OXFORD 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 Naomi Eilan and Johannes Roessler
CONTENTS Introduction Mechanisms of control Epistemological issues 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. 2. 3. 4.
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 experiments 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 difficult 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 unbutton 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 n10vements, 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 question 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 Il1ysteries that come up when we attempt to explain the nature of control. One is the question 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, leaving 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 constitutive 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 introduct0ry 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 considering 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 bafflement 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 particular 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 departure James's description of going upstairs to change for dinner and finding himself getting into his nightclothes and climbing into bed instead. Reason labels this a 'double capture error', where attention is captured by some internal preoccupation, 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 comprehensive 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 anything like the way we do the pathological cases. That may be so, but the explanation 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 descriptions. 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 conflicts 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 deciding 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 personallevel 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.
The Two-Level Theory Much work in the psychology of action is informed by what is known as the ideomotor conception of action. On this conception, actions, in contrast to movements that can be explained by appeal to SR psychology or pure physiology, are controlled by representations of their perceivable results (James's 'bare idea of a movement's sensible effects': see 189Gb: 522), in the sense that the representation is causally responsible for the movement and may serve to correct errors during execution. (Note that the ideon10tor conception of action is to be distinguished from the notion of ideomotor actions, alluded to a moment ago. The former is intended to subsume all actions, ideomotor as well as 'willed' actions. For detailed discussion of both the ideomotor conception of action and ideomotor actions, see Wolfgang Prinz's chapter in this volume.)
8
Naomi Eilan and Johannes Roessler
Now, explanations of events by appeal to the goal-representation that causes them are more inclusive than intentional explanations as characterized above, in at least two respects. First, on the face of it, goal-directedness can apply to the behaviour of subjects where appeal to reasons for acting is for some reason inappropriate, but which is none the less of a kind to which SR explanations do not apply. Second, consider the mechanisms involved in executing even simple actions, like lifting a cup. Information-processing accounts of how we succeed in implementing our intentions will appeal to the operations of a variety of subsystems, the operation of which will need to appeal to represented goals. And here, too, the ascription of reasons for action is otiose and inappropriate. (We will discuss some of the issues raised by such accounts in Section 4.) The sense of paradox in the absent-minded case was derived fron1 an assumption that intentional action is action for a reason, from the subject's perspective, and that the subject should be credited with two conflicting intentions, and acts on one of them only, without having abandoned the other. But once a weaker notion of action is introduced, there is n10re room for manoeuvre. In the absentminded case, and in pathological cases, we can re-describe either or both of the intentions in terms of goal-representations, and then go on to give a purely causal, mechanistic (i.e. non-rational) account of how it comes about that, in particular circumstances, the goal informing a person's behaviour may be overpowered, causally, by a different goal, giving rise to an action incompatible "vith the first goaL On this kind of account, our bafflement is due not to a sense of paradox, but to unfulfilled causal expectations based on generalizations from the usual case. More specifically, the following two claims are common ground between Perner and Humphreys & Riddoch, though, as will emerge below, they develop the second claim in rather different ways. First, intention-thwarting behaviour, such as in the examples just given, is controlled by a low-level information-processing mechanism, whereby stimuli automatically activate well-learned, habitual activities. (The familiar sight of the junction activates the action of turning left; the sight of the right-facing handle prompts the right hand to be selected for grasping the cup.) Second, intentions do not control behaviour (directly' but by modulating the lower-level mechanism. To illustrate, Humphreys and Riddoch suggest that if you form a prior intention, say, to go to the supermarket before driving home, you thereby prin1e, or pre-activate, a particular response to a particular stimulus (e.g. the response of turning right at the junction where you normally turn left). Whether or not your intention will be effective depends on whether, when you reach the junction, the excitation of the primed response is greater than that of the response habitually activated by the stimulus. So the low-level mechanism serves two explanatory purposes in this picture: it is responsible for unintended habitual actions; and it is the mechanism by which intentions get a purchase on behaviour. Humphreys and Riddoch note an interesting parallel between this approach and late selection theories of perceptual attention. According to the latter, perceptual
Mechanisms and Epistemology
9
analysis of stimuli, including categorization, operates in parallel, independently of deliberate attention. So processing, not only of physical features such as spatial position, but also, crucially, of the identity of stimuli, is unselective. Similarly, the two-level approach to action control maintains that habitual stimulus-response associations are processed independently of deliberate attention. Accordingly, when an agent fails to. modulate the responses automatically activated by stimuli, whether due to absent-mindedness or to brain damage, her behaviour will be controlled by stimuli rather than by her intentions. An action such as lifting a cup, then, may be caused directly by the perception of the cup, or by the perception being primed by a prior intention. And both chapters suggest that in the first case we should describe the action as unintentional' not just in the familiar sense in which we talk about unintended aspects of intentional actions, but in the radical sense that it is an action that has no intended aspect at all-an action that is intentional under no description. So, as Perner argues explicitly, we have to abandon the view, which has been influential in the philosophy of action, that for an event to be an action just is for it to be intentional under some description. Now, that view has been challenged before. For example, according to O'Shaughnessy (1980), unthinkingly tapping a rhythm is a 'sub-intentional' action. But one important difference between the two challenges is this. Sub-intentional actions, according to O'Shaughnessy, are causally explained by desires-he is not suggesting that such behaviours fall outside the purview of common-sense psychological explanation. But on Perner's view, common-sense psychology simply has nothing illuminating to say about double capture errors. Neither intentions nor desires are relevant. So we need an informationprocessing notion of habitual action to fill the explanatory gap in the common-sense scheme. Moreover, the sense of paradox we set out with is, on this account, the consequence of a mistaken over-extension, and, thereby, mistaken application, of our notion of intentional action.
Common Sense and Mechanism It is precisely at this point that a defender of the first line of response to -the paradox will want to step in. The first line, recall, was that our notion of intentional action, so long as it is not given an incorrect theoretical gloss, can cope with dissolving the paradox generated by slips of action. A defender of this approach will insist that, contra Perner, such actions are intentional under some description. They might be compared with cases where we act without any deliberation, but clearly intentionally, in response to perceived affordances (e.g. accelerating when perceiving a traffic sign signalling the end of a speed restriction). Recall our earlier example of a double capture error: you turn left despite the fact that this thwarts your intention to go to the supermarket. On Perner's view, when we say that you took the left turn 'automatically', we show some awareness that the
10
Naomi Eilan and Johannes Roessler
explanation of your action is to be given not in terms of your practical reasons, the aims you were pursuing, but in terms of a mechanism that (on this occasion) operated independently of your objectives. On the alternative interpretation, the term 'automatic', in this context, should be read as spelling out how you acquired the intention to turn left: the intention was not the result of active deliberation; rather, you let yourself be 'saddled with' it by uncritically taking at face value the perceived practical significance of a stimulus. 1 Of course, this interpretation does not make the impairnlent to the unity of agency disappear-it suggests, precisely, that in the cases under consideration agents thwart their own intentions by acting with a different intention. But it may help to understand why there is nothing paradoxical about such impairments. The sense of paradox stems from the assumption that agents are active all the way down, as it were-active, that is, not just in executing but also in acquiring intentions. Once we acknowledge that intentions can be acquired unintentionally, we will no longer be baffled by breaches of the unity of agency. As James put it, 'not only is it the right thing at the right time that we thus involuntarily do, but the wrong thing also, if it be an habitual thing' (1890a: 114). On the other hand, given that rational agents are certainly able actively to form and revise their intentions, if they put their mind to it, we may still think of such cases as exceptions to the norm, where the norm is a matter, not of causal generalizations from the usual case, but of normative principles, such as a coherence, that are constitutive of rational agency. Clearly even on this alternative interpretation, the resources of common-sense psychology in explaining your absent-minded left turn remain limited. We can explain your action in terms of your intention, but will find little to say, other than referring to habit and absent-n1indedness, about the explanation of your intention. At this point an inforn1ation-processing theory might yield further illumination: for example, the unselective processing of habitual stimulus-response associations might be invoked to explain why the stimulus caught your attention. So the debate is between, on the one hand, theories on which informationprocessing mechanisms are held to give an exhaustive causal account of an action, and, on the other hand, theories on which information-processing explanations work in tandem with common-sense rational explanations. A sin1ilar (though far bigger) issue arises with respect to the explanation of actions informed by prior intentions. As noted earlier, Perner and Humphreys & Riddoch agree that the 'lower level' of control is also highly relevant here: intentions 1 This kind of view might also be motivated by philosophical concerns with the nature of rulefollowing. Crudely, the thought would be that we must make room for the notion of habitual intentional 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 perspectives. 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 overseerhomunculus 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 suggests 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, representations 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 information-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, dissociable systems. Just this argument is made in Humphreys and Riddoch's chapterentitled '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 commonsense 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 informationprocessing theory is required to avoid the conclusion that conscious will is an illusion. (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
Control and Knowledge To return to our paradox, it is worth drawing attention to the possibility that the different kinds of example may not all be amenable to the same kind of treatment. We have focused on the case of everyday slips of action, where there is clearly room for debate as to whether there is a description under which a double capture error may be an intentional action. But consider the standard symptoms of Anarchic Hand syndrome. The case against thinking of these movements as intentional actions is stronger than with non-pathological absentmindedness, as patients manifest both their awareness of what is going on and their determination to stop it by trying to arrest the movement using their good hand. So perhaps in this kind of case it is undeniable that we need an informationprocessing conception of action. (Compare Peacocke's suggestion that the movements constitute actions that do not have an agent-that are not someone's actions.) Again, the case of ES may call for yet another kind of explanation. There is no doubt that ES's movements in the 'cups' experiment are intentional actions-they are clearly intentional under the description 'picking up the cup'. But how about the description 'using the right hand', in a case where this violates the instructions (and her intention to con1ply with them). Humphreys and Riddoch suggest that the source of the problem is that the mechanism by which prior intentions prime the lower-level system is damaged; the intentional specification of the target (the cup) is effective, but the representation that should modulate response selection is 'degraded'. It might be said that this amounts to being unable to act intentionally as far as response selection is concerned. On the other hand, one might insist that perceiving the cup's handle as pointing to the right gives ES a good reason to use the right hand (it 'affords' picking up with the right), which would suggest that the response selection is intentional, albeit, owing to her pathological absent-mindedness, at variance with her prior intention. Consider now an example described by Frye and Zelazo of 3-year-olds' inability to play the following kind of game. The children are presented with two cards, for example, a red triangle and a blue circle. The task is to match a series of cards to these targets by some specified dimension (e.g. by colour). After a few rounds they are asked to switch to a different dimension (e.g. shape). It appears that 3-year-olds have great difficulty with this game: they tend to continue to sort by the first dimension, despite grasping the rules of both the shape game and the colour game, and being able to play each one separately, and despite being invited to switch to the shape game. The question, then, is, given the intention to cooperate with the examiner, why do 3-year-olds persistently refuse to switch dimensions? Initially, one might suppose that when asked to switch from the colour gan1e to the shape game children form the right intention, but then fail to inhibit an incorrect habitual response, the kind of explanation we may want to give in both pathologies and absent-minded cases. In fact, however, 3-year-olds' behaviour
14
Naomi Eilan and Johannes Roessler
seems to be perfectly deliberate, as can be seen, for example, from their comments on others' performance on the task, in which they insist that others too should persist in sorting by colour rather than shape. This is what they think, on the face of it, they should be doing, and it is relative to that that they judge their own and others' success and failure. Yet on separate occasions, they exhibit everything one would expect if they do know what should be done when the instruction is to sort by shape. So the developmental cases of failures of control seen1 to raise specific issues, and a special kind of challenge. While in the case of ES, for example, we may resort to a combination of common-sense psychology (to explain the use of the right hand in picking up a cup that affords picking up with the right) and neuropsychology (to explain ES's inability to select responses in line with her prior intentions), neither of these kinds of explanations seems capable of casting n1uch light on 3-year-olds' performance in the card-sorting experiment. Crudely, their behaviour seems to resist explanations in terms of some kind of (perhaps pathological) absent-mindedness; rather, it seems to force us to revise our conception of what it is for a mind to be present. In her commentary on Frye and Zelazo's chapter, Jennifer Hornsby argues that from the point of view of common-sense psychology the 3-year-olds' behaviour is inexplicable. The common-sense principle that renders it such is, she suggests: 'If you know what you should do and are able to do it then (in the absence of any tendency not to do it), you will do it.' There is good evidence that they do know what they should do, that they have no tendency not to play the game, and yet they don't do it. But if we take their behaviour to be governed by the common-sense generalization, we must say that they do not know what they should do. So, at the post-switch stage we must treat the child as both knowing and not knowing what she should do. This, in turn, suggests, that if their behaviour is to be at all explicable, we must turn to something other than common-sense psychology, an empirical theory that will explain these results. In effect, what Frye and Zelazo propose is a resolution of this paradox by ap-peal to a developmental theory on which the children are unable to work out what they should do, relative to the intention to cooperate with the examiner. What they cannot do is formulate the kind of switching rule required for success in the task. In more detail, on the developmental account Frye and Zelazo propose, the middle of the second year is a crucial milestone in the sense that children for the first time control their behaviour by the use of 'conditional rules' linking means and ends. Put differently, they engage in practical reasoning. Frye and Zelazo place this change in the context of a general developmental progression from domination by the 'exigencies of environmental stimulation' to a more autonomous mode of control, which they characterize in terms of the progressively more complex rules children are capable of employing in forming intentions. With respect to the developmental change manifested in the card-sorting task, they argue that -------------------------------
-------------~
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 intention 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-yearaIds' 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 behaviour 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 intentional 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 puzzleraising con1petition between two reasons for action only if we assume that subjects 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 connection 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 control, some account will have to be taken, in whatever way, of this intuitive connection in our common-sense psychology. This is the issue to which we now turn.
3. EPISTEMOLOGICAL ISSUES The problem of agents' knowledge was put on the philosophical agenda by Elizabeth Anscombe in her classic Intention (1957). According to Anscombe, it is a defining characteristic of intentional action that the agent has non-observational knowledge of what she is doing (not just of what she is intending, or trying, to do). AnscoITlbe's account of the source of this knowledge is best approached by looking at the case of knowledge of future actions. Suppose you are planning a visit to the
16
Naomi Eilan and Johannes Roessler
supermarket and someone asks you what you are going to buy. The obvious way to answer that question is to disclose the content of your shopping-list-to make a prediction of your future behaviour on the basis of deciding, or having decided, what to do. Put differently, you respond to a theoretical question about the future by answering a specific practical, or deliberative, question. 2 This contrasts sharply with an answer based on theoretical evidence, as when you predict that you will probably end up with a box or two of Turkish delight because you always do. Anscombe's central claim is that non-observational knowledge of our own current actions is available for the same reason, and in the same way, as knowledge of our own future actions. Roughly, I know about my current (intentional) actions without any need for observation or other kinds of evidence because they are the upshot of my practical reasoning. We can think of Anscombe's account as an attempt to chart a middle course between what, in her view, are two unattractive extremes. Traditionally it has been held that we have knowledge of (certain aspects of) our own actions 'from within', or from the first-person perspective, on a basis that differs from the kinds of evidence we use in finding out about others' actions. One extreme position would be to reject the very idea of a first-person/third-person asymmetry in relation to knowledge of agency-to assimilate the agents' perspective to that of an outside observer. At the other extreme lies the traditional account of that asymmetry. On this view, knowledge 'from within' is to be equated with introspective knowledge, the only possible objects of which are thought to be 'inner' conscious states or events. Knowledge 'from within' of bodily actions turns out to be introspective knowledge of such things as acts of the will or tryings. (Anscombe ridicules this as an appeal to 'a very queer and special sort of seeing eye in the middle of acting' (p. 57).) The general idea behind her middle course is that agents' knowledge is sui generis in the following sense: Agents' Knowledge. The agent of an action is aware of what she is doing in virtue of controlling her action, rather than on the basis of observation or introspection. 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 producer 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 movements 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 (abandoning 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 intentional 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 movements. 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, spontaneous 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 distinction 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 knowledge 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 situation' 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 accessible 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 strategies 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 experience 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 entitled 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
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 according 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 experience' and the nature of the experience of ownership.
4. ACTION AND PERCEPTUAL CONSCIOUSNESS:
DISSOCIATIONS AND CONNECTIONS Suppose there is a cup to your left and you lift it intentionally. If there is a kind of control of action that is essentially bound up with subjects being aware of what they are doing, it ought to be possible to say son1ething systematic about the connection between (a) the description under which your action is intentional (e.g. lifting a cup, lifting a heavy cup, lifting a cup on your left, lifting a cup with a handle turned to the right, etc.); and (b) the content of your awareness (being
22
Naomi Eilan and Johannes Roessler
aware of the cup, its location, shape, etc.). And it ought to be possible to spell out why and when the content of your awareness is critical for the intentional control of the action, as described. Work presented in the chapters by Tony Marcel and Marc Jeannerod, in which control and awareness come apart in various ways, provides material aplenty both for formulating common-sense intuitions about the nature of the connection between intentional control and awareness, and for putting them to the test. Let us have before us, first, two examples of the work they describe. 1. Marcel. Subjects' arm and hand were occluded by a surface that had a number of lights placed on it. They were asked to move their hand to a particular position identified by reference to one of the lights on the surface. Crucially, subjects' proprioceptive experience of the position of their arm was illusory (this was the result of an experimental technique involving vibrating a muscle tendon). So subjects experienced their hand as being to the left of the light when it was actually to its right. Despite this abnormal set-up subjects had no difficulty complying with the instruction-they moved their hand to the correct position. However, many were mistaken about the direction in which they moved-they claimed, say, to have moved their hand to the right when in fact they moved it to the left. 2. Jeannerod. Subjects were instructed to draw lines in the sagittal direction with a stylus on a digital tablet. The output of the stylus was shown to them on a computer screen seen in a mirror, itself placed so as to mask the subject's hand. On some trials, a bias was introduced in the output of the digital tablet, such that the line seen in the mirror deviated from the sagittal direction. The subjects therefore had to correct for the bias, and they were consistently able to do so. However, their reports of the direction in which they moved showed that they were unaware of this: they reported having moved in the sagittal direction.
Most people would find these results, at least initially, surprising. And there is little doubt that the surprise is relative to some set of expectations about the contribution made by conscious experience and knowledge to the control of action. Very roughly: we think that moving the hand to the left, in Marcel's experiment, for example, is something subjects do intentionally. Yet they are either unaware, or having an erroneous experience, of the direction of their movement. The surprise confirms the common-sense intuitions about a link between awareness and control; but the experiments seem to confound them. In fact, however, the experiments show first of all that articulating these common-sense expectations opens up a minefield of interconnected issues. And it is only as we try to unravel them that the question of whether the experiments indeed confound common sense can be addressed. In what follows we layout some of these issues and their relation to empirical findings, under two headings: awareness of bodily movement in intentional action; and the role of perceptual consciousness in the control of intentional action.
Mechanisms and Epistemology
23
4.1 Awareness of Bodily Movements in Intentional Action
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 succeeded 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 primarilyaware 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 proprioceptive 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 intentionally 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 intentional 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 'automatic 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 nonconscious 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 contents 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.
The Content ofAgents' Knowledge One way of putting the epistemological problem raised by these experiments is this. The dissociations between control and perceptual experience seem to drive a wedge between two elements of what we normally think of as rational control of action. One is the idea that a rationally controlled action exploits representations of objects, and their properties, that offer means for accomplishing the agent's objectives. In this sense, the leftward hand movement, in Marcel's vibro-tactile illusion experiments, is clearly rationally controlled. The other element is the idea that an action takes place because the agent thinks there is a good reason for performing it. In this sense, the leftward hand movement is, apparently, not rationally controlled. The accurate information regarding the hand's location exploited in controlling the movement appears to be non-conscious, and hence not accessible to the subject. So it is hard to see how it could have provided the subject with a reason to move leftwards. Now one kind of response to the experiments is to say we should associate agents' awareness firmly with rational control in the second sense (i.e. with agents'
Mechanisms and Epistemology
25
practical reasons). This is to hold on to Anscombe's intuition that agents' knowledge has its source in their rational control (see Sect. 3). Holding on to that intuition, though, con1es at a price. The price is that the scope of agents' knowledge may turn out to be narrower than we might have thought. Thus, the spatial parameters of object-directed movements, for example, their direction, may turn out to be normally inaccessible from the 'first-person perspective'. An alternative response is to insist that the content of agents' awareness is quite rich. In that case, though, we have to be prepared to sever the direct explanatory link between rational control in the second sense and knowledge of agency 'from within'. Let us spell out the two options in slightly more detail. A version of the second option is pursued by Christopher Peacocke in his commentary on Marcel's chapter. In Peacocke's view, it would be a non sequitur to argue that if a given aspect of an action is controlled at the 'automatic level', then the agent will not be aware of that aspect 'from the inside'. Briefly, for Peacocke, the philosophical significance of the experimental evidence is this. His general approach to knowledge of action 'from within' is that such knowledge is based on the phenomenology of action, which he takes to be a n1atter of a distinctive nonperceptual experience, it's seeming to the agent that he is doing this or that. Now, the content of the phenomenology of action, he argues, depends on what the subject is trying to do, and it is tempting to think the content of the trying exhausts the content of the phenomenology. The experin1ental evidence, Peacocke argues, shows that this is not so. While subjects in Marcel's experiment certainly tried to move their hand to the location of the light, it is not the case that they tried to move their hand either to the left or to the right. So to explain subjects' (illusory) impression of moving their hand to the left (when they moved it to the right), we have to bring in materials other than the content of the subject's trying. Peacocke envisages a subpersonal explanation, drawing on the illusory representation of the position of the hand. The general lesson he draws from the experimental work is that the content of the phenomenology of action is richer than that of the subject's trying (or that of the intentions informing the action). Advocates of the first option, in contrast, may take their lead from the following remark of Anscombe's: 'In general, as Aristotle says, one does not deliberate about an acquired skill; the description of what one is doing, which one completely understands, is at a distance from the details of one's movements, which one does not consider at all' (1957: 54). This passage suggests that Anscombe might well have been quite sympathetic to Jeannerod's claim that agents' awareness is 'remarkably limited'. However, the claim that when picking up a cup you are not aware of the direction in which you are moving seems to be counterintuitive, to put it mildly. There are two things that a defender of this option might say to make it more palatable. One is that of course normally when you grasp a cup, you will be able to give a reasonably accurate description of the spatial parameters of your movement, given that you can see what is going on. The other
26
Naomi Eilan and Johannes Roessler
is that even without observing the movement you can give an accurate descriptive identification of its spatial parameters, for example, 'I am reaching out in the direction of the cup'. (Even subjects in Marcel's experiments would be able to identify the direction in which they are moving descriptively; they could tell us 'I am moving my hand in the direction of the light, whatever that is'. What they are mistaken about is whether to move towards the light is to move to the right or to the left.) Furthermore, you could use that descriptive identification, together with a judgement about the egocentric location of the cup, to infer in which direction you are moving. The point is simply that the direction of your movement, egocentrically specified, is not part of the content of your non-observational, non-inferential awareness of what you are doing. Whichever of these lines one takes, there is an important point they have in common. They both suggest that the conception of an action that figures in the agent's deliberation, or intention, or forms the content of her trying (in short, the conception under which the agent exercises rational control, in the second sense) is son1ewhat removed from the details of her bodily movements. Then the obvious question, relevan t to both responses, is: how far removed? Does the agent think of 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 taking 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' (perceptually presented) light; they are not aware of moving 'this' (perceptually presented) 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 9
4 An interesting exception is the case of deafferented subjects. As Marcel points out, IW, the deafferented 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, intuitively' 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 representations 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 assignment, 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 induction 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
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 perceive 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 accelerating. 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 proposal, see Ch. 18.)
4.2 The Role of Perceptual Consciousness in the
Control of Intentional Action
The Epiphenomenalist Challenge One thing the experiments certainly do show is that it is a mistake to assume that perceptual information involved in the control of action must be the content of a conscious perceptual experience. This is a thought that has become familiar from the much-discussed case of blindsight. Blindsighted subjects do not experience the size of an object in the blind field, yet when they are told to reach for the object, visual information about its size plays a causal role in controlling the pre-shaping of their hand. A prin1a-facie extreme reaction to the experiments is to say that as far as action is concerned we may as well all be blindseers. In fact, this is a natural way to read David Milner and Melvyn Goodale's following claim: 'Of course one may consciously perceive many features of an object at the very same time as one examines it, reaches for it, or n1anipulates it. The studies in this chapter support our contention, however, that such perceptual processing is quite independent of the visual processing that actually governs those actions' (1995: 177-8). On the strongest possible reading, the passage from Milner and Goodale amounts to the claim that conscious visual experience is wholly irrelevant relative to causal explanations of object-directed actions. Actually, the empirical data regarding dissociations between perceptual experience and action control do not even provide a prima-facie case for this sweeping verdict. They concern a limited set of aspects of bodily actions, for example, the direction of hand movements. There is no evidence, for example, that when you choose to pick up a green smartie out of an array of differently coloured smarties, your colour experience plays no causal role. However, there is a more specific epiphenomenalist thesis that does seem to be supported by the evidence, and that sets the agenda for John Campbell's
30
Naomi Eilan and Johannes Roessler
chapter. This is the thesis that our experience of the location of an object plays no causal role in setting the spatial parameters of actions directed at the object. The significance of the thesis, for Campbell, is this. According to a traditional, and plausible, view, there is a fundamental connection between object-directed intentional action and perceptual demonstratives, ways of identifying objects such as 'this [perceptually presented] melon'. In the basic case, you need to be able to identify an object demonstratively to be able intentionally to act on it. (If your only way of identifying the target of your action is 'the nicest melon in the market', you will not be able intentionally to act on that object.) How should we explain the connection? An initially plausible suggestion is this. The distinctive feature of a perceptual demonstrative is that you only understand it if you perceptually experience the object referred to; and perceptual experience of an object is precisely what enables you intentionally to act upon the object, to determine the spatial parameters of a movement that will enable you to manipulate the object in the way you would like to. Our epiphenomenalist thesis, however, seems to undermine that explanation. It suggests that conscious experience is not in fact what sets the spatial parameters of your movement. Campbell's analysis of the problem here draws on the theory of two streams of visual processing presented by Milner and Goodale (1995). The theory features three distinctions: • an anatomical distinction between two main pathways of visual processing, the ventral stream and the dorsal stream; • a functional distinction between two uses of visual information: its use for the control of object-directed action (pragmatic) and its use for the recognition of objects (semantic); • a phenomenological distinction between conscious and unconscious visual processes. Milner's and Goodale's suggestion was that these distinctions map onto each other as follows. It is visual information used for the recognition of objects (in their terms, for 'perception'), processed in the ventral stream, that involves conscious experience; visual information used for the control of action, processed in the dorsal stream, operates outside conscious awareness. The central notion in Campbell's solution to the question of how to explain the link between perceptual demonstratives and action is that of selective perceptual attention. Very crudely, the idea is that conscious attention to an object is required to identify the target for the non-conscious information processing that determines the spatial parameters of object-directed actions. It is what enables the semantic system to 'communicate' with the pragmatic system. The subject's conscious attention determines which object is to be acted upon; then the pragmatic system takes over, setting the spatial parameters for the movement. The upshot is that your conscious experience of location does have some limited causal role to
Mechanisms and Epistemology
31
play: it enables you to consciously attend to the object, and thus to define the target for the non-conscious information processing that controls the details of your movement.
Attention, Affordances, and Consciousness Marcel's point mentioned earlier about the focus of attention in object-directed actions, and its bearing on the content of agents' awareness, is based primarily on reflection on the phenomenology of action. In this context, attention is conceived simply as a central structuring feature of our conscious lives, and, as such, as familiar to us as anything. As James put it, 'Everyone knows what attention is' (1890a: 403). The thrust of Campbell's chapter is that information-processing psychology serves to deepen our understanding of this very same phenomenon. This way of appealing to information-processing psychology to supplement comn10n-sense psychology is similar in spirit to a natural reading of the approach taken by I-Iun1phreys and Riddoch, in their account of the way in which lower-level control underpins habitual action (where this general approach is contrasted, on the one hand with approaches that seek to replace common-sense psychology with information-processing accounts, and, on the other, with approaches that think of common-sense psychology as employing a set of concepts that are closed to further en1pirical expansion). However, there is a crucial difference in detail between the ways in which this marriage of the two psychologies is achieved, which turns on the way the notion of 'affordance' is used. For Campbell, the perception of affordances is automatic in Jeannerod's sense, that is, essentially non-conscious. To be in a state in which one is correctly described as perceiving affordances only is to be in a state in which the input system has been causally hooked up with the action system independently of the mediation of any semantic processing. On this kind of account, the Milner and Goodale challenge is to find some role for conscious perception, which is, by definition, the perception of something other than affordances. His solution is spelled out in his treatment of blindsight. On his account, the blindsighted subject perceives some of the affordances provided by the object, that is, opportunities for action offered by an object in the blind field, some of its dispositions to behave in certain ways in response to certain actions. What she lacks is an understanding of why the object affords these actions. In Campbell's terms, she lacks knowledge of the grounds of the affordances-the categorical properties in virtue of which it affords particular actions. For example, a blindsighted subject may perceive that something affords picking up in a certain way, but she does not perceive the categorical properties of the object-its size, shape, or location-in virtue of which it lends itself to being picked up in that way. Knowledge of the categorical properties, Campbell suggests, requires conscious perceptual experience. And it is this, in turn, which is needed for intentional action, action for a reason.
32
Naomi Eilan and Johannes Roessler
For Humphreys and Riddoch, in contrast, the perception of affordances is automatic only in the sense that affordances are picked up by the lower level of control, invoked to explain habitual action. There is nothing in their account that rules out the consciousness of the affordances. Their work is part of a project of giving detailed empirical substance to Gibson's claim that we perceive affordances; and their use of the notion of affordances is probably closer in spirit to Gibson's own account, on which affordances are described as what we notice when perceiving. According to Gibson, 'what we perceive when we look at objects are their affordances, not their qualities. We can discriminate the dimensions of difference if required to do so in an experiment, but what the object affords is what we normally pay attention to.' He also makes the developmental claim that 'the affordance of an object is what the infant begins by noticing. The meaning is observed before the substance and surface, the colour and form, are seen as such' (Gibson, 1979: 134). Campbell's analysis of blindsight presents a direct challenge to such claims. Perceptual states guide behaviour either by representing the affordances of things, or by representing the grounds of affordances. If the former, then there is no reason to think that the states are conscious. If the latter, the states will be conscious, but what will be experienced is the categorical ground of the affordance, not the affordance itself. The heart of the argument here turns on the idea that dispositions cannot be consciously perceived. A defence of the idea that perception of affordances can be conscious would involve insisting, inter alia that dispositions can, contra Campbell, be the objects of direct conscious experience, and this is what is experienced in the case of habitual actions. Why should we be interested in pursuing this line of thought? One of many reasons can be brought out as follows. On Campbell's account, consciousness comes in only for the sake of theoretical understanding and reasoning. This, in turn, explains what is distinctive about intentional action: it is informed by such reasoning. This approach will disappoint those whose intuition is that there is some kind of explanatory link between consciousness and action. A specific thought here, discussed by Tom Baldwin in Chapter 8, is that perceptual states are conscious in virtue of their perspectival content-they represent things relative to the subject's point of view, as being to the left, or within reach-and that it is of the essence of such contents that they can be used in guiding the subject's purposive activities. What the above discussion brings out, at the very least, is that if there is something important in this idea that conscious perspectives are linked to action, considerable care is needed in making good this intuition. A second, less abstract reason for wanting to establish a connection between affordances and perceptual consciousness would be to explain the nature of a more primitive level of action control, the kind of action control we find in non-human animals and young infants. For example, there is evidence that by 15 weeks infants perceive whether an object is within arm's reach or too far away: they will only reach for a desirable object if it looks to be within reach (see Neisser, 1993).
Mechanisms and Epistemology
33
Arguably, though, it would be far-fetched to credit the infant with an understanding of the grounds of that affordance (of the fact that the object's distance from herself does not exceed the length of her arm). We may not want to credit 15-weekolds with a reflective understanding of their dimensions and abilities, or, for that matter, with any first-person thinking at all. Intuitively, an infant's grasp of what it is for an object to be within reach would seem to be entirely practical. As John Campbell put the point in earlier work, infants, or animals, may grasp the causal significance of an object's being within reach entirely in terms of its implications for their own actions. This does not require grasping the notion (x is within reach of y', a notion that can be used in thinking about things being within reach of agents other than oneself. Their understanding that something is within reach may be (causally indexical', in the sense that it consist in the disposition intentionally to reach for objects perceived to be within reach (Campbell, 1993). A major issue facing the second route is whether it is entitled to help itself to the notion of intentional behaviour. The issue, in crudest outline, is this. Intentional behaviour, it might be said, is behaviour informed by a grasp of the causal relation between means and end. We move in a circle, on this view, if we appeal to a creature's intentional reaching in explaining her grasp of the causal significance of an object's being within reach. The causal belief that reaching for an object will bring one's hand into contact with it is part of what makes the behaviour intentional. This conception resonates with the account of the developmental sequence offered by Douglas Frye and Philip Zelazo in Chapter 11. In their view, young infants' behaviour is not intentional at all, given that at this stage the infant is not aware of how a particular goal is to be reached. Properly intentional behaviour, they argue, does not emerge until the middle of the second year. At this age children show surprise, not only in cases where an action fails to produce its expected outcome, but also when a desired outcome is brought about by an action that bears no intelligible relation to it (see Frye, 1991). Such surprise reactions may be taken to n1anifest the beginnings of a concern with causalexplanatory relations between means and ends-an understanding of the grounds of affordances. (See also Gopnik and Meltzoff, 1997, ch. 5.) Yet, it is difficult to resist Michael Lewis's description of infants' behaviour as intentional. Lewis found that infants as young as 2 months rapidly learn to exploit a contingency, set up by an experimenter, between arm pulls and an interesting display, showing intense joy at the appearance of the display (and anger when the association is broken). Control subjects, who were also presented with the display but lacked any control over it, did not show these emotional reactions. On the face of it, nothing could be clearer than that experimental subjects were striving to get the display to re-appear by pulling their arm. On the other hand, Lewis is surely right to regard 2-month-olds' intentions as different in kind from 18-month-olds'. And the difference may be precisely a matter of the different ways in which the causal significance of the contingency between means and ends is grasped.
34
Naomi Eilan and Johannes Roessler
Coming back now to the problem of affordances: one thing Campbell's challenge makes very vivid is that if we do want to appeal to the notion of affordances to explain human development and to argue for some kind of constitutive connection between action and conscious points of view, then, at the very least, we have to solve simultaneously for what the perception of affordances consists in and for the sense in which behaviour informed by the perception of affordances only can be intentional.
5. THE SENSE OF OWNERSHIP
What is involved in being aware of an action as one's own action? Is ownership of one's actions experienced, or inferred, or invented? And what do we mean by ownership here in the first place-what is it for an action to be the action of a particular agent? We want to approach these questions by focusing on two specific issues. The first concerns the relation between awareness of ownership and awareness of the content of an action-between agents' knowledge of what is being done and their knowledge of who is doing it. Is is possible to be aware of an action (from within', without experiencing, or thinking of, the action as one's own? Some of the pathologies of action studied by psychologists suggest an affirmative answer; but it is far from clear what we should make of this. The second issue concerns the relation between (as Jerome Dokic puts it) the fact of ownership and the sense of ownership, or between the metaphysics and epistemology of ownership. A natural thought here is that the metaphysical issue is the prior one: if we want to clarify the nature of agents' awareness of ownership of actions, we had better first gain some understanding of what it is for an action to be the action of a particular agent. But there are also strong intuitive grounds for thinking that there are constitutive interconnections between the the metaphysical and the epistemological question. In attributing an action to someone we have in mind something richer than ownership of a moving body. (When you squash a spider as a result of being thrown out of bed whilst asleep, this is not an action of yours in the sense in which we ordinarily talk about ownership of actions.) The obvious proposal is that to be the agent of an action you need to exercise some degree of control over the action. But how should we characterize the relevant kind of control? One suggestion might be that it is an essential feature of the kind of control that goes with ownership that it provides the agent with a sense of ownership. At least, it might be said, this is true for intentional actions (in contradistinction to idle tongue n10vements or foot-tappings). Recall the case of Anarchic Hand syndrome. In the light of patients' reactions to the movements of their (anarchic hand', it seems intuitively compelling that these are not the patients' own actions. But note that part of the reason we find this compelling is simply the fact that patients do not experience the movements as their own actions.
Mechanisms and Epistemology
35
One question, then, is whether the crude intuition of a necessary connection between the fact and the sense of owernship can be sustained. If so, we face the question of how to explain the connection.
Immunity to Error through Misidentification In philosophy a common way of approaching the subject of self-awareness is to examine a range of distinctive ways we have of acquiring knowledge of our own properties, the mark of which is taken to be the phenomenon of immunity to error through misidentification relative to the first person (see Shoemaker, 1994; Evans, 1982). Suppose you see someone kicking a football and judge: John has kicked the football into the hedge. The following is a distinct epistemic possibility. You are right that someone has kicked the football into the hedge, but wrong in your belief that it is John. This is possible because your judgement that it is John is itself based on an identity judgement of the form 'this person is John', an identity judgement that can go wrong. In other words, your judgement is subject to error through misidentification. Now this possibility is not restricted to judgements about other people's actions, as the following example (borrowed from Patrick Haggard's chapter) brings out: 'Ow!' squeaked something. 'That's funny,' thought Pooh. 'I said "Ow!" without really oo'ing.' 'Help!' said a small, high voice. '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!'
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 immunity 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 single 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 taking 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 question. As Tony Marcel points out, patients with Anarchic Hand syndrome are proprioceptively 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) movement, 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 proprioceptive 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 'anarchic' 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 consistent 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) unqualified 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 supposition 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 distinction 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 distinction 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 informing much of this work is that to find out who is the agent of an action, it is necessary 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 etlerent 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' consciousness 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 question 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 ownership. 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 judgement '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 schizophrenic 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 ownership of an action requires being aware of the action as something that is controlled 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 understanding. 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 representations 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 perceptual 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 matter, 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 awareness 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 version of the second strategy, on which the sense of ownership does not constitutively involve any causal understanding. However, O'Shaughnessy argues that awareness of trying does not occur
Is the Sense of Ownership Pre- or Post-motor? Earlier we mentioned Brian O'Shaughnessy's thought experiment concerning the case of total failure (see Sect. 3 above), where an attempt at bodily action is comprehensively foiled by sudden paralysis. What O'Shaughnessy draws from the thought experiment is that one can be aware of a trying in advance of, and indeed in the absence of, any bodily movement. The first part of this claim bears an interesting relation to experimental work presented by Patrick Haggard. The starting point for Haggard's work was the classic study by Benjamin Libet, alluded to at the outset, where subjects were instructed to make a hand movement at a
Mechanisms and Epistemology
41
binding, Marcel to the egocentric content of motor specifications, and Peacocke to the nature of trying), they are all agreed that the sense of ownership is 'pre-motor'. Haggard's work raises two immediate questions. One is how to interpret the two kinds of judgements-exactly what is their content and their basis. The other question is to what extent the findings license any general conclusions-whether conclusions relating to the nature of agents' knovvledge or (in line with Libet's original concerns) regarding the causal power of the conscious will. An interesting background question here concerns the relation between these two kinds of conclusion. What is the basis of M judgements? Haggard's account suggests that in one sense, the basis is simply a certain stage in pre-motor processing, devoted to the selection of particular movements. It might be said, though, that if this is the whole story about the basis of M judgements, we can hardly think of such judgements as expressive of any kind of knowledge. Given that the infornlation processing giving rise to the judgements is presumably remote fronl consciousness, the judgements would be no more than blind hunches without any rational basis from the subject's point of view. Now, of course, there are independent grounds for questioning whether the judgements can be said to express awareness, or knowledge, of the movement. The problem is not just that subjects get the time of the onset of the movement wrong (albeit only slightly). More importantly, we would expect a state of awareness of a movement to depend causally on the movement itself-which would seem to require the reverse temporal sequence to the one Haggard hypothesizes. (At one point, Haggard hinlself conlments that 'people think they move before they actually do', allowing that the thought nlay fall short of knowledge.) A possible way of solving simultaneously for both of these problems might be to suggest that to get right the basis of M judgements we need the notion of trying, as it has been articulated by philosophers. So Haggard's experinlental work on real subjects might be said to dovetail with the lesson O'Shaughnessy draws from his thought experinlent. The idea would be that M judgements are based, causally but also rationally, on the conscious experience of trying (where this could either mean: the conscious event of trying itself, as O'Shaughnessy would argue, or: the subject's introspective awareness of her trying, as Peacocke maintains). We then face the question of what to make of subjects' tendency to rely on their awareness of trying in answering questions about the onset of their movement. One view might be that this is simply a mistake we tend to make-if perhaps an understandable one, given that, if O'Shaughnessy is right, 'trying' and 'moving one's body' can be descriptions of the very same event. An alternative view would be that awareness of trying does normally constitute a sound basis for judgements about bodily action, despite the fact that the trying precedes the bodily movement. This is Peacocke's view. As briefly mentioned in Section 3, Peacocke argues that the experience of trying warrants judgements about bodily actions in virtue
42
Naomi Eilan and Johannes Roessler
of the fact that a trying reliably (and indeed constitutively) gives rise to the relevant bodily movement. Should we conclude that the existence of tryings, conceived as conscious events that constitute a source of knowledge of actions 'from the inside', has been experimentally established? The case is perhaps less clear-cut than its advocates make out. Haggard takes it that subjects' judgement that they moved their hand at t reflects an awareness they had at t of moving their hand now (where t occurs 86ms before movement onset). But it is far from obvious that this is so. Another possibility might be that subjects' knowledge that the movement occurred arises from some post-motor source (e.g. perceptual experience of the movement), and that they are simply mistaken as to when the movement started. A possible explanation of this latter mistake might then be that subjects overestin1ate the degree to which the movement conformed to their intention (perhaps they intended for the movement to start at t). So there is certainly room for further debate about the epistemological role of tryings. The wider issue raised by Libet's work is how it bears on our commen-sense conception of action control. A natural gloss on Libet's findings is that although we do have knowledge of our actions 'from the inside', in an important sense such knowledge is nevertheless the kind of knowledge spectators have of events beyond their control. Consciousness of both intention and action looks like a mere epiphenomenon of the various stages of pre-motor processing unfolding in our brains. (As Haggard puts it, a 'tabloid paraphrase' of Libet's results might be that 'your brain knows you are going to move before you do'.) Now, Haggard himself argues that his more specific account of the content of W judgements-on which such judgements express an awareness of intention that arises from the computational process of movement selection-suggests a possible role for consciousness. Awareness of intentions for specific movements nlay 'allow a movement to be evaluated prior to its actually being executed'. The idea would be that while Libet may be right that 'freely willed' actions can be 'unconsciously initiated', that does not make the agent's consciousness of intentions causally irrelevant. Becoming conscious of an intention may give you a kind of veto over it. This still leaves us with the epiphenomenalist thesis concerning the initiation of 'freely willed' actions. There is an illuminating parallel here with the debate surrounding the role of non-conscious perceptual information in setting the spatial parameters of object-directed actions, touched on above (see Sect. 4.2). Recall Campbell's suggestion that the causal role of perceptual consciousness resides in the subject's conscious perceptual attention, defining the target of a movement whose detailed parameters are then specified on the basis of non-conscious information. A corresponding point about Libet's experiments would be that the causally relevant contribution of the subjects' conscious intentions lies in their prior decision to move their hand 'sometime', delegating the task of deciding exactly when to move to some subpersonal mechanism (which may then inititate
Mechanisms and Epistemology
43
pre-motor processing before the subject becomes aware of the present-tense intention to move). Now it might be said that this point is less effective than Campbell's point in the spatial case, given that Libet's subjects were explicitly asked to move their hand 'at a time of their own choosing'. But note that in effect this means 'at an arbitrary time', a time chosen in the absence of any grounds for choosing one time rather than another. (On Libet's (as it were) existentialist conception, the paradigm of a (freely willed' action would be an acte gratuit rather than an action demanded by the agent's conception of her practical reasons.) Under these circun1stances, delegating control to subpersonal mechanisms may be precisely a way of choosing when to act.
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 phenomenology 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 further 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
any materials prima-facie absent in those circumstances, such as introspective awareness. Marcel's own radical conclusion is that what he calls a minimal sense of ownership is provided by the spatial content of movement specifications (the specifications which, according to Haggard, form the content of subjects' response to Libet's W question). The self enters the representational scene as the origin of the egocentric frame of reference utilized in movement specifications. Actually, as Marcel himself emphasizes, the self remains in the background here: it figures only as an implicitly represented origin of a point of view (implicit in such representations as 'to the left' or 'in fronn, rather than as an explicitly represented object. Suppose we grant that the egocentric content of motor specification constitutes a form of implicit presence of the self in the phenomenology of action. As Christopher Peacocke argues in his commentary, this is not the same as accepting that it constitutes a sense of ownership, a sense of being the agent of an action. On the face of it, given that ownership of actions seems to be a causal notion-the notion of someone's intentions controlling a movement-it is indeed hard to see how the implicit presence of the self qua purely spatial origin of an egocentric frame of reference can be sufficient for a sense of ownership. (Note that Marcel himself discusses the role, and importance, of a sense of the effectiveness of one's intentions, under the heading of 'the long-term sense of agency', i.e. a conception of oneself as an agent with causal powers, which, he suggests, forms a necessary background to the 'occurrent' sense of agency.) Where does this leave the claim that awareness of ownership of actions is a case of immersed self-awareness? There are a number of options. One would be to hold on to Marcel's claim that the self figures only implicitly, and argue that this may be sufficient for a causal sense of ownership. Consider the case of young infants' enjoyment of controlling interesting effects, mentioned in Section 4 above (for extended discussion see Chapter 13 by Michael Lewis). Infants' emotional reactions to interesting effects under their control seem to demonstrate some kind of awareness of controlling things; yet, intuitively a sense of control of this kind does not require an explicit representation of the self as the agent of an action. (The intuition is pursued in more detail in the chapter by Roessler.) But of course this is a special case: we are talking about agents who may not even possess an explicit representation of the self, and whose actions are arguably not intentional in the full sense of the word. Suppose we focus on agents whose sense of ownership does involve an explicit representation of the self. A major issue arising from Marcel's discussion is whether such explicit awareness of ownership can be immersed in the sense of the second ingredient in Marcel's threefold characterization of immersion-namely, in the sense of absence of introspective awareness. One response is to deny the very coherence of a non-introspective sense of ownership. For example, if you are committed to a 'higher-order thought' theory of consciousness, then the very idea of a phenomenology of action (let alone
Mechanisms and Epistemology
45
a sense of owership) without reflective awareness of mental states will strike you as incoherent. (See Josef Perner's chapter.) It might also be argued that the problem is an artefact of a tendentious characterization of introspective awareness, implying that such awareness requires some active contribution on the part of the subject ('reflecting on one's mental states'). Given that picture, the argument might go, it is of course easy to elicit intuitions to the effect that a sense of ownership of action does not consitutively involve introspective awareness. But there may be alternative conceptions of intro~ spective awareness. For example, O'Shaughnessy argues that trying is a conscious event that causes knowledge of itself, in a way that is not mediated by any further event of the subject's noticing the trying. On this view, introspective awareness itself might be said to occupy the background of attention: you can be introspectively aware of your mental states even when wholly in1mersed in some activity. Both these responses insist that introspective awareness (of intentions, or tryings, or both) is a necessary condition of the sense of ownership of actions; so the second ingredient in Marcel's conception of immersion (absence of reflective consciousness) is inconsistent with a sense of ownership. An alternative option is to distinguish between two senses in which one can be said to be aware of an intention, and to claim that one of them is consistent with immersion. In one sense, being aware of an intention is a matter of being introspectively aware of having a certain psychological property, that of intending to do something. Being aware of an intention, in this sense, is to know the answer to a theoretical question such as 'what are my current intentions?' Alternatively, awareness of an intention may be construed as the kind of awareness you express when you answer the practical question 'What shall I do?' This latter question seems to call for an answer based on practical reasons, rather than for a report of introspective knowledge. Connectedly, it is natural to think of the awareness expressed when you answer that question as 'practical' in Anscombe's sense: it does not aim to fit the world (as factual knowledge does), but for the world to fit with it (as intentions do). (To insist that these are different questions is of course not to deny that they may be intimately related: see Moran (200 1) for an important discussion of the nature of their relation.) The distinction suggests a different orientation on Marcel's points about immersion. Being absorbed by an activity may be precisely a matter of having a practical, rather than a theoretical, awareness of the intention informing the action: the focus of your attention is on the question of what to do, rather than on what your psychological properties are. Both O'Brien's and Roessler's accounts of agents' awareness can be interpreted along these lines. On O'Brien's view, you know what you are doing in virtue of doing it with a sense of control, where this means that you act on the basis of an assessment of the options-that is, on the basis of an answer to a practical question. In Roessler's account, a crucial part is played by agents' use of perceptual experience to answer practical questions. Intentions acquired in this way, he argues, provide reasons for specific perceptual
46
Naomi Eilan and Johannes Roessler
expectations, which help to explain the sense of control that can be integral to Our perceptual awareness of intended effects. The distinction between a theoretical and practical awareness of intentions also bears on the background question, which Marcel's concern with immersion helps to throw into relief: in virtue of what does action awareness involve an awareness of the self at all? If we think of the sense of ownership as arising from introspective awareness of intentions, the self comes into play, primarily, qua owner of introspectively accessible mental states; an action is experienced as mine if it issues from an intention of which I am introspectively aware. On the other other hand, if we associate the sense of ownership with a practical awareness of intentions, we may locate the source of self-awareness in the first-personal character of practical deliberation, in the fact that in asking what to do we are asking 'what shall I do?' Given that answering the deliberative question seems to involve evaluating your practical reasons, there is, on this latter view, an essential normative ingredient in the sense of ownership. Crudely, to be aware of an action as one's own is to be aware of it as controlled by one's own practical reasons. (Compare O'Shaughnessy's remark that the answer to the question' "Which was the agency: me or my reasons?" , is this: ' "me and my reasons", i.e. "me, i.e. my reasons" , (1980: 341).) Now, Marcel, for one, emphasizes the normative aspect of the relation between agents and the actions for which they are (not just causally) responsible. But he considers that there is a tension between awareness of an action as one's own, in that sense, and what he calls non-detached immersion of the self in action. We will not pursue the matter here. What the point illustrates is how closely issues in the phenomenology of action are intertwined with problems concerning the (causal and rational) explanation of action.
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
GIBSON, J. J. (1979), The Ecological Approach to Visual Perception. Boston: Houghton Mifflin. GOLLWITZER, P. (1996), 'The volitional benefits of planning', in P. Gollwitzer and J. Bargh (eds.), The Psychology ofAction. New York and London: The Guilford Press. GOPNIK, A., and MELTZOFF, A. (1997), Words, Thoughts, and Theories. Cambridge, Mass.: MIT Press. HINTIKKA, J. (1974), 'Practical vs theoretical reason-An ambiguous legacy', in S. Korner (ed.), Practical Reason. Oxford: Blackwell. HUME, D. (1978), A Treatise of Human Nature, ed. L. A. Selby-Bigge. Oxford: Oxford University Press. JAMES, W. (1890a), The Principles ofPsychology, i. London: Macn1illan. --(1890b), The Principles ofPsychology, ii. London: Macmillan. JEANNEROD, M. (1997), The Cognitive Neuroscience ofAction. Oxford: Blackwell. KORSGAARD, C. (1996), 'Creating the kingdom of ends: reciprocity and responsibility in personal relations', in her Creating the Kingdom of Ends. Canlbridge: Cambridge University Press. LIBET, B. (1985), 'Unconscious cerebral initiative and the role of conscious will in voluntaryaction', Behavioural and Brain Sciences, 8: 529-66. McDoWELL, J. (1992), 'Meaning and intentionality in Wittgenstein's later philosophy', Midwest Studies in Philosophy, 17, 40-52. MILNER, A. D., and GOODALE, M. A. (1995), The Visual Brain in Action. Oxford: Oxford University Press. MONSELL, S. (1996), 'Control of mental processes', in V. Bruce (ed.), Unsolved Mysteries of the Mind. Hove: Erlbaum. MORAN. R. (2001), Authority and Estrangement. Princeton and Oxford: Princeton University Press. NEISSER, U. (1993), 'The self perceived', in U. Neisser (ed.), The Perceived Self. Cambridge: Cambridge University Press. NORMAN, D., and SHALLICE, T. (1986), 'Attention to action', in R. Davidson, G. Schwartz, and D. Shapiro (eds.), Consciousness and Self-Regulation. New York: Plenum. O'SHAUGHNESSY, B. (1980), The Will, ii. Cambridge: Cambridge University Press. PERNER, J. (1991), Understanding the Representational Mind. Cambridge, Mass.: MIT Press. REASON, J. (1984), 'Lapses of attention in everyday life', in R. Parasuraman and D. R. Davies (eds.), Varieties ofAttention. Orlando, Fla.: Academic Press. SEARLE, J. (1983), Intentionality. Cambridge: Cambridge University Press. SHOEMAKER, S. (1994), 'Self-reference and self-awareness', in Q. Cassam (ed.), Self-Knowledge. Oxford: Oxford University Press. SPERRY, R. W. (1950), 'Neural basis of the spontaneous optokinetic response produced by visual inversion', Journal of Comparative and Physiological Psychology, 43: 482-9. VON HOLST, E., and MITTELSTAEDT, H. (1950), Das Reafferenzprinzip (Wechselwirkungen zwischen Zentralnervensystem und Peripherie)', Naturwissenschaften, 37: 464-76.
2
The Sense of Agency: Awareness and Ownership of Action Anthony Marcel
It is perfectly true, as philosophers say, that life must be understood backwards. But they forget the other proposition, that it must be lived forwards. Kierkegaard (1843), in Dru (ed. and trans., 1938), The Journals ofSoren Kierkagaard The problem in that game is that she wasn't living with the ball. Virginia Wade, commentating on the Wimbledon Women's Tennis Finals, June 1998 My deeds upon my head. Shylock, The Merchant of Venice,
IV.
i, by William Shakespeare
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 responsibility, 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 clarify 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 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.
J
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 accompanied 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 assumptions. 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. 1.2 Kinds of Self Consciousness and Sense of Agency There are several senses and kinds of self consciousness.! Some are primarily awareness of oneself as a mental entity, others are primarily of oneself as physical. Some are long-term (of something persisting over time), others are more occurrent and present-tense. Some are detached awarenesses of oneself, others are more immediate or immersed. ,./ One form of bodily self consciousness is a sense of agency, of oneself as the agent of action. I will specify more what is meant by this below. The sense of agency with which I am primarily concerned here is of something physical, but the question that will raise itself throughout this chapter is the extent to which this is also of a mental self. By a physical self I mean an entity which exists in the J physical world and has physical effects via its physicality. By a mental self I refer to non-physical realization, for example, one's intentions experienced as one's own. If it turns out that experience of one's intentions plays a central part in one's experience of oneself as the agent of physical actions, then one's sense of agency is of oneself as at once a physical and mental entity. But the main point is that when I am aware of my actions and experience them as mine, I thereby experience myself: an experience of myself as agent. There is a further distinction between two kinds of a sense of self or self consciousness, and thereby between two kinds of a sense of agency. One kind of a sense of self is a long-term sense of oneself. One form of this is a sense of oneself as a distinct persisting entity in either the physical or social world, another is the core content of autobiographical memory. In the former, the self is an explicit experiential or conceptual object; in the latter, the self may be more implicit as the bearer of or the point-source of the remembered experiences which traverse time. Another kind of a sense of self is of something occurrent, for example something to which an experience is happening, to whom someone else is talking, the location of current bodily sensations. In this kind of sense of self, the self tends to be 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 consciousnesses 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 experience of oneself as the doer of whatever activity, but also to what Gibsonian psychology 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 manifests 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 awareness 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 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.
