home, my quasi-automatic driving may nevertheless
require a schematic representation in consciousness of the best way home (given the circumstances). I
would say that my driving counts as a voluntary action in so far as it entails such a conscious repre-
sentation, even if it is not very detailed. See Perner (Ch. 10, this volume) for a taxonomy of action. One
point of disagreement with him will pop up in Section 9 below.
336 Jerome Dokic
conscious of the separate moments of these top-down and bottom-up processes,
which take place partly at the subpersonal level. What she does consciously
perceive, though, is the development of her action as being at the service of the
smooth attainment of the goal. (Whether this requires reflection in the sense that
the goal is explicitly represented as the object of a pro-attitude will be discussed
in Section 9.)
One might concede that perceiving an action as developing through the agent's
consciousness of the goal is phenomenologically different from perceiving a mere
behavioural schema coming to fruition, but complain that it does not explain the
first-personal character of our experience of control. Once again, though, the dif-
ference between the first-person case and the third-person case is not to be found
at the level of the contents of the respective experiences. Given the transparency
of bodily experience, we can have external perception as of someone else's action
being controlled. The difference lies in non-representational, architectural facts
about bodily experience. What makes an experience of control first-personal is its
implicit reflexivity-more precisely, the fact that in this case, the consciousness of
the goal which governs the development of the action and the awareness of the
action itself as it unfolds are inseparable.
Of course, it follows that bodily experience of controlled activity does not sit
comfortably on either side of the traditional divide between perception and
action. It is a case of exercising our bodily will which at the same time gives us a
sense of what we are doing. Consciousness in acting, because it essentially involves
bodily experience, yields some consciousness ofthe action. Controlling one's move-
ments is one way of knowing what one is doing, even though it is an entirely
practical way.
The second objection I would like to consider starts with the contention that the
proposed analogy between the sensation case and the action case is crucially
imperfect. Whereas knowledge of our sensations is based on experience of these
sensations, it is not clear that knowledge of our actions is based on experience of
these very same actions. An action is identical with a trying, which is at least partly
a non-spatial psychological event. Perceptual experience, though, is always as of
spatial states of affairs. It follows that the model of direct observation is no more
appropriate to actions than to pain experiences. Arguably, we directly perceive,
'from the inside' or not, voluntary activity of the body. When I act, my body feels
active, and I can direct my attention to various aspects, phases, and patterns of this
activity. However, what I feel in my hand when I am moving it cannot be identi-
fied with my action of moving my hand. An action is ascribed to a whole agent,
and if the question of its spatial location arises at all, we should say that it derives
The Sense of Ownership 337
from that of the agent herself or from her wider situation. In contrast, different
types of activity can be located in different regions of body-space, so that I
can in principle focus on the activity of a specific part of my body, for example,
my arm stretching out. Moreover, two simultaneous actions of the same type tend
to fuse. If I try to move my left hand and at the same time try to move my right
hand, I will end up trying to move both hands. If there is one agent here, there is
only one overarching trying, not two. This is not so with bodily activity. The
same type of activity can be simultaneously located in two different regions of
body-space.
Although actions cannot be directly perceived, externally or 'from the inside',
we can perceive something which does not fall short of the fact that someone is
acting, that is, bodily activity. Bodily activity, though not strictly identical with
the action, is a constitutive part of it. It would be a mistake, then, to identify
bodily activity with mere physical movement that can be fully described without
mentioning the presence of an agent. In some cases at least, my arm and hand are
controlled and guided at a psychological level, in a way which (as we have seen)
cannot be correctly understood otherwise than as an essential contribution to the
realization of a goal-directed action. ,Bodily activity, in the sense presupposed
here, is essentially voluntary. Feeling as if my hand is moving is not the same as
feeling as if I am moving it. In the former, but not in the latter case, I am aware of
sensations which fall short of the fact that I am engaged in an intentional action.
Three relevant facts should thus be kept apart: l3
(i) that I stretch my arm (an action)
(ii) that a part of my proper body-my arm-is active (a case of bodily
activity)
(iii) that there is a physical movement of arm-stretching (an objective,
'intransitive' movement)
When I act, bodily experience presents me with a fact of type (ii), and often indi-
rectly with a fact of type (iii) (I am aware that there is a physical arm-movement
by experiencing 'from the inside' my arm stretching out), but not with the action
itself. Actions are not among the values of the variable F in the proper description
of bodily experience (4). What is plausible, though, is that bodily activity is
an essential component of physical action (which does not mean that any
physical action is a trying to move one's body),
13 Hornsby (1997: 102-10, esp. 107) draws a similar distinction between two sorts of movements
involved when there are actions. There are 'colourful' movements that are intelligible only to someone
who uses action concepts, and there are neutral or 'colourless' movements which have nothing to do
specifically with actions. This corresponds to the distinction made here between facts of type (ii) and
facts of type (iii). What is less clear is why Hornsby insists on there being only a causal relation between
actions and the corresponding colourful movements, as if they could exist exactly as they are in the
absence of any acting.
338 Jerome Dokic
9. THE ROLE OF INTROSPECTION IN
AWARENESS OF ACTION
14 More precisely, the level corresponding to Norman and Shallice's Supervisory Attentional
System.
15 As I shall suggest in the concluding section, there might be a more basic use of action concepts
which does not require reflection on the underlying pro-attitude, like in 'X is doing this', where 'this' is
a demonstrative grounded on perceptual experience of bodily activity.
340 Jerome Dokic
experiential and belief-independent, the latter involves introspection and is belief-
dependent. 16
As a consequence, there are at least two cases where the identification of
perceived bodily movements as one's own can go wrong:
(i) The capacity of experiencing controlled activity 'from the inside' is
impaired. The subject perceives externally (i.e. not 'from the inside') I
16 Does the belief-independence of our basic sense of ownership undermine our previous claim
that this sense can be defined in terms of primitively compelling transitions to judgements? I do not
think so, for what is primitively compelling can be resisted. For instance, I can find the transition
between my visual experience of a Miiller-Lyer figure to the judgement that the two lines are unequal
primitively compelling, while in fact I know better and judge that they are equal.
17 In fact, schizophrenic patients have perceptual deficits which also disrupt,· at least partly, the
basic sense of ownership; for further thoughts, cf. Campbell (1999), Jeannerod (Ch. 5, this volume),
and Proust (Ch. 14, this volume).
The Sense of Ownership 341
10. CONCLUSION: INTROSPECTION VERSUS
INTERNAL PERCEPTION
In this concluding section, I would like to say a few more words about the rela-
tionship between internal perception and non-perceptual, introspective know-
ledge of our thoughts and experiences. Shoemaker famously opposed the classical
conception of introspection as a perceptual or quasi-perceptual faculty. His main
objection against that conception is worth recalling here. He notes that under cer-
tain conditions, it is impossible to be 'self-blind' to one's own beliefs and desires;
having them somehow entails being aware that one has them. The relevant con-
trast between objective or intentional perception and introspection is that in the
latter case, 'the reality known and the faculty of knowing it are, as it were, made
for each other-neither could be what it is without the other' (1994: 245). So one
does not perceive one's own beliefs and desires 'from the inside' because their very
existence entails our awareness of them, and vice-versa-the logical gap between
appearance and reality has not been provided for.
Shoemaker extends his observation to pains and exercises of the will. As it is
impossible to be self-blind with respect to pains, it is impossible to be unaware of
one's own actions, at least to the extent that they are intentional. I think that
Shoemaker underestimates the distinction between introspecting a non-spatial
thought or experience and being aware of a property of oneself through bodily
experience. Let us grant Shoemaker's argument that introspection is not a species
of objective perception, not least because the notion of an introspective illusion
does not in general make much sense. 18 Bodily experience is a different matter,
since its objects are clearly presented in space, can attract attention, and can be
illusory in some respects. For instance, in the illusion of the Japanese hand, it may
seem to me as if I am raising a finger of my left hand while it is really one of my
right hand's fingers which is intentionally moving.
As O'Shaughnessy (1980, ii: 75) remarked, there is no such thing as the episte-
mological relation between a man and his psychological properties. Immediate
self-knowledge of one's own thoughts and experiences is one thing, perceptuai
awareness 'from the inside' of properties of one's bodily self is another thing. It
might well be that in the former case, the reality known and the faculty of know-
ing it are made for each other. The match is not so perfect in the latter case,
though, since various kinds of illusions are possible. I suggest, then, that we
recognize internal perception as a genuine, albeit certainly eccentric form of
intentional perception which discloses psychological properties of oneself
(see Table 15.1).
18 Even though I agree with Shoemaker that introspection is not a perceptual capacity, I do not want
to commit myself to his functionalist analysis of introspection. The notion of 'ascent routines' inspired
by Evans (1982: 227-8) and developed by Gordon in various essays (see, for instance, his 1995) might
be one of the keys to understanding how (non-perceptual) introspection works.
342 Jerome Dokic
TABLE 15.1. Bodily experience as an eccentric form of percep-
tual awareness
Possibility of Possibility of
(self- )blindness illusion
Despite the impossibility of being self-blind with respect to the states of affairs
presented in bodily experience, the latter shares an essential core with external
perception, so that it is legitimate to picture it as a minimally objective form of
awareness.
To sum up, there is a distinction between two kinds of psychological phenom-
ena. Some of them can be directly perceived, whereas others need interpretation
and inference, or (in one's own case) introspection. Pains and intentional move-
ments (and perhaps emotions) belong to the first category, whereas experiences
and propositional attitudes like beliefs, desires, and intentions belong to the sec-
ond category. I think that the ultimate ground of this distinction is metaphysical.
The ontological furniture of the mind comprises spatially located psychophysical
phenomena and non-spatial psychological events or attitudes.
Note that because we can directly perceive some psychological phenomena, it is
arguable that we can have some unreflective, demonstrative understanding of
them. It may be that the capacity to perceive these phenomena does not require a
sophisticated battery of mental concepts in the form of an articulated theory of
mind. In some cases, what is perceived when I act or when I observe someone else
acting can be conceptualized simply as 'I am/She is doing this'. It is not obvious
that concepts of non-spatial psychological attitudes are mobilized here. My
thought exploits the perceptual presence of a psychological phenomenon, and
is thus dispensed from explicitly identifying the specific attitudes (like intentions,
desires, and beliefs) essentially involved in the action. In contrast, introspection
necessarily deploys concepts of mental attitudes. Introspection is not a perceptual
capacity, so it is impossible to conceptualize one's thoughts in a demonstrative
way. One has to mobilize concepts of propositional attitudes in order to specify
what is being introspected. That is why introspection is necessarily explicitly
reflective.
The Sense of Ownership 343
REFERENCES
ANSCOMBE, G. E. M. (1959), Intention. Oxford: Blackwell.
BERMUDEZ, J.-L., MARCEL, T., and EILAN, N. (eds.) (1995), The Body and the Self
Cambridge, Mass.: MIT Press.
BREWER, B. (1995), 'Bodily awareness and the self', in Bermudez et al. (1995).
BURGE, T. (1993), 'Content Preservation', Philosophical Review, 102: 457-88.
CAMPBELL, J. (1998), 'Le Modele de la schizophrenie de Christopher Frith', in Grivois and
Proust (1998).
--(1999), 'Schizophrenia, the space of reasons and thinking as a motor process', The
Monist, 82: 4.
DELLA SALA, S., MARCHETTI, c., and SPINNLER H. (1994), 'The anarchic hand: a fronto-
mesial sign', in F. Boller and J. Grafman (eds.), Handbook ofNeuropsychology, 9.
