Thought and Language ROYAL INSTITUTE OF PHILOSOPHY SUPPLEMENT: 42
EDITED BY
John Preston
CAMBRIDGE UNIVERSITY PRESS ...
113 downloads
1063 Views
3MB Size
Report
This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!
Report copyright / DMCA form
Thought and Language ROYAL INSTITUTE OF PHILOSOPHY SUPPLEMENT: 42
EDITED BY
John Preston
CAMBRIDGE UNIVERSITY PRESS
PUBLISHED BY THE PRESS SYNDICATE OF THE UNIVERSITY OF CAMBRIDGE
The Pitt Building, Trumpington Street, Cambridge, CB2 1RP, United Kingdom CAMBRIDGE UNIVERSITY PRESS
The Edinburgh Building, Cambridge CB2 2RU, United Kingdom 40 West 20th Street, New York, NY 10011-4211, USA 10 Stamford Road, Oakleigh, Melbourne 3166, Australia © The Royal Institute of Philosophy and the contributors 1997 Printed in the United Kingdom at the University Press, Cambridge Typeset by Michael Heath Ltd, Reigate, Surrey A catalogue record for this book is available from the British Library
Library of Congress Cataloguing in Publication Data
Thought and Language/edited by John Preston p. cm.—(Royal Institute of Philosophy supplement: 42) ISSN 1358-2461 (42) Papers presented to the annual conference of the Royal Institute of Philosophy, University of Reading, Sept. 1996. Includes bibliographical reference and index. ISBN 0-521-58741 7 (pbk.) 1. Language and languages—Philosophy—Congresses. 2. Thought and thinking—Congresses. I. Preston. John, 1957- . II. Series. P106.T495 1997 401—DC21 97-42270 CIP r97 ISBN 0-521-58741-7 paperback ISSN 1358-2461
Contents Preface
v
Notes on Contributors
vi
Introduction: Thought as Language
1
JOHN PRESTON
Seeing through Language
15
DONALD DAVIDSON
'The only sure sign...': Thought and Language in Descartes
29
JOHN COTTINGHAM
Words and Pictures
51
JOHN HYMAN
Social Externalism and Conceptual Diversity
77
ANDREW WOODFIELD
The Explanation of Cognition
103
JOHN R. SEARLE
Thought Without Language: Thought Without Awareness?
127
L. WEISKRANTZ
Philosophy, Thought and Language
151
HANS-JOHANN GLOCK
The Flowering of Thought in Language
171
W. V. QUINE
Talking to Cats, Rats and Bats
177
K. V. WILKES
Analyticity, Linguistic Rules and Epistemic Evaluation
197
CHRISTOPHER HOOKWAY
111
Contents How to Do Other Things With Words
219
DANIEL C. DENNETT
Bibliography: Twentieth-Century Philosophical Texts on Thought and Language
237
Index
245
IV
Preface The papers in this volume, except for the introduction, were all presented, in the order in which they are published here, to the annual conference of the Royal Institute of Philosophy, held at the University of Reading in September 1996. The conference organiser and the Royal Institute of Philosophy would like to thank the Humanities Research Board of the British Academy for a British Conference Grant which went towards the travelling expenses of our U.S. speakers. I should also like to thank Yvonne Stanford, for her invaluable secretarial help in preparing for the conference. John Preston, (conference organiser), Department of Philosophy, The University of Reading
Notes on Contributors John (Nottingham is Professor of Philosophy at the University of Reading Donald Davidson is Professor of Philosophy at the University of California, Berkeley Daniel C. Dennett is the Director of the Center for Cognitive Studies at Tufts University Hans-Johann Glock is Reader in Philosophy at the University of Reading Christopher Hookway is Professor of Philosophy at the University of Sheffield John Hyman is Reader in Philosophy at the Queen's College, Oxford University John Preston is Lecturer in Philosophy at the University of Reading W. V. Quine is Emeritus Professor of Philosophy at Harvard University John R. Searle is Professor of Philosophy at the University of California, Berkeley L. Weiskrantz is Emeritus Professor of Psychology at the University of Oxford K. V. Wilkes is Senior Tutor in Philosophy at St. Hilda's College, Oxford University Andrew Woodfield is Reader in Philosophy at Bristol University
VI
Introduction: Thought as Language JOHN PRESTON Western philosophy has a long-standing interest in the relationship between thought and language. This is not least because languageuse and our mental capacities are so central to our human self-conception, as well as to the ways in which we have tried to think about other beings. Retrospectively, it is possible to identify certain broad traditions in the philosophical study of thought and language, traditions which also have their representatives in psychology and linguistics. In this introduction I shall focus on one such tradition, the one sometimes known as 'lingualism', in so far as it bears on the papers brought together in this volume.1 In the Theaetetus, Plato has Socrates answer the question 'What do you mean by "thinking"?' by characterising thought as 'A talk which the soul has with itself about the objects under its consideration' (189e). On such a conception, there is a logical or 'internal' connection between thought and language: thought just is the discourse of the mind with itself (see also Plato's Sophist, 263e). While this is not the only Platonic account, it is perhaps the one which has borne the most fruit. Most accept that there is some kind of intimate (even necessary) connection between thought and language. But is it, as the lingualist supposes, that thought must always take place 'in' language? Construing thought as inner language certainly promises to explain some very important things, notably, several much-vaunted parallels between thought and language. Thoughts and utterances exhibit semantic parallels: not only are both candidates for meaning, reference, and truth-or-falsity, but the thought that p and the indicative statement 'p', produced in otherwise identical situations, must have the same meaning, reference and truth-value.2 Against mystics who suppose that there can be ineffable thoughts, lingualists insist that whatever we are capable of thinking, we are in principle capable of saying, and vice versa. Furthermore, there are alleged syntactic parallels: the number of thoughts we can have (if this notion makes sense), their syntactic complexity, and the systematic relationships between those thoughts which are possible for a given thinker, are all mirrored in similar features of things one can say. ' I should perhaps stress that this is only one possible way to divide up the territory. 2 Aside from exotic exceptions like that in which 'I am now stating that so-and-so' takes the place of 'p', of course.
John Preston Lingualists, as we shall see, also make much of the fact that their account appears to be the only one which can make sense of contemporary cognitive-psychological theorising. Theories from cognitive psychology represent thought largely as a matter of computational operations on mental symbols at a 'sub-personal' level. Computer science shows us how computational manipulations might consist in the processing of linguistic tokens. To suppose, as the lingualist does, that thought is language, would therefore seem to solve problems and offer rewards at a stroke. Unsurprisingly, Plato was by no means the only philosopher prior to the 'cognitive turn' to embrace a lingualist approach. Rene Descartes, who did much to re-orient the course of philosophy in the seventeenth century, did not straightforwardly identify thought with language. But John Cottingham's paper in this volume details the ways in which Descartes' conception of thought follows Plato's, as. well as the ways in which it anticipates the twentieth-century computational conception. Cartesian 'ideas' and thoughts are publicly accessible, since they are linked with the ability to use language correctly. Descartes' dream of constructing an easily learnable formal language which would allow the ordered expression of all possible human thoughts was later pursued by Leibniz (1677), whose great project, the ars combinatoria, was to construct a perfect language in which thoughts would receive their most perfect (efficient and transparent) expression. For Plato and Descartes, what really does the thinking is the rational (part of the) soul, rather than the whole human being. But this does not mean that 'immaterialism' is the most important commonality here: in fact, it seems to be of tangential relevance to lingualist views. Cottingham, for example, suggests that Descartes argued himself into immaterialism in a way which might not have persuaded him had he known what we now know about the brain. Contemporary lingualists certainly take pains to make their views compatible with moderate forms of naturalism, the view that psychological phenomena can be explained in ways acceptable within the natural sciences, and materialism, the view that the phenomena adverted to in such explanations must be physical. But there still remains the question of whether and how the brain can be credited with the semantic capacities these lingualists suppose it to have. Theories of Judgement At the beginning of the twentieth century, the relationship between thought and language was debated in a new context. In his seminal
Introduction: Thought and Language essay 'Uber Sinn und Bedeutung' (1892) and his 1894 review of Edmund HusserPs Philosophie der Arithmetik, the German logician Gottlob Frege insisted upon distinguishing between the 'objective content' and the 'subjective performance' of thinking. The former, that which can be thought, is capable of existing without a thinker, and of being the common property of several thinkers. The latter, the psychological episode in which the former is 'apprehended' or grasped, must have a bearer. The objective content of thinking Frege called 'the thought' {der Gedanke), and he identified this with the 'sense' of a sentence, that which is capable of being true or false. Always resisting the attempt to blur the boundary between psychology and logic, Frege's opposition to 'psychologism' came, in his later essay 'The Thought: A Logical Inquiry' (1918), to re-instate Platonism. There he expressed the relation between thought and language using a metaphor that has come to be familiar: The thought, in itself immaterial, clothes itself in the material garment of the sentence and thereby becomes comprehensible to us. We say a sentence expresses a thought. (Frege 1918, p. 20; see also p. 26n) From the supposition that the thought expressed by a sentence is both immaterial and non-psychological, together with his conviction that although psychological episodes need bearers, thoughts do not, Frege drew the conclusion that over and above the things of the 'outer world' (material objects) and those of the 'inner world' (psychological phenomena) we must recognise a 'third realm', whose contents cannot be grasped by the mind until they are dressed in language. On this conception, although 'thoughts' can and do exist independently of our grasping them, thinking consists in grasping them with a special mental capacity, and judging (that is, thinking that so-and-so is the case) consists in taking the thoughts thus grasped to be true. The view is lingualist not because it represents thoughts as linguistic (it need not), but because it construes thinking as coming to stand in a relation to 'objects of thought', these objects being the 'senses' of sentences, those things which are true or false. This period in the history of our subject has been subjected to the closest of scrutiny by Michael Dummett and others. From it, Dummett has recently drawn conclusions about the nature and trajectory of philosophy itself. In this volume, the period in question and Dummett's interpretation of it are treated in the essay by HansJohann Glock. Glock rejects Dummett's claim that the difference between 'analytical' philosophy and philosophies from continental Europe influenced by phenomenology can be traced to a contrast
John Preston between the philosophy of language and the 'philosophy of thought'. He traces the idea that the basic task of philosophy is to analyse thought back to its nineteenth-century roots, showing how the Kantian concern with representation was transformed by the 'linguistic turn' of the early twentieth century, instigated primarily by Wittgenstein. Bertrand Russell, although he does not sit comfortably in the lingualist tradition, codified an associated perspective on psychological phenomena in a way which has influenced us down to the present day. To think, assume, believe, know, expect, remember, desire, hope and fear that so-and-so, Russell argued, are each matters of standing in different psychological relations to propositions. To believe that p is to believe the proposition that p. But although, according to Russell, 'it seems natural to say one believes a proposition and unnatural to say one desires a proposition ... as a matter of fact that is only a prejudice (Russell 1956, p. 218).' Generally, each having of a thought consists of an object (the proposition to which it is directed) and an attitude (the manner in which the subject is disposed towards the object). Russell, notoriously, changed his mind about what propositions are, conceiving of them (and their 'constituents') sometimes as linguistic, sometimes as non-linguistic. But he did hold firm to the thesis that to have the thought that p is to have an attitude towards the proposition p. He therefore dubbed all these psychological verbs 'prepositional verbs', and the phenomena they pick out have come to be known as 'propositional attitudes'. To know what someone thinks, on such a conception, is correctly to identify the proposition which is the (abstract) object of their attitude. Although this conception has a very prestigious pedigree, doubts about it are expressed here in the articles by John Searle and K. V. Wilkes. Russell's assumption that when one believes that p what one believes is the proposition that p has also come under fire from Alan White (1972, 1979), who argued that Russell conflated two different kinds of accusatives of psychological verbs. On the one hand, there is what is believed when someone believes a person or a story (the 'object-accusative' of the verb). On the other hand, there is what is believed when one believes that p (the verb's 'intentionalaccusative'). In the first kind of case, thinking is a relation which obtains only if both terms of the relation exist. In the second, what is believed (namely, that p) need not obtain at all. The fact that, in his contribution to this volume, John Hyman deploys a closely related distinction in his study of the resemblance theory of pictorial representation demonstrates that the issue has relevance beyond the study of linguistic representation. Hyman defends the resem-
Introduction: Thought and Language blance theory by showing that it can be separated from an unacceptable theory of visual perception with which it has been associated. He challenges the rival semiotic theory of pictorial representation put forward by Nelson Goodman, which seeks to exploit a supposed parallel between pictures and words, arguing that Goodman's version of this theory invokes the very conception of visual perception which both Hyman and Goodman eschew. The lingualist tradition came to something of a head in Ludwig Wittgenstein's Tractatus Logico-Philosophicus (first published in German in 1921). In a September 1916 entry from his Notebooks, Wittgenstein had already remarked that [I]t is becoming clear why I thought that thinking and language were the same. For thinking is a kind of language. For a thought too is, of course, a logical picture of the proposition, and therefore is just a kind of proposition. (Wittgenstein, 1979, p. 82) This line of thought is pursued in the Tractatus, the logical foundation of which is the 'picture theory' or 'model theory' of representation, according to which a proposition is both an expression of a thought (3.1) and a logical picture of reality (4.01). On the Tractatus conception, a thought, too, is a logical picture of facts (3), meaning that the thought that p must consist of psychological elements arranged in the same way as the elements of the propositional sign 'p'. To think is to create psychological representations which are isomorphic with possible states of affairs. This view would commit Wittgenstein to the picture of language earlier endorsed by John Locke, for whom the primary function of language is to communicate thoughts by using perceptible signs to effect a correlation between the 'mental state' of the speaker and that of the hearer. Wittgenstein, notoriously, said almost nothing about the constituents of these psychological representations. In replying to a letter in which Russell had asked him about the constituents of a thought, Wittgenstein retorted I don't know what the constituents of a thought are but I know that it must have such constituents which correspond to the words of Language. Again, the kind of relation of the constituents of the thought and of the pictured fact is irrelevant. It would be a matter of psychology to find out. (Letter of 19 August, 1919). In contrast with Russell's early theory, which construed judgement as a relation between a judging subject A and a proposition p, Wittgenstein held that for A to judge that p is for a psychological
John Preston fact to be true of A - a fact which, if p were true, would be isomorphic to the fact that p. One of the problems with this view is whether any such relation could possibly constitute judgement (or assertion, or thought): whether having a mental constituent with certain psychological properties is either necessary or sufficient for being truly said to think that so-and-so is the case. Although Wittgenstein did insist that thoughts do not consist of words, he admitted that they comprise 'psychical constituents that have the same sort of relation to reality as words' (ibid.). However, the postulated relationship between these two kinds of pictures, language and thought, is not transparent. Wittgenstein put this in terms of the same metaphor Frege had used: Language disguises thought. So much so, that from the outward form of the clothing it is impossible to infer the form of the thought beneath it, because the outward form of the clothing is not designed to reveal the form of the body, but for entirely different purposes. (4.002) The task here envisaged for philosophy itself is to clarify thoughts, which are otherwise cloudy and indistinct (4.112), and to set limits to what can and cannot be thought, by 'working outwards through what can be thought' (4.114). The realms of the thinkable and the sayable coincide (5.61). Contemporary Lingualism in Cognitive Science A very different but widespread contemporary conception of philosophy, facilitated by W. V. Quine's well-known critique of the analytic/synthetic distinction, has it that philosophy is continuous with linguistics and psychology in forming part of an amalgam known as 'cognitive science' (roughly, the intersection of artificial intelligence, cognitive psychology, the neurosciences, Chomskyan linguistics, philosophy of mind and parts of related fields such as anthropology and sociology). Somewhat surprisingly, the Tractatus conception of thought, especially the idea that thinking is a kind of language, can be seen as an ancestor of this contemporary account, since the latter also centres around the idea of a language of thought. This lingualist view received its canonical defence in Fodor (1975). In this book, Fodor aimed 'to resurrect the traditional notion that there is a language of thought' (ibid., p. 33), in order explicitly to provide an underpinning for cognitive theorising. He set out, for the first time, the philosophical and methodological presuppositions of this kind of psychology, arguing powerfully that
Introduction: Thought and Language contemporary cognitive theorising clearly presupposes not only that there must exist a language of thought, but also that cognition consists in computational operations upon sentences of that language, and that the language in question could not possibly be one which the subject had learned. This 'computationalist' view of cognition, at which John Searle's critique in this volume is aimed, portrays thinking as, or as the outcome of, the computational manipulation of mental symbols on a level below that of conscious awareness. Our knowledge of the operation of contemporary digital computers is supposed to give us a way to understand thinking itself. Searle complains that such computationalist explanations of cognition leave out the subject's consciousness. He sets out the features which characterise rule-governed, intentional behaviour, and argues that too few of them are respected by the explanations cognitive science offers. Further, he suggests that although cognitive scientists would like to portray their activity as continuous with the natural sciences, the phenomena they study do not satisfy a necessary condition for the objects of natural-scientific explanation: they are crucially observer-dependent. The fully fledged language of thought hypothesis is that thinking consists, quite literally, in computational operations performed upon sentences of mentalese, an internal language with which thinkers are innately endowed.3 For a creature to think, on this view, is for it to have rational symbol-manipulations occurring in its mental medium. (In a later collection of articles (1981, p. 1), Fodor admits that the theory of mind he intends to defend looks a lot like that of Descartes.) The mind is conceived of as a set of interlocking 'modules', characterised not in terms of their structure, or of the material they are 'realised' in, but in terms of their functional interrelations. Their functioning consists in the processing of information encoded in linguaform mental representations. T o believe that p is the case, for example, is to have a sentence which means 'p' in one's 'belief box' (to use Stephen Schiffer's memorable expression). Likewise, to hope, fear or desire that so-and-so, is to have the appropriate sentence in one's relevant mental module. Notice how clearly this gives expression to an updated Russellian perspective, with sentence-tokens (which are materialistically respectable, since they can supposedly be instantiated in brain states) standing in for problematic Russellian 'propositions'. If, as Fodor claims, our only remotely plausible accounts of minds and their capacities, or even our best such accounts, do presuppose that minds are such systems, 3
For an excellent introductory critique of this account, see McGinn's (1982), Ch. 4.