J
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? 1.3 Observational and Non-observational Knowledge: Are there Special kinds of Knowledge of the Self? It has been suggested by both philosophers and psychologists that there are dif-
/
/
I
ferent kinds of knowledge or knowing and that some of these are either specific to or much more a property of knowing certain aspects of oneself. Among the various meanings of 'special' kinds of knowledge, three usages are relevant here, about whose overlap one needs to be careful. First, one may mean a channel of information that informs about ourselves but not of the world or of others. Second, one may mean a kind of knowing that applies to oneself but not to others or to the world. Third, one may mean that there is a certain kind of knowing that is privileged, either in that it cannot be erroneous or in that it transparently carries with it all its relevant decompositions or causes, and that this privilege is a property of certain kinds of self knowledge. The term 'proprioception' is an example of the first usage. Unfortunately it has different senses and some of these are not restricted to self perception. It usually refers to all those channels of afferent nerves that derive from receptors sensitive to both the interior and the periphery of the body, as opposed to those channels that derive from receptors sensitive to distal stimulation. Yet some of those channels inform one both of one's body and of the outer world. Nerves responsive to mechanical pressure at or beneath the skin underlie the experiences of both tactile sensation and haptic perception, and which of these is the primary content of one's experience at a time is determined by attentional focus. Likewise for those nerves responsive to surface temperature. Complementarily, authors as diverse as Gibson (1979) and Merleau-Ponty (1961) have pointed out that we can have information about ourselves from those receptors and information channels normally thought of as exteroperceptive. Vision, touch, and hearing not only intrinsically inform one of aspects of oneself (e.g. egocentric self-location, possible actions) but can also be deployed to perceive ourselves from the outside, so to speak: that is, I can see and touch much of my body. None the less there exist bodily receptors that allow me to know of certain properties of myself that I cannot directly know of others, and certainly not by the same means. Thus kinaesthetic receptors primarily inform one of one's own movements. They only inform one of one's dancing partner's movements indirectly. (See Eilan et al., 1995, for a fuller discussion of proprioceptive awareness versus proprioceptive information.) The second usage of knowledge special to oneself necessarily overlaps with the first. My sensations, thoughts, intentions, and phenomenal experience are available from the inside only to me. But further, they are potentially available to me
The Sense ofAgency
53
in a way distinct from that of many aspects of the outside world. My knowledge of them can be non-inferential, and it can be non-observational. Just as Gibson (1979) pointed out that aspects of the world are perceptually available to us directly without need for inference, even more so features of my own conscious mental states are directly available to me (though not always explicit in their presentation and not always easily accessible or easily attended) in the sense that those of others are not. There are several usages of the term 'non-observational' knowledge. However, the one with which I am concerned lies in a contrast with the way knowledge is gained about the world through exteroperception. My knowledge that there is a desk before me is justified by my seeing or feeling it; and this can be one way that I know that my body has moved (through proprioceptive feedback or kinaesthesis). However, it has been proposed by O'Shaughnessy (1980) that I can know that I have acted without recourse to such perceptual channels, since they inform me about the objective world while 'what at any moment I am doing is, from my point of view, not a part of the world (which we all inhabit), but of my world ... ' (pp. 3-4; emphasis in original). O'Shaughnessy illustrates this vividly in a passage (pp. 31-2) also cited by Russell (1996): If one is to relate as observer to anything then one has to be 'without it', whereas if one is 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 observation .... 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 observation: 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.
r-
Such characterization is at least the start of a theory of the epistemology and / phenomenology of awareness of action, of felt agency, and of felt ownership of action. However, all I wish to do at this point is to suggest that this kind of knowledge is one kind of knowledge we can have of our action. For the moment, I envisage that we can have either kind of knowledge of our action without assuming one or other to be the norm and without necessarily endorsing O'Shaughnessy's assertion of the phenomenological consequences of the alternative. The point is that on one approach (or in one mode) we may ask what the relations are between awareness of our own actions and the sense of ownership of them, while on another approach (or in another mode) we take them theoretically as integral to each other. The third usage of the notion of special knowledge is knowledge that is privileged, in that it is transparent, necessarily veridical, or cannot be mistaken. It is tempting for obvious reasons to associate this characteristic with each of the other two senses of a special kind of knowledge. Indeed, it has been claimed that if one's knowledge of bodily posture is via proprioception one cannot be mistaken as to whose body one is informed about (Evans, 1982). However, from the point of view of psychology, it is far from obvious that proprioceptive self-knowledge is
54
Anthony Marcel
error-free or that non-observational knowledge is necessarily true; both can be seen as empirical matters. Even if our knowledge of our intentions were errorfree, this itself would not guarantee true knowledge of the ensuing actions, since actions are almost always more specific than intentions, are sensitive to occurrent physics, and are implemented by lower-level mechanisms. With regard to these special kinds of knowledge, proprioceptive awareness of one's action is what concerns us here. However, the extent to which such awareness is non-observational and to which it is veridical of the action itself are central questions. One aspect of this is whether ownership can come apart from awareness of action and the conditions under which it can do so, if any. Indeed my starting position, as will be seen below, is to attempt to treat awareness of and felt ownership of action as separate issues. 1.4 What is a Sense of Agency? Relationships among Components
J ""\I
By a sense of agency I mean a sense of oneself as an actor or a sense that actions are one's own. It is not just the sense of oneself as causal (which would imply separateness of self and actions), but the sense of oneself as the source of doings, apart from their willed or intended nature. This involves feeling responsible for actions that are one's own. Even if causal structure for an action is traced further back, where the buck stops is oneself. This is why one apologizes for treading on someone's toe even when unintended. This is at the heart of many legal issues. A problem arises at this point as to the appropriate starting-point. Given that we are concerned with awareness of action and with self consciousness, which of these should one take as the point of departure or the explanandum? Mainstream empirical psychology until quite recently has been generally unconcerned with self and self consciousness, though this is less true of social and clinical psychology. It is partly the integration of developmental and pathological issues that has returned self consciousness as a topic. In addition, for reasons deriving from the information-processing cybernetic tradition, mainstream psychology tacitly assumes all awareness as observational. (This is not surprising if issues of self and self consciousness are not addressed.) Although I shall later argue that it is mistaken, I shall start from a mainstream psychological approach, first focusing on awareness of action and then separately on ownership of action. It will help to restate the alternatives with regard to the relation between awareness of action, agency, and felt ownership. Given the definitions I offered of agency and action, to perceive an action as an action is to perceive it as agentive. However, (a) it is possible that to perceive agency of an action is separate from perception of who is the agent. Alternatively, (b), while (a) may be true for perception of the actions of others, it is possible that to perceive agency of one's own actions is to perceive their ownership. That is, for whatever reason, awareness of actions that happen to be one's own may carry with it, intrinsically or extrinsically, their
The Sense ofAgency
55
ownership. This may be due to the special kinds of knowledge indicated above that we have of ourselves, or to the content of the awareness. These conjectures concern (i) what is and is not separate from what else, and (ii) whether and to what extent self awareness is different in kind from awareness of other things, for example, whether it is (usually) non-observational. It is worth briefly mentioning alternative ways that psychologists consider in which perception of action, of agency, and of ownership, in self and others, go together or are separate. The intentional nature of an action or its agentive, autonomous nature can be a perceptible characteristic of the action or movement itself, rather than relying on attribution, and this can be perceptually available equivalently in the actions of oneself and of others. In addition to movement initiation unaccompanied by other spatio-temporally adjacent movement or impact, the dynamics of movement acceleration, trajectory, and termination themselves reflect agency and intentionality, and force-changes over time in a voluntary or controlled movement are different from those in reflex and mechanically caused movements. If such agency is visually perceptible in others' behaviour, presumably such a characteristic is just as perceptible proprioceptively in one's own behaviour, since the dynamic manner of movement is transmodally available. On this approach, there is nothing special to the perception in oneself of agency per se, and the identity of the agent may be carried by awareness of the sensory modality carrying the information: that is, if it is by vision then it can be oneself or another, if it is by kinaesthesis it must be oneself. This approach is in direct contrast to those which make primary for ownership the awareness of one's intentions themselves. Humphrey (1992) uses the term 'authorship' in discussing ownership of actions, body parts, and sensations. He sees this as starting from volitions and intentions as intrinsically owned and suggests that 'if such instructions are necessarily mine, it follows that the actions that / result from them are also necessarily mine' (p. 145). Note that according to. Humphrey self-ownership of actions is an inference from observing actions that result from (correspond to) intentions, of which one is aware. Ownership of/ action is not immediate but inferred, and this is done by observation. But to the extent that authorship and intention are experienced as separate from an action .f that is produced or experienced as produced, they are not part of the action itself. Indeed, authorship is a fitting term here, since an author is separate from the finished text she produces. Frith's (1992) position also links ownership to the comparison of feedback with intention, though the comparison of movement feedback with a copy of the intention or of predicted movement is apparently non-conscious, and it is only the result of the comparison which is conscious as ownership or disownership. Why a mismatch should not simply yield awareness of an unintended action or a failed intention is unclear. According to Frith (personal communication), a mismatch leads to disownership when individuals are
56
Anthony Marcel
unable subsequently to attend to their intention. This seems rather ad hoc and requiring reflection. The above illustrates positions among psychologists on what, among the phenomenological components in question, is intrinsic or extrinsic to what else. They tend to view ownership as inferential and observational. I want to relate this to the kinds of self consciousness laid out at the start. Jorge Luis Borges in his essay 'Borges and l' (1970) describes how the writing that he produces and the author of these public products (public in that they are part of the objective world) continually slip away from 'him', the present self. Once written, they are no longer part of his present sense of being. They are part of someone he knows to be himself but not part of someone he feels as himself. This is not a matter of memory in this Borges essay, but the experienced pastness of memory and of intention are also themselves distancing factors. The contrast between actions experienced as one's present self and Humphrey's and Frith's conceptions illustrates the distinction between kinds of self consciousness (immersed versus detached), and indicates two things that are meant by 'an "immediate" sense of agency or ownership': non-mediated and present to one. For the purposes of adducing empirical data, I shall approach self awareness as observational. But what should one conclude if awareness of one's own action (as agentive) and the sense of who the owner or agent is can come apart? This would mean either (a) that awareness of action is not non-observational, or not intrinsically so, or (b) that it is only observational in certain cases. If those cases are restricted to what we may reasonably count as pathological or involving a special attitude, then we are faced with the constant problem of what to infer from pathology, neurological or other: whether a psychological dissociation reveals a basic separation that is hidden by normal integrated functioning or whether it reflects an abnormal mode or some compensatory attempt to deal with a dysfunction. Since I shall be dealing with things that come apart, this question attends the whole of this chapter and will be returned to at the end.
1.5 What is the Content of Awareness of Action? The question 'what is the content of awareness of proprioceptively known actions?' focuses on whether awareness of an action performed by oneself is sufficient to give one the sense that it is one's own action, and if so, how. Many psychologists seem to assume that one is necessarily aware of the content of one's phenomenology, in its entirety and its exact form. However, this is far from certain. Dennett (1991), among others, has argued that, in several ways, we are not authoritative or incorrigible as to our conscious experience. This is particularly true when a fine grain of time is involved. In addition, quite simply the constraints of attention make it hard to be aware of all of one's phenomenology at a time.
The Sense ofAgency
57
I have argued more radically that there are a number of examples, both normal
and pathological, where people are unaware of their phenon1enology or are unable to be aware of it (Marcel, 1993). More simply, one may be generally aware of something without knowing exactly in what the experience consists, as many examples of emotion experience illustrate (Lambie and Marcel, 2002) I may be aware of my performing a certain action without being able to tell the exact forn1 the content of that awareness takes. Let us consider four possible relations between the awareness of the action and the experience of its ownership. Merely .introspecting will not tell which is (or are) the case. In fact, trying to attend ana1ytically to what you are aware of may well change that content. I assume that the immediate phenomenological content of awareness of an action is certainly not in linguistic terms. However, expressing various possible forms of such content in linguistic terms serves to clarify the distinctions. 1. As only an unowned gerund/participle (x-ing). The agent is not in the phenomenology at alL In such a case there are three ways that agency or ownership may be assigned. (a) Awareness of action is compared with awareness of intention and a match specifies action as meeting intention. Ownership of action relies on awareness and ownership of intention. (b) A copy of intention, of motor commands or predicted outcome is non-consciously compared with feedback from the action (von Holst and Mittelstaedt, 1950). How a positive match leads to experienced ownership of the action ren1ains to be specified and also seems to require awareness and ownership of intention. (c) One's only awareness of an agent is inferred, for example as cause. If no one else is perceived as agent one might experience oneself as agent as a default. 2. As a composed continuous present tense ('1 am x-ing'). What is meant by (composed' is the integration in a structure of different components, where the whole is decomposable. This in itself could take two forms.
(a) Additive, where there is an awareness of action and an awareness of the
agent of the action and the two are unified in awareness. For example: Awareness of x-ing (as an action) agent/owner/author)
+ Awareness of I (as
(b) Intrinsically Structural, where the slots within the structure are permanent.
This is like a sentence where what fills anyone slot is paradigmatic and the relation between slots is syntagmatic. Anyone slot in1plies others in specific relations, even if they remain unfilled. I x (present simple)/ I am x-ing (continuous present)
In such cases the agent is in the phenomenology and is in it as a separate entity from the action. However, this may be just a property of certain kinds of language. Whereas in English indicative moods of verb realization need explicit specification of subject, in Italian or Latin they do not (vado, cogito).
58
Anthony Marcel
3. As an 'acting as mine'-Agent as predicate, that is, as a participle with agent attached, for example explicitly 'x-ing-I' or implicitly 'x-ing (-1)'. In this case, the content of awareness is not propositional but a subject-predicate relationship. I use the terms here as adapted from linguistics by Barnard and May (1999) to indicate a property or relationship that is assigned to an entity. I am taking whatever is the focus of experience as the subject and whatever is assigned to it as the predicate. If the focus of experience is an action and the agent is a descriptor, the action is subject and the agent is predicate. In this case, although the subject cannot exist without a predicate, the value or identity of the predicate is unspecified by the subject itself. My analogy is perception of a red ball. A ball always has a colour but it does not have to be red. 4. Where a property of bodily action itself intrinsically specifies owner or agent; for example, egocentric content implicates point(s) of origin: '1 as source'. Awareness of action itself may specify the agent, though this may only be true of proprioceptive awareness of one's own actions. Perception of the action of someone else does not per se specify the agent. But in awareness of one's own action the content of the awareness may in some way specify that it is one's own. Any person can be a third-person, but indexically only one person can be a first-person. What is crucial could be either the way in which one is aware, for example, proprioception, or what is known thus, for instance, a property of action such as its putative perspectivalness or felt effort. This would seem to apply especially to when we are involved in action.
2. AWARENESS OF ACTION 2.1 What is an Action?
Before discussing empirical data on awareness of action some definition of action is needed, not only in order to be clear about what I am referring to and to delimit its scope, but also to guard against solving the problem by definition. In discussing action I shall be concerned almost exclusively with bodily action. I distinguish an action from a movement and from a mere 'reaction', and I do so in terms of several criteria which are themselves question-begging or insufficiently specified. (1) There is a degree of discreteness or categorialness. By analogy with speech, an action has 'lexicality'. It is not that it has semantics or is stored in an
internal 'lexicon', but that it can be segmented, even though it may flow smoothly into and out of other movements. (2) This is linked to the point that an action is under a description, since what the action is defines its end-points. But also, in taking a step forward, I may tread on someone else's foot. Yet a consequence or entailment of an action is not part of the action itself, which in the case above is 'taking a step forward'.
The Sense ofAgency
59
(3) An action has some goal, end-point, or effect. This is usually what defines or
individuates an action, though to a perceiver it is a matter of interpretation. Was the goalkeeper diving or falling? I am evading mentioning intention, though it will enter later. However, I distinguish the notion of an intention as something of which an action is a realization from the intentional notions of directedness and of content. Quite apart from being realizations, actions are defined as such by their directedness, having definable end-points. An action has content in that, whatever the movements or their specific character, they constitute a semantic category, even if their interpretation is arrlbiguous (manual exploration versus caressing). (4) There is some degree of voluntariness. This is necessary if one is to distinguish an action from a habit and from whether it is caused in the mechanical sense. Often postural adjustments as reactions to changes in force or optic flow approximate to a kind of reflex. One problem is how to count cases of reactions that are under voluntary control to the extent that one can inhibit them if one wishes. A movement or behaviour will be seen as an action to the extent that it is 'agentive', by which is meant that it is selfgenerated and at one's will. Even if someone else commands an action, one's compliance and execution of the action are ultimately a matter of one's own choice. This is connected to the criterion that what is done is not merely caused mechanically by some other entity. Whatever the antecedent events, the doing originates in the actor so that one might relate it to the antecedent events by reason rather than by cause. This will of course cover explanation by desire. The distinction I have in mind is captured by the perceptual discrimination people make when watching animations of geometric 'figures (Heider and Simmel, 1944) between an object moving passively apparently as the result of another object colliding with it and an object moving where there are no equivalent temporo-spatial conditions for its initiation. People see movement in the latter case to be 'selfgenerated' or autonomous: that is, as an action. These criteria increase ordinally in their sufficiency for a behaviour to be counted an action. Whether they go together is debatable. It is partly for this reason that I have left out intention as a defining criterion. Later I shall discuss the phenomena of Anarchic Hand and Utilization Behaviour, which Della Sala (2000) has described as actions without intentions. If what we describe as an action can be generated without what we count as an intention, the latter cannot be criterial for the former. The second, third, and fourth criteria above seem to implicate intention in the definition of action. However, intention can precede action, and appears empirically to do so even in those cases where there seen1S no temporal gap. Even if intentions are necessary causes or generators of what we count as actions, I do not count then1 as definitionally criterial since they are not part of
60
Anthony Marcel
the action itself. However, it may be thought that we know what action we are performing by reference to an intention. We therefore need to consider how aware we are of our intentions. 2.2 (Un)Awareness of Intentions
To what extent are we aware of our intentions? One reason for asking this question 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 psychologists 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 nonconscious 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 phonological (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. 2.3 Unawareness of 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 psychology 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 monitor 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 possibility 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. 2.3.1 Vibro-tactile Illusions of Limb Position
I will describe two experiments of mine. It is the second which is the most pertinent to awareness of action, but to understand it I will relate the first one and the background to them. If a muscle tendon at the joint of a limb is vibrated at certain frequencies (around 100Hz) it induces a reflex movement of the limb. Vibration of the biceps tendon at the elbow induces flexion and that of the triceps tendon induces extension. However, if the reflex-induced movement of the arm is prevented from occurring there occurs the illusion that the elbow is moving in the manner opposite to the reflex (extension for the biceps, flexion for the triceps) and there also occurs the illusion that the joint and the dependent limb segment (here the forearm and hand) is in the position it would occupy if the muscle was indeed stretched (Hagbarth and Eklund, 1966; Goodwin et al., 1972). Especially when you cannot see your stimulated arm your hand feels to be in a position very different from its actual position. The position it feels to be in can be shown either by report using a visual map or by mimicking of the felt position by the unperturbed arm. It is interesting that if, at the initial maximal point of the illusion, instead of asking for report of the location of the hand of the stimulated arm, one requires the person to grasp the wrist with their other hand, they do so correctly, and yet do not usually realize (immediately) the disparity. This was reported by Lackner and Taublieb (1983), but only as part of the method for examining the effect of grasp on the illusion, and they apparently failed to see the potential importance of this dissociation: the location of the body part which is the target for grasp accurately controls the grasp movement yet is different from the consciously felt location which at that moment fails to control the movement even though one supposes that it entered into the intention! My colleagues and I have explored this phenomenon, to investigate why matching the felt position with the other arm reflects the illusion but prehension does not. We carried out the experiment exactly as before (with the stimulated arm visually occluded), but at various times after onset of vibration we presented one of three signals which required the subject (a) to report the felt position of the wrist of the stimulated arm by reference to a visual map of pericorporal space, or (b) to grasp it
The Sense ofAgency
63
110
-+-
100 c:::: 0
'';:;
'Vi
0 0-
~
~
-0-
90
Reported Pointing Grasping
80 70
E
60
V'l
50
~
-0-
OJ
~
40
C'O
20
.01 OJ "0 OJ 01
OJ
~
30 10 0 -10 -7 -6 -5 -4 -3 -2 -1 -seconds
0
2
Point of initial maximal illusion
3
4
5
6
7
8
9
10
+seconds
Time preceding/following initial maximal illusion FIGURE
2.1. Results of first vibro-tactile experiment
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 individuals, 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
Subject draws intended/performed movement \
..
!(
"
·tJl
. .
, ,
,':
..
Real position
FIGURE
'\
Illusory experienced position
2.2. Experimental set-up in second vibro-tactile experiment
Note: LEDs: light-emitting diodes.
The Sense ofAgency
65
immobile or released to allow n10vement of the forearm in an arc. The arm is occluded by a horizontal surface, on which are a number of lights in an arc that traces the possible trajectory of the hand beneath. The arm is passively positioned while the subject is attending to movement and location of their other arn1 such that they do not have a good awareness of where the relevant arm is, since joint position information is carried mainly by dynamic receptors and decays after a few seconds. The angle of the elbow and whether the biceps or triceps is stimulated is varied. Also, on some trials, the vibrators are applied such that there is no illusory movement, to provide controls and catch trials. After stimulation, individually adjusted for frequency and duration, one of the lights goes on and the person, in different conditions, either (a) has to draw with their other, free, hand the movement they will have to make to get their hand under the light and then, at a further signal (at which the armrest splint is released), move their hand to under the light, or (b) move their hand to under the light and then immediately draw the movement they have just made, or (c) do all three. Incidentally, under normal conditions it is easy to move your hand to a visually n1arked location when you cannot see the hand. The pre-movement drawing was intended to reflect conscious intention, that done after the movement to reflect awareness of the movement. If one simply asked each person to do all three on each trial, the latter drawing might simply be influenced by or be a copy of the first drawing. As it turned out, the relevant results for the third condition are almost identical to those in the other two. The most interesting trials are for those central targets when the hand is to one side of the target light but feels to be on the other side, since on those trials it is not just extent but direction of movement which differs for real and illusory pre-movement hand location. Table 2.1 shows a dramatic result. First, the drawn pre-estimate of the movement, which we take to reflect the intention, always follows the illusion; that is, it is in the wrong direction and of the extent appropriate to the felt location. Second,
2.1. Effect of illusion on movement estimate, movement, and awareness of movement on critical trials in vibro-tactile experiment 2: Percentage of subjects (a) drawing necessary movement, (b) showing actual movement, and (c) drawing movement made, according to the postural illusion or correctly, on trials when target location is between real and felt locations
TABLE
Intended movement (drawn before)
According to illusion (%)
Correct (0/0)
90-100
0-10
Actual movement
o
100
Report of movement (drawn after)
60-70
40-30
66
Anthony Marcel
by contrast, the actual movement of the hand is always accurate in direction (in 2 out of 30 people on about 3% of trials it was not) and almost always accurate within normal limits in extent. Third, immediately after moving, between 60 and 70 per cent of individuals draw their movement in exactly the same direction as their predrawn estimate/intention on equivalent trials without noticing at all that they had moved in the opposite direction, while the rest, between 30 and 40 per cent draw their just-made movement correctly. I say between two percentages because as one adds more subjects the proportion varies up and down slightly. All of the latter subjects commented in one or another way that there was something peculiar, or that while they had intended to move in one direction they had the impression that they had moved in the other. After the main experiment, all those who were largely unaware of their actual movement were told about their behaviour and were again run in a short version of the experiment. Half of them still showed no improvement. That is, they continued to experience the direction of their voluntary movement wrongly and it must have carried such conviction that they were unable to adapt their reports of it even though they knew about the illusion. There are several messages to take from this study. First, if pre-trial drawings of the necessary movement validly reflect conscious intention, then conscious intention is based on felt limb location. Second, the implementation of movement is based on target location and the real position of the limb. Third, while some people have good conscious access to their actual movement, the majority have poor or no conscious access to it. Some of the latter, even when they understand their behaviour, find it very difficult or impossible to gain such access. For them, their actual movements implementing voluntary actions seem 'cognitively encapsulated', that is, they have no direct conscious access to them. The question arises as to what is the basis of awareness of one's movement. It is implausible that it is information from the movement itself, since there is no reason to suppose that any feature of the movement phase produced the illusion. Likewise it is unlikely to be an inference from the felt initial position, because when people were correct they mostly remarked on the disparity between intention and movement. Presumably then, the basis of awareness of action is or involves either intention itself or some specification for movement prior to movement itself. Christopher Peacocke (personal communication) has pointed out that on each trial what was required of the subjects and thus the primary focus of their intention was to put their hand in a location: the location specified by the light. Therefore direction, and even perhaps extent, of movement was not their principal concern or focus of attention. This is valid. In the vast majority of normal actions the intention is framed at a higher level of generality than the movement itself. Since actions of necessity exceed in specificity the intentions that they realize, they will inevitably have details unenvisaged in the intentions. Exceptions to this are actions where great care or self-monitoring is needed, often the case in
The Sense ofAgency
67
psychology experiments. Since, in this particular case, on every trial at some point subjects had to draw the necessary to-be-performed or just-performed movement, they are unlikely to have ignored direction altogether, unless one cannot attend to both location and direction. If the apparent unawareness of movement direction were a matter of limited focus of attention (on target location or on movement direction), then one would also expect that when informed of their behaviour all subjects would have been able in the follow-up session to improve their post-movement drawn reports to at least some extent. As stated, many could not. However much detail of one's actions one takes responsibility for post hoc, this is quite separate from and does not entail that one is directly aware of such details at the time.
2.3.2 Other Cases of Unawareness ofAction One n1ight think that the lack of proprioceptive awareness (or of accurate awareness) of actual n10vement under the conditions of the vibro-tactile illusion may be due to the unnatural proprioceptive situation, for example, of proprioceptive conflict. But other data make this unlikely. These come from another interesting example of proprioceptive unawareness of the details of our actions. I have recently reported (Marcel, 1998) that when GY, a much-researched subject with blindsight, attempts to grasp objects of differing shape, size, and orientation in his blind field, his hand and wrist are appropriately pre-shaped prior to final grip closure on the object, indicating non-conscious visual processing of the relevant attributes. We wanted to see if he could infer the shape of the object from the shape of his hand just before the grasp closure. This was a question raised by Dan Dennett (personal communication). We carried out the original experiment where he had to reach for and grasp differently shaped objects in his blind field, but (a) signalled him to stop when his hand was shaped, and (b) the object was removed. The results were that if GY is prevented from actually grasping the object at the last moment but when his hand is shaped appropriately, he cannot tell what shape his hand is in. Although neither GY nor the situation are quite normal, there is no proprioceptive illusion here. This suggests that we may have very poor direct proprioceptive awareness of the spatial arrangement of a limb or effector during an action. Obviously it is extraordinarily difficult to carry out the equivalent experiment with normally sighted people. (Unless they are prevented from seeing the target object consciously, they can infer hand shape from object shape.) The reason that this is interesting is that if we try to recognize haptically an unseen object which the hand encompasses, we are fully aware of the shape of the grasped object. And we are aware of object shape by dint of the proprioceptive information from the hand as to manual disposition. It is tempting to conclude that in action our awareness is focused on the distal object or world rather than on the details themselves of our own action, and that it is extremely difficult
68
Anthony Marcel
if not impossible to become aware of the latter. This may be a matter of individual difference or training, as the vibro-tactile results suggest. In the vibro-tactile experiment, people did have an intention which was conscious in the sense that they could draw the intended movement. In GY's case, he claims that he has no specific conscious intention in terms of a particular shape, size, orientation, or location of object, only to reach and grasp in an unspecified way. In the former case, subjects were aware of their action but mistakenly; in the latter case, there was no awareness of hand shape. Could it then be that any proprioceptive awareness of action that we have is more closely tied to our intention than to real feedback from movement? The answer to this will playa central role below. A further example of unawareness of current action, in this case inaction, can be interpreted to suggest that one's intention can certainly dominate any kinaesthetic. feedback. This comes from people who become hemiplegic, usually after stroke, and are anosognosic for, unaware of, their paralysis. A large subset of such patients, when asked to move their plegic arm or leg are concurrently unaware of their failure to move, and without any evidence of hallucinatory experience of movement (Marcel and Tegner, 1995; Marcel et aI., forthcoming). When blindfolded and asked to raise their plegic arm they happily tell you that it is at shoulder height when it has not moved at all. Both in the cases that we investigated and as reported by Bisiach and Geminiani (1991), this unawareness of movement failure at the same time as the attempt to move can occur with preserved kinaesthesis and preserved monitoring ability. Bisiach and Geminiani (1991) point out that anosognosic plegic patients believe they have moved their paralysed limb 'even when the pathways that would carry kinaesthetic information are spared and the patient can perceive passive movements'. It may be that such patients are unable to attend to the proprioception (or lack of it) from the plegic limb. Indeed, this interpretation is supported by the effects of neck vibration or of caloric stimulation of the vestibular system on orientation of attention and consequent temporary remission of the anosognosia. If so, the patients are left only with their intention and/or effort. It is relevant to present concerns that a number of such patients disown their plegic limb, and it is possible that whatever underlies lack of ownership of the limb, even where full disownership is not manifested, underlies lack of awareness of movement failure rather than the reverse. Taken together, the three phenomena of unawareness of action discussed here suggest that something close to intention for action might be the basis for apparent awareness of action itself. 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 movement 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 direction 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 vibrotactile 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 parabolic 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 learning 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 agreement 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 distortion). 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 reaction 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 effectors. 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, proposing 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 specifications n1ay underlie initial awareness of action, we are not necessarily aware of them as such: that is, not under that description. 2.5 Candidate Constituents of Awareness of Action
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 proprioceptive 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
2.2. Necessity and sufficiency of possible constituents of awareness of action: summary of the arguments and evidence
TABLE
Awareness of
Necessary/Sufficient?
Relevant evidence
Proprioception
Necessary? No Sufficient? No
IW-no proprioceptive awareness Anosognosia for Plegia
Intention
Necessary? No Sufficient? Perhaps
Absent-mindedness, etc. Anosognosia for Plegia
Effort
Necessary? No Sufficient? Yes
Effortless actions Unintended responses to force changes, etc.
Motor Specifications Necessary? (not experienced Sufficient? Yes as such) Conclusion
Libet and Haggard experiments (TMS)initial awareness of action covaries with pre-motor effector selection
No singular answer evident
Nor can proprioceptive awareness be sufficient for us to experience action. First, both reflexes and passive movement are proprioceptively experienced but not experienced as actions. Second, among patients who are anosognosic for hemiplegia, a subset are unaware of their failure to move when they attempt an action even with preserved proprioception and kinaesthesis (Bisiach and Geminiani, 1991; Marcel and Tegner, 1995; Marcel et al., forthcoming), despite Levine's (1990) claim that proprioceptive loss is a necessary component. Incidentally, proprioceptive awareness may be sufficient for ownership of movement, since we do experience reflexes and passive movements as our own. 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 occasions 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 thalamic 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 intention to do so, if not contradicted (at least not consciously contradicted) by feedback 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 voluntary 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.
Experiences associated with Motor Specifications
Libet's finding (Libet et al., 1983; see Libet, 1993, for review) that people's awareness of a voluntary action precedes the movement itself by up to 40 to 80 ms has led to claims that what people are aware of in awareness of an action is the motor command rather than feedback or any peripheral information. However, Jonathan Cole in an unpublished study (personal communication) magnetically stimulated the motor cortex of IW who has no proprioceptive awareness. This produced movements, but IW was completely unaware that he had moved. Therefore, either one does need feedback for awareness of movement, since IW lacks awareness of motor feedback, or the content of movement awareness is prior to motor commands themselves. Haggard's data discussed earlier is entirely consistent with the latter view. His own conclusion is that the relevant stage underlying initial awareness of action is that of motor specifications. By motor specifications I am referring to the preparation and organization of forthcoming movements referred to by Haggard and Magno (1999) which they associated with the Supplementary Motor Area and which Haggard and Eimer associate with the Lateralized Readiness Potential and the selection of effectors. This is downstream of intention but prior to activation of primary motor cortex controlling the movements themselves. I prefer to refer to experience associated with such activity rather than experience of such activity, because I doubt that in such experience we are aware of the specifications or commands as such or under that description. I doubt it because while it is common parlance to speak of awareness of effort, intention, and movement, awareness of motor specifications or commands is never spoken of by lay people. The nature of such putative experience I will try to suggest later. I am reticent to comment on the necessity of such experience for awareness of action because I am unaware of any relevant data. However, Haggard's data do suggest that such experience is sufficient for awareness of action. Subjects' initial awareness of their actions was tightly tied to onset of the associated LRP and was delayed when TMS was applied over the relevant area. As I said before, Haggard's data only relate to initial awareness of action. Therefore other factors may contribute to non-initial awareness. Effort is an obvious candidate. But so is proprioception or kinaesthesis. My argument against its sufficiency was based on its absence in patients anosognosic for plegia failing to
The Sense ofAgency
75
serve as awareness of the absence of moven1ent. But this argument is weak for two reasons. First, it is an argument from awareness of absence of movement, which may not play the same role as awareness of presence of movement. Second, the patients are pathological, and it may be that proprioceptive awareness in them fails to override other compelling evidence that they have moved as intended. The reason for rehearsing these points is that it may be that awareness of action can depend on more than one source, even if one source is more predominant or more usual. It may also be that one kind of source is more related than others to ownership of actions. Indeed, it is worth exploring whether consideration of experienced ownership of action bears on this. We now turn to this.
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 ownership 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 anecdote 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 circumstance 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 voluntary and involuntary and between intended and unintended. Indeed, 'lynch mob' induction is not automatic: being in a critical or detached attitude reduces its likelihood 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 disown 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 distinguish, 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 description, 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 phenomenon 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 disowned actions are unintended or seem so to the patient, and it is this that gets confused 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 weakening 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 question 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 different 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 occasionally 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 anarchic 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 phenomenon 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 phenomenon 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 sufferer 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 phenomenology 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 mention 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 necessarily 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 control 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 apparent 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 awareness 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 stimulation 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 temporary 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 certain 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 individual 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 schizophrenics 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 disownership 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 disownership 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 disownership 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 theoretically 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 internal 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 supposition has been raised to axiomatic status by some authors (Shoemaker, 1968; Evans, 1982) and termed 'immunity to error through misidentification relative to 3
I am grateful to Chris Frith for discussion of this.
The Sense ofAgency
81
the first-person pronoun'. The thesis is essentially that any knowledge about properties that is gained in certain ways cannot be known to apply to an individual without ipso facto knowing that the properties apply to oneself. While for Shoemaker this holds only for introspectively gained knowledge about mental states, Evans argued that there are ways of acquiring knowledge about our physical properties that are immune to error in the same way, one being through internal proprioception. Since patients' awareness of the anarchic action need not be via external visual or haptic perception, it would seem that Anarchic Hand disproves this thesis.
4. IS THERE A DISTINCTIVE PHENOMENOLOGY
OF AGENCY AS OWNERSHIP OF ACTION? A MINIMALIST HYPOTHESIS 4.1 A Distinctive Phenon1enology?
Is there a distinct, positive phenomenology of ownership of action or of agency? I return here to the question of the content of awareness of action. The issue is not the same as what it is to have a sense of agency. It is whether there is something distinct and explicit in one's immediate phenomenology. One can ask the former question and still ask in what does it consist. It seems plausible that in infancy there is something phenomenally distinct. When an infant for the first time accidentally touches and sets in 1110tion a mobile hanging above its cot, usually its attention is engaged, it shows pleasure, and it attempts to repeat the occurrence (Watson and Ramey, 1987). Unfortunately, this occurs long before infants have the linguistic and episodic memory capacities to relate the experience in so far as they have one. One presumes that what is experienced as new is the infant's causal effectiveness, though under what description it is experienced and whether it is experienced as its own causal effectiveness are open questions. None the less it is plausible that something is distinct about such an experience. However, it is not at all clear that after infancy this is explicitly experienced, unless it fails. We may become so used to our effectiveness that it becomes perceptually recessive such that we do not notice it (just as that to which we become habituated gains attention only by its absence), or perhaps there simply is nothing extra to awareness of action that could be noticed, other than by inference or reflection. In Anarchic Hand there is a positive otherness of the anarchic actions. To be sure, this is partly experienced in the alien will implicit in the difficulty of suppression of the anarchic action. But apart from and prior to this, patients experience the action itself as alien. The patient I saw was aware of the anarchic actions performed by his left hand even when out of his sight, and before attempting any
82
Anthony Marcel
suppression he said 'There it goes again'. But he felt as if someone else was doing the actions, probably due to their wilfulness and directedness. He did not experience the actions as unowned. Indeed, I know of no cases of an action merely being experienced as unowned (though movements may be so experienced). If there is no positive phenomenology of ownership or agency, is selfownership experienced by default? This is to posit that self-ownership is assumed, unless there is an experience of alienness or a proprietorial dispute. For the vast majority of actions this is unlikely, since one is never usually in any doubt as to ownership of one's own actions. However, it is not always the case and cannot be rejected out of hand. Two kinds of instance come to mind. In joint coordinate action where the forces from the individuals are sufficiently matched to eliminate resistance, such as handshaking, it is often phenomenologically unclear what is one's own contribution or that there is a separate contribution from oneself. If one does not actually move one's hand but merely allows the other to do the shaking, the absence of one's contribution then becomes clear. In actions that meet with resistance, the experience of effort gives one a sense of one's own bodily self as physically causal. (This contribution of experienced effort to ownership is separate from its role in awareness of action itself.) The other kind of instance of unclear ownership is where one performs actions in a state known as a 'flow experience' (Csikszentmihalyi, 1978). In activities where one has some level of skill and when one is immersed and focused, it is sometimes the case that one has the experience that one is not doing the actions oneself but that 'one is being done by the activity'. This is what is described in Zen practices (Herrigle, 1988). In such states there is an experiential indeterminacy as to the source of the actions. There is little data on this in the conventional psychological literature, but I myself have certainly experienced this when playing tennis, skiing, and performing on stage. There seem to be three relevant conditions. A minimal level of skill or practice or anticipation is required, such that otherwise separate actions lead smoothly into one another. One adjusts so that one takes advantage of rather than overcomes resistances. One is immersed in rather than reflective or detached about the activity. If, apart from these kinds of case, there is a distinctive phenomenology of ownership of action, in what does it lie? Even if felt effort can contribute, it cari be discounted as necessary for two reasons. First, the same reasons that felt effort is not necessary for awareness of action (cases of effectively zero resistance), apply to ownership. Second, in Haggard's experiments on initial awareness of action (see earlier) the actions were experienced as the subjects' own but the experience was derived from a point before any stage of action that could produce experience of effort. Russell (1996) proposes that experience of ownership of actions lies in the experience of bodily control (p. 175). Since Russell is not specific, this can be explicated in several ways, none of which is satisfactory. First, in one reading it is circular. If he is saying that I experience my bodily movements as under my (voluntary) control, it begs the question of the phenomenology of 'my bodily control'
The Sense ofAgency
83
and how the actions are linked to it. As an alternative for the experience of bodily control, he gives 'the experience of trying to do things that usually come off but which may not'. If this version brings in the awareness of intention as explanatory, it is problematic since awareness of intention has been shown above to be unnecessary for ownership of action (though it might be sufficient). Another reading of Russell's proposal rests on a subpersonal cybernetic comparator mechanism (von Holst and Mittelstaedt, 1950). At one level of such a mechanism, intention or predicted end-state is compared to perceived end-state; at another, a copy of motor commands (efference copy) is compared to kinaesthetic feedback. But whenever such a mechanism is drawn on to explain something (e.g. the experienced stability of the world in the course of eye movements) the mechanism and component representations are assumed to be non-conscious and subpersonaL In all cases, but especially in the case of the experience of ownership, it is quite unclear how and why the operation of the mechanism generates the conscious experience. If there is a mismatch, such a process could underlie an orientation response or an experience of untowardness, or if there is a match, it could underlie the lack of any experience of untowardness. But it is unclear how it explains any positive experience of ownership or control of action. To the extent that Russell elides or fails to notice this problem, he is far from alone. In fact, he highlights the problen1. He writes (p. 85) 'action-monitoring is ... a sub-personal mechanisn1. In this section, however, we are located squarely at the personal level, at the level of "what it is like to be an agent"'. It is possible that there is no single or essential phenomenological content in which lies experience of agency or of ownership of action. If so, it may be because the experience of agency or ownership of action is derived from and smoothly applicable only to clearly distinguishable cases: a privately premeditated act versus being pushed. Humphrey (1983) points out that our limited set of concepts are derived from and apply to certain distinctions but often have to be applied to cases which do not fit these. When people say or deny 'I did it' the meaning comes from clear referents, but they are often trying to describe new experiences unfitted to the originals. This may be true in the example of 'contagion' but particularly in acquired pathologies. The experiences of Anarchic Hand, schizophrenia, and brain stimulation, an10ng others, probably do not fit the cases to which our language normally applies, and we should not necessarily take the meaning of phenomenal descriptions as stable. Related to this, Kirsch and Lynn (1999) review data showing that one can easily influence the experience of oneself or another as the source of an action, and that many ideomotor responses and automatisms are compellingly attributed to external agents. They suggest that the experience of voli~ion or agency is the result of an attribution based on situational cues, cultural beliefs, and consistency of behaviour with goals. However, before conceding to a pluralist view, there is a candidate for an essential content which is derived from what I have reviewed so far and which I am not aware of having been proposed before.
84
Anthony Marcel
4.2 Perspectivalness of Source as Minimal Ownership of Action There are three things that have been mentioned that, taken together, are suggestive. The first is that awareness of an action does not necessarily derive from the action itself. Second, Haggard's data suggest (a) that initial awareness of action is underlain by specifications for bodily action that follow intention but precede action itself, and (b) that the Supplementary Motor Area plays some role in this. Third, in the one strong case of disownership of action itself, Anarchic Hand, the common location of the lesion site, to the best of our present technical knowledge, is the Supplementary Motor Area. We might use this neuroanatomical coincidence as converging evidence. One other finding can also be put in place at this point. If one stimulates the Supplementary Motor Area either magnetically or intracranially (the latter before neurosurgery without general anaesthetic), people report urges and attempts to move, but do not necessarily show movement itself or any awareness (even illusory) of movement (Fried et aI., 1991). The specifications for bodily action that follow after intention have one aspect that might itself specify ownership in a certain way. In so far as motor specifications are in spatial terms, they must entail coordinates, even if only implicitly. It is logically necessary that such specifications are in egocentric coordinates, if we allow points on the body to be considered as points of origin for a spatial frame of reference. The starting-points of the body parts that are to implement the action must be specified in a common reference frame with the targets, so that trajectories or movement accelerations and decelerations are produced. The targets of such movements must also be specified with respect to the starting-points of the relevant body parts or effectors. This is true even if final adjustments are achieved by closed loop monitoring of the movement. The only spatial description common for all body parts and for external locations is an egocentric one. Given that we can confidently point at a seen location and can do so without visually monitoring our hand, points of origin (eyes, shoulder, hand) must be trigonometrically transformable one to another. In this sense the 'ego' of an 'egocentric' reference frame is whatever is the body part, and its indexical location, that stands in an intentional relation to a target, and this can be at multiple points simultaneously. This would give, in the normal phenomenology, a perspectivalness of the source of the action, that is, spatial points of origin and intention. Such points of origin are the motor equivalent of the perspectival point of origin in perception. Indeed in perception there cannot be a single point-source, as in the supposed cyclopean eye, since one supposes the same perspectivalness to hold for haptic perception as it does for vision. And when we look at an object which we are feeling manually, since we have a unified experience of the visual and haptic percepts, the location perceived by each modality must be the same, and therefore in a common frame of reference. To return to action, one supposes that this sense of the source of an action is implicit rather than explicit in the phenomenal content and is perceptually
The Sense ofAgency
85
recessive, just as is the perceiver implicitly specified by the perspectivalness of visual content. That is, in immediate phenomenology the actor is a spatial source, not a substantive or conceptual entity. The actor is substantive in this sense only as a body that has extension and whose articulation can be altered with physical forces. This would seem to be a minimal (and physical) sense of ownership of action. I suppose that the perspectival content of action specifications is reflected in the corresponding phenomenology that one experiences as action itself. It is important to note that in this account the content itself of awareness of action specifies ownership of the action (where it is self-ownership). Nothing in this proposal is at odds with .the contribution of later aspects of action itself to awareness of action. But it emphasizes what is a primary and minimal content of such awareness, and one that is close to what makes something an action, its source. The phenomenology of Anarchic Hand must also be dealt with by this account. In Anarchic Hand, the action itself is experienced from the inside via proprioception. But any effort is experienced only from the outside as resistance to attempts to subdue the action with the unaffected hand or other body parts in contact with the anarchic hand. In whatever way the anarchic actions are generated, they bypass the Supplementary Motor Area. One supposes that in lesions to the Supplementary Motor Area the patient is deprived of the phenomenology of this source of action, which is the normal phenomenology of intentional actions. Though the patient may experience perspectivalness of the perceptual content of proprioception, he or she will lack the perspectivalness of the source of the action. This account is also consistent with features of anosognosia for plegia. In failed attempts to move the plegic arm, patients in whom proprioception is intact appear to experience the intended but unperformed movement. In many such cases, the experience or the unawareness of failure seems genuine, whatever other symptoms are shown. Presumably, the experience of such patients is underlain by the specifications for the movement rather than by the intention. Whether any positive movement hallucinations in such patients are underlain by such specifications or by motor commands uncancelled by feedback is unanswered at this time. If such specifications are for movement and if they are to playa role in explaining the results of the vibro-tactile experiments, how is it that in those experiments the actual movements were in what might appear to be the opposite direction to such specifications? There is good reason to suppose that, whether they select and prepare effectors for movement only in terms of end-point locations or in terms of direction, the actual implementation of goal-directed movements is modulated at a lower level. Much evidence leads us to believe that the precise way in which intended endestates are achieved is determined by low-level constraints and mechanisms without executive intervention or knowledge (Turvey, 1977; Kelso, 1995). Since this applies to trajectories and forces, it may well apply also to direction. This in itself would account for the movement direction. Concerning the erroneous awareness of the movement direction, first, it is difficult to gain access
86
Anthony Marcel
to such low-level mechanisms and information; and second, in such brief actions it is difficult to switch attention fast enough from the task of making the movement with regard to the target to the task of monitoring its lower-level directional implementation. Three points can be made about the present hypothesis. First, consider situations where the type of proximal bodily movement is qualitatively different from the distal desired movement. This exists in situations where the person uses button presses to control a device which produces or displays spatial changes or vice versa, as occurs in some computer games and industrial machines. 4 Where there is a very small temporal lag, presumably the intention is under the description of the distal end-state but the ownership of the action is still underlain by the motor specifications. This issue pertains to many situations and must involve also all those parameters relevant to causal attribution, including temporal relations, consistency, and so on. Second, the present account clearly does not apply to ownership of non-bodily actions, such as mental actions: shifting attention, planning, thinking of something, solving a problem. It might be desirable to have a common account for ownership of all kinds of action, but I see little reason why it should necessarily be so. Indeed, this is relevant to the third point, which is that there are various things to which the experience of ownership applies and where what underlies that experience differs. Sensations, body parts, intentions, thoughts, as well as actions, are all usually experienced as owned. In each case, they appear to be capable of being experienced as disowned. (In the case of thoughts, it is unclear the extent to which schizophrenics disown thoughts themselves or their source. In asserting the insertion of a thought they are not denying that they themselves are having it.) But in each case, the situation under which disownership occurs is different. And in each case, what it means to say that it is owned can be seen to be different. 'Being a part of' may apply to body parts, while 'being the state of' or 'being appropriately located' may apply more to sensations, and 'being the author of' may apply more to thoughts or intentions. Adequate discussion of this topic is beyond the present scope. All that is relevant is to point out that ownership in the experience of agency is not necessarily underlain in the same way nor does it necessarily amount to the same thing as in other cases of felt ownership. The question posed at the outset was whether the content of awareness of action specifies its ownership, and if so in what way. The answer I have proposed is that in the case of one's own action, a property of the phenomenology of bodily action itself intrinsically specifies the owner or agent. That property is the egocentric points of origin of the action specifications for an egocentric space. This proposal does not rule out other sources of experienced ownership or
4
The relevance of this situation was brought to my attention by Bob Kentridge.
The Sense ofAgency
87
agency. Indeed, I suggest that other sources mentioned also playa part. But it does seem to be not only the initial source in an action but also the simplest both in terms of conception and in terms of what is required of the organism. 4.3 Perspectival Source as Ownership and Non-observational Knowledge We can now return to a couple of related issues raised in the Introduction with regard to the issue of special kinds of knowledge and their relation to awareness of action. In raising the idea of non-observational knowledge of one's own actions, a passage from O'Shaughnessy was quoted, in which he characterized nonobservational knowledge of action as one's being 'within' the action. The suggestion that felt ownership of action lies principally in awareness of the perspectival source of action puts the owner within the action, or at least within the content of the immediate phenomenology of action, and is a non-observational conception of awareness of action, where the felt owner is not separate from the action as one is aware of it. In this way, phenomenologically one can both be aware of an action and be a part of it. Indeed, the empirical inaccuracies and lack of veridical awareness of action itself discussed here go exactly with a non-observational knowledge of one's own action, since although ownership is carried as the source of the action, one does not normally observe the details of such action itself. It is not just that Haggard's data suggest that initial awareness of action precedes action itself, but that awareness of what follows of the action itself need not be observational, especially when one is attentionally immersed in, rather than detached from, what one is doing. (Experiments on this would need to be restricted to single trials per subject, since once one has asked what the subject knows of an action this puts him into a different relation to his actions.) But what this emphasizes is that while normal awareness of one's action may be special in being non-observational, it is not privileged in the sense of being immune to error. To respond to another point, what I have proposed about Anarchic Hand is that the pathological condition makes the person an observer of their own action. What one feels from the inside in Anarchic Hand is kinaesthetic feedback from unwanted and unintended movement; one has no awareness of the movement specifications that carry perspectival source. The agency of the action is only felt from the outside, as its wilfulness. This suggests that one only has observational knowledge of one's own actions in particular states. In the pathological case, it is due to a restriction caused by removal of normal control. In the non-pathological case(s), it is by adoption of a certain attentional attitude, namely, by taking a detached stance in inspecting one's proprioceptive feedback. Anarchic Hand does not 'reveal nature cut at its joints', but rather puts the actor in a different, 'abnormal', relation to his anarchic actions.
88
Anthony Marcel 5. THE LONG- TERM SENSE OF AGENCY AS BACKGROUND
At the outset two kinds of an immersed sense of agency were distinguished, an immediate, occurrent sense of being the agent of an action or of acting and a long-term sense of oneself as an agent. What I have discussed in the body of this chapter is the immediate occurrent sense of agency. However, I wish to suggest that it could not take its full phenomenological form unless it relied on the longterm sense of agency. In order to say why and how this is so, it is necessary to give some indication of what that long-term sense consists in, even if quite briefly. My supposition is that the long-term sense of agency consists in several aspects. Some of these aspects are only usually revealed if they are negated. One of these is to take oneself as a source of effect. There is a difference between experiencing changes in the world as brought about by 'this body' and changes as brought about by 'me: Changes brought about by 'this body' not only pertain to unintended effects (accidentally knocking over an object), but are a condition of spatial existence itself. Changes brought about by 'me' apply to intentional action. To experience oneself as an effective willing agent requires the correlated awareness of effect, intention, and appropriate effort. In addition, changes in my body alone are differentiated into those that are brought about by me and those that are not. There are numerous cases where one's will does not effect change, where either one's will does not impinge on the world or where the world resists (including the unresponsiveness or unreliability of one's own body in paralysis and cerebral palsy respectively). As Russell (1996) suggests, these cases may serve to establish the world and oneself as separate. However, to have a sense of oneself as causally effective at one's will, one has to have a sense that one's actions are reliably effective. This consists not just in my intentions being usually successful, but also in some of them being consistently unsuccessful. No matter how hard I try to do some things, such as jumping over my house, they are never successful. I know my physicallimitations. This is a function of the combination of the physics of the world and the physics and physiology of my body as they each are in certain conditions. Being in zero gravity or in water changes things, but reliably so. Such a sense· of reliable effectiveness also depends on one's movements conforming reliably to intentions, which under normal circumstances they do. As Gibsonians point out, perceptual input specifies the spatiality of the surrounding world in terms of possibilities for action and indeed in terms of one's action; that is, the input specifies what one would have to do with how much force to get to a location, to move or lift an object, and so on. But this depends on one's unthinking sense of the possibilities and constraints of one's bodily action. This sense is unthinking because such expectations are tacit. This is part of "practical knowledge': It might be thought that such a sense of oneself as agent would be compromised in infants with cerebral palsy, which is congenital, since such individuals
The Sense ofAgency
89
from birth have an extreme lack of motor control. 5 However, it has to be said that the degree of lack of control varies both across time in an individual sufferer and between individuals. Also, some actions are always exempt from the effects of the disease, for example, opening and shutting of eyes. Therefore, while assessment of such individuals' sense of agency is interesting and relevant, the thesis of reliable effectiveness probably still holds. All this may seem obvious. But there are consequences of the negation or loss of one's reliable effectiveness. When the reliability of outcomes is undermined, as it is in techniques used in concentration camps (Bettelheim, 1961) or to subvert resistance (Shallice, 1972), one loses an essential aspect of incentive, the assumption that actions can be successful. The result is a passivity, what is termed 'Learned Helplessness', and is characteristic of profound depression (Abramson et al., 1978). Under these conditions, clinical patients have a reduced sense of self and reduced will. In this way, a sense of agency can be seen to be not just cognitive but conative and affective, with hedonic quality. People in such a state lose desire and appetite. Interestingly, the converse of this is often experienced in euphoria. In this case, one often has a sense of vastly expanded possibility or ability in which one experiences little effort. Whereas in depression one's immediate phenomenology is of diminishment, in euphoria it is of augmentation. It could be that the occurrent and the long-term senses of agency are independent of each other. However, there are two considerations which suggest otherwise. The first is that the immediate sense of agency is linked to the long-term sense of agency in ways that consist in the former being not merely a knife-edge atemporal experience. First, in carrying out most actions we have a non-reflective sense that we can do them that is derived from our practical knowledge. This sense of capability involves also a projection of possible actions. The past and future are thus implicitly felt in action. Second, our sensitivity to the 'aspectuality' of actions gives one a sense of agency in and over time. The 'aspect' of an action is that which, separate from tense, differentiates initiation, completion, and continuity of an action (I start to X, I have x-ed, I am x-ing), punctate versus extended action (hit versus swim), temporal immediacy versus distance (I am on the point of x-ing versus I will x). McCormack and Hoed (1999) point out that children's language shows that they are veridically sensitive to aspect before tense. Third, in acting voluntarily, even if our intentions are not always explicit to us, we have a sense of reasons, motives, and goals, which bring in the past and future of each action. (See Gallagher and Marcel, 1999, for fuller discussion.) A further consideration is my suspicion that the occurrent sense of agency would not have its normal phenomenology without the long-term background. One might cognize that 'I am the one doing this'. But what is normally experienced is more than that, a conviction of one's doing it. Certainly this is probably fed by 5
I am grateful to Jill Boucher for bringing this issue to my attention.
90
Anthony Marcel
awareness of one's intention and will in effort. But in one of the cases where one lacks such a conviction, namely, schizophrenia, two other things usually obtain, a state of reflective detachment engendering doubt and a long-term sense of passivity. The other case where one lacks the conviction that one is the agent of action is in the 'flow experiences' mentioned above (Csikszentmihalyi, 1978; Herrigle, 1988). In a way, this is the opposite state to schizophrenia, a sense of effortless possibility and a state of immersion where there is no separation between oneself and the world. And to feel oneself to be an agent requires that one experiences oneself as both part of and separate from the world. These two cases, then, suggest conditions for agency, a sense of oneself as distinct and simultaneously a sense of that entity, myself, being a source of causal effectiveness, that is, connected to the world. It is into that persisting entity that the occurrent phenomenology of the indexical source of present action fits. It fits into it in the following way. The executions of intentions that provide the perspectival source of actions are what constitute practical knowledge, knowing how to do something; and it is just this knowing how that is what is reliably effective over time. My suggestion is that the normal conviction of oneself as agent derives from the fact that current ownership of an action comes from that which translates intention into movement, and this is part of one's fund of practical knowledge, which itself is what constitutes one's sense of oneself as a persisting agentive entity. Existentially speaking, 'doing' is an expansion of oneself in time, where oneself is the doing. It is this that is captured in Borges's and O'Shaughnessy's descriptions cited toward the beginning, and it is this that returns me to my initial quotations. The first two emphasize the necessity of one's non-reflective immersion in one's action, the second two demand that one's actions constitute oneself. Since this chapter is long and discursive, let me summarize its main points. I have proposed that the occurrent immersed sense of agency lies in the experience of ownership of actions, and this is given in the content of awareness of action itself rather than being separate. The content of awareness of action consists in the specifications for action, and the owner is given as their spatially perspectival source. Such specifications follow intention but precede motor commands. Given this, our awareness of our actions can often be inaccurate. Both the sense of agency and the kind of self consciousness proposed here are minimal. The normal sense of ownership is due to our non-observational awareness of action; pathological disownership of action is due to the person's awareness of action being observational. The occurrent sense of agency relies on a long-term sense of agency which depends on reliable causal effectiveness and practical knowledge.
REFERENCES L. Y., SELIGMAN, M. E. P., and TEASDALE, J. D. (1978), 'Learned helplessness in humans: critique and reformulation', Journal ofAbnormal Psychology, 87: 49-74.
ABRAMSON,
The Sense ofAgency BARNARD, P.
91
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 unconscious 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, extension, partiality and disunity of bodily unawareness', Cortex. MERINGER, R., and MAYER, C. (1895/1978), Versprechen und verlesen: eine psychologischlinguistische 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 understanding 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 something. (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 comments 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
1. THE SIGNIFICANCE OF THE
EMPIRICAL DATA, AND A TWO-LEVEL THEORY Tony concentrates largely on awareness of action. The significance of the psychological phenomena seems to me to extend beyond the phenomenological, to the issue of what it is to be an action (the metaphysics of action, if you like). The significance also extends to the relation between the nature of actions on the one hand, and our conceptualization and consciousness of them on the other. Tony says that the movements of the Anarchic Hand are actions; and there are grounds for agreeing. These movements are not reflexes, and they must be under the control of states and initiating events with some form of content. The apparent action has a goal, which is in some way subpersonally represented. When the Anarchic Hand is engaged in such a complex action as unbuttoning a shirt, 1
I treat some ancillary issues raised by Tony's chapter in an appendix.