DIENES, Z., and PERNER, J. (1999), 'A theory of implicit and explicit knowledge', Behavioral
and Brain Sciences, 22: 735-808.
EVANS, G. (1982), The Varieties ofReference. Oxford: Clarendon Press.
FALVEY, K. (2000), 'Knowledge in intention', Philosophical Studies, 21-44.
GORDON, R. (1995), 'Simulation without introspection or inference from me to you', in
M. Davies and T. Stone, Mental Simulation. Oxford: Blackwell.
GRIVOIS, H., and PROUST, J. (eds.) (1998), Subjectivite et conscience d'agir. Paris: Presses
Universitaires de France.
HORNSBY, J. (1997), Simple Mindedness. Cambridge, Mass.: Harvard University Press.
HUTCHISON, W. D., DAVIS, K. D., LOZANO, A. M., TASKER, R. R., and DOSTROVSKY, J. O.
(1999), 'Pain related neurons in the human cingulate cortex', Nature Neuroscience, 2 (5):
403-5.
JEANNEROD, M. (1994), 'The representing brain: neural correlates of motor intention and
imagery', Behavioral and Brain Sciences, 17: 187-245.
--(1997), The Cognitive Neuroscience ofAction. Oxford: Blackwell.
KINSBOURNE, M. (1995), 'Awareness of one's own body: an attentional theory of its nature,
development, and brain basis', in Bermudez et al. (1995).
McDOWELL, J. (1986), 'Criteria, defeasibility, and knowledge', in J. Dancy (ed.), Perceptual
Knowledge. Oxford: Oxford University Press.
MARTIN, M. G. F. (1995), 'Bodily awareness: a sense of ownership', in Bermudez et al.
(1995).
MERLEAU-PONTY, M. (1962), Phenomenologie de la perception. Paris: Gallimard.
NORMAN, D. A., and SHALLICE, T. (1986), 'Attention to action: willed and automatic
control of behavior. Center for Human Information Processing', in R. J. Davidson,
G. E. Schwartz, and D. Shapiro (eds.), Consciousness and Self-Regulation, iv. New York:
Plenum.
O'SHAUGHNESSY, B. (1980), The Will: A Dual Aspect Theory, 2 vols. Cambridge: Cambridge
University Press.
--(1995), 'Proprioception and the body image', in Bermudez et al. (1995).
PACHERIE, E. (1997), 'Motor images, self-consciousness, and autism', in J. Russell (ed.),
Autism as an Executive Disorder. Oxford: Oxford University Press, 215-55.
--(2000), 'The content of intentions', Mind and Language, 15 (3): 400-32.
PEACOCKE, C. (1992), A Study of Concepts, Cambridge, Mass.: MIT Press.
344 Jerome Dokic
PEACOCKE, C. (1999), Being Known. Oxford: Clarendon Press.
PERNER, J. (1998), 'The meta-intentional nature of executive functions and theory of mind',
in P. Carruthers and J. Boucher (eds.), Language and Thought. Cambridge: Cambridge
University Press.
PERRY, J. (1990), 'Self-Notions', Logos, 11: 17-31.
RIZZOLATTI, G., FADIGA, 1., GALLESE, v., and FOGASSI, 1. (1996), 'Premotor cortex and the
recognition of motor actions', Cognitive Brain Research, 3: 131--41.
ROESSLER, J. (1998), 'Comments on Perner: Intention, Control and Theory of Mind',
Warwick Conference on Action.
RUSSELL, J. (1996), Agency: Its RoLe in MentaL Development. Hove: The Psychological Press.
--(1998), 'Les Racines executives (non modulaires) des perturbations de la mentalisa-
tion dans l'autisme', in Grivois and Proust (1998).
SHOEMAKER, S. (1994), The First-Person Perspective and Other Essays, Cambridge:
Cambridge University Press.
16
The Epistemology of Physical Action
Brian O'Shaughnessy
1 The Will: A Dual Aspect Theory, Cambridge: Cambridge University Press (1980).
346 Brian O'Shaughnessy
either they are one and the same event as the act itself or else one event of 'doing'
of a psychological kind lies buried within an ontologically more complex second
event of 'doing'-even though 'each' doing emerges immediately out of one and
the same token act-progenitor phenomena like intention, decision, desire, and so
on. The theoretical inelegance of this account, together with a series of other con-
siderations, led me to the theory which identifies all physical actions with a trying
or attempting which is of psychological ontological status.
(2) The final element in the dual aspect theory, more exactly the variety of dual
aspect theory advanced by me in The Will, is concerned with the act considered
from a physical point of view. Thus we know that a necessary condition of an act
falling in the so-called 'basic act' mode under a description like 'arm raising', is
that the event of arm rise be linked via act-mechanistic neuro-muscular regular
links with earlier stages of one and the same motor process. This fact led me to the
following claim, which I advanced at several places in the book. 2 Namely, that the
event of acting and/or willing is one and the same as the full activation of
the motor-system. Now it should be emphasized that the claim is not that the
event of willing activates the motor-system. Rather, it is that the immediate aet-
progenitors, act-desire and act-intention, activate that system, and that this activation-
syndrome falls simultaneously under the concepts of 'try', 'strive', 'do', 'will', and
finally by the end of the motor-process also under 'arm raising' or 'walking' or
'talking' as the case may be. The claim I made was therefore that no matter how far
the activation of the motor-syndrome had progressed, provided it was immedi-
ately caused by act-desire and intention it already fell under the psychological
concepts of 'willing' or 'striving' or 'doing'-which is to say that something (or
another) had already been done by whatever moment one selected, even if the
motor-process ground to a halt in the very first stages.
One fills in the rest of the theory in the following way. Each enlarging sector of
the motor-process, which falls already under 'try', 'strive', 'do', and 'will', proceeds
to cause the next stage of the process, and finally to cause the (so to say) crown-
ing event of (say) arm movement. Then although whatever slice of the process one
likes to select falls already under the concept of ,willing', provided the slice in ques-
tion stretches back to the beginning of that process, and although it proceeds to
cause the next stage and ultimately arm movement, this is not to say that the event
of willing causes arm movement. Rather, the theory claims that the event of will-
ing physically develops, in naturally appointed causal manner, to the point at
which it incorporates the event of limb movement, and completes itself in so
doing. A natural, ubiquituous, and perfectly normal complex physical phenom-
enon is at that moment realized, which falls at the same time under a set of active
concepts like 'try', 'move arm', 'ring bell', and so on. Then to repeat: since willing is
2 See The Will: A Dual Aspect Theory, ii, chs. 11-13 and ch. 16, 286. Cambridge: Cambridge
University Press (1980).
The Epistemology ofPhysical Action 347
identical with arm moving, and arm moving actually incorporates arm move-
ment, we cannot say that willing causes arm movement. We cannot say it, even
though arm movement owes its existence through a sequence of causal relations
to the phenomenon of willing. The developmentalist character of the theory,
together with the criteria for event-identity, disallow such an interpretation.
(3) I shall not here attempt to defend this theory, even though I am convinced
that there exists strong argumentative support for it, of the kind elaborated in
chapters 11 to 13 of The Will. I am concerned rather with the epistemological
implications of the theory. We shall see that they have a bearing upon the
mind-body problem. Now I have described the theory as 'dual aspect' in type, but
it seems to me important that we understand the sense in which the expression is
applied in this context. It is, I think, essential that we distinguish it from a sense
that is primarily ontological. Thus, certain non-reductionist versions of
Physicalism could perfectly well be described as 'dual aspect' theories. Some
philosophers would say of certain cerebral phenomena that they fall at one and
the same time under psychological concept like (say) 'seeing' and under physical
cerebral concepts. This theory is ontologically dual aspect in one sense of the
term. And it has epistemological implications that confirm this description, which
derive from the primary ontological claim. For example, if one could manage to
identify the psychological phenomenon in cerebral terms, the subject of (say) a
visual experience would in principle be able to see through mirrors an event
which he immediately experienced as a visual experience with such and such a
content. He would be able to see an event that was the seeing of itself. Physicalism
in one of its forms is therefore ontologically and by implication also epistemolo-
gically dual aspect in one sense of that expression.
The dual aspect theory of bodily action which I set out earlier does in fact imply
an epistemological claim of exactly the same character. But the theory differs in
two interesting respects. For one thing, it arrives at such a position without
depending on Physicalist Theory. It so to say constructs its own local Physicalism
for bodily actions, on the basis of arguments which have no application elsewhere
in the mind. Thus, it involves in the particular all that the Physicalist position
claims in general terms. The dual aspect theory after all asserts, amongst other
things, that bodily actions are psychological events and also physical events. And
if this is true, it must in principle be possible for an active subject to see or feel an
event which he immediately experiences as action, just as it is on the Physicalist
account. In each case an epistemologically dual aspect thesis follows on the heels
of an ontological thesis.
However, it is not merely on the grounds of justification that a dual aspect the-
ory of physical action parts company with Physicalism. It differs also significantly
in its epistemological content, going well beyond the above epistemological claim.
Indeed, it is specifically this property which justifies the application of the epithet
'dual aspect' in the special sense reserved for physical actions. What I have in mind
348 Brian O'Shaughnessy
is the fact that according to this theory the physical character of a particular
psychological event, to be exact uniquely the event of physical action, is given
a priori in physical terms. The dual aspect theory of physical action is constitutively
specific in physical terms as Physicalism is not. The epistemological part of the the-
ory asserts that there exist specific a priori-given physical directives for identifying
a psychological event in physical terms. We merely single out the entire process of
mechanism-activation that culminates in the event of limb movement, the theory
being that in doing so we are picking out a psychological event of type 'willing'.
Dual aspect action theory is the only doctrine I can think of which on purely philo-
sophical grounds offers a specific recipe for identifying a psychological event in
physical terms. Not only does Physicalism not do so, it is a theory which is perfectly
consistent with all sorts of quasi-Cartesian interiorist analyses of physical action in
which the act is identified at one and the same time with an interior event and with
unspecified purely cerebral phenomena-as well as with the above theory.
(4) Now so far I have spoken mainly of the physical content of dual aspect the-
ory, and thus by implication of third-person epistemology in the case of bodily
actions. Naturally such forms of epistemological access are available also to the
agent, relating to himself so to say in the third-person mode. In whatever way others
can see and hear and even tactilely feel the movements of his limbs, so too can the
agent himself. However, one mode of sense-perceptual access is reserved for the
agent alone, namely the proprioceptive mode. And it is clear that this mode of
awareness must be of central importance so far as physical action is concerned,
indeed in a general sense must be essential to the very phenomenon of physical
action. True, the use of a mirror in shaving demonstrates that the epistemological
feedback in physical action can be distributed amongst the senses, and it may even
be that a particular bodily act might occur in the absence of proprioceptive aware-
ness of the body. However, it seems certain they could not in general do so.
Proprioception is in general essential, and occupies a unique position in physical
act situations. After all, if the mind is to be able to physically act, it needs nothing
more than proprioceptive awareness of the presence and position of a limb over
which the bodily will has sway. We cannot say the same of sight or any other sense.
This asymmetry is irreducible.
Now if the dual aspect theory of action is correct, or indeed if various other
analyses of physical action happened instead to be correct, the act should be given
to the agent in sense-perceptual mode. And in fact it is a near truism that it is so
given. There can be little doubt that the sense-perceptual experience of one's own
actions is of vital importance to an agent as he acts. One has merely to consider
what would be the effect upon one's physically active life of a complete proprio-
ceptive blackout: for most creatures in most situations it would lead to a quasi-
paralysis. While sense-perceptual experience of one's actions does not inform one
that one's will has been active, it plays an essential role in enabling action. It gives
The Epistemology ofPhysical Action 349
to the bodily will its target-object, functions generally as a corroborative causal
condition of one's knowing that one's endeavours are so far successful, and
informs one of what next to do to complete one's active project.