John Preston we must take the contemporary lingualist case very seriously indeed.4 The essays in this volume by Donald Davidson, Hans-Johann Glock and Daniel Dennett raise certain problems for this account. Dennett, for example, argues that although we do talk silently to ourselves, and we do find ourselves thinking thoughts not framed in the words of any natural language, these are sophisticated phenomena rather than the foundation of cognition. He stresses the biological role of public words as tools, instruments which allow us to structure information, to stretch our faculties and even to recast the resources within our brains. The language of thought hypothesis, by contrast, is shown up as distinctly wnbiological, and the analogy with a particular computer architecture on which it is premised is challenged along the way. More recently, Fodor's arguments for a 'representational theory of mind' have been joined by an argument which in no way relies on psychological science. It is now claimed that the feature of 'propositional attitudes' known as intentionality itself forces upon us a theory of mental representations. The intentionality of thought consists in the fact that verbs of propositional attitude are about or directed, in a curious way upon, the situation specified in the proposition they contain. The curious feature is that those situations need not actually obtain: that is, one can believe that p, hope to 0, or fear the x, without p ever being the case, 0ever occurring, or x ever existing. The argument is then that for a mental phenomenon to be intentional (in this special sense) just is for it to be representational: a belief that p represents the world as being such that p is the case. So we are allegedly committed, simply by recognising the phenomenon of intentionality, to a theory of mental representations. The remarks in Searle's paper on the concept of information, when transposed into the key of 'representation', contain materials with which to criticise this argument. The Problem-Solving Approach to Thinking Whether this contemporary lingualist picture falls victim to the same problems which flawed the Tractatus is an excellent question. 4
This version of lingualism might also be able to explain the familiar 'tip-of-the-tongue' phenomenon - mentioned in several papers in this collection — in which one apparently finds oneself searching for words in which to frame one's antecedently existing thoughts. The idea would be that to have something on the tip of one's tongue is to have had the thought in question, but to be temporarily unsuccessful in translating it from mentalese into one's own natural language.
Introduction: Thought and Language Wittgenstein himself eventually moved to a perspective in which questions like 'How do we use the expression "to think", and its cognates?' and 'Might there be only a "family resemblance" between instances of thought?' displaced his previous concern with the metaphysical preconditions for the possibility of thought. The later sections of dock's essay, in which the lingualist notion that thought must take place 'in' a medium is criticised, but which also seek to demonstrate the 'internal' relation between thought and language, exemplify this approach. In a related vein, it has been argued that Wittgenstein's later perspective cuts the ground from underneath cognitivist theorising and the philosophical picture embroiled in it: see Hacker (1993), ch. IX. This Wittgensteinian perspective has more in common with what Christopher Hookway identifies as the 'problem-solving' approach to thinking. Here, postulation of mental representations is apparently reined in (to varying degrees), in favour of a focus upon public problem-solving activities and the analysis of problem-solvers' protocols. Like Quine, Hookway emphasises the way in which language amplifies and extends our problem-solving abilities, allowing us to cast our problems in an external form which makes them more tractable. He suggests that adoption of the problem-solving perspective would transform our understanding of the debate over the analytic/synthetic distinction, allowing us to re-evaluate Rudolph Carnap's resistance to Quine's critique. Subsequently, he argues that examination of the norms which govern reflective thought reveal that the relation between questions and answers, problems and solutions, is not a purely semantic matter, but involves pragmatic and contextual considerations. Since the problem-solving approach lends itself readily to experimental investigation, it has been popular among psychologists, as the fascinating experiments detailed in Lawrence Weiskrantz's paper in this volume bear witness. Weiskrantz is sceptical both about certain philosophers' strictures on the possibility of thought without language, and about others' tolerance as to what counts as thought. He proposes that we take 'thinking' to refer to activities like problem-solving, and argues that if we do so, then observations and experimental studies show these activities to be possible in the absence of natural language. But he hesitates to extend the thesis to conscious thought, 'awareness', since although human subjects with cognitive defects have the capacity to manipulate images and symbols (in some sense), they cannot bring their manipulation of these items to consciousness. Experiments show that they know how to perform the tasks in question, but also that they cannot be said to know that they can do so, since these subjects will respond by sin-
John Preston cerely denying that they have the capacities in question. The phenomenon known as 'blindsight', which Weiskrantz himself has done so much to illuminate, and which has captivated the imagination of many philosophers, serves as only one example of this predicament. The problem-solving approach also has other philosophical adherents, most notably within the tradition of pragmatism. The founding fathers of that tradition, as well as their successor George Herbert Mead and more recent thinkers influenced by them, such as W. V. Quine, Donald Davidson and Daniel Dennett, all agree in linking the presence of thought to the satisfaction of some publicly observable criterion (language, physiological processes, or problemsolving behaviour itself). Quine, in his paper in this volume, gives a sophisticated account of language-learning along lines compatible with the strictures of the pragmatist tradition, strictures which he has already famously employed in his discussion of 'radical translation'. Here, both he and Davidson stress the intertwining of thought and language, and detail a series of scenarios and advances which may have brought us to where we are today, a possible history of thought and language in the species, reproduced in miniature within the development of the individual organism. In his contribution, Davidson also extends his well-known critique of the idea that there could be rival conceptual schemes, showing how this idea is linked with the unacceptable metaphorical notion that we see the world 'through' language. For Quine, Davidson, and Dennett, our interaction with the world is direct, and language is not an representing intermediary but a set of tools which we use to cope with incoming sensory information. Davidson, for his part, compares language with a mode of perception in a way that bears comparison with Aristotle's discussion of thought and perception in De Anima 427a17. William James, one of the founders of the pragmatist tradition, famously characterised the 'Self of selves' as consisting mainly in a collection of 'peculiar motions in the head or between the head and throat' (1890, ch. X). In this respect, the pragmatist approach has something in common with psychological behaviourism. John B. Watson, one of the founders of the latter movement, identified thoughts with 'the action of language mechanisms' (1919, p. 316), and B. F. Skinner, behaviourism's principal exponent, although he rejected the identification of thinking with sub-audible talking, associated thought with the probability of verbal behaviour. Lingualism, evidently, can appear in very different guises. Language of thought theorists, however, would undoubtedly reply that in so far as we have any idea of how creatures solve prob10
Introduction: Thought and Language lems, we simply must present such activity as consisting in the computational manipulation of mental representations. In support of this, they might well refer to the heartland of artificial intelligence, where accounts of problem-solving (and games-playing, planning, etc.) are, of course, exclusively computational. Thought Without Language? Lingualism apparently derives little support from the ways in we conceptualise the behaviour of non-language-users since, arguably, they can be credited with certain kinds of thoughts (beliefs, desires, fears, etc.) in the absence of language. Possible cases of thinkers without language would include: pre-verbal infants, non-human animals and physiologically damaged humans (discussed here in the papers by Wilkes and Weiskrantz). Contemporary lingualists, however, preserve the intuitive judgement that such creatures think, since, according to their view, the behaviour such creatures exhibit itself necessitates the supposition that they are (unconsciously) manipulating mental representations. In 1982, however, Donald Davidson published a provocative article in which he argued that only language-using creatures can have beliefs, or any other 'propositional attitudes'. Davidson (1982) softens us up by first arguing that there couldn't be much thought without language. He starts from the fact that we ascribe propositional attitudes 'holistically': either a creature has a rich network of such attitudes, or it has none. All the propositional attitudes require a background of beliefs. But in the case of non-language-users, it is impossible to say exactly which propositional attitudes they have: we have no conception of how to tell whether such creatures have certain specified beliefs, desires, etc., rather than other, very closely related ones. The complex patterns of behaviour needed to justify the attribution of specific propositional attitudes are present only in language-users. Davidson's argument proper is then that in order to have a belief (or any propositional attitude), it is necessary to have the concept of a belief, and in order to have this, it is necessary to have language. To have the concept of belief is to have the concept of a state of an organism which can be either true or false. In order to be able to think in this way, one must have grasped the subjective/objective contrast. One must be capable of identifying some pair of situations in the first of which one had a false belief, and in the second of which one acquired in its stead a true one. But the only way of revealing command of this contrast is by means of linguistic communication. Those who are tempted to reply to 11
John Preston Davidson that it is not necessary to have the concept of belief in order to have beliefs will now have to contend with another aspect of the argument, developed in the present volume: that having concepts can in no way be separated from having fully fledged propositional thoughts. The 'Content' of Thoughts and Utterances Contemporary philosophers have also been much exercised by the question of what fixes the 'content' of thoughts, that is, what makes my thought that p is the case a thought that p, rather than that q. The best-known theories on offer here are 'internalism', according to which the content of thoughts is fixed by what goes on in one's consciousness, and 'externalism' for which, by contrast, the content of thoughts is fixed by features of the (physical or social) environment. The latter version of this view - social externalism - is the subject of Andrew Woodfield's paper. Woodfield defends the idea that the contents of 'high-level' conceptual thoughts are fixed by the public meanings of the words in which those thoughts are expressed. He then argues that this view has important implications for psychology. Notably, he proposes that it supports a version of the Sapir-Whorf hypothesis, the hypothesis of 'linguistic relativity', according to which this kind of thinking is structured by the language(s) one has learnt to speak. Woodfield's view of concepts as intellectual norms or social traditions and Davidson's idea of concepts as ways of classifying, as well as their strategies for resisting conceptual relativism, might profitably be compared. A neglected alternative to both internalism and externalism, defended here by Glock, is the view that what makes a thought the thought it is (though not in the causal sense of bringing about that thought) is how the thinker would explain it. Notice, however, that this Wittgensteinian view, taken by itself, has the same consequences as Davidson's argument, that non-language-users cannot be thinkers. A rather different approach, championed by K. V. Wilkes, challenges the assumption that thoughts and utterances must have a determinate content in the first place. Wilkes argues that, to the contrary, there are plenty of cases in which what was meant by the speaker (or thinker) is simply indeterminate, that although answers to the question 'What did he mean?' must be restricted to some extent, we ought not to assume that such questions have a single right answer. (Quine's 'indeterminacy thesis' reverberates here). Like Searle and Hookway, Wilkes seeks to impugn the idea that 12
Introduction: Thought and Language beliefs, desires, hopes and fears are rightly characterised as attitudes toward propositions. If these more sceptical approaches to the problems surrounding 'mental content' are tenable, the conclusion that thoughts cannot be ascribed to non-human animals might well be placed in doubt. Conclusion How does lingualism fare in the currents of contemporary debate? Although there is no general animus against representational theories of mind among the contributors to this volume, none of them subscribes to the language of thought hypothesis, and none of them explicitly equates thought with language. Opinions on computationalism and the merits of contemporary cognitive psychology are strongly divided. However, some contributors endorse positions closely related to the core lingualist thesis. Davidson and Glock, for example, see the ability to entertain complex thoughts (at least) as restricted to language-users. Many of the contributors take up and elaborate the theme, familiar from Vygotskian psychology, that words fulfil a crucial role in thought and problem-solving activity, and some accept the identification of thought with problem-solving. Indeed, perhaps the broadest commonality among the essays in this volume is in treating thought from the 'problem-solving' point of view, rather than from the subjectivist, individualist, 'internalist' perspective usually, but perhaps inappropriately, linked with the name of Descartes. References Davidson, D. 1982. 'Rational Animals', Dialectica 36, 317-27. Reprinted in E. Lepore and B. McLaughlin (eds), Actions and Events: Perspectives on the Philosophy of Donald Davidson. Oxford: Blackwell, 1984 Fodor, J. 1975. The Language of Thought. Cambridge MA: Harvard University Press 1981. Representations: Philosophical Essays on the Foundations of Cognitive Science. Sussex: Harvester Press Frege, G. 1918. 'The Thought: A Logical Inquiry'. Reprinted in P. F. Strawson (ed.), Philosophical Logic. Oxford University Press, 1967 Hacker, P. M. S. 1993. Wittgenstein: Meaning and Mind, Part I: Essays. Oxford: Blackwell James, W. 1890. Principles of Psychology. Reprinted, Cambridge, MA: Harvard University Press, 1981 Leibniz, G. 1677. Dialogue on the Connection between Things and Words Reprinted in P. P. Weiner (ed.), Leibniz Selections (New York: Scribners, 1951). 13
John Preston McGinn, C. 1982. The Character of Mind. Oxford University Press Russell, B. A. W. 1956. 'The Philosophy of Logical Atomism' (1919). Reprinted in Logic and Knowledge. London: Allen & Unwin Watson, J. B. 1919. Psychology from the Standpoint of a Behaviorist. Philadelphia: J. B. Lippincott White, A. R. 1972. 'What We Believe', in N. Rescher (ed.), Studies in the Philosophy of Mind: American Philosophical Quarterly Monograph Series, No. 6. Oxford: Blackwell 1979. 'Belief as a Propositional Attitude', in G. W. Roberts (ed.), Bertrand Russell Memorial Volume. London: Allen & Unwin Wittgenstein, L. 1979. Notebooks, 1914-1916, 2nd edition. Oxford: Blackwell
14
Seeing through Language D O N A L D DAVIDSON We see the world through language; but how should we understand this metaphor? Is language a medium that simply reproduces for the mind, or accurately records, what is out there? Or is it so dense there is no telling what the world is really like? Perhaps language is somewhere in between, a translucent material, so that the world bears the tint and focus of the particular language we speak. All these attitudes have, or have had, their apostles, but none of them seems to me more than half right, and none captures what is most important about language. Language is certainly a convenient human skill which we use in coping with one another in our common terrestrial setting. Without it we would not think of things, as we do. But it does not follow, of course, that we never perceive how the world really is, as Kant thought, or that every view is necessarily distorted, as Bergson and many others have held. There might be an argument for this view if it were possible, in principle at least, to isolate some unconceptualised given which could be shaped by the mind, for then it might make sense to imagine a multitude of structures within which the given could be shaped. Without the idea of such a given, however, it is hard to divine what it is that wants shaping, and few of us now are taken by the idea of an unprocessed given. Do we understand what we mean by a real alternative to our conceptual scheme? If a scheme could be decoded by us, then it would not, by this very token, be all that different from ours except, it might be, in ease of description here or there. If we could explain, or describe, in a convincing way, how an alternative scheme deviates from ours, it would again be captured in our system of concepts. This is not to deny that some people have conceptual resources not available to everyone. Biologists, aeronautical engineers, solid state physicists, musicologists, cartographers, molecular biologists, selenographers and psychoanalysts all command vocabularies and theories many of us do not. In our more restricted way, we common types also have our specialities: our own list of proper names, with their uniquely contextualised references, our private endearments and verbal twists, our own mispronunciations and malapropisms. And there is no denying that some dialects reveal sexist, nationalist or racist features, while all human languages are rife with barely concealed anthropomorphism. These are, if we like to say so, dif15
Donald Davidson ferences or provincialisms in our conceptual schemes. But they are variants or features we can explain to one another, or could, given enough time, adequate attention and sufficient intelligence on both sides. Not everyone can grasp the concepts of quantum mechanics — I can't — but the language of relativistic quantum physics doesn't constitute a different conceptual scheme. It's just a suburb, though an exclusive one, of the universal scheme which assumes an ontology of ordinary macroscopic objects with their ordinary properties. The trouble with the idea of genuinely incommensurable languages and conceptual schemes is not that we couldn't understand them, but that the criteria for what would constitute a scheme incommensurable with ours are simply unclear. Perhaps we think we can imagine a culture where creatures communicate in ways we are permanently disabled from penetrating. But speculating on this possibility hardly advances the case until we decide on our criteria for communication. Fluent exchange of information, purposeful interaction? But how are these manifested? The only ends we recognise are the same as, or analogous to, our own. Information as we know and conceive it has a propositional content geared to situations, objects and events we can describe in our homespun terms. Of course it happens in exotic settings that we recognise that we are witness to intelligible conversation, though grasping nothing that is said. It also is possible that the parling parties have agreed (in a language we could learn) to use an apparently unbreakable code. What these folk say has a translation we could understand, but we can't discover it for ourselves. But in such situations we have good grounds for believing that what we do not grasp we could learn to decipher: the parties look to be people like ourselves, and we are justified in thinking we descry known purposes and activities. These cases pose no problem for the anti-conceptual relativist; there is a serious problem only where no translation is possible. It is in this supposed case of genuinely incommensurable languages where I sense unintelligibility in the supposition. Perhaps we don't really understand the concept of radically different ways of thinking and talking. Still, doesn't it make sense to hold that our actual languages mould our perception of the world to such an extent that what we take in is always distorted? What is obvious is that our language is rich in facilities that match our interests, and lacks the means for easily expressing what is orthogonal to those interests. Our basic vocabularies trace out the vectors which point in the directions in which we naturally generalise; apart from discussing why this is so, we have no interest in emeroses and the classes of things that are gred, bleen or grue. These are concepts at least expressible in terms to be found in our growing dictionaries 16
Seeing through Language in fact, if 'grue' is not already there, I'm sure it soon will be. But there are endless classes for which we have no term, no matter how complex. Is this distortion? If it is, it is not language that is responsible. Of course, language reflects our native interests and our historically accumulated needs and values, our built-in and learned inductive dispositions. But this fact hardly supports the claim that language seriously distorts or shapes our understanding of the world; the influence, such as it is, goes the other way. The most we are entitled to say is that as individuals we inherit culturally evolved categories we personally did little to devise. In this case, language does not distort; rather, society gives us a leg up on coping with the environment it partly constitutes. What we should resist is the claim that it is truth that is bent or distorted by language. My language may (or may not) have something to do with my, or society's, interest in the question whether the clearing of tropical forests is speeding the destruction of the ozone layer, but language has nothing to do with the truth of the matter. Plenty of concepts are vague, but putting grey areas and ambiguity aside, most of our declarative utterances are simply true or false, not true then but false now, not true for me and false for an inhabitant of the Tuamotu Atolls, not partly false and partly true. Our languages do not distort the truth about the world, though of course they allow us to deceive ourselves and others, according to predilection. I started out by describing three possible poses that have been struck with respect to the role of language in our thinking about the world. One was that language is opaque, hiding the real thing from us. I rejected this view. A second was that language is a translucent medium, leaving its own character written on everything in its domain. This seemed trivially true at best, an exaggeration of the simple and natural fact that language reflects our interests and our needs. There remains the idea that language is transparent, a medium that can accurately represent the facts. Alas, we know that this too is an idea with no cash value. Nouns, names and predicates may refer to, or be true of, one or more things, but they cannot, by themselves, represent facts or states of affairs. Only sentences can do this, and no one has discovered a way of individuating facts or states of affairs in a way that would help to explain which fact a given sentence represents. If, in saying that language represents facts, we mean no more than that we can use sentences to describe objects and events, no harm is done. This is, after • all, just a fancy version of the platitude that some sentences are true and some false. But we deceive ourselves when we talk of language representing reality (or anything else) unless we can usefully specify the entities represented. 17