96
Christopher Peacocke
this representation of the goal will be involved in controlling the sequence of sub-actions which comprise the whole action. The movements of the Anarchic Hand are also sensitive to information, including perceptual information or its precursors, that is deployed and is available in the subject's psychological economy. But what the Anarchic Hand does is not, nor is it the result of, anyone's trying to do it. Nor are there reasons that explain the actions of the Anarchic Hand. An action of the Anarchic Hand is not something done for a reason. If the Anarchic Hand is engaging in action, characterizations of action that require that actions be, or be ' caused by, a subject's tryings are too strong. Perhaps someone will insist that this thin notion of an action is not genuinely action if it is not produced, or does not consist in, someone's trying to do something. In fact, for the points I am about to make, it would not matter if that is true. The important point is that there is a nonempty category of events which are controlled by representations of their goal, and which are not caused by, and do not consist in, anyone's trying to do something. The fact that this category is instantiated has at least three consequences for philosophical and psychological theories of action. (a) If the initiating events which cause the movements of the Anarchic Hand
are not anyone's tryings (or do not lead to tryings), what more is required for an initiating event to be, or to lead to, a trying? The natural answer is that it must be an event of a kind that is characteristically under the control of a subject, and can over time be appropriately influenced by that subject's plans and practical reasoning. This implies that the subject who acts-the agent-must be capable of persistence through time. Even if there can be actions of which no one is the agent, any action which is, or is suitably caused by, a trying, must have an agent. This claim is at the level of the metaphysics of action. This metaphysical claim will also provide further support for a variant of Thesis (7), Longer-Term Dependence, if in one way or another having a sense of agency depends on being an agent (in some possible circumstances). The short term depends upon the long term here, both at the level of agency, and at the level of phenomenology. (b) Claims about distinctions at the level of metaphysics do have repercussions for mastery of those distinctions. If the movements of the Anarchic Hand are controlled by some representation of a goal, then to be responsive to the distinction between those events which are so controlled and those which are not is not necessarily yet to have any notion of a persisting agent and the nexus of mental states characteristic of persisting agents. To have some notion of action is merely to have some grasp of the idea of the goal of the apparent action, and of causation of the apparent action by some state with a content specifying that goal. The case of the Anarchic Hand shows that having a conception of the goal or point of the apparent action falls short of having the conception of
Awareness, Ownership, and Knowledge
97
a persisting subject who tries and acts. On the one hand, this thin notion of action may make for an easier route into the acquisition of the notion of a persisting subject. Its very thinness means that it does not presuppose possession of the notion of an agent and the agent's continuing psychological states. On the other hand, it means that one must be very careful, in designing tests for the presence of a notion of action in an infant human, or an animal, that one discriminates evidence that shows the presence of the thin notion of action from something that can exist only as the action of persisting agents who try and act. (c) Anarchic Hand also bears on the proposals in the section Tony entitles 'Perspectivalness of Source as Minimal Ownership ofAction', which lead up to Thesis (6), the thesis of Minimal Ownership as Egocentric Origin. Tony observes that motor specifications for bodily action must be given in egocentric terms. 'The only spatial description common for all body parts and for external locations is an egocentric one' (p. 84). He suggests that this egocentricity 'might itself specify ownership in a certain way' (p. 84). ' ... in immediate phenomenology the actor is a spatial source, not a substantive or conceptual entity' (p. 85). We can distinguish the task of giving an account of ownership itself from the task of giving an account of the phenomenology of ownership. On the account of ownership itself, I do not see how it can be a sufficient condition of the existence of an owner of the action that the action in question is explained by events and states that involve use of contents framed in egocentric terms. When the Anarchic Hand takes leftovers from a neighbour's dinner place, as reported by Della Sala, this must involve use of egocentric frames. 2 It can hardly be a stereotypical action uncontrolled by information about the egocentrically identified location of the food. A neighbour's food is not always in the same direction or layout from the hands and body. So what distinguishes the egocentric information used in the unowned cases, the actions of the Anarchic Hand, from owned actions? In his account of the phenomenology of owned action, Tony notes that the consciousness is of being the source of the action, where the source is given in perspectival, egocentric terms. Of a patient suffering from Anarchic Hand, Tony writes 'Though the patient may experience perspectivalness of the perceptual content of proprioception, he or she will lack the perspectivalness of the source of the action' (p. 85). I agree with these phenomenological descriptions; but the extent to which they can be explanatory depends very much upon how the notion of the 'source' of an action is to be understood. 2 S. Della Sala, 'Clinical aspects of Anarchic Hand: disowned actions', paper presented at 16th European Workshop on Cognitive Neuropsychology, Bressanone, Italy; cited in Chapter 2, this volume.
98
Christopher Peacocke
There are stronger and weaker readings of the notion of the 'source' of an action. On the stronger reading-which I do not think Tony intends, but which I mention to set it aside-the notion of the source of an action is simply that of its agent. This reading would deprive the descriptions of explanatory status, both at the level of ownership and at the level of the phenomenology of ownership. An impression of being the agent cannot be what makes it correct that one is the agent (and indeed such impressions can be false-see below). Equally, at the level of phenomenology, an experience of being the source in this stronger sense simply redescribes the experience of ownership, and does not explain substantively what this impression consists in. On a weaker reading, all that is meant by 'source' is the body in whose movements the action consists, and whose states and generated forces control the action. This weaker reading seems closer to Tony's intentions when he writes 'The , actor is substantive in this sense only as a body that has extension and whose articulation can be altered with physical forces. This would seem to be a minimal (and physical) sense of ownership of action' (p. 85). How does this weaker reading fare on the explanatory tasks? On the non-phenomenological issue of what it is for an action to have an agent, this weaker sense is powerless to explain why the Anarchic Hand's actions have no owner. For those actions are certainly actions of 'a body that has extension and whose articulation can be altered with physical forces'. On the phenomenological issue, it is true and important that there is a non-proprioceptive way of being aware of one's own, owned actions in which one is aware that one's body is the source, in this weaker sense, of one's movements. It seems to me, however, that this is so because the agent is aware of his trying to act in a certain way. This suggests a different orientation on the whole issue; and also makes me want to enter a reservation on Tony's general position on minimal ownership of action. While ordinary non-proprioceptive consciousness of owned bodily action certainly does involve an egocentric frame of reference, that is so because those are bodily actions. There is equally a distinctive awareness of mental actions-they include calculating, thinking, reasoning, imagining. This awareness is an awareness of them as one's own; and in these cases there equally seems to be an awareness of trying to do something. It seems to be an awareness of trying that is common to all awareness of actions, bodily and mental, as one's own. Egocentric content of oneself as physical source is restricted just to the bodily actions, and is not present in all cases of awareness of an action as one's own. I myself think that there must be some level at which there is a uniform account of ownership that applies to all actions, bodily and mental. I do not think that the notion of ownership is ambiguous. Nor is it at all tempting to hold that 'I' has two different referents in such sentences as 'I walked along the path and I worked out a solution to the problem'.
Awareness, Ownership, and Knowledge
99
Tony's own vibro-tactile experiments establish, as he forcibly argues, that when it seems to the agent, in the distinctive, non-proprioceptive way, that he is acting, this is not causally explained by the agent's acting that way. I would add that the experimental results must also very substantially constrain acceptable formulations of the thesis that the distinctive phenomenology of action is explained by the contents of the agent's tryings. The experimental results establish that the following propositions are not simultaneously true: (a) the distinctive phenomenology of its seeming to the agent that he is ¢-ing
is explained by his trying to ¢; and (b) his trying to ¢, where ¢-ing is a bodily action, involves his issuing motor
specifications for what is in fact in the circumstances a ¢-ing. In the case of a subject with the vibro-tactile illusion described by Tony, it seems to the subject that he is moving his arm in a certain direction, anticlockwise. Some subjects, even after being told about their actual movement in the experiment, 'continued to experience the direction of their voluntary movement wrongly' (p. 66). Suppose such a subject were also trying to move his arm in that same direction, and were issuing motor specifications for moving in that direction. Then-assuming no inability to issue and act on commands meeting these specifications-his arm would move in that direction. Yet it does not move in that direction. It moves in the opposite direction, clockwise. We must conclude from this information, assuming no inability, that a simple theory of action-awareness of ¢-ing consisting of the conjunction of (a) and (b) is false. Any acceptable formulation of the thesis that the phenomenology of action is explained by the content of tryings cannot be so simple. The action-phenomenology in the case of catching a ball, described by Reed, McLeod, and Dienes also provides a counter-example to the conjunction of (a) and (b).3 Tony writes that in action, a subject catching a ball does not merely believe, but experiences, his gaze as rising then falling (p. 69). If this is true, we have another counter-example to the conjunction of (a) and (b) if we take ¢ to be the action type of trying to raise and then to lower his gaze. Presumably these agents would perform that action if they tried to. But although their phenomenology is as of their raising then lowering their gaze, that is not in fact what they are doing when they have the phenomenology of so doing. A minimal, and attractive, revision to the trying-based thesis would be a twolevel theory which distinguishes two contents. One content captures a way the action is given in the agent's phenomenology in acting (it need not be the unique such way it is given). This content is part of the specification of what it seems to the agent, from the inside, that he is doing. In the vibro-tactile experiments as described by Tony, it seems to the subject that he is moving his arm in the 3 N. Reed, P. McLeod, and Z. Dienes, 'Implicit learning and motor skills: what people who know how to catch a ball do not know', unpublished paper.
100
Christopher Peacocke
direction which is, in fact, anticlockwise. So this first content would in that example be 'I am moving my arm anticlockwise'. (I assume for simplicity that the subject is conceptualizing directions as clockwise or anticlockwise: the assumption could be relaxed in a more symbol-laden exposition.) The other content is more closely related to, but not identical with, the content of the motor specification which is operative in explaining his action. As Tony notes, subjects in general need not have any awareness of such low-level matters as the content of motor specifications (which may have to do with the positions of specific limbs, or even muscles). But if a motor specification is a motor specification for a movement type which, in the circumstances, is a given kind of action, the subject can have some conception of what that action kind is. This kind may feature in the content of the subject's tryings, and this is the second level of content I wish to distinguish. In the case of our vibro-tactile subject, one such content would be 'I am moving my hand to that location, just under the light'. At the level of motor specifications, some subjects in these vibro-tactile experiments are reliably successful, in the sense that they do reliably succeed in the stated task, that of moving their hand underneath the illuminated light. When there is reliably successful action, even in the circumstances of the vibro-tactile illusion, at least some of the representations in the specifications or motor commands must be ones that are correctly implemented. They will have contents which are somehow computed from the contents of tryings. (I assume we do not have a case of coincidentally cancelling errors.) On the evidence we have, these correctly implemented tryings seem to concern place. The subject issues commands to move in a certain way that will in fact be a movement of his hand to a certain place, and reliably does so move. The illusion concerns the direction in which his hand is moving, and in which it needs to move if it is to move to that place. Presumably the subject's illusion about his hand's initial position, or the causal grounds of that illusion, interacts with his correct representation of the target to produce these illusions about the direction of movement. Whatever the detailed mechanism and computational processes involved, both levels of content are essential to the correct description of the phenomena. If we omit the level of the content of tryings which help to determine motor content representations involving location that are correctly implemented in action, we would have no explanation of the success of the actions. If we omit the contents involving direction, we would have no explanation of the illusion about the action. It is natural to speak of two levels here, because the content of one level is much more closely related to (though is still strictly distinct from) the level of motor specifications, and therefore much more closely related to execution; while the content of the other is more a matter of how the subject happens to conceive of actions which meet the content given at that first level.
Awareness, Ownership, and Knowledge
101
One revision of the trying-based thesis about the phenomenology of action that takes into account these two levels of content would run as follows:
Two-Level Theory. When someone has the distinctive non-proprioceptive phenomenology from the inside that he is cfJ-ing, then: (a) there is SOllIe action-type 0/ such that the phenomenology of his cfJ-ing is explained by his trying to 0/; and (f3) concerning the action-type 0/: there are representations which cause it to seem to the agent that an action of that type is a cfJ-ing. Here cfJ and l/J correspond to the two levels of content, and cfJ and l/J will be distinct in the problem cases we have been discussing. In the vibro-tactile case, 0/ is the lowerlevel, more executive, content of moving one's hand to the location under the light;
102
Christopher Peacocke
be that he uttered the intended word. If we are prepared to hold that in every such case, there is an unconscious trying to utter the word actually uttered, then the (a)/(f3) characterization can stand, understood as applying to unconscious intentions as well as conscious intentions. Otherwise, that characterization must be relaxed, to allow not only the case in which the agent is trying to lfJ, but also the case in which there is some motor specification of a moven1ent type which need not be anything the agent is intending to perform. The more general two-level formulation of the conditions for the distinctive awareness of cP-ing would run: (A) there is some action type q; or motor specification M such that the phenomenology of his 4>-ing is explained by his trying to lfJ, or by his having motor specification M; and (B) concerning the action type l/J or motor specification M: there are representations which cause it to seem to the agent that it will respectively be or result in a cP-ing. I should also enter some caveats about the relations between trying and 'nl0tor specifications'. The notion of a motor specification in general can cover two different kinds of case, one of which is executive, and the other is not. If I imagine in great detail playing a particular piece of m.usic on the piano in a quite particular way, I may be imagining from the inside making the very movements I would be making in the imagined state of affairs. In some sense, my imagining must involve motor specifications (or imagined motor specifications). But these are not executive, and can certainly not be regarded as either identical with, nor as realizing, any genuine tryings (as opposed to imagined tryings) on the part of the subject. It is motor specifications of a kind that commonly lead to motor commands that are the realizations of the agent's tryings.
2. THE FIRST PERSON, OWNERSHIP, AND AWARENESS OF ACTION FROM THE INSIDE
Can an action which involves a movement of one's own body be experienced as not being one's own action? There is one reading of this question on which Anarchic Hand establishes the possibility. As Tony writes, 'What I wish to draw from Anarchic Hand at present is that it is possible for what we could call an action to be experienced as another's (or not one's own)' (p. 79). In Anarchic Hand, however, as Tony immediately goes on to note, 'awareness from the inside of relevant intentions, effort and will are lacking'. It is by proprioception or by visual or tactile perception that the subject is aware of the actions of his Anarchic Hand. This then in1mediately raises the issue which constitutes the other reading of the opening question of this paragraph: can an action of which the subject is aware from the inside be experienced as not one of his own? If not, why not? If so, how?
Awareness, Ownership, and Knowledge
103
This issue is an issue about whether an action that is experienced in one wayfrom the inside, not by external perception and proprioception-must also be experienced in another-as one of one's own. The issue is distinct from the question of whether an action experienced from the inside must also be one of one's own. Even if it need not be one of one's own actions (as I suggest below), it may be that it still has to be experienced as being such. It does seem that any experience of an action from the inside is also an experience of that action as one's own. I suggest an explanation of this apparent datum. The explanation draws on three explanatory premises: a Constitutive Fact; a fact about the nature of awareness of action from the inside; and a further hypothesis about the determination of the content of awareness of action. The Constitutive Fact is that a subject's trying to 4> involves, constitutively, his trying to make it the case that he 4>'s (where this 'he' refers to a first-person notion or way of thinking of himself). This Constitutive Fact about trying is of course itself in need of deeper philosophical explanation and elucidation, but it is all we need for present purposes. When a trying to 4> is successful, the subject does 4>, and is the agent of the 4>-ing in question. 4 The fact about awareness of action from the inside is that this awareness has representational content. In enjoying such awareness, it seems to the subject that the world is a certain way. There is such a thing as taking such an apparent awareness at face value; and, as many of the examples in Tony's chapter show, there is such a thing as being misled by it. Possession of specifically representational content (as opposed to intentional content that is not representational) is a feature awareness of action from the inside shares with perceptual awareness and, I would argue, with occurrent emotions. We must sharply distinguish between the content of a trying and the content of an apparent awareness from the inside of a successful trying taking place. The trying itself does not involve representational content in this sense. Its 'direction of fit' requires the world to fit with the mind. By contrast, the apparent awareness from the inside of a successful trying does have representational content in the present sense. It follows that an experience from the inside of successfully trying to 4> will be an experience which has representational content, and the correctness of this representational content implies the success of the agent's first-person trying to bring it about that he himself 4>'s. The further hypothesis which will then explain our apparent datum is that this implied content is, in such cases, absorbed into the representational content of the awareness itself. That is, the awareness of the agent 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 egocentric 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 perceptual 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 representational content, in the sense discussed above. These first two facts in the perceptual 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 representational 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 content 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 typedistinct 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 specification area, that he himself is doing it. Aside from the case of overdetermination, only one of these impressions will be correct.
AW~A.RENESS FROM THE INSIDE PROVIDE KNOWLEDGE OF ACTION?
3. HOW DOES
Tony's experimental results, and the literature he cites, make it plausible that in a range of cases, the distinctive apparent awareness from the inside of moving one's arm is not caused, even in part, by one's arm's moving. Yet in some such cases, it is very plausible that you know you have nl0ved your arm, without reliance on proprioception, and without observing your arm. Some of these cases seem to be ones of normal, non-experimental, non-pathological situations. I can know that I have signed my name on the fornl without looking as I sign; and it is quite implausible that my proprioceptive and tactile experiences really distinguish and represent it as being the case that I have signed that particular word. My judgement that I have signed my name seems to result from my taking the representational content of awareness of action from the inside at face value. Our question is: how can this yield knowledge, if the phenomenology is not explained by the occurrence of the bodily movements in which the signing consists, if in normal cases the phenomenology from the inside would be the sanle even if no such bodily movement occurred? This question need not be an expression of real scepticism, an expression of the attitude that such awareness from the inside does not yield knowledge. The philosophical question, which is equally pressing for someone who is not sceptical, is this: why does apparent awareness from the inside of 4>-ing entitle a thinker to judge that he is 4>-ing? We want to know why there is such an entitlement, why relying on such apparent awareness can lead to knowledge of one's own actions. Is there an entitlement because trying to ¢ normally reliably causes the bodily movements that ¢-ing requires? This suggested explanation needs some restriction. Trying to write down the description of a perpetual-motion machine does not in any relevant circumstances reliably result in the writing-down of the description of a perpetual-motion machine. Let us suppose this reliability theorist finds some pertinent restriction of the reliability claim. There are still
106
Christopher Peacocke
many well-known objections to the view that reliability is sufficient for entitlement and knowledge. 6 I will not rehearse or review the discussion of that issue in the philosophical literature at this point in these comments. I will proceed on the view, which needs extended independent discussion, that purely reliabilist views of entitlement have not succeeded, and have not done so because reliability by itself cannot capture the element of rationality that entitlement involves. Actually, even in the case of ordinary perception, I do not think that pure reliability (in the absence of reasons for doubt) is, by itself, what makes perception yield knowledge. So although the phenomena cited by Tony make the issue of entitlement vivid, there would still be live issues about entitlement even if bodily movements were part of the causal explanation of the distinctive phenomenology of action from the inside. The extra pressure added by Tony's discussion is that it is also not even an initial answer to the epistemological questions to say 'Whatever the correct account of entitlement in the case of perceptual knowledge may be, it will apply in the case of action too'. The state of affairs perceived causally explains the perception, but the occurrence of the inscription of the signature does not causally explain the distinctive awareness from the inside in the action case. Any parallelism between the perceptual and the action cases cannot be at such a superficial level. I will argue that there is a deeper parallelism between the epistemology of action and of perception. The parallelism lies in the nature of the entitlement to judge in each case. We can distinguish three levels of characterization of the entitlement relation.? At the first level, the level of instances, we have instances of the entitlement relation. True propositions at this level state that a thinker in such-and-such states and circumstances is entitled to judge a certain content. At the second level, we have true descriptive generalizations about instances of the entitlement relation. At the third level, there are philosophical explanations of why the generalizations at the second level hold. Here is a plausible generalization at the second level about the entitlement to make judgements in which one takes certain contents of perceptual experience at face value. Thinkers are entitled to judge an observational content which features in (or, if perceptual experience has non-conceptual representational content, is an immediate consequence of) the representational content of their experience in the absence of any reason for doubting the veridicality of experience. If we are to understand the nature of this entitlement, we should ask why this generalization holds. Part of the explanation-a first step towards a fuller 6 For one statement, see 1. Bonjour, The Structure of Empirical Knowledge (Cambridge, Mass.: Harvard University Press, 1985). 7 I plan to discuss these issues in much more detail in a forthcoming work on entitlement. There is 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 observational 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 parallelism 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 somewhat 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 experiences 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 argument above, but are still of independent interest. (a) Prior Intention, Trying, and its Precursors
Tony marshals a number of considerations against the thesis that consciousness of intention is a constituent in the distinctive awareness we have of our own actions (pp. 60-1). These considerations all seem good to me, but none of them applies to conscious trying or its immediate precursors. Prior intentions are standing states, whose effective realization involves the subject trying to do something. Such tryings are particular events, rather than states. One may, as Tony says, forget one's intentions, and so in a sense forget why one is doing something. But as one wonders why one was going to the cupboard, one still knows that one is trying to open the cupboard door. The subjects in the Hecaen et al. experiments mentioned by Tony were conscious of the stimulation-induced action of clenching and unclenching the fist, and indeed had no prior intention to perform such actions. 9 But it is quite a plausible description of the case that the stimulation caused a trying that was conscious, and was a constituent of the consciousness of the induced action. When one consciously tries to resist a sudden gust of wind, to stay upright, there need not be a prior intention so to resist: but one is consciously trying to resist. At one point, Tony writes, after reflecting on Haggard's experiments, 'what is relevant may rather be something that is closer to action itself than intention ... ' (p. 73): I suggest that it is the event of trying, or one of its immediate precursors, which meets that condition.
9 H. Hecaen, J. Talairach, M. David, and M. Dell, 'Coagulations limitees du thalamus dans les algies du syndrome thalamique: resultats therapeutiques et physiologiques', Revue de Neurologie (1949), 81:
917-31.
Awareness, Ownership, and Knowledge
109
This difference from Tony would disappear if he were to revise his views only slightly, and say that by 'intention' he means intention in action, rather than prior intention. In fact, one's intention in action seems not to be different from what it is one is trying to do. There is no difficulty in saying that in resisting the gust of wind, one's intention in action is to stay upright.
(b) A Counter-Example to Immunity to Error through Misidentification? Tony writes that Anarchic Hand 'disproves' Evans's thesis that certain self-ascriptions are immune to error through misidentification (p. 81). Evans held that certain ways of gaining knowledge of our bodily properties are immune to such errors in the following sense.!O Consider an utterance 'Someone's legs are crossed, but is it my legs that are crossed?' When the first component expresses knowledge gained in one of these ways, the utterance does not appear to make sense. Evans included proprioception as one such way of gaining knowledge, and this is the target of Tony's claim. The first question to ask about Tony's objection is: what is the property F for which it is supposed to 'make sense' to say 'Someone is F, but am I the person that is F?' As Jennifer Hornsby, and others, have rightly insisted, we must always distinguish the property of moving one's own hand from the property of being someone whose hand moves.!! Let us take each of these properties in turn, to see if it gives a counter-example to Evans's thesis. Does the subject in Anarchic Hand have knowledge that someone is moving his own hand? He cannot know that, since it is not true. The actions of Anarchic Hand are not the actions of anyone: they have no agent. So this property cannot provide a case in which the subject knows that someone is moving his own hand, and wonders whether it is he who is doing so. The first conjunct of a relevant utterance of the form 'Someone is F, but am I the person that is F?' is false. Hence this is not a counter-example to inlmunity. Do we do better if we take as the property that of being someone whose hand moves? Then we do indeed have that the subject in Anarchic Hand knows, via proprioception and not by external visual or haptic perception, that someone's hand moves. But the subject does also know that it is his hand that nl0ves. Here the second conjunct, following the 'but' in the utterance Evans asks us to consider, is not something it 'makes sense' to ask in the context. So again, there is no counter-example to immunity. It may be replied that Evans's formulations are not sacrosanct. We do not have to remain au pied de la lettre, and Tony is right that there is a counter-example to sonlething that is very close to immunity to error through misidentification as defined by Evans. For the subject in Anarchic Hand can know by proprioception that there is an action of moving an arm, and can wonder whether it's his own action. Certainly, it does not seem to be his own. That is important, but nevertheless the letter of Evans's formulations captures a significant philosophical distinction. The formulation framed in terms of the question 'Someone is F, but am I the person who is F?' concerns specifically properties of subjects, persons, rather than actions characterized without reference to persons. Evans was interested in the idea that there are ways of coming to know about one's mental and bodily properties that do not involve what he called 'an identification component'. There seems to be 'no gap' 10 The following formulations are those given in G. Evans, The Varieties of Reference (Oxford: Oxford University Press, 1982), at 220-1. 11 J. Hornsby, Actions (London: Routledge and Kegan Paul, 1980).
110
Christopher Peacocke
(Varieties, 221) in a range of cases between knowing that the property F is instantiated_ and this is a property of persons, so it is instantiated by a person-and knowing that F is instantiated by oneself. If the above reasoning is correct, Evans remains right. There is no property of persons for which the phenomenon of Anarchic Hand opens a gap between knowing, through proprioception, that some person has that property, and knowing that one has that property oneself.
4
Conscious Awareness of Intention and of Action Patrick Haggard
1. INTRODUCTION 'Ow!' squeaked something. 'That's funny: thought Pooh. 'I said «Ow!" without really oo'ing: 'Help!' said a small, high voice. '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)
This conversation offers a good example of the problem of 'efferent binding'. Efferent binding refers to the process of associating intentions with the actions or physical effects that they produce. Pooh cannot bind the physical action of shouting to his intention to shout. In this case, he is right to do so, because the shout was Piglet's rather than his own, and he detects a mismatch. But normally, efferent binding proceeds straightforwardly and unconsciously: we have no difficulty in discriminating things we did from things we did not do. In this case it breaks down because an action occurred which might plausibly have been caused by Pooh's intention, but in fact was not. Interestingly, the failure of efferent binding is worrying for Pooh, and appears to threaten the stability of his sense of self. In this chapter I want to (1) discuss the role of efferent binding in conscious awareness of action, (2) review some experimental evidence regarding conscious awareness of intentions and of actions that participate in efferent binding, and (3) speculate on the importance of efferent binding for our sense of agency and, accordingly, for our sense of self. Why is efferent binding necessary? One account within the motor control literature claims that efferent binding is required for the cancellation of self-induced
112
Patrick Haggard
sensory signals. An intentional action, such as to press a light switch, leads to a barrage of sensory information, such as extension of the arm muscles, pressure on the fingertip, and light on the retina. These consequences are entirely predictable, and, once I have decided to switch on the light, may be of little interest. Therefore, there is no requirement for the brain to process information about these events in detail. However, the same sensory information may be highly important in other circumstances. The central nervous system (CNS) must suppress information about things that I have intended, while processing information about unexpected environmental events. It seems likely that the cerebellum compares copies of the efferent motor command with afferent sensory information. When the intention predicts the action, the two signals cancel, and no further processing is required. If the afferent signal cannot be cancelled by the intention, then an unexpected environmental event has occurred (Blakemore et aI., 1998). This account of efferent binding focuses on afferent physiology: the role of efferent binding is to make perception easier. Efferent binding saves our perceptual systems from having to interpret the barrage of sensory information that follows our every action. However, such theories miss the possibility that efferent binding may be important for action, as well as for perception. Efferent binding may be important for action in at least two ways. First, efferent binding may be important during motor learning. Imagine that I intend to perform an action, but my action produces an error. Clearly, I should learn not to make the same mistake next time, but learning this raises the problem of error assignment. Specifically, I can only learn to correct an error if I can associate it with the corresponding intention. The erroneous feature of the action has to be attributed to an erroneous aspect of the intention for learning to occur. Put another way, the 'error' is only defined with respect to my intention-action unit. For example, the fact that something slipped out of my hand is an error if I am washing dishes, but not if I am tidying things into the rubbish bin. Error detection and correction thus require consequences to be associated with the intentions that cause them. Without efferent binding, then, it seems unlikely we could learn to improve our actions. Second, efferent binding may be important in the construction of the agentic self (Baumeister, 1998). An individual who has behavioural goals, and who performs skilled acts to achieve them, must have the ability to relate the content of their intentions to the actions and environmental consequences that they produce. Consciousness of intention and action may have an important psychological role in allowing individuals to acquire a conscious sense of themselves as agents. This aspect of self may be acquired during development as a by-product of the process of motor learning (see above). One evolutionary purpose of consciousness of intention, on this view, becomes the construction of an appropriate sense of self. The consciousness we have of our intentions and our actions need only be strong enough and sufficient for these two conscious states to be bound together in a process of efferent binding. The phenomenology of intention and action need
Action Awareness
113
not be compelling (Fourneret and Jeannerod, 1998). What matters is that the consciousness be sufficient to sustain efferent binding of intentions to actions. The efferent binding process itself must be sufficient to sustain learning, and to sustain construction of self. A further consequence of the efferent binding approach is to reorder the traditional philosophical priorities in this area. The central philosophical question about action has been whether conscious free will exists. That is, how can T control my body? On an efferent binding view, questions about conscious free will become relatively unimportant. Instead, the important question becomes how does my will or intention become associated with the ~ctions that it causes, with the person that caused them.
2. AWARENESS OF INTENTION Libet et al. (1983) is the classical study in this field. Libet et al. developed an experimental 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 judgements, 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 comparing 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 intentional 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 preparation 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 intentional 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, beginning 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 timings 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 conscious 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 conscious 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 movements 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 psychophysical judgements, or in physiological measurements of brain activity, this difference will not be discussed further. Cortical activity was recorded continuously 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 calculated 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 components 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 Wawareness 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 Early awareness trials Late awareness trials
Q)
"0
.~
a. E
'"
c..
c:
-1500
-1000
-500
+400 ms Keypress
FIGURE
4.1. Grand average readiness potentials
Note: Averaged separately for trials with early Wawareness (thick line) and for trials with late Wawareness (thin line). Source: From Haggard and Eimer (1999).
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 deflection of the LRP, when cortical activity related to the specific movement is developing, 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 identifying 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
Patrick Haggard
118
Lateralized readiness potential (LRP) Median split on W awareness values Wearly
-2000
-1500
-1000
1
-500
w
"0
.~
a. E
'"
Cl..
"" --l
Early awareness trials Late awareness trials
FIGURE 4.2.
Grand average LRPs
Note: Averaged separately for trials with early Wawareness (thick line) and for trials with late Wawareness (thin line). The average time of early and late W judgements, and the average corresponding LRP onset times, are labelled and arrowed. Source: Adapted from Haggard and Eimer (1999).
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 consciousness 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 implemented, 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 planning 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 movement patterns of normal subjects. Specifically, movement patterns are more variable 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 button 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 acceptable motor alternatives with which to achieve it. It is interesting, therefore, that the conscious awareness of intention appears to be associated with the computationally 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 subjects' 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 anticipatory 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 finding 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 anticipatory 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 ,.....
g
121
100
Q)
E
'p
'"'"Q) c -100 2:
'"~
...'" Q)
..Cl
:.:::;
-300
FIGURE
3 Number of letters in sequence
5
4.3. Effects of sequence length on awareness of movement onset
Notes: Results of the experiment of Haggard et al. (1999). The perceived time of a key press becomes more anticipatory as the number of elements in the sequence increases,
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 perceived 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. Day et al. (1989) produced a robust and well-localized example of such selective
I
influence on motor processing stages. They delayed simple reaction time movements 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 contains 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 warningstimuIus 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 signal, to prevent subjects anticipating. On other control trials, no TMS was delivered. 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
-------
0--0
MI stimulation Fez stimulation
~'-6 Occipital stimulation
250 r-.
I
V'l
200
~ l-
e... u
OJ
150
~
OJ
O'l
C
';;" til
Qj
100
Cl
I"",
50
8,-
------·---6
0 Reaction time FIGURE
judf:jed reaction time
4-4. Effects ofTMS on RT and movement awareness (mean and standard error from seven subjects)
Notes: Delaying effects of TMS on motor responses ('actual reaction time') and on conscious awareness of action (,judged reaction time'). Source: Adapted from Haggard and Magno (1999).
(Goldberg, 1985). This pre-motor area has been strongly implicated in the generation 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 subtracted subjects' actual reaction times in control (no-TMS) trials from their reaction 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' judgements of when they pressed the button revealed that less than 100 ms of the TMSinduced 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 consequence 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 stimulation 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 conscious 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 psychophysiological 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 premotor 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 stimulus 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 behavioural 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 perceived 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 agentic 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 generated 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 binding 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 foundations 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 movement', Experimental Brain Research, 107: 254-66. --LESCHZINER, G., MIALL, R. c., and STEIN, J. F. (1997), 'Local learning of inverse kinematics 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 voluntary 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 chapter, 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 second), which leaves only a little time for top-down control. Models accounting for such characteristics have used the notion of optimization principles, that is, principles 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 compose the grip (e.g. thumb and index finger) reaches a maximum value which exceeds object size. The important point, however, is that maximum grip amplitude 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 system 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 estimated 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 biomechanics 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 finger 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 cylinder) was visually 'simple' in that it afforded an infinite number of possible positions 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 orientation of such an object may cause a major reconfiguration of the grasping movement directed at that object (Stelmach et al., 1994).
80
~
~ 70 ~ 60 :;- 50
_
40
~ .....:J
t
40
-i1
30
~
~ 20
o
Y
i
I T (a)
10
8:
0
o
-0-
Small
--e- Middle -0-
Large
o -lOLl_,-,-_ _-,--_ _,-,--_ _-:-_ _--,-,,_ _----:_
x
w
:$i
(b)
-10
0
10 20 30 Object position (degrees)
40
80
~ 70
~
60 :;- 50
--0-
-0-
Small Middle
~Large
t 40
!~~
~ ~-
~ 10
o
8: (c)
~
'-"
[]J' • •~.
:'
0
•
0_ 10 ---:-:---::----:-;:----,:;-;0----,;;;-----:;",--10 0 10 20 30 40 Object position (degrees)
FIGURE
(d)
5. L Orientation of opposition axis during grasping movements
Notes: Fig. 5.1 (a). The diagram in (a) illustrates averaged XY trajectories of the hand grasping the same object (a 6 cm diameter upright cylinder) placed at different locations (from 40° right to 10° left with respect to body axis). I, T, W, trajectories of index finger, thumb, and wrist, respectively. Note change
in orientation of the axis between final position of index finger and thumb (opposition axis), as a function of object position. Curves on the left are plots of orientation of opposition axis as a function of object position computed within two different frames of reference: object-centred frame of reference (b) and head-centred frame of reference (c). Object-centred frame of reference: the angle of the opposition axis with respect to the reference axis changes for each position of the object. Head-centred frame of reference: the angle of the opposition axis remains more or less invariant. In each diagram, the three curves refer to cylinders of different diameters (circles, 3 cm, squares, 6 cm, diamond, 9 cm). (From Paulignan et al., 1997.)
Photographs (d) illustrate critical changes of hand configuration when varying the orientation of an object. Source: Data from Stelmach et al. (1994).
A Cognitive Neuroscience Approach
131
1.1 The Anatomical Theory of a Duality of Visual Pathways
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, correlatively, 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 concentrate 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 occipitoparietal route (the dorsal visual pathway) specialized for visuospatial processing and, by extension, for processing object-oriented movements; and an occipitotemporal route (the ventral pathway) for object recognition. Subjects with lesions located in specific areas of the parietal lobe exhibit a typical impairment in objectoriented behaviour with their contralesional arm. They misreach object location and are unable to form the proper grip: grip size is improperly calibrated, orientation 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 mechanisms 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 supported 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 orientation (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 accuracy. 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 trajectory 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).
1.2 A Model for Automatic Control of Movements The system for automatic control of movements has been described using a simple feedforward model. Roughly speaking, such a model includes a representation of the goal of the forthcoming movement (the desired state), which can be used as a reference for completion of the movement, and which generates output signals for executing it. During execution, the reafferent signals (i.e. signals arising as a consequence of the movement itself) are checked against the desired state. Any mismatch resulting from this online comparison process is in itself a signal for immediate corrections. The reafferent signals used for the comparison are of several types. Visual signals arise from limb displacements and from concomitant displacements of visual objects; proprioceptive signals arise from joint or skin receptors, and so on. Similar models were invented about fifty years ago by physiologists (e.g. von Holst, 1954) to account for the distinction between selfproduced and externally produced changes in the external world. According to these authors, if perceived changes were correlated with self-generated output signals, they were registered as consequences of one's own action. If not, by contrast, they were registered as originating from an external source. The correlation between outgoing signals and the resultant incoming signals is thus an unambiguous feature of self-generated changes.
134
Marc Jeannerod
The goal of the desired action must be kept in the memory for a sufficient duration so as to allow the comparison to operate. As stressed above, if the comparison reveals a mismatch with respect to the desired state, an immediate correction may be generated. If, on the other hand, no mismatch occurs and the desired state is reached, the memory can be erased. The model offers the possibility that the comparison process could also take place ahead of execution itself: a representation of the anticipated desired state as it would occur if it were executed and of the reafferences that it would generate is built and matched with the initial internal model. If this comparison does not anticipate any mismatch, then the system would proceed to the next step (e.g. Wolpert et al., 1995). Such a view, based on engineering models, opens new possibilities for understanding the functions of motor representations. 1.3 Evidence for Lack of Awareness of Automatic Actions These mechanisms are supposed to operate automatically, that is, they should remain outside the agent's subjective experience. This prediction was tested in an experiment specifically designed to investigate the degree of awareness which subjects can access about the details of their own movements. In this experiment (Fourneret and Jeannerod, 1998), subjects were instructed to draw lines in the sagittal direction with a stylus on a digital tablet. The output of the stylus was shown to them on a computer screen seen in a mirror, itself placed so as to mask the subject's hand (Fig. 5.2(a». On some trials, a bias was introduced in the output of the digital tablet, such that the line seen in the mirror appeared to deviate from the sagittal direction (to the right or to the left) and by a given angle. The subject therefore had to deviate his tracing in the opposite direction and by the same angle in order to fulfil the instruction of drawing in the sagittal direction. At the end of each trial, the subject indicated verbally (by selecting a line on a test card) in which direction he thought his hand had actually moved. The results were twofold: first, the subjects were consistently able to trace lines that appeared sagittal, that is, they accurately corrected for the bias. Second, they gave verbal responses indicating that they thought their hand had moved sagitally, hence ignoring the actual movements they had performed (Fig. 5.2(b». Thus, normal subjects appear to be unable to consciously monitor the signals generated by their own movements. Instead, the entire set of mechanisms outlined above (e.g. desired state, memory storage, comparator, etc.) operates on an implicit mode, without conscious counterpart. Subjects, when they are requested to make a conscious judgement about what they did, simply tend to adhere to the goal, but cannot access the way it has been achieved. A different situation arises in cases of failure of the movement, that is, when the goal cannot be reached. In this case, according to the above speculations, because the mismatch between the desired state and the expected result is not corrected,
A Cognitive Neuroscience Approach
A<-------,
lli
135
'What I see' ..-.J
areir~:~g mirror
Pen stylus
Graphic
tablet
I
._
(0) Fixed seat
(b)
Onset of right perturbation (10·)
Starting-point
(n=9)
10
1:
C1
5 6 7
9
ii:
8
+
~
c: 0
Q.
~
6
""i;
-e >'"
pepc!'~
~~CJ?T +
5
.3
4 +
2 (c)
-10 -7 -5 -2
2
5
7 10
(d)
FIGURE 5.2. Poor conscious monitoring of components of one's own actions Notes: An experiment using the apparatus displayed in (a) was used to assess subjective awareness of actions. Subjects drew on a graphic tablet straight lines in the sagittal direction between two targets appearing on a computer seen in a mirror. Subjects saw the line tracing appearing in the mirror as a red line. As shown in (b), a bias could be introduced (dashed line) such that the subjects, in order to keep a straight line, had to orient their movement in the opposite direction. At the end of the trial, subjects were asked to indicate verbally on a chart (d) in which direction they thought their hand had moved during the trial. Results appear in (c), where the verbal responses were plotted as a function of direction and angle of the bias (-10° to - 2°, to the left, 10° to 2°, to the right). As the figure shows, subjects consistently indicated a direction corresponding to about 2°, even when their movements erred by a larger angle. Black crosses indicate the correct responses if subjects had perceived the direction of their movements. Source: From Fourneret and Jeannerod (1998).
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 subjects 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 strategy, 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).
2. WHAT IS CONSCIOUSLY REPRESENTED IN ACTIONS The next problem, therefore, is to identify the level at which the agent becomes aware of his goals, intentions, and desires or, more simply, is able to produce a conscious judgement about his action. According to a widespread conception in psychology, the conscious level of cognitive activities should pertain to a unitary symbolic representation, common to many modalities (e.g. linguistic, perceptual, motor, etc.). Requin (1992), for example, considered that, at the highest level, motor actions would be represented in a non-motoric, holistic, and symbolic mode and thus, could give rise to a conscious experience. This formulation departs from the concept of a motor representation used here, for several reasons. The first reason is that motor cognitive activity seems to be endowed with modular properties. Studies on motor memory show that it relies on mechanisms distinct from those of semantic memory, for example. This can be verified by comparing retention in subjects instructed to learn a list of action sentences ('open a book') either after mere verbal presentation by the experimenter, or after pantomimed performance of each action by the subjects. It was found that performing the actions remarkably improves recall and recognition of these actions. This enactment effect, however, allows for a good recall only of action-relevant information,
A Cognitive Neuroscience Approach
137
not contextual information associated with the actions (e.g. in sentences like 'smoke a pipe in the garden'). Conversely, if instructed to mentally visualize the action during the learning phase, subjects retain contextual information better than action-related information (see Engelkamp, 1995). This is clearly in contradiction to the idea that consciousness of action could be an attribute of a purely speculative 'symbolic level'. The empirical evidence is more in favour of the idea that mental representations consist of modality-specific structures than it is of a propositional-computational mode of functioning (e.g. Paivio, 1986). The second set of reasons for rejecting a single holistic system for conscious representations is that it contradicts many data obtained with cognitive neuroscience methods, showing the discontinuity between the automatic level and the conscious level. Our working hypothesis here will be that the motor representation is a recursive structure, in the sense that goals which account for automatic execution of individual movements are embedded into a broader goal which accounts for the unfolding of the whole action. The memory system in which this ultimate goal is stored to build a representation of the whole action must have characteristics different from those of the short-term storage enabling the performance of individual movements. The problem in this section, therefore, will be to determine at which level of integration of individual movements into the whole action these characteristics begin to change. Before we can address this point, however, we have to introduce a more complete definition of the term action, a definition which did not appear necessary as long as only execution was considered. Action as it is understood here should not be limited to its overt appearance (the production of a set of observable movements). It also includes the covert aspect which corresponds to the internal representation of its goal. Indeed, we postulate that the two aspects, covert and overt, of an action are closely related to each other, such that they are parts of the same continuum. This postulate has important logical consequences, namely, that an overt action necessarily involves a covert counterpart, but that a covert action does not necessarily involve an overt counterpart. This asymmetry has raised one of the most critical issues in classical psychology, about how to infer the properties of the covert part in the absence of any behavioural manifestation (see Jeannerod, 1999). The problem at this point is to determine the content of this conscious level. It is not expected to be a mere 'reading' of the content of the automatic level, where the optimization rules, the biomechanical constraints, and the pragmatic properties of objects are represented. Indeed, as already shown, the content of the conscious level does not include the complete set of details of what has actually been performed and how it has been performed. Introspectively, the agent seems to have access only to the general context of the action, its ultimate goal, its consequences, and the possible alternatives to it. In the next two sections, experimental situations where the consciously available content of an action can be objectively tested will be described. By using the paradigm of motor imagery, we will show
138
Marc Jeannerod
that the actual content of a mentally imagined action extends far beyond what is consciously available. Experiments based on mental chronometric analysis will be used to probe this content and experiments using neuro-imaging techniques will provide some preliminary arguments as to how it is built. 2.1 Chronometric Analysis of Motor Images Mental chronometry relies on subjects' responses, based on their subjective experience. Analysing the pattern of these responses allows making inferences about the cues that subjects potentially use to give their responses. To take a typical example, consider the mental simulation of the classical Fitt's task of hitting alternatively two targets. The rate of hitting is paced by a metronome, beating at an increasing frequency. The subject is instructed to warn the experimenter whenever he or she feels unable to keep hitting the targets as the metronome frequency increases. Sirigu et al. (1996) found that the critical frequency at which subjects failed to hit the imaginary targets was determined by the difficulty of the task, that is, when targets were smaller or placed at a longer distance from each other: this is exactly what happened when the same subjects executed the action of hitting real targets, as predicted by the so-called 'Fitt's law'. In the mental task, all that the subjects (consciously) know is that, when metronome frequency increases, they experience an increasing difficulty in imagining themselves hitting the targets. What they do not know is that their responses follow the same regularity as if they were actually hitting visual targets. Another example of the existence of an implicit content in motor imagery is provided by a series of experiments based on conscious judgements of feasibility of an action. In an experiment performed by Frak et al. (2001), subjects were shown a glass of water from above with an indication of where the thumb and the index finger should contact it. Subjects, without performing the action, had to judge (by pressing on different keys) whether the action of raising the glass and pouring the water into another container would be easy, difficult, or impossible for each set of finger positions. Their pattern of responses followed the limitations that the geometry of the upper limb would have imposed on real motor performance, which implies, although they received no instruction to do so, that they unconsciously simulated the movement before giving the response. Their response times (within 1,500 to 2,000ms) increased with the estimated difficulty of the task. Furthermore, a similar increase in reaction times was observed when the same subjects actually performed the task of grasping the glass, indicating that the same implicit evaluation process found in mentally simulated movements also takes place during execution of these movements. This result is in line with other observations showing that recognition of the handedness of a visually presented hand depends on similar covert sensorimotor processes (Parsons, 1994).As a rule, these experiments tend to converge on a common finding, namely, that the time
A Cognitive Neuroscience Approach
139
give the response reflects the degree of mental rotation needed to bring one's hand into a position adequate for achieving the task. This implicit process is the motor counterpart of the classical mental rotations or displacements used for giving responses about visual objects (Shepard and Metzler, 1971). However, unlike 3-D shapes, which can be rotated at the same rate in any direction, rotation of one's hand is limited by the biomechanics of the arm joints. According to Parsons et al. (1995), response times thus reflect biomechanically compatible trajectories, at the same rate as for executed movements. Similar effects on response times have been observed in making judgements on how to use hand-held objects or tools. Our implicit knowledge about actions influences the way we cognitively process the visual world. In other words, low-level constraints, which normally pertain to the level of automatic execution, are also represented at the level of simulated actions, but they remain implicit. to
2.2 Functional Anatomy of Motor Representations A second set of information about the organization of motor representations can be drawn from neuro-imaging experiments. Although these experiments do not bear directly on the content of motor representations nor on how this content can be accessed consciously, they nevertheless provide new insights into their internal structure. Studying the pattern of brain activity during the process of an action being generated, either limited to its covert part, as in intending or mentally simulating, or also including overt motor performance, reveals that activated areas partly overlap during different modalities of representation. During mental simulation of movement of the right hand, activity increases in several areas directly concerned with motor behaviour. At the cortical level, the primary motor area, area 6 in the inferior part of the frontal gyrus, and area 40 in the inferior parietal lobule are activated on the left side. Subcortically, the caudate nucleus is activated on both sides and the cerebellum on the left side only. Another focus of activity is observed in left prefrontal areas, extending to the dorsolateral frontal cortex (areas 9 and 46) (Decety et al., 1994). Finally, the anterior cingulate cortex (areas 24 and 32) is bilaterally activated, as is SMA (Stephan et aI., 1995). Besides mental simulation, there are other modalities of consciously represented actions, like intentionally selecting a motor pattern among several possible alternatives. Brain activation in this condition also involves the left dorsolateral prefrontal cortex, the anterior cingulate region (Frith et al., 1991), as well as premotor and parietal cortices (Spence et aI., 1997). In addition, a similar pattern of activation can be found in the case of completely implicit motor representations, like those generated by observing tools (Perani et aI., 1995) or even hearing action verbs. The implicitness of covert actions like mentally simulated ones opens interesting perspectives on other modalities of motor cognition. Indeed, another form of
140
Marc Jeannerod
implicit mental operation has recently been added to the realm of motor imagery: this is the cognitive process related to recognizing and understanding actions observed from other individuals. A series of PET experiments were performed to explore brain activity during observation of actions performed by others. In Grafton et aI.'s (1996) experiment, subjects were instructed to carefully observe simple meaningful actions (prehension movements performed by the experi- 1 menter's right hand). Brain activation in this condition was compared with a control condition where the subjects merely looked at visual objects. The activated , areas were mostly located in the frontal lobe, where the SMA and the lateral area 6 in the precentral gyrus were involved, in the inferior frontal gyrus (area 45), and in the parietal lobe (area 40). Grafton et al.'s study, however, did not ' allow a clear distinction to be made between different modalities of acquiring information related to the observed action. It is likely that the subject's cognitive strategy during observation could influence the distribution of cerebral activation. In order to evaluate this possibility, Decety et ai. (1997) gave their subjects specific instructions during the observation of different types of actions. When the subjects had received the instruction to memorize the actions with the purpose of later imitation, the SMA and the ventral premotor cortex were activated. A bilateral involvement of the dorsolateral prefrontal cortex was also found in this condition, in agreement with previous studies concerning the planning of voluntary actions (Frith et aI., 1991) and the mental simulation of actions. When the instruction was to observe the actions with the purpose of recognizing them, by contrast, only the parahippocampal gyrus in the temporal lobe was activated. This pattern of activation supports the idea that actions performed by others can be understood by an observer to the extent that they can be 'simulated' by that observer (see Gallese and Goldman, 1998). This strategy of putting oneself in the shoes of the agent would represent the basis for a broad spectrum of cognitive functions, starting from understanding others' actions, up to learning by observation and imitation. The notion of simulation is in line with many of the results described earlier in this chapter, like, for example, the implicit simulation of action during handedness recognition. Thus, observing an action would activate within the observer's brain the same mechanisms that would be activated were that action intended or imagined by that observer. In turn, this implicit representation in the brain of how movements are generated influences the interpretation of actions observed by other people (see Jeannerod, 1999,2001). A hypothesis accounting for how a representation of an action performed by someone else could be built has been put forward by the Rizzolatti group, based on monkey experiments. In these animals, a population of neurons located in the premotor cortex can be activated both when the animal performs a certain type of movement and when it observes that same movement performed by a conspecific (mirror neurons, Rizzolatti et aI., 1995). According to the above simulation
A Cognitive Neuroscience Approach
141
theory, these neurons might fulfil the role of a representational system encoding observed actions within a format compatible with their execution by the observer. A number of behavioural phenomena could be explained in this way. Among those are motor facilitation phenomena, whereby an action can become contagious among a group of people (yawning and laughing are typical examples). Whether this rather primitive mechanism would also hold for extracting the cognitive significance of an observed action is an open possibility. The main conclusion to be drawn from this section is that motor cognition relies on cortical patterns of activity specific to each function (e.g. mentally simulating, intending, understanding) but partly overlapping between functions. Motor representations, although they can be built or activated in several different ways (e.g. mental simulation of an action originates from a certain type of motor memory; action understanding relies on a special type of perception for biological motion, and so on) share common mechanisms. The automatic level and the 'conscious' level are not independent from each other: they are different aspects of a common process.
3. ATTRIBUTION OF AN ACTION TO ITS AUTHOR: THE AGENCY LEVEL This section will emphasize a paradox in motor cognition. Whereas subjects execute actions, the content of which they remain essentially unaware, they seem to have no problem in correctly attributing th,ese actions to themselves or to an external agent. Besides having explicit, albeit limited, knowledge of the content of . their own actions, normal individuals are good at judging whether they were at the origin of an action or not. Although such a judgement may often be trivial, this is not always the case. Apart from pathological conditions (see below), there are in everyday life ambiguous situations which obviously require a subtle mechanism for signalling the origin of an action. Among those are situations created by interactions between two or more individuals (e.g. joint attention, matched actions, or mutual imitation, for example). There are also situations, usually pertaining to the domain of man-machine interactions, where the cues for an agency judgement about an action can be degraded (e.g. telemanipulation, virtual reality systems, etc.). Schematically, a basic distinction can be made between situations involving observations of actions by other individuals and situations involving self-generated actions (executed or not). In the first case, the incoming information is exclusively based on visual perception. It has been shown that visual information arising from 'biological' movements (e.g. those produced by living individuals as opposed to mechanical devices) contain kinematic features that render them 'intentional' and make them easily recognizable, even by very young infants
142
Marc ]eannerod
(Dasser et al., 1989; Johansson, 1977). This visual information, according to the mirror neuron concept, would ultimately activate motor neurons, with the consequence that the observed action would be understood as a potential action of the self. Now consider the second case, that is, when the action is imagined or intended. Here, the available information is entirely internal to the subject, based on activation of those structures that will represent a desired state. The processing of this self-generated information also ultimately reaches the motor areas: as stressed earlier, the pattern of cortical activation in the two cases is similar and may create confusion. It has to be assumed, however, that the creation of a desired state, although it may activate neurons which are also activated during observation of action (e.g. the mirror neurons), will activate many other populations which will not participate in the situation of mere observation. Finally, it is only when the self-generated action comes to execution that the distinction from an observed action becomes unambiguous. A set of signals become involved, with the possibility of comparing the self-generated signals with coordinated reafferent signals, visual or proprioceptive. This comparison, in principle, should not be possible during action observation, where no reafferent signals are generated. Pathological conditions such as compulsive imitation in frontal patients, where the observed action is readily transformed into execution (Lhermitte, 1983), may represent an exception to this rule: this behaviour suggests that an observer monitoring an action performed by someone else is never far from also being the agent of that action. 3.1 An Experimental Study of Agency
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 disentangle this ambiguity to produce an agency judgement. To this end, an experimental 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 experimenter'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 similar 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 randomly 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 verbal 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 performance degraded in the 'difficult' trials where they saw the experimenter's hand performing 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 generated 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). 3.2 Pathological Dysfunction of Agency One possibility for substantiating the above hypothesis is to examine pathological conditions which affect consciousness of action and self-consciousness. One class of symptoms displayed by schizophrenic patients seems to be closely related to dysfunction of the above mechanisms. These symptoms include insertion of thought, auditory-verbal hallucinations, delusion of reference, delusion of alien control. Those are false beliefs which lead to a feeling of depersonalization by impairing the distinction between the self and the external world (e.g. Schneider, 1959). The impairment of such patients in correctly attributing an action to the proper agent is demonstrated by the frequent occurrence of auditory verbal hallucinations. Neuro-imaging studies in hallucinating and deluded schizophrenic patients provide an interesting framework for interpreting these symptoms. During verbal hallucinations, an abnormal activation of primary auditory cortical areas is observed: in other words, patients 'hear' their inner speech as if it were the voice of someone else (Dierks et ai., 1999). A similar phenomenon was observed during the generation of spontaneous arm movements. Spence et ai. (1997) monitored the cortical activity with PET in schizophrenic patients experiencing delusional control. During the scan, the patients were required to voluntarily move a joystick and to freely select the direction of the movement. Most of them reported vivid experiences of alien control when performing the motor task. Brain activation
Marc Jeannerod
144
---------~
E
Video camera 2
TV screen
(0)
24 20 16 c
'" 12 '6 OJ
2: 8 Non-delusional
4
o Exp. Same (b)
FIGURE
Exp. Ditt.
Subject
Hand presented
5.3. Misattribution of action in schizophrenic patients
Notes: Normal and schizophrenic subjects were shown for 5 seconds their hand or an alien hand in a mirror (a). The hand seen in the mirror could be that of the subject (Subject), or that of an experimenter performing an action different from that of the subject (Exp. Different), or an action the same as the subject (Exp. Same) (see (b)). After each trial, subjects had to respond whether the hand they saw was theirs or not. In the most difficult situation (Exp. Same), normal controls misattributed the alien hand as theirs in about 25% of cases. This proportion rose to more than 80% in schizophrenic patients with delusion of control. Non-deluded patients were significantly more accurate than the deluded ones. Source: From Daprati et al. (1997).
A Cognitive Neuroscience Approach
145
was found to be increased, with respect to normal controls, in a cortical network including the left premotor cortex and the right inferior parietal lobule and angular gyrus, at the level of areas 40 and 39. This right parietal hyperactivity in deluded subjects is particularly interesting: it is noteworthy that lesions at this level frequently result in altered awareness (neglect) for the contralateral limbs and space, and denial of the disease (anosognosia); conversely, transient hyperactivity (during epileptic fits for example) may produce the impression of an alien phantom limb. This result, together with those obtained for verbal hallucinations, might be interpreted as a deficit in cortico-cortical inhibition (possibly of prefrontal origin, Frith, 1996), normally suppressing activity in critical areas (like temporal and parietal cortex) during self-produced action. Lack of inhibition in these regions would therefore lead to incorrect agency judgements with a tendency to misattribute actions to an external agent. The pattern of misattributions in schizophrenic patients is twofold. First, hallucinating schizophrenic patients may show a tendency to incorporate external events in their own experience, or to interpret environmental cues as specifically directed to themselves. Accordingly, they may misattribute their own intentions or actions to external agents. During auditory hallucinations, the patient will hear voices that are typically experienced as coming from a powerful entity trying to monitor and control his or her own behaviour. In cases of delusion of alien control, the patient may declare that he or she is being acted upon by an alien force, as if his or her thoughts or acts were controlled by an external agent. The reverse pattern of misattribution can also be observed. In this case, patients are convinced that their intentions or actions can affect external events, for example, that they can influence the thought and the actions of other people. As a consequence, they tend to misattribute the occurrence of external events to themselves, as if the effects of actions of others were interpreted through the intentions of the self. Daprati et al. (1997), using the same paradigm as above in groups of schizophrenic patients, found a dramatic increase in the rate of incorrect responses in the 'difficult' trials. The error rate was 80 per cent in a group of schizophrenics with delusional experiences, whereas in a non-hallucinating group, it was only 50 per cent (Fig. 5.3(b)). The fact that all patients gave nearly correct responses in the other two cases (the error rate remained within 1-7%) excludes the possibility that the effect observed in the 'difficult' trials could be due to factors unrelated to the task, such as lack of attention.