Now despite these valuable and indeed essential epistemological contributions
on the part of sense-perceptual experience of one's limbs to the successful
performance of physical action, such a mode of awareness fails to deliver the full
measure of information that is necessary for successful intentional physical action.
For one thing, sense-perceptual experience can do no more than perceive the outer
segment of the act, it cannot perceive the full extent of the deed, in particular has
no awareness of its inner physical boundaries, and for these reasons cannot be
perceptually individuative of its perceptible event-object. But secondly, sense-
perceptual experience of physical action is unable to detect the sheer occurrence of
activeness of character. What it does perceive, namely movement in part of the
body where the will has sway, is evidence that action is going on. But no more than
that. It is in the first place not proof, and secondly it is neither needed nor used by
the agent to know that his will is active. While it is at all times for the agent an infer-
ence that bodily action is going on, which is to say that successful action is occur-
ring, it is no inference that willing is occurring, that he is trying or attempting and
so forth, which is of a wholly different order from the above and immediate.
4. EXPLANATORIAL ISSUES
Lucy O'Brien
1. INTRODUCTION
This chapter was mostly written during leave taken under the Research Leave Scheme of the Arts and
Humanities Research Board. I am very grateful to them. I have benefited from very useful comments
from participants of meetings in Fribourg, Girona, Manchester, and Warwick (the latter in a seminar
arranged as part of the AHRB Project in Consciousness and Self-Consciousness). Special thanks to the
participants in a graduate research seminar on self-knowledge at VCL; and for comments, to Alan
Brown, David Levy, Paul Noordhof, Mark Sacks, and Barry Smith and an anonymous referee for this
volume. Particular thanks are due to the editors for their comments, interest, and patience.
1 There was a flurry of papers on the subject in the early 1960s. See O'Shaughnessy (1963);
Anscombe (1963); Donnellan (1963); Broadie (1967); Olsen (1969); Danto (1963). Since then there
has been Velleman's important discussion in his Practical Reflection (1989). See also Dunn (1998) on
this. It was only after writing this chapter that I came across Johannes Roessler's valuable contribution
to this issue. See Chapter 18, this volume.
2 See Cassam (1994), MacDonald et al. (1998), and Shoemaker (1996) as key sources for the work
3 I argue elsewhere that we should take bodily actions as prime and as not amenable to reductive
analysis in terms of intentions and bodily movements. Although no reductive analysis of actions is
available, a more modest account of actions is.
360 Lucy O'Brien
does not seem to be the case that all our actions could as a matter of brute fact
be beyond our ken.
Let us go through these features in a bit more detail:
1. Authoritive. I am going to take first-person authority with respect to a given
subject area to imply that there is the possibility of an epistemic first/third-person
asymmetry with respect to that area. Here the asymmetry must be taken to lie not
simply in the fact that the subject can in central cases know more or better than
others, with respect to that area, because they are around a lot or are more inter-
ested. Rather, the subject can know more or better because they know in a way
which is in principle unavailable to others. So let us say:
X is first-person authoritative with respect to a fact, p, about X iff in central cases
of X's judgements concerning p, we can say that X is in a better position than
others to know p, because X knows p in a way in principle unavailable to others.
Given this understanding of first-person authority it is, I think, plausible to claim
that we are authoritative with respect to our actions. This authority is exemplified
by the fact that when a subject acts, and so moves their body, we take the subject
to be authoritative relative to others, about whether the subject acted in so mov-
ing or not. Consider the case where I intentionally raise my arm. Given that there
is an action of me raising my arm, it will almost always be the case that I am able
to know that I raised it, and know in a way that others do not. Others will look to
me to know whether I raised my arm voluntarily. If I tell them that it was a vol-
untary act on my part, and not an involuntary, unwilled, movement they will
presume me to be right. Of course, it is possible that I am deluding myself. It is also
possible that I fail to be sincere and have reasons for wanting the person to think
that, contrary to the facts, I acted voluntarily. However, for the main, and in the
normal case, it will be presumed that I am in a better position to say when I am act-
ing, and when I am not, than they are. We might not take a subject as authoritative
about her action described as a movement of her body. However, as long as we keep
clear the distinction between authority with respect to movements of the body
(arm risings), and authority with respect to actions (arm raisings), then it is clear
that there is a first/third-person asymmetry with respect to the latter. 4
Let me emphasize that, on this understanding, first-person authority with
respect to our own actions does not mean that it is not possible for us to be wrong
about whether we acted, and for another to be in a better position than me to
know whether I acted.
We are not infallible with regard to our actions and can clearly think that we acted
while having failed-due to some motor failure, say-to do so. On an occasion
5 Physical actions are not the only psychological phenomena we describe in terms of their con-
sequences. We describe mental actions in terms of their consequences-we say things like 'Your decision
not to go to the party was a decision not to meet NN' even when the subject being addressed does not
know NN or that NN was to be at the party. Note that in a not dissimilar, but in a much more restricted
way, we ascribe people beliefs in terms of their implications or presuppositions. ('Your belief
that women are foolish is a belief that your own daughter is foolish', 'Your belief that water is wet is a
belief that H 20 is wet', for example.) Perhaps the way we very often re-describe the objects of percep-
tion, beyond any capacity the viewer has to recognize the objects as falling under those descriptions,
comes closest to the action case. ('He was looking at a genuine Goya, priced at £10, and did
not buy it.')
362 Lucy O'Brien
I did when what I did comes under the description of unintended consequences.
However, it does not follow from this that I did not have any authority about what
I did relative to descriptions which are more basic. If there are descriptions of
actions which are somehow basic, we can make our claim a claim about our
authority over our basic actions.
One suggestion might be that if we take intentions and/or tryings as necessary
for action, then there will always be a description of my actions in terms of what
I intended or tried to do, and that description may be thought basic, relative to
other descriptions. And, it may be said, I do seem to be authoritative about what
I have done, when what I have done is described under the description drawn
from what I intended or tried to do. However, things are more complex than this
suggests. Our intentions and our tryings can also appear to fall under competing
descriptions. So we can say: 'You think you are intending/trying to ring the door
bell of no. 6, but you are not, that's no. 4's bell you are intending/trying to ring'.6
Perhaps we should think of these as de re ascriptions the truth of which, like de re
belief ascriptions, often transcend our capacity to accept them as true. The claim
could then be that the statement of authority should be taken to be utilizing de
dicta ascriptions.
I do not think that this is the right way to go. First, it is far from obvious that
intentions or tryings, as causal precursors, are necessary for actions. It is possible,
and in my view plausible, to hold that that there can be deliberate actions which
are not preceded by any intention to act. One might think that intentions should
be understood as effective ways of storing conclusions of practical reasoning for
the future, which are not needed in cases where an action itself is the conclusion
of an exercise of practical reason. It is also possible to deny that we try to act
whenever we act, even when we succeed in acting. One might think that tryings to
act are kinds of degenerate or failed actions. Settling these questions falls beyond
anything that can be accomplished here, but their openness raises a concern with
the suggestion mooted.
Second, it is implausible to say that I am in general authoritative with respect
to my actions when those actions are described in terms of what I de dicta
intended or tried to do. Consider a subject intending or trying to get a ball in the
corner pocket in a game of snooker. Suppose that the subject acknowledges that
this is the right way to describe what they are intending or trying to do. Imagine
they strike the ball and the ball rolls into the corner pocket. While we do want to
say that the subject is authoritative about whether they acted, it seems implausible
to claim that the subject is authoritative, in contrast to others, about whether they
are getting the ball in the corner pocket.
The trouble with descriptions in terms of what a subject is intending or trying
to do is that they seem to avert to the subject's main purpose or motive in doing
7 Note that we have here a close parallel with our authority over our perceptions. I know author-
itatively that I am seeing, and I know what I am seeing, but my authority over the latter seems only to
extend to descriptions in terms of basic observational concepts.
364 Lucy O'Brien
a grasp of the possible basic ways that we might act, and given that we choose to
act in one way rather than another, relative to that grasp, then we seem have the
materials to account for a way of knowing what we have done that could not be
available to anyone other than the agent, and which does not call on our perceptual
faculties to provide immediate grounds. So, if this picture is right, the suggestion
that a subject who acts is authoritative relative to others about their actions-both
in terms of what they have done and of whether they have acted-becomes com-
pelling. Further, if this picture is right, we have some explanation for the second
feature identified above of our knowledge of our actions.
2. Relatively a priori. Perhaps the most notable and problematic feature of
our knowledge of our own actions is that it appears to be immediately available
to the subject who is engaged in acting, and who is aiming to answer the question
as to whether they are acting. Such knowledge does not seem to have to await the
testimony of our perceptual faculties. 8 We do not seem to need to feel our mus-
cles clenching and our arn1 rising, or to see the trajectory of our hand, in order to
know what we are doing. Perceptual knowledge is, without doubt, required as part
of the background that makes action possible. However, it seems that in order to
know what we are doing in the case of an individual action we do not need to per-
ceive simultaneously, either via bodily awareness or our other five senses, what we
are doing. 9 In this way, our knowledge of our action is relatively a priori: that is,
it is not acquired, or justified, via perceptual evidence in a given case, although it
does rest on the obtaining of background conditions which we are entitled to take
as met, and which themselves may garner support from perceptual information.
3. Relatively transparent. It is not just that I do not seem to have to perceive
that I am acting in order to know that I am acting, nor just that I am authoritative
when I do judge that I am acting. Also the subject's actions, like other psycholo-
gical phenomena, seen1 to have a certain conditional epistemic availability to the
judging subject. We can obviously fail to know that we are acting, as when we are
acting absent-mindedly or are repressing what we are doing. But it does not seem
to be the case that our actions can be, as a matter of brute fact, beyond our ken. It
is, I think, very hard for us to imagine an agent who is capable of askingthern-
selves the question <What am I doing?' not being able normally to answer the ques-
tion correctly. This would be to imagine an agent capable of reflexive thought,
voluntarily carrying out one action rather than another, and yet not knowing that
8 Pace AnscoITlbe I do not think we need to think of knowledge of our bodies through bodily aware-
ness as non-perceptual. I therefore mean to include bodily awareness among those sources not directly
necessary for knowledge of what we are doing.
9 Experimental psychological data tends to confirm this in so far as it finds that subjects judge
themselves to be acting before their bodies can be observed to be moving. If they had to observe the
movements of their bodies in order to ground their judgements, we would expect their judgements to
come much later.