Donald Davidson Is anything now left of the metaphor with which we began, the figure that has language as something through which we view the world? No: as a metaphor it misleads. Language is not a medium through which we see; it does not mediate between us and the world. We should banish the idea that language is epistemically something like sense-data, something that embodies what we can take in, but is itself only a token, or representative, of what is out there. Language does not mirror or represent reality, any more than our senses present us with no more than appearances. Presentations and representations as mere proxies or pictures will always leave us one step short of what knowledge seeks; scepticism about the power of language to capture what is real is old-fashioned scepticism of the senses given a linguistic twist. We do not see the world through language any more than we see the world through our eyes. We don't look through our eyes, but with them. We don't feel things through our fingers or hear things through our ears. Well, there is a sense in which we do see things through - that is, by dint of having - eyes. We do cope through having language. There is a now-metaphorical point to my title. There is a valid analogy between having eyes and ears, and having language: all three are organs with which we come into direct contact with our environment. They are not intermediaries, screens, media, or windows. Perhaps we are influenced by the idea that a language - especially when its name is spelled with a capital, as in 'English', 'Croat', 'Latvian', 'Inuit' or 'Galician' - is some sort of public entity to one or more of which each of us subscribes, like the telephone service, and which therefore really is extraneous to us in a way our sense organs are not. We forget there is no such thing as a language apart from the sounds and marks people make, and the habits and expectations that go with them. 'Sharing a language' with someone else consists in understanding what they say, and talking pretty much the way they do. There is no additional entity we possess in common any more than there is an ear we share when I lend you an ear. Of course there are differences between being able to converse with others, and being able to see. We develop sight early and without social prompting; the conditions for acquiring a language are more complex, and mastery develops later. But it does develop amazingly fast once things get going. The phonemes of our mother's tongue seem to have a start in utero (Locke 1993), but sentences emerge only after a year or two. By three years, most children glibly generate sentences, and have the basic grammar of their environment right. The average six-year-old commands about 13,000 words, and a good high-school student may know 120,000. The 18
Seeing through Language window for learning all this is brief; after eight or so, practically no one can learn to speak a new language (first or second) like a native. There seems little reason to doubt that we are genetically programmed in fairly specific ways to speak as we do; every group and society has a language, and all languages are apparently constrained by the same arbitrary rules. Tribes we consider primitive have languages as complex and complete as those of developed cultures. We tend to think speech is radically different from the senses, partly because there is no external organ devoted just to it and partly because of the diversity of languages. But these differences are superficial. Speech, like the sense organs, has its specialised location in the brain; as a result, brain damage can cause loss of the ability to use language without destroying general intelligence. And, more significantly, all languages apparently share structural rules despite the surface variety. The evidence for this is partly the discovery of universal constraints on grammars. There is also the astonishing fact that children brought up hearing nothing but pidgin, which is a highly simplified invention of adults thrown together and lacking a common language, soon elaborate the pidgin into a Creole as developed and complicated as French or Turkish. These bits of information come from various sources, some of them influenced by Noam Chomsky, but particularly from Steven Pinker's recent book, The Language Instinct, which brings them persuasively together. Pinker concludes that 'Language is not a cultural artifact that we learn the way we learn to tell time...Instead, it is a distinct piece of the biological makeup of our brains' (Pinker 1995, p. 18). My opening metaphor seems to chime with Pinker's remark, 'When we are comprehending sentences, the stream of words is transparent, we see through to the meaning...automatically' (p. 21). No wonder that, following Chomsky, he calls language a 'mental organ' (Chomsky 1980, pp. 138-9, Pinker 1995, p. 307). What, though, does all this have to do with the relation between language and thought? According to Pinker, Fodor and a number of others, the extraordinary ease with which language develops, added to the apparent existence of linguistic universals, shows that what is innate — that is, genetically programmed — is an internal language, the language of thought, or mentalese. According to the theory, this inner language is not learned, but emerges as part of our genetic heritage, and it is prior to any spoken language. The salient point is that the existence of mentalese does not depend on the development of language, but vice versa. Thus, given the universal grammar that is wired in, 'the connectedness of words...reflects the relatedness of ideas in mentalese. This solves the problem of taking an interconnected web of thoughts in the mind and encoding them as a string 19
Donald Davidson of words...For the child, the unknown language is English (or Japanese...or Arabic); the known one is mentalese' (Pinker 1995, pp. 101, 102, 278). In this last quoted passage, Pinker is comparing the child to Quine's radical translator, with the difference that for Pinker the child doesn't need to work things out; the child simply knows right off what idea in mentalese is represented by the words he hears. Pinker has no doubt about the priority of mentalese. Do we think in English or Cherokee or some other language, he asks, '[O]r are our thoughts couched in some silent medium of the brain - a language of thought, or "mentalese" - and merely clothed in words whenever we need to communicate them to a listener?' (Pinker 1995, p. 56). He votes for the silent medium. The arguments are various: we often know what we think but can't find the words; we sometimes recognise that what we said was not what we meant; it is a silly myth that conceptual schemes can differ widely; and there is the universal grammar. This general view is now accepted by many linguists and cognitive scientists, but it seems to me the arguments for it are flawed and the conclusions confused. It is no more philosophically significant that major aspects of language ability are wired in than that colour and brightness contrasts are carried to the brain by the optic nerve rather than processed by some higher cognitive machinery. It is worth emphasising that language aptitude is part of our natural equipment, and not a tool we contrived for coping with problems of understanding, calculation and communication. I like the analogy with the sense organs, and therefore the implication that language is not something that comes between us and reality; it can't come between, since it is part of us. But postulating a language of thought spoils what is attractive about this picture; if the language of thought is what is part of us, then our spoken language is an intermediary between thought and what thought is about, and what is genetically engineered does threaten to hide or distort the world in much the way Kant thought the architecture of the mind does. The arguments for the existence of a language of thought prior to, or independent of, a socially engineered language are feeble. The fact that we sometimes cannot find words for what we want to say has simpler explanations than the postulation of a pre-existing internal but wordless message striving to find its translation into a spoken idiom. It is enough to suppose that we sometimes cannot access words or phrases we already know, or even that, already having a language, we are able to think of new things that need saying. The idea of inborn constraints on syntax, for which Chomsky has argued so vigorously, is, as he has shown, supported by impressive empirical evidence. There may remain, for some of us, a question to 20
Seeing through Language what extent these constraints are an artefact of our means of describing alien languages in our own; but the evidence is convincing that this is by no means all there is to it. Not only do we come into the world equipped for language acquisition, but we know something about the constraints on what comes naturally. What conclusions can we draw about real languages, or thought? What is needed in the way of genetically engineered linguistic aptitude to explain the relative ease of language acquisition and the similarities in actual spoken languages has little, if anything, to do with the contents of our thoughts or utterances. What we are born with, or what emerge in the normal course of early childhood, are constraints on syntax, not semantics. There is no reason to suppose that ideas, concepts or meanings are innate if this is taken to mean anything more than that people have come to have languages and thoughts that reflect the needs and interests of human animals. Nor is it surprising, given our common heritage, that our thoughts and utterances are mutually intelligible - up to a point, of course, since success in interpretation is always a matter of degree. This is not to say that constraints on syntax may not generate structural constraints on semantics; though it is not easy to think how, in detail, the argument would go. In any case, my contention is not that what we think and say is not constrained by our genes; it is the weaker claim that we are not born with anything like a contentful language. Evolution has made us more or less fit for our environment, but evolution could not endow us with concepts. Nature decided what concepts would come naturally, of course; but this is not to say the mind knew in advance what nature would be like. Something more should be said about the 'argument from the poverty of the stimulus', the argument that much of what we come to know about the language we speak must be inborn because we acquire an accurate (though mostly unconscious) knowledge of the grammar, vocabulary and even semantics of our mother tongue on the basis of such sparse evidence. I have already expressed my doubts about semantics, at least if semantics has to do with reference and truth. But recent research also throws doubt on the idea that pre-linguistic infants lack the ability to learn very rapidly and accurately from limited and incomplete input. Experiments reveal that eight-month-old children learn to segment speech into words on the basis of nothing but statistical relationships between adjacent sounds after only two minutes, 'suggesting that infants have access to a powerful mechanism for the computation of statistical properties of the language input' (Safran, Aslin, and Newport 1996, p. 1926). In other words, not that much has to be wired in; learning has an important role to play in entry into the realm of speech and thought. 21
Donald Davidson These are important matters of degree. What matters, for present purposes, is that once in place, language is not an ordinary learned skill; it is, or has become, a mode of perception. However, speech is not just one more organ; it is essential to the other senses if they are to yield propositional knowledge. Language is the organ of propositional perception. Seeing sights and hearing sounds does not require thought with propositional content; perceiving how things are does, and this ability develops along with language. Perception, once we have propositional thought, is direct and unmediated in the sense that there are no epistemic intermediaries on which perceptual beliefs are based, nothing that underpins our knowledge of the world (Davidson 1983). Of course, our sense-organs are part of the causal chain from world to perceptual belief. But not all causes are reasons: the activation of our retinas does not constitute evidence that we see a dog, nor do the vibrations of the little hairs in the inner ear provide reasons to think the dog is barking. 'I saw it with my own eyes' is a legitimate reason for believing there was an elephant in the supermarket. But this reports no more than that something I saw caused me to believe there was an elephant in the supermarket. Sometimes we have sensations, and we may, on occasion, refer to them as reasons for beliefs. But sensations, or their wispy messengers - percepts, sense-data, the like - do not constitute reasons, though the belief that we heard a noise, or witnessed a fiery streak in the sky, may be a reason, when coupled with appropriate subsequent information, for thinking we heard an explosion or a saw a plane crash. The reason sensations, percepts and sense-data cannot provide epistemic support for beliefs is simple: reasons have to be geared conceptually to what they are reasons for. The relation of epistemic support requires that both relata have propositional content, and entities like sensations and sense-data have no propositional content. Much of modern philosophy has been devoted to trying to arbitrate between the unconceptualised given and what is needed to support belief, but we now see that there is no chance of success. The truth is, nothing can supply a reason for a belief except another (or many another) belief.1 Perceptual beliefs are formed at first spontaneously. They are simply caused by what goes on that we can see, hear, touch, taste and smell. We have no control over the onset of such beliefs, except as we can move our bodies to put ourselves in the way of reception. Control sets in once a belief is caused; another look can correct the 1 This is a view I argued for in Davidson (1983), p. 50. John McDowell, in his recent John Locke Lectures, accepts the claim that reasons must have propositional content, but rejects the idea that only beliefs can be reasons for beliefs (McDowell 1994, p. 63). 22
Seeing through Language first impression, a moment's reflection can cancel the idea that we are seeing a long-dead friend. What we see can wipe out what we thought we heard. Comparing observations is what we credit scientists with, but we all do it, if not so systematically or methodically, all the time. In the end, it is perceptions we have to go on, but on the basis of perceptions we build theories against which we evaluate further perceptions. I take for granted that the perceptual beliefs we cannot help forming, however tentatively, are themselves heavily conditioned by what we remember, what we just a moment ago perceived, and the relevant theories we have come to accept to one degree or another. Beyond the skin there is mindless causality, but what gets bombarded is a thinking animal with a thoroughly conditioned apparatus. There is no simple relation between the stimulus and the resulting thought. Given the mostly inscrutable complexity of this relation, why should our empirical beliefs, even the perceptual ones, be trustworthy? If the only rational grounds for a belief are other beliefs, what role can nature play in determining the contents of beliefs? The question is akin to the question about language. It is one thing to insist that language, like perception, is unmediated, but this insistence makes it seem impossible to account for the contents of our observation sentences, just as it seems impossible to account for the contents of our perceptual beliefs. The two problems are obviously intertwined, for what accounts for the contents of thoughts must also constitute at least part of the explanation of why observation sentences have the contents they do. What makes these problems pressing is the question how beliefs, if epistemically supported by nothing more than other beliefs, can independently, or as a collection, be connected with the world. Whether or not these questions can be satisfactorily answered, I am pretty sure an answer to one will turn out to be, or at least lead to, an answer to all. One place to begin is by asking how the sentences directly tied to perception get their content. We may call these perceptual sentences. There is no reason to think all perceptual sentences are simple, or that they are the same for everyone. Not necessarily simple, since some of us learn to know directly, just by looking at the glass, that stormy weather is ahead; we hear certain sounds and know someone has said that the tide is up; we look in the cloud chamber and remark that we have seen an electron. In these cases we can give reasons for the beliefs we formed directly; we can explain why what we saw begot the belief it did- And not the same for everyone. Some people don't perceive that stormy weather is ahead because they haven't learned to read a barometer. Each of us has a unique repertoire of people recognised at a glance. 23
Donald Davidson Perceptual sentences have an empirical content given by the situations which stir us to accept or reject them, and the same goes for the beliefs expressed by those sentences. But what reason is there to suppose this content is appropriate? Even someone with a going language into which a new sentence can be fitted can learn to affirm a sentence in situations in which it is true without understanding it. Someone with no understanding of physics could easily come to utter the sentence 'There goes an electron' as the streak appears in the cloud chamber, while having little idea what an electron is. Understanding the sentence depends on prior theory, without which the content would be totally unlike what we think of as the meaning. But isn't theory, in a sense that extends theory to cover tacit understanding, isn't theory always needed for the conditioning of sentence to circumstance to yield the right content? Only someone knowledgeable about sailing ships could recognise on sight that he sees a brig and not a brigantine, though he might use the words in the right situations (a brig is a two-masted ship square-rigged fore and aft, while the latter differs in having a fore-and-aft rigged mainsail). Even a simple sentence like 'That's a spoon' (perhaps boiled down to 'Spoon!') if understood requires knowledge of what spoons are for, that they are persistent physical objects, and so on. So there must be more to content than is conveyed by saying it is given by the situations that stir us to accept or reject sentences appropriate to the situations. People do not acquire the gift of tongues by themselves; they are tutored intentionally or by accident by parents, playmates, teachers and Sesame Street. In the process, ostension, or what amounts to it, plays a major role. But how does ostensive teaching differ from the tutoring nature provides in any case? As far as I know, we are not born preferring berries that are blue to berries that are red. But the lone gatherer will be taught by nature to prefer the blue; they are much more apt to be nourishing and sweet. Perhaps an even keener mind will discover that the berries birds eat are almost always human fodder too. Mistakes make their mark: those big, gaudy, but poisonous berries will be avoided next time. What's the difference between this ordinary process of conditioning and the reward-andpunishment conditioning of sentence to situation that makes ostension the successful method it is? Corrections, whether administered by teacher, parent, playmate or nature, can in themselves do no more than improve the dispositions we were born with, and dispositions, as Wittgenstein emphasised, have no normative force. A slippery road is disposed to cause cars to skid, but we do not, if we are sensible, hold this against the road, though we may alter its character to suit our purposes by 24
Seeing through Language spreading sand or salt. Animals are different in that they are pleased and pained and so their behaviour can be altered by means not available with roads. But the point remains: we improve the road, from our point of view, by spreading sand or salt; we improve the child, from our point of view, by causing pleasure or pain. In neither case does this process, by itself, teach road or child the distinction between correct and incorrect behaviour. To correct behaviour is not, in itself, to teach that the behaviour is incorrect. Toilet training a child or a dog is like fixing a bathtub so it will not overflow; neither apparatus nor organism masters a concept in the process. We may be inclined to think that concept-formation is more primitive than entering the world of proriositional attitudes, in particular, of beliefs. But this is a mistake. Unless we want to attribute concepts to butterflies and olive trees, we should not count mere ability to discriminate between red and green or moist and dry as having a concept, not even if such selective behaviour is learned. To have a concept is to classify objects or properties or events or situations while understanding that what has been classified may not belong in the assigned class. The infant may never say 'Mama' except when its mother is present, but this does not prove conceptualisation, even on a primitive level, unless a mistake would be recognised as a mistake. Thus there is in fact no distinction between having a concept and having thoughts with propositional content, since one cannot have the concept of mama unless one can believe someone is (or is not) mama, or wish that mama were present, or feel angry that mama is not satisfying some desire. I stress the connection between concepts and thoughts only to make the point that concept-formation is not a way station between mere dispositions, no matter how complex or learned, and judgment. What must be added to a meaningless sound, uttered at moments appropriate for that same sound, uttered as speech, to transmute the former into the latter? It is not enough that the meaningless sound has been reinforced in the past and is now uttered because of its magical powers, or the fact that cats miaow to be fed would count as meaningful speech. What then? I am under no illusion that I can provide anything like an analysis; perhaps there is no answer that does not lead in a circle, for a non-circular answer would tell us how to account for intensionality in non-intensional terms. But I do think it is just here that language adds a necessary (though not sufficient) element. We remarked just now that ostension cannot, by itself, do the job, because of the aid it needs from a prior grasp of how language works. It is worth reflecting, though, on the initial phase of ostensive learning. At the start, there could be no point in the learner 25