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 responsible 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 neuroimaging 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 influence 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 disruption 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 produce 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 perception of spatial features of objects', Neuroimage, 10: 114-24. FOURNERET, P., and JEANNEROD, M. (1998), 'Limited conscious monitoring of motor performance 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 auditory 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 displacement', 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 interaction 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 coordination', 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:
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 position 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 manmade 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, 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
v.,
I
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 opposition 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 sensorimotor 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 discussed 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 demonstratives 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 grazing side by side, and take careful aim at mine. At the last moment, the two donkeys 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 donkey 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 identified 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 demonstratives 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 voting 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 impossible 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 identification 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 singlingout 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 interpret 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 subject 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 successful in my guesses or attempts. It still seems evident that, since I still cannot consciously 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 identification. At most, I am using some such identifier as 'the window in which my interlocutor 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 identification of the object under which my action is intentional.
2. THE GROUNDING THESIS This still leaves us with the question: why is there a special connection between demonstrative identification and action? If we think, as I have suggested we should, that demonstrative identification of an object depends on conscious attention to the object, then this can be reduced to the question: why is there a special connection between conscious attention to an object and action? One natural thesis is this:
Grounding: The meaning of a perceptual demonstrative is grounded in those aspects of perceptual experience that set the parameters for my action (how far I move, in what direction, and so on). A term like 'that window' does not have the same meaning whenever it is used. On anyone occasion on which I use it, its meaning has to be filled out by the way in which I am perceiving the object. My current experience of the object gives me a
Demonstratives in Action-Explanation
153
kind of picture of the object-its shape and size, how far away it is, and so on. According to the Grounding Thesis, this picture of the object is the meaning that the term has on that occasion. And, by the Grounding Thesis, the way in which I am aware of the object-the picture of it that I consciously have-determines the coordinates of my action, it sets the details of the way in which I reach for the thing. The idea is that the perceptual information in which the meaning of a perceptual demonstrative is grounded gives a kind of descriptive meaning of the demonstrative, and this perceptual information sets the parameters of my action-how far I move, in what direction, and so on. What explains the fact that perceptual demonstratives have their special link to action is that the meaning of the demonstrative is given by the pictorial content of the experience, which is what controls my movements. You might be encouraged in your adherence to Grounding by an uncritical reading of Alan Allport's conception of visual attention as 'selection-far-action'. He describes it as follows: Any goal-directed action requires the specification of a unique set of (time-varying) parameters for its execution-parameters that determine the outcome as this particular action rather than any other: as this particular vocative or manual gesture, this particular directional saccade, and so on. Consider now what is required if those parameters are to be controlled by sensory (say, visual) information. Suppose that visual information has to guide manual reaching, for example, to grasp a stationary object or to catch a moving one. Clearly, many different objects may be present in the visual field, yet information specific to just one of these objects must uniquely determine the spatiotemporal coordinates of the end-point of the reach, the opening and closing of the hand, and so on. Information about the position, size and the like of the other objects in view, and also available must not be allowed to interfere with (that is, produce crosstalk affecting) these parameters-though they may need to affect the trajectory of the reach in other ways. Consequently, some selective process is necessary to map just these aspects of the visual array, specific to the target object, selectively onto the control parameters of the action. (Allport, 1989: 648)
To think that this amounts to an endorsement of Grounding, though, you would have to suppose that the visual information that is being used in setting the parameters for your action must be part of the content of your experience; and we shall see that this need not be so. Notice already that in the case of a blindsighted subject who manages to act successfully with respect to one rather than another object in the blind field, there is selection-for-action in Allport's sense, even though the visual information that is being used to set the parameters of the action is not consciously available. You might suppose, though, that the only possible role for conscious demonstrative identification of the object in explaining action is the one stated by Grounding: that the experience contains the information to be used to set the parameters for the action. But there is another possible role for the experience, there is an alternative to Grounding. The role of conscious
154
John Campbell
attention to the object may be to single out just which object is to be acted on, without yet directly setting any of the parameters for the action. That setting of parameters may be done at a level remote from consciousness. An analogy may be helpful. Suppose we go back for a moment to the example of acting intentionally on someone by ringing up that person. The analogue of 'setting the parameters of your action' here is determining which particular numbers to dial. So you might argue that explicit knowledge of the telephone number is needed to execute this type of intentional action on a person. But suppose you have a telephone which stores Marianne's number. In that case, all you need to do is to press a button marked 'Marianne' and the machine will set the parameters for your action. You do not need to know the relevant digits yourself. Even if you had a mistaken belief about what Marianne's number was, this procedure would still let you successfully ring Marianne up. If we thought this a better model for the role of demonstratives in action-explanation, the conscious experience itself need not be what sets the parameters for your action, and experience of the object will have some quite different role in explaining the action. In particular, conscious attention to the object may have the role merely of identifying the target of your action, with the parameters of the action being set not at the level of conscious experience, but by subpersonal visuomotor systems remote from consciousness.
3. PROBLEMS WITH GROUNDING: TWO VISUAL SYSTEMS To see why Grounding is problematic, we should look at the distinction between two visual pathways, sometimes referred to as the 'action' and 'perception' pathways. The distinction is partly motivated as an interpretation of the anatomical distinction between dorsal and ventral visual pathways. One way of setting out the distinction between action and perception pathways is provided by Jeannerod (1997) (see Fig. 6.1). The suggestion is that the various processing streams can be divided into (a) the computation of direction and distance of the object, given in body-centred coordinates, for activation of reach, together with (b) size, shape, and so on being computed in allocentric terms, for activation of grasp. Both of these processing streams belong to the 'action' pathway. This is contrasted with the 'perception' pathway, in which (c) properties of the object such as size, shape, colour, and so on are bound together and the object categorized by the visual system. The 'action' pathway is a relatively low-level system, remote from consciousness, used in the fine control of motor movement (Goodale and Milner, 1995). Once we have this distinction in place, the natural proposal is that intentional action on an object must involve the 'perception' pathway, which underpins awareness of the object (Goodale and Milner, 1995). For there to be intentional action, there must be an involvement of high-level identification of the object involved, provided by processing in the 'perception' pathway.
Demonstratives in Action-Explanation
,.,::::.-------...1
Distance
~'--- _ _
155
Activation of reach
----l
Activation of grasp
Object
Identification Visual processing FIGURE
6.1. Diagrammatic representation of central processes involved in object-oriented behaviour
Notes: Extrinsic attributes of an object (related to its spatial position) are processed in body-centred coordinates, for activation of the reach. Its intrinsic attributes are processed in a different, parallel pathway for activation of the grasp, also pertaining to the dorsal visual pathway and to the posterior parietal areas. Shape analysis, using the same object primitives, is effected in both the dorsal pathway for visumotor transformation and in the ventral pathway for perceptual identification. Only a few components of visual processing (e.g. size cues, depth cues, etc.) are mentioned. Others are symbolized by the empty box. Semantic knowledge stored in the ventral pathway can improve visuomotor transformation using connections between the two pathways. Source: From Jeannerod (1994).
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 carbonmonoxide 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 capabilities. 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 correlated 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 display 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 difference 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 correctly 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 consciously 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 experience 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 target, 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 managing 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 notwithstanding (Goodale, 1996; Agliota et al., 1995). Once again, it seems that for ordinary 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 determined 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 particularly basic about perceptual demonstratives in the explanation of intentional action?
4. THE BINDING THESIS I think that we can give a better account of the role of perceptual demonstratives than the Grounding Thesis offers. Let us look again at the distinction between 'action' and 'perception' pathways in vision. The phenomenal experience of an object depends on 'perceptual' processing, and in particular it depends on the binding of the elementary attributes of the object into a single entity. But the 'action' pathway does not depend on binding. The point is that the various properties represented on the 'action' pathway contribute one by one to the control of action; each individually contributes to the motor configuration of the hand (Jeannerod, 1997: 77). Of course, since the planning of action does require that it
Demonstratives in Action-Explanation
159
be attributes of one and the same object that are used to contribute to the motor configuration of the hand, there must be a kind of 'virtual binding' even in the pragmatic control of action. But the point is that each attribute thus chosen contributes separately to the control of action; it is not some higher-order property of the bound whole object that is controlling the movement. If you identify an object on the basis of semantic processing-if you think, for example, 'that hat is my hat'-and choose to act on it, there is the problem that the 'action' system has to swing into play to pick that very hat, the one that has been experienced on the basis of semantic processing. There is another binding problem here, which is: how do we make sure that the 'action' and 'perception' systems are dealing with one and the same external object? I think that the natural proposal here is the one that Jeannerod makes. The picture he has in mind is that there may be early processing areas shared by the two visual pathways in which the visual primitives and spatial localization are presented on the same map (Jeannerod, 1997: 80). In effect, the suggestion is that it is location that is the binding principle. Jeannerod talks about 'the classical Aristotelian notion that what arises from the same point in the external world pertains to the same object' (p. 80). So on this view, 'attentional mechanisms would play a role in binding different modes of representation into a single, higher-order one' (p. 80). I said earlier that conscious attention to an object is what you use in interpreting a perceptual demonstrative referring to that object. We also saw that demonstratives seem to have a special role in action-explanation. Consequently, our problem has been to explain what the relation is between conscious attention and the capacity for intentional action on the object. The present proposal is that conscious attention to the object will include some awareness of the location of the object, and that the target for processing by the visuomotor system can be identified as: 'the object at that location'. For this to be successful, the experience of location would not have to be accurate. Moreover, even if the initial experience of location was accurate, there could be subsequent changes in the location of the object which were not picked up by the 'perception' system but which were compensated for by the 'action' pathway in setting the parameters for action on the thing. It is not critical to this picture that location should be the key factor that conscious attention uses to set the target for the visuomotor system. The point is that, however in detail it is done, the role of conscious attention to the object is not directly to set the parameters for action on the thing, but to provide enough information about the thing to define a target for the visuomotor system. Once the target is defined, it is up to the visuomotor system to set the parameters for action on the object. One natural analogy here is with a heat-seeking missile. You do have to begin by pointing it in roughly the right direction, so that it locks on to the right object. But once that has been done, lower-level mechanisms take over to keep it locked on to the object.
160
John Campbell
The Grounding Thesis held that the role of conscious experience of the object in action was to set directly the parameters for action on the object. I am now suggesting that that is a mistake. Rather, the role of consciousness is this: conscious attention is what defines the target of processing for the 'action' system, and thereby ensures that the object you intend to act on is the very same as the object with which the 'action' system becomes engaged. I shall call this the Binding Thesis:
Binding: Conscious attention is what defines the target of processing for the 'action' system, and thereby ensures that the object you intend to act on is the very same as the object with which the 'action' system becomes engaged. Accepting the Binding Thesis is, of course, quite consistent with rejecting Grounding; it allows that the properties being computed in the 'action' pathway may be at variance with the properties being computed in the 'perception' pathway which you experience the object as having. The role of conscious attention is only to ensure that both sets of computations concern the properties of the same object. The role I am envisaging for conscious attention to the object, then, is to find the target of the information-processing, which need not itself be conscious, involved in acting on demonstrative propositions about the object. The reason why location matters in your experience of the object is that your conscious attention to the object has to be identifying the target for the benefit of these information-processing systems. And location may be one of the ways in which the information-processing systems identity their target. And this whole procedure may work even though your experience of the location of the object is not particularly accurate. Goodale and Milner (1995) suggest that it is the 'perception' pathway, rather than the 'action' pathway, that sustains conscious experience of the object. We can see why it should be experience of the object that matters for knowledge of the reference of a demonstrative by going back to the case of a blindsighted subject. Goodale and Milner's analysis is that in such a case, there is still input to the 'action' system from areas other than the primary visual cortex, whereas there is no significant input to the 'perception' system. A blindsighted subject with a functioning visuomotor system may be able to perceive the propensities of the objects around her to respond in various ways to her actions; she may be able to perceive some of the 'affordances' provided by the objects, in Gibson's sense (Gibson, 1979). But what is such a subject missing? If the subject reaches and grasps successfully, the subject none the less does not know why the reaching and grasping has been successful. The subject has been right in thinking that there are these affordances there, but does not know why the world has afforded just this and that. Someone who has conscious experience of the object, though, experiences the categorical object itself; experiences, that is, the reason why just that reaching
Demonstratives in Action-Explanation
161
and grasping would be successful. And that ultimately, I suggest, is why knowledge of the reference of a demonstrative requires specifically conscious attention to the object. What experience of the object provides is knowledge of the object whose categorical properties are the grounds of the affordances that the blindsighted subject may be able to perceive. Mere perception of affordances is not enough for knowledge of which object is being referred to.
5. THE SENSE OF AGENCY I want to conclude with some remarks about our knowledge of our own actions, remarks that are consequential upon the role of demonstratives in actionexplanation. It is often said that I seem to have knowledge of my own actions in a way that is different to the way in which I have knowledge of your actions. I have knowledge of my own actions 'from the inside' whereas I have knowledge of your actions only as an external observer of them. But what does this difference come to? I think there is a difference in the way in which I have knowledge of the objects of my own actions, from the way in which I have knowledge of the objects of your actions. Suppose that you and I are confronted with a wagon of apples. I watch you reaching for an apple. Straight off, it seems that at best I can have a hypothesis about which particular apple you are reaching for. It is only a hypothesis that you are reaching for this apple rather than that. So I could be making either of two types of mistake. I could be wrong about whether you are reaching for anything at all; perhaps you are just slowly toppling over. Or I could be right in thinking that you are reaching for something, but just wrong about which particular thing it is. It could be that you are after all reaching for some other apple than the one I think you are after. Suppose I think that I know that you are reaching for 'that particular apple'. I could in principle be confronted with evidence that would make me doubt whether it is indeed that apple that you are reaching for, while leaving intact my initial evidence that you are indeed reaching for something, even if it is not that apple. If that doubt is correct-if you are indeed reaching for something, only not the apple I have in mind-then I have, in the terminology of Shoemaker (1984), made an error of identification. Although my judgement that you are reaching for that apple is mistaken, the mistake is, as it were, confined to being a mistake about which object is in question. It is a mistake of identification only, rather than a mistake which undermines the whole judgement. On the other hand, consider my own judgement about which apple I am reaching for. Here I seem to have some authority over which object it is that I am reaching for. It does not seem to be possible that I should think I am intentionally reaching for one apple, but actually be intentionally reaching for something else. It is, of course, possible that I should accidentally pick up the
162
John Campbell
wrong apple; but that is not the same thing as reaching intentionally for it. And it ' is possible that I think I am reaching for something, when in fact I am sound asleep; but here the error is not confined to my identification of the object On which I am acting. In Shoemaker's terminology, my judgement about which apple I am reaching for is immune to error through misidentification. This phenomenon is, though, confined to knowledge of the object of my action which identifies the object demonstratively. Consider again Austin's example of the donkeys. When I make a mistake, take careful aim and fire at your donkey, I will say, 'I am shooting my donkey', and this judgement really does involve an error of identification. In contrast, when, in the same situation, I say, 'I am shooting that donkey', my judgement is not subject to error through misidentification. As Austin pointed out, there might be an accident, the donkeys might move at the last minute so that I shoot one donkey, but not the one I identified demonstratively. In that case, my judgement, 'I am shooting that donkey', involves an error. But the error is not an error of identification alone. Because of the unexpected movement of the donkeys, I am not in a position to know that I am shooting anything at all. So the mistake here is not local to the subject term; it is not an error of identification. I think that this knowledge of the object of one's action, immune to error ~ through misidentification, is connected to one's sense of being the agent of one's action. I think this point illuminates such puzzling phenomena as Anarchic Hand syndrome. This kind of case is sometimes described as one in which the hand executes unintended yet purposeful actions. But this kind of description, natural though it is, is paradoxical. How can the movements of the hand be purposeful, if they are unintended? There is only one person around who could be the agent of the action, so if the movements of the hand are purposeful they simply must be intended by the agent. We can, though, think of the movements of the hand as involving actions that are intended by the subject, but which are not recognized by the subject as intentional: the subject has no sense of agency in connection with those actions. Suppose that you are subject to this syndrome. You may form a motor instruction to pick up a particular glass, but not realize that you have issued that motor instruction. So you simply watch your hand move towards the glass. In that case, you can form the hypothesis, 'I am going to pick up that glass', but in that case you really are only forming a hypothesis about which glass is going to be picked up. So you are demonstratively identifying the glass, and the meaning of the demonstrative may be grounded in those aspects of perception which are being used to control the action. None the less, from your point of view, it seems that you could make an error of misidentification in making your judgement about which glass is being intentionally acted on. You might be right, as you watch the hand, that some glass is going to be picked up, but just wrong about which one it is. The sense of possession that a schizophrenic might have could lead to a similar upshot. The schizophrenic who feels that he has been possessed arid that his hand
Demonstratives in Action-Explanation
163
is being moved towards a glass may form a hypothesis about which glass it is that is about to be picked up. But again, it is only a hypothesis. From his point of view, it seems that he might be right that some glass is about to be picked up, but wrong about which one it is. He has a demonstrative, 'that glass', whose meaning is grounded in those very aspects of perception that are being used to control the action. But his hypothesis, 'that glass is about to be picked up', could involve an error of identification. So both these cases stand in contrast to the ordinary case, in which I fire at the donkey and know 'I am shooting that donkey'. Here my judgement cannot involve a mistake of identification. I may have missed, I may have hit something else altogether. But if I acted intentionally on anything, it was that donkey that I acted on. So I cannot know that I acted intentionally on something, but be wrong in thinking it was that donkey that I acted intentionally on. You might say that the pathological cases are not after all so different from the ordinary case here, for the pathological cases are after all not cases in which the subject knows himself to have acted intentionally on anything. The whole point about the Anarchic Hand is that I do not know that these movements are my actions; the whole point about being possessed is that I do not know that these movements are my actions. So these are not cases in which I know that I have acted intentionally on something, only I might be wrong about which thing it is. Rather, these are cases in which the action is not ascribed to me, but to some external agency: the hand itself, or the malign alien in the case of the schizophrenic. The note on which I want to end is that it seems to me that we ought to think of the self-ascription of an action, the sense that one is the agent of an objectdirected action, as a consequence of the fact that one has knowledge of which demonstrated object is being acted on that is immune to error through misidentification. There are not two separate things here, the fact that in the pathological cases the patient has only a hypothesis about which object is being acted on, and the fact that in the pathological cases the patient does not feel himself to be the agent of the action. Rather, it is the fact that in the pathological cases the patient has only a hypothesis about which object is being acted on that explains why the patient does not feel himself to be the agent of the action. The actions of which I am the agent just are those actions of whose targets I have knowledge which is immune to error through misidentification.
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 reaching 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 further 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 intentional and volitional basis of action. They date back to Lotze's and James's discussions 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 condition 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 physiological 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 development 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 appropriate 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 (volitional) 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 situations 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 discussion 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 mysterious 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 actionand, 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 transparent 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 movements 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 functional 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 concerned. 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 performance (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 mechanisms 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 imitative 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 outcome (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, spontaneous 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:
compatible:
SI
Sr
RI
Rr
incompatible:
X
When the assignment is spatially compatible, stimuli and responses share a common spatial feature (both left or both right), whereas they will always exhibit two different spatial features in the incompatible assignment (right and left or left and right). As has been shown in a large number of experimental studies, response performance for compatible assignments is clearly superior to incompatible assignments, and this holds for both response times and error rates (e.g. Fitts and Biederman, 1965). It is not easy to account for an effect like this in terms of practice and contiguity. Rather, the effect suggests that responses can become prespecified by stimuli on the basis of shared features. This is not far from claiming that perception induces action by virtue of similarity. What exactly could these shared features refer to and what precisely could be the basis of the correspondence between stimuli and responses that generates the effect? This has been studied for the Simon task, a close relative to the standard compatibility task (Simon, 1969; Simon and Rudell, 1967). In this task, two response keys are assigned to the two hands, and two possible stimuli can occur, for example, a high-pitched and a low-pitched tone (H and L, respectively). The task requires to press the left of the two keys in response to one stimulus (say, L) and the right-hand key in response to the other (say, H). Stimulus identity (Hvs. L) is therefore the relevant stimulus dimension. Yet, in addition to this, the position of the stimulus is varied as a further dimension (left vs. right of a fixation mark). This dimension is irrelevant in the sense that the participant is instructed to focus on stimulus identity and completely ignore stimulus position. Hence, the four resulting pairings of stimulus identity and position can be classified according to correspondence between stimulus and response positions. Positions do correspond when stimulus L is presented on the left-hand side or stimulus H on the right-hand side. They do not correspond in the other two cases. In this task, too, reaction times are short in the case of correspondence and long in the case of non-correspondence. This is the Simon effect, and again one can think of it in terms of shared features in perception and action: responses are fast when stimuli and responses share the same spatial position but slow when the positions are different. Yet, there are several levels at which the correspondence could be effective. First, when, for example, the left-hand key is pressed in response to a left-hand stimulus the motor command for the hand movement is generated in the same brain hemisphere in which the sensory code for the stimulus has been generated
170
Wolfgang Prinz
immediately before. Also, the response is generated by the left hand, that is, by an effector attached to the body hemisphere that corresponds to stimulus position. This may be summarized under the notion of anatomical correspondence. Second, correspondence may be claimed between the location at which the stimulus is presented and the location at which the action is performed. This may be called locational correspondence. In the standard Simon task, locational correspondence is always confounded with anatomical correspondence. There is, however, an easy way to unconfound them by having subjects perform the task with their hands crossed (Simon et aI., 1970; Wallace, 1971). If anatomical correspondence is the critical factor, one has to expect that the effect is completely reversed in this condition. In contrast, it should be unaltered if locational correspondence counts. The results of the experiments are clearly in favour of locational correspondence, suggesting that the functional origin of action induction resides at a high level of representation at which environmental events and actions (= stimuli and responses) are localized in extracorporal space. One may even go one step further and distinguish between the location of the action itself and the location of the action's goal, provided one can separate the two. In the standard paradigm where the action is a simple key press there is no way to separate the two because the action (= pressing down the key) and the goal ( = the key being pressed down) share the same location. What happens when one dissociates actions and goals? In this case, in addition to locational correspondence, intentional correspondence might be effective as well-that is, correspondence between stimulus location and goal location-irrespective of response location. Hommel (1993) ran an experiment to test this view. In his experiment, the hands were left uncrossed. What was crossed instead were the links between the two response keys and two additional lights that were triggered by operating the keys. As a result, when the left-hand response key was pressed, a feedback light on the right-hand side would go on, whereas a feedback light on the left-hand side would go on when the right-hand response key was pressed. The task was a Simon task, with stimulus identity (pitch) as the relevant and stimulus location as the irrelevant dimension (high vs. low-pitched tones, coming from loudspeakers mounted on the left vs. right-hand side). There were two groups of subjects. In the control group, participants were instructed to press the left key in response to low-pitched tones and the right key in response to high-pitched tones and ignore the lights altogether. In the experimental group the same task was administered under a different instruction. This time participants were instructed to switch on, as fast as possible, the light on the right-hand side in response to lowpitched tones and the light on the left-hand side in response to high-pitched tones. The purpose of the design was to implement two different intentional sets: one in the control group that refers to a goal residing in the movement itself (= key press) and one in the experimental group that refers to a goal beyond the movement (= light onset). In the control group where the lights served no particular function
Experimental Approaches to Action
171
the standard Simon effect was obtained. However, in the experimental group, where the lights served the function of action goals, the location of the movements was virtually irrelevant. Instead, the critical factor was the correspondence between stimulus location and goal location: responses were fast when the stimulus and the target light shared the same location (despite the fact that the response was always located on the opposite side). Conversely, responses were slow when the stimulus and the target appeared at different locations (despite the fact that the location of response corresponded to that of the stimulus). Two major conclusions seem to emerge from this research. First, action induction seems to be based on representations which code stimuli and responses as environmental events in extracorporal space. Second, it seems that action induction is more dependent on shared features between stimuli and goals rather than between stimuli and movements. 1.3 Common Coding In summary, then, we see the classical sens.orimotor framework faced with two major challenges: to provide functional roles for similarity and for goals. In order to meet these challenges, we need to come up with a novel framework that blends elements from the sensorimotor and the ideomotor stance. The framework of common coding is meant to offer solutions to both of these challenges (cf. Hommel et al., 2001; Prinz, 1984, 1987, 1990, 1997a, b). Similarity: Commensurate Coding
How can a functional account be given of the fact that certain patterns of stimulation may, under certain conditions, induce certain patterns of action by virtue of similarity? Basically, there can be no role for similarity without commensurate coding. The notion of common coding suggests that, somewhere in the chain of operations that lead from perception to action, the system generates certain derivatives of stimulation and certain antecedents of action that are commensurate in the sense that they share the same system of representational dimensions. Such shared dimensionality is a prerequisite for any form of similarity-based processing. Commensurate coding is a strange notion for the sensorimotor framework for perception and action. This framework relies entirely on separate coding. Stimuli and responses are, by definition, distinct and incommensurate, and this applies to their internal representations as well: stimulus codes stand for patterns of stimulation arising in receptor systems and movement codes stand for patterns of excitation travelling to effector systems, and there is no way these codes could be compared with each other, or matched. What is required, instead, is a mapping device that translates stimulus codes into movement codes. The metaphor of translation which is in fact abundantly used in the reaction time literature (Welford, 1968, 1980; Sanders, 1998; Massaro, 1990) stresses the incommensurability between stimulus codes and movement codes: there is one
172
Wolfgang Prinz
representational language for stimulus information, organized in terms of sensory dimensions and there is another language for movement information, organized in terms of motor dimensions. The operation mediating between the two translates from one language into the other: it bridges the gap between perception and action by creating links between incommensurate entities. The claim I want to defend is not that the sensorimotor framework is mistaken. Instead, I maintain the weaker claim that it is incomplete. Quite obviously, there must be routes from perception to action where incommensurate codes are translated across separate representational domains. Yet, to account for action induction, there must also be routes where stimuli and responses meet in a common representational domain-represented through commensurate codes and speaking the same language so that they can be matched to each other without translation. In this domain, action codes can be the same as (or similar to) perceptual codes with respect to some of the representational dimensions and different or dissimilar with respect to some others. Therefore, perception and action codes will overlap to various degrees and, as a consequence, they may induce each other by virtue of their overlap, or similarity. What codes could be functional in the supposed common domain? A number of options can be considered, varying on two dimensions (d. Prinz, 1984, 1990, 1992). One dimension refers to the functional locus at which perception and action meet. The other refers to the nature of the information contained in the supposed commensurate codes. As to the first dimension, the functional locus at which perception and action meet may either be placed on the perceptual or the action side. On the one hand, one can think of the meeting place as being located in a sensory or perceptual coding domain where stimuli and responses are both represented in terms of stimulusrelated coding dimensions. This view posits a perceptual basis for action representation. For instance, one may assume that action and perception can talk to each other directly because actions are represented through their sensory or perceptual consequences. On the other hand, one can also think of the meeting place as being located in a motor- or action-related domain of coding where stimuli and responses are both represented in terms of response-related coding dimensions. This view posits an action basis for perceptual representation. For instance, one may assume that perception and action can talk to each other directly because perceptual events are represented through movements or actions that lead to them. s 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 gestures 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 corresponding 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 commensurate 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 correspondence 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 representation of an action that a person performs him or herself will be entirely incommensurate 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 commensurate, 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 representational 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.
Goals: Action Codes Commensurate coding of events and actions notwithstanding, our approach is still sensorimotor in its basic functional logic, as long as we do not provide a functional role for goals in action control. Therefore, we need to add to it an ideomotor component, yielding an extended framework with functional roles for both external and internal causes of action. Goals may vary along a number of dimensions. For instance, they can be concrete (like e.g. catching a rolling ball by a particular grip) or abstract (like e.g. winning a game of chess). Further, they can refer to resident versus remote effects of action. For example, one can raise one's arm just in order to perform this particular gesture (for instance, in dancing) or in order to open a window, or in order to say hello to somebody else. In these three cases, the same movement is used to achieve three entirely different goals. In the first case, the goal is resident in the movement itself, whereas it is remote in the other two cases. Remote goals can in many cases be realized through a number of different movements. For instance, one can open a window either by raising one's arm or just by asking somebody else in the room to open the window. Since the same movement can realize several goals and the same goal can be realized through several movements, it is reasonable to assume that goals and movements are represented independently of each other. This leads us to propose a distinction between two basic constituents of action codes, namely, goal codes and movement codes. Further, we propose that action codes subserve two basic functions, evaluation and control. Outcome evaluation is readily explained if one assumes that movement code activation tends to go along with goal code activation and that goal codes serve as a reference for the evaluation of the outcome actually observed. More delicate is the issue of control, that is, how goal codes can playa causal role in movement control. As stated above, this requires the reversal of the links established between movements and goals, that is, working back from goals to movements suited to realize them. The standard solution to this problem has been to postulate anticipatory goal codes and assume that they are furnished with the power to control the execution of movements suited to realizing these goals. In the literature, there have been various attempts to solve this seeming puzzle, like Ach's 'determining tendencies' (Ach, 1905), Hull's 'fractional antedating goal response' (Hull, 1943), or Greenwald's (1970) 'ideomotor mechanism'. These mechanisms share the notion that actions are controlled by anticipatory goal codes which have
Experimental Approaches to Action
175
emerged from previous learning of relationships between movements and their outcomes. Without going into the details of the mechanisms proposed in the literature, the nature of this learning may be illustrated by a simple thought experiment (d. Prinz, 1997b). Suppose that whenever a certain movement is performed, the performing system is capable of keeping track of the movement's resident and remote effects up to a certain level of remoteness. When the same movement is .repeated several times, some of these repetitions will yield similar effects, others will yield different effects. Close, or resident, effects will tend to be more similar to each other across repetitions than remote effects. As a result, the total pattern of movement effects will take the shape of a divergent fan. That fan specifies the variety of effects that have been observed to follow from performing this particular movement, including relatively constant close effects and more variable remote effects. Suppose, furthermore, that divergent fans are available for a large variety of movements, each of them representing possible effects of the respective movement. Naturally, these fans will overlap to a considerable extent. This is because the large number of events can be effectuated by a variety of movements. This is particular true of more remote effects. For example, as mentioned above, the closing of a window can be effectuated through a large variety of movements, including, for example, the raising of one's arm in a particular way-but also, for example, through asking somebody else in the room to close the window. As a result, one can account for the pattern of linkages between movements and their effects in two ways. One way is to conceive it as an assembly of strongly overlapping divergent fans, individuated on the basis of the movements and specifying the variety of possible effects. The other way is to conceive it as an assembly of convergent fans, individuated on the basis of potential action effects and specifying the variety of possible movements that may lead to these effects. These action fans emerge and develop irrespective of how the movements come about. The sole requirement is a powerful network, capable of learning contingencies between movements and effects. Our framework proposes that action codes are built upon knowledge about relationships between movements and effects, as stored in convergent fans. The critical requirement to add is that the linkages built into action fans can be used either way-that is, not only in accordance with temporal order but in the backward direction as well. If such backward computation is warranted, goal codes are furnished with the power to activate their movement codes, that is, elicit movements through which the goals can be attained. 7 7 The structure of action codes may be functional not only in action control but in ordinary perception 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 activated. 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 perception. 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 representational 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 common representational system. On the basis of its inherent representational dimensionality 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 keysa 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 the first trial derives from the fact that no such priming occurs at all. The difference 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 characteristics 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 performance 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 I
,
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 relationships in the interaction between stimuli, goals, and actions as well as the putative operations in the underlying codes. 2.1 Instructed Reactions In most experimental tasks, actions come about as re-actions to certain stimuli, and they do so by virtue of assignment rules as fixed in the experimental instructions. For instance, in a typical response selection task, participants are instructed, to respond, for example, with one of two key presses, R l and R2 , in response to one of two stimuli, 51 and 52' respectively. Once the task set has been implemented this way, individual stimuli, 51 or 52' are then presented in an unpredictable order and participants are required to deliver the appropriate responses as fast as possible. The task thus requires the selection of responses to given stimuli on the basis of a prespecified rule.
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 manual gesture, that is, lifting. The stimulus was always provided by a hand on the screen which was the mirror 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 disappeared. Under symbolic instructions iconic incongruency was still effective but under iconic instructions only a slight effect of symbolic incongruency was preserved. 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 cueing 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 corresponding stimulus gestures as compared to non-corresponding gestures. Further, in one of our experiments we made an attempt at separating two factors that were confounded 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 intermediate 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 performed is completely prespecified and predetermined. Moreover, they suggest a functional role for similarity on (at least) two independent dimensions, one referring 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
In a further set of experiments we made an attempt to study action induction under conditions where there is no explicit task set that demands certain actions in response to certain stimuli. What we wanted to study instead was the spontan- , eous occurrence of movements and their relation to the events going on in the actor's environments. Such spontaneously occurring movements have sometimes been called ideomotor movements (cf. Prinz, 1987). As the term suggests, these' movements have often been discussed in close relationship to ideomotor approaches to action. Still, the descriptive term (to which I refer here) must not be confounded with the theoretical concepts addressed above. Ideomotor movements may, under certain conditions, arise in a person who is observing the course of certain events. Classical examples of ideomotor action are body movements induced by watching other people's actions. For instance, while watching, in a slapstick movie, an actor who walks along the edge of a plunging precipice, observers may often be unable to sit still and watch quietly. They will move their legs and their arms or displace their body weight to one side or the other. Further, as mentioned above, ideomotor movements may also be induced by watching physical events resulting from actions. For instance, while following the course of a bowling ball, people are frequently unable to resist these movement tendencies. The involuntary, or even countervoluntary nature of ideomotor actions has placed them among the curious phenomena of mental life. Moreover, the fact that they are instrumentally completely ineffective makes them even more mysterious (cf. Prinz, 1987). Quite obviously, ideomotor movements fall under the rubric of action induction phenomena. So far, however, it has been an open question how the pattern of body movements that is induced in the observer is related to the course of events that induce them. Basically, two answers to this question have been suggested. The classical answer believes in perceptual induction, that is, induction based on similarity between the events perceived and the movements induced. This answer was already inherent in James's Ideomotor Principle, according to which the mental act of representing certain movements (like perceiving them) will always induce a tendency to perform the same or similar movements. Accordingly, perceptual induction invokes that the observer tends to repeat in her actions what she sees happening in the scene. It considers ideomotor actions a special class of imitative actions-special in the sense of lacking an underlying intention to imitate. A competing answer is offered by intentional induction. This principle relies on intended rather than perceived events. It holds that the observer tends to perform actions that are suited to realizing what he wants to see happening. In other words, he is believed to act in a way that would be suited to reaching certain intended goals if his movements were effective. In a way, then, this principle considers ideomotor actions a special class of goal-directed, instrumental actions-special in the sense of being instrumentally ineffective.
Experimental Approaches to Action
183
We developed a paradigm that should allow us to study the relative contributions of perceptual and intentional induction (Knuf, 1998; Knuf et al., 2001). The task was modelled after the logic of the bowling-ball example. Participants saw a ball moving towards a target on a screen, either hitting or missing it. At the beginning of a trial, the ball was shown at its starting position at the bottom, and the target position was shown at the top. Starting positions and target positions were always chosen such that the ball had either to travel in a north-eastern or north-western direction in order to hit the target. Participants triggered the ball's computercontrolled travel and observed its course. The ball's travel was divided into two periods, instrumental and induction. During the instrumental period (which lasted about one second) participants could manipulate one of the two objects by corresponding joystick movements, depending on condition. In the ball condition, joystick movements acted to shift the ball to the left or the right (after which it would continue travelling in the same direction as before). By this means participants could shift the ball's trajectory and have a chance of hitting the target. In the target condition, joystick movements acted to shift the target to the left or the right in an attempt to give it a chance of getting hit. (Initial motion directions were chosen such that the ball would never hit the target without correction.) We reasoned that this task should allow us to study ideomotor movements occurring during the induction period (which followed the instrumental period and lasted for about two seconds). We examined how joystick movements occurring during this period (where they are no longer effective) were related to the happenings on the screen. Perceptual induction predicts the same pattern of joystick movements for both conditions: they should always point in the same direction as the ball motion (leftwards with the ball travelling north-east, rightwards with the ball travelling north-west). Intentional induction predicts a more complex pattern. First, it leads one to expect that systematic joystick movements should only occur on trials with upcoming misses but not with upcoming hits. On upcoming hits, participants should be able to see, or extrapolate, that the ball would eventually hit the target, so that no further instrumental activity was required to achieve the goal. On upcoming misses, participants should likewise be able to extrapolate that the ball would eventually miss the target-which should then induce ideomotor movements performed in a (futile) attempt to affect the further course of events. More specifically, the details of these attempts should depend on two factors: the object under initial instrumental control (ball vs. target) and the side on which the ball is expected to miss the target (left vs. right misses). In the ball condition (where the ball is under initial control), joystick movements should act to push the ball towards the target (i.e. rightward in the case of a left miss, and leftward in the case of a right miss). In the target condition (where the target is under initial control), joystick movements should act to push the target towards the ball (leftward in the case of a left miss and rightward in the case of a right miss).
184
Wolfgang Prinz
The results of our experiments lent strong support to intentional induction but not to perceptual induction. First, the direction of ball movement (north-west vs. north-east) did not appear to be a major determinant of the direction of induced movements. Second, on trials with upcoming hits, induced movements were virtually absent. Third, on trials with upcoming misses, pronounced induced movements emerged, whose directions were dependent on both the object under initial control (ball vs. target) and the side of the upcoming target miss (left vs. right), exactly in line with the pattern predicted by intentional induction. These findings suggest that, at least in our paradigm, ideomotor movements are much more strongly governed by representations of intended than of perceived events. Further experiments have shown that perceptual induction may in some cases be effective, too. For instance, when one looks at ideomotor movements induced in effectors that are not instrumentally involved in joystick control (like head and foot movements), one sometimes sees perceptual induction, too, suggesting that non-instrumental effectors tend to follow the ball's travelling direction. Intentional induction was, however, also effective in head and foot movements. Accordingly, a more comprehensive view will need to encompass both (weak) perceptual induction and (strong) intentional induction.
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 automatically 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 considerations and experimental observations both suggest a novel conceptual framework 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 special 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 performance', 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
Within the empiricist tradition perception is conceived as essentially passive in comparison with exercise of the active will: Locke writes 'For in bare naked Perception, the Mind is, for the most part, only passive'.l According to pragmatists this contrast is overdone and leads to the mistakes of 'the spectator theory of knowledge' which are to be corrected by giving due weight to the connections between perception and agency: Dewey writes that 'Experience, in other words, is a matter of simultaneous doings and sufferings'.2 As ever within philosophy there are truths on both sides of this dispute; the difficult task is to strike the mean.
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 contradictory 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, 1975; 1st pub. 1689), II. IX. 1. Emphasis in original. 2 J. Dewey, 'The Need for a Recovery of Philosophy' (1917), reprinted in The Essential Dewey, i ed. L. Hickman and T. Alexander (Bloomington, Ind.: Indiana University Press, 1998),49. Emphasis in 3 L. Wittgenstein, Philosophical Investigations (Oxford: Blackwell, 195:3), ii, p. x. original.
Perception and Agency
189
Once I have described what I believe the facts concerning some situation to be, I do not need to add a further description of the situation itself; my description of that was given by my description of my beliefs about it. This way of explaining Moore's paradox reverses the direction of that suggested above while confirming the underlying point. For the redundancy of our self-ascriptions of belief rests on our commitment to the truth of what we believe. This commitment is intimately linked to the role of beliefs as reasons for beliefs and action. Even though what we assume, we assume to be true, we do not take our assumptions to provide us with reasons for belief or action, and this is reflected in the fact that in making an assumption we do not commit ourselves to its truth. Furthermore, although the connection between rationality and truth is most obvious in the case of theoretical reasoning, it is the role of beliefs as reasons for action which is more fundamental to our commitment to their truth. For since the conclusions of theoretical reasonings are just further beliefs, our commitment to the truth of the beliefs we employ as reasons for them remains dependent on our commitment to the truth of the conclusions we draw from them. Thus an explanation of our commitment to the truth of beliefs is deferred rather than grounded by concentrating on their role in theoretical reasoning. By contrast, in the case of practical reasoning we employ our beliefs as premises in reasonings whose conclusions we attempt to put into action. So in this case, our commitment to the truth of these beliefs is not dependent on our commitment to the truth of other beliefs inferred from them, and we can readily understand it as manifested in our willingness to act. This sets up a connection between belief and agency (I shall return to it at the end of this chapter). Equally, however, the redundancy of belief shows that the connection here has to be at arm's length only. For where a thinker regards the truth of a proposition p (that, e.g., the sun will shine tomorrow) as not up to him, he is committed by the redundancy of belief to regarding the truth of the proposition I believe that p as similarly not up to him. For suppose, on the contrary, he could envisage himself deciding to believe that the sun will shine tomorrow while recognizing that his decision will not affect the actual state of the weather. Only a little reflection is required to exhibit such a decision as inherently irrational: the redundancy of belief implies that when deliberately bringing it about that he believes that the sun will shine tomorrow he is doing something which will commit him to the truth of the proposition that the sun will shine tomorrow; but his recognition that his decision can make no difference to whether the sun will shine tomorrow implies that he has no reason for making this commitment. It only follows from this that rational thinkers cannot regard their own beliefs as subject to their will. It does not follow that these beliefs may not in fact be in some respects subject to their will; indeed we are surely all of us prone to wishful thinking. But this fact about ourselves cannot show itself to us. We may acknowledge that it often would suit the internal economy of our feelings to be able to maintain Our belief in something-for example, the belief that our children are honest. But faced with the question as to whether we believe this, the redundancy of
Thomas Baldwin
190
first-person ascriptions of belief forces us to confront the simple question 'are my children honest?', which is a question whose answer is not to be found by appealing to our feelings about the matter, but by reference to the behaviour and dispositions of our children. Thus we have to look to the world to find answers not only to questions about the world, but also to questions about our beliefs.
2
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 perceptions 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 empiricist 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, 4
L. Wittgenstein, Philosophical Investigations, 190.
Perception and Agency
191
haven't I voluntarily broken off the visual impression of it?-But I will be told that that sort of thing isn't called 'banishing the picture of the hand'! Certainly not; but where does the difference lie? One would like to say that the will affects images directly. For if I voluntarily change my visual impression, then things obey my will.
The point here seems to be that although my ability to modify the 'things' I perceive, for instance, my hand, is a way of voluntarily changing what I perceive, this is not what the hypothesis of §79 was supposed to concern. Instead we are supposed to imagine a case in which someone can change their 'visual impressions' without doing so by altering the things they see. In the next three sections Wittgenstein argues that this supposition is incoherent: 92 But what if visual impressions could be controlled directly? Should I say, 'Then there wouldn't be any impressions, but only images'? And what would that be like? How would I find out, for instance, that another person has a certain image? He would tell me.-But how would he learn the necessary words, let us say 'red' and 'round'? For surely I couldn't teach them to him by pointing to something red and round. I could only evoke within myself the image of my pointing to something of the sort. And furthermore I couldn't test whether he was understanding me. Why, I could of course not even see him; no, I could only form an image of him. Isn't this hypothesis really like the one that there is only fiction in tl1e world and no truth? 93 And of course I myself couldn't learn or invent a description of my images. For what would it mean to say, e.g., that I was forming an image of a red cross on a white background? What does a red cross look like? Like this??-But couldn't a higher being know intuitively what images I am forming, and describe them in his language, even though I couldn't understand it? Suppose that this higher being were to say, 'I know what image this man is now forming; it is this: ... '-But how was I able to call that 'knowing'? It is completely different from what we call 'knowing what someone else is imaging'. How can the normal case be compared with the one we have invented? If I think of myself in this case as a third person, then I would have absolutely no idea what the higher being means when it says, with regard to someone who has only images and no impressions, that it knows which images that man has. 94 'But nevertheless can't I still imagine such a case?' The first thing to say is, you can talk about it. But that doesn't show that you have thought it through completely. (5 o'clock on the sun).
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 describing 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 perceptual 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 providing evidence concerning a real world, the argument does not depend on a simple-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 fundamental 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. 3
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 considerations advanced so far that our epistemology has to be based on passive observation. 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 simply 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 methods to test them. All that follows from the conclusion that perception is not subject 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 relationships 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 waysfor 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 perceptual 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 5 6
R. G. Collingwood, The Idea ofHistory (Oxford: Clarendon, 1946),269. Wittgenstein, Philosophical Investigations, 194.
194
Thomas Baldwin
organization, which, once achieved, leads to the experience of the 'dawning of an aspect'. In these cases, then, the will can only switch us between ways of seeing or hearing which it has not itself created. Cases of this kind involve a form of perceptual attention, and this is a more general feature of perceptual consciousness in which the will is often engaged. Again, the basic activity of the will in this area is one of selecting among a variety of perceptual inputs that are independently available to the subject. In simple cases the selection is of one sense-modality, for example, hearing, at the expense of others; but usually it is a matter of concentrating on one situation that is currently perceived in preference to other currently perceptible situations (e.g. attending to a lecture instead of noticing the people passing outside the window). The role of perceptual attention in this respect is central to the perceptual process. Like most organisms we are assailed all the time by sensory inputs which make available to us potential information about our environment and ourselves that far exceeds our limited capacity to extract the implicit information or do anything with it. A perceptual system which lacked a means of concentrating its limited resources on just some of its sensory inputs would be unable to respond coherently in a complex environment; instead of keeping track of a prey or learning from a spoken message it would be continuously distracted by other sights, sounds, and smells. No doubt there are simple organisms which lack this capacity. But I think Naomi Eilan is right to argue that perceptual consciousness comes only with perceptual attention/ that a subject has a 'point of view' only where it is able to attend selectively to its sensory inputs, and thus develop a coherent and ordered representation of those aspects of its current environment in which it locates itself. In many cases, of course, perceptual attention is involuntary: movements at the periphery of the visual field typically 'catch one's eye'; and we all know the infuriating way in which the attempt to conduct sustained conversation at a party can be disrupted by involuntary switches in auditory attention to a conversation that is just audible in the distance. Is it then wrong to think that the will makes a central contribution to perception through its exercise in directing perceptual attention? These cases show that the will is not omnipotent in this area-we can try, but fail, to attend to something. Perceptual attention remains in some respects involuntary, doubtless for evolutionary reasons: our ancestors needed the ability to respond promptly to potential dangers that arise outside the situation currently attended to. None the less, as rational beings we need to be able to harness our capacity for perceptual attention to our own ends, and we do this precisely by developing the ability to direct it ourselves. Learning to take control of our capacity for perceptual attention, in so far as we can, is a bit like learning to control our breathing, in so far as we can. 7 N. Eilan, 'Consciousness and the Self', in The Body and the Self, ed. N. Eilan (Cambridge, Mass.: MIT Press, 1995),337-57.
J. Bermudez, A. Marcel, and
i.
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 thesis 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 follows 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, intuitively, 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'disembodied'. 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 altogether 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 8 9
10 11
Q. Cassam, Self and World (Oxford: Clarendon, 1997). P. F. Strawson, Individuals (London: Methuen, 1959), ch. 2. B. Russell, Collected Papers, viii, ed. 1. Slater (London: George Allen & Unwin, 1986), 150. Ibid. 255.
196
Thomas Baldwin
show that it is altogether independent of an intuitive awareness of our embodiment. My own view is that some such intuitive awareness is implicit in perceptual consciousness-even auditory consciousness-because of the way in which perceptual consciousness incorporates an implicit recognition of the causal role of the physical objects of perception. But spelling this out is a complex and contentious matter and independent of my concerns here. 12 4
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 general, 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 representations 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 representations are such as to enable them to contribute to the determination of behaviour 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 perceptual 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 'subjective' 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 subject 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 subjective 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 purposive 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 sensory states have representational content only in so far as they are capable of contributing 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 contribution 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 something 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 structure 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 perception 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 considering 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 practical 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 substantive 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 classify correctly requires the practical ability in question. Admittedly, in the case of some sophisticated but distinctive pieces of equipment, such as an electron microscope, 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 identify 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 concepts. 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 'common 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 pragmatist 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 circumvented 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 perception 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 constraints 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 Glyn W Humphreys and M. Jane Riddoch
1. INTRODUCTION
Neuropsychological studies have played an important role in guiding our understanding of cognitive function. For example, studies of selective deficits in particular tasks, following brain lesions, provide constraints on models of cognition, demonstrating functional modularity between processes underlying the tasks. In many instances, the deficits reflect impairments to specific forms of stored representation or in gaining access to those stored representations-even when the representations are normally accessed automatically (examples here would include deficits in object recognition and in the recognition of printed and auditory words; see Humphreys, 1999; Coltheart et ai., 1980; Hall and Riddoch, 1997). In other cases, though, the deficits do not reside in automatic processes but rather in control processes that either modulate access to stored representations or that modulate the consequences of the access process. A classic illustration of this is socalled 'utilization behaviour', following damage to the frontal lobes. Patients displaying utilization behaviour may act upon stimuli present in their environment irrespective of the tasks they are asked to perform or even irrespective of the physical constraints of the situation. Lhermitte, for example, placed pairs of spectacles in front of one patient who proceeded to place each pair on his face-even though this involved placing one pair on top of the others (see Lhermitte, 1983)! Such behaviours appear to represent extreme examples of loss of task-based or instructional control over action. Task constraints appear neither to prevent irrelevant stimuli accessing learned associations nor to halt the learned associations from being enacted. These deficits in control processes can provide important information about the nature of intentional behaviour. In this chapter, we will review neuropsychological evidence indicating that the intentional control of behaviour can break down in a number of ways. The dissociations between the different forms of intentional control show that self-agency is not unitary but instead involves a number of separable components. We suggest that it is only by identifying these components that we will begin to understand the nature of intentional behaviour. 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.
2. ACTION SLIPS AND THE INTENTIONAL CONTROL OF ACTION In everyday life, our actions usually correspond to aspects of our general intentional will. The intentional control of action occurs even in instances where psychologists have shown that the behaviour is targeted at a stimulus that the subject 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 processing, 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 disorders, from visual neglect (Vallar et al., 1991), to achromatopsia (Humphreys et al., 1992) and prosopagnosia (Bauer, 1984), where patients cannot make intentional discrimination responses to the stimuli involved. Further examples are balance 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 information 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 intentional 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 circumstances (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) mechanisms, 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 control 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 nevertheless 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).
Glyn W Humphreys and M. Jane Riddoch
204
Intention Stimuli:
Start on familiar Familiar landmarks - - - - - - . . route
~
~~ :r-----------------, Turn left at : Ll~~_:~?_~
c=J Primed by intention
C::J Not primed by intention
j
Do not turn left at junction
Response set FIGURE
9.1. Outline framework for how intention may modulate our taking a novel turn on a familiar road
Supervisory attentional system
Stimulus - - - - e t
Hierarchy of action
Response
Contention scheduling system
FIGURE
9.2. Illustration of the Norman and Shallice (1986) framework
The approach is incorporated into models such as that proposed by Norman and Shallice (1986). Norman and Shallice distinguish between a 'contention scheduling 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 'lateselection' theories of selection (cf. Deutsch and Deutsch, 1963). Late-selection theories propose that all stimuli are processed to a high level, with stored representations 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 stimulusresponse 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 exposition). 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 processing 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 identify 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) proposes 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 provide 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 volitional 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 distinct 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 distinction 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 mortified 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 movements 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 appropriate 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 associated 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 familiarity 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, intentional 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 stimulusresponse 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 contrasting factors. For instance, the left hemisphere (controlling the right hand) seems affected by learned stimulus-response associations (see above). The right hemisphere (controlling left-hand responses) may be dominant under conditions of spatial uncertainty; hence inappropriate left-hand responses were made frequently 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 representations 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 overlearned 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. 4. THE DISTINCTION BETWEEN OBJECT SELECTION AND RESPONSE SELECTION FOR ACTION The errors made by ES are consistent with her having an impairment in taskbased, intentional control of behaviour. In fact, the responses are reminiscent of utilization behaviours made by patients with frontal lobe damage (see above),
210
Glyn W Humphreys and M. Jane Riddoch
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 associated 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 righthand 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 consistent 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' network 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 Stimulus selection
211
Response selection
(Resources)
---1
(Resources)
o FIGURE
Primed by intention
i_-:~
Response
31
Not primed by intention
9.3. Illustration of a framework in which there are separable processes for object selection and action selection
Notes: Object selection is modulated by templates that specify attributes of the target. Action selection is modulated by intentional activation of behaviours in a response set. Different levels of stimulus analysis (perceptual, meaning-based, and so forth) are supplied by independent resources.
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 represented 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' experiments 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 distractor, 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:
FIGURE
9.4. lllustration of 'distractor capture'
Notes: Task requires reaching to a target cup (in black) where the distractor (in white) falls in the reach trajectory.
Source: After Humphreys and Riddoch (2000).
Fractionating Intentional Control
213
selected following correct selection of the target (and activation of the associated response, based on the orientation of the target's handle). If the response to the target was already programmed then the distractor, selected concurrently with this, simply took over the programmed action; distractor errors then arose. These results on both target and distractor errors remain consistent with the distinction we have drawn between, on the one hand, an intentional set for the properties of stimuli and, on the other, an intentional set for possible responses. These two means of implementing intentional behaviour are functionally isolatable following brain damage. 6. ON THE RECRUITMENT OF RESOURCES FOR STIMULUS AND RESPONSE SELECTION We have already discussed the case of patient FK, who had abnormal difficulty in overcoming learned associations between stimuli and actions when required to make relatively novel, task-based responses. This difficulty was apparent not only in the 'cups experiments' but also in a range of other tasks; for example, FK showed large interference effects in the so-called 'reverse-Simon' task in which, ignoring the meaning of a target word (LEFT or RIGHT), he had to press a right or left button according to the word's location on the left or right of a computer screen. In this task his responses were markedly slow when the required response was incompatible with the irrelevant dimension of the stimulus (e.g. making a left button press to the word RIGHT presented on the left of the screen). This is to be expected due to FK's difficulty in intentionally overriding learned response information (the word's meaning) activated by a selected object (the target). Now, earlier we discussed Lavie's (1995) notion that selection can be biased towards early (perceptual) or late (response-based) processes depending upon the difficulty of the task. There is a bias towards early selection under conditions of high task load. Kumada and Humphreys (2002) examined the effects of task load on FK's ability to ignore (select out) irrelevant information in the reverse-Simon task. They contrasted FK's performance when the target word was presented alone, with his performance when the target was presented in a cluttered visual display with irrelevant X's present. The results were counter-intuitive. Rather than worsening FK's performance, the irrelevant X's improved it! FK was better able to ignore the meaning of the target, and to respond selectively to its location, when the X's were present than when they were absent. How can we account for this result? The reverse-Simon task requires that (i) the target is selected, and (ii) responses are assigned according to the target's location and not its associated meaning. We have argued that FK was able to select a target stimulus (part (i», but he was impaired at applying an intentional bias to favour responses to the target's location rather than its meaning. Despite this, his performance could be biased towards location information by increasing the task load. One way of conceptualizing this is in terms of resources being allocated to
214
Glyn W Humphreys and M. Jane Riddoch
enhance different levels of processing of a stimulus: its perceptual properties, its meaning, and so forth (see Fig. 9.3). Increasing the task load recruits resources available for perceptual processing, and consequently there is stronger activation of responses linked to these perceptual representations under load conditions. Our point in raising this result here, though, is to contrast this successful effect of task load on perceptual processing with FK's poor intentional control of response selection. Effects of task load occur in a patient unable to perform intentional selection processes normally. From this we infer that increasing the task load recruits resources for perceptual processing in a relatively automatic manner, irrespective of whether a subject can bias processing by intention. 7. INTENTIONS AND ERROR MONITORING 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 problem 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, however, remains able to detect a flagrant anarchic hand movement because this transgresses even a degraded intentional goal. 8. ARE SOME ACTIONS NECESSARILY INTENTIONAL? We have argued, and presented neuropsychological evidence, that many actions can be effected unintentionally-including, for example, grasp responses to target stimuli. 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 unintentionally. Now, when discussing evidence from the 'cups' experiments, we noted that task-inappropriate responses made by ES's right hand reduced when pointing 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, pointing 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 unimpaired 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 implementation of different intentions takes place through separable neural systems.
REFERENCES BAUER, R. M. (1984), 'Autonomic recognition of names and faces in prosopagnosia: a neuropsychological 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 templates 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 selection 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 SelfRegulation: 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: saccade 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 interference', Cognitive Neuropsychology, 15: 645-84. --HUMPHREYS, G. W., and EDWARDS, M. G. (2000a), 'Visual affordance and object selection', 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 unilateral 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, hemianesthesia, 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 control (Norman and Shallice, 1986) in order to distinguish intended actions from non-intended actions and those from mere movements or happenings. My specific 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 phenomena while automatic actions belong to a different cluster. The former cluster comprises conscious awareness and attention, verbal justification, moral responsibility, 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. 1.1 Causal Theory of Action The classical proposal of a causal theory (Davidson, 1963; Brand, 1984) was that a person's doing something is an action if it is caused by a mental event, the content of which relates appropriately to the performance of the action. Three shortcomings of this theory have characteristically been noted (e.g. Searle, 1983). (1) For many actions there does not seem to be (a consciously aware)intention preceding them. (2) The theory cannot explain the lack of intentional action in cases of causal deviancy. (3) Since the mental state is not part of the action but just a prior causal trigger the theory cannot account for the phenomenological difference between actions and other bodily movements. To solve these problems Searle (1983) has proposed a distinction between prior intentions and intentions-inaction. However, the critical point is that the intentional component identified by intentions-in-action is integral and synchronous to the execution of an action. In fact, there is no real need for two different types of intentions. We only need one 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 differentiation 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 preoccupation 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 intentionally, 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.