On Knowing One's Own Actions 365
they are acting. It seems to me, however, that we cannot, in Shoemaker's phrase,
envisage a creature who is simply self-blind with respect to all their actions in this
way. There seems rather to be a necessary and conceptual connection between a
subject acting and its knowing what it is doing. lO
There are two different kinds of action, which perhaps ought to be separated
here. Many have argued that we need to distinguish non-intentional actions-
actions done for no reason-from intentional actions. If there is such a distinc-
tion then we might imagine that while our intentional actions are likely to prove
accessible to us in the way characterized, our non-intentional ones may not. l l Let
us count absent-minded finger tappings as non-intentional actions of mine. Am I
epistemologically disassociated from such actions to a degree that makes the claim
that I could be totally self-blind with respect to them look plausible? It is clearly
true that I can be tapping my fingers without noticing. However, to the extent that
it is plausible that there is genuine agency in such cases, by which I mean that I
can be said to be controlling the action, I must normally be able to come to know
what I am doing. (And note that it is far from clear that it is plausible that we have
genuine agency in such cases, rather than just that my body is caused to move by
a purposive system I am not responsible for.) Consider a case where a subject is
acting-counting using the fingers of her right hand, say-and then acts in a way
incompatible with carrying on doing the former-picking up her cup of tea with
her right hand, say. If we want to say that the subject herself is genuinely control-
ling both actions, rather than that some subpersonal system is, we want to explain
how the second action relates to the former in the case where the subject is sup-
posed to have no possibility of accessing the former. If there were a general disas-
sociation for the subject between the perspective from which the question 'What
am I doing?' gets answered, and the perspective from which she carries out her
actions, we would have the possibility of a subject deciding to pick up her cup of
tea wondering if her action could bring to an end or disrupt, or indeed be dis-
rupted by, other actions she was unaware of carrying out. But turning the subject's
actions into possible external impositions in this way seems to get things quite
wrong. If what I am doing can be said to be controlled by me, I must at least have
the power to initiate it or to will it to cease when I have reason to do so. The con-
trol and regulation of my actions as the actions of a unified agent seem to require
10 This connection is noted by Olsen (1969) and Broadie (1967). Note that someone who denied
that our knowledge of our actions was relatively a priori might still accept that they must be relatively
transparent. They might hold that it is a condition of something's being an action that it be generally
known to the agent, but hold that our actions are known to us on the basis of perceptual information.
They would then have to say that when the subject is blind to their movements, their movements can-
not be actions.
11 See e.g. O'Shaughnessy (1980, ii, ch. 10) and Ginet (1990: 3). By a non-intentional or subinten-
tional action is meant not just an action which is not intended under one description but an action
that is intentional under no description.
366 Lucy O'Brien
this. And surely if I have the power to initiate or to stop what I am doing, then
what I am doing must normally be in some way accessible to me. Thus for an
action to be within a subject's control, the subject must be capable of knowing
what they are doing. And given the ability to ask the question 'What am I doing?'
the subject's awareness of what they are doing must normally feed into an answer.
With these features of our knowledge of our action identified, I want now to
explore some possible accounts we might give of our knowledge of actions and to
sketch out the approach I think most promising. Some of the accounts explored
do not sit well with the intuitively plausible features identified, and some have
other shortcomings. However, nothing I say here about the accounts I set aside in
favour of the one I think most promising will be sufficient to show that they could
not be made good, or could not explain, or explain away, the features identified
above. My hope, nevertheless, is that enough critical work will have been done to
motivate a consideration of the account I think we should go for.
3. POSSIBLE SOLUTIONS
Certain theorists about belief have attempted to solve the problem of how it is that
we have relatively a priori knowledge of our beliefs, despite their relational char-
acter, by claiming that what we took to be a unified phenomenon is in fact a dual
component oneY Thus our knowledge of our beliefs on this account divides into
two parts: knowledge of a narrow component, of which, as subjects, we have a dis-
tinctive kind of knowledge-knowledge by introspection, and knowledge of an
external component not knowable by the subject in any distinctive way. The dual
component theorist tries to explain how it is that our knowledge of our beliefs has
the features it has, despite being relational, by adopting a dual component theory
of belief and an account of our knowledge of each element. If this is a natural
move to make in dealing with our knowledge of our beliefs, it is an even more
natural move to make in dealing with our knowledge of bodily action. It has been
common to think of actions as the combination of at least two separate components:
a psychological component-an intention, or trying, and a non-psychological
one-a movement of the body. If this were to be our view of actions then we could
say that the features that we have taken to be features of our knowledge of
our actions are in fact only features of our knowledge of our tryings or intentions.
Our knowledge of our intentions is authoritative, a priori, and transparent,
but our knowledge of our bodily movements is not. As Donnellan puts it: 'What
this suggests is that our knowledge of our own intentional actions is complex, that
12 Putnam's original paper (1975) suggested this move. It is also developed in McGinn (1982).
On Knowing One's Own Actions 367
it divides up, so to speak, into an element of "direct awareness" ... and other
elements to which observation is relevant'.13
To try to explain the knowledge we have of our actions by claiming that it is in
fact knowledge of two more primitive components is not, I think, the right direc-
tion to move in.
First, such an account makes a critical part of my knowing what I am doing a
matter of perception. Given that we have perceptual access to others' bodily move-
ments, we might think that it fails properly to capture the first/third-person-third-
person asymmetry involved in the first-person authority we have found to be a
feature of our knowledge of our actions. Also, such a dual component account
clashes with the seeming aprioricity of our knowledge of our actions. The account
has it that my knowledge that, for example, I am raising my arm depends not only
upon my knowing immediately that I intended or am trying to raise my arm, but
also upon perceptual information that my arm is rising. In so doing, it fails to
explain a feature of our knowledge of our actions we found plausible. Moreover, it
threatens to conflict with the feature of relative transparency. If my knowledge of
my actions depends directly upon a perceptual component in this way, then given
the possibility for widespread brute error that perceptual capacities leave room for,
we should expect to find such a possibility in the case of action. It may be that the
dual component theorist can explain why we must normally be able to know what
we are doing despite adopting a partially perceptual model. They may argue that
since our normal conditions for initiating action require information about the
nature of our bodies, via perceptual feedback, perceptual breakdown would in fact
rob me of my capacity to act, by robbing me of a grasp of what possible things I
could do. If this were the case, there would not be the possibility of actions to which
I was self-blind in the case of perceptual breakdown. However, even given percep-
tual breakdown, we may think that a subject would in fact be able to continue act-
ing for a short time, relying on past information for a grasp of the possible actions
it could carry out. And it is not plausible that they would be unknowing about such
actions simply in virtue of that breakdown. The theorist might also have some
other reason for saying that there is a conceptual dependence between our acting
and our knowing that we are acting, so that there would be no actions in the case
of perceptual breakdown. Prima facie, however, it is hard to see what their reason
might be. But the reason cannot be the attractive one that there is something about
the conditions under which we act which is already apt to provide us with know-
ledge of the action. This cannot be their reason because, on their account, to know
what I am doing I have to wait to receive perceptual information which is not
available at the onset of the action about what my body is doing.
Second, the dual component model seems to be partly motivated by
the assumption that an account of the knowledge we have of our intentions or
13 Donnellan (1963).
368 Lucy O'Brien
14 Of course we can also come to have knowledge of the movements of our bodies from observing
and listening to others-someone starting can tell me that I have touched them, or someone can
verbally inform me that I have moved-but such sources are clearly not operative in the basic case.
On Knowing One's Own Actions 369
awareness. It has not been doubted that this knowledge is set against a background
which makes it unnecessary for me to check to see whether my motor system is in
working order. Or that having reason to think that my motor system is playing up
is very likely to result in an appeal to the testimony of my senses to check whether
we acted. Nor is it doubted that keeping up an accurate schema of what actions are
possible, and so being able to continue acting at all, will require perceptual feed-
back on the position of my limbs and so on. However, none of this gives us reason
to think that any such feedback is part of my immediate grounds for my know-
ledge that I am raising my arm.
We can imagine our dual component theorist, in response to the second prob-
lem raised above, rejecting the assumption that our knowledge of our intentions
or tryings will involve direct awareness of some component, narrowly construed.
What account the dual component theorist comes to offer of our knowledge of
our intentions or tryings will depend on what account she thinks right for
mental phenomena in general. So, instead of taking our knowledge of our mental
states and actions to be due to some kind of internal perception, she might think
it due to some reliable mechanism. Or she might think that our intentions or
tryings can function as reasons for their own ascription. Whatever account is
adopted (and it may be an account that embraces an externalist individuation of
intentions and tryings), it may then be combined with an account of our know-
ledge of our bodily movements, and perhaps our knowledge of the relation
between our bodily movements and our intentions or tryings, to give an account
of our actions. However, it needs to be noted that once the theorist has embraced
the idea that intentions or tryings depend upon external elements, much of the
motivation for giving an account of our knowledge of our actions by thinking of
actions as divided into two components falls away.
However, recalling the third problem, even given adaptations of the kind dis-
cussed, if one operates with an understanding of actions as unified phenomena,
then, however compelling in themselves the component parts of the account
offered are-the account of our knowledge of our intentions, of our bodily move-
ments, and so on-the resulting account of action is unlikely to be accepted.
So far we have supposed that our dual component theorist is committed to a
dual component theory of action and, thus, to a dual component account of our
knowledge of our actions. Might it not be possible for a theorist to meet the third
problem by agreeing that our actions are unified psychological phenomena, but
holding that our knowledge of them nevertheless comes in two parts? Here the
suggestion would be that I have knowledge of my intentions or tryings, however
such knowledge is construed, and also knowledge of my bodily movements,
through bodily awareness and/or ordinary perception, and that these two kinds of
knowledge are what enable me to know what I am doing. On this account, my
knowledge of what I am doing will be essentially indirect. According to this ver-
sion of the dual component account, intending or trying is at most a necessary
370 Lucy O'Brien
condition of, but not a component part of, an action. So, on this account,
my knowledge of my intending or trying is knowledge only of a necessary
condition-perhaps of a causal antecedent of my action-not of my acting. And
any observation of my body moving will not in itself give me knowledge of what
is in fact my action. If we held a unified account of the nature of bodily action, but
adopted a dual component account of our knowledge of such actions, we would
in effect adopt an account on which our knowledge was always inferred from the
knowledge that necessary conditions for my acting were satisfied. Given the con-
tinued presence of a perceptual condition, such an account would again conflict
with the relatively a priori nature of our knowledge of our actions claimed above,
and threaten to conflict with the transparency claim. However, identifying this
possibility does raise the question of whether our knowledge of our actions is
some kind of inference from knowledge of necessary conditions, combined with
background conditions.
15 If my trying to do something is not supposed identical with, or part of, my acting when success-
ful, then in order to preserve the assumption that my acting is my doing something we would have to
assume that trying is not all I have to do to act. But if we assume this, then it seems that the gap
between tryings and actions will be much the same as that between intentions and actions.
16 I take this description from Peacocke (1998, 1999).
372 Lucy O'Brien
the relation between the phenomenon and the knowing of it is a rational relation,
either because it is partly constitutive of the relata, or because it is taken to have
some wider rational significance. Shoemaker argues that it is not just a matter of
fact that belief ascriptions are authoritative, and beliefs relatively transparent, it is
rather a conceptual, and so necessary, truth that they are. It is such a conceptual
truth because it is partly constitutive of what it is to be a belief that a creature with
the belief and capable of self-ascription, will normally be able to self-ascribe it.
Burge argues that what makes the connection between our beliefs and our self-
ascriptions capable of sustaining knowledge is the role such a connection plays in
our nature as rational beings. He argues that our nature as critically rational sub-
jects depends upon the obtaining of such a relation and takes this fact to confer a
kind of external warrant for the self-ascriptionY
Similar embellishments could clearly be proposed for our account of our know-
ledge of our actions. We might take actions to be self-intimating as a matter of
conceptual fact. So movements of our bodies that we were normally unable to self-
ascribe would not count as actions. We might also claim that an ability to be author-
itative about its own actions is a prerequisite for .a critical agent, and take this to be
sufficient to confer the status of knowledge on to the agent's self-ascriptions.
Whilst both suggestions have considerable plausibility about our natures as
agents, the no-reasons theorist faces the same question in this case as he faces in
the case of belief and that is: how do these truths secure knowledge for the
subject when they might seem to suppose it?