Donald Davidson questioning the correctness of the teacher's ostensions. The learner may or may not be learning how others in some linguistic community speak, but the learner can discover this only later. In the private lesson, a meaning is being bestowed on words quite apart from any use those words may have at other times and with other people. If we think of ostension only as the teaching of a socially viable meaning we miss the essential lesson, which is that for the learner ostension is not learning something already there. The learner is in at a meaning baptism. If we ignore the difference between passing on an established meaning and the creation of a new one, the difference between teacher and innovator fades, and with it what distinguishes teacher and learner. Paring down the scenario even further, we can imagine a sort of proto-ostension before there is the general grasp of language that allows us to get more out of ostension than goes into it. In this elementary situation we can study some of the necessary conditions for the development of thought and language. These include the fact that all people generalise naturally in much the same ways. They avoid bitter tastes and loud, sudden sounds; they seek the sweet and the quiet. Learning requires three generalisations: the learned association of fire and hurt requires two, and the learning is displayed in the similarity of the responses: we avoid hurt by avoiding fire. Before there can be learning there must be unlearned modes of generalisation. Before there can be language there must be shared modes of generalisation. The sharing of responses to stimuli found similar allows an interpersonal element to emerge: creatures that share responses can correlate each other's responses with what they are responses to. Person A responds to person B's responses to situations both A and B find similar. A triangle is thus set up, the three corners being A, B, and the objects, events, or situations to which they mutually respond. This elaborate, but commonplace, triangular interaction between creatures and a shared environment does not require thought or language; it occurs with great frequency among animals that neither think nor talk. Birds and fish do it as well as monkeys, elephants and whales. What more is there to linguistic communication and developed thought? The answer is, I think, two things that depend on the basic triangle and emerge from it. The first is the concept of error, appreciation of the distinction between belief and truth. The interactions of the triangle do not in themselves generate this appreciation, as we see from the example of simple animals, but the triangle does make room for the concept of error (and hence of truth) in situations in which the correlation of reactions that have been repeatedly shared 26
Seeing through Language can be seen by the sharers to break down; one creature reacts in a way previously associated by both creatures with a certain sort of situation, but the other does not. This may simply alert the nonreactor to an unnoticed danger or opportunity, but if the anticipated danger or opportunity fails to materialize, a place exists for the notion of a mistake. We, looking on, will judge that the first creature erred. T h e creatures themselves are also in a position to come to the same conclusion. If they do, they have grasped the concept of objective truth. With the second, final, step, we move in a circle, for we grasp the concept of truth only when we can communicate the contents - the propositional contents - of the shared experience, and this requires language. The primitive triangle, constituted by two (and typically more than two) creatures reacting in concert to features of the world and to each other's reactions, thus provides the framework in which thought and language can evolve. Neither thought nor language, according to this account, can come first, for each requires the other. This presents no puzzle about priorities: the abilities to speak, perceive and think develop together, gradually. We perceive the world through language, that is, through having language.2 References Chomsky, N. 1980. Rules and Representations. New York: Columbia University Press Davidson, D. 1983. 'A Coherence Theory of Truth and Knowledge', in D. Heinrich (ed.), Kant oder Hegel? Stuttgart: Klett-Cotta Locke, J. L. 1993. The Child's Path to Spoken Language. Cambridge, MA: Harvard University Press McDowell, J. 1994. Mind and World. Cambridge, MA: Harvard University Press Pinker, S. 1995. The Language Instinct. New York: Harper Collins Safran, J. R., Aslin, R. N. and Newport, E. L. 1996. 'Statistical Learning by 8-Month-Old Infants', Science 274, 1926-28
2
I thank Barry Smith and Ernest Lepore for very helpful suggestions and corrections. 27
The only sure sign../: Thought and Language in Descartes JOHN COTTINGHAM Introduction Some people like to think that the modern discipline of philosophy has little if anything to learn from the history of the subject, but in reality the philosophical inquiries of each generation always take shape against the background of an implicit dialogue with the actual or imagined ideas of past thinkers. Many of our current debates on the relationship between thought and language bear the imprint of what the 'father of modern philosophy' said, or is supposed to have said. A basic presupposition of the 'Cartesian' metaphysical framework as normally interpreted is the idea of thought as something inner, hidden, private. We start, each of us, from the inside, from our own internal reflections and cogitations, and then by a problematic and circuitous route move outwards, to the public world of communication. The validity of the private perspective of the Cartesian meditator has had a curious dual fate in the twentieth century: the majority of philosophers (from Ryle and Wittgenstein onwards) have found it, in one way or another, deeply suspect; but in our non-philosophical culture it still strikes a responsive chord. Most people, unversed in philosophy, would probably respond favourably to the words of T. S. Eliot in The Waste Land: 'we think of the key, each in his prison'. 1 Words, on this picture, are attempts to clothe the true inner core of thinking, private to each of us. 'Cartesian privacy', it seems, has not lost its popular hold. Descartes himself, however (though this is often forgotten), devoted most of his career not to metaphysics but to science. There is strong evidence, for example, that much of the famous metaphysical argument in part four of the Discourse was cobbled together, in I am most grateful to Max de Gaynesford, Hanjo Glock, and David Oderberg for stimulating discussions of an earlier version of this paper; thanks are also due to various participants at the Royal Institute of Philosophy conference on Thought and Language for a number of helpful comments. 1 'We think of the key, each in his prison/Thinking of the key, each confirms a prison...' The Waste Land [1922], part v, lines 413-14, in Eliot (1963), p. 77. 29
John Cottingham a fairly hasty fashion, and hurriedly inserted when the main scientific part of the work was ready for the press.2 And I shall be suggesting later on that we best understand Descartes if we realise that scientific inquiry is the engine that drives his philosophy, even when he invokes seemingly very metaphysical items like souls. At all events, when wearing his scientific hat, Descartes ended up with a quite different perspective on the relationship between language and thought from that which is implied by the model of the mind as a secret prison. In his scientific correspondence and in the Discourse (which was of course a preface to his first published collection of scientific essays), Descartes unequivocally advanced the claim that there is no thought without language; and in arguing for this claim he treats language, throughout, as an objective, interpersonally fixed phenomenon, subject to firm 'external' criteria for what can count as its genuine instantiations. And hence he maintains we are quite mistaken in attributing any kind of thought to non-human animals, given that they lack genuine language. The fate of this Cartesian thesis has been exactly the opposite of that which has befallen the doctrine of the privacy of the mental. This time, it is the philosophers who have tended to embrace it, while our non-philosophical culture has found it uncongenial. 'No thought without language' is a slogan which many philosophers, following the lead of Donald Davidson and others, have found highly persuasive. But the 'lay public' tends to recoil at the restriction of thought to language-users: oysters and sponges, to be sure, don't think (some weirdos might be tempted, even here, to add 'probably'); but the cat scrabbling at the door of the fridge, then rubbing itself round its owner's legs, or the dog fetching its lead and depositing it by the master's armchair, then staring mutely up into his face — these are instances which the pre-philosophical intuitions of the great majority would classify as involving thought: Felix thinks there is catfood in there, and is asking you to get it out; Fido has figured out how to get you to take him for a walk. The two Cartesian claims so far mentioned - the privacy of thought and the linguisticity of thought - hardly look compatible. Or at any rate, if thought is entirely and essentially private, then 2
Descartes wrote to a correspondent that the metaphysical section of the Discourse was 'the least worked out section in the whole book...I did not decide to include it until I had nearly completed the work and the publisher was becoming impatient' (letter to Vatier of 22 February 1638, AT I 560; CSMK 85-6). In this paper, 'AT' refers to the standard FrancoLatin edition of Descartes: Descartes (1964-76). 'CSM' and 'CSMK' refer to the standard English translation: Descartes (1985) and Descartes (1991) respectively. 30
Thought and Language in Descartes there seems to be no conclusive reason for restricting it to languageusers: for aught we know, cats and dogs might have an inner mental life. And conversely, if thought is analytically linked to linguistic competence, then language-use becomes a public criterion for the presence of thought, undermining the alleged essential privacy of the mental. We thus appear to have two pairs of opposed options, arising from the disagreements just sketched. In the first place, with regard to thought, its privacy can either be (la) upheld, 'Cartesian'style, or (lb) denied, in line with much contemporary philosophy. On view (la) Descartes was on to something important, that there is a crucial sense in which thinking is, in part at least, a subjective, private phenomenon, in principle accessible only to the thinker; and our popular culture has correctly preserved elements of this fundamental Cartesian insight. On view (lb), Descartes fathered a monumental error, the fallacy of essentially private thought, which has deeply infected our popular culture, but which it is the task of modern philosophy of mind and language to eradicate. In the second place, with regard to language, we have two opposing positions: (2a) linguisticism - Descartes was right, and our contemporary lay culture wrong; thought is indeed confined to language users; and, opposing this, (2b) a more 'generous' view, prevailing in our contemporary culture (and indeed going back to Montaigne in the renaissance), but characteristically denied by many modern philosophers: that thought extends beyond the restricted domain of genuine language users. The positions are set out for convenience in the following table: Thesis
view
'Cartesian'
prevailing view in modern philosophy
prevailing 'l'lay' view
la
privacy of thought
y
n
y
lb
publicity of thought
n
y
n
2a
linguisticity of thought
y
y
n
2b
thought extends beyond n language use
n
y
The strategy of this paper is as follows. Firstly, with respect to the first two rows in the table, I want to side with the 'philosophical' rather than the lay position, defending (lb). But I propose to offer a reinterpretation of the actual Descartes according to which he is not, in fact, lumbered with (la). In other words, I will try to 31
John (Nottingham show that the label 'Cartesian privacy', despite its widespread use, is a highly misleading one, since it does not accurately correspond to what the actual historical Descartes held. Secondly, with respect to the second two rows, I shall end up suggesting something closer to the 'lay' as opposed to the 'philosophical' position: thought may indeed extend beyond the domain of language. This will involve taking issue with some of the actual Descartes' conclusions; but, paradoxically, I shall maintain that the arguments he deploys on this subject would, if developed more fully, have permitted him to take a more generous (and less counter-intuitive) view of the realm of thought. The Myth of Cartesian Privacy Why has Descartes' name been so automatically associated with the idea of the privacy of thought? The word 'private' is never used by Descartes in any of his discussions of mental phenomena. And to his seventeenth-century contemporaries it would, I think, have seemed quite bizarre to interpret him as some kind of introverted, subjectivist metaphysician, preoccupied with the supposedly internal domain of the mental. In the vast majority of his writings, Descartes approaches the phenomena of human experience from the perspective of a natural scientist, searching for perfectly objective 'external' schemas of explanation. In Descartes' early work on the nature of man, what strikes the reader is not the use made of a subjectively accessed 'soul', but the extent to which appeals to the soul are declared to be redundant. The Traite de I'homme, composed in the early 1630s, advances a wholesale mechanistic reductionism,3 whereby a vast range of human activities are ascribed to the operations of a self-moving machine which, like a 'clock or an artificial fountain or mill' has the power to operate purely in accordance with its own internal principles, depending solely on the disposition of the relevant organs (AT XI 120; CSM I 99). Descartes insists that it is not necessary to posit any 'sensitive or vegetative soul' (of the kind favoured by the scholastics), just as he refuses to acknowledge any principle of life apart from the internal fire of the heart - a fire which has the same nature as the fires to be found elsewhere in inanimate objects (AT XI 202; CSM I 108). 3 Though I cannot develop this here, Descartes' mechanism, in my view, should be construed as reductionist, but not, as is sometimes supposed, eliminativist; that is, he offers scientific explanations of phenomena such as animal sensation, but does not need to deny to animals the ascription of, for example, pain. See further Cottingham (1978a). 32
Thought and Language in Descartes The list of functions to be explained in this way, without any reference to soul, is remarkably comprehensive: digestion of food, the beating of the heart and arteries, the nourishment and growth of the limbs, respiration, waking and sleeping, the reception by the external sense organs of light, sounds, smells, tastes, heat and other such qualities, the imprinting of ideas of these qualities in the organ of the 'common' sense and the imagination, the retention or stamping of these ideas in the memory, the internal movements of the appetites and passions, and finally the external movements of all the limbs which aptly follow both the actions and objects presented to the senses and also the passions and impressions found in the memory, (ibid.) To those brought up on standard conceptions of 'Cartesian dualism' it will be something of a surprise to see how far the list extends beyond what we might think of as 'pure physiology'. What is declared to be capable of explanation without reference to any inner 'soul' is not just functions belonging to the autonomic nervous system such as respiration and heartbeat, but - on the face of it at least — 'psychological' functions like sense-perception and memory, passions and sensations like fear and hunger, and even, apparently, voluntary actions such as running. When a sheep sees a wolf and runs away, Descartes was later incredulously asked, are we really supposed to believe that this can occur in the absence of any kind of 'sensitive soul'? His answer was unequivocal: yes. And he went on to insist that, in the case of humans too, a mechanistic explanation was quite sufficient to explain even such waking actions as walking and singing, when they occur 'without the mind attending to them'. 4 In his first published psychological essay, La Dioptrique (1637), Descartes provides an explanation of vision which, despite what the modern use of the label 'Cartesian' might lead us to expect, systematically attacks the picture of a private soul contemplating images. 'We must take care not to assume, as some philosophers do, that in order to have sensory awareness the soul must contemplate certain images...resembling external objects'.5 Instead, vision is explicated in terms of a scientific model which can properly be 4
'animo non advertente' (Fourth Replies, AT VII 230; CSM II 161). For more on this passage, and the significance of the final qualification, see J. (Nottingham, 'Cartesian Dualism: Theology, Metaphysics and Science', in Cottingham (ed.) (1992), pp. 236ff. 5 La Dioptrique [1637], AT VI 112; CSM I 165. There are complexities in Descartes' arguments, associated with his debate with the scholastics, which cannot be examined here. For more on the details of Descartes' attack on the 'pictorial' hypothesis, see Hyman (forthcoming). 33
John Cottingham called functionalist. When the blind man finds his way around using a pair of sticks, his awareness of shapes and distances has nothing to do with the contemplation of private images; rather, it is a matter of the sticks being 'caused to moved in different ways, setting in motion the nerves in his hand and then the regions of his brain where these nerves originate' (AT VI 114; CSM I 166). What is true for the blind person is true for the sighted person, except of course that the relevant molecular motions occur in the optic nerves and
Figure 1
the brain, instead of in the nerves of the hands and the brain. You might say (again trying to foist the 'Cartesian' label on to Descartes): 'well, here he is talking merely of the physiological causes of vision, not vision itself - vision itself must be a matter of images displayed in the private domain of the mind.' But very little that Descartes says supports this kind of private theatre view. The crucial point about the blind man analogy is that it tries to move us away from the idea of the private theatre of images. Look at the information-processing system, Descartes is saying; observe the causal flow of the transmission of neural impulses, and you will see that talk of the reception of images is way off the mark - and once you have freed yourself from that misleading picture by reflecting on the blind person case, you will be ready to discard it in the sighted case as well. Admittedly, Descartes is cryptic about precisely how the relevant modes of sensory awareness come about as a result of the operation of this functional processing: coining an awkward phrase which generated decades of fruitless metaphysical theorising after his 34
Thought and Language in Descartes death, he simply says that the brain events 'occasion' the soul to have sensory awareness, by a divine or 'natural' institution. 6 What exactly Descartes meant by this is a highly complex question: interpreted one way it points forward to a Leibniz-style conception of pre-established harmony between body and soul; interpreted another way, it takes us forward to the Malebranchian notion of divine monocausality. Descartes probably means neither, and is possibly doing no more than signalling that the relationship between the brain events and the mental events cannot be construed in terms of efficient causation - and in fairness to Descartes, we have to admit that this much, at least, is true. What Descartes is quite clear about is that sensory awareness is not a matter of images being transmitted to the soul: even when he maintains, for physiological reasons, that images are formed (on the retina, or the pineal gland), he explicitly rejects the homunculus model which would lead to an infinite regress of further sets of inner 'eyes'. 7 The brain (or the pineal gland) is emphatically not construed as a 'fax machine to the soul', in Daniel Dennett's delightfully scathing phrase;8 that may be a just lampoon of the picture we have come to regard as 'Cartesian', but it is not the picture offered by Descartes himself. Yet whatever the correct interpretation of Descartes' early scientific work may be, you might feel that once we come to the metaphysics of his mature period we are unavoidably confronted with a truly 'Cartesian' picture of thought as an essentially private, firstperson-singular process. But the reality is by no means so straightforward. In the book which Descartes offered as the canonical presentation of his metaphysical views, and which he tried to get adopted as a university textbook (I refer to the Principles of Philosophy, 1644), virtually the entire metaphysical argument is presented in the public, intersubjective domain. The narrative (though hardly any commentators, so far as I know, have even remarked on this) is conducted not in the first person singular, but the first person plural. 'Quoniam infantes nati sumus...', says the opening sentence: 'Since we began life as infants, and made various judgements concerning the things that can be perceived by the senses before we had the full use of our reason, there are many preconceived opinions that keep 6
For the notion of 'occasioning', see AT XI 144; CSM I 103. Compare also Comments on a Certain Broadsheet [1648], AT VIIIB 359; CSM I 304. For the idea of a 'natural' institution see Optics, AT VI 130; CSM I 167. 7 'We must not think that it is by means of resemblance [between image and external object] that the picture causes our sensory awareness of these objects, as if there were yet other eyes within our brain by means of which we could perceive it' {Optics, Discourse 6, AT VI 130; CSM I 167). 8 Dennett (1991), p. 106. 35
John Cottingham us from knowledge of the truth' (Part I, article 1). Later on, to be sure, the famous Cogito argument makes its appearance, but it is introduced, again, in an interpersonal context: 'In rejecting, and even imagining to be false - everything which we can in any way doubt, it is easy for us to suppose that there is no God and no heaven, and there are no bodies, and even that we ourselves have no hands or feet, or any body at all. But we cannot for all that suppose that we, who are having such thoughts, are nothing' (article 7, emphasis added).9 Elsewhere, in the Search For Truth (composed sometime in the 1640s),10 Descartes explores the Cogito in an even more overtly interpersonal mode, that of a two-way dialogue, and the relevant verbs occur not in the first but the second person: 'You cannot deny that you have doubts', says Eudoxus to Polyander; 'rather it is certain you have them, so certain in fact that you cannot doubt your doubting. Therefore it is also true that you who are doubting exist' (AT X 515; CSM II 410). This is not, moreover, just a matter of how Descartes conjugates his verbs (though that, to be sure, is highly significant). For if Descartes really considered thought to be an essentially private domain, directly accessible only to the subject, one would surely have expected his reconstruction of the foundations of knowledge to have addressed the so-called problem of other minds: how can I be sure that others, apart from myself, really feel pain, or have genuine consciousness, in the way I do? Yet nowhere, in the entire Cartesian corpus, is that dreary stock-in-trade of modern epistemology courses even remotely touched on, much less discussed. Descartes was not, in fact, an 'epistemologist' at all, in the neurotic modern sense of one who tries to engage with that absurd philosopher's dummy, 'the sceptic'. 'No sane person', he brusquely observes in the Synopsis prefixed to the Meditations, 'has ever seriously doubted that there really is a world or that human beings have bodies' (AT VII 16; CSM II 11). The ordo essendi, the order of reality which structures his system, is from first to last an objective, divinely guaranteed order, in which all flows from the commands of the supreme being - 'the laws (to quote the Discourse) which God 9