2. THE CASE FOR DUAL CONTROL 2.1 Non-intended Actions 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 movement 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 order to differentiate it from Searle's case of 'unintended' action.
220
Josef Perner
But that is not the case I have in mind. My concern is with 'action slips' (Reason, 1979). One of the many examples is the following (p. 73): 'I meant to get my car out, but as I passed through the back porch on my way to the garage I stopped to put on my Wellington boots and gardening jacket as if to work in the garden.' Clearly, putting on one's Wellington boots and gardening jacket is more than mere movement. It is an action, but one that was not intended; not intended at all rather than not intended under a particular description. Non-intended actions of this kind have no place in Searle's analysis. Moreover, if we now contrast this case with the case where the very same action is carried out as absent-mindedly except that the larger (prior) intention was to go into the garden, then the same action counts as intentional. It seems that the experience of action while acting is the same in both cases. If, according to Searle, the experience of acting is determined by the intention in action then we have the curious situation of having to say that the non-intended action slip has as much intention in action as when I intend to go into the garden. To drive this point home, imagine yourself driving home fully engaged in a conversation with your passenger. You absent-mindedly switch gears, steer your car and take the left turn, direction home, at the point where, to the right, you'd turn to the supermarket. Despite doing all these actions quite absent-mindedly (and some would say without conscious awareness) we cannot say that they were done non-intentionally, for you will claim that you meant to do everything the way you actually did. By contrast, on another occasion everything is exactly the same except that you are supposed to stop at the supermarket before going home. However, being deeply engrossed in your conversation, you take the usual left turn, direction home, instead of turning right to the supermarket. I think we can agree that at the time of taking this left turn there was no difference in terms of intention in action, and in terms of phenomenal experience of how it felt (if anything at all) 'to take that left turn' between the earlier occasion when you meant to drive straight home and the occasion when you were supposed to turn right to the supermarket but failed to do so. Only when your spouse asks you for the groceries do you discover your mistake and you will say something like, 'Darn it, I didn't mean to turn left, I wanted to go to the supermarket'. The important point here is that in both cases the behaviour has the same intention in action and, therefore, qualifies as action rather than mere movement, on several grounds. Phenomenally the two actions appear to have the same feel and both seem to be coordinated by the same goal representation as part of motor representations (a la ]eannerod).4 However, there is nothing in Searle's theory that 4 It is natural to assume that 'non-intentional' actions are as much governed by some motor representation 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. 3. A MODEL OF DUAL CONTROL 3.1 Objectives Asatisfactory theory needs to account for several terminological and phenomenal distinctions, and do justice to empirical data. Here is a list of desiderata to be addressed in my proposal: Conceptual and phenomenal distinctions. There are several relevant distinctions used in different research traditions to be integrated in a common framework. For instance, as mentioned, philosophers distinguish mere movements (or events) from actions, which tend to be taken as intentional. But then there is the obvious need to distinguish intentional from non-intentional actions. Animal psychologists (Heyes and Dickinson, 1993) distinguish responses (or habits) from actions. Experimental psychologists contrast automatic with controlled processes (Shiffrin and Schneider, 1977). Neuropsychologists (Shallice, 1988; Frith, 1992; Jahanshahi and Frith, 1998) make a similar contrast between involuntary and willed, controlled or voluntary actions, and social/clinical psychologists (e.g. Kirsch and Lynn, 1999) make a distinction between automatic or non-volitional responses and volitional responses under voluntaryS control. Consciousness and verbal report. The higher level of control is requmng conscious awareness and often enabling verbal reportability. We think we are conscious of and typically able to talk about our intentional, in particular our willed, voluntary actions in order to justify them, but we are often not able to do this for automatic, involuntary actions which often go without conscious recognition. An account of intentional action should give an explanation of these clusters of type of action, conscious awareness, and effability. Tasks requiring executive control. The theory should explain why a series of tasks have been claimed to need higher-level executive control by Norman and Shallice (1986) in their SAS model (supervisory attentional system) and by Baddeley (1996) in his model of the central executive within working memory. Slack in intention judgements. Why are we often unclear and wrong about whether we are or are not authors of actions and events? Contrary to our deepest 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 ;
convictions, increasing evidence of a schism between actual control and con- ; sciously perceived control leads to an unfortunate tendency to see consciousness \ as an epiphenomenon: our conscious intentions are but perceptions based on internal sensations of innervations (Jeannerod, 1997: 181 ff.), or even worse, post hoc concoctions of what we sense ourselves initiating and of what we see ourselves actually doing (e.g. Nielsen, 1963: 'alien hand' experiment).6 Consequently there are theories (e.g. Wegner and Wheatley, 1999) that see conscious will not as an initiator but as an attribution after the fact. These have gained strength on the basis of recent experimental evidence. Action organizing goals can be subconsciously implanted (Bargh and Chartrand, 1999). There is over-attribution of will (Wegner and Wheatley, 1999) if a thought about an event (e.g. making the cursor stop at a particular object) occurs temporally close to that event (cursor stops at the object) even when it is objectively clear that some other person has caused the event (confederate made the cursor stop). There is action projection to external agents of one's own actions as in the case of 'facilitated communication' (Burgess et al., 1998). That is, it can be demonstrated that people emit unconscious behaviours (guide the handicapped person's attempts at communication) that further some goal (give the correct answer to a question) that one doesn't think one is pursuing (one doesn't want to give the answer, just help the handicapped to express it). One objective of the dual control theory helps avoid the conclusion that conscious will is an illusion. It does so by giving enough slack between our conscious intentions (will) and what our action-producing system (automatic control) actually does and yet at the same time preserves the sense that our conscious will sets (is causally responsible for) the frame within which our actions are produced. A complete theory should also account for the sense of self, and of agency (the self in charge of action), cognitive limitation of conscious/willed action, the attribution of moral responsibility for action, and other concerns, which I will not address. 3.2 My Proposal-In Brief 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 limited 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. 6 This is to be distinguished from the neuropsychological 'alien/anarchic hand' syndrome (Della Sala et aI., 1994; Parkin and Barry, 1991) where one of the patient's hands behaves in an unruly way in conflict with what the other hand does.
The Case ofNon-intentional Action
223
The functioning of the higher and lower systems and their interaction are characterized by the following. (1) The higher level implements action schemata at the lower level. (2) The action schemata of the lower level autonomously initiate and control the execution of actions. (3) The execution of action is checked against the representations at the higher level (monitoring). This model owes much to the dual control system proposed by Norman and Shallice (1986) with the important difference that it overcomes one of the critical shortcomings in specifying the higher-order system (SAS: supervisory attentional system). As Baddeley (1996)-who sees his notion of central executive in working . memory as a close relative of the SAS-has pointed out, the SAS is theoretically void and tacitly assumes the existence of some homunculus. My proposal tries to fill that gap in terms of a difference in explicitness of representation (predication implicit vs. fact explicit) and in type of control (vehicle vs. content control).
The Lower Level Actions are governed by action schemata. As a concrete example, I take a card sorting test for children (Zelazo et aI., 1996). For instance, the child has been told that cards with something red on them have to be filed into the left box, cards with something green on them into the right box. Once this action schema has been established, the child will sort cards correctly if he or she has the goal to sort cards. Importantly, action schemata are not just stimulus-response associations. They also depend on the activation of some goal. In fact, even control of actual hand movements to a target shows this complexity. As Jeannerod (1997) suggested, schemata (motor representations) not only represent the external conditions but the goal of the action and the bodily movements to achieve that goal. There must be monitoring of the action by computing deviations of the unfolding movement from the predicted path by a forward model of how bodily force translates into movement (Wolpert et aI., 1995; as cited by Jeannerod, 1997, fig. 6.3). One characteristic of action schemata is the way in which they represent the environmental conditions, goals, motor responses, and the link between these parts. They represent it predication implicitly-a notion that Zoltan Dienes and I (Dienes and Perner, 1999) have introduced to capture a particularly important facet of how knowledge can leave various aspects of a known fact implicit. I present here a refinement of the basic idea applied to the simple action schema, 'green card: put it into left box'. This schema can be in two states. It can be tacit (nonactivated) or occurrent when it is activated by a green card being presented. Let me at first concentrate just on its antecedent condition of a green card being presented. In its tacit state the 'green card' can be said to explicitly represent the property of being a green card, since there is a clear distinction between the representation 'green card' and, say, 'red card'. There is no representation of any instances that have the property of being green cards. The fact that there are instances of something being green cards, one might suggest, is represented when the antecedent condition
224
Josef Perner
gets activated by a presentation of an actual green card. And one might suggest, that the fact that 'green card' gets activated-rather than 'red card'-represents the fact that being a green card is predicated of the presented instance. However, if one takes that view one has to realize that representing the instance of something being presented and the predication of the property of being a green card to this instance remains indistinguishable from just thinking about a green card without one being presented. In other words, the predicating being a green card of a presented instance remains implicit in the causal effects of the presented instance on the mind-brain I (representational vehicle). We call it predication-implicit because there is no separate distinction in the representational medium that serves the function of representing the fact that a predication is being made. Similarly the entire action schema 'green card: put it into left box' can be said to represent the complex event property of green card being put inside the left ' box. However, the fact that there are particular instances ofwhich this property is being predicated is left implicit in the causal effects that presentation of a green card has on the activation of the schema and the schema on the execution of the desired action. Also the fact that these actions are desired events is left implicit in the causal history of reasons for why this schema was set up. Another feature of action schemata is the kind of control they use. An established action schema passes control from the antecedent condition to the consequent, that is, the action representation. It does so purely on the grounds of an associative link that has been formed in the representational vehicle. Although one can say that this link has representational content, that is, it represents the existing action demands on the organism, and that this content is responsible for the existence of the link, this content need not be evoked for routing control once the link has been established in an existing action schema. This kind of control I have called vehicle control (Perner, 1998). And I define the lower level of control as being restricted to predication-implicit representations and to vehicle control. Norman and Shallice (1986) in their dual control model include at the lower level not only control within each schema but also control among schemata (contention scheduling). Schemata are connected by inhibitory links to competing schemata and by excitatory links to supportive schemata. The inhibitory links are designed to make sure that the dominant, most activated schema is not interfered with by other schemata that might get some activation. The important point here is that once these across schema connections are in place, it is the connections among the schemata as representational vehicles that suffice for implementing this contention scheduling. Now let us look at the learning through conditioning of new schemata and what kind of control that requires. A general learning mechanism typically functions in such a way that if a haphazard response leads to success, then the learning algorithm strengthens those associations that were responsible for the successful response, that is, the link that associated the green card with the action of putting it into the left
The Case ofNon-intentional Action
225
box. Other links not exercised on this last successful trial, like associating the green card with putting it into the right box, become diminished by the learning algorithm. Such a learning algorithm exerts control of a higher order in that it changes the control structure implemented by the individual action schemata. It directs its control of strengthening or weakening schemata on the basis of whether these schemata had been active or not at the time of the last, successful or unsuccessful, action. Having been activated recently is a feature of the representational vehicle independently of its content. Now, again, one could argue that the learning mechanism chooses on the basis of recent activation because that vehicle feature represents that the action represented by the schema is most likely to be responsible for the successful outcome. Hence it directs its control on the basis of the representational content of having been recently activated. And again, although this is true for evolution that developed learning algorithms, it is irrelevant once the algorithm is in place. Once in place it directs its control on the basis of the vehicle feature of recent activation regardless of what that feature might indicate about the world. In sum, the control exercised by established action schemata, the contention scheduling among competing schemata, and the acquisition of new schemata through feedback learning are based on predication-implicit knowledge and vehicle control.
The Higher Level
The higher level is defined by the necessity to entertain predication- and factexplicit representations and to exercise content control over the lower level. Paradigm examples that require the higher level of control are the acquisition of a novel schema through verbal instruction, planning, reasoning, or hypothesis testing. For instance, when given verbal instructions: 'Put the green cards into the left box', the child has to represent the meaning of these instructions. This cannot be done in a predication-implicit way that makes the predication of represented properties (e.g. 'green card') dependent on the causal effect of a presented instance, since there is no instance being presented. Hence the representation of these instructions needs to be predication-explicit, making reference to hypothetical presentations of instances of green cards. That also means that not only predication to instances but also the fact that these are not real but only hypothetically considered instances needs to be made explicit. Making the reality status (fictive vs. factual) explicit we have called fact-explicit (Dienes and Perner, 1999). The same explicitness is, of course, also required for planning, reasoning, and entertaining hypotheses before one can come to a conclusion which action sequence best to employ. Implementing verbal instructions, implementing actions as a result of reasoning and planning, and testing hypotheses all require content control. For instance, when given verbal instructions to put every green card into the left box, one has
226
Josef Perner
to install a new (or sufficiently strengthen an existing weak) action schema by strengthening the association between the input condition 'green card' and the action 'put card into left box'. As in the case of learning by conditioning, there needs to be higher-order control. In the case of conditioning, the higher-order control problem consists in translating the feedback information (+, -) into strengthening certain schemata and weakening others. As discussed above, the appropriate schemata can be found on the basis of vehicle features, for example, strongly or weakly activated during the production of the last action. In the case of verbal instructions, however, the higher order control problem consists of translating the verbal instructions into installing or strengthening a particular action schema. The important difference is that-unlike in the case of conditioningthere are no pure vehicle features that would allow the higher-order control process to find the correct schema. The only target specification is the schema with the content specified by the verbal instructions. Control that needs to identify its target in terms of content I call content control. Since content in itself has no causal power, it cannot direct any control. What is required is a provision in the representational vehicle that allows identification of different representational vehicles in terms of sameness of their content. Content control is also required for implementing new schemata on the basis of reasoning, planning, or hypothesis testing, since the fact-explicit representation of hypothetical events needs to be translated into corresponding action schemata, of which the parts are only identified by their representational content.
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 changing 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 triggering 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.
4. MEETING THE OBJECTIVES 4.1 Conceptual and Phenomenal Distinctions Different types of movement and action can be distinguished depending on whether the actual action produced by the lower level does or does not conform
The Case ofNon-intentional Action
227
to the higher level's prescriptions, and whether the higher level does or does not monitor the lower level: Movement: A bodily movement not due to an action schema, for example, I slip with my elbow off the table. Response/Habit: 7 Movement governed by a schema that is goal oriented but there is no goal representation involved. For instance, if a rat learns to turn right at the T-maze junction because it gets rewarded with food at the end of that branch, then that behaviour is goal oriented (McFarland, 1989, used the term 'goal .seeking') in the sense that the rat does turn right because it leads to a reward. However, the rat could turn right at this junction without any representation of the food waiting at the end (though there are data from goal devaluation studies (Heyes and Dickinson, 1993) suggesting that rats do have goal representations for instrumentally conditioned responses but not for innate ones, e.g. scratching when it itches). Action: Movement due to action schemata governed by a represented goal. Non-intentional (automatic) action: The lower system produces an action that is not envisaged by the higher order system, for example, driving straight home when one wants to go to the supermarket before going home. Intentional action: The higher system sets up the lower system. When the external conditions arise, the lower system executes the appropriate action. (a) Absent-minded (automatic) intentional action: The higher system is not involved in triggering the action or in monitoring its execution, for example, driving home as planned but deeply engrossed in conversation thereby not consciously registering any of the driving actions. (b) Conscious (automatic) intentional action: The higher system is not involved in triggering the action but it does monitor its execution, for example, driving home and taking quick appropriate action that registers after the fact. (c) Willedlcontrolled/voluntary action: The higher system represents for the particular case what ought to be done, triggers the schema, and monitors the execution. This constitutes a case of'willed or controlled action' (e.g. Jahanshahi and Frith, 1998), for example, I have to lift my arm at some self-determined time; or I say to myself in my mind, 'this is the crossing to the supermarketI have to turn right', and I do turn right. 7 In animal psychology (e.g. Dickinson, 1994) the term 'habit' is used for a stimulus-response association 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.
Josef Perner
228
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 controlled 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 consciousness. 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 promiscuously 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 selfascription 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 consciously 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 blindsight. 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 higherorder 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 information 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 contrast, 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 blindsight 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 perspective 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 consciousness 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 perspective. 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. Factexplicitness 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 dreaming 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 higherorder control, we have gained a justification for the intuition that voluntary control 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 conscious 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 distinguishing <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 higherorder 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 firstorder 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 COlllmunicate 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 practical understanding of different experiences of the world. So we see that the 1'/zorder-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 requirements 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 representation (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 consciousness. 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 children. I kept singing while thinking about more important matters. When interrupted 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 representational 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. 4.4 Executive Control Dual control systems are suggested by neuropsychological research. Starting with Luria (1966) and Bianchi (1922), it was found that patients with frontal lobe insult experience a loss of control over their actions (Milner, 1964) and have difficulty with certain tasks. Norman and Shallice (1986) interpreted these difficulties as resulting from the impairment of the higher level of control within their dual control model where automatic control by contention scheduling is contrasted with willed control by the supervisory attentional system (SAS). I therefore refer to the kinds of tasks that Norman and Shallice list as requiring the higher level as' SAS tasks'. In fact, it is this list of impairments that gives empirical meat to their claim about the existence of the SAS. SAS 1. SAS 2. SAS 3. SAS 4.
planning/decision-making troubleshooting novel/ill-learned action sequences dangerous or technically difficult actions
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 reportersdue 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
SAS 5. overcoming strong habitual response tendencies or temptation (e.g. Stroop) A similar approach has been taken in the working memory tradition (Baddeley and Hitch, 1974) where a central executive (CE) is identified as the coordinating agent within working memory and which is considered as a close relative of the SAS (Baddeley, 1996). The nature of the CE, too, has been characterized by a list of tasks to which I refer as the 'CE tasks' (Baddeley, 1996; Baddeley and Della Sala, 1998): CE 1. CE 2. CE 3. CE 4.
coordination of two independent tasks (dual task performance) generation of random numbers attention switching and focusing retrieval and manipulation of information from long-term memory
I have specified the two levels of control independently on theoretical grounds. In particular, the higher level was specified as that level which requires factexplicit representation and content control. I now need to show that fact-explicit representation and content control is required for SAS and CE tasks in order to explain why these tasks become problematic when the higher-level control system is impaired. To this end SAS and CE tasks can be taken together and regrouped into three categories for easier discussion of relevant features. 1. Some tasks require that new action schemata be established or that established tasks be given additional support if not well enough established (SAS 3: novel or ill-learned) or because the established schema is not strong and precise enough for the purpose on hand (SAS 4: dangerous or technically difficult). Vehicle control becomes successful if the relevant connections have been firmly established for each task and once the appropriate relationship among tasks has been established. Hence, in all cases where the established relationship among tasks is not adequate, additional content control is needed. This also applies to the first CE problem, coordination of independent tasks. Although each component task is established and could run by itself on vehicle control, the task of coordinating the two established tasks is novel (or else it would be a single task) and requires content control. 2. When established tasks do not function correctly (SAS 2: troubleshooting) or interfere with what one tries to do (SAS 5: overcoming habits) then one needs additional, content-directed control. The same is required for CE2 and 3: random number generation and attention switching. Both these cases require content control in order to stay away from the established, i.e. randomness requires avoidance of following any regularities and switching attention consists in overcoming the forces that would naturally maintain attention on the established task. 3. Planning and decision-making (SAS 1) consist of entertaining and generating predication- and fact-explicit representations of possibilities and then picking one possibility. Moreover, the generation of possibilities is supposed to be content
236
Josef Perner
governed, that is, projection of unusual future states that can be reached from the present state by possible actions. It needs to be more creative and go beyond generating the usual train of thought (which can be vehicle controlled). As we see, the tasks that are deemed to require executive control are tasks that need content control and, consequently, according to argument, they need factand predication-explicit representations which are accessible to conscious awareness and verbal report. 4.5 Slack in Intention Judgements Why are we often unclear and wrong about whether we are or are not authors of actions and events? Searle's intention-in-action theory leaves little room for such errors. In his theory, every intentional action has its intention as an integral part and so leaves little room for the existence of non-intended actions or for misattribution of agency to ourselves when we were not the authors of an act (e.g. Wegner and Wheatley, 1999). The dual control model gives more room for non-intended actions and more room for misattribution, since the lower system is given leave to operate on its own within the specifications stipulated by the higher system. The attribution of agency, whether it was me who intentionally instigated and .executed this action, must be based on a judgement whether the action and its results are in line with the higher level's stipulations. The relevant information may be a mix of direct innervation sensations (as reviewed by Jeannerod, 1997) and proprioceptive and visual sensory feedback about the action. So, for a willed/controlled action, where the higher system gives the impetus for action and closely supervises the action execution, misattribution errors are rare. And indeed, such errors can be experi- . mentally demonstrated only with the help of most unusual manipulations of visual feedback as in Nielsen's (1963) 'alien hand' experiment, or by drawing attention away from the action execution by various tricks as in Wegner and Wheatley's (1999) Ouija board experiment.
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 perception has come under strong criticism (Armstrong, 1991; Burge, 1991) as overintellectualizing 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 selfreferentiality 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 supporting particular schemata. The higher level thus implements (or helps implement) 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 predicationimplicitly 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 12 An implementation schema could be considered a procedural (predication-implicit) counterpart to implementation actions and implementation intentions studied by Gollwitzer (1999).
238
Josef Perner
of control (e.g. when S occurs then A needs to be taken in order to achieve 0) and takes the action of installing a functional procedure (S-A-O) with the goal that this procedure then produces the action A at the appropriate time. Now this implementation schema has as part of its content that the implemented action schema be causally responsible for actions. If not causally self-referential, it is causally referential.
5.2 Intentions in Action and Causal Deviancy I want to briefly elaborate the example of causal deviancy given by Searle to show that it then provides a difficulty for Searle's theory of intentions in action but can be accommodated by the dual control model. The example was of the young man who wants to kill his uncle and does so unintentionally even though it was the preoccupation with his desire to kill that was causally responsible for running over the pedestrian that happened to be his uncle. The intention-in-action theory can account for the lack of intentionality because there is no intention-in-action that has as its content that the particular driving action is to achieve the desired goal. The dual control model can account for the lack of intentionality because the driving actions were not governed by an action schema that was implemented by the young man's higher-level desire to kill his uncle. Now to distinguish between the theories I elaborate the example. Assume our young, uncle-hating man is very methodical. He knows how difficult it is to coldbloodedly kill a relative. So he registers in a course for professional killers. Following course wisdom, he then practises for months on foggy nights in front of his uncle's favourite pub to run over his uncle with his car. After reaching perfection, he stalks his uncle to the pub with the intention of carrying out his well-practised routine on the way home. While in the pub he has a complete change of heart, sees the error of his ways, and-in my terminology-disavows his well-practised killing schema. Then on the way home he spots his uncle and, despite having been disavowed-the by now automatized killing routine takes over and the uncle is dead before the young man fully realizes what he has done. This was not an intentional killing. This lack of intentionality is difficult to explain on the basis of intentions-in-action since the very same act would count as intentional if the young man had not had a change of heart in the pub. It is not intentional on the dual control model, since the uncontrolled action schema has been disavowed by the higher-level control system. Well-practised action often precedes our thoughts as nicely documented by Wakefield and Dreyfus (1991: 265), quoting the Boston Celtic basketball player Larry Bird describing the experience of passing the ball in the heat of a game: 'A lot of times, I've passed the basketball and not realized I've passed it until a moment or so later.' The difference from our unfortunate, hypothetical young
The Case ofNon-intentional Action
239
man is that Larry Bird's passing throws conform to his declarative intentions and, therefore, they count as intentional.
6. SUMMARY
My main concerns were to account for the existence of non-intended actions within a causal theory of action. A dual control model can satisfy this concern. Intentional action is defined by the match between what the lower level produces and what the higher level stipulates should be done. The higher level of control I defined as requiring predication- and fact-explicit representations and exercising content control (changes in the flow of control are aimed at vehicles with a particular representational content) and the lower level as control by representational vehicle. The need for predication- and fact-explicit representations provides an account for why intentional action is seen as conscious and verbally reportable. To get a tighter fit between consciousness, verbal report, and content control I argued for a minimalist 11/2-order-thought theory of consciousness, which makes the perspective-bound meta-knowledge provided by predication- and fact-explicit representations (instead of a full-blown second-order thought) sufficient for consciousness.
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 frantomesial 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 development and cognition in human infants: Development and functional specialization of the cortex, Agelonde, France, September 2000.
The Case ofNon-intentional Action
241
GOLLWITZER, P. W. (1999), 'Implentation intentions: strong effects of simple plans', American Psychologist, 54: 493-503. GORDON, R. M. (1995), 'Simulation without introspection or inference from me to you', in M. Davies and T. Stone (eds.), Mental Simulation: Evaluations and Applications. Oxford: Blackwell, 53-67. HAAKE, R. J., and SOMERVILLE, S. S. (1985), 'Development of logical search skills in infancy', Developmental Psychology, 21: 176-86. HEYES, c., and DICKINSON, A. (1993), 'The intentionality of animal action', in M. Davies and G. W. Humphreys (eds.), Consciousness: Psychological and Philosophical Essays. Oxford: Blackwell, 105-20. JACOBY, 1. 1. (1991), 'A process dissociation framework: separating automatic from intentional uses of memory', Journal ofMemory and Language, 30: 513-41. JAHANSHAHI, M., and FRITH, C. D. (1998), 'Willed action and its impairments', Cognitive Neuropsychology, 15: 483-533. JEANNEROD, M. (1997), The Cognitive Neuroscience in Action. Oxford: Blackwell. KIRSCH, 1., and LYNN, S. R. (1999), 'Automaticity in clinical psychology', American Psychologist, 54: 504-15. LABERGE, S. (1985), Lucid Dreaming. Los Angeles: Tarcher. LURIA, A. (1966), Higher Cortical Functions in Man. New York: Basic Books. MCCORMACK, T., and HOERL, C. (1999), 'Memory and temporal perspective: the role of temporal frameworks in memory development', Developmental Review, 19: 154-82. McDONOUGH, 1., MANDLER, J. M., McKEE, R. D., and SQUIRE, 1. R. (1995), 'The deferred imitation task as a nonverbal measure of declarative memory', Proceedings ofthe National Academy of Sciences USA, 92: 7580-4. McFARLAND, D. (1989), 'Goals, no-goals and own goals', in A. Montefiori and D. Noble (eds.), Goals, No-goals and Own Goals: A Debate on Goal-Directed and Intentional Behaviour. London: Unwin Hyman, 39-57. MASANGKAY, Z. S., MCCLUSKEY, K. A., McINTYRE, C. W., SIMS-KNIGHT, J" VAUGHN, B. E., and FLAVELL, J, H. (1974), 'The early development of inferences about the visual percepts of others', Child Development, 45: 357-66. MELTZOFF, A. N. (1988), 'Infant imitation after a I-week delay: long-term memory for novel acts and multiple stimuli', Developmental Psychology, 24: 470-6. MILLIKAN, R. G. (1984), Language, Thought and Other Biological Categories. Cambridge, Mass.: MIT Press. MILNER, B. (1964), 'Some effects of frontal lobectomy in man', in J. M. Warren and K. Akert (eds.), The Frontal Granular Cortex and Behaviour. New York: McGraw-Hill, 313-31. NAGEL, T. (1974), 'What is it like to be a bat?', Philosophical Review, 4: 435-50. NIELSEN, T. (1963), 'Volition: a new experimental approach', Scandinavian Journal of Psychology, 4: 225-30. NORMAN, D. A., and SHALLICE, T. (1986), 'Attention to action: willed and automatic control of behavior'. Center for Human Information Processing Technical Report No. 99. Reprinted in revised form in R. J. Davidson, G. E. Schwartz, and D. Shapiro (eds.), Consciousness and Self-Regulation, iv. New York: Plenum, 1-18. O'SHAUGHNESSY, B. (1991), 'Searle's theory of action', in E. Lepore and R. Van Gulick (eds.), John Searle and his Critics. Oxford: Blackwell, 271-87.
242
Josef Perner
PACHERIE, E. (2000), 'The content of intentions', Mind and Language, 15: 400-32. PARKIN, A. J., and BARRY, C. (1991), 'Alien hand sign and other cognitive deficits following' ruptured aneurysm of the anterior communicating artery', Behavioural Neurology, 4: 167-79. PERNER, J. (1991), Understanding the Representational Mind. Cambridge, Mass.: MIT Press. I A Bradford book. --(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, 270-83. --(1999), 'Metakognition und Introspektion in entwicklungspsychologischer Sicht: Studien zur "Theory of mind" und "Simulation"', in W. Janke and W. Schneider (eds.), 100 Jahre Institut fUr Psychologie und Wiirzburger Schule der Denkpsychologie. Gottingen: Hogrefe, 411-31. English Version on internet: http://www.sbg.ac.at/psy/people/perner/ docs/wuerzburg.doc. --and DIENES, Z. (2003), 'Developmental aspects of consciousness: how much theory of mind do you need to be consciously aware?', Consciousness and Cognition, 12: 63-82. --and RUFFMAN, T. (1995), 'Episodic memory an autonoetic consciousness: developmental evidence and a theory of childhood amnesia'. Special Issue: Early memory. Journal of Experimental Child Psychology, 59 (3): 516-48. PIAGET, J. (1937), The Construction of Reality in the Child. New York: Basic Books. REASON, J. (1979), 'Actions not as planned: the price of automatization', in G. Underwood and R. Stevens (eds.), Aspects of Consciousness. London, New York, Toronto, Sydney, and San Francisco: Academic Press, 67-89. ROSENTHAL, D. M. (1986), 'Two concepts of consciousness', Philosophical Studies, 49: 329-59. - - (2000a), 'Consciousness, content, and metacognitive judgments,' Consciousness and Cognition, 9: 203-14. - - (2000b), 'Sensory qualities, consciousness, and perception', unpublished manuscript, City University of New York. SEARLE, J. (1983), Intentionality. Cambridge: Cambridge University Press. SHALLICE, T. (1988), From Neuropsychology to Mental Structure. Cambridge: Cambridge University Press. SHATZ, M. (1978), 'Children's comprehension of their mothers' directive questions', Journal of Child Language,S: 39-46. SHIFFRIN, R. M., and SCHNEIDER, W. (1977), 'Controlled and automatic human information processing: II. Perceptual learning, automatic attending, and a general theory', Psychological Review, 84: '127-90. TYE, M. (1995), Ten Problems of Consciousness: A Representational Theory of the Phenomenal Mind. Cambridge, Mass. and London: MIT Press. WAKEFIELD, J., and Dreyfus, H. (1991), 'Intentionality and the phenomenology of action', in E. Lepore and R. Van GuIick (eds.), John Searle and his Critics. Oxford: Blackwell, 259-70. WEGNER, D. M., and Wheatley, T. (1999), 'Apparent mental causation: sources of the experience of will', American Psychologist, 54: 480-92.
The Case ofNon-intentional Action
243
WILLATS, P. (1997), 'Beyond the "couch potato" infant: how infants use their knowledge to regulate action, solve problems, and achieve goals', in G. Bremner, A. Slater, and G. Butterworth (eds.), Infant Development: Recent Advances. Hove, East Sussex: Psychology Press, 103-35. 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. WOLPERT, D. M., GHAHRAMANI, Z., and JORDAN, M. I. (1995), 'An internal model for sensorimotor integration', Science, 269: 1880-2. ZELAZO, P. D., FRYE, D., and RApus, T. (1996), 'An age-related dissociation between knowing rules and using them', Cognitive Development, 11: 37-63.
11
The Development of Young Children's Action Control and Awareness Douglas Frye and Philip David Zelazo During the first half-dozen years of life, there are regular-and often remarkable-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 perform 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 possibility 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 intentions 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 potential 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 implications for children's intentional action across a wide variety of situations.
1. AN EXAMPLE OF AN EARLY AGE-RELATED
CHANGE IN INTENTIONAL ACTION A task that requires a small set of well-defined responses can help to expose how young children's action control and awareness develop because changes in skill can be held constant. One task devised for this purpose is the dimensional change card sort (DCCS). The DCCS is an executive function task that bears some resemblance to the Wisconsin Card Sorting Task (WCST) which has been used extensively with adults and older children to detect differences in functioning associated with damage to areas of the prefrontal cortex. Unlike the WCST, however, the DCCS is a rule-use task rather than a rule-learning task. As such, children are explicitly told which rules to follow; thus, to succeed, they must simply keep the rules in mind and act upon them. The DCCS presents children with two target cards-for example, a red triangle and a blue circle-each of which is affixed to a tray. The children are asked to match a series of test cards to the targets. The test cards-for example, blue triangles and red circles-are designed to match a different target when they are sorted by one dimension (e.g. colour) than when they are sorted by the other (e.g. shape). Children are told the correct rule for sorting on every trial. During a preswitch phase, they are first encouraged to match a small number of cards according to one dimension. Then, during a post-switch phase, then are asked to switch to the other dimension. So, for instance, they may be introduced to the colour game first and told on each trial, 'If it's red, put it here; if it's blue, put it there'. Each time they are then presented with a test card that they are allowed to sort. After several cards, they would be instructed to switch to the shape game and told • each time, 'If it's a triangle, put it here; if it's a circle, put it there'. The primary developmental change found in this task occurs between 3 and 5 years of age (Frye et al., 1995). Whereas 5-year-olds switch their sorting, 3-year-olds typically continue to sort by the first dimension even when they are told the rules for sorting by the second dimension on every trial. This pattern occurs regardless of which dimension is presented first, and it is not limited to the dimensions of
246
Douglas Frye and Philip David Zelazo
colour and shape. Nor does it seem to depend on a specific wording of the rules. Finally, the developmental pattern agrees with a variety of other executive function tasks that, like Simon Says, do not involve sorting cards (Carlson and Moses, 2001). At first glance, the limits in action control exhibited by the young children in the DCCS seem to be compatible with the inhibition account mentioned previously. The 3-year-olds could know what they want to do but be unable to inhibit the outdated, prepotent response (Dempster, 1992; Diamond and Taylor, 1996). In Simon Says, the prepotent response is to comply with the instructions being given. It could then be difficult to inhibit that prepotent response tendency when an instruction is not preceded by the authorizing phrase 'Simon says'. The preswitch responses in the DCCS are not prepotent prior to the task, but they become so during the procedure: whichever game children are asked to play first (i.e. sorting by colour or shape) becomes prepotent and then dominates the child's actions during the post-switch phase. There are other findings with the DCCS that seem to conform with an inhibition account. The DCCS reveals a distinct abulic or knowledge-action dissociation (Zelazo et al., 1996). After 3-year-olds have perseverated on the first dimension of the card sort, they are none the less able to demonstrate an understanding of the rules that they fail to use. For example, when children persist in sorting by colour even though they have been told the correct rules to sort by shape, they can be asked to point to answer the questions, 'Where do the triangles go in the shape game? And where do the circles go?' Three-year-olds almost invariably respond to these knowledge questions correctly. Nevertheless, they continue to perform the incorrect sorting action. When they are told to go ahead and sort a given card according to the shape game ('Okay, good, now play the shape game. Where does this triangle go?'), they still perseverate and incorrectly match according to colour. This dissociation appears to provide the most compelling support yet for the hypothesis that young children understand what they are meant to do on the DCCS but are unable to control their actions accordingly. Specifically, the dissociation data seem to indicate that children know the response they should make but instead carry out the old, prepotent one, as if they know exactly what they are supposed to do, and even try to do it, but simply cannot inhibit their prepotent sorting responses. According to this interpretation, the ability to overcome the prepotent response would not depend on changes in children's intentions (i.e. their understanding of what they should do). Rather, it would depend on the strengthening, or even the establishment, of an inhibitory mechanism (e.g. White, 1965). Despite the prima-facie appeal of a response-inhibition interpretation of the dissociation effect, there are several other findings from research with the DCCS that question its validity, or at least suggest that inhibition cannot apply at the level of the response. First, notice that 3-year-olds switch flexibly between the two pre-switch rules. For example, when sorting in the initial game, they may sort
Children's Action Control and Awareness
247
several red cards into one box on one side and then have no difficulty switching to sort a blue card into the box on the other side. In fact, the positions of the target cards on the trays can be exchanged during the first game, and 3-year-olds will continue to sort without missing a step. These observations indicate that however the 3-year-olds are construing their responses, it is not simply on the level of 'this card goes on the left' because they are not bound to that response when either the test card or position of the initial target cards changes. These observations also show that a simple response-inhibition account cannot be correct. Sorting several cards to one side ought to establish a prepotent response, and a simple inhibition . account would predict that 3-year-olds should perseverate, putting all of the cards into the same box. However, 3-year-olds do not perseverate on a single response or a single rule. Instead, from pre-switch to post-switch, they perseverate on a pair of rules for sorting according to a dimension (e.g., colour or shape). There are additional findings regarding the abulic dissociation in the DCCS that further constrain any interpretation of the phenomenon. Although verbal responses may be intrinsically more flexible than manual ones (Luria, 1961), so that it could be suggested the dissociation depends on a change in response modality, in fact, it is not necessary for the sorting and question answers to be in different response modalities for the dissociation effect to occur. Rather, the DCCS can be modified so that the children only make verbal responses indicating where each card should be placed. In these circumstances, 3-year-olds still perseverate after the game changes, and they still give correct verbal answers to the dissociation questions (Zelazo et al., 1996). This pattern rules out the simple inhibition explanation that the dissociation effect only occurs because the change in modalities allows the child to bypass the original prepotent responses. One further set of observations that employs an error-detection procedure with the DCCS raises even more serious worries about the adequacy of children's understanding of what to do in the task. Instead of having 3-year-olds sort the cards themselves, Jacques et al. (1999) had them observe a puppet playing versions of the DCCS, and asked them to evaluate the puppet's performance. Thus, it was possible to determine if the children knew what should be done without requiring that they execute the necessary actions. The experiment was arranged so that when the rules were changed to the new game, the puppet either perseverated on the old dimension or switched correctly to the new one. In a third condition, the puppet switched on its own even though the new rules had not yet been issued. The results showed that when the puppet perseverated, 3-year-olds incorrectly said it was doing the right thing. When the puppet switched correctly, they incorrectly said the puppet was doing the wrong thing. However, when the puppet switched on its own (i.e. in the absence of a rule change), they correctly said that the puppet was wrong. An additional comparison showed that individual children's evaluation of the puppet's performance was related to their own proficiency on the usual DCCS.
248
Douglas Frye and Philip David Zelazo
The results of the error-detection study make it unlikely that 3-year-olds know perfectly well what to do, but just have difficulty stopping themselves from executing a prepotent response in the DCCS. If they knew what to do, then they should have been able to detect the puppet's errors. Further, if young children fail the usual DCCS because their tendency to repeat well-established actions interferes with the formulation or execution of new ones, then eliminating the requirement that children sort cards on the pre-switch phase should have allowed them to judge successfully on the post-switch phase. Thus, these results are a strong sign that inhibitory control of responses is not what is at issue in the task. Whatever change occurs that allows children to switch their actions in these circumstances, it does not seem to be simply a matter of acquiring the ability to execute appropriately formulated intentions. These results lead us back to the possibility that the3-year-olds do not understand what they should be trying to do in the DCCS, despite their answers to the knowledge questions in the dissociation study. Indeed, the error-detection pattern suggests that the 3-year-olds' own sorting errors were in accord with what they thought was right. For both the puppet's and their own actions, they seemed to think that continuing to match by the old dimension was the appropriate thing to do. Five-year-olds, in contrast, recognize for both the puppet and themselves that the new rules require a change in action. If this interpretation is correct, then it raises two questions: (1) What are the characteristics of 3-year-olds' intentions that lead them to think they are performing correctly in the DCCS?, and (2) How can their mis-actions be reconciled with their accurate question answering in the dissociation experiment? 2. A COGNITIVE COMPLEXITY AND CONTROL ACCOUNT In the broadest of terms, Simon Says and the DCCS ask or instruct the child to carry out an intentional action. In other words, they set a goal for the child or specify the satisfaction conditions that the child's action should fulfil (Searle, 1983). The child must provide a means to the goal or formulate an intention to act in a way that meets the satisfaction conditions that have been set. These characteristics are what make it useful to analyse executive function tasks, including Simon Says and the DCCS, as instances of problem solving (Zelazo et al., 1997). The specific characteristics of the DCCS will be used to explore what sort of change in intentional action is occurring during this period of development. Subsequently, a particular explanation of the change, taken from Cognitive Complexity and Control theory (Frye et al., 1995, 1998; Zelazo and Frye, 1997, 1998), will be given. The large variety of findings with the DCCS can help to determine what it is about the young children's intentional action that changes. The empirical findings are essential because even with a well-defined task like the DCCS there are many ways in which children could be construing their intentions in it. Their goal in the
Children's Action Control and Awareness
249
task could be as general as 'match the cards' or as narrow as 'get this card in this tray'. Similarly, they could be relying on descriptions of actions that vary from 'indicate this choice' to 'use these two fingers to move the card from here to there'. These specifications are likely to be critical for understanding 3-year-olds' characteristic action patterns. It can reasonably be assumed, probably without contention in developmental psychology, that if adults act intentionally, then 3-year-olds do so as well. Intentional action appears to emerge during infancy (please see Sect. 3). Thus, the explanation of 3-year-olds' characteristic intentional action will not depend on whether or not they can employ means and goals (they can), but rather on whether they can adequately specify the appropriate means and goals in particular problem-solving situations. One possibility is that 3-year-olds in the DCCS are able to adopt exactly the goals specified by the experimenter's instructions. In the colour game, then, they would know something like, match by red and blue. When the rules change to shape, they would need to recast the goal to match by triangle and circle. There is strong evidence that 3-year-olds can formulate either of these goals because they can easily play either game if it is presented first. However, to account for the basic findings of the DCCS, this approach would have to assume that the 3-year-olds are also 'captured' by the old means. In other words, they have the correct goal after the change, but they employ the wrong means. This explanation can be seen to be a variant of the 'correctly know what they want to do but cannot accomplish it' approach. The error-detection findings that 3-year-olds think the puppet is acting correctly when it perseverates after the change would seem to rule out this account. Although 3-year-olds may not adopt the correct goals for both games, it may nevertheless be reasonable to assume that they adopt the right ones for the first game. After all, they sort correctly in the first game, using the features given in the instructions (e.g. red and blue). So, it could be that after they have successfully acted in the situation, they are unable to recast the goal when the instructions change. According to this view, after the change, they would actually be making a good choice of means, but in pursuit of the old goal. This possibility accords well with the error-detection results. The approach supposes that children's understanding of the satisfaction conditions for the actions in the task continues to incorporate the first instructions even after these instructions have changed. If so, then the puppet's perseverative responses after the change would continue to meet those satisfaction conditions, and thus appear to be right. The puppet's correct responses when the instructions change would appear to be wrong because these responses violate the original satisfaction conditions. And, of course, the comparison condition of the puppet switching without the change in instructions would also appear to be wrong because the satisfaction conditions really are violated. Even though the incorrectly specified goal account helps to explain the errordetection findings, it leaves the dissociation results untouched. In fact, the dissociation effect seems prima facie to contradict this account. The dissociation results seem to demonstrate that 3-year-olds know how to achieve the second game's goal
250
Douglas Frye and Philip David Zelazo
even after playing the first. For example, when asked where the triangles and circles go in the shape game, they answer correctly. The most straightforward interpretation of this finding is that the knowledge questions indeed show that the child knows how to achieve success in the second game. Still, knowing how to achieve a goal and adopting that goal are two different things. The child could truly know how to play the second game and yet not change the specification of the goal when the instructions change. An example may help to illustrate this possibility. Adults can usually be counted on to accomplish their daily drive to work. However, if there is an early morning meeting at a different campus, there is a chance that the average academic will drive to the office rather than appear at the meeting. The mistake can occur even though the person was informed of the meeting and knows the directions to the other campus. In fact, it should be possible to demonstrate the person's knowledge of the correct route by asking where the given roads lead. Nevertheless, the outcome of the drive is likely to be the office car park. This example indicates that, although the person can formulate either goal (go to the office, go to the meeting) and can have knowledge of how to accomplish the out-of-the-routine goal, this knowledge does not ensure success. When thinking about going to work, the person must adopt the goal of getting to the meeting and keep that goal in mind at the time of acting, in order to select the existing information about how to reach that destination. Similarly, when trying to match the cards in the DCCS, 3-year-olds are capable of forming the goal of playing either game (they sort correctly in whichever game is presented first), and they have the knowledge needed for the new one (they demonstrate this knowledge on the knowledge questions). However, they may fail to switch because they fail to adopt the new goal and select the appropriate means. The dissociation results show 3-year-olds have the relevant knowledge to accomplish either goal. The errordetection results show they lack the knowledge to adopt the new goal when the situation changes. With adults, it seems likely that if they were reminded of the goal ofattending the meeting, they would then have no trouble with the change. Cognitive Complexity and Control theory (CCC) suggests that the 3-year-olds' action pattern in the DCCS is not the result of a momentary memory lapse. Instead, it proposes that there is a developmental improvement in young children's ability to change goals deliberately in specific situations, and that the complexity of the action plans and goal formulations required by the situation will matter. Obviously, 3-year-olds can shift their goals in a relatively simple situation such as sorting a red one and then sorting a blue one in the colour game. It is only when the situation becomes sufficiently complex-two different games with two choices each-that 3-year-olds' pattern of acting fails. Furthermore, the difficulty will only arise when there is an inherent conflict in the situation. Switching goals in the driving example is challenging because the goals are related in that they
251
Children's Action Control and Awareness
/\
'" ,
,
,
51
IF
/\ IF
a1 (A)
THEN
a2
I I
c1
(B)
c2
/\
/\ a1
a2
IF
I(C) I(D) c2 c1
52
a1
/\ a1
a2
I (B) I
I(C) I(D)
c1
c2 c1
(A)
THEN
a2
c2
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 satisfaction 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 satisfaction 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 postswitch 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 discussed 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 understanding of their own and others' mental states. One primary change is the understanding 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 determine '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.
3. AN OUTLINE OF THE DEVELOPMENTAL SEQUENCE The present proposal that there are age-related changes in the complexity of children's action plans and goal specifications may help to explain changes in children's action control in the period from 3 to 5 years; however, it is also important to situate these changes within a developmental sequence. This sequence should establish the constituents of intentional action that result in the characteristic patterns of responding shown by the 3-year-olds. To that end, the CCC account has described a sequence of changes in intentional action that occur over the first five or six years of life. A separate model, Levels of Consciousness (Zelazo, 1999, 2000; Zelazo and Zelazo, 1998), outlines a corresponding series of changes in awareness, each of which is associated with an improvement in intentional action. Because action control has been emphasized thus far, the Levels of Control (LOC) approach will be employed to sketch the changes preceding those observed on the DCCS. According to the LOC model, during infancy and pre-school, there are four major age-related increases in the highest level of consciousness that children are able to muster in response to situational demands. These increases are brought about by a functional process of recursion or re-entrant signalling (cf. Edelman, 1989; Elman, 1990) that allows the contents of consciousness at one psychological moment to be reprocessed and considered together with other contents of consciousness. Each degree of recursion has important consequences for the quality of subjective experience, the potential for recall, the complexity of knowledge structures, and the possibility of action control. As regards action control, each degree of recursion permits an increase in the complexity of children's action plans and goal specifications, thereby extending the range of what the child can do. In general, the progression moves consciousness further away from the exigencies of environmental stimulation in what might be called psychological distance (cf. DeLoache, 1993; Dewey, 1985; Sigel, 1993), and allows for the formulation of increasingly complex, and more decontextualized discursive patterns of reasoning and acting. According to this model, abulic dissociations occur (under conditions of interference) until incompatible pieces of knowledge are integrated into a single, more complex action plan via another degree of reflection. In the absence of integration, the particular piece of conscious knowledge that controls manual or verbal responding is determined by relatively local associations. In the LOC model, infants are assumed to be endowed initially with minimal consciousness (Zelazo, 1996), which is meant to be the simplest, but still conceptually coherent, kind of consciousness that can be imagined. It permits the baby to be aware of the world-say of a specific item in the world, a rattle-but not to
254
Douglas Frye and Philip David Zelazo
reflect on that awareness. Or, in other words, the very young baby will have a conscious experience of the world, but will not be conscious of that experience as an experience, and will not be conscious of him- or herself as the subject of that experience. Minimal consciousness supports behaviour, but this behaviour will be relatively stereotypical and rooted in pre-existing behavioural routines or reflexes that are associated with broad classes of stimuli. Hence, contact with a rattle will be likely to elicit sucking as the rattle is assimilated to the broad class of 'suckable things'. Although experience with particular stimuli can change the behaviours I associated with these stimuli-so that, for example, the pattern of sucking certain stimuli can change over time and be coordinated into a higher-order routine, as Baldwin (1968) and Piaget (1952) described in detail (see also Cohen, 1998)-the action will not be intentional. From the baby's point of view, the behaviour does not have satisfaction conditions. Indeed, there is no goal specification at all. The object simply triggers the behaviour, and the baby is only aware, in succession, of the trigger and then of the behaviour itself. The first major change in the developmental sequence occurs between 9 and 12 months of age. The change is modelled by feeding the contents of minimal consciousness back into minimal consciousness via a single recursive loop, which _ results in a higher level of consciousness referred to as recursive consciousness. The content of this level of consciousness is a description or label for an experience of an item in the world. The label, in turn, can trigger an action pattern apart from the item itself. The most obvious example is that babies can first manually search for a completely out-of-sight object at this point (Piaget, 1954). The label serves as a proxy for the object itself, and when it guides behaviour, it can be considered a goal. Note, however, that the child's awareness here is only of the desired object, and not of the relevant action pattern (i.e. the means for obtaining the goal) until that action pattern is triggered. Consequently, the action pattern can be wrong when the situation changes, as in the case of the A-not-B error, which occurs when babies search in the wrong place for a toy they have seen moved from one hiding place to another. There is also direct evidence of this distribution of awareness, in that infants who have been bringing about a particular outcome will be surprised if the outcome unexpectedly changes, but will not be surprised if some unrelated action they perform begins unexpectedly to produce that outcome (Frye, 1991). In Searle's (1983: 85) terms, these are intentional actions with satisfaction conditions that are not self-referential. The infant counts success as enough, and does not yet recognize that the particular action must be the one that brings about, or is directed towards, the outcome. The next significant change occurs about a year later. An additional cycle of recursion allows an awareness of both a goal and also one of the child's actions that can stand in relation to the goal. Essentially, the 18-month- or 2-year-old can form a conditional rule which represents that the performance of an antecedent action (a means) will result in a given outcome (a goal). There is empirical evidence that
Children's Action Control and Awareness
255
the child's awareness of the constituents of intentional action change at this age. Children at the end of infancy are not only surprised when the outcome of an action unexpectedly changes, but they are also surprised when an action that should not bring about an outcome nevertheless does (Frye, 1991). Moreover, when they simply see a novel action but do not witness the intended outcome, they are still able to attempt to bring about the outcome themselves (Meltzoff, 1995). These findings, which show that children expect that only certain actions will produce particular outcomes, indicate the children are now aware of the action, the outcome, and the relation between them. As such, their own intentions can be considered self-referential in Searle's (1983) sense. Although the child cannot yet perform complex actions that involve more than a single action and outcome, awareness of the three fundamental constituents of intentional action (a means, a goal, and the conditional relation between them) is present. The next change occurs between 2 and 3 years of age, and elaborates upon the preceding achievement. Even though 2-year-olds can routinely formulate simple intentional actions, those actions remain largely independent of each other. That is, children do not yet explicitly consider the relations or similarities among different actions. The LOC model stipulates that the next cycle of recursion allows children to relate pairs of actions, or form a pair of rules for acting, that can then be considered in contradistinction to each other. Being able to sort by red or blue in the DCCS is an example of such a rule pair. Zelazo and Reznick (1991) demonstrated this development by having young children sort pictures into two categories (e.g. things that are found inside a house and things found outside). Three-yearolds had no difficulty with the task. In contrast, in spite of being able to label the pictures correctly for inside versus outside, 2.5-year-olds typically only sorted by one of the two rules. Usually, they started to sort correctly, for example, by placing a snowman in the correct box, but then tended to perseverate on putting test cards into that first box (Zelazo et al., 1995). According to the current approach, they were acting to comply with the instructions, but were underspecifying the goal. Being able to sort correctly requires explicitly integrating 'inside' and 'outside' into the satisfaction conditions of the actions. Forming the rule pair of 'if found inside, put here; if found outside, put there' is one way of representing the needed specification, and in order to form that rule pair, children would have to reflect on the rules and explicitly consider the relation between them. The characterization of the change between 2 and 3 years helps to illustrate what else is involved in the subsequent change between 3 and 5 years that leads to Success on the DCCS. Both changes are hypothesized to require a further specification of the goal of the actions. At 3 years, children are able to form a rule pair for sorting by red or blue and another rule pair (considered independently of the first) for sorting by triangle or circle. The recursive step that is theorized to occur between 3 and 5 years, however, makes it possible for the child to step back from these two rule pairs, consider them in contradistinction, and select between them. That is, the additional degree of recursion in awareness allows them to appreciate
256
Douglas Frye and Philip David Zelazo
the hierarchical structure of the task and formulate a corresponding hierarchical system of rules for sorting. As a result, whereas 3-year-olds can only switch between values (e.g. red, blue), 5-year-olds can switch between dimensions (e.g. colour, shape). In line with this interpretation, the setting conditions that come to be added to the specification of the goal are general (e.g. 'in the colour game', and 'in the shape game'). These setting conditions make it possible for 5-year-olds to select the correct actions (and judge someone else's selection of them as being correct) when presented with the change in instructions, but they also have other implications. Because they are general, they extend the range of action beyond the situation. For example, they make it possible for 5-year-olds to recognize perspectives in theory of mind tasks, just as they recognize dimensions in the DCCS. Thus, in the change-of-Iocation false belief task, the addition of setting conditions not only allows children to predict where Maxi will look upon his return, but it also allows ' the understanding that any inference that involves where Maxi thinks the object is will be similarly affected. This example illustrates how the change to a higher level of reflective awareness (i.e. a higher level of consciousness) could increase the child's understanding of both more complex and also more widely decontextualized actions.
4. IMPLICATIONS FOR WHAT THE CHILD CAN DO INTENTIONALLY The age-related changes in children's awareness of the constituents of action expand the range of actions that children can perform intentionally. This developmental claim is similar to, but distinct from, claims that can be made regarding the consequences of knowledge acquisition and skill learning. When children acquire knowledge, such as knowledge of the rules of chess, or learn a new skill, such as how to ride a bicycle, there will of course be changes in children's ability to act intentionally. Unlike these changes, however, the developmental changes in action that are made possible by age-related increases in children's level of awareness will not be restricted to a single domain. The shift in children's awareness of action between the ages of 3 and 5 will be used to illustrate this possibility, and the effects of the shift will be considered for two domains-physical causality and moral reasoning. In both cases, the shift allows children to perform counterstereotypic as well as stereotypic actions. Physical causality. To examine the consequences of the shift between 3 and 5 on children's ability to act upon the physical world, Frye et al. (1996) gave pre-schoolers a simple physical device and asked them to make it work in different ways. The device was a covered ramp with two input holes at the top of the ramp and two outlet holes at the bottom. These holes were connected by two sets of channels.