With respect to the first suggestion, we want to say that, while it may be true that
certain processes or activities only count as actions if they are epistemically acces-
sible to the subject in a certain way, this does not mitigate the need for an independ-
ent account of the way in which they are episternically so accessible. With respect to
the second, it is hard to suppose that the self-ascription counts as knowledge merely
in virtue of the fact that it is required to be knowledge for the self-ascribing subject
to function as a critical agent. The essential role played by the self-ascription within
the critical agent may confer some warrant on it. However, we would expect some
episternic relation to hold between the action and its self-ascription independently of
that role, and expect it to explain, at least partly, why the self-ascription counts as
knowledgeable. That explanatory task still seems pressing. IS
20 Evans (1982: 225): (1 get myself in a position to answer the question whether I believe that p by
putting into operation whatever procedures I have for answering the questions whether p.'
21 An account such as this can easily give way to a kind of psychological instrumentalism-rather
than taking it that there really are beliefs constituted by the practices of self-ascription, we might think
that there are only the practices of self-ascription themselves, and that the beliefs ascribed are useful
interpretational fictions.
On Knowing One's Own Actions 375
that our self-ascriptions are epistemically groundless and lacking in justification.
Nevertheless, the justification a subject has for their belief that they are doing IjJ
cannot lie in the fact of their doing 1jJ. Rather, it must lie in the fact that the belief
is held in circumstances in which, given the subject's desires for self-knowledge, the
belief will ensure that they are motivated to do 1jJ. While it seems right to say that
such nesting of a belief can provide evidential support for the belief, it is hard to
accept that this is the way epistemic support figures in the standard case.
Another important cause for concern is that this account works only by assum-
ing that intentions, understood as beliefs about what we are going to do, are neces-
sary for action. That makes the account unusable to someone who thinks we need
form no such belief about our future action as a result of practical reason. If we hold
the position mooted before, that we can sometimes just act as the conclusion of a
process of deciding how to act, then we are not going to want to hold that a belief
about what we are going to do is necessary for our knowledge of that action.
What an account such as that offered by Velleman does seem to get right, how-
ever, is that in cases of psychological self-knowledge, the knowledge we have of the
phenomenon is somehow simultaneously given with its occurrence, rather than
merely occasioned by it. As such, it seems to respect the features of our knowledge
of our action intuitively identified. If we cannot reconcile this fact with the naive
thought that we know what we are doing partly because we are doing it, then we
may be forced to adopt such an account.
22 The suggestion is, of course, drawn from Peacocke's (1999) account of our knowledge of our
beliefs.
376 Lucy O'Brien
Well, we may say at this point that it is not simply Smith's raising his arm that
can function as the reason for Smith's self-ascription of the fact that he is raising
his arm. Rather, it is centrally those actions Smith is consciously engaged in which
stand to function as the reasons for their own ascription. The crucial question
then becomes: how should we understand the sense in which the action is one
that Smith is consciously engaged in? We might try two ways: (i) raising an arm,
say, might seem to give us consciousness of the arm. This does not seem to help.
(ii) We might take a conscious action to be an action that we make a judgement
about. But, then we have gone a full circle in our explanations.
Let us pick up a suggestion of Peacocke's about how to characterize what is
involved in the consciousness of a conscious attitude. The key suggestion made by
Peacocke is that a state is conscious, in the relevant sense, if it occupies our atten-
tion. He tells us that to understand properly what is meant by a state occupying
our attention, it is important that we distinguish it from the case of a state being
the object of our attention.
So, how are we to understand what it is for a state to occupy, without being the
object of, our attention? Peacocke's discussion at this point does not offer a fur-
ther explication. He identifies a number of distinct mental phenomena that can
be said to occupy without being objects of attention-perceptions, conscious atti-
tudes, tryings, and actions-but does not attempt any further account.
However, we clearly need more of an account, and such an account may prove
to be easier to come by in the case of actions than in the case of Peacocke's con-
scious attitudes.23 The things that we do strike us immediately as the paradigm of
things that occupy, without being the objects of, our attention. They are the para-
digms of the things known by engagement or participation, rather than because
they function as the objects of some scrutiny.
How then might we understand what it is for an action of mine to occupy my
attention? That is, how should we understand what is going on when we say that
an action is conscious in virtue of the subject's engagement with it? I am going to
assume that dynamic entities such as actions and events have properties, and that
an action's being conscious is a matter of the action having a certain property.
I am also going to assume that we can make a distinction between the relational
and intrinsic properties of an action or event. So, to take an unproblematically phys-
ical example, we would, I think, want to say of the rolling of the pink ball into the
pocket of the table, that its 'being a fast rolling' is a relatively intrinsic property of
the event. On the other hand, its 'being a rolling which occurred after the sipping
of the beer' would seem clearly to be a relational property of the event. It is of
course hard to state any principled difference between relational and intrinsic
properties (especially since it is likely to be a partly contextual matter whether
23 I think that what Peacocke calls a conscious belief is in fact a kind of mental action; and that the
account required of what enables our beliefs to function as reason for their own ascriptions will be
parasitic on an account of what enables our judgements to function as the reason for their own ascriptions.
On Knowing One's Own Actions 377
a property is considered intrinsic or not). But roughly, in order to fix the intrinsic
properties of an event, say, we need consider only the event itself along with standing
background conditions, and can bracket other independent events that may be
occurring. What this suggests is that, for our purposes, we can think of a property
as an intrinsic property of an action if it is a genuine monadic property of the
action (even if requiring relational individuation). And we can think of a property
as a relational property of an action if it stands to be analysed as a dyadic property,
that is, as a relation.
Keeping in mind this distinction between the intrinsic properties and the
relational properties of a dynamic entity such as an action, and the assumption
that being conscious is a property of an action in virtue of which it occupies our
attention, we can ask whether such consciousness is to be understood as a
relational or as an intrinsic property of the action. What answer we give to the
question depends upon the model we have for what it takes for an action to
occupy our attention. Let me consider two possible models.
We might think that when we engage in an action in such a way that the action
occupies our attention, what we have is a kind of complex action involving two
more basic actions. First, we have the action of doing l/J and second, we have the
action of attending to doing l/J. So our doing l/J in a way that makes it occupy our
attention is a matter of its having the relational property of being part of a com-
plex action which involves not just a doing of l/J but also an action of attending to
a doing of l/J. Here doing something consciously is a matter of its standing in a cer-
tain relation to something else I do. To adopt this as an account of what is involved
leaves us, of course, requiring an account of what is involved in the subject attend-
ing to doing l/J where l/J occupies attention rather than being the object of it.
In contrast, we might think that when we do something in a way which occu-
pies our attention what we have is a single action done in a certain way. Rather
than there being two actions, the doing of l/J and the attending to the doing of l/J
there is rather just the one action: the doing of l/J in a certain way. Here the sug-
gestion is that for an action to be conscious, in the sense that it can stand as the
reason for its own ascription, is for it to have an intrinsic property.
It is very natural to assume that for a mental state or activity to be conscious in
a way that makes it accessible to its subject-for it to be conscious in a way that
makes it poised to stand as the reason for its own ascription-is for it to have a
certain relational property, for it to bear a relation to some further act on the part
of the subject. In particular, it may be thought that it can only have such a prop-
erty in virtue of some further act of attending to their actions, for example, on the
part of the subject. This very natural thought is, in my view, a mistake-at least as
far as our activities, as opposed to our mental states, gO.24
24 It is in fact my view that passive mental states such as beliefs and perceptions are only conscious
in this sense (as opposed to being conscious in the sense of making the subject conscious of objects in
the world) because they bear certain relations to non-passive states such as judgings and perceivings.
378 Lucy O'Brien
Instead, I want to suggest, we should take the kind of consciousness exhibited
by mental activities, which occupy without being objects of attention and are able
to stand as reasons for their own ascription, as intrinsic to those activities and not
requiring any further mental act on the part of the subject. The kind of con-
sciousness inherent in my actions should be thought of as a nl0nadic property of
the action rather than a dyadic one.
However, with only this much in place, we are still left with two questions:
What way of acting makes an action conscious, and so occupy rather than stand
as the object of our attention?
Why should an action that occupies our attention qualify for knowledgeable
seif=-ascription? What grounds my taking the action I am consciously carrying
out to be my action?
What seems clear is that an adequate answer to the first question should deliver
an answer to the second question. If an action's being conscious makes it fit to
stand as the reason for its own self-ascription, then what it is for the action to be
conscious should make it clear why it does ground such self-ascriptions. In
essence, my suggestion will be that I act consciously when I engage in my action
as something I control. Further, that engaging in an action as something I control
is engaging with the action as my action, and so involves a primitive fornl of self-
awareness. Because of this my conscious actions are apt inlmediately to ground
self-ascriptions.
Actions strike us as the paradigms of mental phenomena that can occupy or
engage our attention without being an object of it. Quite intuitively, our actions
are those things we know, not by observing them, or by reflecting about them, but
rather by actively engaging in them. We, of course, carry out many actions as a
IIlatter of habit, or relatively automatically and unattentively, but when we act
consciously we seem to act with a sense of guiding our action, with a sense of
control. A natural suggestion as to what distinguishes an action carried out in a
way that occupies our attention, from an action carried out in a way that does not
occupy our attention, is that an action which occupies our attention is· one the
agent carries out with a sense of control.
But what is it for an agent to act with a sense of control? This is obviously a
large question, but to act with a sense of control must, at least be, to be aware of
guiding our actions, to experience our actions as those we initiate and those we
have the power to stop. We can either take this kind of awareness as primitive, or
attempt further explication. If such explication of what it is to act with a sense of
control is available it will, I think, most likely rest on our acting with an awareness
of the means by which we control or guide our actions. In essence, we control our
actions by acting on the basis of an evaluation of the possibilities open to us. So,
one suggestion would be that we experience our actions as controlled when we
act on the basis of our evaluation of possible actions, grasped as pO,ssible. On this
On Knowing One's Own Actions 379
suggestion, to experience an action as controlled is for the action to be the result
of a process of evaluation of the options available, grasped as options. Conscious
bodily action would then involve the agent having a grasp of the possible ways
that they could move their body as basic actions, and carrying out one action
rather than another on the basis of an assessment of their options. The agent's
grasp of which actions are available as basic actions will be based on a general
grasp of the ways in which they can move their body, which itself will be based in
ways that they have moved it in the past. It will also be based on a particular grasp
of the position of their body at the time of action, which itself will be based on the
ways in which they have most recently moved it. Let us take a case of an agent with
only two options: suppose that an agent has just one barely functioning arm that
they can move in just one way. If the agent grasps the two options of 'moving arm
up' and 'not moving arm up' as things that could be done, and acts directly on the
basis of an assessment of these options as 'to be done' or 'not to be done', the sub-
ject seems to have what is needed to act with a sense of control. However, if this is
what is involved in acting with a sense of control, then we have reason to think that
acting with a sense of control gives us some awareness of my action as my action.
It is certainly the case that only the agent of an action can be engaged in their action
in this way. Only the agent of action can act immediately on the basis of an assess-
ment of options. But we also have reason to attribute to the agent awareness not
only of the action through this kind of engagement with it, but also an awareness
that it is their action. Acting directly on the basis of an assessment of options,
grasped as options, means I have grasped that action as an option for me. An agent
may not be able to self-ascribe something as an option, because it may not have the
requisite first-person concept. However, acting directly on the basis of an assess-
ment of whether to carry out an action or not, given a grasp of the choices, must
either presuppose, or give me, a grasp of the action as an option for me.