Principles of Philosophy [Principia philosophiae, 1644], Part I, art. 7 (emphasis supplied). Later in this passage, of course, the famous dictum 'cogito ergo sum' is introduced (in the first person); but it is presented 'interpersonally', as it were, as 'the first and most certain piece of knowledge to occur to anyone who philosophises in an orderly fashion' (AT VIIIA 7; CSM I 195, emphasis supplied). 10 The date of the (unfinished) Recherche de la Verite is uncertain. It may have been written about the time of the Meditations, but may well have been composed during Descartes' ill-fated stay in Stockholm in 1649-50. 36
Thought and Language in Descartes has established in nature and of which he has implanted notions in our hearts'. 11 But what of the Meditations'? 'Solus secedo', says Descartes in the opening paragraph: 'I am here quite alone.' But withdrawing into solitude, something most of us do from time to time in order to philosophise, need not imply any radically solipsistic view of the essential privacy of thought. It is, of course, true, that the Meditations purport to follow not the ordo essendi, the objective order of reality which Descartes accepted, but rather to take the reader along an ordo inveniendi, an order of discovery, which starts from the self and proceeds on to God, and then to science and the external world. In so far as this is supposed to provide a free-standing foundation for knowledge, as countless commentators have eagerly pointed out, it doesn't work, it can't work. My purpose is not to defend Descartes on this old score, or to rehash the problems of circularity which he was caught in, and which were so ably uncovered by Arnauld and Descartes' other contemporary critics. But what is clear, even in the most apparently solipsistic mode of the First Meditation, is that despite what may be called the epistemic privacy of the scenario, Descartes does not, and cannot, subscribe to the notion of semantic privacy: his exercise could not even be formulated on the basis of the kind of private assignment of meanings which Wittgenstein famously attacks. 'Whether I am awake or asleep, two and three added together are five, and a square has no more than four sides'.12 The objective constraints of language and meaning are in operation from the outset, and become ever more prominent as the argument of the Meditations develops. The basic argumentative structure of the Meditations involves an investigation of the 'ideas' which the meditator finds within him. Now despite the frequently heard complaint that Descartes 'privatised' or 'psychologized' the mind and its objects, if we look at Descartes' own use of the term 'idea', we find instead that it has strong links with the classical and medieval usage, stemming ultimately from Plato, in which idea is a formal, not a psychological notion. In the Third Meditation, indeed, Descartes comes close to explicitly Platonic terminology, when he traces the source of our ideas to a primary idea associated with an archetype: 'although one idea may perhaps originate from another, there cannot be an infinite regress here, but eventually one must reach a primary idea, the cause of which will be like an archetype' (Third Meditation, AT 11 Discourse on the Method, part v (AT VI 41; CSM I 131). 12 The doubts subsequently raised by the introduction of the demon are, in my view, much weaker in scope than is often supposed. See Cottingham (1976), reprinted in Moyal (ed.) (1991), vol. II, 129ff. 37
John Cottingham VII 42; CSM II 29). When asked to provide a definition of the term 'idea', Descartes distinguishes an episode of thinking ('that which is within us in such a way that we are immediately aware of it') from an idea, which is 'the form of any given thought' (Second Replies, AT VII 160; CSM II 28). Considered from a psychological point of view, the nature of an idea is 'such that of itself it requires no reality except what it derives from my thought of which it is a mode, i.e. a manner or way of thinking' (Third Meditation, AT VII 41; CSM II 28). But considered from a logical point of view, it has a certain representational content. 'Considered simply as modes of thought, there is no inequality among my ideas - they all appear to come from within me. But in so far as ideas are considered as representing different things, they differ widely [in that some contain more "objective reality", or representational content, than others]' (Third Meditation, AT VII 40; CSM II 28). It may be seen from passages like this that Cartesian ideas are in some respects much more like publicly accessible concepts than private psychological items. Since the occurrent modifications of my consciousness belong to me, rather than to you, you of course cannot have my thoughts, in this sense; but it is implicit in Descartes' whole project, and the route he offers to validate the foundations of knowledge, that two people can be said to have the same idea in so far as their thoughts have a common representational content. It is significant that Descartes links the possession of an idea to the ability to use a linguistic term correctly: 'whenever I express something in words, and understand what I am saying, this very fact makes it certain that I have an idea of what is signified by the words in question' (Second Replies, AT VII 160; CSM II 113). He is thus careful to distinguish an idea, in his sense, from the scholastic notion of a phantasm in the corporeal imagination: 'it is not only the images depicted in the imagination which I call "ideas"; indeed, in so far as these images are in the corporeal imagination, that is are depicted in some part of the brain, I do not call them "ideas" at all; I call them "ideas" only in so far as they give form to the mind itself (Second Replies, AT VII 160-1; CSM II 113, emphasis added). Thus anyone who has a conception corresponding to the expressions 'God' and 'soul' must eo ipso know what is meant by the 'ideas' of God and the soul, 'namely nothing other than the conception he has' (Letter to Mersenne of July 1641, AT III 392; CSMK 185). Thomas Hobbes, perhaps wilfully, misunderstood Descartes on this point trying to assimilate Cartesian ideas to his own theory of corporeal images, or quasi-pictorial impressions - an interpretation which Descartes brusquely dismissed, in the Third Set of Replies printed with the Meditations (Third Replies, AT VII 180-1; CSM II 126-7). 38
Thought and Language in Descartes Ideas in Descartes have a publicly accessible structure; they are not dependent on the private vagaries of individual psychology, but relate to 'immutable and eternal essences which are not invented by me or dependent on my mind' (Fifth Meditation). That is why Descartes' disciple Nicolas Malebranche, extrapolating from the master, but not departing from his model in any important respect, located ideas in the divine intellect: in Malebranche's slogan we 'see all things in God'. This was regarded by some of Malebranche's contemporaries as little more than a curiosity; in Locke's dismissive phrase, 'tis is an opinion that spreads not, and is like to die of itself or at least do no great harm.' 13 But Malebranche's development of the Cartesian model is in crucial respects an accurate development of Descartes' thinking, and it has the signal merit of resisting the later tendency to privatise or psychologise all ideas.14 At all events, as far as Descartes is concerned, the semantic privatisation charge can be firmly rebutted. The reflections of the Cartesian meditator, properly understood, relate to ideas construed as objects in the public domain, objectively determined forms of thought whose structure is grounded in an independent reality, accessible to all. That reality, to be sure, is one we nowadays tend to think of as a social construct, evolving over time through the developing rules of language, whereas Descartes conceives it in more Platonic fashion as a set of immutable and eternal archetypes. But that difference, important though it is for understanding how our contemporary worldview has become more secularised, more anthropocentric, is irrelevant to the public, intersubjective status of our concepts. So far from wanting to privatise thought, Descartes confided to his friend and editor Mersenne that he had a dream of formulating a universa mathesis - a public formal language possessing the same kind of objective structure as a mathematical symbolism, whereby 'all possible thoughts might be arranged in order like the natural order of the numbers', thus enabling us to 'express all the other things which fall within the purview of the human mind'.15 If we look at the argu13
A comment made by Locke some three days before his death; cited in Jolley(1990), p. 81. 14 The seemingly bizarre theory of vision in God turns out, on Nicholas Jolley's suggestive interpretation, to be a corollary of his firm separation of the province of logic from that of psychology; 'to say that we directly perceive ideas in God is to say that we directly perceive items in logical space'. Jolley (1990), p. 87. 15 Letter of 20 November 1629, AT I 81; CSMK 12. The phrase universa mathesis comes in Descartes' early treatise on method, the Rules for the Direction of our Native Intelligence [Regulae ad directionem ingenii, c.
1628], Rule Four (AT X 378; CSM I 19). 39
John Cottingham mentative structure of the Meditations, it is clear that the meditator's thoughts are supposed by Descartes to conform to just such objective patterns of meaning - they cannot on pain of the incoherence of the whole project, belong in some kind of private semantic domain (that very phrase, for good Wittgensteinian reasons, can be seen to be a contradiction in terms). Cartesian privacy, at least in respect of Descartes' views on the structure of thought, is a myth. The Relation of Thought to Language: Philosophical Analysis and Scientific Evidence I now turn to some of Descartes' scientifically driven work on the topics of thought and language, beginning with his views on the states and processes attributable to animals. Descartes had an enduring scientific interest in non-human animals, dating from the days when he lived in Kalverstraat (the butchers' quarter) in Amsterdam, and regularly ordered veal carcasses to be brought to his house for dissection.16 'I cannot share the opinion of Montaigne and others', he subsequently wrote, 'who attribute thought to animals': I know that animals do many things better than we do, but this does not surprise me. It can even be used to prove that they act naturally and mechanically, like a clock which tells the time better than our judgement does. Doubtless when the swallows come in spring, they operate like clocks. The actions of honeybees are of the same nature; so also is the discipline of cranes in flight, and of apes in fighting, if it is true that they keep discipline. Their instinct to bury their dead is no stranger than that of dogs and cats who scratch the earth for the purpose of burying their excrement; they hardly ever actually bury it, which shows that they act only by instinct and without thinking.17 Later in this passage (from a letter to the Marquis of Newcastle) Descartes considers a possible argument of his opponents: since the 16
Later on, working on the problem of the circulation of the blood, Descartes performed vivisections on dogs and rabbits, any possible ethical scruples allayed by his belief that the squeaks of the victims were of no more significance that the fact that a church organ makes a certain sound when you press one of the keys (cf. AT IX 165; CSM I 104). Descartes' stance on whether non-human animals have feelings or passions (as opposed to thoughts) is, however, complex and problematic; see Cottingham (1978a), reprinted in Moyal (ed.) (1991), vol IV, pp. 323ff. " Letter to Newcastle of 23 November 1646 (AT IV 573, 575-6; CSMK 302, 304). For Montaigne's views, see his 'Apology for Raymond Sebond', Essais [1580], II, 12; translated in Montaigne (1987). 40
Thought and Language in Descartes bodies of animals are so like ours, might not some elementary kind of thought be 'attached to these organs' ? Descartes rejects this argument on the grounds that if similarity of organs is a reason for allowing the possible presence of thought, then it would be a reason for assigning a rational soul to all animals: 'there is no reason to believe it of some animals without believing it of all, and many of them, such as oysters and sponges, are too imperfect for this to be credible'. This seems a bit of a muddle; Descartes' opponents could surely just deny that oysters and sponges are anatomically similar in the relevant respects, and thus resist the implied reductio. In fact Descartes is being a little mischievous, or at least brusque, here, since he was well aware of the difference between creatures with pineal glands, and those without; the latter operate merely in a reflex fashion, via the flow of the so-called animal spirits, while the former, as we have seen from the earlier quotation about animal memory, sensation and purposive activity, are capable of far more complex kinds of behaviour. So he is being distinctly cavalier in lumbering his opponents with the position that once you attribute thought to a dog you may be committed to attributing it to an oyster. However that may be, his denial of genuine thought, that is intellection,18 to all non-human animals, is clear (as I have argued elsewhere, his position with respect to sensations and passions is a great deal more complicated). As far as anatomical arguments are concerned, Descartes of course notoriously rejected the whole idea of a physical organ of thought, and this may partly explain his summary treatment of the Montaigne lobby in this passage. But to understand the rationale for his position, we need to look at a far more powerful argument he has in his armoury, one which hinges not on anatomical similarities and differences, but on functional output. The stronger argument is developed in a letter to another Englishman, the Cambridge Platonist Henry More, written three years later: In my opinion the main reason for holding that animals lack thought is the following. Within a single species some of them are more perfect than others, as humans are too. This can be seen in horses and dogs, some of which learn what they are taught much better than others; and all animals easily communicate to us, by voice or bodily movement, their natural impulses of anger, fear, hunger and so on. Yet in spite of all these facts, it has never been observed that any brute animal has attained the perfection of 18
For Descartes' 'narrow' use of the term 'thought' (pensee, cogitatio), see Cottingham (1978b), reprinted in Moyal (ed.) (1991), vol. II, 288ff. 41
John Cottingham using real speech, that is to say, of indicating by word or sign something relating to thought alone and not to natural impulse. Such speech is the only sure sign of thought hidden within a body. All
human beings use it, however stupid and insane they may be, even though they may have no tongue and organs of voice; but no animals do. Consequently this can be taken as a real specific difference between humans and animals. (Letter to More of 5 February 1649, AT V 278; CSMK 366) Why is speech so important? Descartes' reasoning strikingly anticipates the line taken in our own time by Noam Chomsky.19 It hinges on the observation that a machine, or a bete machine, is essentially a stimulus-response device. You may be able to train a magpie to say 'bonjour', Descartes observed elsewhere, but each word will be a fixed response to an external stimulus causing a given change in the nervous system.20 As Descartes put it in the Discourse: We can certainly conceive of a machine so constructed that it utters words...corresponding to...a change in its organs (e.g. if you touch it in one spot it asks what you want of it, and if you touch it in another spot it cries out that you are hurting it). But it is not conceivable that such a machine should produce arrangements of words so as to give an appropriately meaningful answer (pour re'pondre au sens) to whatever is said in its presence, as even the dullest of men can do. (AT VI 56; CSM I 140) In short, the human language-user has the capacity to respond appropriately to an indefinite range of situations, and this capacity seems generically distinct from anything that could be generated by a 'look-up tree' or finite table correlating inputs with outputs. Descartes' language argument, properly construed, is not an argument about how linguistic competence provides plausible evidence for the occurrence of an essentially private process. It is based on an analysis of what it is to think - namely that it involves an 19
Cf. Chomsky (1968). Compare Descartes: 'Since reason is a universal instrument which can be used in all kinds of situations, whereas [bodily] organs need some particular disposition for each particular action, it is morally impossible for a machine to have enough different organs to make it act in all the contingencies of life in the way in which our reason makes us act' (Discourse, part V, AT VI 57; CSM I 140). By 'morally impossible' Descartes means impossible for all practical purposes. 20
Letter to Newcastle of 23 November 1646, AT IV 574; CSMK 303. Descartes does, however, add the curious comment that the word so produced will be the 'expression of one of the bird's passions, e.g. the hope of eating'. For Descartes' not entirely consistent stance on animal passions, see above, note 16, and Cottingham (1993), s.v. 'animals'. 42
Thought and Language in Descartes indefinitely rich, stimulus-free capacity to respond to 'all the contingencies of life': it is based, in other words, on a gap between input and output, as observed 'from the outside'. Genuinely linguistic competence (vera loquela), the creation every day of new sentences, unlinked to specific behavioural stimuli, is something which could not in principle be produced by a purely mechanical automaton. There is nothing whatever in these arguments that invokes the picture of thought as a mysterious inner process accessible only to the subject. But what of the phrase from the letter quoted a moment ago: 'speech is the only sure sign of thought hidden in a body'? 'Loquela est unicum cogitationis in corpore latentis signum certum.' This could be read as implying a commitment to one form of 'Cartesian privacy' - the idea of thought as an essentially hidden process, of which speech is merely the outward sign. Whether this is right hinges on the exact meaning of the Latin participial adjective 'hidden' (latens). We may be tempted to read it as indicating something in principle unobservable, and something accessible only to the subject. But that is by no means the only possible meaning of the term. The compelling idea of the scientist as an investigator of something 'hidden' (latens), had been presented some thirty years earlier by Francis Bacon (whom we know Descartes had read),21 in the Novum Organum (1620). The true 'work and aim of human knowledge', wrote Bacon, is to investigate and uncover the latens schematismus — to search for the 'hidden schematisms', or micro-processes and configurations of matter, which are responsible for the behaviour of all observed physical phenomena.22 'Latent' here does not imply anything occult or mysterious: the explanatory structures uncovered by empirical science are 'hidden' only in the sense that they are not readily observable at the macro level. What the scientist does, starting from careful observation of the phenomena, is to theorise about the possible fine structures which might be responsible for what is observed, with the eventual goal of bringing what is initially hidden into the light. This, of course, is precisely the aim of Descartes in all his scientific work. To explain magnetism, fire, the beating of the heart, growth, respiration and (as we have seen) even vision, and purposive behaviour like the sheep's running away from the wolf, Descartes offered explanations in terms of the minute interactions of particles too small to be observed with the naked eye. The incredibly fast-moving jostlings of subtle matter, he reasoned, are the unobserved causes responsible for what we call light; the whirrings of tiny screw-shaped molecules are responsible for mag21 Cf. letter to Mersenne of 10 May 1632, AT I 251; CSMK 38. 22
Francis Bacon Novum Organum [1620]. Book II, §1, in Bacon (1905), p. 302. 43
John Cottingham netism; the pneumatic pressures of animal spirits are responsible for reflex behaviour in humans and animals; the events in the nervous system and brain are responsible for more purposive action. In the case of reasoning and language, to be sure, Descartes saw no way forward for the physical scientist. He did at least consider (in the Discourse) the possibility of a corporeal 'instrument' (instrument) of reason. But what made a physical realisation of such an instrument hard for him to envisage was, at least partly, a matter of number and size - of how many structures of the appropriate kind could be packed into a given part of the body. Descartes' anatomical dissections of the brain and nervous system had revealed the harmonious operation of tiny structures which he believed had considerable explanatory power. But they had also, so he believed, established the essential underlying simplicity of those structures. Everything going on in the heart and brain, the nerves and muscles, and the 'animal spirits' worked by means of elementary 'push and pull operations', not in principle any different from the simple operations of cogs and levers and pumps and whirlpools that could be readily inspected in the ordinary macro-world of 'medium-sized hardware'. Everything happened according to the laws of mechanics, the same laws that operated always and everywhere in the universe (AT VI 54; CSM I 139). And Descartes could not envisage the brain or nervous system as being capable of accommodating enough mechanisms of the requisite simplicity to generate sufficiently complex responses to constitute genuine thought or linguistic behaviour. So he was driven to suppose that the hidden schematism responsible for thought was something mysterious and incorporeal (a non-explanation, of course, since there is no reason to suppose the problem of complexity is somehow solved simply by positing an immaterial substance). Had he been alive today, as I have elsewhere argued, he might well have abandoned his dualistic view of the mind. For the language argument in the Discourse hinges on the practical impossibility of a physical mechanism possessing a sufficiently large number of different parts (assez de divers organes) to facilitate the indefinite range of human responses to 'all the contingencies of life' (AT VI 57; CSM I 140). Could such an argument survive the modern discovery of the staggering structural richness of the microstructure of the cerebral cortex, composed as we now know, of over ten billion neural connections? Well, Descartes might still have maintained that a purely physical structure could not generate the relevant kind of plasticity and innovativeness necessary for genuine linguistic output; but his view of what matter might or might not do was coloured by a very crude conception of material stuff as pure geometrical extension, so there must be an element of speculation in trying to transfer his 44