Children's Action Control and Awareness
257
One or the other set of channels could be blocked off so that marbles inserted into an input hole either rolled straight down the ramp from top to bottom, or rolled
diagonally across the ramp from the top on one side to the bottom on the other. The children were asked to place the marble in the ramp so that the marble would come out of a given outlet hole. Because the ramp operated in two distinct configurations, it could be made to be formally equivalent to the DCCS with there being a 'straight down' condition and an 'across' condition. That is, in the straight condition, to make a marble exit on the left one must insert it on the left, and to make it exit on the right one must insert it on the right. In contrast, in the across condition, the rules are reversed: if out left, then in right; but if out right, then in left. .A light was used to signal what configuration the ramp was in. Children operated the ramp in one of the configurations (e.g. straight, light on) and then it was switched to the other (e.g. across, light off). Pre-schoolers' performance on the task was comparable to their performance on the DCCS. Whereas 5-year-olds could act to produce the desired outcome in either configuration, the 3-year-olds only succeeded in the stereotypical circumstance of the marble rolling straight down. When the marble rolled across the ramp, they still insisted on trying to insert it directly above the place where they wanted it to emerge. To assess whether this problem was really one of intentional action, it was important to determine whether children had the knowledge necessary to understand the workings of the ramp. In the task, children were initially given a demonstration of the ramp in both configurations, and on every trial they were told the rules for how the ramp would operate, just as they were told the rules in the DCCS. The best indication that children understood the workings of the ramp, however, came from an experiment in which one of the input holes was covered over and children were asked to predict where the marble would emerge when it was put into the one available input hole. In this situation, 3-year-olds could make accurate predictions about the marble's path. They could predict when the marble would roll straight down or across depending on whether the light was on or off. Thus, in the one-input version when 3-year-olds had to use a pair of rules involving two goals (e.g. if light on, emerge left; if light off, emerge right), they were able to do so. However, in the two-input version, the physical situation reached the complexity of two pairs of rules involving two pairs of actions that have overlapping outcomes (iflight on, insert left, emerge left; iflight off, insert right, emerge left). In this instance, 3-year-olds were no longer able to choose where to insert the marble to make the device operate as they wished. And, when the situation became this complex, they failed in the characteristic way of expecting the more typical outcome to occur-that the marble would always roll straight down. The findings suggest that with situations of this complexity 3-year-olds will have difficulty with any causal sequence in which there is an exception or change from the way things usually work (cf. Das Gupta and Bryant, 1989). Because of this difficulty, they will not be able to plan the actions needed to initiate these
258
Douglas Frye and Philip David Zelazo
sequences. From our perspective, the developmental change between 3 and 5 years described by the CCC account allows 5-year-olds to extend their intentional actions to appreciate and manipulate non-canonical causal sequences. Morality and action. Moral action furnishes a good example of how the developmental changes in children's ability to specify the constituents of action will enlarge the range of socially relevant actions they can consider. Zelazo, Helwig, and Lau (1996) required 3- and 5-year-olds to use information about an action in order to predict an agent's behaviour towards an animal and then make moral judgements about that behaviour. In the normal situation, the animal experienced pleasure from being petted and pain from being hit. In the non-canonical situation, these causal relations were reversed. Children were further told that the actor was either a nice or a mean person. Following a series of confirmation questions that ensured that children understood each scenario, participants were first asked a behavioural prediction question. For example, in one story involving Sally, who was said always to be nice, the question was phrased as follows: 'Now, Sally knows that mugwumps are weird. She knows that a mugwump cries when you pet it and 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 behavioural 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 understood 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 indicates 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
I
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 children 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 generally beyond them. In contrast, 5-year-olds are able to specify the satisfaction conditions 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 cognition', 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 representational 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 development: 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 function: 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 prediction 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
Children's Action Control and Awareness: Comment on Frye and Zelazo 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 interest 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 theoreticians 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
'Theory of mind' is often used as if it was more or less equivalent to 'commonsense psychology'. To explain now why r avoid the term 'theory of mind' may help both to indicate what can be meant by 'common-sense psychology' and to emphasize the difference r think there is in this area between commonsensical and theoretical accounts (such as CCC and LOC). Frye and Zelazo say about 'theory of mind' that it 'has been taken to mean children's understanding of their own and others' mental states: But 'a theory of mind' has not usually been supposed to stand for something that children distinctively possess. It is usually meant as a name for what people putatively acquire in childhood in I am grateful to the editors for helpful comments on a draft of this chapter.
264
Jennifer Hornsby
virtue of which, beyond childhood, they possess and are able to gain psychological knowledge about others.' Theory of mind, so understood, is evidently something in whose acquisition and evolvement developmental psychologists will be interested. Yet I suggest that the term 'theory of mind' can be prejudicial to debate-in three different ways. In the first place, 'theory of mind' can seem contentious as a label for what one needs in order to have understanding of others' mental states. For it is a controversial question whether it is correct to assimilate the capacity to deploy psychological concepts to the possession of a theory. As is very well known to workers in the field, it is often argued that this is not correct. 2 Second, there is a good question about the extent to which someone's being able to find others psychologically intelligible is a counterpart of their being psychologically intelligible themselves. Some philosophers argue that these two things are bound to go hand in hand. Thus capacities to think things, to know things, and to want things run alongside capacities to know what others think, or know, or want. If this is so, then it will be misleading to talk about psychological development as if it was simply the acquisition of a capacity to understand others; but this is what is sometimes suggested by speaking of the acquisition of a theory of mind. Third, there is a good question about the extent to which agency and mental states are separable. Adult human beings possess mental states (as well as being capable of knowing others' mental states); adult human beings do things intentionally (as well as being capable of knowing what others do intentionally). These things seem to be related. Indeed, capacities to acquire mental states on the basis of interactions with the world, capacities to acquire mental states on the basis of interactions with others, capacities to reason theoretically, capacities to reason practically, capacities to act in the world on the basis of reasons seem all to be interconnected. The idea that there are interdependencies between them all is what a thesis of the holism of the mental can convey. When 'mind' is glossed as 'mental states' (ef. n. 1), 'theory of mind' suggests a divorce between agency and mentality. Speaking of 'theory of mind' can then seem to disallow the holism. 3
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 possesses 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 understandings 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 concepts 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 inseparable 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 specifically 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. 2. COMMON-SENSE PSYCHOLOGY AND DEVELOPMENTAL PSYCHOLOGY In his (1999) Frye sets out by saying that 'theory of mind and executive function ... seem very different'. He tells us that 'it is not apparent why the development of mental state understanding would be connected to the control of one's own actions'. But I hope it is apparent now why someone's understanding of others' 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: someone who takes the intentional stance towards people is someone towards whom that stance can appropriately 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 commonsense 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 developmental stages that children pass through. What seems certain, before any theorizing 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 commonsensically 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 gradual 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 psychological 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 persons. 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 possession 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 learning 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 commonsense 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 connection, 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 experience than they are for being common-sense psychological subjects. s But the present 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.) 3. COMMON-SENSE PSYCHOLOGY AND 3-YEAR-OLDS' DCCS PERFORMANCE I suggest that from the point of view of common-sense psychology, the 3-year-olds' performance on the dimensional change card sort (DCSS) tests is inexplicable. Here is a generalization which I think we take to be true of commonsensically explicable psychological beings, and which the 3-year-olds, in the studies described by Frye and Zelazo, flout: 'If you know what you should do and are able to do it, then (in the absence of any tendency not to do it), you will do it.' (No doubt this needs more qualification;6 and with its hand-waving talk of 'tendencies', it can hardly be thought of as a basic principle of common-sense psychology. But my present concern is only to elicit what is perplexing in 3-year-olds' performance, so that my claim now need only be that common-sense psychology commits us to something a bit like this generalization.) Now consider the 3-yearold who has just sorted by shape, and is told to sort by colour but continues to sort by shape. She knows what she should be doing, because she will tell you if she is asked. She is able to sort by colour, because if she had initially been told to do so, then that is what she would have done. Moreover, since children are cooperative on the task, there is nothing to suggest that she has any tendency not to sort by colour. So why does she not now sort by colour? Commonsensically there seems to be nothing to say-except that she is only 3. And this of course is just to admit that 3-year-olds are not yet full-fledged psychological subjects. One is led to inconsistency if one attempts an adult common-sense psychological treatment of the 3-year-old. For we want to say that the child knows what she should be doing on the DCCS task. But if we take her performance to be governed
5 Evidently much here is controversial; and no doubt my own view that only creatures with language 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 judgements 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 commonsense 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 commonsense 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 extraordinarily 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 supposition, 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 commonsense 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'.)
4. THEORY IN DEVELOPMENTAL PSYCHOLOGY 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-yearolds 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.
5. THE COGNITIVE COMPLEXITY AND CONTROL ACCOUNT CCC proposes that younger children's faulty performance on the DCCS and other tasks should be thought of as a kind of underspecification. [T]here is a developmental improvement in young children's ability to change goals deliberately in specific situations. ... 5-year-olds are able to change their actions [after a switch in the DCCS test] ... because they are better able to specify the goal that they are attempting to reach. (Ch. II)
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 specification 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 completeness, 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 underspecification 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 underspecified 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 common-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 explanation 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. 6. THE LEVELS OF CONSCIOUSNESS ACCOUNT 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 psychological 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 consciousness'? 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 information 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 deliverances 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 commonsense 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 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. DAVIES,
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
HORNSBY, J. (l997a), 'Dennett's Naturalism', in her Simple Mindedness: In Defense ofNaive Naturalism in the Philosophy of Mind. Cambridge, Mass.: Harvard University Press, 168-84. --(1997b), 'Causation in intuitive physics and in commonsense psychology', in her Simple Mindedness: In Defense of Nai've Naturalism in the Philosophy of Mind. Cambridge, Mass.: Harvard University Press, 185-94. STICH, S. P. (1983), From Folk Psychology to Cognitive Science. Bradford Books. Cambridge, Mass.: MIT Press.
13
The Development of Self-Consciousness Michael Lewis
1. THE DEVELOPMENT OF SELF-CONSCIOUSNESS
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 capacities affects the onset of self-conscious emotions, such as embarrassment, shame, guilt, and pride.
2. THE NATURE OF A REFLECTING SELF The adult self is made up of a variety of different aspects, functions, and structures which occasionally work in harmony (see Wylie, 1961, for an historical review of this idea). For example, I certainly know, as I sit here writing, that I have a plan to write this chapter and an outline which I have made to help formulate my thoughts. It is clear that I have intentions and desires and presumably the ability to carry out the task of thinking and writing. Yet the very acts themselves seem to emerge from me almost effortlessly. Indeed, if I focus my attention on them, I find that doing so interrupts the very act that I am performing. It is clear, then, that this self of mine-the body and the mind-that carries out this task does not need and, in fact, may be hindered by my paying attention to myself. A self is necessary to formulate, at least sometimes, what it is that I wish to think about, but does not appear to be involved in the process that actually carries out the task of thinking.
276
Michael Lewis
Consider the example: We give a subject the problem of adding a 7 to a sum of 7s that preceded it (e.g. 7 + 7 = 14 + 7 = 21 + 7 = 28, etc.). It is clear that as We carry out this task, we cannot watch ourselves do the arithmetic. It would seem that one aspect of the self has set up the problem and another will solve it; and it is likely that the first will evaluate the result of what the second did. Let us consider self-deception. How is it possible for a self to deceive its self? It would appear to be a logical impossibility, but only if we believe that a self is a single thing. A self as a single thing could not deceive its self. If, however, we conceive of a self in the manner that Freud (1959) did, one that consists of several aspects or features, then we would be able to argue that one part of the self can deceive another part. The idea of self-deception suggests that the one way to understand the self is to assume the position that there are multiple aspects to the self, which may mean that the self is a modular system-an idea applied to brain structure and process (Gazzaniga, 1988). It is clear that whatever the self may be, it is a complex multiaspect sort of 'thing' or 'process'. This multi-aspect self has been considered in many different ways. In early writing, I have referred to it in terms of 'subjective' versus 'objective' self-awareness (Lewis and Brooks-Gunn, 1979a, b; see also Duval and Wicklund, 1972) or the machinery of self versus the idea of'me' (Lewis, 1994, 1995a). Given this idea of a multi-aspect adult self, how are we to treat the idea of the development of self? From a developmental perspective, not all these aspects exist at birth or even develop at the same time. If they did, there would be little that develops. Thus, it is essential when studying the development of the self that we first agree to the general principle that the term 'self' in and of itself imparts little meaning since it does not specify particular aspects of a self. If investigators talk about the existence of a self at birth or even at 3 months (Gopnik and Meltzoff, 1994; Watson, 1994), they may mean something very different than what others might mean when they talk about the self as evolving in the middle of the second year oflife (Lewis, 1992, 1995b; Lewis and Brooks-Gunn, 1979a). What I should like to do is to make a distinction between different aspects of the self and to argue that the aspect that we, adult humans, refer to as ourselves is, in fact, a rather unique aspect of self, one that we share with few other species (the exceptions being the great apes and, perhaps, porpoises and whales). This aspect of the self develops somewhere towards the middle of the second year of life (Lewis, 1994). It may grow out of other aspects of the self which appear earlier or it may have little connection to them-being related only as part of a developmental function of emerging skills associated with maturational processes. More important, however, is the need to make clear, in both our conceptions and our language, that the functions of this late maturing aspect of self not be assigned to earlier aspects of the self.
The Development of Self-Consciousness
277
2.1 Features of Selves Let me give you an example. In a study of infant learning and emotion, infants of 8 weeks of age are placed in a situation in which their arm pulls result in a reward: in this case a picture appears accompanied by some music for 2 seconds. Each experimental session included a 2-minute baseline during which we were able to determine the baseline or ongoing rate of arm movement. Infants then received a learning phase of contingent stimulation in which the audio-visual stimuli were activated by each arm pull. All infants learned the task within the first 3 minutes of the learning period. When learning was achieved, a 2-minute extinction phase occurred, followed by a second 3-minute learning phase. During extinction, no event was presented after an arm pull. Rates of arm pulling throughout the session were computed as the total number of arm pulls per minute. Facial movements were coded from videotapes of the infants using the Maximally Discriminative Facial Movement Coding System (MAX; Izard, 1979). Coders sampled the videotape segments of each subject using a frame-by-frame analysis of the videotape for each of three facial regions: brows, eyes, and mouth. After coding each component, facial expressions were identified by MAX formulas and their frequency tabulated for each minute of the session. We describe only two-joy and anger face-which could be coded with over 90 per cent agreement between judges. Infants were assigned to the experimental and yoked-control conditions. The experimental subjects' arm pulls resulted in the event occurring whereas the control subjects received the same amount of the event as did the experimental subjects, but it was not related to their arm-pull behaviour. For them, there was no possibility of associating a cause and effect. Look first at the arm-pull data for each age group (see Fig. l3.1a). Notice that control subjects showed no change from the base period to the learning, extinction, and learning phases. Not so for the experimental subjects: to begin with, the infants who could cause the event to go on significantly increased their arm-pull behaviour. Of particular interest are the subjects' responses once the association between arm pull and event ceased to work (extinction). Notice that when the arm pull no longer caused the event, arm-pulling behaviour significantly increased rather than declined over the period of disassociation. Once the extinction phase was over, the infants returned to the rate of arm pulling they showed during the first learning phase. These differences were all highly significant. Now let us turn to the emotional behaviour. There was little joy during the base phase and no change for the control subjects (see Fig. 13.1 b). The subjects who learned showed increases in joy during the initial learning phase, a total decline during extinction, and renewed joy once the second learning phase began. Angry expressions follow a reverse pattern (see Fig. l3.1c). There is little anger during the base or during the initial learning
Michael Lewis
278
f ! ~C_~_~ _ _2_m_o_nt_h_s
1
L1
B
EX
L2
Phase of learning
(0)
2 months
I
f~j - -=~= = = =~_: . . . .:. . 2_C~: : .P:.:N_ 2.0
l...-_ _
B
L2
V>
O)c
Co "'.-
o~
2.0 1.5 I
C .... 1.0 ",c..
2
FIGURE
EX
2 months
....
(c)
L1
Phase of learning
(b)
><
EXP
~
0.5~ o L--_...:::::::::::=-...- ===-----="=====----'=_ EX B L1 L2 Phase of learning
13.1. Two-month-olds' behaviour in a contingent learning task
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 outcome, 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 epistemological 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 distinction 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 knowsis 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 newborn 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 cannot 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 selfprocesses 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 selfawareness 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 selfawareness; 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 maturation 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).
3. THE RELATION BETWEEN SELF-RECOGNITION, PERSONAL PRONOUN USAGE, AND PRETEND PLAY 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 consciousness. 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 embarrassment. 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 selfpretending, 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 spontaneous pretend play, mirror self-recognition, and personal pronoun use in a longitudinal 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 themselves 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 TABLE 13.1. Proportion of children who showed self-recognition, personal pronoun use, and/or pretend play first in development Behaviour (a) One behaviour by itself Self-recognition Personal pronoun use Pretend play (b) Two behaviours at the same time Self-recognition and personal pronoun use Self-recognition and pretend play Personal pronoun use and pretend play (c) All three behaviours at the same time
Per cent
48 12 5
23 3
o 9
284
Michael Lewis
children who showed self-recognition, personal pronoun use, and pretend play first in development by (a) one behaviour by itself or (b) two behaviours at the same time. The data are quite clear; self-recognition as a single behaviour emerges first in 48 per cent of the children; self-recognition and personal pronoun use at the same time emerges only in 23 per cent of the children. The findings point to a developmental change in self-representation in the second year of life that involves the onset of a theory of the mind (ef. Gallup, 1991; Leslie, 1987; Povinelli and Eddy, 1996) or at least the ability to represent the self as an object of knowledge, that is, the appearance of the self that knows that it knows or knows that its play is not real (Lewis, 1995a; Lewis and Ramsay, 1999). That there was a close sequential relation between the onsets of self-awareness and pretend play suggests an underlying developmental relation between the abilities as opposed to independent developmental milestones. If the abilities had been independent, there would not have been a consistent sequence in their acquisition. Self-recognition, personal pronoun use, and pretend play all indicate the capacity for self-representation. It is apparent that with development, self-representation increasingly becomes a more complex and multifaceted phenomenon that progressively includes other cognitive and evaluative aspects of self-knowledge (e.g. Lewis and Brooks-Gunn, 1979a, b). None the less, the present findings suggest, in terms of emergent time, the primacy of self-recognition in the formation of a complex self-representation. Of the three self abilities assessed in this study, selfrecognition was the one most likely to emerge first in development. Thus, the present findings suggest that physical self-recognition provides the core aspect of the self-representation that continues to develop beyond the second year of life. Consistent with the present findings is recent work that indicates children's emerging understanding of a theory of mind by the middle of the second year of life. Meltzoff (1995), for example, reported findings indicating the ability of 18-month-old toddlers to understand the intentions of others. Similarly, Asendorpf and colleagues (Asendorpf and Baudonniere, 1993; Asendorpf et al., 1996) found increases in imitative play linked to the presence of self-recognition in 20-month-old infants. Piaget's (1952, 1954, 1962) early theory of sensorimotor intelligence did not emphasize the development of the self although it would appear that self-awareness is a necessary aspect of various representational abilities including pretence. In this regard, Bertenthal and Fischer (1978) assessed the degree of correspondence between the development of object permanence and the development of various mirror-directed behaviours including the mark-directed behaviour indicative of self-recognition. While they expected considerable decalage in development across these two domains, to their surprise they found a strong correspondence between them. This strong relation is consistent with a central, perhaps organizing role for self-knowledge in Piaget's system of the six successive sensorimotor stages
The Development of Self-Consciousness
285
(d. Lewis and Brooks-Gunn, 1979b; Mounoud and Vinter, 1981). Even more important for our purposes is the finding that with the emergence of this cognitive ability, the child becomes capable of intentions of an adult-like form. As we shall see in the next section, the creation of this meta-representation allows for the development of emotions such as shame and pride which in turn help the child in making plans and in choosing which actions they wish to engage in. In this regard, achievement motivation becomes part of intentions.
4. SELF-AWARENESS AND EMOTIONS Perhaps nowhere is it more obvious that a meta-representation of the self or the idea of me is necessary for human behaviour than in the emotional domain. Darwin (1872/1965), in his discussion of the emotional life of humans and animals, makes clear that emotions can be distinguished by the role of the self. For example, in discussing blushing, Darwin believed that blushing was caused by how we appear to others or as he put it, 'the thinking about others thinking about us ... excites a blush' (p. 325). What is clear is that the emotional life of humans must involve the self. While I have discussed this issue at great length elsewhere (Lewis and Michalson, 1983; Lewis, 1992, 1993a, b, 1995a, 1999), I should like to make several specific points in this regard. 4.1 Cognition as the Eliciting Condition While any discussion of eliciting events in relation to emotion involves all emotions, I will restrict my comments to the class of emotions that I have called self-conscious evaluative emotions, in particular, pride and shame. To understand the problems attendant on the notion of eliciting events, the two major models that link emotions to their context or to the events likely to elicit them need to be considered first. The first of these models maintains that emotions, especially those called the 'primary' or 'basic' emotions, can be elicited by non-cognitive events. These events, which may be internal or external in nature, act in some direct way to elicit a class of behaviours called 'emotional'. We can think of this one-to-one correspondence as akin to proposing specific emotional receptors that respond to specific stimuli. Support for such specificity can be found in the perception literature. For example, Gross, Rocha-Miranda, and Bender (1972) report that in monkeys some cells in the visual cortex are so finely tuned that they respond only to specific shapes and objects. In other words, specific features or locations have been reserved for specific eliciting events. Because of this one-to-one correspondence, little or no cognitive function is necessary. However, to the degree to which sensory mechanisms have a cognitive component, this argument is made difficult. Nevertheless, excluding the cognitive processes involved in
286
Michael Lewis
sensory processing, information from eliciting events may act directly in eliciting emotion (Zajonc, 1980). LeDoux (1989) has reported a series of studies on rats that indicate that the amygdala, rather than the visual or auditory cortex, may be capable of learning emotionally based information. This suggests that some emotional processing takes place in brain regions other than the cortex. I do not believe that this means that cognitive processes do not take place, only that they take place at different levels of brain functioning. In any event, theories that hold to this immediate oneto-one correspondence between a class of elicitors and an emotional response have been called by Hochschild (1979) 'hiccup-like theories of emotion'. To give examples of elicitors that are commonly used to characterize such immediate processes, fear is triggered by loss of balance, falling, or visual looming, startle by a loud noise, and joy by seeing the face of a beloved person, whether it is the mother, in the case of an infant, or the significant other, in the case of an adult. Thus, while we might talk about emotions as the consequence of agency, or that they occur around intentions-for example, joy and anger in the 4-monthold infant as a consequence of action-these are part of the system and exist as components of the machinery of self. While conscious processes and intention may not be necessary for some classes of emotion, they are when we consider more complex emotions. In the case of jealousy, envy, empathy, embarrassment, shame, pride, and guilt, it is very difficult to think of some one-to-one correspondence between specific environmental elicitors and the production of such emotions. These emotions generally require that the organism make a comparison or evaluate its behaviour vis-a-vis some standard, rule, or goal. Thus, for example, pride occurs when one evaluates one's behaviour against a standard and finds that one has succeeded, whereas shame or guilt follows when such evaluation leads to the conclusion that one has failed. Thus, the elicitation of this class of emotions involves elaborate cognitive processes, and these elaborate cognitive processes have, at their heart, the notion of self, agency, and conscious intentions. Cognitive processes must be the elicitors of these complex emotions (Lewis, 1992, 1995a, b). It is the way we think or what we think about that becomes the elicitor. These cognitive processes and intentions, therefore, playa vital role in eliciting these types of emotions. Whether self-cognitions are involved in all emotions remains in question. If we believe, following James (1890), that emotions are both physiological events and experiences of these events, then self-cognitive processes must be involved in the experiential aspect of emotion. Because of the confusion between physiological events and our experiences of them, I have found it useful to draw a distinction between emotional states-a hypothetical construct of bodily changes that may occur at any level and include neuromuscular changes in the face as well as central or autonomous nervous system responses-and emotional experiencesour interpretations or experiences of the state. Elsewhere, I have. argued
The Development of Self-Consciousness
287
(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. 4.2 The Role of Self Stipek et al. (1992) make it clear that the cognitive process of 'self-evaluation' involves the concept of self. It must be pointed out, however, that pride and shame are only a limited representation of the class of emotions that involve the self. Empathy, for instance, also involves the self since it involves putting oneself in the role of another. As some have indicated, empathic behaviour as measured in older subjects does not seem to emerge until a referential self emerges (Bischof-Kohler, 1988, 1994). In addition to empathy, the emotions of guilt, embarrassment, and jealousy also involve the self. Thus, it is quite clear that there exists a wide range of emotions that, by definition, involve the concept of self. My colleagues and I have undertaken a series of studies that attempt to explicate the developmental sequence of self-referential behaviour and to relate the emergence of this capacity to emotional development. I have attempted to clarify in some detail those specific aspects of self that are involved in shame and pride, two among the other selfconscious emotions (Lewis, 1992). In particular, I have suggested that self-evaluative emotions involve a set of standards, rules, or goals (SRGs) that are inventions of the culture that must be transmitted to the child. By 2 years of age children are able to demonstrate that they have incorporated, in some fashion, the SRGs of their parents. Incorporation represents the ability of children to carry the rules learned without an adult being present. What is intriguing is that this process appears to start at extremely early ages. We have no trouble believing that children can learn these SRGs either through direct reinforcement or through discovery. However, what is apparent from the work of Stipek et al. is that the process of incorporation starts quite early. The second point to stress is that the child must be capable of accepting responsibility for his or her behaviour: if children are unable to perceive that they are the actors or the producers of a particular set of behaviours, then, in fact, there would be no basis for evaluation. For example, if I win the lottery I would be happy but not proud. I am not proud because I am not responsible for this 'chance' event. Self-evaluation, therefore, implies not only a standard, rule, or goal, but also the realization that it is one's own action that leads to the success or failure. Finally, the self also enters in the comparison between one's action and one's standard and
288
Michael Lewis
in the decision to adopt an emotional response. Two distinct processes seem to be involved. I can, for example, evaluate my behaviour against my SRGs and conclude that I have succeeded or failed. However, this will not lead me to either pride ., or shame unless I am prepared to believe that I am responsible for that success or failure. In the attributionalliterature, this has been considered as the distinction between an internal and an external attribution (Weiner, 1986). The same is true for shame, and this process is at work even with very young children. Lewis et al. (1992) presented children of about 3 years of age with a series of easy and difficult tasks. Most subjects succeeded at the easy tasks, and some also succeeded at the difficult ones. We found that children were significantly more likely to show pride when they succeeded at the difficult rather than at the easy task. Shame was likewise affected in that the children were much more likely to show shame when they failed at an easy task than when they failed at a difficult one. Still another issue to consider has to do with the evaluation of one's self in terms of specific or of global attribution. 'Global self-attribution' refers to the whole self, while 'specific self-attribution' refers to a specific feature or action of the self (for a similar distinction, see Dweck and Leggett, 1988; see also Weiner, 1986). If one internalizes (i.e. accepts) the fact that one has failed a particular standard, rule, or goal and makes a global attribution, one is likely to feel shame. However, if the attribution is made about a specific rather than a global feature of the self, one is apt to feel guilt or regret. Let me give an example of this difference. Imagine that you have written a research paper and submitted it for publication. It is returned rejected. Besides the involvement of the self in constructing goals and evaluating behaviour, there are several other processes involved in determining your reaction. The first of these has to do with internal versus external attribution. You can assume that the reviewers did not know what they were talking about and, thus, refuse to accept the rejection as a failure having to do with you. Alternatively, you can accept it as a failure for which you are responsible. If you do not accept the rejection as failure on your part, you may simply send the paper off to another journal. If you do accept it as failure, you still have a second attribution to make, namely, to determine through processes yet to be considered whether this reflects a global or a specific attribute of your failure. If you make a global evaluation, such as, 'I am not a good scientist; you are much more likely to feel shame than if you make a specific attribution, such as, 'I should have conducted an analysis of covariance rather than what I did'. The response of global self-attribution and, therefore, shame is likely to lead to the cessation of activity-that is, the body collapse as described by Stipek et al. (1992). Shame has been associated with body collapse as well as with a lack of repair and reparation (Lewis, 1992). In the case of our hypothetical colleague, under a global, internal attribution of failure, he or she is likely to take the manuscript and put it in a drawer, never to be looked at again. In contrast, failure that is internalized and specific is likely
I
The Development of Self-Consciousness
289
to lead to reparation since a specific feature in need of repair has been identified. In such a case, our hypothetical scientist is likely to revise the article and send it off again. We can see, therefore, that the role of self-processes in this class of emotions is quite elaborate, involving (1) knowledge of standards, rules, and goals; (2) incorporation of these SRGs; (3) evaluation of one's behaviour vis-a-vis the SRG; (4) distribution of the blame to oneself or to others; and (5) attribution to either a global or a specific self. In almost each one of these processes, a concept of the self and self-processes needs to be considered. 4.3 Socialization of Self-Conscious Emotions By what socialization process does a standard, rule, or goal become incorporated by the child? I have argued elsewhere (Lewis, 1992) that the motives for incorporation may have to do with the self-concept itself; that is, once an organism is capable of constructing an idea of itself, this idea includes not only such characteristics as feature representation but also, in this case, representations of its own beliefs, ideas, and so on. This is not far removed from the view of many self theorists, starting with James (1890), who held to the idea that the concept of oneself has within it not only the existential self (one's existence) but also a categorical self, the latter consisting of all those characteristics that one attributes to oneself. Beliefs, once they are acquired and once one has a self, are by necessity part of the self. Thus, although one may be concerned over one's parents' view of oneself or about the possible implication of one's actions, the idea of having a self automatically means that we have standards, rules, and goals. Which of these apply for a given culture or family needs to be learned, but having them does not: they are intrinsically a part of ourselves; they constitute a part of the self's definition. One self-process involves comparing one's behaviour to an SRG and concluding whether one has failed or succeeded vis-a-vis that standard, rule, or goal. Socialization rules playa direct role here; receiving a B in a course may be viewed as a success by one child but as a failure by another. For example, I know a patient who reported that whenever he took an exam his father was always critical, no matter how well he did. He recalled coming home with a 97 on a chemistry test and having his father ask what happened to the remaining three points. This example illustrates how the evaluative process of comparing one's behaviour against a standard can be influenced by socialization processes. The strict demands made by socializing agents may produce what has been called in the psychoanalytic literature an overdeveloped superego. Another aspect of socialization has to do with whether or not people make global or specific self-attributions and whether these evaluations are positive or negative. Given the proposition that shame is a consequence of a global rather than a specific self-attribution concerning failure, it is likely that socialization
290
Michael Lewis
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 contempt 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 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. ALESSANDRI,
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 experience', 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 development', 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.), SelfAwareness 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: NorthHolland, 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 selfconscious 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 language', 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 imitation 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 developing 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 discovery 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 presupposes 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 selfknowledge 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 question 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 controversial 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 attributing 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.
1. UNDERSTANDING INTENTIONS AND ACTIONS:
A PERCEPTUAL THEORY How does an agent gain knowledge about her own acts? How can she identify her own current intentions, and how does this capacity relate to identifying intentions as driving others' observed behaviour? It does not seem prima-facie plausible to claim that both kinds of attribution of an intention are effected in an even remotely similar way. A received view maintains that intentions are identified in self through non-observational knowledge, while another person's intentions are recognized through perception-cum-inference (Anscombe, 1957). Intention does not seem to be the only kind of mental event or disposition to be attributed differently according to its particular target-self or other-: pains, beliefs, desires, emotions also seem to allow a similar contrast between first- and thirdperson attributions. A common argument for invoking a non-observational source of knowledge is that there is no apparent mediation, in this kind of case, between evidence gathered and a fact to be known: one directly grasps one's own physical posture or intention to do P, without apparently having to identify typical sensations for being in that posture or having that intention. The notion of'knowledge without observation', however, raises the problem of understanding how new knowledge can be gained when no sensory information is collected. How is an agent in a position to recognize, for example, that her leg is bent, or that she intends to drink a glass of water? The immediate or direct character of the kind of knowledge one has of one's own limb position, of one's own pain or intention, is compatible with the fact that some type of informational access is provided to the relevant property. A plausible alternative to the non-observational knowledge approach consists in suggesting that a subject becomes aware of her being in such and such a state through perception. After all, many clear-cut cases of perceptual states also have such a direct relation to their own contents. When, for example, an observer sees a patch of red, she does not first identify the sensation she has, then correlate it with some
298
Joe lle Proust
external property, to infer that there is some red object out there. The external fact is what she directly sees. A second ground for rejecting the non-observational view is linked to the demands it places on the relation between being in a mental state and attributing that state to oneself. According to this view, one cannot be in pain, have an emotion, form an intention, without simultaneously forming the belief that one has it.! One might object that pains, emotions, physical postures, intentions often fail to be acknowledged by their bearer; they may be currently active in an individual (i.e. be contextually triggered and control her behaviour in a way essentially similar to conscious pains, emotions, physical postures, and intentions), while being, at least for some time, undetected. These two arguments suggest it may be worthwhile to try articulating in more detail a theory of the kind of access that an organism has to her own proprioceptive and intentional states, thus grounding knowledge of one's current deeds, intentions, and well-being in specific informational facts, which would in turn be stored in memory. Sensory perception, generally speaking, consists in collecting information about an objective, independently existing world. Let us briefly summarize the ingredients which have to be present for some event or property to be perceived. (a) A perceiving subject, using specialized captors, extracts spatial or qualitative
bits of information on a property, event, or state of affairs, such as: (i) '0 is presently at P'; a has Q, R, S (colour, shape, sound, texture ... properties).
(ii)
(b) Such perceptual knowledge is represented in a phenomenological way; in other words, a distinct qualitative feeling is normally associated with the perceived properties. Seeing involves in standard cases having a visual experience in which a state of affairs is being presented to the seer. 2 (c) On the basis of the information extracted by the senses, perceptual judgements produce a conceptual categorization of the perceived events or objects through which a singular percept is subsumed under some concept: 'this is a horse', 'this is a ringing of the bell', and so on. (d) Perceptual judgements may be either veridical or illusory: to be veridical, a perception must present the world as seeming to be what it actually is, and the fact that the world has the presented property must be what causes the perceiver to have this perception. In such a causal theory of perception, which we will have to take here for granted, the existence of a causal link between the world and a subject's perceptions makes her awareness objective. When the relevant perceptual mechanisms are not reliable, or when the situation perceived is ambiguous or specifically contrived so as to produce See Shoemaker (1996). Such a presentation conveys perceptual content in an analogical, non-conceptual mode; on this notion of a non-conceptual content of perception, see Evans (1982), Peacocke (1992a, b), Crane (1992), Bermudez (1995,1998); among opponents, Armstrong (1968), McDowell (1994). 1
2
299
Perceiving Intentions
systematically deviant subjective effects, the observer may produce wrong judgements induced at a perceptual, rather than at a conceptual level. (e) A basic claim of perceptual realism, on which non-idealist philosophers agree, is that the objects of perception are independent from the fact that they are being perceived. That there is a tree in my visual field is true by way of there being a tree at a certain time and location, and my being in a position to see it, and the latter capacity being in part caused by the fact that there is a tree there. In no way is the existence of the tree at this location contingent on my own capacity to perceive it. In other words, the existence of perceived events necessarily lies, in a de re mode, outside the scope of perceptual attributions. Any perceptual event thus essentially involves a counterfactual truth, to the effect that For some particular object 0 existing at time T and location L, had a non-instrumented perceiver P not been appropriately located at Tin the relevant vicinity of L, P would not have perceived O. Given this type of analysis, proprioceptive states seem to fall naturally within the scope of perception. 3 The existence and mechanisms of a specialized sense, sometimes called 'muscular sense', allowing a subject to identify the types of postures which he currently has, have been documented by pathological cases such as deafferented patients, as well as by experimental work on muscle vibration. Specialized sensory channels are used to appreciate one's own kinaesthetic posture. Muscular proprioceptive messages from various body segments (most notably, from eye, neck, and foot muscles) help a subject keep track of her own posture and gaze direction. 4 Although terms like 'observation' and 'perception' as employed in ordinary psychological idiom may sound inadequate to express this kind of informational extraction and processing in postural judgement, this case is vision-like in all the main respects. Proprioception, like vision, allows facts and objects of spatial and qualitative varieties to be grasped (the subject's head being twisted to the left; her legs being bent, etc.); it is phenomenological, in the sense that there is something it is like to experience that one's leg is bent, or that one's head is twisted to the left. Moreover, a subject who feels her legs bent picks up sensory information to form a judgement about an independent physical fact. The latter is independent in the sense that the position of a limb is not constituted by the awareness of the subject whose body is involved. One may fail to know one's limb position for various reasons, including deafferentation, experimental manipulation of one's tendons and muscles, and so on. It may on the other hand seem that one's limb is in a different position than it actually is. Illusory proprioceptive states can also be produced experimentally by tendon vibration. A motionless subject may in those modified conditions, for example, feel her head, trunk, or whole body rotate. S 3 5
For a full defence of this view, see Martin (1995). Cf. Roll et al. (1990: 549-65).
4
See Roll et al. (1991).
300
Joe/Ie Proust
The cause of perceiving that one's legs are bent is that some specific muscles, tendons, 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 environment, 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 perceived, 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 mechanisms deal with the dynamical pattern of an observed movement. Humans can easily discriminate a biological motion from a non-biological, mechanical movement, even when the cues are limited to a few luminous dots located on the moving 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). 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 intentional 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 independent 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
I
301
Perceiving Intentions
without recogmzmg (that 5 does A). Perceiving a biological movement as an action, the objection goes, would require applying the whole explanatory structure 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 transition. 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 concept 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
6
7 10
On these syndromes, see Lhermitte (1983), Lhermitte et at. (1986). Harding and Golinkoff (1979). 8 Gergely et at. (1995). Bretherton et at. (1981).
9
Penson (1984).
302
JodIe Proust
attack, or a court display, as an unmistakable dynamic pattern of behaviour associated with some specific target event. This type of perceptual format is exemplified in ritualized behaviour, whose goal is to influence the motivational states of the conspecifics in some predetermined way. The detailed teleological structure of ritualized behaviour would not be intelligible if there were no perceptual system able to extract the relevant intentional pattern expressed in an animal's behaviour (Lorenz, 1983; Zahavi and Zahavi, 1997). Here again, the claim should not be understood as involving a mental attribution of intentions. An animal which extracts the information about its predator's intention does not need to have the corresponding mental concept of intention, any more than an animal able to use conspecifics' signals needs to have the full mental concept of communication. What is claimed is only that living creatures had better be in a position to extract perceptual information about typical kinds of events involving conspecifics' or predators' agency. Even though they lack the mental concept of intentionality, they must be ready to acquire the perceptual knowledge that a specific kind of action has been launched against them, or is to involve them otherwise crucially. Thus they should be able to perceive the action in some dynamic, action-specific format, and to categorize it according to its consequences. 11 In this strategy, it is thus accepted that perception of animate movement constitutes, in the human case, a preliminary step to an understanding of action in fully intentionallmental terms, and, in the animal case, a specific domain of registration that may offer, in time, resources for teleological categorization. A second strategy, however, defends the view that the transition from perceiving an animate movement to making a perceptual judgement (where an action is seen 'as an action') does not require a contribution from some external faculty (such as language, learning, etc.). In this view, infants and non-human animals have the capacity to judge in some implicit way that an action of some crude type is performed. This strategy has been adopted by those who defend a modularist conception of theory of mind development, such as Leslie (1994) and Baron-Cohen (1995): detecting intentions is taken to be a specialized informational mechanism that responds automatically to perceptual inputs with predetermined features. There is no need in this view to distinguish 'perceiving an action' and perceiving it 'as an action', for the very format in which the action is perceived implies that it is dealt with as an action, which the subject must adequately identify and respond to. The two objections presented above are addressed by admitting that the protoconcept of action that is automatically triggered by perceptions of animate movements needs to be enriched and corrected in order to be part of a wider and more reliable inferential capacity. This proto-concept is generally restricted to categorizing a dynamic pattern as leading to a particular goal. Modularists typically hypothesize II We will presuppose here that non-human animals can use concepts to predict regularities in their environments. See Proust (2000b).
Perceiving Intentions
303
that additional modules, such as SAM and above all ToMM,12 offer the kind of resources allowing a child to explain (in mental terms) why an agent performs an action with such and such an intentional content. We do not need at this point to arbitrate between these two strategies to address the objections above. What is clear is that, in both cases, 'intentions' are taken to be grasped in a different way, from an early phase in which intentional actions are discriminated (in a sensory way, in the first view, or categorized through protoconcepts, in the second) to a later phase during which their representations are embedded within a general theory of rational agency. These considerations should warrant a specific terminological use that will be adopted in the remainder of this chapter. What we tried to establish above is that the word 'intention', which refers usually to a representation conveying the mental conceptual content of an action and driving its execution, needs to be extended to what perception affords one when another agent acts intentionally. By claiming that an intention, rather than the corresponding action, is perceived, we want to emphasize that perception offers all the evidence needed, in many ordinary cases, for judging-without inference-not only what an agent does, but what she is up to. In fact, the very anticipatory and teleological nature of this kind of perception makes it plausible to say that what is seen is an intention, and not only a developing action. This kind of use is not isolated: although emotions in other agents are also mental states with an intentional content, they are taken to be perceived early on, even when the corresponding concept is lacking. We now want to suggest that it also makes sense to say that an agent perceives her own intentions in a similar way as she does others'.
2. PERCEIVING ONE'S INTENTIONS
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 BaronCohen (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 considered 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, nondeliberate 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 occurrent 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 affording 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 attentional 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 features, 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, proprioception, 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 features 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
13
14
See Feinberg (1978). See Searle (1983) and Proust (forthcoming b) for a defence of this view.
306
Joe lie Proust
able to assemble as a single temporally extended batch those changes contingent on the agent's behaviour.l 5 Now let us turn to condition (b). The kind of access that a subject has to her own intention through the dynamic and qualitative structure of the input is also phenomenological, in the sense that there is a distinctive feeling associated with being the agent of a specific action. First, the way one's body moves is felt differently when the subject is active or passive. The distinctive sense of effort involved in voluntary actions, whether interpreted by Wundt (1903) as a 'sense of innervation'-namely, as caused by efferent commands to the muscles-or understood by James as an image of peripheral sensations, seems to offer an important signal for perceiving actions differently when performed by self or passively observed. This distinctive feeling seems, however, to be variable in intensity according to the subject's mood, and may even disappear in 'depersonalized' conditions sometimes occurring, for example, in depression and schizophrenia. As we will show below, schizophrenic patients may be said to have defective mechanisms for perceiving intentions.l 6 At least in some cases, they seem unable to capture perceptually the dynamic pattern of their actions, thus failing to recognize their underlying efficient intention, just as an agnosic might look at an object composed of a blade and a handle, while failing to recognize a knife. Beyond the specific impression related to a sense of effort, an agent also has distinctive qualitative feelings associated with both passive and active movements including, as James shows, both the 'remote effects of the movement'-that is, the external events it produces-and the 'resident effects of the motion': Not only are our muscles supplied with afferent as well as with efferent nerves, but the tendons, the ligaments, the articular surfaces, and the skin about the joints are all sensitive, and, being stretched and squeezed in ways characteristic of each particular movement, give us as many distinctive feelings as there are movements possible to perform. (James, 1890, vol. 2: 488)
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 gesturesalso 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 sensory 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.
3. ARE INTENDING AND PERCEIVING ONE'S INTENTIONS ONE AND THE SAME EVENT? Let us start with the first of these difficulties. How can an intention, that is, some event or state of the 'internal' variety, be perceived by an agent as some external property in the environment? If intention is a propositional attitude with a particular content, the only kind of perception that seems to make sense for it is introspection. The very idea of looking to the outer world to perceive one's own intentions simply seems to make no sense at all. Let us approach this difficulty from a different tack. As a content of thought, an intention may be analysed by using a classical distinction between what one thinks, what one thinks about, and what one thinks about it (Prior, 1971). When one thinks that grass is green, one thinks, about grass, that it is green. Now an intention is by definition a representation. It differs from other thoughts not through its intentional content per se, but through its direction of fit: world-to-mind as in desire rather than mind-to-world as in belief (Searle, 1983). Let us examine more closely the intentional content of an intentionY When I intend to do P, where P, say, consists in carrying an object 0 from Ll to 12, three dimensions of thought are similarly to be distinguished: what I think is (1) 'I intend to do P'.
17 The expression 'intentional content' in all its occurrences in this chapter means 'semantic content' rather than the content of an intention.
308
Joe lle Proust
What I think about when I think (1) is some definite property of my target 0, that is, the final state of the object's location: (2) 0 in L2. What I think about it is (3) that, in order to be in L2, a movement displacing 0 from Ll to L2 is to be performed. 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 realizing 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 perception (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-toworld 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 metaphysics 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 premotor 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 gestures, 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 distinct 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?
4. HOW IS PERCEIVING RELATED TO INTENDING? 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 intention 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 movement). What is preserved is the instruction (3) deployed over time by the action.
Joe lie Proust
310
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 different 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 movement 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 predict 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 situation in the world that responds to her own intention (with a world-to-mind direction 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 perceiving 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 generalize 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 oneself: 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
5. INTENTIONS MISPERCEIVED: TYPES OF ORDINARY FAILURES Given that there is a capacity for consciously perceiving a developing pattern in the world as the substance of one's intention, a semantical problem arises. When shall we say that an agent identifies correctly a dynamic pattern as her intention? Might not the present approach encourage scepticism about intentional content, by allowing an agent to perceive a variety of coherent patterns as self-generated, or reciprocally by preventing her from perceiving her own intention in case the expected effects on the outer world are incongruent or lacking-in particular, when her intention is not efficient? Let us suppose that the agent first forms intention 11, initiates the corresponding bodily movement, but, as a result of an interferent context, does not pursue the intended action: she rather engages in some other intention 12 (e.g. she goes to her bedroom to pick up a book, but her intention is interfered with by the prepotent context, and she finds herself undressing (cf. James, 1890)). Or, while cooking, an agent opens the first available door (which turns out to be to a fridge) instead of the desired door (of the microwave oven, for example). Which intention, in the present theory, will the agent perceive-the initial interrupted action or the interfering one? The answer to this question is important: the rationality of the agent depends on her ability to correct, and therefore first to detect, a mismatch between the actual dynamics of the scene presented as her realized intention and her intention itself. How might an agent discriminate perceptually 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 happens 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 exercising 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 precondition for evaluating one's course of action. As was suggested above, the conceptual 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 anticipation what is to be done in some context allows her to evaluate how far the selfinduced 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.
21
Cf. Proust (2002).
22
James (1890, ii: 497).
Perceiving Intentions
313
In some cases, the subject may falsely believe that her intention corresponds to what she can see, whereas in fact what she sees differs from what she does. In an experiment by Fourneret and Jeannerod,23 for example, an agent perceives her intention as drawing a line straight ahead (because this is what she can see on a monitor screen), when in fact she draws a line at an angle (responding to an experimental bias affecting the feedback, unknown to her). In such a case, the intention controlling the subject's movement coincides neither with what the subject can perceive nor with her conscious intention. This case, however, does not reflect a failure of the subject's intention detector. It rather expresses a conflict-experimentally orchestrated-between the visual and proprioceptive components in the perception of an intention. A subject might give more weight to vision than to proprioception in appreciating her own intention because the current goal to be reached, in its conceptual content, is a visual property (the drawn line as seen on the monitor screen must be vertical). This asymmetry between vision and proprioception associated with the general dominance of vision in humans misleads the subjects into misperceiving their efficient intentions as that of drawing a line straight ahead. A sceptic might object to this theory that, even though an agent may perceive her current intention, when she actually entertains one, and conceptually recognize it on the basis of perceptual evidence, she might also experience many other predictable stimuli as an expression of her intention. For why after all would not the clouds, say, follow the path which I intend them to? To address this concern adequately, we need to show that an agent is able to discriminate perceptually external events contingent on her action from those which occur independently of her wilful activity.
6. SELF-GENERATED/EXTERNALLY GENERATED PREDICTABLE STIMULI: THE SEPARABILITY PROBLEM It is part of the phenomenology of perception that an agent expects to perceptually identify objects and events of a given kind (where a 'kind' may be identified either conceptually, or through specific sensory or spatial cues). In any type of perception, perceivers evaluate whether things are, or fail to be, as expected. The subject may rely on a conscious emotional signal, a specific feeling, to discriminate between highly predictable, familiar stimuli, or conversely, stimuli that are unusual in their shapes, colours, frequencies, and so on. The intensity of this feeling on a familiarity/novelty scale is triggered by the degree of informational match between an incoming stimulus and stored perceptual regularities. 23
See Fourneret and Jeannerod (I 998) and Slachevsky et al. (2001).
314
Joe/Ie Proust
A predictability signal cannot, however, be sufficient to determine that some perceived change is a result of one's own intention. One might even worry that the feeling of predictability in fact constitutes an additional source of confusion rather than a distinctive cue in perceiving intentions; for if this information was all that was available, an agent would have much trouble discriminating the selfversus the externally generated equi-predictable events. A traditional response to this worry is that an agent feels the world differently when the perceived external changes are contingent on her action or independent from it. This 'sense of effort', also called 'sense of innervation',24 was subject to a hot controversy in the nineteenth century: is the effort felt when initiating an action a result of the feeling of resistance from the world, or a direct expression of the causality of consciousness?25 In a functional perspective, recent empirical data suggest that there is a grain of truth both in a peripheralist and in a centralist view of sense of effort: on the one hand, this feeling is directly related to world-directed perception; on the other hand, such a perception is compared with a feedforward model of the developing dynamical context (Blakemore et aI., 1998; Blakemore and Decety, 2001). The sentiment of self-induced predictable stimuli might depend inter alia on the intervention of a neural structure-the superior colliculus-which, in the case of a self-generated action, integrates sensorimotor (visual or auditive) information, while failing to be activated when predictable stimuli are not self-generated. 26 In this theory, a subject does not need to consciously represent the postulated feedforward model of her action to be able to identify a change in the environment as self-produced. A distinctive sense of the effort applied on bodily movements is the felt counterpart of the neuronal connections allowing her to identify which perceived environmental changes are self-producedY
7. INTENTIONS MISPERCEIVED: TYPES OF PATHOLOGICAL FAILURES If the discussion above is on the right track, two different kinds of misperception of one's intentions should occur, according to whether the content of the intention, or the intentional marker, is respectively misrepresented. The first kind of problem is common, affecting in particular subjects sensitive to interference (aged subjects). In these common cases already described above, To be distinguished from the sense of muscular tension: see Roland (1978). A detailed analysis of the arguments offered respectively by peripheralists and by centralists cannot be developed here. See James (1890, vol. 2, ch. xxvi) and Jeannerod (1983, ch. vi), for a review of this controversy. See Gandivia (1987), for a recent defence of a centralist perspective. 26 Stein et al. (1995), Blakemore et al. (1998). 27 Another aspect of the conscious experience of agency emphasized in Blakemore et al. (2001) is the attenuation of self-induced auditory or somatosensory sensations. This is why one cannot tickle oneself. 2. 25
Perceiving Intentions
315
an agent fails to keep track of her goal while acting, and as a consequence misperceives her initial intention. A contextual overlap may lead an agent to be perceiving intention II while in fact the motor sequence was initiated by intention 12. Such a mishap does not result, as we saw, from a perceptual failure, but rather from an instability of the intentions driving behaviour combined with the ambiguity of the reafferences (equally compatible with 11 and 12). A less ordinary case, characteristic of schizophrenic delusion of control, arises when an agent correctly detects that an intention has developed, individuates its content correctly, but fails to perceive it as self-generated. In those cases, the agent lacks the impression of 'being in charge'; she may further perceive incorrectly that some particular action of hers has been performed through her, rather than by her. 28 Note that the problem originates at a perceptual level, and not at a conceptual one: the perceiver eventually fails to attribute to herself an action because she initially failed to perceive the marker of effort while she acted, not because her access to mental concepts is defective. 29 Although such an agent is able to perceive the intentional content of her current action, and acts intentionally, she does not perceive it as self-generated, but rather as other-generated (for lack of an internal feedforward model of her action, if the supposition above is empirically correct).30 A symmetrical delusion in schizophrenic patients consists in perceiving others' intentions as their own. In such a case, a patient may be induced to take up someone else's role by completing or echoing the perceived action. Again, the difficulty may arise from the inability to discriminate, among predictable stimuli, those that are contingent on one's own intentions. 31
8. PERCEIVING ONE'S INTENTIONS AND SELF-KNOWLEDGE If the view defended here is right, then it sheds a new light on the role of perception in so-called introspective knowledge, more exactly as a precursor of self-knowledge. As many philosophers have claimed,32 a belief concerning the self cannot be gained perceptually as if the self were a thing one observes in one's environment. There is 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 discriminate 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 selfgenerated 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 selfblindness about intentions is incoherent. According to him, the potential incoherence 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 disavowing 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 necessary 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 (conceptual) 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 information concerning the observed agents. Concepts of types of action, as well as inferences 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 independent 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 himself' 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 informational 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 understanding of intention', Nature Reviews/Neuroscience, 2 (Aug.): 561-7.
35 36
Cf. Evans's Generality constraint (1982: 209). See Bermudez (1998) for a full defence of this concern
318
ladle Proust
BLAKEMORE, S.-f., RIlES, G., and FRITH, C. D. (1998), 'How do we predict the consequences of our actions? A functional imaging study', Neurospsychologia, 36 (6): 521-9. BRETHERTON, 1., MCNEW,S., and BEEGHLy-SMITH, M. (1981), 'Early person knowledge as expressed in gestural and verbal communication: when do infants acquire a "theory of mind"?" in M. E. Lamb and 1.. R. Sherrod (eds.), Infant Social Cognition. Hillsdale, NJ: Erlbaum, 333-73. BURGE, T. (1991), 'Vision and intentional content', in E. Lepore and R. Van Gillick (eds.), John Searle and his Critics. Cambridge: Mass.: Blackwell, 195-214. CAMPBELL, J. (1998), 'Le modele de la schizophrenie de Christopher Frith', in H. Grivois and J. Proust (eds.), Subjectivite et conscience d'agir, approches clinique et cognitive de la psychose. Paris: Presses Universitaires de France, 99-113. --(1999), 'Schizophrenia, the Space of Reasons and thinking as a motor process', The Monist: Cognitive Theories ofMental Illness, 82 (4): 609-25. CAREY, D. P., PERRETT, D. 1., and ORAM, M. W. (1997), 'Recognising, understanding and reproducing action', in F. Boller, J. Grafman, and G. Rizzolatti (eds.), Handbook of Neuropsychology, 2nd edn. (Amsterdam: Elsevier), 11: 111-29. CRANE, T. (1992), 'The nonconceptual content of experience', in T. Crane (ed.), The Contents of Experience. Cambridge: Cambridge University Press, 136-57. DAPRATI, E., FRANCK, N., GEORGIEFF, N., PROUST, f., PACHERIE, E., DALERY, J., and fEANNEROD, M. (1997), 'Looking for the agent, an investigation into self-consciousness and consciousness of the action in schizophrenic patients', Cognition, 65: 71-86. EVANS, G. (1982), The Varieties ofReference. Oxford: Clarendon Press. FEINBERG,1. (1978), 'Efference copy and corollary discharge: implications for thinking and its disorders', Schizophrenia Bulletin, 4: 636-40. FENSON, 1.. (1984), 'Developmental trends for action and speech in pretend play', in 1. Bretherton (ed.), Symbolic Play. New York: Academic Press, 249-70. FOURNERET, P., and fEANNEROD, M. (1998), 'Limited conscious monitoring of motor performance in normal subjects', Neuropsychologia, 36 (11): 1133-40. FRITH, C. D., BLAKEMORE, S.-J., and WOLPERT, D. M. (2000), 'Explaining the symptoms of schizophrenia: abnormalities in the awareness of action', Brain Research Reviews, 31: 357-63. GANDIVIA, S. C. (1987), 'Roles for perceived voluntary motor commands in motor control', Trends in NeuroScience, 10: 81-5. GERGELY, G., NADASDY, Z., CSIBRA, G., and BIRO, S. (1995), 'Taking the intentional stance at 12 months of age', Cognition, 56: 165-93. GIBSON, E. J., OWSLEY, C. J., and JOHNSTON, J. (1978), 'Perception of invariants by fivemonth-old infants: differentiation of two types of motion', Developmental Psychology, 14: 407-16. HARDING, C. G., and GOLINKOFF, R. M. (1979), 'The origins of intentional vocalizations in prelinguistic infants', Child Development, 50: 33-40. HAVILAND, J. M., and LELWICA, M. (1987), 'The induced affect response: lO-week-old infants' responses to three emotional expressions', Developmental Psychology, 23: 97-104. JAMES, W. (1890), The Principles ofPsychology, 2 vols. London: Macmillan. JEANNEROD, M. (1983), Ie Cerveau-machine: Physiologie de la volonte (Paris: Fayard); trans. D. Hurion, The Brain-Machine (Cambridge, Mass.: Harvard University Press, 1985).
Perceiving Intentions
319
_(1988), The Neural and Behavioural Organization of Goal-Directed Movements. Oxford: Clarendon Press. _(1999), 'La Contribution de la neuroimagerie fonctionnelle a la comprehension de la psychose schizophrenique', Bulletin de I'Academie Nationale de Medecine, 183 (3): 477-85. Seance of 23 Mars 1999. JOHANSSON, G. (1977), 'Studies on visual perception of locomotion', Perception, 6: 365-76. LESLIE, A. M. (1994), 'ToMM, ToBy and Agency: core architecture and domain specificity', in L. Hirschfeld and S. Gelman (eds.), Mapping the Mind: Domain Specificity in Cognition and Culture. New York: Cambridge University Press. LHERMITTE, F. (1983), "'Utilisation behaviour" and its relation to lesions of the frontal lobes', Brain, 106: 237-55. --PILLON, B., and SERDARU, M. (1986), 'Human autonomy and the frontal lobes: part 1: imitation and utilization behavior', Annals ofNeurology, 19 (4): 326-34. LORENZ, K., (1983), Das sogenannte Bose-Zur Naturgeschichte der Agression, Munich: Deutscher Taschenbuch Verlag. McDOWELL, J. (1994), Mind and World. Cambridge, Mass.: Harvard University Press. MANDLER, J. M. (1992), 'How to build a baby: II conceptual primitives', Psychological Review, 99: 587-604. MARTIN, M. G. F. (1995), 'A sense of ownership', in J. L. Bermudez, A. Marcel, and N. Eilan (eds.), The Body and the Self Cambridge, Mass.: MIT Press, 267-89. MELTZOFF, A. N. (1993), 'Molyneux's babies: cross-modal perception, imitation and the idea of an external world', in N. Eilan, R. McCarthy, and B. Brewer (eds.), Spatial Representation. Oxford: Blackwell. --and MOORE, M. K. (1995), 'Infants' understanding of people and things', in J. L. Bermudez, A. Marcel, and N. Eilan (eds.), The Body and the Self. Cambridge, Mass.: MIT Press, 43-69. PEACOCKE, C. (1992a), 'Scenarios, concepts and perception', in T. Crane (ed.), The Contents ofExperience. Cambridge: Cambridge University Press, 105-35. --(1992b), A Study of Concepts. Cambridge, Mass.: MIT Press. PREMACK, D. (1990), 'The infant's theory of self-propelled objects', Cognition, 36: 1-16. PRIBRAM, K. H. (1978), 'Movements and acts: distinguishing their neurophysiology', The Behavioral and Brain Sciences, 1: 158. PRIOR, A. N. (1971), Objects of Thought. Oxford: Clarendon Press. PROUST, J. (2000a), 'Awareness of Agency: Three Levels of Analysis', in T. Metzinger (ed.), The Neural Correlates of Consciousness. Cambridge, Mass.: MIT Press. --(2000b), 'Can non-human primates read minds?', Philosophical Topics, 27 (1): 203-32. --(2002), 'Imitation et agentivite', in J. Nadel and J. Decety (eds.), Imiter pour decouvrir l'humain. Paris: Presses Universitaires de France, 189-216. --(2003), 'Action', in B. Smith (ed.), John Searle. Cambridge, Mass.: Cambridge University Press. ROLAND, P. E. (1978), 'Sensory feedback to the cerebral cortex during voluntary movement in man', The Behavioral and Brain Sciences, 1: 129-71. ROLL, J.-P., GILHODES, J. c., ROLL, R., and VELAY, J.-L. (1990), 'Contribution in skeletal and extraocular proprioception of kinaesthetic representation', in M. Jeannerod (ed.), Motor Representation and Control: Attention and Performance, xiii: 549-65.