We can summarize the line of thought using the following argument:
1. An agent acts with a sense of control when they carry out their action on the
basis of an assessment of the options, grasped as options, of acting one way
rather than another.
2. If an agent acts directly on an assessment of the options, grasped as options,
of acting one way rather than another, they manifest awareness of the option
as an option for them.
3. Therefore, acting with a sense of control is acting with self-awareness.
There is of course much more to be said about how to understand these various
claims. In particular, it would I think emerge from a proper consideration of them
that there will be various kinds of grasp that can be attributed to a subject of their
options, and that there will be correlated notions of self-awareness. There will be
agents who may at times conceptualize what actions are possible as basic actions
in their deliberations over their reasons for acting one way rather than another,
380 Lucy O'Brien
and who at other times have only a more primitive grasp of their options and who
evaluate them in less sophisticated terms. A creature who manifestly experiments
and revises their strategies for getting what they want may be thought to grasp but
not conceptualize their options. There may also be creatures who function as
rational systems in such a way that it seems right to say that they in some sense act
on the basis of an assessment of their options, but for whom we do not want
to say that they grasp their options as options, and so who do not act with
self-awareness.
If the above suggestion has anything going for it, we have the beginnings of an
account of our knowledge of our own actions. We are able to act consciously, and
acting consciously is acting with a sense of control. Acting with a sense of control
is acting in a certain way: it is acting directly on the basis of an evaluation of the
possible ways of acting, grasped as possible actions. Acting directly on the basis of
an evaluation of the possible ways of acting, grasped as possible actions, is acting
with self-awareness of a prinlitive form. So, when my acting consciously acts as
the reason for my self-ascription of the action, my self-ascription is knowledge-
able because it rests on an awareness of what I am doing.
I have said that acting consciously, which is acting in a certain way which
grounds our self-ascriptions, involves a form of self-awareness. This may give rise
to objections. It might be said that properly speaking such consciousness cannot
be a form of self-awareness as it does not involve a capacity for first-person refer-
ence. 25 Certainly, if being self-aware is understood to imply a capacity for first-
person reference, then this cannot be a form of self-awareness. However, it is form
of awareness which is such that a suitably cognitively equipped subject-a subject
with grasp of the first person and the concept of an action-will immediately be
able to self-ascribe the action they are conscious of in this way. A form of aware-
ness which is self-indicting in this way clearly needs to be distinguished both from
our awareness of the world and things around us, an awareness that may in many
cases be quite independent of our ability to ascribe anything to our selves, and
from our self-ascriptions themselves. I do not mind whether we call this primitive
self-awareness, or sonlething else. The point is to identify the need for and nature
of the phenomenon.
By way of conclusion, let us see how the account sketched fits with the features
of authority, relative aprioricity, and transparency introduced at the beginning of
the chapter as intuitive marks of our knowledge of our actions.
Given the claim that, in central cases, it is the acting in a certain way that
grounds one's knowledge of one's actions, we have an explanation of first-person
authority. The agent whose conscious or engaged action stands as the reason for
her self-ascription will be first-person authoritative over her actions. The agent of
25 This is what, for example, Castaneda would say. See C~t~n~~ 111~~2~ l=-~ 1 _
On Knowing One's Own Actions 381
an action will know in a way unavailable to others whether she is acting because
only the agent acts directly as a result of her assessment of the possibilities avail-
able, understood as possibilities. Furthermore, the agent will not only in general
know that she is acting when she knows in this way that she is acting, she will also
know what she is doing. As long as she has no evidence to the contrary, the agent
is entitled to the assumption that her motor systems are working properly and
that she has, on the basis of past action, a veridical grasp of both the general and
particular possible basic actions open to her. In bringing about a movement of her
body as the direct result of an evaluation of the ways she might have moved it,
grasped as ways she might move it, the agent can be said to know what she is
doing. She can know what she has done in a way not dissimilar to the way she
could know which object she has picked out if she had a grasp of the possible
objects available and picked one, rather than another. However, it is important to
emphasize a couple of points made earlier. When it is claimed that our actions can
act as the grounds for our knowledge of them, it is not being claimed that there is
knowledge of them under any description. So, while the moving of my hand may
be the potting of the pink ball, my knowledge that I am so moving it does not by
itself give me knowledge that I am potting the pink. The claim of first-person
authority with respect to our actions is to be understood as relative to certain
descriptions which could be regarded as basic. Authority is also compatible with
the possibility of mistaken basic action ascriptions, of raising my arm, for example.
My motor system could malfunction in such a way that my arm does not move,
or that it lowers rather than rises, or I could for one reason or another have a
non-veridical grasp of the possible actions open to me as basic.
It should by now be clear how our knowledge of our actions is being taken to
be compatible with the feature of relative aprioricity. There is obviously a consid-
erable role that is played by our proprioceptive and other perceptual faculties in
maintaining and updating a subject's grasp of the possible ways she can act.
However, when an agent with a grasp of the possible actions available carries out
a single basic action, she need not avert to the testimony of her senses to know
what she is doing. While my perceptual faculties are clearly required to give me
knowledge of the things I might do, they are not required to give me knowledge
of which, out of the things I might do, I am doing. All that is required to give me
knowledge of that, given the appropriate background, is to do it.
What of relative transparency? Given that acting consciously has been under-
stood to be acting on the basis of an evaluation of one's options, grasped as
options, we have reason to think that any agent who acts while asking themselves
what they are doing will act consciously. An agent who acts on the basis of an
assessment of their options while considering what they are doing, will be an
agent who is asking what options for action are being taken while assessing those
options. That must mean that the agent acts on the basis of an assessment of their
options, understood as options. Given that acting consciously will, in the absence
382 Lucy O'Brien
of any repressive mechanisms, be sufficient to ground a knowledgeable
self-ascription, an agent who acts, and who asks themselves what they are doing,
will know what they are doing.
REFERENCES
Johannes Roessler
1. INTRODUCTION
When we perform intentional bodily actions we are normally aware of what we are
doing. Of course, we can be mistaken about important aspects of our actions, and
our actions can have many surprising features. For example, in a Peanuts cartoon,
Lucy is shown tearing up what she takes to be Linus's blanket, but it turns out that
the blanket is in fact her own. In a sense Lucy does not know what she is doing. But
of course her lack of knowledge is far from total. She knows, for example, that
she is tearing up 'this' (perceptually presented) blanket. What this example illus-
trates is that typically, at least some of the descriptions under which an action is
intentional-some of the descriptions that figure in the content of the agent's
intention-are descriptions under which the agent knows what she is doing.
Much discussion in the philosophy of action has focused on the claim that
agents' knowledge of their intentional actions is non-observational. On one read-
ing, this claim is relatively uncontroversial. We do not normally acquire know-
ledge of our current actions by self-observation; being aware that you are tearing
up a blanket is not normally a matter of judging of someone you observe that
'this' (perceptually presented) person is tearing up a blanket. But some have
defended a stronger reading, according to which agents' knowledge is not based
on perceptual experience at all. For example, Elizabeth Anscombe maintains that
what she calls practical knowledge is akin to the knowledge of someone 'directing
a project, like the erection of a building which he cannot see and does not get
reports on, purely by giving orders' (1957: 82). On this reading, the claim may
seem counter-intuitive. Surely perception has some role to play in informing us of
the intended, or unintended, effects of an activity, and, thereby, of what we are
doing. Anscombe explicitly rejects this intuition. In her view, intentional action
really involves 'two knowledges': perception-independent 'practical' knowledge of
what one is doing, and perceptual knowledge of 'what is happening at a
given moment to the material one is working on' (1957: 89). Very crudely, on
Anscombe's picture, a claim to practical knowledge is not based on any kind
of evidence; rather, it is the conclusion of a piece of practical reasoning.
I am very grateful to Naomi Eilan, Christoph Hoerl, John Campbell, and Tony Marcel for many help-
ful discussions of issues raised in this chapter. Earlier drafts were presented at seminars and workshops
in Sheffield, Barcelona, Oxford, and Fribourg. Many thanks to the audiences for their comments.
I would also like to thank an anonymous referee for OUP for valuable suggestions.
384 Johannes Roessler
Correlatively, if a claim to such knowledge turns out be mistaken, the mistake is
'not one of judgement but of performance'. Many of those who find this 'two-
knowledge' view unattractive have opted instead for a 'two-component' view. As
K. Donnellan put it, 'knowledge of our own intentional actions is complex', divid-
ing up 'into an element of direct awareness, to be assimilated to the examples of
pain and anger, and other elements to which observation is relevant' (1963: 407).
The idea is that such knowledge draws on two distinct sources: introspective
knowledge of intentions (and perhaps acts of the will or tryings) and perceptual
knowledge of bodily movements and their effects, where the integration of the
elements will take the form of an inference, along the lines of 'I am trying to raise
my arm. My arm is rising. So (very probably) I am raising my arm.'
The two approaches share an important assumption. They both agree that what
perception in action yields, at least most immediately, is knowledge of events con-
ceived as mere happenings-knowledge of 'what is happening to the material one
is working on', or knowledge of one's bodily movements, where it is left open
whether the happenings constitute anyone's successful intentional action. So if we
wish to accommodate the intuition that perception can inform us of what we are
doing (as Anscombe does not), we have to think of perception as a source of evidence
from which we may infer what we are doing. In what follows I will suggest an
alternative view. On this view, in perceiving intended movements, or their
intended consequences, we sometimes directly perceive facts about our own activity,
such as 'I am tearing up this blanket'. The source of such self-awareness is neither
self-observation nor introspection, but perception, as it is used in controlling
intentional actions. This use, I will argue, must be distinguished from the use of
perception in observing mind-independent facts. (So on this view, there is an
important sense in which the slogan that agents' knowledge of their actions is
non-observational is correct, though this is not the sense in which it is commonly
understood.) A natural way of motivating this alternative view would be to launch
a direct attack on the assumption that we only perceive mere happenings. My
approach here will be more indirect-I will be mainly concerned to develop the
alternative view, rather than to defend it against competitors, though, I hope, by
the end, some of its attractions will be evident.
I begin by considering, and raising some difficulties for, the suggestion that
awareness of action is a case of ecological self-awareness in Gibson's sense, a case
of 'co-perceiving oneself' in perceiving the environment. The central claim I pur-
sue in the rest of the chapter is that the difficulties can be overcome by looking
more closely at the role of perceptual attention in action.
There is evidence that infants take pleasure not just in the interesting effect they
produce, but in their producing it. The contingency between their own active
movements and the interesting effect seems to be a crucial ingredient in what
makes the activity enjoyable. (See Lewis, Ch. 13, this volume.) This suggests that,
in some sense, young infants are sensitive to the fit between schema and move-
ment, and thus aware of their agency. But in what sense?
I want to suggest that the key to this question is provided by Piaget's remark
that secondary circular reactions 'essentially tend towards repetition' (1953: 154).
Infants' awareness of controlling a perceived event appears to be closely related to
a particular practical reaction-that of repeating the action responsible for the
event. It is not as if infants first satisfy themselves that they have successfully
completed one action, and then direct their attention to some other affordance.
Rather, perceiving the success of one round of rattling and perceiving an
affordance for another seems to come to the same thing.
What this suggests is that infants may perceive their actions in terms that are
causally indexical, in John Campbell's sense (Campbell, 1993). What makes a term
causally indexical is the way in which the causal significance of the property in
question is grasped. For example, perceiving an object as being within reach may
just be a matter of being disposed to reach for the object, given suitable desires.