Thought and Language in Descartes arguments to the context of our far richer contemporary physics. However that may be, Descartes' general quest, I am arguing, was for latent (in Bacon's benign sense) structure capable of explaining all behavioural phenomena; unable, in the case of thought, to conceive of a physical structure capable of doing the job, he was driven to posit a non-physical entity - the rational soul. But there may seem to be a major obstacle to this line of interpretation: Descartes' rational soul, the res cogitans responsible for thought, is surely an entity which the reasoning of the Meditations shows to be entirely hidden in a more suspect sense, accessible only from the inside, through the subjective reflections of the individual meditator. And does not this take us right back to the idea of a 'latent' soul, not in the benign Baconian sense, but in the more suspect sense familiar from modern attacks on the mind as an essentially private theatre? Well, the term 'thought' is unquestionably used by Descartes to 'include everything that is within us in such a way that we are immediately aware of it' (Second Replies; note again, by the way, the first person plural — 'us'). But this encapsulates a relatively harmless truth: that when we are conscious we are immediately aware of thoughts, feelings and sensations - something not many even of the most materialistic philosophers have been tempted to deny. You may, if you wish, describe this in terms of each individual's immediate transparent access to the contents of his or her mind; but even here, though I cannot develop this here, there are complications. Just to digress very briefly on this, when we look at Descartes' later work on ethics and psychology, he departs significantly from the official dualism of the Meditations, and suggests that a proper understanding of our human nature involves recognition of the extent to which we are not just angelic minds inhabiting bodily mechanisms, but what he calls 'genuine human beings' (vert homines)23 — creatures whose deepest and strongest feelings are intimately tied up with structures and events that are concealed from us as 'thinking beings'. This, moreover, is not just a matter of causes, but of the very character of those feelings. Descartes, the very thinker who is so often accused of having a nai've theory of the perfect transparency of the mind,24 is actually telling us in his later writ23
See letter to Regius of January 1642, A T I I I 493; C S M K 206. It is fairly common, for example, for commentators on the Freudian notion of the unconscious mind to say that Freud broke from the Cartesian thesis that mentality and consciousness are equivalent; cf. Gardner (1993), p. 207. Margaret Wilson attributes to Descartes 'the doctrine of epistemological transparency of thought or mind' (1978, p. 50), though she does go on to discuss certain aspects of Descartes' philosophy of mind which are in potential conflict with that doctrine (cf. p. 164). 24
45
John (Nottingham ings that our emotional life as embodied creatures, as human beings, is subject to a serious and pervasive opacity. Cartesian transparency, rather like Cartesian privacy, is a label that embodies very great oversimplifications.25 But even if we ignore Descartes' complicated theory of the human being, and restrict ourselves to his metaphysical conception of the pure mind, the res cogitans, the fact that our thoughts are transparently presented to consciousness does not entail that we have some kind of privileged access to our nature as thinking things. Admittedly Descartes, in a series of very shaky arguments in the Second Meditation, claims that mind is better known than body. But Malebranche, again seeing the implications of the master's work more clearly than perhaps Descartes always grasped them, astutely pointed out that these arguments only established the existence of thinking, not its underlying nature. 'Je ne suis que tenebres a moi meme': 'to myself I am but darkness, and my own substance seems something which is beyond my understanding'. 26 Yet even in Descartes, there is a realisation that much of our mental activity involves far more than is transparently accessible to consciousness. He believed, for example, that the soul always thinks, even in sleep - yet had to admit that the evidence for this, in terms of subjective awareness, is very slight.27 But the doctrine of continuous mental activity hints at the powerful idea, never properly developed by Descartes himself, of latent structures and processes whose operation qualifies as genuine thinking, even though it is not fully accessible to the subject. How often, on waking up in the morning, does the scientist find that a tightly knotted puzzle has begun to unravel, or the novelist that the 'blocked' final chapter has begun to take shape? Thinking, genuine creative thinking - it is hard to see how this could be reduced to mere mechanical crunching - surely goes on beneath the surface of what we consciously monitor. Even if we confine ourselves to what happens when we are awake, it is clear that the thoughts traced out by our conscious intellect are 'produced' by a staggeringly complex process of whose workings, as we ponder and deliberate, we are largely unaware. Descartes himself, though he never developed the point, implicitly recognised that the simple nature he called a res cogitans was actually far too thin and meagre a conception to have genuine explanatory power in explaining what we are. In the Third Meditation, Descartes recognises the 25
For more on this issue, see Cottingham (1996), pp. 200ff., and Cottingham (forthcoming). 26 Meditations chretiennes et metaphysiques [1683], ix, 15; in Malebranche (1959-66). See further Cottingham (1988), pp. 154-5 and p. 220, note 67. 27 Cf. Conversation with Burman [1648], AT V 150; CSMK 336. 46
Thought and Language in Descartes utter weakness and dependency of the thinking ego: the Cartesian meditator, qua thinking thing, is forced to acknowledge that it depends on a power far greater than itself, so much so that it could not even continue to exist from moment to moment without being supported by the unseen substance that sustains and preserves it (Third Meditation, AT VII 49; CSM II 43). Descartes, of course, was referring to the supreme power of God, without which everything would slip out of existence. But there is a secular analogue. The rational thinking self, as it ponders and deliberates from moment to moment is, as Descartes saw, something extraordinarily thin and fleeting - a series of isolated moments of cogitation: 'I am, I exist, that is certain. But for how long? For as long as I am thinking. For it could be that were I totally to cease from thinking, I should totally cease to exist' (Second Meditation, AT VII 27; CSM II 18). The fragility and momentariness of the Cogito requires divine conservation to give it an enduring identity. Translating this into a less metaphysical and more biological idiom, we might say nowadays that beneath the thinking episodes of which we are immediately aware, sustaining their continuity and grounding their identity, is an awesomely powerful nexus of psycho-physical structures and processes; it is these that provide us with our enduring sense of self, with the memories and patterns of recognition that enable us to locate ourselves in a story with a history, with the imaginative and creative drives that urge us to recover our past, and take us forward to the future - in short with everything that makes us truly ourselves as thinking things. Notice that this continuing activity is the very reverse of a private, immediately present, subjective process. For, in the first place, much of it (as Descartes sometimes glimpsed) is, in large part, not present to inner awareness at all; and in the second place the criterion for its occurrence is something public and intersubjectively accessible, the resulting, publicly expressed, output. To conclude: I hope it is clear that I have not been attempting in this paper to defend Descartes' theory of the mind, or even to suggest he has a consistent or finished theory, but rather to suggest that if we read him carefully we can see that he is groping towards a host of insights that are ignored if we focus exclusively on the official metaphysics that is the stock-in-trade of introduction to philosophy courses. What emerges about the relation between language and thought? Descartes, as we have seen, came close to treating language as criterial for thought - as necessary and sufficient for its occurrence. He seems clearly correct in arguing that it is sufficient. 'Loquela cogitationis certum signum': the presence of genuine language is a sure sign of the presence of thought (perhaps, indeed, 47
John Cottingham this comes near to being a tautology). But is it necessary? If I am right, and Descartes' more considered stance on these matters takes us away from the notion of thought as a hidden, occult process, inaccessible from the outside, then he ought to have allowed (as indeed he did allow in many of his writings) that whether cats and dogs can be said to think must be a function of the complexity of the output they produce - a function of whether such output can be reduced to a finitely determined, mechanical pattern of response. Whether cats and dogs do qualify as thinking in this sense remains far from settled. The examples I mentioned at the outset - those of the cat demanding its feed, or the dog requesting a walk - seem on reflection to involve an output which is far too thin, in relation to the relevant stimuli, to decide the matter: they are quite compatible with the judgement that attributing genuine thought is merely sentimental anthropomorphising on the part of the pet-owners. It might be possible to devise more subtle experiments, perhaps examples of the kind of inferential, hypothesis-testing behaviour allegedly observed in the case of Chrysippus' dog (finding no scent in path A, and then immediately setting off on path B) - experiments which (though they would have to be more complex than the Chrysippus example) might uncover the kind of output warranting the ascription of at least rudimentary thought. But this is an empirical issue, as the general thrust of Descartes' scientific arguments on this question unmistakably allow. There are other cases nearer home, however, which seem clearly to show that linguistic output is not necessary for thinking. Consider the composer, trying out something on the piano, then flinging himself into an armchair, brow furrowed, his whole demeanour exuding intense concentration. After five minutes he rushes to the piano again, and modifies the melodic or harmonic sequence. I cannot imagine anyone but a philosopher denying that thinking is going on during such intervals. And it seems, moreover, to be thinking that operates in a manner entirely free from linguistic activity, from any kind of propositional conceptualising. Perhaps a musicologist could give linguistic expression to what the composer is trying to do, or to 'say' (to use a dubious metaphor), but such musicological expressions, though themselves instances of thought, seem at several removes from the actual musical thinking that is going on. And plainly someone could be a brilliant composer while being musicologically speaking utterly inarticulate or illiterate (compare some of the great early jazz composers). In general terms, human composers are, of course, always language-users, and it may be contingently true that no one could be a composer unless they were an active participant in a genuine language-using community. 48
Thought and Language in Descartes But it seems logically possible to imagine inhabitants of a distant planet, sustained perhaps by an entirely benign and superabundantly nutritious environment, who lacked linguistic competence, in the Descartes/Chomsky sense, yet produced musical compositions which we immediately recognised as far superior to those of Mozart or Wagner. I suggest it could not reasonably be denied of such beings that they engaged in highly complex forms of thinking. The truth of the 'no thought without language' slogan thus remains open; but those, like Descartes, who assert it are on to something vitally important, that thinking is not something in principle inaccessible from the outside, but is essentially bound up with complexity of information-processing, and signalled by complexity of output when measured against what is inputted. Our modern philosophical and scientific inquiries into the nature of thought clearly still have a long way to go before we develop a conceptual and empirical schema adequate to understanding, much less explaining, what is going on when thought occurs. But if we are to make progress in devising such a schema, we would do well to leave behind the 'Cartesian' paradigm, and be ready to learn more from the non-Cartesian legacy of Rene Descartes.
References Bacon, F. 1905. The Philosophical Works of Francis Bacon, ed. J. M. Robertson (reprinted from the translation of Ellis and Spedding). London: Routledge Chomsky, N. 1968. Language and Mind. New York: Harcourt, Brace & World Cottingham, J. 1976. 'The Role of the Malignant Demon', Studia Leibnitiana 8, 257-64 1978a. 'A Brute to the Brutes? Descartes' Treatment of Animals', Philosophy 53, 551-59 1978b. 'Descartes on Thought', The Philosophical Quarterly 28, 208-14 1988. The Rationalists. Oxford University Press 1993. A Descartes Dictionary. Oxford: Blackwell 1996. 'Cartesian Ethics: Reason and the Passions', Revue Internationale de Philosophic 196, 193-298 • (Forthcoming). Philosophy and the Good Life: Reason and the Passions in Greek, Cartesian and Psychoanalytic Ethics. Cambridge University Press Cottingham, J. (ed.) 1992. The Cambridge Companion to Descartes. Cambridge University Press Dennett, D. C. 1991. Consciousness Explained. Boston: Little, Brown Descartes, R. 1964—76. CEuvres de Descartes, 12 volumes, ed. C. Adam and 49
John Cottingham P. Tannery, revised edition. Paris: Vrin/CNRS. (Referred to here as 'AT') 1985. The Philosophical Writings of Descartes, Volumes I and II, ed. J. Cottingham, R. Stoothoff and D. Murdoch. Cambridge University Press. (Referred to here as 'CSM') 1991. The Philosophical Writings of Descartes, Volume HI, The Correspondence, ed. J. Cottingham, R. Stoothoff, D. Murdoch and A. Kenny. Cambridge University Press. (Referred to here as 'CSMK') Eliot, T. S. 1963. Collected Poems 1909-1935. London: Faber & Faber Gardner, S. 1993. Irrationality and the Philosophy of Psychoanalysis. Cambridge University Press Hyman, J. (Forthcoming). 'Art and Illusion in Descartes' Optics'', unpublished. Jolley, N. 1990. The Light of the Soul. Oxford: Clarendon Press Malebranche, N. 1959-66. CEuvres Completes, ed. A. Robinet. Paris: Vrin Montaigne, M. de. 1987. An Apology for Raymond Sebond, trans. M. A. Screech. Harmondsworth: Penguin Moyal, G. (ed.). 1991. Descartes: Critical Assessments. 4 vols. London: Routledge Wilson, M. 1978. Descartes. London: Routledge
50
Words and Pictures JOHN HYMAN Pictures have always played a prominent role in philosophical speculation about the mind, but the concept of a picture has itself been the object of philosophical scrutiny only intermittently. As a matter of fact, it was studied most intensively in the course of a theological controversy in the Eastern Roman Empire, during the eighth century - which is a sufficient indication of its marginal place in the history of philosophy. Perhaps this is because pictures have never produced in us the theoretical paralysis which Augustine famously associated with time, but have on the contrary generally seemed too unproblematic to deserve much time from philosophers. Even today, after several decades of accumulating theory, philosophers with no stake in the matter are often impervious to its charm. I feel some sympathy for this attitude, because the task of explaining the nature of depiction is, I believe, one which calls for the refinement rather than refutation of our first thoughts about it. But a precise understanding of depiction is both a necessary prolegomenon to a significant part of aesthetics, and a useful prophylactic against confusion in the theory of the imagination. Besides, there is also the pleasure of the chase, which J. L. Austin nonchalantly appealed to many years before the Research Assessment Exercise was inaugurated. Introduction Although abstract paintings are sometimes called pictures, I shall follow the restrictive convention which stipulates that every picture depicts, is a picture of, something or other. A picture is a kind of artefact. In general, kinds of artefact clocks, lamps, knives, chairs - are distinguished by their function. A clock is a device for keeping time, a lamp for shedding light, and so on. Of course clocks and lamps can serve many different purposes. For example, a clock can be used to decide whether a record has been beaten and a lamp can be used to attract insects; and either can be used as a door-stop. But a clock or a lamp can be used for a certain purpose because it is a clock or a lamp if, and only if, it serves this purpose by keeping time or shedding light. Hence, these are their definitive functions. In many cases an artefact's function will dictate the kinds of material it can or cannot be made of, and its form or shape. For example, a knife cannot be spherical or made of sponge, 51
John Hyman and a sail cannot be made of chicken-wire. There are also many kinds of artefact which are distinguished by form or size, distinctions which may correspond to differences in function, but need not do so - for example, kinds of spear or sword: halberds, pikes and bills, cutlasses, sabres and scimitars. Finally, simple machines are distinguished not only in terms of their function, but also in terms of their manner of operation - not only what they do, but also how they do it. Pulleys and capstans, for example, are both machines for lifting weights, but they work differently. A picture is not a machine; but the definition of a picture, like the definition of a pulley, will need to explain what a picture does and how it does it. The philosophical literature contains answers to both questions. Answers to the first - What do pictures do? - include the following: They show us things. They give us the illusion of seeing things. They make us think of things. They stand for or denote things. Historically, the most common answer to the second question How do they do it? - has been: by means of resemblance. However, in Languages of Art, Nelson Goodman rejects this answer, and argues instead that pictures depict by means of semantic and syntactic rules. Languages of Art deservedly remains a focus of debate three decades after its publication, partly because of its rhetorical panache, but principally because its arguments are ingenious and its conclusions original, not to say perverse. It may be helpful to lay out some representative answers to the two questions in a table, in order to display the route which led to Languages of Art: What do pictures do?
How do they do it?
Beardsley
A picture 'shows the latest model Chevrolet, a plaid shirt...something that exists or might exist outside the picture frame'. (1981, p. 267)
By means of a close resemblance in shape and colour; for in order to depict a Y, a picture must must contain an area 'that is more similar to the visual appearance of Fs than to objects of any other class', (ibid., p. 270)
Hospers
A picture is a symbol and therefore refers to or stands for the objects it represents. (1946, p. 29)
By means of a 'close natural relationship, namely likeness or resemblance, between symbol and thing symbolized', (ibid., p. 40)
Goodman
A picture stands for or denotes its subject. (1981, p. 5)
By means of a system of semantic and syntactic rules.
52
Words and Pictures The table is not arranged chronologically. Beardsley's Aesthetics was first published about a decade after Hospers' Meaning and Truth in the Arts. But it conveys a schematic picture of the relationship between the three theories more effectively than a chronological arrangement would. For as we can see, Beardsley and Hospers disagree about the 'what' of pictures, but agree about the 'how'; whereas Hospers and Goodman agree about the 'what', but disagree about the 'how'. In order to see which way philosophy has moved since Languages of Art, it is best to take a long step back, and to consider how Descartes fits into the picture: Descartes A picture 'enables the soul to have sensory perceptions of the various qualities of the objects to which it corresponds'. (1985, p. 166)
By the operation of mechanisms which physiology can investigate,
Descartes' idea is that the function of a picture is to produce psychological episodes of a certain sort, namely, 'sensory perceptions of the various qualities' of the things it depicts. So instead of supposing that the distinctive nature of the experience of looking at a picture - what makes this experience different from the experience of looking at visible things of other kinds - is simply a function of the distinctive nature of pictures themselves, he makes the opposite supposition: that the distinctive nature of pictures should be understood in terms of the distinctive character of the experience of looking at them. And what is distinctive about this experience is that it is not, or not exclusively, a sensory perception of the various qualities of the picture itself, but of the forest or battle or storm that the picture represents. A picture is therefore something whose essential nature is disguise: a flat pattern of ink or pigment, contrived with the express purpose of frustrating the desire to see it simply for what it is. This is, in effect, an illusion theory of pictures: the function of a picture, in Descartes' view, is to produce an illusion. Today almost everyone with an opinion agrees that the illusion theory is not viable; but the general approach that Descartes pioneered is more popular today than any other, partly because of the influence of Ernst Gombrich's Art and Illusion, which has done even more than Languages of Art to make pictorial representation a lively topic in philosophy, and partly because Goodman won few converts, but persuaded many that the resemblance theory was irreparable. There isn't a consensus about how exactly the experience of looking at a picture should be defined, but there is widespread agreement that defining this experience is the key to the 53
John Hyman theory of pictorial representation. Paradoxically, the most faithful Cartesian in the twentieth century was Gilbert Ryle, who was therefore - in a sense - ahead of his time.1 In what follows, I shall begin by discussing the doctrine that depiction is explained by resemblance, and two main arguments against it, the first propounded by Descartes, in his treatise on optics, and the second by Goodman, in Languages of Art. Secondly, I shall examine and criticise Goodman's own theory of depiction. Finally, I shall defend a positive proposal about the nature of depiction - although it is much less than a complete answer to the theoretical problems raised by pictorial art - and examine its connection with Descartes' and Goodman's arguments and conclusions, and with the doctrine they opposed.