320
Joe lle Proust
ROLL, J.-P., ROLL, R., and VELAY, J.-L. (1991), 'Proprioception as a link between body space and extra-personal space', in J. Paillard (ed.), Brain and Space. Oxford: Oxford University Press, 112-32. RYLE, G. (1949), The Concept of Mind. London: Hutchinson & Co. SEARLE, J. R. (1983), Intentionality. Cambridge: Cambridge University Press. SHOEMAKER, S. (1996), The First-Person Perspective and Other Essays. Cambridge, Cambridge University Press. 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 (3): 332--40. STEIN, B. E., WALLACE, M. T., MEREDITH, M. A. (1995), 'Neural mechanisms mediating attention and orientation to multisensory cues', in M. S. Gazzaniga (ed.), The Cognitive Neurosciences. Cambridge, Mass.: MIT Press, ch. 43. WUNDT, W. (1903), Grundzuge der Physiologischen Psychologie, 5th edn. (Leipzig: Englemann); trans. E. B. Titchener, Principles of Physiological Psychology (New York: Macmillan, 1904). ZAHAVI, A., and ZAHAVI, A. (1997), The Handicap Principle. Oxford, Oxford University Press.
15
The Sense of Ownership: An Analogy between Sensation and Action 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 particu1arly 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 fundanlentallevel 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 essentially restricted to one's own case? In this chapter, I try to answer this question and 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 common form, which does not presuppose that internal and external perception introduce different modes of perception, or even that they involve altogether different ways of experiencing the world. Rather, what grounds the sense of ownership 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 experience, 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.
2. THE TRANSPARENCY PROPERTY For anyone not already in the grip of a philosophical theory, bodily experience, that is, awareness 'from the inside' of one's body and its properties, shares a number of central features with external perception (like vision and audition). To begin with, bodily experience has an irreducible spatial content. I am aware of bodily sensations as being located at various parts of my apparent body. Normally, when I feel pain, I am able to locate the hurting body part on the basis of my bodily experience only; I do not need extraneous information (not given in the very process of feeling pain) about the pain's location. Because bodily experience has a spatial content, it is conceivable that I feel two qualitatively identical but numerically distinct pains at the same time. In bodily experience, I can direct my spatial attention successively to various parts of my felt body. As visually attending to an object is typically a way of gathering further information about it, directing my attention to a state or process presented-in bodily experience is a way of knowing more about it.
l'
The Sense of Ownership
323
Moreover, bodily experience can be illusory. The phantom limb experience is hallucinatory, because one seems to perceive a part of one's body which does not exist any more. Referred pain is another case of illusion, although the correct explanation of the illusion requires a more detailed account of bodily experience. Do I experience the wrong body part as hurting, or is it only the location of the hurt body part which is misrepresented? Another point often noted is that I can make a mistake as to the type of my bodily sensation-for instance, I can mistake a mild pain for a sensation of contact or pressure. All these observations clearly favour a perceptual account of bodily experience. I shall assume here that this account is superior to the familiar Cartesian alternative, according to which we project non-spatial psychological properties onto the external world, not least because the notion of 'projection' used in this connection is rather obscure (cf. Brewer, 1995; for instance, how can I project exactly similar pains in distinct arms if space does not play any role in individuating pains ?). So here is a first attempt at describing the general form of bodily experience: (1) Experience (a particular body part is F)
Statement (1) describes a perceptual experience enjoyed by a particular subject, whose objective content (what is perceived) is described inside the brackets. On the present suggestion, what we experience 'from the inside' is the instantiation of a property F in some part of our body. The property F can be a feeling, like 'hurts' or 'tickles', but as we shall see in the second part of the chapter, it can also be a case of bodily activity, as when my action involves the activity of a particular limb, say my hand grasping a glass. Another possibility, which I am not going to explore here, is that the property F is purely dispositiona~ like many of the properties one has access to through proprioception, which gives us a space of possible bodily movements at a time.! Although the phrase 'a particular body part is F' is supposed to capture what is perceived by the subject (at the level of reference), it does not necessarily correspond to how she would herself describe the perceived state of affairs. She would rather say something like 'My shoulder hurts', using the first-person concept. However, it would be slightly misleading to use the latter phrase to characterize the content of bodily experience. That my shoulder hurts is the content of a judgement based on bodily experience, but as Hume observed, it is not dear that there is any constituent in the state of affairs disclosed by bodily experience which must be described with the first-person concept. There is nothing in the content of internal perception that tells me that I am the one who is in pain. So 'This shoulder is hurting' fares equally well as a literal description of what is perceived 'from the inside' in a particular situation. I What is at stake here is a conception of proprioception according to which what this species of 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 special 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 bodily experience can be the intentional object of external perception, in particular 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 different) 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 difference, 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 similar 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 experience 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.
I
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 'perceptual 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 acknowledge 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.
3. THE SENSE OF OWNERSHIP Our discussion so far suggests that the transparency of bodily experience is best explained on the hypothesis that it is itself a special case of external perception. There is some tension, though, between this hypothesis and the fact that bodily experience gives us a sense of ownership. Whatever property we can be aware of 'from the inside' is instantiated in our own apparent body. Bodily experience seems to be necessarily short-sighted, so to speak, since it cannot extend beyond the boundaries of one's body. The very idea of feeling a pain in a limb which does not seem to be ours is difficult to frame, perhaps unintelligible. What seems to be an essential property of bodily experience has no analogue in external perception. Of course, vision can extend beyond the boundaries of one's body. So how could bodily experience be a case of external perception? To see whether this tension can be resolved, we need a more detailed account of the sense of ownership. Martin (1995) suggested that what is at stake is rather a 'sense of boundedness', which has nothing directly to do with self-awareness, and can be accounted for entirely at the level of the spatial content of bodily experience, independently of the nature of the property F. Consider visual perception. The boundaries of a seen object always fall within the limits of the visual field. There are always visible points outside the boundaries of the object. Contrast bodily experience. The boundaries of the perceived object (the body) are coextensive with the limits of the somatosensory field. There are no (bodily) sensations to be experienced outside the boundaries of the felt body. What we conceive as a sense
326
]Crome
Dokic
of ownership is really the feeling that anything perceived through bodily experience will fall within one of one's boundaries, and so will be experienced as properties of a single body. This need have nothing to do with the properties F which are perceived as exemplified in various parts of the body. Rather, it concerns the spatial structure of bodily experience as opposed to other sensory modalities. In fact, Martin's comparison between bodily experience and vision is not quite right. Martin observes that there is a modal contrast, in bodily experience, between regions where one could currently feel sensation and regions where one could not. This contrast is allegedly responsible for what we call 'the sense of ownership'. However, vision introduces a similar contrast. As Merleau-Ponty observed, the space which vision gives us access to is not bounded by the visual field itself; it is perceived, although not of course through visual sensations, as extending beyond it, for instance, behind the subject. In this case too, there is a modal contrast between regions where one could currently perceive visual qualities and regions where one could not. One might concede the point, but insist that there is still an important difference between bodily experience and vision. Our bodily experience can confront us with a bounded object (the felt body) which fills the entirety of the somatosensory field. Admittedly, this is not the usual case. More often, one feels proper parts of one's body as figures against a background of other sensations, including bodilyones (cf. Kinsbourne, 1995: 217-18 and O'Shaughnessy, 1995: 183). In any case, the visual field is never perceived as a bounded object; rather, visible objects are always perceived within it. This difference between bodily experience and vision, even if it turns out to be a contingent one, may be what actually grounds our sense of ownership. However, once the crucial concession is made, it is much less clear that the relevant difference between internal and external perception should be traced back to the spatial contents of these experiences. The fact that all bodily feelings are perceived as properties of a single object has of course repercussions on the spatial organization of bodily experience, but such organization cannot explain by itself why we have a sense of boundedness in bodily experience but not in vision.
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 psychological 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 interpretation 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 generally, we should distinguish between the scarcely intelligible idea that we directly perceive 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 expressions 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 perceive, 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. THE FACT OF OWNERSHIP If (4) does not deal directly with the sense of ownership, it explains the fact of ownership which underlies it. The fact of ownership is that any property which is (veridically) perceived 'from the inside' necessarily is a property of one's own body. The explanation proceeds in two steps. The first step relies on the consciousness-consuming nature of properties F, and on the implicit reflexivity of
The Sense of Ownership
329
bodily experience. It can be argued that in so far as instantiated properties experienced 'from the inside' require my conscious experience, they form a subset of my psychological properties. At least, no one else's instantiated psychological properties have the power of essentially attracting my consciousness. The second step stems from the observation that properties F are not perceived luerely as being spatially located. They are perceived as being spatially located in specific body parts filling those locations (cf. Brewer, 1995: 299). On the present account, it is crucial that the primary objects of bodily experience are properties F. It is through these properties that a body part is perceived, and can be demonstratively identified as this body part. So if we perceive any body part 'from the inside', it must be the one which in fact exemplifies F. There is no question of bodily experience being illusory just as far as properties F are mistakenly ascribed to an independently identified body, which might be someone else's. It follows that in bodily experience, the perceived body parts themselves are parts of my body. The peculiar situation in which bodily experience takes place guarantees that the presented arm cannot be someone else's limb. 5 Veridical placing of properties F in material objects or parts of them makes them parts of one's real body. My hand is a part of my body because I can feel sensation or bodily activity in it. Of course, veridical bodily experience is only a sufficient condition of body ownership, not a necessary one. There are parts of one's body that are inaccessible to bodily experience (and possibly to consciousness in general). Nevertheless, my body is at least partly defined as whatever thing instantiates the set of properties F I am perceiving 'from the inside'. One will object that (4), unless perhaps (5), cannot be used to explain the sense of ownership. All that is explicitly represented in bodily experience is that bodily properties are exemplified at particular locations in objective space. The difference between observing and feeling pain is traced to a fact which is not represented in the subject's experience. So how can it be phenomenologically significant? 5 In exceptional cases, such as that of Sianlese twins, it seems that the same body part can be perceived '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 impenetrable, 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 illusory, 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 perception, 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 subject's finding the transition from a particular case of bodily experience to a judgement 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 warranted 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. proprioceptively), 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 feeling 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 experience, 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.
6. THE ANALOGY WITH ACTION In the first part, I argued that the intimate knowledge we have of our own sensations 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 indeterminate). 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 subject'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 reflexive? 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 bodily 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 normal case) and warranted.
7. THE EXPERIENCE OF CONTROL In this and the following sections, I would like to discuss two independent objections to the forgoing picture of first-person knowledge of actions. The first objection is that awareness of one's bodily activity, understood as voluntary or controlled, cannot be necessary for knowledge of one's physical actions. The second objection is that awareness of bodily activity cannot be sufficient for such knowledge. Let's start with the first objection. As Anscombe (1959) famously remarked, I normally know without observation what I am actually doing. Now 'observation' can be taken in at least two ways. Anscombe herself thought that such knowledge is not based on experience at all. On a different reading, the phenomenon Anscombe was trying to capture is that I know of my own action without relying on perceptual experience independent of my intention to do whatever I am doing. This leaves room for the view that my knowledge 'without observation' is nevertheless based on experience, though not one which is separate from my very intention in acting. The distinctive knowledge we have of our own actions does not seem to be anything like the knowledge we have of our own names or dates of birth, which is
334
Jerome Dokic
clearly not grounded on any (contemporary) experience. To know what I am actually doing, I have to know not only why I am doing it (the goal), but also how I am doing it (the route I take to attain the goal). This knowledge is normally based on bodily experience. I know that I am intentionally stretching my arm to grasp a glass before me at least partly because I am aware of my body 'from the inside' as being involved in some recognizable activity, which includes an arm movement.l° If by 'introspection' is meant a non-perceptual way of knowing one's mental properties, the privileged knowledge we have of our own physical actions is not in general purely introspective. On the contrary, it is internally related to some perceptual awareness of the corresponding bodily activity. Suppose, for the sake of argument, that it was not so related, but was instead conceived as introspective knowledge of one's trying considered as a fully psychological event. For instance, I am introspectively aware that I am trying to grasp the glass before me. I might simultaneously experience bodily activity which culminates in the glass being grasped, but the latter experience is separate and independent of the former awareness. The problem with such a picture is that once knowledge of action is severed from experience of bodily activity, the agent cannot bring them back together except by inference or association. The agent's situation would be such that, having launched an action, she would have to identify which observed changes in the world are genuine realizations of her attempt. Such identification would necessarily be based on defeasible evidence, whatever it might be. This does not seem to be the general case. Normally, we know in an indefeasible way that our body is active by way of our trying to do something. In fact, we very often know more than this; we are able to recognize immediately some perceived movements as constitutive parts of our action. This can only be explained by allowing that knowledge of many of our actions essentially involves some experience of bodily activity. Since such experience is perceptual-like, it can be illusory. Someone can experience bodily activity in her seeming arm even if the arm is not where it seems to be, is paralysed, or there is no arm. In so far as awareness of one's action is based on bodily experience, one can always be wrong about what one is doing. One could then argue that when one's action fails or is cancelled at the last moment, one has a conscious (perhaps reflective) access to a motor image (cf. Jeannerod, 1994; Pacherie, 1997) which fully grounds one's knowledge of what one was trying to do, without any help from bodily experience. Perhaps this is true. However, when one's action is not a complete failure, and essentially involves some bodily activity, there is always more specific knowledge about what one is (intentionally) doing to be gained from bodily experience. The goal of an action is often fixed in advance of any bodily movement, and does not normally change during the course of the 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 representation 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 consciousness. 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, requiring 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 representations. 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 representation, 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 difference 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 movements is one way of knowing what one is doing, even though it is an entirely practical way.
8. ACTION VERSUS BODILY ACTIVITY 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 identified 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 indirectly 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
According to the second objection, knowledge of one's physical actions cannot be fully grounded on awareness 'from the inside' of one's bodily activity, in so far as ' one cannot simply identify actions with such activity. Indeed, it can be argued that in typical cases, knowledge of one's actions involves not only bodily experience, but also external perception and introspection. Consider first the role of external perception. I know that I am about to grasp that (seen) glass, but what I am aware of 'from the inside' is only structured bod- ' ily activity. In particular, I do not experience the glass 'from the inside', I understand this activity as part of a determinate goal-directed action because I perceive it as being systematically related to other aspects in my perceptual space, like the glass. What grounds my knowledge of what I am doing is my whole perceptual experience of the situation, bodily and external. If this part of the objection is conceded, knowledge of our own actions is indeed based on both internal and external perception. We would have to qualify again Anscombe's remark that we know of our own actions 'without observation'; after all, I have to observe the glass to know that I am grasping it. An alternative move would be to distinguish between what enables me to do something and what justifies my knowledge of what I am doing (d. Falvey, 2000, who expands on a distinction drawn by Burge, 1993). My visual perception of the glass is what enables me to grasp it the way I do, but it is not part of my justification for the claim that I am grasping that glass. From an epistemological point of view, knowledge of my action is exclusively grounded on bodily experience of controlled activity. After all, bodily experience should give me a sense of what is the goal of the activity, which in turn cannot be understood otherwise than by reference to the perceived glass. Let's turn to introspection-the second element supposedly missing from our account of first-person knowledge of actions. What is the role of introspection, conceived as a way of being non-perceptually aware that one is in a specific mental state, in our knowledge of our own actions? Might introspection be entirely lacking from the state of being aware of one's action? To begin with, in order to know that I am about to grasp the glass, I do not need to reflect explicitly on the fact that the experienced bodily activity is dependent on my visually perceiving the glass; it is enough to understand it as being dependent on the features and spatiallocation of that glass. Perner (1998) has convincingly argued (cf. also Dienes and Perner, 1999: 744-5) that controlling actions is a content-based process, as opposed to a process which is merely sensitive to representational vehicles. However, he also claims that enough aspects of content must be explicitly represented to indicate that the action is desired. In other words, when the action involves a higher-order level of
The Sense of Ownership
339
control,14 the agent must represent the motivating pro-attitude as such (with its specific world-to-mind direction of fit), on pain of confusing its non-factual content with factual content (whose direction of fit is reversed). This requires introspection, which is essentially a reflective or meta-representational act. However, it is not clear that an explicit reference to the psychological pro-attitude ('I am trying to do X') is the only way of avoiding the confusion. As Roessler (1998) pointed out, one can be unreflectively aware of one's goal, as something that should be done (and is not done yet), without being reflectively aware that one has a certain goal. In fact, as there is a non-reflective or immersed way of being aware of a limb hurting, without thinking about the experience of pain, there is a non-reflective or immersed way of being aware of one's goal-directed activity, which does not involve thinking of one's activity as part of an intentional action-at least to the extent that such thinking requires a reflection on the underlying pro-attitude. 15 Yet at the same time, perceiving pain 'from the inside' can ground a self-ascription of a pain experience even if the pain is not strictly identical with the non-spatial experience which discloses it. Analogously, perceiving one's bodily activity 'from the inside' can ground a self-ascription of action (and so a reflection on the underlying pro-attitude) even though, as I have suggested, such activity is not strictly identical with the action. Indeed, it would be more appropriate to identify the action itself, and thus the trying, with the non-spatial experience of bodily activity. There is no question of denying that some higher-order actions have essentially a reflective component. My claim is only that the distinction between actions that have such a component and actions that do not have it does not coincide with the distinction between controlled and automatic actions. There are actions which are entirely unreflective, and yet require consciousness of them on the agent's part, which makes them controlled according to our definition. Another important fact which is compatible with the present account is that knowledge of one's action is in general much enriched by introspection. Our actions typically presuppose a more or less tacit background of propositional attitudes like beliefs, desires, and prior intentions. A fuller awareness of one's actions should include a knowledge of the way they result from these underlying attitudes, which are often dispositional (Campbell, 1998). Such knowledge is certainly at least partly introspective, and arguably also participates in the common sense of ownership. However, I think that we should distinguish between the basic sense of ownership associated with bodily experience, and the more sophisticated sense of ownership associated with the agent's recognition that her action flows from long-term beliefs, desires, and prior intentions. Whereas the former is 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 beliefdependent. 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') goal-directed bodily activity in what is otherwise correctly perceived, for instance, through proprioceptive sensations, as her own body. (ii) The capacity of experiencing controlled activity 'from the inside' is preserved. Yet, the subject refuses to identify herself with her action because she does not perceive it as being in line with her long-term propositional attitudes. In the first case, the subject lacks a sense of ownership in the basic sense; she does not have the impression of controlling the activity 'from the inside'. In fact, it is not clear that there is any intentional action to be ascribed to the subject in this case. Yet she experiences goal-directed movements in her own body. A speculation is that this is what happens with patients struck by the Anarchic Hand syndrome (d. Della Sala et al., 1994, and Marcel, Ch. 2, this volume, for a description of this syndrome). The second case is very different. Here, the subject has a sense of ownership in the basic sense, but lacks a sense of ownership in the more sophisticated, beliefdependent sense. Still, there is an intentional action in these cases, although the agent denies responsibility for it, implicitly rating her impression of controlled bodily activity as a (perceptual) illusion. If I understand Campbell (1998) correctly, this might be what happens in some cases of schizophrenia where the patient does not recognize the perceived action as hersP Let's take stock and answer the second objection raised above. Experience 'from the inside' of one's bodily movements, perhaps with non-introspective, perceptual experience of their relations to the rest of the environment, gives us a core sense of what we are doing, in a way which is often enough to ground knowledge of one's physical action. So even though one does not, strictly speaking, perceive one's action 'from the inside', knowledge of it inherits essential features from the grounding bodily experience.
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).
I
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 relationship between internal perception and non-perceptual, introspective knowledge 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 certain 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 contrast 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 epistemological 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 knowing 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.
Jerome Dokic
342
15.1. Bodily experience as an eccentric form of perceptual awareness
TABLE
Possibility of (self- )blindness
Possibility of illusion
External perception (visual, tactile, etc.)
Yes
Yes
Internal perception (bodily experience)
No
Yes
Introspection (non-perceptual self-knowledge)
No
No
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 phenomena. Some of them can be directly perceived, whereas others need interpretation and inference, or (in one's own case) introspection. Pains and intentional movements (and perhaps emotions) belong to the first category, whereas experiences and propositional attitudes like beliefs, desires, and intentions belong to the second 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 frontomesial 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 mentalisation 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 THIRD-PERSON EPISTEMOLOGICAL
IMPLICATIONS OF DUAL ASPECTISM (1) In my book The WillI I put forward a dual aspect theory of physical action along the following lines. The theory begins with the claim that whenever we perform an action an event of the type <will' occurs, and that the act in question is identical with that event. Before this particular assertion can be determinate we need to know the sense with which <will' is used. The ordinary language term that is generally invoked to explicate the concept is
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 considerations, 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 aetprogenitors, act-desire and act-intention, activate that system, and that this activationsyndrome 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 immediately 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) crowning 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 question 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 willing 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 phenomenon 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 movement, 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 epistemologically 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 theory 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 theory 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 philosophical 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 theory, 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 awareness 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 proprioceptive blackout: for most creatures in most situations it would lead to a quasiparalysis. 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, senseperceptual 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 inference that bodily action is going on, which is to say that successful action is occurring, 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.
2. THE FIRST -PERSON EPISTEMOLOGICAL IMPLICATIONS OF DUAL ASPECTISM (1) This brings me to a third element of epistemological shortfall on the part of
sense-perceptual experience of one's own physical actions. We have seen that to the agent the physical act is usually mediately perceptible by sight, and near universally immediately perceived proprioceptively. Then while proprioceptive experience is attentively immediate, even though causally mediated by kinaesthetic sensations, the brand of immediacy realized does not exemplify what one might describe as 'mental immediacy': that is, immediacy of the kind encountered with sensations and experiences and beliefs, and the fact that simple unproblematic proprioceptive illusions are possible as with the other senses confirms the fact. Here we have a further epistemological lack on the part of sense-perceptual experience of physical action, to be added to the fact that it is neither individuative of its event object, nor identificatory of its activeness of character. It is clear that each of these properties are essential elements of the normal awareness of one's own actions. One needs to be aware of the act qua a single event, and to be aware of it qua active in type, and one needs moreover to do so in a mode of immediacy that enables the phenomenon in question to be immediately related to the rest of one's mental life. And in fact such an experiential identification of the act satisfying these requirements is a reality. While it is not possible to do so sense-perceptually, if the
350
Brian O'Shaughnessy
dual aspect theory of action is correct then at the time of action identification of the act occurs in mental terms under the concept of 'will' which at one stroke individuates its object as a single event, types that event as active, and does so in a way that has a right to be described as mentally immediate. Now in the normal COurse of events such an immediate awareness of the act in mental terms is accompanied by kinaesthetic awareness of the act, and this fact has 1 believe caused some philosophers to deny the very existence of an active mental event: the kinaesthetic phenomenon occupying centre-stage mentally and tending to hide the mental element of activeness. Nevertheless, there can be little doubt that in an exceptional situation of total failure an active mental event would be openly encountered. For example, suppose that in some normal-seeming circumstances one tried to move a limb-which failed to budge an iota. Looking back, reporting that '1 tried to move the limb at that moment', one would be referring to an active event which came to consciousness without the usual kinaesthetic sense-perceptual accompaniments. This exceptional situation lays open to view a mental and active event which is present on normal occasions of success, and which tends on such occasions to be concealed by one's kinaesthetic experience. Then the content of the dual aspect theory of physical action can be expressed in terms of the aforementioned active and epistemological experiences. The claim by the theory is that the active event of which one was immediately aware on the occasion of total failure, is both present in normal situations of active success and is one and the same event as that given in non-individuative and non-identificatory form in sense-perceptual experience which is at best attentively immediate in a mode which falls significantly short of the immediacy encountered in the mind. (2) 1 wish now to consider the nature of the experience that is given in this retrospective fashion to the active subject on the occasion of total failure. I want at this point to ask a few questions concerning the active mental event of which retrospectively one seems to be aware at such times. The first thing to be considered is, whether the event one refers to in saying 'At that moment I tried to move the limb' was experienced. Now if tlle event actually occurred one would know so. How would one know? Certainly not inferentially. In particular, not on the basis of an inference of the following kind: 'I know that I firmly intended to act at that instant, nothing happened in the limb at that instant, therefore 1 must at that instant have tried to move it': that is, an inference based on the existence of a firm absolutely contemporaneous intention to act. While the inference is valid enough, it is not used by the subject as his way of knowing of the event. The knowledge of which we speak is immediate in a sense inconsistent with the use of inference. Meanwhile it is not quite like the brand of immediate cognitive access that we have to many non-experiential mental phenomenon: say, to the knowledge that one intends visiting Spain next month, the kind of immediacy relating us to the present existence of (say) our intentions and
The Epistemology ofPhysical Action
351
beliefs. Thus, while one has immediate access to one's beliefs, one can none the less readily imagine momentary inaccessibility of a belief or item of knowledge in a way one cannot the sheer existence of present experience. True, one might selfdeceive concerning the character and object of an experience, but it is difficult to imagine sheer ignorance of the existence of intense present experience. No phenomenon is epistemologically closer than present experience. Then my suggestion is that this is how we stand in relation to the above example of completely failed trying. And in fact it is surely truistic that the event in question was experienced. It is obvious that it would find a mention if one were cataloguing the contents of one's inner experiential life of the last few minutes. It would find a mention as the intention of visiting Spain would not. (3) The obvious conclusion is, that the completely failed event of trying to move the limb was experienced at the time it happened. What I want now to discover is the character of that experience. To begin, it helps to consider the spatial properties of this trying event as given to the experiencing active subject, and to compare them in this regard with sensations. Now in general we may say of sensations that they are experienced as in body-sites, and generally as in body-relative physical space. For example, we experience a twinge of pain as a toothache, a tickle as on one's forehead, whatever their actual literal physical site. This property of sensations enables us to have two qualitatively indistinguishable sensations at the same point in time: say, a dull ache in one's left thumb and the same in one's right thumb. While one cannot have two such sensations simultaneously in the same body-site, since body-sites individuate them, one can simultaneously have them in different sites. The spatial framework individuates them in the first place, and the temporal framework does so secondly. Consider now the event of trying to move an arm whose psychologicality was revealed nakedly to view because of the complete absence of kinaesthetic experience. We shall see that the situation here is significantly different in the above respects from that realized with sensations. Thus, where was that psychological event, of which one was immediately aware, experienced as taking place? One is perhaps inclined to say one experienced it as being at the felt location of the arm. However, it seems to me that to do so would be to confuse the site of the targetobject for the will, with the event of willing that is directed to that place. Then whereas two qualitatively indistinguishable sensations can appear simultaneously in consciousness, thanks to the fact that we individuate sensations spatially, there is no way a failed attempt to move a limb could do the same. There is no way such an event with such an object could be given simultaneously at different places in experienced body-space. This is not because it is given at the felt bodysite of its target-limb and spatially individuated through that criterion, which is the situation in the case of bodily sensations. Rather, it is because there cannot be two such simultaneous events. I can simultaneously fail to move left and right
352
Brian O'Shaughnessy
arm, but I cannot simultaneously fail to move one and the same arm in two different presentations. The event in question is exhaustively specified by time, type, and object-content. Space finds no place in that specification. This event is not experienced as in body-space. Then since experiences are not and perceptibles are given in space, this strongly suggests the event is an experience and not a perceptible. (4) How would one describe this event? That is, what description would one give of the phenomenon one recalls when one recollects the event of striving as nakedly revealed to view on an occasion of total failure? Let us suppose this event occurred completely unexpectedly. Then of what would one be aware at the moment of failure? To be sure, one would be aware of affective experiences of the type of surprise, shock, alarm, and suchlike. However, these are merely consequences of the occurrence, products of one's awareness of the event. But what would one say as a description of the experience itself? Simply, 'I tried to move my arm-in a forward direction-at normal enough speed' (and perhaps also a reference to intensity). But what of quale? What of felt or experienced character? But what conceptual room is there for variation of this sort? And in fact what room for quales of any kind? After all, we have an object-directed phenomenon of type 'try', a given object, sayan arm movement with a limited few properties, and that is all. This event seems wholly exhausted by its type, content, and perhaps also intensity. While it has the accidental attribute of happening when it does, everything else seems essential: take away the type, change the object in any respect, and it is simply a different animal. This strangely simple phenomenon is the event that normally occurs in the mind when ordinary successful examples of limb moving occur. In itself it is no different on the two occasions. Of course there will be a few differences when it takes place in the exceptional context of complete failure. For example, the farther reaches of the act will not eventuate, the driving intention will fail differentially to express itself, and so on. However, these differences do not imply that the event of which we are mentally aware when ordinarily we move a limb has significantly different mental characteristics from those listed above. The event in question will still be exhausted by type and content: that is, when brought under purely psychological concepts. The concept of quale, of phenomenological character or 'feel', will still lack application. For to repeat: what this event psychologically is seems exhaustively given when we conjoin the type, strive, with the intended object, limb movement of such a variety. In short, the general character of the phenomenon nakedly revealed in the pathological situation of total failure, puts on display the mental character of the normal phenomenon of success. (5) One final observation before I draw a few deductions from the above. I have suggested that the event filtered out for view by complete failure, of type trying/striving/attempting and directed to limb movement, enters the catalogue of phenomena when one lists the contents of experiential consciousness, of
The Epistemology of Physical Action
353
the 'stream of consciousness' of the moment. Then although if dual aspectism is true that one and same event will on occasions of success be proprioceptively experienced, which is to say proprioceptively perceived and in that specific perceptual sense 'felt', it is not 'felt' in some purely psychologically immediate mode. We feel our pains when we are aware of them, which is to say when we perceive them, but I do not think it true that in the situation of total failure we feel the event of trying to move the limb in question. Thus, a mental event occurs which is directed in some way to a part of the body whose physical presence one proprioceptively feels, but the awareness of that mental event is not a mentally immediate experience of the type: feeling that phenomenon. It seems on the face of it to be no more than the having of it. This fact suggests that the awareness in question must be non-perceptual in type. And it lends weight to the supposition that the concept of quale lacks application to this phenomenon-since any distinctive quale of bodily striving ought, one assumes, to fall under the concept of 'feel'.
3. DEDUCTIONS FROM THE FOREGOING OBSERVATIONS (1) These observations suggest an answer to two fundamental questions: What is the general type of the psychological event of trying to move a limb? What is the general type of the awareness of that event? The following considerations clarify the nature of the problems expressed in these questions. I claimed that the event of trying, filtered out for view in purely psychological form in a situation of total act-failure in the complete absence of kinaesthetic experience, is experienced by its agent. One knows of its occurrence, not through inference from the firmness of an absolutely contemporaneous intention and not just mentalistically immediately (in the way one knows one's beliefs). One knows of that trying mentalistically immediately, but in the rather more special mode one knows of present experience. 'Then the foregoing observations in Section 2 shed light on what we mean when we say of that trying that it was experienced. And they shed light on the nature of our awareness of the event of trying. For it is not prima-facie obvious whether these latter are the same thing. It could on the one hand be that an experience that was active in nature occurred and immediately caused knowledge of itself. Alternatively, it could be that an active psychological event occurred that imn1ediately caused an experience of awareness of itself, and that the awareness experience caused knowledge of its active psychological object.
(2) The two possible model situations are as follows. (a) At a certain point in time a thought crosses one's mind, causing knowledge
of itself.
354
Brian O'Shaughnessy
(b) At a certain point in time a sensation occurs, causing an event of noticing of
itself, which in turn causes knowledge of the sensation object of noticing. Now in either case knowledge occurs which is based on experience, but only in the latter case are there two phenomena which could be said to be experienced, viz. the sensation and the perception of the sensation. In one case an experience occurs and it is of such a nature as to engender knowledge of itself, in the other case an experience takes another psychological item as its experienced object, and is such as to engender knowledge of that psychological object. The question we need to decide is: which (if either) of these two models fits the situation in which complete motor-failure filters out for view the active event of trying or striving or attempting in purely psychological form? What the question boils down to is. something very simple. Namely, is the active psychological event that was laid open to view itself an experience? Or is it instead the immediate extensional object of an awareness experience? Was the experience that led to knowledge active in nature? Or was it instead of an epistemological character, viz. the perception of one's own willing? (3) It seems to me that the data uncovered in Section 2 point to the former answer. They suggest that the active psychological event revealed in the special situation of failure simply is an experience. That is, that when one casts one's mind back to the experiential situation and says 'I tried', one is recalling one and not two phenomena given immediately in experience. Thus, it suggests that there exists an experience which is one and the same thing as trying to move a limb. An obvious conclusion, one might think, but I disagree. What is obvious once one has engaged in the requisite thought-experiments, is that on such occasions one is in the presence of an active psychological event that is in some familiar sense given immediately in experience. But it is not obvious in what way it is thus given. My suggestion is, that the active event is given in experience merely in being an experience. The aforementioned data which lend support to this reading of the situation consist in facts of the following kind. Namely, that the event laid open to view by the absence of all kinaesthetic sensory experience was: (i) (ii) (iii) (iv)
active in nature, not given as located in body-space, not given as of an experienced spatial extent, not capable of being simultaneously accompanied by a second psychological phenomenon of identical type and object, (v) not endowed with a phenomenology, being (vi) entirely determined by type and object-content (and perhaps intensity).
Properties (ii) to (iv) characterize experiences generally, and much else that is mental but not an experience. But they do not characterize the· immediate
The Epistemology ofPhysical Action
355
psychological objects of awareness experiences. To return for a moment to the example of sensations mentioned earlier: while the two simultaneous qualitatively indistinguishable pains in one's two thumbs are each experienced in spatial terms, the experiences of those pains are not so experienced. This is in line with the supposition that the active psychological phenomenon is of the type, experience. (4) At this point I move on from the above arguments in support of this claim. These arguments have taken the form of characterizing the way the event of striving is given experientially to its agent in situations of total failure. I shall turn soon to explanatorial considerations which point to the same reading of the situation. However, before I do so there are a few general observations worth making on this question, stemming from the nature of mental action. The first is, that none of the above questions can be seriously raised in the case of mental action. For one thing, theories of the will are never cast in terms of mental action: they are invariably encountered in the form of a theory concerning the origin of active bodily movements. Moreover, there is no highly exceptional situation in the mind that would count as an analogue of total active failure in the case of bodily action, and even if there were it would not act as a filtration agency laying open to view an active mental entity whose existence has been questioned by some. In any case, such a device is simply not needed as support for the doctrine of the reality of the mental will. This is because mental willings are an unchallengeable category. No one seriously doubts that there exist mental occurrences which are active in nature, nor can many people be disposed to offer analyses of these phenomena in which that active character is reduced to the simultaneous presence of an initiating intention and its desired goal event. Everyone knows that we do such things as try to listen, and succeed and fail in varying degrees; that we try to remember names, and sometimes succeed and sometimes not; and it is clear that the trying in question is an irreducible event of trying. Then the point I want to make, which is relevant to the main issue at present under discussion, is that mental acts of all kinds, whether listenings or visualizings or thinkings, are all of them experiences and pretty plainly so. For one thing, they are phenomena which are individuated in terms of a purely temporal framework. Thus, they are such that two incidents of trying to remember the same name, phenomena which from the point of view of the experience were qualitatively indistinguishable, must have occurred at different times. Moreover, when we come to report the occurrence of the active experience in question, we do so merely on the basis of having had the experience: we do not base our claim upon the existence of a further distinct awareness-experience. And they find a place in any narrative of the 'stream of consciousness' at the time. And so on. Then although there are significant differences between willings of the latter mental ilk and those directed to immediately felt limbs, the fact that the mental variety is of the type experience is evidence that the bodily variety is also. True, the
356
Brian O'Shaughnessy
differences between these two examples of the will are great: the bodily phenomenon being primitive in the extreme, since such willings must be present at the very lowest point of the psychological life of animals of the very simplest kind yet probably differ in no significant respect from the bodily willings of rational selfconscious creatures. (For I can see no reason for supposing our bodily willings to be different from those of other animals.) By contrast, exercises of the mental will encompass a great deal of the inner life of self-conscious subjects. Nevertheless these phenon1ena each give immediate expression to act-desire and intention, are subject to both control and n1eaning-giving at the hands of their intention source. And they are each of active type. And so on. I mention these facts in support of the theory that the event of striving that is encountered in physical act situations is of the type experience.
4. EXPLANATORIAL ISSUES
Explanatorial considerations support this interpretation of the facts, as I hope now to show. The main explanatorial issue to be resolved is, how at the time of action an agent knows that his will is operative, i.e. that he is doing something or another (at the very least). The question is, whether his knowledge is an immediate effect of willing or arises through the mediation of an awareness experience. The answer should indicate the answer to the central question of this chapter: nan1ely, whether the active mental event (Le. willing) is an experience. If the knowledge is an immediate effect of the phenomenon of willing, then the willing must surely be an experience; if it is experientially mediated, it cannot be. I shall approach this problem by asking a preliminary question concerning our knowledge of future action. Let us suppose a man is about to perform an intended action, for example is racing across a room to extinguish a flame. What explains his believing an act is about to occur? The obvious answer is, the presence of an intention. It entails such a belief, and simultaneously causes it. However, the desire that the flame cease, and knowledge that it is within his power to effect it, must also determine his knowledge of its impending occurrence. After all, the intention is to perform an act with a desired and possible end in view: the commitment to future action is not a blind commitment. It is not as if the inducements to action cause a commitment to future action which simply puts those inducements behind it, so to say a rigid and unthinking pointing of oneself action-wards! Thus, both the intention as well as the belief and act-desire cause knowledge of future willing. Then what happens to these causal powers at the time of action? Consider specifically the intention. If the intention causes knowedge of willing immediately before willing occurs, why should it lose this power as willing comes into existence? And yet it is clear that the event ofwilling causes knowledge of itself, and does so through (in some as yet undetermined mode) being experienced by its
The Epistemology of Physical Action
357
agent. Then what cognitive causal scope can there be left for the intention at the time of action? So one might be inclined to say. But in fact the cognitive situation during action is no different in the relevant respects from that encountered with future action. For just as knowledge of future intended action is not through being aware of a blind commitment to future action, so knowledge of present willing is not through being aware of willing in sheer isolation: it relates us to a willing to which of necessity we are intentionally committed. And to a willing we desire to do and believe possible. In short, these causal agencies-experiencing the event of willing (in some as yet undetermined mode) on the one hand, and the intention, desire, and belief on the other-are not in the nature of opposed possibilities, nor do they relate to one another as contributory conditions of the occurrence of act and cognition. They are all of them necessary conditions both of the present intentional action and our knowledge of it. No one of them could be causally operative in engendering such knowledge in the absence of the others. The question we must now settle is: what is the 'as yet undetermined mode' in which willing is experienced? The question this in turn naturally raises is: under such circumstances what cognitive role would be open to a distinct experiential event of the type, awareness-of willing? That is, what cognitive role would be open to an experiential event of the type of perception of the will? It seems to me that in such a situation none remains, that the explanatorial space is already fully occupied by intention and belief and act-desire (and in a different mode by willing itself): they guarantee knowledge of present intentional willing without need of a further agency. The explanatorial context in which willing occurs is such as to nullify any cognitive role for perceptual experience. However, when perception occurs a cognitive role must be realized, whether or not it is used. The deduction I think we should make is, that since no cognitive role is available for putative perception, the possibility of immediate perception of the will must automatically be nullified. This adds explanatorial weight to the previous arguments in support of the claim that we 'just know' of willing when it occurs. That is, for the inexistence of any distinct awareness experience. Thus, the act of will generates knowledge of itself simply through being an experience: awareness of willing is no different in this regard than (say) awareness of one's own thoughts and emotions. The overall conclusion is, that the act of will is of the generic type, experience.
17
On Knowing One's Own Actions Lucy O'Brien 1. INTRODUCTION
Given the recent debates about self-knowledge and first-person authority it is surprising that there has not been more discussion about our knowledge of our actions. l It is surprising because our knowledge of our own actions seems, prima facie, to share many of the features of our knowledge of beliefs and perceptions that have given rise to these debates. 2 At least this is what I want to suggest in this chapter. Indeed, perhaps the rnain motivation for this chapter is to place discussion of our knowledge of our own actions firmly alongside our knowledge of other psychological phenomena. I will outline TvVhat seem to be intuitively plausible features of our lmowledge of our actions, and consider what account we might give of such knowledge that respects those features. A suggestion as to what form an account of our knowledge of our actions should take will be offered. These tasks will constitute the explicit content of the chapter. However, it is also my hope that discussion of the knowledge we have of our own actions will serve some further aims. I hope it will serve to extend the range of possible sources for self-knowledge. By 'self-knowledge' I mean not just knowledge of that thing which is in fact the knower, but knowledge of ourselves as ourselves. The account sketched here promises to give us a way of knowing of our own actions that is unavailable to anyone other than the acting subject. An assumption that forms part of the background to this chapter is the assumption that actions are as primitive a psychological phenon1enon as beliefs 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 that constitutes these debates. .
On Knowing One's Own Actions
359
and perceptions. 3 And it aims to consider what account we should give of our actions so understood. In doing this, the chapter aims to take its place in a move away from a general passivism about the psychological, a move away from the idea that the basic case of a mental phenomenon is a mental state or disposition. A Martian reader of contemporary philosophy of mind would be quite likely to think that we do not really believe that we judge or act. Or at least she would think that we understand such things as psychologically non-basic: as structured groups of more basic psychological, and possibly non-psychological, states. While there are many who claim themselves to be non-reductivist about psychological phenomenon, the reduction or elimination of mental activity in favour of mental passivity is in fact widespread.
2. FEATURES OF OUR KNOWLEDGE OF OUR OWN ACTIONS Let us consider a simple case of action. I want to catch my friend's attention and I raise my arm in order to do so. Now consider my knowledge of my action of raising my arm. Intuitively, such knowledge is puzzlingly easy to come by. I seem to know directly and authoritatively that I am voluntarily raising my arm in a way that I do not know others' actions directly and authoritatively. Further, when I know that I am raising my arm, I seem to know it no later than when I have started to raise it. In particular, I do not seem to have to await perceptual information, for example, that muscles are contracting and that my arm is rising in order to know it. Also, wanting to know what I am doing seems to be all that is required in order for me, in normal cases, to know what I am doing. If I wonder what I am doing when I am raising my arm, I seem to be immediately supplied with an answer. My knowledge of what I am doing seems, normally, to be immediately available given that I am acting. Let us separate these features of our knowledge of our action under three headings and say that intuitively our knowledge of our actions appears to be:
First-person authoritative. An agent seems to be authoritative over her ovm actions, and in a way that she is not over other's actions. Relatively a priori. Our knowledge of our actions appears to be spontaneous and to be given immediately upon acting. It does not seem to require any investigation or to be based on evidence. Relatively transparent or self-intimating. Our actions, like other psychological phenomena, seem to have a certain conditional epistemic availability to us. It 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 interested. 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 moving 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 voluntary 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 acting, 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 4 It should be noted that given bodily awareness-kinaesthesia in particular-it might be thought that we also have first-person authority about whether our bodies moved. It seems to me that the nature of the authority attached to such cases is complex and is weaker than the phenomenon identified, but I will not discuss it further here.
On Knowing One's Own Actions
361
where such a thing occurs, another may be in a position to put me right. But note that we have here a parallel with perception, and possibly belief. We claim a firstperson authority with respect to our own perceptions, and beliefs, even though we can think that we are perceiving something or having belief about something when we are not. Further, given the possibility of self-deception and unconscious action, we might fail to know that we are acting when we are. In such cases, another subject who sees us moving may be better able than us to know that we are acting. My deep desire to catch someone's attention may result in my voluntarily dropping my handkerchief, but the desire may be a sufficiently uncomfortable one for me to entirely disavow the action, and for me to think that it was involuntary. Nevertheless, it seems clear that given certain background conditions which we are entitled to take as met (such as that the subject's body is functioning normally or that there are no special repressive mechanisms in place) we can assume a firstperson and third-person asymmetry over the question of whether the subject acted. Authority with respect to Knowing that I Acted vs. Knowing what I Did Let us suppose that it is agreed that we do seem to know that we have acted in a way that is in principle unavailable to others, and agreed that this seems to give us a certain kind of epistemic authority over our actions. It may nevertheless be said that this does not give me authority over what I have done. To know that, it may be said, I have to have recourse to my perceptual faculties. And if my knowledge of what I have done is grounded in perception, then any first/third-person asymmetry in my knowledge of what I have done will be due to the fact that only I have perceptual access through bodily awareness to the activities of my own body. It will not be special to action, but rather will be a feature of my knowledge of bodily movement more generally. Well, I can clearly be ignorant, relative to others, of many of the things I have done. The tendency to describe our actions in terms of their effects-effects that may well be unknown to us-means that we very often can be said not to know what we have done. Given that actions can be described in terms of their consequences, and given that I can be ignorant of the consequences of my actions, I can be ignorant of my actions under such descriptions. It is perhaps mainly for this reason that we overlook the authority that we have over own actions. S So, it is important to admit that I did not seem to have any special authority about what 5 Physical actions are not the only psychological phenomena we describe in terms of their consequences. 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 perception, 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.')
Lucy O'Brien
362
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 6
The example comes from O'Shaughnessy (1980, ii: 85).
On Knowing One's Own Actions
363
what they are doing. It seems, however, that while the subject will also be authoritative about what her purpose in acting is, and authoritative about the fact that she is acting for the purpose of doing one thing rather than another, she may not be authoritative about whether her purpose came off. Consider, in contrast, a subject's relation to her basic actions. Basic actions are those actions a subject can carry out directly, without having to do anything else; they are the actions that a subject needs to do in order to do anything else. I think it is plausible to suppose that the descriptions which correspond to a subject's basic actions will be descriptions in terms of bodily movements, and so plausible to suppose that the subject is authoritative with respect to bodily movement descriptions such as 'raising my arm' or 'lifting my foot'. I am, however, not going to aim to settle here the question of exactly which actions are basic actions. The point I want to urge here is that, whichever actions we think are basic actions, we are justified in supposing that an agent will have a non-conceptual grasp of the possible ways in which they are able to act directly. For, if an agent had no such grasp of the possible ways that they could act directly, it is hard to see what sense we could make of the agent determining to act in one way, rather than another, in order to realize their aims. Given that the ability to carry out a basic action is presupposed by the ability to carry out any action, we are justified in supposing that it is a precondition of any action that an agent has a grasp of the possible ways they can act, which are in this way basic. If the above is right, then I think we have some way of seeing how it could be that an agent can be first-person authoritative not only about the fact that they have acted but also about what they have done. The suggestion is that knowledge of what one is doing, and not just knowledge that one is acting, comes from ones acting against a grasp of possible things one could have done as basic actions. It is my opting to carry out this basic action, rather than that, which, assuming I have the requisite concepts, will give me the description under which I am authoritative with respect to the action. Further, although I may be authoritative with respect to a basic action ljI, and I may intend or try to do
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 compelling. 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 muscles 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 perceive 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 psychological 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 askingthernselves the question <What am I doing?' not being able normally to answer the question 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 awareness 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 actionsactions done for no reason-from intentional actions. If there is such a distinction 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 controlling 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 supposed to have no possibility of accessing the former. If there were a general disassociation 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 disrupted 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 control 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 cannot be actions. 11 See e.g. O'Shaughnessy (1980, ii, ch. 10) and Ginet (1990: 3). By a non-intentional or subintentional action is meant not just an action which is not intended under one description but an action that is intentional under no description.
Lucy O'Brien
366
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 3.1 The Dual Component Model 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 character, 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 distinctive 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 direction 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 movements, we might think that it fails properly to capture the first/third-person-thirdperson 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 perceptual breakdown, we may think that a subject would in fact be able to continue acting 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 knowledge 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
tryings, an account which was supposed to explain the distinctive features of action self-ascription, can non-problematically be provided. The assumption seems to be that our intentions or tryings will involve no external component and will be capable of being the objects of direct awareness. However, first, we would need to make sense of such direct awareness of any supposed internal components of our actions, and, second, we have no good reason to think that the contents of our intentions or tryings will be any more independent of their relations to external elements of the environment than the contents of our beliefs and perceptions. It is highly plausible to think that a creature inhabiting an arm-free world cannot even try to raise an arm. Third, the model can seem unsatisfactory for much the same reason that a dual component theory of belief, and of the knowledge we have of our beliefs, has struck theorists who think of belief as a unified psychological state as unsatisfactory. If we think of actions as unified psychological phenomena, rather than aggregative constructions of psychological and non-psychological components, then we are not going to adopt the view outlined above. As I have said, I am not going to offer a defence of this view of actions here but merely consider what account of our knowledge of them will seem plausible for someone who accepts such a view. Considering how a dual component theorist will respond to the problems raised brings out the fact that there are in fact many different forms of dual component account. There are as many different forms as there are combinations of the different accounts one could give of the nature and knowledge of a person's knowledge of their intentions or tryings, on the one hand, and of nature and knowledge of the movements of their body, on the other. The dual component model envisaged above is only one version of a dual component account. In dealing with the first problem raised above, it is important to stress that a contemporary dual component account is likely to hold that our knowledge of our actions depends not just on our five senses, but also centrally on bodily awareness,!4 And given a view on which bodily awareness is necessarily awareness only of my body, there is room to say that, even in the perceptual component, there is a first/third-person asymmetry. So, there is room to say that first-person authority can be reflected in both components of the dual component account. However, we still have a problem with relative aprioricity-it will still be the case that I need to await the deliverance of my perceptual capacities in order to know whether I have raised my arm. This has seemed to be phenomenologically counter-intuitive. It rather seems that my raising my arm, say, gives me ceteris paribus the knowledge that I moved my arm without my having to monitor feedback from bodily 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 feedback 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 knowledge that I am raising my arm. We can imagine our dual component theorist, in response to the second problem 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 knowledge 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 discussed, 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 movements, 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 version 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 continued 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. 3.2 The Inference Model The suggestion that we appeal to an inference to explain our knowledge of our actions can only be the suggestion that we make inferences about our actions from either our observations of our bodies or from our knowledge of the precursors of actions: intentions or tryings. The idea in the former case would have to be that I come to know what I am doing from making an inference on the basis of perceptual information, through bodily awareness and the other senses, about the movements of my body. It seems clear that, if the remarks made above-against the dual component theorist's idea that part of our knowledge of our own actions flows from our sensory access to our bodily movements-were well taken, then this account will also be held in doubt. It will sit well with neither the seeming aprioricity of our knowledge of our actions, nor with their relative transparency. What of the suggestion that we infer that we are acting from our knowledge that we are intending to act, or trying to act, together with an assumption that our motor systems and so on are all in working order? Well, first, it is very important to stress that trying, in this context, has to be understood, much like an intention, as an independent precursor of my actually acting. If the picture is that my trying is part of my action, then it would conflict with our unity thesis for actions. If the supposition is that my trying is my acting then, whilst there may be objections to the identification, the attempt to give an account of the trying would be to give an account of the action. Given this clarification, on this version of the inference model I have know-ledge of my actions via direct knowledge only of my intentions or tryings and the assumption, say, that my motor system is in working order. Consider the account as told for intentions (and given the idea of my trying as an independent precursor of my acting, I think that what will be true for intentions will be true for
On Knowing One's Own Actions
371
tryings).t 5 On this account, I know that I am doing ljJ indirectly as an inference from knowing that I intend to ljJ and that my motor system is in working order. The first thing that strikes us about such an account is that it does not fit our naIve understanding of what is going on when we know we are acting. It does not seem to me that I know I am raising my arm as a result of inferring that I am from my knowing that I intend to raise my arm; it rather seems that I know that as a result of my raising my arm. On this account, no further epistemic support is given by my doing anything. Further, it seems implausible that the subject need have beliefs about their motor system in order to be credited with knowledge of what they are doing. But even putting aside these considerations, without more being said, the grounds given in the account do not seem to be sufficient to ground knowledge that I am ljJ-ing. Maybe I usually do what I intend to do, but all too often, even if I could do what I intend to do, I do not. It would seem that my intention to ljJ, plus assumptions about my motor system working properly, may plausibly ground a justified hypothesis that I willljJ, but it does not seem able to ground my knowledge. The above inference models do not seem easy to accept. There are more sophisticated variants of the latter form of the inference model. These, however, fall under what I call the anticipation model which will be discussed later on. 3.3 The No-Reasons Judgemental Model 16 In reviewing accounts of self-knowledge with respect to belief, Peacocke identifies what he terms 'no-reasons accounts'. According to no-reasons accounts, our beliefs about our beliefs count as knowledge although they are held on the basis of no reasons. In particular, the first-order belief does not constitute, on such an account, my reason for my self-ascribing my belief. The simplest of such accounts has it that as a matter of fact agents like us, due to some internal mechanism, only self-ascribe beliefs that they actually have. Given that a self-ascription of the belief that p is a reliable indicator of a belief that p, if we accept reliablism about knowledge, it can be said that by such selfascriptions agents like us express knowledge of their beliefs. The equivalent account for the self-ascriptions of actions would have it that as a matter of fact we tend to be reliable about what we are doing, due to some non-epistemic feedback mechanism, and in virtue of that we are capable of knowing what we are doing. Such simple reliablist accounts are acceptable to few in making the connection between the psychological phenomenon, and the knowing of it, an entirely brute connection. Such accounts tend, therefore, to be embellished in such a way that 15 If my trying to do something is not supposed identical with, or part of, my acting when successful, 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 selfascriptions 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 subjects 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 knowledge 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 selfascribe would not count as actions. We might also claim that an ability to be authoritative 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 accessible to the subject in a certain way, this does not mitigate the need for an independent 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 3.4 The Anticipation Model One obvious way to make true our beliefs about an area of which we have control, is to bring about what we believe rather than have our beliefs conform to the facts See Burge (1996). These are in effect Peacocke's criticisms of Shoemaker and Burge. See Peacocke (1999, ch. 5). Despite this worry about Burge's account, it will be obvious that much else that is said in this chapter is influenced by the discussion in his (1996), and also in his (1998). 17
18
On Knowing One's Own Actions
373
as they independently are. Our actions are, paradigmatically, an area over which we have control. One suggestion for why we seem to be authoritative about what we are doing, and why we are able to know what we are doing without any process of evidence gathering, is that we do what we believe we are going to do. Velleman, in his Practical Reflection, advances a sophisticated development of this suggestion. Velleman advances the basic thesis that our knowledge of our own actions has the features it has because we normally do what we believe we are going to do. However, his account also aims to meet the question as to why we should be motivated to do such a thing-why do we not just wait and see what we do and form our beliefs about what we are doing accordingly? The answer is taken to lie in a deeply rooted desire for self-knowledge, in particular, a desire to know what we are doing. Given a desire to keep up with what one is doing we will be motivated to bring it about that we accurately foresee what we are going to do, and that motivates us actually to do what we foresee doing. Furthermore, Velleman identifies intentions with self-fulfilling expectations: to form an intention is to form a belief about what I will do, that I am poised to make true on account of my desire to know what I am doing. Given a framework in which I do what I expect to do, a most effective way for me to get myself to do something-which is after all what an intention is-is to expect that I will do it. Velleman's account has the attractive feature that it explains the transparency of our actions and the spontaneous, relatively a priori, nature of the knowledge we seem to have of them, while also allowing that the authority we have over our selfascriptions is not groundless. He distinguishes between having knowledge that is adduced from evidence and knowledge which is supported by, though not adduced from, evidence. We are, on his account, justified in claiming to know what we are doing, not because we form the belief about what we are doing on the basis of evidence, but because we form the belief within a framework which we know makes the belief true. Self-knowledge is a kind of justified invention. Before considering this account further, it is worth nothing that it too has a parallel account in the literature on the self-ascription of belief. Accounts of belief selfascription have it that the reason our self-ascriptions tend to be authoritative, while appearing spontaneous or groundless, is that what in part determines the beliefs we have is what we are inclined to self-ascribe. If our beliefs are in this way ascription-dependent-we believe what we take ourselves to believe-then there is no epistemic gap between our belief that p and our taking ourselves to believe p that needs to be bridged by reasons or evidence.l 9 We need to be careful, however, to distinguish three distinct claims that may be conflated. First, few would doubt that our self-ascriptions are determinative of our beliefs merely in the sense that answering the question: 'Do I believe p?' will tell me, almost always, whether
19
See Wright (1989) and (1998).