Grasp of the causal significance of the fact that the object is within reach may be
exhausted by grasping its implications for one's own actions. It need not involve the
ability to give causal explanations of one's own, or others', actions citing the prop-
erty. Indeed, one need not be able to think of objects as being within reach of
someone else at all. Most importantly, causally indexical notions do not require
the ability to think of oneself as an object, or to have first-person thoughts. One
may perceive information that is implicitly 'self-specific', such as that an object is
'within reach', without the ability to think of the object as being 'within my reach'.
This latter kind of thought would involve the notion of x being within the reach
of y, and that notion need not be available to someone with a purely practical
grasp of what it is for an object to be within reach.
Intentional Action and Self-Awareness 387
My suggestion is that infants' grasp of the causal significance of their own
actions may be practical in just this sense. They may be aware of controlling an
event by perceiving it in causally indexical terms, as being repeatable, or
continuable-that is, as an affordance for further action. Their awareness may
thus be unreflective in the sense that (a) they may not be able to give causal
explanations of events in terms of their actions (such as 'the rattle is shaking
because I am shaking it'), and (b) they may not be able to think of themselves as
an object with causally relevant properties. (Perceiving an event as repeatable does
not require thinking of it as something 'I can repeat'.)
If this account is along the right lines, though, we should revise our interpreta-
tion of Neisser's proposal. The claim that the ecological sense of agency amounts
to perceptual knoweldge of transitive facts now looks problematic. If the fact that
I am shaking the rattle is to be evident to me, I need the conceptual capacities
required to grasp that fact. And these capacities include both a conception of
myself as an object, and causal concepts like 'to shake something'. This is not to
say that we should reject Neisser's claim that infants are aware of their actions 'as
their own'. The point is just that he fails to distinguish between a reflective and an
unreflective reading of what such awareness consists in. On the unreflective read-
ing, it is a matter of perceiving a distinctive kind of affordance, rather than of
grasping a first-personal fact. It seems to me that this distinction helps to under-
stand young infants' emotional experience. Preyer described infants' pleasure in
controlling things as 'the proud feeling of being a cause' (quoted in Gibson, 1993:
36). But the more recent literature tends to draw a sharp distinction between
pleasure in causality and pride (see Stipek, 1995, for a review). Perhaps the most
obvious difference is that pleasure in causality is tied to current activities, whereas
pride in achievements is typically experienced after the event. More importantly,
pride is a self-evaluative emotion-it involves being aware of, and evaluating, a
first-personal fact. In contrast, pleasure in causality seems to arise from experi-
encing an activity, rather than from evaluating the fact that one is engaging in
the activity. So it may arise from an experience of agency in which the causal
significance of the action is represented in causally indexical terms. l
So far I have focused on infants' sense of controlling events in their environ-
ment. But how about their perception of their own bodily movements? According
to one (admittedly, folk-psychological) interpretation, there is no significant
difference between the two cases at all: 'The new-born child does not realize that
his body is more part of himself than surrounding objects, and will play with his
toes without any feeling that they belong to him more than the rattle by
his side' (Maugham, 1998: 51). On the face of it, this picture conflicts with
1 This account suggests that explaining infants' pleasure in causality does not require attributing to
them a non-conceptual explicit representation of the self, as advocated by Bermudez (1995). For a
discussion of the relation between the development of self-evaluative emotions and of self-awareness,
see Roessler (2002).
388 Johannes Roessler
Neisser's suggestion that young infants have a sense of themselves as a 'bounded,
articulated and controllable body' (1988: 39). His suggestion is supported by evid-
ence that infants can distinguish their own moving legs from the moving legs of
another baby, when both are seen on a real-time TV screen (Bahrick and Watson,
1985). Even at this early stage, according to Neisser, infants perceive the body as
the self. They do so, he suggests, in virtue of the distinctive kind of control they
exercise over bodily movements: 'those objects and movements I can inevitably
and consistently control are parts of me' (1988: 39). But I think the conflict
between the two views is merely apparent. It is one thing to say that infants per-
ceive self-specific information (including information that a particular object is
their own body), it is another thing to say that they have an explicit representation
of the self. Infants' grasp of the causal significance of an object's being their own
body may be purely practical-they may perceive bodily movements as a distinc-
tive kind of affordance, as directly and consistently controllable, say, without
being able to reflect on the fact that these are movements of their own body.
In summary, it appears that the ecological account of the sense of agency does
not really speak to the epistemological question raised in the last section. One
response to this result may be scepticism about the idea that agents directly per-
ceive transitive facts. The unreflective experience of agency, it might be said, may
be a case of direct perception. But when it comes to an explanation of reflective
knowledge, this picture provides no help. A reflective awareness of the match
between intention and movement, according to the sceptic, requires a prior
awareness of the matching items: introspective awareness of my intention to raise
my arm and perceptual awareness of the intransitive fact that my arm is rising.
But it seems to me that this reaction cannot be right. I think we should hold on
to the basic insight of the ecological account, that perceptual information may
yield direct self-awareness in virtue of its role in controlling actions. What we
need is an explanation of how perception can simultaneously playa practical and
an epistemological role. To see how this is possible, I suggest, we need to focus on
the role of attention in action.
parameters need to be specified: when and where to move, which object to grasp,
how hard to push, how far to reach, and so on. To a large extent, the parameters
of our actions are specified by skills, as Neumann puts it, involving well-learned
motor sequences as well as routines of picking up relevant perceptual informa-
tion. But frequently this leaves some parameters under- or overspecified, and in
this case selective attention is required to add the missing specifications.
I suggest the role of attention here is that of selecting perceptual information to
answer practical questions. If this is right, the difference between the theoretical
and practical use of perception cannot correspond to different kinds of inten-
tional content. For to the extent that a perceptual experience comes as the answer
to the subject's question (whether theoretical or practical), its content will be
propositional. Rather, the difference is a matter of different directions of fit. When
perception is used to answer theoretical questions, its content will be the content
of a state of knowledge or belief. When it is used to answer practical questions, its
content will be the content of an intention.
Suppose this account of the role of attention in action (to be elaborated in a
moment) is along the right lines. Then, I suggest, it is natural to think of the dis-
tinction between attentional and non-attentional parameter specification as a dis-
tinction between two kinds of causal explanations. When perceptual experience
provides the subject with answers to her practical questions, it yields an under-
standing of why a particular parameter specification is the right one-why it is the
right thing to reach (thus far', or to play (that fast'. The subject's action, under these
descriptions, can be explained in terms of her reasons, and it will be intentional
under these descriptions. In contrast, when a skilled typist hears the word (spring',
for exan1ple, she will be able to use that information in selecting which fingers to
move to which position; but she will do so, as we say, automatically, without being
aware of any reason for selecting a particular finger or position. Accordingly, while
her action will be intentional under the description (typing the word «spring"', it
will not be intentional under any description detailing the movements of her fin-
gers. Neither the subject herself, nor, indeed, common-sense psychology in gen-
eral, will have anything very illuminating to say about her action described in
terms of finger movements, except, of course, that the subject is a skilled typist
trying to type the word (spring'.
There are a number of difficulties with this simple distinction between rational
and non-rational aspects of action control, however. For example, there are phe-
nomenological reasons for thinking the distinction is too simple as it stands. And
there is experimental evidence that can make the distinction look problematic.
But let me begin with the central epistemological question: how does perceptual
392 Johannes Roessler
experience provide answers to the subject's practical questions, and in what sense,
if any, does it do so 'directly'? A natural initial thought here is that to the extent
that perception contributes to the subject's grasp of practical reasons, it does so in
virtue of providing her with answers to theoretical questions, for example, informa-
tion about means to achieve her ends. I think while this is not incorrect, it is
misleading in so far as it fails to do justice to the ways in which theoretical and
practical issues can be intertwined. Let me illustrate this with the help of two
examples.
(1) Suppose you are driving a car and intend to take the next left turn. So you
form the intention 'I will turn left at that junction' when you see the junction
coming up. Intuitively, the transition from that intention to the intention
'I will turn left now' is not a matter of establishing a fact-'I have now reached
that junction'-in the light of which you decide to adopt the latter intention.
Rather, perceptual experience seems to control your intention immediately. You
look to see when to turn left.
This example brings out the sense in which, phenomenologically, perceptual
experience may engage with practical questions directly. Note, though, that this
does not conflict with the letter of our initial thought. You do pursue the practical
question by looking for factual information; it is by generating a rational belief
that perception contributes to the rational control of your action. The point is just
that no sooner have you acquired the belief than you intend to act on it. No
further practical decision is needed-the decision to act on the perceived
information has been made in advance. This is what makes it natural to describe
your perceptual expectations in practical terms: you expect to see when to turn,
not just when you will reach the junction.
(2) Suppose you want to establish whether you are tall enough to touch the ceil-
ing. One method would be to measure the distance between the floor and the ceil-
ing, and to compare it with that between your toes and your fingertips. Another
possibility would be to try to touch it. But there is an even quicker way to find the
answer. You may simply look at the ceiling to see whether it 'affords' touching. The
latter question is one which our visual experience will answer directly and, in gen-
eral' accurately: it informs you whether an object is within, or without, reach. We
can call this a reflective use of affordances. This reflective use is certainly not the
only way in which perceived affordances help to control actions. For example,
infants act on perceived affordances in a way that involves no first-person thought
at all. And of course we frequently exploit affordances as a matter of habit, with-
out scrutinizing them. To use an example that features prominently in an experi-
ment of Humphreys and Riddoch (Ch. 9, this volume), if you perceive a cup with
its handle pointing to the left, you may well, for that reason, use the left hand to
pick it up, without necessarily acquiring a belief to the effect that it is easier, or
more convenient, for you to grasp the cup with the left than with the right hand.
Nevertheless, there is an important role for the reflective use of affordances. It is
Intentional Action and Self-Awareness 393
often said that there are certain norms of rationality to which intentions (in con-
tradistinction to mere desires) are subject. For example, it is rational to form an
intention only if you think you will (or at least may) be able to execute it. So the
reflective use of affordances-its use to answer questions about one's practical
abilities-provides a basis for rational intentions.
On the face of it, though, it is far from clear what makes perceptual beliefs
about one's practical abilities rational. How is it that simply by looking at the ceil-
ing you may come to know that you should be able to touch it? Arguably, your
confidence is grounded on something-it is different from the case of a blind
hunch, say, about what time it is. Yet if we focus just on the content of your visual
experience it is not easy to see how your belief differs from a blind hunch.
Intuitively, someone could have a visual experience with the same content as
yours, but come to believe that the ceiling is definitely out of reach. One sugges-
tion here is that the reflective use of affordances involves a kind of motor imagery.