The Resemblance Theory The truth may be simple to express but difficult to expound; and we sometimes discover, in the course of expounding it, that the simple formula we began with is as apt to mislead as to inform. The proposition that resemblance explains how pictures depict will turn out to be of this sort; and whether we stick to it, as the best epitome of a subtle matter, or prefer to be less economical with the truth, will matter relatively little, once we are able to provide a more substantial explanation. One reason for treating the simple formula with caution is that we often believe the right thing for the wrong reasons; and what may seem to be a good reason for believing that resemblance explains depiction is actually no reason at all. What I have in mind is the fact that some pictures are good likenesses. This is how the train of thought might run: a portrait is one paradigm of a picture, and although a portrait can be used in various ways - to commemorate the dead, to negotiate or celebrate a marriage, to express allegiance or to reinforce authority - it cannot be deemed a successful portrait unless it is a good likeness, that is, unless it resembles its subject. And what is true of portraits is true of other pictures too. A successful picture of a girl looks like a girl, and a successful picture of an apple looks like an apple. These resemblances, no less than the 1
Ryle's comments about depiction occur in a chapter devoted to the imagination, and although the two philosophers make strange bedfellows, these comments mark a point of congruence between Descartes' theory of perception and Ryle's theory of imagination. See Ryle (1949), ch. 8, esp. pp. 253-4. 54
Words and Pictures resemblance between Smith and his portrait, will be obvious to any spectator who knows what girls or apples or Smith look like. If these thoughts seem plausible, it is because we find it easy to misidentify the things which resemble each other when pictures are good likenesses.2 If, for example, I am struck by a close likeness or resemblance between Renoir's Girl with a Falcon and the postman's daughter, the resemblance which strikes me is not between the postman's daughter and a canvas smeared with pigment, but between the postman's daughter and the girl in Renoir's picture. Similarly, Frans Hals's portrait of Johannes Hoornbeek resembles his portrait of Herman Langelius more closely than it resembles its subject; but the man we can see in the picture resembles Hoornbeek, or so we believe. It may be tempting to object that the man in the picture cannot be said to resemble Hoornbeek, since he is Hoornbeek; but this objection betrays an important confusion, for the phrase 'the man in the picture' can be used in two different ways: to describe a picture, what it shows us and what we see or fail to see in it; or to refer to a man whom a picture portrays. 3 Thus, 'The man in Hals's picture is dead' is false if the phrase is being used in the first way, to describe Hals's picture, but it is true if it is being used in the second way, to refer to Johannes Hoornbeek. It is normally clear in which of these two ways a phrase like 'the man in the picture' is being used. For example, 'The man in the picture has a book in his hand' normally means 'The picture depicts a man with a book in his hand', in which case the phrase 'the man in the picture' is being used in the first way; but 'The man in the picture was a professor of Theology at the University of Utrecht' means 'The picture portrays a professor of Theology at the University of Utrecht', so here the phrase 'the man in the picture' is being used in the second way. Needless to say, we can imagine circumstances in which it might be difficult to be sure how such a phrase was being used, but it would be a simple matter to find out. We can call the first way of using the phrase 'the man in the picture', the one which corresponds to the paraphrase using the verb 'depicts', 'internal', and the use of the same phrase which corresponds to the paraphrase using the verb 'portrays', 'external'. And now it should be obvious that Hals's portrait is a good likeness, not if the canvas resembles Hoornbeek, but if the man in the picture, in the internal use of the phrase, resembles him. This resemblance is 2
This and the following paragraph draw substantially on and modify Hyman(1992). 3 Cf. Anscombe (1965), p. 159. 55
John Hyman not what makes the picture a portrait of Hoornbeek; but it is what makes the portrait a good likeness. For convenience, we can say that a portrait is a good likeness if its internal subject resembles its external subject.1' In sum, we need to take care not to confuse the marks on the surface of a picture and its internal subject. The resemblance which entitles us to describe a portrait as a good likeness is not a resemblance between the man or woman portrayed and the marks on the surface of the picture, but between the man or woman portrayed and the picture's internal subject; and if an apple in a picture is unmistakably a Cox's Orange Pippin, it is the picture's internal subject, the apple we can see in it, and not the patch of canvas which depicts the apple, that must resemble other Cox's Orange Pippins more closely than it resembles Bramleys. Magritte's celebrated painting Le Trahison des Images illustrates the confusion. The artist seems to be claiming that every picture, by its very nature, is a deceiver, a false friend. But the truth is less dramatic: 'Ceci n'est pas une pipe' is true if the pronoun refers to the painting, and false if it refers to its internal subject. Whether Magritte knew it or not, the painting turns on an equivocation, not a paradox. This argument does not disprove the resemblance theory; but it removes one unsatisfactory reason for endorsing it, and it also suggests that a more explicit statement of the theory than the simple formula we began with may be useful: not resemblance tout court, but a resemblance in shape or colour between the disposition of pigments on a picture's surface and the picture's internal subject explains depiction, if the resemblance theory is right. There have been two major attacks on the resemblance theory: one by Descartes; the other by Goodman. Both make use of an analogy between pictures and words in a language. Descartes was not especially interested in the nature of depiction in its own right. He denied that resemblance explains depiction 4 The distinction between pictures that have an external subject and pictures that don't is drawn by many philosophers who write about the subject. However, it is commonly thought that the external subject of a picture must be someone or something which actually exists, and therefore cannot be a fictional character or thing. For example, Malcolm Budd, who calls pictures which have an external subject 'relational pictures' (1995, pp. 66ff.), claims that 'a [relational] picture must stand in a certain relation to an actual thing', describes relational pictures as pictures of 'real individuals' (ibid., pp. 66, 73, italics added), and argues that pictures of fictional characters - to put it more crudely than Budd does — merely pretend to be relational pictures (ibid., p. 72). The idea that the external subject of a picture cannot be a fictional character is, I believe, mistaken; but it is a difficult issue, and I shall not examine it here. 56
Words and Pictures because he believed that this doctrine was implicated in a theory of visual perception which he was bent on undermining. Nevertheless, his argument is worth examining carefully. This is what he says: in no case does an image resemble the object it represents in all respects, for otherwise there would be no distinction between the object and its image. It is enough that the image resembles its object in a few respects. Indeed the perfection of an image often depends on its not resembling its object as much as it might. You can see this in the case of engravings: consisting simply of a little ink placed here and there on a piece of paper, they represent to us forests, towns, people, and even battles and storms; and although they make us think of countless different qualities in these objects, it is only in respect of shape that there is any real resemblance. And even this resemblance is very imperfect, since engravings represent to us bodies of varying relief and depth on a surface which is entirely flat. Moreover, in accordance with the rules of perspective they often represent circles by ovals better than by other circles, squares by rhombuses better than by other squares, and similarly for other shapes. Thus it often happens that in order to be more perfect as an image and to represent an object better, an engraving ought not to resemble it. (Descartes 1985, pp. 165-6). Descartes makes three points here. The first is that an image cannot be a perfect facsimile of the things it represents; the second is that an image need only resemble what it represents in a few respects; and the third is that even this resemblance is often imperfect. Then he illustrates the second and third points with the example of an engraving. The first point has often been made. Socrates argues for it with a nice example in Plato's Cratylus (432b-c): 'An image', he says to Cratylus, 'which reproduced all the qualities of the thing imitated would no longer be an image.' And he continues as follows: Suppose that some god makes not only a representation such as a painter would make of your outward form and colour, but also creates an inward organisation like yours, having the same warmth and softness, and into this infuses motion, and soul, and mind, such as you have, and in a word copies all your qualities, and places them by you in another form. Would you say that this was Cratylus and the image of Cratylus, or that there were two Cratyluses? Although it is obvious which answer Socrates expects, it is less obvious that either is satisfactory. However, what matters more is that 57
John Hyman the general proposition about images has not been demonstrated, and seems to be false; for many pictures depict other pictures, and in such a case, part of the depicting picture will resemble the depicted picture in its 'inward organization' as well as in its 'outward form and colour', as long as the same sorts of materials are used in both cases. Perhaps we can accommodate this exception by modifying the general proposition, for it seems plausible that an entire image cannot resemble an object it represents in all respects. But since many parts of images depict objects or parts of objects, the modified proposition will not shed much light on the nature of depiction, at least until some further work is done to tease out its significance. Descartes' second point is that an image need only resemble what it represents in a few respects. Socrates' argument has a bearing on this point also, because it implies what was proposed earlier - that the few respects in which an image must resemble what it represents, if the resemblance theory is right, are 'outward form and colour'. However, it is noticeable - although not surprising, given his theory of colour - that Descartes chose to illustrate the point with the example of an engraving, ignoring colour altogether.5 Descartes' final point is that even in respect of shape the resemblance between an image and the object it represents is often imperfect. He supports this with two observations, one concerning the shapes of bodies and the other, two-dimensional shapes. First, he points out that 'engravings represent to us bodies of varying relief and depth on a surface which is entirely flat'. So, for example, the parts of an engraving which depict a cylindrical tower and its conical roof cannot possibly resemble these things in shape, because a figure on a plane cannot be cylindrical or conical. And then he adds that 'in accordance with the rules of perspective [engravings] often represent circles by ovals better than by other circles...and similarly for other shapes'. Thus, even part of an engraving which depicts something with a two-dimensional shape - say, the facade of a building or the surface of a table - will not have exactly the same shape as the thing it depicts, if the artist has used foreshortening. I think this criticism is interesting, and rather effective. Just how effective it is will become clear in the sequel. I turn now to Goodman. Goodman tries to disprove the resemblance theory by attacking a theory of vision which he believes it presupposes. Following Gombrich, he calls this theory the 'myth of the innocent eye'. It is propounded by Berkeley in his New Theory of Vision and defended by Russell in Our Knowledge of the External 5
58
On Descartes' theory of colour, see Hacker (1987), ch. 1.
Words and Pictures World: what we see immediately and directly, the visual appearances that things present to us, are always simply a patchwork of colours. We identify the things themselves and their various qualities, by interpreting this brilliant panorama in the light of experience. (As Berkeley (1975, sec. 130) put it: 'in a strict sense, I see nothing but lights and colours, and their several shades and variations...') Goodman argues that if, as the resemblance theory implies, a picture were a copy of the visual appearance of the things it depicts, an artist would need, somehow or other, to strip away the varnish of thought and interpretation which clothes his - as everyone's - perception of things, in order to record the pristine patchwork of colours which lies beneath it. But since this operation, which Goodman (1981, p. 8) caustically describes as 'purification rites' and 'methodical disinterpretation', is impossible, the resemblance theory is false. I find this argument unsatisfactory because it imputes guilt by association. The theory of perception Goodman attacks is definitely confused. It implies that perceptual judgements - such as the judgement that there is a book before your eyes, or a window to one side of you - are the results of inferences, and a faithful report of the evidence on which these inferences are based would need to be couched in a vocabulary restricted to terms which refer to shapes and colours. Hence it implies that every reasonable perceptual judgement we make arises out of a visual experience which we are able to describe faithfully in these austere terms, since the evidence for the judgement would otherwise be beyond our grasp. But we are not generally able to provide this sort of description of our visual experience; hence our experience and the judgements which arise out of it cannot be related in the way the theory implies. It is easy to miss this point, if we imagine that a visual experience itself, rather than a description of what we see, could be the data or evidence on which an inference was based. But in fact the phrase 'sense-data' is misleading. A visual experience is not the premise of an argument, which can be taken as true and argued from: hence it cannot be evidence in the required sense.6 One can of course imagine the resemblance theory being combined with this patchwork theory of vision, the myth of the innocent eye. Indeed there is no need to imagine it, because it was until recently quite common, not only among theorists but among painters too. Monet, for example, offered the following advice: 'When you go out to paint, try to forget what objects you have before you, a tree, a house, a field, or whatever. Merely think, here is a little square of blue, here an oblong of pink, here a streak of yel6
Cf. Ryle(1971), p. 346. 59
John Hyman low, and paint it just as it looks to you, the exact colour and shape, until it gives your own naive impression of the scene before you'.7 Nevertheless, Goodman fails to show that one cannot endorse the resemblance theory without implicitly accepting the myth of the innocent eye, and it is doubtful whether this can be shown. An advocate of the resemblance theory needs to say something about appearances. But why are we entitled to suppose that this confused doctrine about appearances is all that he will find to say? Since there is no obvious reason, since none is given, and since - as will be shown in the sequel - there is much else that can be said about appearances, Goodman's argument is unconvincing. The resemblance theory and the myth of the innocent eye dovetail neatly, and have often been combined, but one cannot condemn a theory because it has been seen in bad company. Goodman's Theory Now I shall examine Goodman's own theory of depiction. The philosophers Goodman reacted against were Susanne Langer, Charles Morris, John Hospers. As we have seen, they proposed a novel answer to the question of what pictures do, although they accepted the orthodox answer to the question of how they do it. Instead of producing illusions, or making us think of things, or presenting us with the appearances of things, they argued that pictures are symbols or signs which stand for or refer to the objects they represent. (Hospers (1946, p. 29) does not explain what 'standing for' is, beyond mentioning, as an example, that the word chair 'refers to or stands for...the kind of object that I am now sitting on'. Morris (1946) explains what standing for is in behaviourist terms.) If both pictures and words are signs or symbols, we shall have to distinguish between them by distinguishing between different kinds of signification. Drawing partly on Peirce's taxonomy of signs, 7
Quoted by Gage (1993), p. 209. Unfortunately, there is no record of this remark in the original French; but it seems very likely that the phrase the translator rendered as 'naive impression' was meant to refer to the colourful mosaic which was widely imagined to be the raw material of visual experience. In England, Ruskin and Fry expressed similar views. Indeed, Ruskin expressed it in the passage in The Elements of Drawing which gave Gombrich the phrase 'the innocent eye': 'The whole technical power of painting', Ruskin wrote, 'depends on our recovery of what may be called the innocence of the eye; that is to say, of a sort of childish perception of these flat stains of colour, merely as such, without consciousness of what they signify...' Quoted by Gombrich (1977), p. 250. 60
Words and Pictures Morris (1946, p. 190) described a picture as a visual and iconic sign, that is, a 'visual stimulus...which is similar in some respects to what it denotes'.8 Hospers argued that a picture is a natural symbol, whose significance does not depend, as the significance of a word does, on stipulation or convention. One of the problems with the semiotic theory is that too many different kinds of thing were said to be iconic signs: onomatopoeic words; sentences which are similar in structure to the circumstances they describe, like 'The fork is to the left of the knife' (written down) or 'The Sanctus comes before the Agnus Dei' (spoken); graphs; maps; paintings; photographs; shadows; footprints; colour samples of the sort one finds in catalogues of artists' materials; and so on. Moreover, it became clear that the terms symbol and stands for were being used too generously to retain much meaning. For example, nimbus clouds were held to stand for rain and a car coming straight towards another car was described as a symbol meaning 'stop', although these were not of course supposed to be iconic signs. But even if it was thought possible to tighten this up, and thereby toss out some of the motley things in the rag-bag of iconic signs, degrees of iconicity still seemed too crude a device to draw significant distinctions between the things that remained. Languages of Art is an avowedly heterodox product of the semiotic tradition in aesthetics. Goodman accepts that pictures are visual signs - they are signs because they stand for or refer to the objects they represent; they are visual signs because what they represent depends on how they look. But he denies categorically that pictures are natural or iconic signs; and that, he says complacently, 'adds up to open heresy' (1981, p. 230). Instead of explaining depiction in terms of a 'close natural relationship' between a picture and what it represents, Languages of Art defends an elaborate analogy between the manufacture and enjoyment of pictures and the use of language. This analogy has existed in one form or another since Plato. It received its most influential expression in the letter Gregory the Great wrote to the Bishop of Marseilles in about 600: 'For the unlettered, and in particular for the heathen, a picture takes the place of a book.' As I mentioned earlier, Descartes drew attention to it in his Optics. And of course it was implicit in the doctrine of the iconic sign. But nobody before 8
Morris also claimed that an iconic sign 'denote[s] any object which has the properties (in practice, a selection from the properties) which it itself has' (1939-40), p. 136; but this implies that if a portrait of Smith is an iconic sign which denotes Smith, then it denotes his twin brother, and that if a picture of an apple is an iconic sign, it denotes every other picture of an apple that it resembles sufficiently closely. 61
John Hyman Goodman supposed that the analogy is sufficiently close to disconnect depiction and resemblance entirely. Goodman begins with a portrait of the Duke of Wellington. The portrait, he argues, like the proper name 'Arthur Wellesley' and the definite description 'the victor at Waterloo', denotes the man it portrays. Like Hospers, Goodman does not explain what he means by denotation. It is the relation between a label and the thing or things it is attached to, a name and its bearer, or a predicate and the several members of the corresponding class - and that is all there is to be said about it. Holbein's portrait of Erasmus denotes Erasmus, and a view of the Matterhorn denotes the mountain. However, a painting of a satyr, although it may be said, with perfect propriety, to depict or represent a satyr, denotes nothing at all, for no satyrs exist to be denoted. In this sort of case, Goodman argues, the verb 'depict' does not signify a relation between two independently existing things; it is part of a complex one-place predicate. A painting of a satyr is not something, a painting, which stands in a certain relation, depiction, to another thing, a satyr. It is merely a certain kind of painting, a satyr-representing-painting, so to speak.9 If Holbein's portrait of Erasmus, like the name 'Erasmus', denotes him, whereas a painting of a satyr, like the word 'satyr', denotes nothing, what is it that makes the two paintings, each of them an intricate disposition of pigments upon a canvas, examples of pictorial representation, whereas the two strings of letters, 'E-R-A-S-M-U-S' and 'S-A-T-Y-R', are linguistic symbols? Not, as Hospers and the other semioticians supposed, a 'close natural relationship, namely likeness or resemblance', or for that matter any other sort of relationship, between the pictures and what they depict: 'Descriptions are distinguished from depictions not through being more arbitrary but through belonging to articulate rather than to dense schemes; and words are [no] more conventional than 9 In Languages of Art, Goodman (1981), p. 21 injudiciously claimed that 'F-representing-picture' is an unbreakable predicate, i.e. that its meaning is not a function of the meanings of its parts and the way in which they are combined. ('Red' and 'round' are trivially unbreakable, because they do not have significant parts; 'red and round' and 'red or round' are not unbreakable. Figurative predicates are often unbreakable, despite having significant parts: for example, one cannot deduce the meaning of 'bluechip' or 'copper-bottomed' from the meaning of their parts.) This claim was criticised by Wollheim (1973) and Scruton (1974), pp. 193-97, and it is certainly false. For if 'sloop-representing-picture' were an unbreakable predicate, it would not be possible to deduce that a sloop-representing-picture is a boat-representing-picture from the fact that a sloop is a boat. Goodman (1972, p. 122) has since conceded that F-representing-picture is a complex one-place predicate, and not an unbreakable one. 62
Words and Pictures pictures...if conventionality is construed in terms of...artificiality' (Goodman 1981), pp. 230-1). The manufacture of pictures, Goodman argues, like the use of language, is a symbolic activity governed by a complex system of semantic and syntactic rules; and it is the distinctive character of the system of rules which guide the manufacture and the use of pictures, rather than a special kind of relationship, such as resemblance, between pictures and what they depict, which makes the difference between a picture and a description in words. A symbolic system is representational, Goodman explains, if any change to a character in any one of various respects, however small the change is, will effect a change in what it represents. This, in a nutshell, is the difference between pictures and written descriptions. One example will suffice. Compare a drawing of a man with a long nose with the corresponding written phrase (figs. 1 and 2). If we change the drawing slightly, say by elongating the part which represents the nose, it will represent a man with a longer nose than it did before; whereas the corresponding change to the phrase will not have the same effect (figs. 3 and 4). Of course, a small change to the written phrase might change its meaning (fig. 5); but only if it