374
Lucy O'Brien
indeed I do believe p. This is just first-person authority. Second, few would doubt that settling an answer to the question: 'Do I believe p?' will not only tell me whether I do indeed believe p already, but will also settle any questions I may have of whether to believe p. As Evans has made clear, our procedures for determining whether we believe p are the same as the procedures for determining whether p.20 However, neither this claim, nor the former one, commits us to the view that my second-order beliefs, my beliefs about what I believe, are in any way epistemically or ontologically prior to my first-order beliefs. Rather, what they make clear is that our beliefs, and our beliefs about our beliefs, are in a kind of concord. Third, and more contentiously, in claiming that our belief ascriptions are determinative of our beliefs we might be saying, not just that the same procedures can be used to answer both questions, but rather that the process of determining what I believe is, in the normal case, what brings my beliefs about. This makes my self-ascriptions ontologically and epistemologically prior to my beliefs, and introduces a kind of psychological idealism in being committed to there being no fact of the matter about what I believe prior to my reflections explicitly about what I believe. 21 Velleman's idea that our knowledge of our actions is somehow invented has most in common with the third claim identified: in both cases our self-ascriptions have the features they do because the self-ascriptions make the facts fit them. However, there are of course differences between actions and beliefs which make a significant difference to the acceptability of the accounts being offered. While actions are single occurrences that can be willed, beliefs are states which cannot. Thus, while we can make sense of a system that brings it about that it does 0/ because it anticipates doing l/J and desires self-knowledge, it is very much harder to make sense of a system that brings it about that it believes p somehow because it self-ascribes the belief p. The latter suggestion makes our self-ascriptions appear ungrounded and our beliefs, therefore, either arbitrary or a matter of will. However, even if not problematically idealist, the trouble with accounts of the sort Velleman proposes, is that they are-as he fully realizes-bound to seem to have things backwards. We seem to have reasons to do things that are not grounded in beliefs about what we take ourselves to be about to do. Indeed, we could have reasons to do things without any capacity for first-person higher-order beliefs. Further, as we noted before, looked at naIvely our beliefs about what we are doing are given epistemic support by what we are in fact doing. Velleman, as we have seen, is keen to emphasize that an account of the kind he offers does not need to hold 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 assuming that intentions, understood as beliefs about what we are going to do, are necessary 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, however, 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. 3.5 Knowledge Through Participation In what follows I will attempt to provide a sketch of what a more plausible account might look like. Let us go back to the naive thought that our beliefs about what we are doing are given epistemic support by what we are in fact doing: we know what we are doing because we are doing it. Now let us suppose given the arguments presented above, that our action cannot function in our justification for its self-ascription in virtue of some observation of it, or inference from its effects or causal antecedents. What of the possibility that the action can function unmediated as the reason for its own ascription? Let us then consider a view according to which we know we are acting when we are, because, and for the reason that we are. 22 Now presented merely as such, the account faces a problem. It provides us with no explanation of how an action, in so far as it need only be world directed, can bear any normative relation to our judgements about ourselves to the effect that we are acting. To put it simply, what is it about Smith's raising his arm that can support Smith's knowledge that he is raising his arm? 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 attention. 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 further explication. He identifies a number of distinct mental phenomena that can be said to occupy without being objects of attention-perceptions, conscious attitudes, 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 conscious 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 paradigms 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 physical 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 complex 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 certain 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 attending 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 occupies 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 suggestion 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 property 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 consciousness 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 selfawareness. 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 subject 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 assessment 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 assessment 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,
Lucy O'Brien
380
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 knowledgeable 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 reference. 25 Certainly, if being self-aware is understood to imply a capacity for firstperson 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 awareness 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 available, 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 considerable 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 understood 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 ANSCOMBE, G. E. M. (1963), 'Two kinds of error in action', journal of Philosophy, 60: 393-40l. BROADIE, F. (1967), 'Knowing that I am doing', The Philosophical Quarterly, 18: 137-49. BURGE, T. (1996), 'Our entitlement to self-knowledge', Proceedings of the Aristotelian Society, 96: 91-116. --(1998), 'Reason and the first person', in MacDonald et al. (1998). CASSAM, Q. (ed.) (1994), Self-Knowledge. Oxford: Oxford University Press. CASTANEDA, H.-N. (1999), The Phenomeno-Logic of the I: Essays on Self Consciousness. Bloomington, Ind. and Indianapolis: Indiana University Press. DANTO, A. C. (1963), 'What we can do', journal of Philosophy, 60: 435-45. DONNELLAN, K. (1963), 'Knowing what I am doing', journal of Philosophy, 60: 401-9. DUNN, ROBERT (1998), 'Knowing what I'm about to do without evidence', International journal of Philosophical Studies, 6 (2): 231-52. EVANS, G. (1982), Varieties of Reference. Oxford: Clarendon Press. GINET, C. (1990), On Action. Cambridge: Cambridge University Press. MACDONALD, c., SMITH B., and WRIGHT C. (eds.) (1998), Knowing Our Own Minds: Essays of Self-Knowledge. Oxford: Oxford University Press. MCGINN, C. (1982), 'The structure of content', in A. Woodfield (ed.), Thought and Object. Oxford: Clarendon Press. OLSEN, C. (1969), 'Knowledge of our intentional actions', The Philosophical Quarterly, 19: 324-36. O'SHAUGHNESSY, B. (1963), 'Observation and the will', journal ofPhilosophy, 60: 367-92. --(1980), The Will: A Dual Aspect Theory. Cambridge: Cambridge University Press. PUTNAM, H. (1975), 'The meaning of "meaning''', in his Mind, Language and Reality, ii. Cambridge: Cambridge University Press. PEACOCKE, C. (1998), 'Conscious attitudes, attention and self-knowledge', in MacDonald et al. (1998). --(1999), Being Known. Oxford: Oxford University Press. SHOEMAKER, S. (1996), The First Person Perspective and Other Essays. Cambridge: Cambridge University Press. VELLEMAN, J. DAVID (1989), Practical Reflection. Princeton: Princeton University Press. WRIGHT, C. (1989), 'Wittgenstein's later philosophy of mind: sensations, privacy and intention', journal ofPhilosophy, 26: 622-34. --(1998), 'Self-knowledge: the Wittgensteinian legacy', in MacDonald et al. (1998).
18
Intentional Action and Self-Awareness 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 illustrates 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 reading, this claim is relatively uncontroversial. We do not normally acquire knowledge 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 helpful 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 'twoknowledge' view unattractive have opted instead for a 'two-component' view. As K. Donnellan put it, 'knowledge of our own intentional actions is complex', dividing 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 conceived 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 pursue 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.
2. INTENDED CONSEQUENCES AS AFFORDANCES According to the ecological approach to perception, in perceiving the environment we co-perceive ourselves (Gibson, 1979). Ecological self-awareness, as Gibson
Intentional Action and Self-Awareness
385
explained it, has two principal aspects. First, the optic flow provides information about the movement and spatial position of the perceiver. Second, we perceive objects as affording actions. For example, we can see whether an object is within reach, a door wide enough to walk through, or a chair the right height to sit on (Neisser, 1993). In perceiving affordances, we perceive things relative to our own dimensions, abilities, and perhaps inclinations. In this sense, information about affordances is self-specific. In a similar spirit, Ulric Neisser has offered the following account of an ecological sense of agency: All of us are aware when we have done something ourselves. When I slap my hand on the desk, for example, I see and hear and feel the consequences of my action. The optic and acoustic and mechanical effects occur just when and as they should, given the control schema that initiated the movement. A very different experience occurs when I passively allow you to pick up my hand and bang it on the desk for me. The sensory consequences may be quite similar, but the coincidence with my own intention is missing. In this context, intention refers to the activation of particular movement control structures. We perceive actions as our own if and only if their consequences are appropriate to the schema by which they were generated. Dancers, atWetes, and others skilled in bodily motion are especially sensitive to that fit. (1993: 9; emphasis in original) Suppose we want to appeal to the concept of ecological self-awareness in explaining the source of agents' knowledge of their current actions. The problem we are facing here can be brought out by distinguishing three ways of characterizing the perceptual experience of intended consequences. (1) I perceive my hand's slapping on the desk. (2) I perceive that my hand is slapping on the desk. (3) I perceive that I am slapping my hand on the desk.
While (1) characterizes my perceptual experience by reference to the perceived event, (2) and (3) tell us which facts are evident to me in perceiving that event. In (2) I am said to be aware of an intransitive fact, a fact about what is happening to my hand; in (3), my experience is described as an awareness of a transitive fact, a fact about what I am doing with my hand. In the passage quoted, Neisser begins by pointing out that the intended consequences of our actions normally include events we perceive. He ends with the claim that in perceiving such events we perceive transitive facts. At least, this would appear to be a natural construal of the suggestion that we perceive actions 'as our own'. The transition is motivated by the thought that agents are sensitive to the fit, with regard to various parameters, between a perceived movement and the control schema which initiated the movement: it is in virtue of this sensitivity that the transitive facts are supposed to be manifest to the agent. What is involved in being sensitive to the fit between schema and movement? Neisser's discussion does not provide much help with this question, but I think we can extract one answer to it from the developmental evidence he uses to illustrate
386
Johannes Roessler
his proposal. It has long been observed that even during the first few months infants take pleasure in controlling interesting visible and (especially) audible events in their environment. In Neisser's view, such behaviours suggest that 'infants are aware of themselves as active agents virtually from birth' (1993: 16). To illustrate, the following is a representative example of what Piaget called secondary circular reactions: At 0; 3 (10), after Laurent has learned to grasp what he sees, I place the string, which is attached to the rattle, in his right hand, merely unrolling it a little so that he may grasp it better. For a moment nothing happens but, at the first shake due to chance movements of his hand, the reaction is immediate: Laurent starts when looking at the rattle and then violently strikes his right hand alone, as if he felt the resistance and the effect. The operation lasts fully a quarter of an hour during which Laurent emits peals of laughter. (Piaget, 1953: 162)
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 movement, 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 property. 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 interpretation 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 reading, 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 understand 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 experiencing 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 environment. 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 evidence 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 perceive 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 distinctive 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 perceive 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.
3. ACTION AND ATTENTION A central element of our ordinary understanding of attention is a distinction between actions we perform more or less automatically and actions that are performed deliberately and, in that sense, occupy attention. We invoke the distinction, for example, in explaining the number of things we can do at the same time: how well you can simultaneously play scales on the piano and do mental arithmetic will depend on the extent to which at least one of these tasks has become automatic for you. We also take it that failure to pay attention can be responsible for everyday slips of action, such as in the often-quoted example from James: 'Very
Intentional Action and Self-Awareness
389
absent-minded persons on going to their bedroom to dress for dinner have been known to take off one garment after another and finally get into bed, merely because that was the habitual issue of the first few movements when performed at a later hour' (1890a: 115). Perhaps most interestingly, we think of attention as something that can disrupt well-practised performances, as when a skilled typist focuses her attention on the Inovements of her fingers. A further ingredient in this common-sense understanding of attention is the idea of a connection between attention and awareness. You may tap your fingers on the table, say, without being aware of doing so; and we associate this lack of awareness with the fact that your attention was engaged with other things. But what should we make of this idea? On what might be called a constitutive reading, paying attention to one's action simply is to be aware, in some sense, of what one is doing. Alternatively, on an explanatory reading, the fact that someone pays attention to her action explains that she knows what she is doing. My aim in the rest of the chapter is to pursue a version of this explanatory reading. In this and the next section, I discuss the role of attention in action. In section 5, I will return to the issue of self-knowledge. Let us begin with the relation between the role of attention in action and its role in perception. The metaphor that tends to be used to characterize the latter is that of an attentional 'spotlight': there is a contrast between a foreground and a background of attention, between those perceptually presented items that are currently illuminated by the 'spotlight' and those which are not (but to which it can normally be moved straightaway if they engage the subject's interest). What is the relation between selective perceptual attention, in this sense, and attention as it is involved in the control of action? I want to consider briefly two accounts of this relation; I will then argue that we should reject the first account in favour of a modified version of the second. 1. In James's influential discussion of voluntary action, the two aspects of attention are made out to be phenomenologically quite different phenomena (James, 1890b; Norman and Shallice, 1986). James characterizes deliberate action as involv-: ing 'an additional conscious element, in the shape of a fiat, mandate, or express consent' (189Gb: 522). He later explains the nature of the conscious experience here by reference to a sense of (mental) effort, the effort needed to sustain the idea of a to-be-performed action in the face of inhibiting tendencies. What is conspicuously absent, in this account of the phenomenology of deliberate action, is any appeal to perceptual experience. For James, as far as perceptual experience is concerned, deliberate action is on a par with what he calls 'ideomotor' actions, actions controlled by the 'bare idea of a movement's sensible effects', for which he offers the following exan1ple: 'Whilst talking I become conscious of a pin on the floor, or of some dust on my sleeve. Without interrupting the conversation I brush away the dust or pick up the pin' (1890b: 522). In James's view, perceptual awareness of an
390
Johannes Roessler
object is necessary even for object-directed ideomotor actions. He concludes that appeal to perceptual experience cannot explain what is distinctive of the phenomenology of deliberate action. 2. A different taxonomy has been proposed by Cedric Evans. According to . Evans, there are two kinds of voluntary attention: interrogative attention-'the attention we pay to an object in order to enlarge our knowledge'-and executive attention-'the attention we have to give those of our performances that require a technique for their execution' (1970: 100). It would be a mistake to interpret this as a distinction between perceptual attention and attention in action. Rather, what Evans has in mind is that there are two different ways in which perceptual consciousness may be structured by attention, corresponding to two kinds of success condition for attention. Success may be defined by the question you wish to answer, or by the intention you wish to execute. So paying attention to a bodily activity is a matter of using perceptual information for a practical purpose. There is a common intuition that the distinction between a theoretical and a practical use of attention corresponds to a difference in the kind of content carried by perceptual experience. Evans puts the point by saying that executive attention need not be 'propositional in form'. Its success is not a matter of 'some facts being noticed' (1970: 101). Similarly, Naomi Eilan (1998) has contrasted the kind of intentional content perception has when it provides the answer to a theoretical question with its content when the subject is immersed in a practical task. In the former case, she suggests, the content is conceptual or propositional; in the latter, it is causally indexical in the sense discussed in the last section-it displays things as, say, being within reach or liftable, where the subject's grasp of the causal significance of this is purely practical. But on the face of it, executive attention too involves using perception to answer questions. It seems natural to describe an orchestra player watching the conductor's cue, say, as looking to see when, and how fast, to play. Or suppose you have to negotiate a narrow staircase with an unwieldy piece of furniture. You will look to see where to move next, when to swerve to the left, and so on. In this kind of case, perceptual experience seems to yield a direct answer to a practical question-a question as to how or when to do something, not just a question as to what is the case. Of course, the subject need not explicitly formulate, or reflect on, the question. What the picture of perceptual experience delivering answers to the subject's questions is intended to bring out is just that perceptual attention is typically guided by the subject's interest in particular kinds of information, an interest she could articulate by formulating questions. It seems to me that using perception to answer practical questions is a central element of what we ordinarily mean by paying attention to bodily actions. To put it in Neumann's helpful phrase, we think of attentional control as a mode of parameter specification (Neumann, 1984, 1990). When you perform a bodily action, a number of
Intentional Action and Self-Awareness
391
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 information. 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 intentional 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 distinction between attentional and non-attentional parameter specification as a distinction between two kinds of causal explanations. When perceptual experience provides the subject with answers to her practical questions, it yields an understanding 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 fingers. Neither the subject herself, nor, indeed, common-sense psychology in general, 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'.
4. PERCEPTION AND PRACTICAL REASONS
There are a number of difficulties with this simple distinction between rational and non-rational aspects of action control, however. For example, there are phenomenological 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, information 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 ceiling. One method would be to measure the distance between the floor and the ceiling, 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 general' 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, without scrutinizing them. To use an example that features prominently in an experiment 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 contradistinction 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 ceiling 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 suggestion 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 experience 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 consists in the subject's selection of perceptual information for the rational control of action, in contrast with the non-rational, 'automatic' (and usually faster) specification of parameters by perceptual and motor skills. In many cases, the application 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 distinction between two ways in which perceptual attention may be involved in the control 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 environment' 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 initiation, 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 distinction James was getting at when he argued that ideomotor actions lack the both phenomenologically and causally relevant
Intentional Action and Self-Awareness
395
Let me end this section with another qualification. I have suggested that the distinction between attentional and automatic parameter specification captures a significant ingredient in the common-sense psychological conception of intentional 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 position. 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 experience 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 direction. (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 psychology. 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 processing that actually governs those actions' (1995: 177-8). This might suggest the radical 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 perceptual 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 connection 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.
5. ATTENTION AND SELF-KNOWLEDGE 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 erection of a building which he cannot see and does not get reports on, purely by giving 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 perceptual 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, mentioned 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
I
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, someone '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 'observation' 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 attending 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 knowledges' 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 attention 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
5
See O'Shaughnessy (1980, ch. 8).
398
Johannes Roessler
thinking once you have started to act? Surely by that time you must have made up your mind. One obvious reponse would be that the intention you have at the beginning of an action often leaves many details to be filled in. But a more important point is the following. Even if you have mapped out all the steps in advance, there is a sense in which executing your intention involves the acquisition of new intentions, at least if we individuate intentions at the level of sense, rather than just at the level of reference. It is one thing to intend to turn left when you reach the junction over there, it is another to intend to turn left now. To execute the first intention you need the second intention, and to acquire the latter you need to get your perceptual experience to tell you when to turn left. The point applies even to very simple actions. If you intend to raise your arm up to a certain point, it is part of your intention to stop raising it when you have reached that point. By attending 'from within the act' you will, when the time has come, acquire the intention to stop raising your arm now. Suppose this account is along the right lines. Why should observing and attending 'from within the act' be conflicting projects? It might be suggested that the conflict arises from the difference between theoretical and practical questions. In so far as you have your mind on your current bodily action, you are concerned to see how to go on, not what is happening to the material you are working on. You aim to acquire intentions, not beliefs. But that suggestion is clearly inadequate. As already indicated, there is no question but that in attending 'from within the act' you do acquire knowledge of the material you are working on. It is by acquiring such knowledge that you acquire the intentions you are interested in. It is the perceived facts that provide you with reasons for acting. (You see when to stop raising your arm by seeing that your hand has reached the right height.) Furthermore, when we consider the case of attending in preparation of an action, rather than attending 'from within the act', we find that observation and practical attending can be integrated perfectly naturally. An orchestra player waiting for her cue will observe the conductor's activities, with a view to acquiring both observational knowledge of the cue and, thereby, the intention to start playing. The problem, then, is to find a way of doing justice to the intuition expressed in O'Shaughnessy's incompatibility thesis, without denying that the perception of intended effects yields perceptual knowledge. The intuitive force of the incompatibility thesis, I think, derives from a distinctive causal ingredient in our notion of observation. It is not just that observation involves a causal dependence of experience on the world. Rather, observation involves an understanding of this dependence, a sense of one's perceptual experience being controlled by the world. If you are engaged in observation, you take yourself to be passive with respect to the facts you perceive in something like the following sense: you may actively focus your attention on a particular object, and you may actively try to obtain the answer to a particular question, but otherwise the content of your experience is entirely controlled by the world, not by yourself. (So establishing your responsibility for
Intentional Action and Self-Awareness
399
a perceived fact would have to be a matter of inference.) In this specific sense, observation is essentially a matter of gaining knowledge of the mind-independent world. This line of thought would suggest that observation and executive attention are at cross purposes in so far as they aim to generate experiences with incompatible causal roles. If you observe where your hand is moving, you aim for the mindindependent facts about the hand's movement to be responsible for the course of your experience (and, thereby, for your belief); if you are attending 'from within the act' of moving your hand, you aim for your unfolding intention, and hence your experience (qua answer to a practical question), to be causally responsible for the direction of the movement. The project of observing your own action, then, would require a kind of bad faith. As observer, you would have to deny responsibility for events which, as agent, you take to be under your control. Does this gloss on the incompatibility thesis leave any room for agents' perceptual knowledge of 'what is happening to the material they are working on'? I suggested that such knowledge plays a crucial role in the practical thinking involved in paying attention to one's actions. But how do we come by such beliefs, if not by observing what is happening to the material we are working on? On the face of it, we are left with two equally unpalatable options: either you have to act blindly, rather than in the light of what is happening to the material; or you have to oscillate between two incompatible modes of attending, devoted to being affected by the world, and to affecting the world, respectively. The assumption which is responsible for this dilemma is that perceptual knowledge is co-extensive with knowledge by observation. It is this assumption which I think we should reject. From the point of view of agency, we are interested to see, or hear, or feel, not what is happening, mind-independently, to the material we are working on, but how we are affecting the material, what difference we are making. And arguably, our perceptual experience speaks to that interest immediately: it tells us to what extent we are succeeding in executing our intention. In other words, it tells us what we are doing. Let me try to bring out how this contrasts with observation. Suppose you find yourself in a shopping-centre, and watch a video screen that shows someone walking down a passage. In this kind of situation, it is of course perfectly possible for you to observe your own actions, consistently with O'Shaughnessy's point. Walking on firm ground makes minimal demands on your attention, so you are free to devote it to finding out what is going on on the screen, and you may not immediately realize that you are the person you are watching. But suppose you suspect that it is you, and wish to verify this hypothesis. The natural way to proceed would be to try to use the monitor to control your movements. For example, you might form the intention to raise 'that arm' (where you are referring to one of the arms of the person on the screen), and use the screen to modulate the intended movement, by looking to see how much further to go, or when to stop. At this point your epistemic relation to the perceived movements undergoes a profound shift. Intuitively, if the screen can be used to guide
400
Johannes Roessler
your actions in this way, your control over the movements on the screen will be immediately manifest to you. In seeing where the arm is moving, you see where you are moving it. Needless to say, this does not mean that you might not be mistaken. Perhaps you are not the person on the screen at all-the correspondence between the perceived movement and your intention might be a sheer coincidence. The point is that, in this case, we would describe your experience as deceptive: it looked to you as if you were controlling the movement. It might be objected that this picture still leaves the agent with the problem of having to assign two incompatible causal roles to her perceptual experience. The causal dependence of perceptual knowledge on the perceived facts is surely a defining element of perceptual knowledge, whether we call it observation or not. So in acquiring perceptual knowledge of what she is doing, the agent has to think of her experience both as affecting the world and as affected by the world (namely, by the facts as to what she is doing). Actually, though, the two roles are perfectly compatible, indeed they are made for one another. It is, partly, by informing you of what progress you are making that your experience helps to control what is happening to the material. So attending 'from within the act' does indeed involve an important element of passivity, but passivity in this sense inconsistent with control over the material. You can take yourself to be a passive recipient of infor- mation as to whether you are succeeding, without disowning responsibility for what is happening to the material you are working on. Still, it might be said, to know that you are moving your arm, you need to know, first of all, that your arm is moving. How should we explain that knowledge, if not by appeal to your sense of being affected by the mind-independent world? But this picture is just what I think we ought to reject. When you attend 'from within the act', your knowledge that the arm is moving cannot be detached from your knowledge that you are moving it. You acquire knowledge of both facts in perceiving the movement with a sense of control. The only sense in which you are affected by the world is that your action gives rise to a perceptual experience of the movement as something that is under your control. How is this 'sense of control' to be explained? The obvious starting-point is Neisser's claim that you perceive a movement, or the effect of a movement, as your own action if it occurs just when and as it should, given your intention. But did we not see earlier, in discussing infants' circular reactions, that this is not sufficient for propositional knowledge of one's actions? We saw that infants seem to enjoy a non-reflective sense of control. But there is a whole complex of differences between circular reactions and actions occupying the subject's attention, on the conception I have been spelling out. The most fundamental difference is this. In so far as perceptual experience is used for the rational guidance of actions, providing answers to deliberative questions, it yields intentions that have a propositional content, a content that stands in inferential relation to the content of some of the subject's desires and beliefs. Such intentions can be explained in terms of
Intentional Action and Self-Awareness
401
the subject's practical reasons. Connectedly, unlike mere desires, they commit the subject to certain beliefs, for example, the belief that she will be able to perform the projected action. None of these features are present in the case of circular reactions. Certainly the latter are goal-directed actions, controlled by the agent's conscious perceptual experience; but they are not controlled by intentions with propositional content. (If we want to speak of intentions here at all, we have to suspend some of the normal connotations of that term.) This has an important bearing on the nature of agents' perceptual expectations. No doubt an infant operating a rattle expects to continue to hear the sound of the rattle. But her perceptual expectations arise simply from being habituated to the contingency between movement and effect. In contrast, in so far as perception is used in executing intentions, the agent has a good reason for particular perceptual expectations. If you intend to raise the arm that is shown on the monitor, you have a distinctive kind of reason to expect to see the arm go up. It is not that you have any evidence that makes this a sensible prediction. Rather, you are committed to bringing it about that the arm goes up, and this commitment itself gives you a reason to expect to see this happen. The question remains: how can you be aware that you are moving the arm, simply in perceiving the movement of the arm? I think there is an illuminating analogy with the following, more familiar question: how can you be aware of the objective, mind-independent world, on the basis of your subjective experience, embodying a particular point of view? A way of putting the problem here is to ask how the following claims can both be true: perceptual experience is, as Strawson puts it, permeated by concepts of mind-independent, persisting, physical objects (Strawson, 1988); yet it essentially represents things in terms that reflect the particular point of view occupied by the subject, for example, as being in front or to the left. How can experience be essentially perspectival, representing things in subject-dependent terms, yet constitute an understanding of how things objectively are? Two ideas that have featured prominently in recent work on this issue are the following. One is that it is a mistake to consider particular experiences in isolation. Again, in Strawson's men10rable phrase, the perception of something as a mind-independent object of a certain kind is 'animated by, or infused with ... the thought of other past or possible perceptions of the same object' (Strawson, 1974: 53). The second idea is that perceiving the world around us provides us with direct knowledge regarding the causal enabling conditions of our experience. It presents us with the world around as one of the factors responsible for our current experience (along with, among other things, our current spatial position). (See Ayers, 1991; Brewer, 1999.) Arguably, the two ideas are not unconnected: it is natural to put the first idea to work in developing the second one. Thus, it may be argued that for the perspectival content of a given perception to constitute a grasp of how things objectively are, the subject needs to understand how things would appear from alternative perspectives, and that, if they did not show the expected
402
Johannes Roessler
appearances, this would cast some doubt on the veridicality of her current experience. Put differently, the subject needs to understand that the content of her experience carries a certain commitment as to how things will (or would) appear if she moves (or moved) around. To the extent that the course of her experience bears out perceptual expectations arising from her understanding of the way things objectively are, it will involve a sense of her experience being controlled by the mind-independent world. The most obvious point of analogy is that perceptual experience used for the rational control of action is also, in its own distinctive way, infused with the thought of other possible perceptions. Perceptual expectations arising from one's current intentions provide one example. Another example is the thought of how things would look or feel if one were to proceed in various directions open to one. Furthermore, in both cases, appeal to such infusion may contribute to an account of a perceiver's 'sense of control', her understanding of the causal relations between world and mind. In the case of objective perception, we find the course of our perspectival experience predictable and intelligible in terms of the concepts of mind-independent objects and properties that permeate it. This helps to explain how it is possible for perspectival experience to constitute a grasp of how things objectively are. In the action case, to the extent that the agent's perception fulfils perceptual expectations arising from her intentions, she will find the perceived facts readily intelligible in terms of the action concepts employed in her intention. Thus, in perceiving what is happening to her material, she will be aware of what she is doing. Again, in both cases, there are philosophical pressures (of a 'broadly Cartesian' nature) towards thinking of our apparently immediate perceptual knowledge as inferential: in the first case, as knowledge based on an inference from immediately accessible subjective experiences to objective reality as their best explanation; and in the second case, as knowledge based on inference from perceptually presented effects to the success of one's trying (again, as their best explanation). It is widely agreed that, in the first case, the pressure should be resisted. I suggest we should resist it in the second case too. Let me end with a comment on a line of thought, specifically concerned with the action case, that might be used to justify yielding to the Cartesian pressure. Suppose we take self-awareness to be a matter (at least in part) of occupying the 'first-person perspective', where this is conceived as being able to acquire noninferential self-knowledge in certain distinctive ways. A notable feature of judgements expressing such knowledge is that they use the first person. Another striking feature is that they are immune to error through misidentification relative to the first-person pronoun. For example, if you think 'I am in pain' (on the basis of feeling pain), then there is no danger of getting wrong who it is you know to be in pain. In contrast, when you judge 'I am bleeding' on the basis of seeing a number of bleeding hands tangled up, your judgement might be wrong on account of your having misidentified the person your experience shows to be
Intentional Action and Self-Awareness
403
bleeding. Now it might be said that perceptual judgements about one's actions are akin to perceptual judgements about bleeding: they are subject to error through misidentification. As I pointed out earlier, it is conceivable that while it looks to you as if you were controlling the movements shown on the monitor, they are in fact controlled by someone else. Supposing that we still think of such judgements as, at least indirectly, expressive of self-awareness, we get a picture on which they are inferred from judgements that are directly expressive of self-awareness, for example judgements about one's intentions, acts of the will, or tryings.6 The problem with this argument is that it assumes, I think mistakenly, that the error in question is one of misidentification. If you commit an error through misidentification, you must be right about something. In the earlier example, you correctly believe that 'this (perceptually presented) hand is bleeding'. It is just that you get wrong whose hand is in question. The assumption, then, is that when your experience misrepresents a movement as being under your control, at least you are right to think 'This person is raising her arm' or 'This arm is being raised by someone'. But this misrepresents the situation. In so far as your experience is used to guide your activity, it gives you no reason for either of the two judgements. To have such a reason you would need to observe the movement, with a view to acquiring knowledge of facts over which you have no control, whereas in attending 'from within the act' you precisely take yourself to control the facts you perceive. Of course when the movement consistently disappoints your perceptual expectations, you will switch to an observational mode of attending, and then you may reasonably make either or both of the two judgements. But at that point your experience will no longer constitute a basis for a self-ascription of the action of controlling the movement. Then what makes the experience of intended effects a first-personal experience? It cannot be first-personal in virtue of the self being amongst the experienced, perceptually presented objects. The material we are working on, and attending to, may be remote from our bodily movements. And even when an action is guided by attending to the body, as in the example of the video screen, the body is not presented to us as the agent of the action. Rather, what makes the experience firstpersonal is that the first person figures in the representation controlling the movement. If you experience an effect as controlled by the content of your intention, you have an experience of yourself controlling it. The first person does not seem to be dispensable here, at least so long as we are talking about actions explained in terms of the subject's practical reasons. If perception is to contribute to the rational control of an action, the agent must be able to use it in answering the 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 awareness 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 ANSCOMBE, G. E. M. (1957), Intention. Oxford: Blackwell. AYERS, M. (1991), Locke, i. London: Routledge. BAHRICK, 1. E., and WATSON, J. S. (1985), 'Detection of intermodal proprioceptive-visual contingency as a potentional basis for self-perception in infancy', Developmental Psychology, 21: 963-73. BERMUDEZ, J. 1. (1995), 'Ecological perception and the notion of a nonconceptual point of view', in J. Bermudez, T. Marcel, and N. Eilan (eds.), The Body and the Self. Cambridge, Mass.: MIT Press. BRATMAN, M. (1987), Intention, Plans, and Practical Reason. Cambridge, Mass.: Harvard University Press. BREWER, B. (1999), Perception and Reason. Oxford: Oxford University Press. BRIDGEMAN, B., KIRCH, M., and SPERLING, A. (1981), 'Segregation of cognitive and motor aspects of visual functioning using induced motion', Perception and Psychophysics, 29: 336--42. CAMPBELL, J. (1993), 'The role of physical objects in spatial thinking', in N. Eilan, R. McCarthy, and B. Brewer (eds.), Spatial Representation. Oxford: Blackwell. CASSAM, Q. (1995), 'Introspection and bodily self-ascription', in J. Bermudez, T. Marcel, and N. Eilan (eds.), The Body and the Self Cambridge, Mass.: MIT Press. DAVIS, 1. (1979), Theory ofAction. Englewood Cliffs, NJ: Prentice-Hall. DONNELLAN, K. (1963), 'Knowing what I am doing', Journal ofPhilosophy, 60: 401-9. EILAN, N. (1998), 'Perceptual intentionality, attention and consciousness', in A. O'Hear (ed.), Contemporary Issues in the Philosophy ofMind. Cambridge: Cambridge University Press. EVANS, C. O. (1970), The Subject ofConsciousness. London: George Allen & Unwin. EVANS, G. (1982), The Varieties ofReference. Oxford: Oxford University Press. FRITH, C. (1992), The Cognitive Neuropsychology of Schizophrenia. Hove: Erlbaum. GIBSON, E. (1993), 'Ontogenesis of the perceived self', in U. Neisser (ed.), The Perceived Self Cambridge: Cambridge University Press. GIBSON, J. J. (1979), The Ecological Approach to Visual Perception. Boston: Houghton Mifflin. JAMES, W. (1890a), The Principles ofPsychology, i. London: Macmillan. - - (1890 b), The Principles ofPsychology, ii. London: Macmillan. JEANNEROD, M. (1997), The Cognitive Neuroscience ofAction. Oxford: Blackwell.
Intentional Action and Self-Awareness
405
KAGAN, J. (1981), The Second Year: The Emergence of Self-Awareness. Cambridge, Mass.: Harvard University Press. MAUGHAM, W. S. (1998), OfHuman Bondage. London: Random House. MILNER, A. D., and GOODALE, M. A. (1995), The Visual Brain in Action. Oxford: Oxford University Press. NEISSER, U. (1988), 'Five kinds of self-knowledge', Philosophical Psychology, 1: 35-59. --(1993), 'The self perceived', in U. Neisser (ed.), The Perceived Self. Cambridge: Cambridge University Press. NEUMANN, O. (1984), 'Automatic processing: a review of recent findings and a plea for an old theory', in W. Prinz and A. F. Sanders (eds.), Cognition and Motor Process. Berlin: Springer. --(1990), 'Visual attention and action', in O. Neumann and W. Prinz (eds.), Relationships between Perception and Action. Berlin: Springer. NORMAN, D., and SHALLICE, T. (1986), 'Attention to action', in R. Davidson, G. Schwartz, and D. Shapiro (eds.), Consciousness and Self-Regulation. New York: Plenum. O'SHAUGHNESSY, B. (1980), The Will, ii. Cambridge: Cambridge University Press. - - (1992), 'The diversity and unity of action and perception', in T. Crane (ed.), The Contents ofExperience. Cambridge: Cambridge University Press. PIAGET, J. (1953), The Origin of Intelligence in the Child. London: Routledge and Kegan Paul. ROESSLER, J. (2002), 'Action) emotion, and the development of self-awareness', European Review ofPhilosophy, 5: 33-52. STIPEK, D. (1995), 'The development of pride and shame in toddlers', in J. P. Tangney and K. W. Fisher (eds.) Self-Conscious Emotions. New York and London: Guilford Press, 237-52. STRAWSON, P. F. (1974), 'Imagination and perception', in his Freedom and Resentment and Other Essays. London: l\1ethuen. --(1988), 'Perception and its objects', in J. Dancy (ed.), Perceptual Knowledge. Oxford: Oxford University Press.
Author Index Aaronson, E. 202 Abramson, L. Y. 89 Ach, N. 174 Agliota, S. 158 Alessandri, S. M. 288, 290 Allport, A. 153 Anderson, J. R. 167n. Anscombe, G. E. M. 15-19, 20, 25, 28,45, 297,333,338, 358n., 364n., 383-384, 396-397 Arbib, M. A. 129 Armstrong, D. M. 236, 296, 298 n. Asendorpf, J. B. 284 Austin, J. L. ISO, 162 Ayers, M. 401 Babelot, G. 232 Baddeley, A. 221,223,235 Bahrick, L. E. 388 Baldwin, T. 21, 32 Baldwin, J. M. 254 Bargh, J. A. 167n.,222 Barnard, P. J. 58 Baron-Cohen, S. 302, 303 n. Barry, C. 222 n. Baudonniere, P. M. 284 Bauer, P. J. 232 Bauer, R. M. 202 Baumeister, R. F. 112 Bender, D. B. 285 Benedetto, E. 252 Bermudez, J. 298n., 317n., 387n. Bertenthal, B. 1. 284, 300 Bettelheim, B. 89 Bianchi, L. 234 Biederman, 1. 169, 208 Bischof-Kohler, D. 282, 287 Bisiach, E. 68, 72 Blakemore, S.-J. 112, 314 Block, N. 228, 231 BonJour, L. 106n. Borges, J. L. 56,90
Brand, M. 218 Brass, M. 179, 180 Bratman, M. 394 n. Bretherton, 1. 301 n. Brewer, B. 28n., 69, 323,327, 329,401 Bridgeman,B. 63,133,157,158,395 Brion, S. 206 Broadbent, D. E. 203, 205 Broadie, F. 358 n., 365 n. Brooks-Gunn, J. 276,281,282,284,285 Bryant, P. E. 257 Burack, J. 252 Burge, T. 107n., 236, 310n., 338, 372n. Burgess, C. A. 222 Butterworth, B. 60 Butterworth, G. 280, 281 Campbell, J. 29-30,31,32,33,34,42-43, 229n., 230, 315n., 339,340,386 Carey, D. P. 311 n. Carlson, S. M. 246, 252 Carruthers, P. 228, 229, 231-232, 233 Cassam, Q. 195, 358n., 403n. Castaneda, H.-N. 380n. Castiello, U. 133 Chartrand, T. L. 222 Chelazzi, L. 210 Cohen, L. B. 254 Cole, J. 26n., 71, 73 Collingwood, R. G. 193 Collins, L. M. 283 Coltheart, M. 201 Cooper, E. E. 208 Crane, T. 298 n. Csikszentmihalyi, M. 82, 90 Danto, A. C. 358 n. Daprati, E. 142, 144, 145 Darwin, C. 279, 282, 285 Das Gupta, P. 257 Dasser, V. 142 David, M. 108 n.
408
Author Index
Davidson, D. 218 Davies, M. 12, 264n. Davis, L. 394n. Day, B. L. 122, 123 Decety, J. 139, 140, 314 Dell, M. 108 n. Della Sala, S. 59, 76, 77, 79, 97, 206, 207, 222 n., 235, 340 DeLoache, J. 253 Dempster, F. N. 246 Dennett, D. C. 11, 56, 67, 104, 234, 265n., 272 Descartes, R. 165 Desmurget, M. 129 Deutsch, D. 205 Deutsch, J. A. 205 Dewey, J. 188, 253 Diamond, A. 246 Dickinson, A. 221, 227 Dienes, Z. 69, 99, 101,222,223, 225, 229, 233,338 Dierks, T. 143 Dokic, J. 20, 34, 35, 39 Donders, F. C. 166 Donnellan, K. 358n., 366, 367n., 384 Dretske, F. 232, 237 Dreyfus, H. 219, 238 Dunn, R. 358n. Duval, S. 276, 282 Dweck, C. S. 288 Eddy, T. J. 284 Edelman, G. 253 Edwards, M. G. 215 Eilan, N. 52,194,228,390 Eimer, M. 74, lIS, 117, 118 Eklund, G. 62 Elman, J. 253 Engelkamp, J. 137 Evans, C. O. 390 Evans, G. 35,36,53,80-81,109, 298n., 315n., 316, 317n., 330, 341 n., 374n. Fallienot,1. 132 Falvey, K. 338 Fein, G. G. 283
Feinberg, 1. 305 n., 315 n. Fenson, L. 232, 301 n. Fischer, K. W. 284 Fitts, P. M. 169 Flash, T. 129 Flavell, J. H. 232 Fourneret, P. 69, 113, 125, 134-135, 313 Frak, V. G. 138 Freese, M. 167n. Freud, S. 77,276 Fried, 1. 84 Frith, C. D. 38, 55, 56, 139, 140, 145, 205, 221,227, 315n., 394 Frye, D. 5,13-14,33,263, 264n., 265, 267-268,270-271 Gallagher, S. 51, 89 Gallese, V. 140 Gallup, Jr., G. G. 284 Gandivia, S. C. 314 n. Gardner, J. M. 232 Gazzaniga, M. S. 276, 280 Geminiani, G. 68,72 Georgieff, N. 146 Gergely, G. 301 n. Gibson, E. J. 178 n., 300, 387 Gibson,]. J. 32,52, 53,160, 178n., 384 Ginet, C. 365 n. Goldberg, G. 123 Goldman, A. 140 Goldstein, K. 206 Golinkoff, R. M. 301 n. Gollwitzer, P. M. 11, 167n. Gollwitzer, P. W. 237 Goodale, M. A. 29, 30, 31, 131, 133;154, 155,156,158,160,395 Goodwin, G. M. 62 Gopnik, A. 33, 276, 281 Gordon, R. M. 230, 341 n. Grafton, S. T. 140 Greenwald, A. 167n., 174 Grodd, W. 282 Gross, C. J. 285 Haake, R. J. 232 Hagbarth, K. E. 62
Author Index Haggard, P. 35,37-38,40,41,42,44, 71, 72,73,74,82,87 Hall, D. A. 201 Harding, C. G. 301 n. Harter, S. 282 Haviland, J. M. 300 Heal, J. 264 n. Hecaen, H. 73,79, 108n. Heider, F. 59 Helmholtz, H. 166 Helwig, C. C. 258 Herrigle, E. 82, 90 Hershberger, W. A. 167n. Heyes, C. 221, 227 Higgins, E. T. 283 Hintikkaa, J. 17 n. Hitch, G. J. 235 Hobson, R. P. 281,282 Hochschild, A. R. 286 Hoerl, C. 89,230 Hoff, B. 129 Hogan, N. 129 Hommel, B. 168, 170,171,176, 178n., 180 Hornsby, J. 5,14, IS, 109, 337n. Hull, C. L. 166, 167n., 174 Hume,D.38 Humphrey, N. 55,56,83 Humphreys, G. W. 5,7,8,10-11,12,13, 24,31,32,79,252,392 Hurley, S. 103 n. Hutchinson, W. D. 324n. Huttenlocher, J. 283 Izard, C. E. 277 Jacoby, L. L. 228, 231 Jacques, S. 247 Jahanshahi, M. 221, 227 Jakobson, L. S. 133 James, M. 132 James, W. 4, 7, 10, 31, 51, 69, 165, 166, 167,182,286,289,306,311, 312, 314n., 388-389,393,394 Jeannerod, M. 22,23-24,25,31,69,113, 125, 154, ISS, 157, 158, 159, 167 n., 220,
409
222,223,227,236,313, 314n., 315n., 332 n., 334, 335, 340 n., 393 n., 395 n. Jedynak, C. P. 206 Johansson, G. 142, 300 Jones, L. 62 Kelso, J. A. S. 85 Kieras, D. E. 167 n. Kinsbourne, M. 326, 329n. Kirsch, 1. 83, 221 Knuf, L. 183 Korsgaard, C. 21 Kumada, T. 213 LaBerge, S. 231 n. Lackner, J. R. 62 Lambie, J. A. 57 Lang, B. 252 Lau, A. 258 Lavie, N. 205, 213 LeDoux, J. 279, 286 Lee, D. N. 202 Leggett, E. L. 288 Lelwica, M. 300 Leslie, A. M. 282, 283, 284, 302 Levine, D. N. 72 Lewis, M. 33, 44, 386 Lhermitte, F. 77, 136, 142, 201, 301n. Liberman, A. M. 172 n. Libet, B. 1,40-41,42-43,44,72,74, 113-117,120,125 Locke, J. 188 Logothetis, N. K. 125 Lorenz, K. 302 Lotze, R. H. 165, 166, 167 Lowe, M. 283 Luria, A. R. 234, 247 Lynn, S. J. 83 Lynn, S. R. 221 MacDonald, C. 358n. Mack,A. 205 Magno,E. 71,74,122,123 Mandler, J. M. 300
410
Author Index
Marcel, A. J. 22,23,24,25,26,27,31,36, 38,39,40-41,43-46,94-95,97-100, 101,102,103, lOS, 106, 108,340,395 Marchetti, C. 77 Marcos, H. 232 Martin, M. G. F. 299 n., 325, 326 Masangkay, Z. S. 232 Massaro, D. W. 171 Maugham, W. S. 387 May, J. 58 Mayer, C. 60 McClintic, S. 287 McCloskey, D. 1. 120 McCormack, T. 89, 230 McCune, L. 283 McDonough, L. 232 McDowell, J. IOn., 298n., 328 McFarland, D. 227 McGinn, C. 366n. McLeod,~D. 69,99,101 Metzler, J. 139 Meltzoff, A. N. 33,232,255,276,281, 284,300 Meringer, R. 60 Merleau-Ponty, M. 52, 323n., 325, 326 Meyer, D. E. 167n. Michalson, L. 280, 282, 285, 287 Mill, J. S. 116 Miller, A. 290 Miller, G. A. 167 n. Millikan, R. G. 237 Milner, A. D. 29,30,31,154, 155,156, 160,395 Milner, B. 234 Mishkin, M. 131, 156 Mittelstaedt, H. 37,57,83 Monsell, S. 2, 4 Moore, M. K. 300 Moran, R. 16n.,45 Morton, J. 120 Moses, L. J. 246, 252 Mounoud, P. 285 Munakata, Y. 251 Miisseler, J. 178 n.
Nagel, T. 231 Neisser, U. 29, 32, 385-386, 387, 388, 400,404 Neumann, O. 390 Newell, R. 167n. Nicolich, L. 283 Nielsen, T. 142, 222, 236 Norman, D. A. 11, 12,204,218,221,223, 224, 234, 332 n., 339 n., 389 O'Brien, L. 19,23,27,39,45 O'Shaughnessy, B. 9, 18-19,20,21,39,40, 41,45,46,53,87,90,219,321,326, 335n., 341, 358n., 362n., 365n., 397, 398,399 Olsen, C. 358n.,365n. Pacherie, E. 136,227, 237,334, 335 Paillard, J. 26n., 69,71 Paivio, A. 137 Parkin, A. J. 222 n. Parsons, L. M. 138, 139 Paulignan, Y. 129, 130 Peacocke,C. 13,19,23,25,28,36,40-41, 44,66, ISO, 298n., 330, 371 n., 372n., 375n.,376 Perani, D. 139 Perenin, M. T. 156 Perner, J. 5, 7, 8,9-12,24,45,252, 335n., 338 Perry, J. ISO, 330 Peterson, S. E. 210 PiagetJ. 232,254,282,283,284,386 Polanyi, M. 69 Posner, M. I. 210 Postman, L. 178 n. Povinelli, D. J. 284 Premack, D. 300 Pribram, K. H. 167 n., 311 n. Prinz, W. 7,26-28,69 Prior, A. N. 307 Proust, J. 18-19,36-37,38-39, 340n. Putnam, H. 199n.,366n. Rabbitt, P. M. A. 61 Rafal, R. D. 202
Author Index Ramey, C. T. 81 Ramsay, D. S. 281,282,283,284 Reason, J. T. 4,60,77,78,203,220 Recchia, S. 287 Reed, N. 69,99, 101 Rees, G. 205 Requin, J. 136 Reznick, J. S. 255 Richardson, J. 119 Riddoch, M. J. 5,7,8-9, 10-11, 12, 13,24, 31,32,79,252,392 Rizzolatti, G. 140, 332n. Robertson, I. H. 215 Rocha-Miranda, C. E. 285 Rochat, P. 281 Rock, I. 205 Roessler, J. 339,358n. Roland, P. E. 311n., 314n. Roll, J.-P. 299n. Rosenbaum, D. A. 129,166n. Rosenthal, D. M. 228,231-232,233 Rossetti, Y. 156 Rudell, A. P. 169 Ruffman, T. 230 Russell, B. 195 Russell, J. 53, 82-83, 88, 335 Ryle, G. 296 Sabini, J. 167n. Sanders, A. F. 166, 171 Schneider, K. 143 Schneider, W. 221 Schubo, A. 178 n. Searle, J. R. 18,118,126,218,219-220, 236, 238, 248, 254, 255, 303, 305 n., 307 Shallice, T. 11,12,89,204,218,221,223, 224,234, 332n., 339n., 389 Shatz, M. 232 Shepard, R. N. 139 Shiffrin, R. M. 221 Shoemaker, S. 35,80-81, 104, 161, 162, 298n., 315n., 316, 330n., 341, 358n., 365,372 Sigel, I. 253
Simmel, M. 59 Simon, J. R. 169,170 Sirigu, A. 138 Slachevsky, A. 135-136, 313 n. Somerville, S. S. 232 Spence,S.A. 139,143 Sperry, R. W. 37 Spinnler, H. 77 Staudt, M. 282 Stein, B. E. 314n. Stelmach, G. E. 129, 130 Stephan, K. M. 139 Sternberg, S. 120,121,166 Stich, S. P. 268 n. Stipek, D. 287,288,387 Stoet, G. 178n. Strawson, P. F. 195,401 Strommen, E. A. 244 Sullivan, M. W. 288 Talairach, J. 108 n. Taublieb, A. B. 62 Taylor, C. 246 Tegner, R. 68, 72 Treisman, A. 205 Tulving, E. 280 Turvey, M. T. 85 Tye, M. 232 Ungerleider, L. G. 131,156 Vallar, G. 202 Velleman, J. D. 358n., 373-374 Vinter, A. 285 Viviani, P. 172 n. von Bartalanffy, L. 280 von Holst, E. 37,57,83, 133 Wakefield, J. 219,238 Wallace, R. A. 170 Warrington, E. K. 132 Watson, J. S. 81,276,280,388 Watson, R. I. 166 Wegner, D. M. 222, 236 Weiner, B. 288
411
412
Author Index
Weiskrantz, L. 202, 280 Welford, A. T. 171 Wellman, H. 252 Wells, H. G. 104 Welsh, M. C. 245 Wheatley, T. 222, 236 White, S. H. 246 Wicklund, R. A. 276, 282 Willatts, P. 232 Wimmer, H. 232, 252 Wittgenstein, L. 21, 188, 190-192, 193,200 Wolpert, D. M. 119,134,223 Wright, C. 373n.
Wugalter, S. E. 283 Wundt, W. 306 Wylie, R. C. 275 Yerys, B. E. 251 Zahavi, A. 302 Zahn-Warle~C. 282 Zajonc, R. B. 286 Zeki, S. 205 Zelazo) P. D. 5,13-14,33,223,263, 264n., 267-268,270-271,272 Zelazo, P. R. 253
Subject Index absent-mindedness, see failures of control action automatic 8-10, 227, 391-396 basic 23-24, 107-108, 346, 363, 379-380 dual-aspect theory of 345 explanation 151,218-219,391 g~al-directed 8, 59-60, 73, 77 immersed 51, 61, 68, 82, induction 168-171,176-179,182-184 intentional 5-7, 32-34, 55, 112, 154, 227,249,255,296-317 passim mental 98,355-356 non-intended 6, 203, 214-215, 219-221, 227,365 object-directed 29-35, 128-133, 150-164 passim, 209-211 phenomenology of 25,31,44,81-87, 99-102,104-105,352 slips of, see failures of control stimulus-driven 77-78, 203, 394 sub-intentional 9 unawareness of 23-29,61-68,101-102, 134-136,214,361,395-396 see also agency, attention, awareness, control, trying, pre-motor vs post-motor awareness, ideomotor affordances 31-34,160,305,384-388, 392-393 agency sense of 17,44,50-56,161-163 (see also awareness of action) perception of-in others and the self 34-46,197,365-366 Alien Hand 76, 206 Anarchic Hand Syndrome 1-15 passim, 34,36-37,76-81,85-87,94-97,102, 109-110,162,206-216 Anosognosia 72-75, 85 Ataxia, optic 156 Attention 19-21, 31-34, 66-69, 87, 205-206,376-378,388-391 and control of action 391-396, 378-380
perceptual 8-9, 26-27, 153-154, 159-161,194-196,389,394 and self-knowledge 396-404,378-380 authorship: see ownership of action automaticity 24,31-32, 128-136 (see also automatic action) awareness of action 13-46, 56-60, 95, 98-110, 119_124,136-138,297-303,331-333, 349_356,358-382,384-388,396-404 of intention 42,45, 55, 60-61, 72-73, 90, 112_119,136-138,303-313,369-371 of movement, see movements, (un) awareness of of trying 103-105, 107-108, 334, 353-356,369-371 introspective 16_19,228-233,296,316, 334,338-342,353-356 Binding thesis 160-161 Blindsight 29,31, 67-68, 151-154, 160, 229,280
card-sorting test (Frye and Zelazo) 223-225,245-253,267-269 causation causal knowledge 38,88-90,256-258 causal self-reference 236-238 commonsense psychology 9-15,23,42, 263-273,388-391 common coding 26-28, 171-179 comparator mechanisms 24, 55, 83, 133-134,197,309,314, consciousness 31-34, 136-141, 221, 376-378,394 access and phenomenal 228-233 higher-order thought theory of 11, 44, 228-233 Levels of Consciousness Theory (Frye and Zelazo) 253-256,271-273 perceptual 21-34,42, 151-165 passim, 194-196,389
414
Subject Index
control Cognitive Complexity and Control Account (Frye and Zelazo) 248-253, 270-271 development of 244-260 passim, 291 dual, see two level theory of failures of 3-5, 13-15,60, 72-73, 77, 202-216 passim, 220-221, 227, 245-256,311-313,364,393-394 two level theory of 7-12, 32, 219-239 passim, 394 sense of 20, 39, 44-46, 333-336, 378-382,387-388,400-404 intentional 22, 202-209,228, 372-375 rational 24-29, 397-404 'cups experiment' (Humphreys and Riddoch) 5,12-13,207-216,252,392 deviant causal chains 218, 238-239 DCCS, see card-sorting test dissociations between conscious perception and action, see sagittal line experiment, vibrotactile illusion experiments, two visual systems hypothesis Ebbinghaus illusion 157 efferent binding 37-41, 111-113 effort, sense of 38,73-74,82,306,310,314 egocentric frames of reference 51, 58, 94-95,97,84-86,129,132,194,197, epiphenomenalism 29-31 executive function/control (see also supervisory attentional system) 221, 225-226,234-236,245-248,265-267
immersion, see action, immersed; self-awareness, immersed vs detached immunity to error through misidentification 35-37, 80-81, 109-110,161-163,330,402-403 intention 244-245,297,356-357,362, 392-393 (un)awareness of, see awareness prior 8-11,108-109,236-239, 303-304,309 in action 236-239, 303-304, 308 chronometric approach to 113-119 IW (deafferented patient) 26, 71-72 Kinaesthesis 55, 68, 350-351 knowledge 'from within' 16, 20, 36, 42, 94, 107, 161-163,338,350,359-366 non-observational 15-21, 52-54, 90, 297-303,333,383-384,397 perceptual 28-29, 190-192, 200, 297-303,398-400 practical 17, 88-90, 383-384, 396-397 procedural vs declarative 69 Moore's paradox 188-190 motor control 133-134,174,-176 cortex 123 imagery 137-139,393 learning 112 specifications 71,74-75,85,100-102, 105,220-221 movements 114, 336-337 (un)awareness of 23-29, 101-102, 120-124
flow experience 82 neglect 211 goal representations 8, 96, 132-133, 137, 166-167,174-176,220-221,250-256, 289,335 ideomotor actions 7, 182-184, 389-390, 394 theory of action 7, 165-167, imitation 75-76, 182
Obsessive-Compulsive disorder 79-80 ownership of actions 34-46, 55-58, 75-87, 97-98, 102-105,236 of bodily sensations 322-331 of the body 328-331 sense of-of actions 34-46,75-87,
Subject Index 97-98,102-105,141-145,315 (see also awareness of action, self-awareness) sense of-of the body 325-328
perception and the will 190-200 and control of action 29-34, 128-130, 152-161,165-179,391-396 and awareness of action 161-163, 296-317,367,396-404 see also consciousness, perceptual perceptual demonstratives 30, 104, 150-163 passim, 300 perspectivalness 51, 58,84-87, 97, 194, 197 pre-motor vs post-motor awareness 40-43,71,74,121-124 proprioception 20, 23, 36, 52-54, 58, 67-68,71-72,81,109-110,133, 322-328,348-351
415
self-awareness 80-81, 103-105, 143-145, 195,275-291 passim, 315-317, 375-382,402-404 ecological 384-385 and emotions 285-287, 291 immersed vs detached 43-46, 56, 90 self-ascriptions of action 141-145, 163, 374-375,375-382 self-ascriptions of belief 188-190, 366, 371,373-374 see also immunity to error through misidentification sensorimotor theory of action 165-167 simulation, mental 139-140 slips of action, see failures of control supervisory attentional system 11, 204, 221-222,234-236 supplementary motor area 5, 71, 74, 94 theory of mind 252, 263-267, 284-285 trying 19,25-26,40-42,45, 96, 99-105, 108-109, 345-57 passim, 362 (see also awareness of trying) two visual systems hypothesis (Milner and Goodale) 30,131,154-161, 395-396
readiness potential 40,74,114-119 reasoning, practical 16,189-190,198-200 reasons practical 45-46, 96, 391-396 epistemic 101, 106-108, 375-382 reflexes 59,62,72 representations, pragmatic vs semantic 30, 131 procedural vs declarative 222-223 Roeloff effect 63, 157
vibrotactile illusion experiments (Marcel) 22,25-26,62-67,85,94, 99-102,395
sagittal line experiment (Jeannerod) 22, 69, 134-136 schizophrenia 36-37, 39, 79-80, 306 delusions of alien control 36-37,39, 143-145,162-163,315-317
will 222,227,350,355-357 and depression 88-89 freedom of 1,42-43, 113-115 W judgement (Libet) 114-119 willing, see trying
utilization behaviour 4, 12,76-78,201