You imagine trying to touch the ceiling, using the content of your visual experi-
ence to control your imagined movements. The ability to imagine performing the
action successfully (without imagining any alterations to your dimensions or
motor abilities) will normally give you good grounds for thinking that you are
able to perform it. Of course, this procedure only works to the extent that the
imaginative exercise is subject to similar constraints to those governing overt
behaviour. But it may be argued that if you are reasonably skilled in motor
imagery, this assumption is largely correct, and you will be aware that it is. 2
The general idea I have been pursuing is that the role of attention in action con-
sists in the subject's selection of perceptual information for the rational control of
action, in contrast with the non-rational, 'automatic' (and usually faster) specifi-
cation of parameters by perceptual and motor skills. In many cases, the applica-
tion of this general idea is quite straightforward. The problem is that there are also
cases where it is not. Recall James's example of his absent-mindedly picking up a
pin while immersed in conversation. A natural description of what is going on in
this example is that his attention is (briefly) captured by the pin. ('Whilst talking
I become conscious of a pin on the floor.') There is little else, though, that can be
taken for granted here. Some would dispute that the action is intentional (under
any description). Some would follow James in maintaining that while the action
may be intentional, it is not 'willed'. But in the current context, the most pressing
question is: should we think of the action as a case of using perceptual experience
for the rational control of action? The case for an affirmative answer is that the
action seems to be both intentional and object-directed, more precisely:
intentional under a description in which a perceptual demonstrative ('this pin') is
used to identify the target of the action. In other words, the action seems to be
controlled by an intention whose content is provided by the agent's perceptual
2 For a review of suggestive experimental work on motor imagery, see Jeannerod (1997, ch. 4).
394 Johannes Roessler
experience. Intuitively, though, to pick up a pin absent-mindedly is to pick it up
without paying attention to what one is doing. On the face of it, then, it is
one thing for action to be guided by conscious perception, but another to occupy
the agent's attention.
As mentioned earlier, it was reflection on this kind of example that led James to
think that the phenomenology of deliberate bodily action cannot be a matter of
any distinctive perceptual experience. The obvious alternative is to draw a distinc-
tion between two ways in which perceptual attention may be involved in the con-
trol of action. In James's example, the intention to pick up the pin is the result of
attention being involuntarily attracted by the pin. In acting deliberately, the causal
dependence goes the other way: it is the agent's current intention that determines
what she attends to, setting questions which perceptual experience is used to
answer. I suggest that in the light of this distinction, my earlier proposal should be
amended as follows: to pay attention to an action is a matter of intentionally
selecting perceptual information for the rational control of the action. In James's
example, selective attention is controlled by a combination of habit and environ-
ment' rather than by the agent's intentions. But this should not blind us to the fact
that perceptual experience contributes to the rational control, or at least initia-
tion, of the action, in so far as it generates the intention to pick up the pin. 3
It might be said that this account leaves out a crucial distinction-the distinc-
tion James was getting at when he argued that ideomotor actions lack the both
phenomenologically and causally relevant <additional conscious element in the
shape of a fiat, mandate, or express consent'. In more recent psychological work on
action, we find a similar distinction between two different types of intentions,
<stimulus intentions' and <willed intentions', generating stimulus-driven or willed
actions, respectively (Frith, 1992). The intuitive force of this kind of distinction is
that, given the agent's passivity in acquiring a stimulus intention, the ensuing
action does not seem to be truly her action at all, unlike willed actions, where the
agent herself is responsible for the intention. It is this intuition which I think we
should reject. The mechanism by which habit and environment conspire to gener-
ate ideomotor actions is to present the agent with a practical reason, say, a pin that
it is both desirable and feasible to remove. An action arising from recognizing such
reasons is no less the subject's own action than actions preceded by protracted
deliberation. (Of course, you may come to regard your disposition to respond to
such apparent reasons as irritating and pedantic. But then what is irritating you is
precisely that it is (unqualifiedly) you who has the propensity to act in that way.)
3 Davis (1979) suggests that absent-minded actions of this kind should be characterized as some-
one doing something intentionally without having the intention of doing it. For the reasons explained
in the text, I find this unhelpful: I take it that intentional actions are actions that are informed by an
intention. But note that this does not commit me to the (probably incorrect) view that if an action is
intentional under a given description, the agent must have an intention to perform the action, thus
described. (See Bratman (1987, ch. 8) for objections to what he calls the (sim~~vie~'l
Intentional Action and Self-Awareness 395
Let me end this section with another qualification. I have suggested that the dis-
tinction between attentional and automatic parameter specification captures a
significant ingredient in the common-sense psychological conception of inten-
tional action. Now, the way I have been putting the distinction makes it sound as
if each parameter of a given action must be determined in either of the two
modes. But I think the distinction is best interpreted as relating not to parameters,
but to descriptions of parameters. Consider the intention to type the word
(spring'. You might say that this intention leaves the spatial parameters of the
action undetermined-it does not tell you which finger to move to which posi-
tion. But it is not as if the intention has no bearing on the spatial parameters of
the action. Rather, what the intention provides is a high-level description of the
spatial parameters: you intend to move your fingers to the positions of those keys,
the pressing of which will result in the word 'spring' getting typed. What a skilled
typist is good at is precisely the translation of this kind of high-level description
of the spatial parameters into particular finger movements.
One reason this point is significant is that it helps to interpret experimental
evidence regarding dissociations between perception and action. 4 One type of
dissociation occurs in cases where the subject experiences a perceptual illusion, yet
her actions are controlled by veridical information. For example, one relevant
finding here is that 'even though a fixed visual target surrounded by a moving
frame appears to drift in a direction opposite to that of the frame, subjects persist
in pointing to the veridical location of the target' (Milner and Goodale, 1995: 162;
see Bridgeman et al., 1981). Similarly, when people are subjected to vibro-tactile
illusions of the position of their arm (i.e. have an illusory proprioceptive experi-
ence of the arm position, induced by vibrating a muscle tendon, and cannot see
the arm), their judgements tend to reflect the illusion, yet often their actions do
not. Thus, subjects have no difficulty in moving their arm to a position marked by
a light, even though this involves moving the arm in the opposite direction to
what, on the basis of the illusory experience, would seem to be the correct direc-
tion. (Note that subjects tend to misidentify the direction in which they moved
their hand: they tend to think that they moved it in what appears to them to be
the correct direction (Marcel, Ch. 2, this volume).)
One might wonder, in the light of such data, whether the notion of attentional
parameter specification, even if it is an important element of the common-sense
picture of intentional action, has not been undermined by experimental psycho-
logy. According to Milner and Goodale, dissociations between conscious visual
perception and visuomotor responses (highlight surprising instances where what
we think we "see" is not what guides our actions'. They go on to suggest that
although one may consciously perceive many features of an object at the same
time as performing actions directed at the object, the perceptual processing
4 For reviews, see Milner and Goodale (1995, ch. 6), and Jeannerod (1997, ch. 3).
396 Johannes Roessler
involved in such conscious perception 'is quite independent of the visual process-
ing that actually governs those actions' (1995: 177-8). This might suggest the rad- I
ical conclusion that visual experience is simply impotent when it comes to fixing
the parameters of object-directed action. But note that in the experiments just
mentioned subjects do successfully execute intentions such as 'I will point at the
location of this [visually presented] target' or 'I will move my arm to the position
of that [visually presented] light'. These are descriptions of the spatial parameters
of the intended actions, and they are made available by the subject's visual
experience of some object. We would not normally expect, furthermore, that per-
ceptual attention is required to determine, on the basis of such intentions, in
which direction to move your hand. So it seems that the data slot naturally into
the common-sense notion of automatic parameter specification. Of course what
is surprising, from the common-sense perspective, is the use of 'implicitly
processed' information here, information that is available for the (automatic)
control of action but not for perceptual judgements. On the other hand, the con-
nection between automatic parameter specification and lack of awareness is
familiar from the way we think of skilled actions such as typing, where indeed it
is commonplace for us to be baffled by 'how we did it'. Part of the reason we find
dissociations between perception and action surprising, I suggest, is that we tend
to underestimate the degree to which even simple actions such as reaching or
grasping involve the exercise of skills.
There is nothing in what I have said so far, I think, that a defender of Anscombe's
account of agents' knowledge would need to object to. It is true that in likening
agents' knowledge to the knowledge of someone 'directing a project, like the erec-
tion of a building which he cannot see and does not get reports on, purely by giv-
ing orders', Anscombe seems to underestimate the extent to which intentional
action, especially object-directed action, depends on perceptual experience. But
this point can be detached from her central thesis that agents' direct knowledge of
what they are doing is 'practical'-based on the subject's practical reasoning rather
than on any kind of evidence. A defender of that thesis might well accept that per-
ceptual experience is an important source of rational intentions (or of 'practical
knowledge' in Anscombe's sense). The claim which she would reject, and which I
wish to defend, is that perceptual experience in action can be a source of ordinary
factual knowledge of what one is doing. It is this claim to which I now turn.
Let me begin with some comments on a further thesis of Anscombe's, men-
tioned briefly at the outset. She claims that object-directed actions typically
involve 'two knowledges': not just 'practical knowledge' of what one is doing, but,
in addition, observational knowledge of 'what is happening to the material one is
Intentional Action and Self-Awareness 397
working on'. I think on one reading, this claim is unobjectionable. If you pay
attention to what you are doing, you will normally acquire knowledge of what is
happening to the material you are working on. Furthermore, Anscombe is surely
right to stress the practical significance of such knowledge. As she puts it, some-
one 'doing or directing anything ... will not go on to the next order ... until he
knows that the preceding one has been executed' (1957: 88). But the term 'obser-
vation' has connotations that suggest a stronger reading of the 'two knowledges'
claim. On this reading, an agent's beliefs about what is happening to her material
are acquired in a particular way, which might be characterized, roughly, as attend-
ing with the intention to acquire knowledge of the mind-independent world. So,
as far as perceptual beliefs about the effects of an action are concerned, the agent's
position does not differ from that of an external observer. Of course, there are
many situations where this account is correct. It is correct, for example, with
respect to effects that emerge when the agent has no longer any control over them.
(You observe the trajectory of a ball you have thrown.) It is also correct with
respect to unintended effects of your current actions. (You observe the coffee
spilling over the floor.) But on the face of it, the strong reading of the 'two know-
ledges' claim glosses over an interesting distinction. Intuitively, when you are
immersed in a temporally extended action, say, negotiating a staircase with an
unwieldy piece of furniture, it seems off-key to describe your attention as being
occupied, even in part, by observing the material you are working on. Adopting a
spectator's standpoint would fatally distract your attention from the task. Rather,
your experience is continuously used to guide the execution of your intention.
This point has been made long ago by Brian O'Shaughnessy in his very suggestive
discussion of observation and the wil1.5 One of his claims is that the project of
observing your own actions makes conflicting demands on your perceptual
attention. As O'Shaughnessy puts it,
a conductor cannot listen as an observer to the music he makes, since he already listens to
it from the standpoint of creator.... Were he to listen as observer, his listening would no
longer be co-ordinated with and logically subordinated to the activity of making music. In
sum: one cannot be listening to the music one makes both from within the act and from
without the act. (1980: 29)
This raises two questions. What is involved in attending 'from within the act'? And
why should attending in this way be incompatible with observation?
On the first issue, the suggestion I have been pursuing is that the role of atten-
tion in action consists in the selection of perceptual information for the rational
control of action, in the use of perception to answer practical questions. Now it
might be said that there is something odd about portraying an agent immersed in
her action as pursuing a practical question. What would be the use of practical
6 Note that someone pressing the current objection is not committed to equating self-awareness
with introspective awareness of mental properties. She might argue that proprioceptive awareness of
one's bodily actions (though not knowledge based on other sensory modalities) is immune to error
through misidentification. See Cassam (1995) for a discussion of the claim that proprioceptive aware-
ness should in fact be regarded as a form of introspective awareness.
404 Johannes Roessler
deliberative, first-personal question of what to do. Part of the point here is that the
agent must have an explicit representation of her own abilities, so she can act on
the basis of beliefs as to how she will be able to produce an intended effect. And
of course she needs to be able to think about what effects she would like to bring
about. To adapt Neisser's suggestion, we experience actions as our own if their
consequences are appropriate to the schema by which they were generated, in so
far as the schema is generated by our own reasons.
REFERENCES