Fig. 1
A man with a long nose Fig. 2
Fig. 3 63
John Hyman
A man with a long nose Fig. 4
A man with a long hose Fig. 5
affects the classification of one or more characters, so that an 'n' becomes an 'h', for example. There is no respect - the thickness or colour of the line, the size or shape of any part of the inscription in which any change, however small, will change the significance of a written phrase. Turning from written to spoken language, we can see that if English were modified in such a way that every difference in intonation, volume and pitch corresponded to a difference in meaning, it would become a representational language, or rather, it would have ceased to be a language and would have become instead a means of making audible representations, distinguishable from pictures only by the fact that they would be constructed out of sounds rather than colours. Hence the distinctive character of a pictorial symbolic system depends essentially upon the 'syntactic and semantic relationships among [its] symbols' (1981, pp. 227-8), and has nothing whatever to do with any putative resemblance between pictures and their subjects. Turning now from exposition to criticism, I shall concentrate on the propinquity Goodman's analysis confers on words and pictures; in fact, I shall focus on one fundamental defect, although I shall approach it by a circuitous route. The structure of the argument will be complicated; so I shall explain it in advance. It has three steps. First, I shall consider one of the two most common objections to Goodman's theory: that it severs the connection between being visible and being depictable. I shall argue that Goodman can meet this objection, but only by conceding that if pictures 'function in somewhat the same way as descriptions' (Goodman 1981, p. 30), the descriptions that pictures are somewhat like are descriptions which are couched purely in terms of form and colour. Second, I shall consider the other of the two most common objections to Goodman's theory: that if I can recognize a certain sort of object then I can be expected to recognize a picture of that sort of object, whereas Goodman's theory implies, on the contrary, that I cannot be expected to recognize a picture of something unless I also know the relevant syntactic and semantic rules. Again, I shall argue that 64
Words and Pictures Goodman can meet this objection, but only by making the same concession: that if pictures 'function in somewhat the same way as descriptions', these descriptions must be ones which are couched purely in terms of form and colour. Finally, I shall argue that the concession is fatal, because it implies that we can only identify the subject of a picture by means of an inference. This approach is basically a matter of asking what kind of rule Goodman's theory postulates, or rather, what its nearest linguistic equivalent would be; and whether seeing what a picture represents could be the result of applying rules of this kind. According to Alberti, 'No one would deny that the painter has nothing to do with things that are not visible' (1966, p. 43). If this seems plausible, then it will be tempting to argue as follows: by insisting that the relation between a picture and its subject is wholly conventional, Goodman loses the resources he would need in order to explain why it is impossible to depict sounds; for it is certainly possible to devise a representational scheme which uses colours arranged on a plane surface to represent sequences of sounds, and where a small modification in the symbol (it would be a kind of graph) affects what it represents. For example, the x axis might represent time; the y axis, pitch; hue might represent timbre; and saturation, volume. If the objection is stated in exactly these terms, it seems to be open to a decisive rejoinder, namely, that it is possible to depict sounds. For example, the drawings by Saul Steinberg which the advertising campaign for Classic FM radio made famous unmistakably depict the sounds of various musical instruments. Furthermore, whatever makes it true that the delicate filigree depicts the sound of a harp, it cannot be a close resemblance between them in shape or colour, or the fact that the drawing produces the illusion of hearing a harp, because there is no such resemblance and no such fact. Hence, far from indicating a weakness in Goodman's theory, the question of whether it is possible to depict sounds reveals a weakness in its principal rivals. Perhaps there is a better way of stating this objection to Goodman's theory. For whereas it is perfectly possible to describe a sound without employing a synaesthetic metaphor, any depiction of a sound will need to be metaphorical; and that is a fact which Goodman's theory cannot accommodate. Once again, an advocate of the resemblance theory can take heart, for if X and Y resemble each other in appearance, the resemblance between them may be a literal one; but if the appearance of X resembles the sound of Y, the resemblance can only be metaphorical. There is, I believe, one way in which Goodman can meet this 65
John Hyman objection, and that is by conceding that the language in which pictures speak to us is only capable of literal descriptions of things with shapes and colours. This would not imply that pictures cannot represent landscapes, men and angels; but that they could only do so in a vocabulary confined to shape- and colour-terms. In Goodman's own terminology, the compliance-classes (or classes of denotata) of characters in a pictorial system of representation would be defined by shape and colour alone. If this concession were granted, the fact that a picture of a sound cannot fail to be a metaphor would cease to be an embarrassment. For although colour-terms can be used to describe sounds - Messiaen was famously fond of giving colourful descriptions of his chords - these descriptions are bound to be metaphorical. Now for the other popular objection to Goodman's theory. The ability to recognize something by its appearance seems intuitively to be connected more closely with the ability to recognize a picture of it than with the ability to recognize a description of it; whereas Goodman's theory implies, on the contrary, that neither connection is closer than the other. This objection surely has some degree of plausibility, but it is difficult to turn it into a compelling argument. It would be nice if we could claim, straightforwardly and unequivocally, that if I can recognize something by its appearance then I must be able to recognize a picture of it; for if depiction is explained by semantic and syntactic conventions, I may be able to recognize something by its appearance without knowing the relevant conventions, and hence without being able to recognize a picture of it. But the straightforward claim is, as it so often is, too bold to be convincing. For example, a dog may be able to recognize its master without being able to recognize his portrait. And if the straightforward claim is false, it isn't easy to modify it sufficiently to make it plausible without weakening it to the point where it is consistent with Goodman's theory. A few critics have tried to do this, and failed. Richard Wollheim, for example, put the objection as follows: it is a grave objection to [Goodman's theory] that it cannot account for the evident fact of transfer. By the term 'transfer' I mean, for instance, that, if I can recognize a picture of a cat, and I know what a dog looks like, then I can be expected to recognize a picture of a dog. But on [Goodman's] view this ought to be baffling. It should be as baffling as if, knowing that the French word 'chat' means a cat, and knowing what dogs look like, I should, on hearing it, be able to understand what the word 'chien' means. (Wollheim 1987, p. 77) 66
Words and Pictures Wollheim's argument is admirably concise, but it is not a cogent objection to Goodman's theory because the theory does not imply that pictures are like names. On the contrary, Goodman argues that the rules which determine what pictures represent are significantly different from the rules which fix the meanings of 'chat' and 'chien', and as we have seen, he is quite clear about what the difference is. So it would be a fairer test of Goodman's theory to ask whether transfer would be a surprising feature of a symbolic system or, in particular, of a representational symbolic system - in other words, a symbolic system in which minute changes of various kinds to a character will change what it represents. In fact, there are many symbolic systems in which transfer occurs. For example, the diagrammatic notation devised for musically illiterate guitarists represents the fingering for each chord by marking a grid with dots, and it seems obvious that if I can recognize the sign for a G major chord and I know the fingering for a diminished seventh on D, then I can be expected to recognise the sign for a diminished seventh on D. So, would it be surprising if transfer occurred in a representational system? I can see no reason to think that it would; and since he fails to provide a reason, Wollheim's argument is unconvincing. He seems to have mistaken a disanalogy between pictures and words for a disanalogy between pictures and symbols in general. But if transfer is an 'evident fact' about pictures, as it seems to be, and if pictures 'function in somewhat the same way as descriptions', as Goodman says they do (1981, p. 30), these will have to be descriptions of a special sort, since transfer will have to be a fact about them too. If I can recognize this sort of description of a cat, and I know what a dog looks like, then I can be expected to recognize this sort of description of a dog. But the only sort of description I can be expected to recognize because I know what a dog looks like is a description of a dog's appearance; and the only sort of description of a dog's appearance I can be expected to recognize because I can recognize a certain sort of description of a cat is a description which makes use of the same vocabulary as this sort of description of a cat. In short, if pictures 'function in somewhat the same way as descriptions', these descriptions must be couched in a vocabulary which has a sufficiently broad competence in the description of appearances to match the full extent of transfer in the recognition of pictures; and which is fully engaged in every description, so that if I understand one description, I can be expected to understand another. When we consider how far transfer extends from pictures of clouds to pictures of cuckoos - it is obvious that this could only be a vocabulary of colour- and shape-terms. So, if a 67
John Hyman picture is like a description, it must be like a description couched purely in terms of form and colour. As we have seen, Goodman says that pictures are like descriptions. If my argument is correct, his theory implies that they are like descriptions of a particular sort, namely, ones which only make use of colour- and shape-predicates. It follows that the theory is demonstrably false; for if pictures were like descriptions of this sort, any attempt to say what a picture depicts which goes beyond a description of their colours and shapes would have to be the result of an inference; but this is false. Goodman (1981, p. 36) explicitly denies that we need be 'aware... of making any interpretations at all' when we perceive what a picture depicts. But in fact his theory implies that we should be aware of making interpretations when we perceive that a picture represents the sun, as opposed (roughly) to something round and yellow, or a girl with a falcon, as opposed to something with such-and-such colours at such-and-such places, just as we would be if we realised that a verbal description in terms of shape and colour was in fact a description of the sun or a girl with a falcon. Once we know a language well, we come to hear or see words, phrases and sentences as having the meaning they do. So there is no obstacle to Goodman's maintaining that we need not be aware of interpreting pictures when we perceive what they represent, despite the fact that they are characters in a symbolic system, so long as we do not outstrip their manifest signification. But as soon as we perceive that a picture depicts something of such-and-such a kind, where 'such-and-such' involves more than shapes and colours, he is bound to admit that we have left behind what visual experience can present us with, by means of an explicit inference. Goodman criticises the resemblance theory for implying that visual experience can only ever present us directly with shapes and colours. Following Gombrich, he calls this doctrine the myth of the innocent eye. But the proposition that perceiving what a picture depicts depends on making inferences is tantamount to applying the myth to pictures; for it is no more plausible to deny that we can see what a picture depicts immediately and directly than to deny that we can see the visible objects in our environment immediately and directly, rather than mere patterns of light and colour. In the final analysis, this is the cardinal principle which Goodman's theory contradicts. If his theory were correct, we would need to be capable of describing what a picture depicts purely in terms of shape and colour in order to be capable of describing it at all. But we don't, any more than we need to able to describe the visible things that surround us in the same austere vocabulary, in order to say what they are. 68
Words and Pictures Art and Occlusion If the relationship between pictures and what they depict is not, after all, a purely conventional one; and if Descartes was right to say that even in respect of shape, the resemblance between an image and the object it represents is often imperfect, is it possible to explain precisely what sort of relationship must exist between a picture and its subject? Or is there perhaps - as Descartes seems to have believed — no answer to be found, until experimental psychology has discovered whatever correlations exist between the forms of images and the 'sensory perceptions of all the various qualities of the objects to which they correspond'? I think it is possible to provide an answer, if we make use of the concept of occlusion shape. The occlusion shape of an object is what some philosophers have called its 'apparent shape' - in other words, an object's outline or silhouette. I prefer the term 'occlusion shape' because it is less tendentious, since it does not carry the suggestio falsi that a contrast is being made implicitly between an object's occlusion shape and its real shape, and hence that an object's occlusion shape is not a perfectly objective visible feature of our physical environment. A circular plate viewed obliquely has an elliptical occlusion shape: it will occlude or be occluded by an elliptical patch on a plane perpendicular to the line of sight. An object's occlusion shape is a function of its actual shape (I shall use the phrase, despite misgivings) and its orientation relative to the line of sight. We can be mistaken about the occlusion shape of an object, and a mistake of this kind can be corrected by measurement and calculation. An object's occlusion shape can also be affected by refraction. A straight stick that is half-immersed in water will have a crooked occlusion shape. Again, this is a perfectly objective matter. The concept of occlusion shape plays the following role in the theory of depiction. Consider an engraving of a man's head. Whether the picture shows its internal subject (as Beardsley thought), stands for it (as Hospers thought), or produces a special sort of psychological episode which is related in some distinctive way to the psychological episode which occurs when someone sees a man's head, it does so because the shape (the actual shape, that is) of the smallest part of the picture which depicts the man's nose or chin is identical to the occlusion shape of the nose or the chin in the picture, relative to the spectator's line of sight. The general principle, which I shall call the Occlusion Shape Principle, is easily proved by means of a thought-experiment. In fact, the thoughtexperiment is an unusual one, because it can actually be performed. The experiment is to try to trace the shape of the part of a picture 69
John Hyman which depicts something - whether it's a house or a tree or a man or a part of his body - by running a finger across its surface, without simultaneously tracing the occlusion shape of the corresponding part of the picture's internal subject - the house or the tree or whatever. Alternatively, one might try to trace the occlusion shape of part of a picture's internal subject without tracing the actual shape of the part of the picture that depicts it. It only takes a moment's reflection to see that neither thing is possible. Hence the occlusion shape of a part of a picture's internal subject (I), relative to the spectator's line of sight, and the actual shape of the smallest part of the picture that depicts that part of its internal subject (D) must be identical. It is important to notice how modest the Occlusion Shape Principle is. First, it tells us nothing about what kind of thing the internal subject of a picture is, as opposed to its occlusion shape relative to the spectator's line of sight. For example, if part of a picture represents a duck's bill or a rabbit's ears, the principle does not tell us which; although it does tell us that if the picture can be seen either as a picture of a duck or as a picture of a rabbit, the duck and the rabbit must have the same occlusion shape, relative to the spectator's line of sight. Second, it says nothing about the relationship between the surface of a picture and its external subject, if it has one. Finally, it does not imply that each part of the internal subject of a picture must have a fully determinate occlusion shape; although it does imply that any such indeterminacy will be precisely matched by a corresponding indeterminacy in the shape of the corresponding part of the picture's surface. But although the Occlusion Shape Principle is modest, it marks a fundamental difference between pictures and inscriptions, for the shape of an inscription alone does not tell us anything whatever about its meaning. Mention of meaning brings me to an objection which might well have been raised when the concept of a picture's internal subject was introduced. The objection may as well set out from a remark in Languages of Art:
From the fact that P is a picture of or represents a unicorn we cannot infer that there is something that P is a picture of or represents... Saying that a picture represents a so-and-so is thus highly ambiguous as between saying what the picture denotes and saying what kind of picture it is. (Goodman 1981, p. 22) The objection is that unless talk about a picture P's internal subject is interpreted as talk about what kind of picture P is, it is bound to involve, at least implicitly, the 'fallacious existential inference' (ibid., n. 19) that there is something that P is a picture of, something 70
Words and Pictures that actually exists; and that the Occlusion Shape Principle implicitly commits this fallacy. If the internal subject of a picture is analogous to the sense or meaning of a description in words, the claim that it has a certain occlusion shape or that it resembles Smith involves a blatant hypostatisation of meanings. The reply to this objection is that attributing an occlusion shape to a part of the internal subject of a picture need not be taken to amount to something more than saying what kind of picture it is. The Occlusion Shape Principle allows us to infer from (i) D depicts something which has the occlusion shape S that (ii) D has the actual shape 5 although of course this must not be taken to imply that any inference is necessary in order to perceive that either (i) or (ii) is the case. But if 'P represents a unicorn' can be understood as saying what kind of picture P is, rather than as expressing a relation between P and something else which exists independently of P, as it surely can, then so can 'P (or D) represents something which has the occlusion shape S'. Hence, neither ' / h a s the occlusion shape S' nor 'I resembles Smith' implies that / is something which exists independently of P, and which might therefore survive its destruction; any more than 'Jones is imagining something which has the occlusion shape S' or 'Jones is imagining a man who resembles Smith' carries the corresponding and equally fallacious implication. A second objection to the Occlusion Shape Principle is that portrait caricature disproves it, for surely the physiognomy of a caricaturist's subject may be wildly distorted on the picture's surface, and moreover the distortion need not conform to any fixed rule. The objection is a weak one, because it turns on a confusion between the internal subject of a picture and its external subject. The internal subject of a portrait caricature is a recognisable distortion of its external subject; but the correspondence prescribed by the Occlusion Shape Principle, between the picture's surface and its internal subject, is not breached. A more telling objection has to do with distortions of a different kind, namely anamorphoses: 'perspectives which, rightly gazed upon, / Show nothing but confusion, - ey'd awry, / Distinguish form'.10 The best known anamorphoses include the famous depiction of a skull at the bottom of Holbein's French Ambassadors (in the National Gallery, London) and a portrait of Edward VI (in the National Portrait Gallery), but the technique was commonly and 10
Shakespeare, Richard II, II.ii.18f. 71
John Hyman less playfully exploited by the artists responsible for Byzantine church decoration, to compensate for distortions which would otherwise appear in pictures as a result of the spectator's eccentric line of sight, or the curvature of the picture's surface. For example, in the Pentecost depicted in the main cupola of the Haghia Sophia in Thessalonica, the legs of the Apostles are deliberately elongated because, whereas the upper parts of the figures lie almost perpendicular to the spectator's line of sight, their lower parts are more nearly vertical, and so the anamorphosis ensures that the apostles' bodies will appear to be well-proportioned when the cupola is seen from below. Examples of this sort seem to disprove the Occlusion Shape Principle, because the occlusion shape of an Apostle's leg, relative to the spectator's line of sight, is not identical to the actual shape of the smallest part of the mosaic that depicts it, precisely because the spectator's line of sight is eccentric. Anamorphosis therefore seems to recommend a different principle, namely that the occlusion shape of / and the occlusion shape of D, both relative to the normal or intended line of sight, must be identical. The phrase 'normal or intended' is of course vital, since the occlusion shape of / will not change merely because the spectator's line of sight, and hence the occlusion shape of D, changes. For example, if a circular part of a picture's surface represents a sphere, it will not cease to do so if the spectator moves to one side of the picture. But whereas the occlusion shape of a sphere can only be a circle, the occlusion shape of a circular part of a picture's surface will become elliptical if the spectator moves to one side." But the same phrase - 'normal or intended' - also indicates what is wrong with the new principle. It simply isn't plausible that a normal or intended line of sight which is sufficiently precise to underwrite the principle is associated with every picture: there would, for example, need to be a normal reading posture precise enough to shame the sternest aficionado of the Alexander Technique, and acknowledged since the invention of the illustrated book. And it is manifestly absurd to suppose that a distinct line of sight is specifically associated with each part of every large picture which is not an anamorphosis. Is it feasible to add a rider to the new principle, stating that the default line of sight is orthogonal to the picture plane? Perhaps; but this would simply amount to a return to the Occlusion Shape Principle, disguised by a cumbrous circumlocution, for the occlusion shape of a two-dimensional form, relative to the line of sight that is orthogonal to the plane on which the form is drawn, is " On some related matters concerning the spectator's line of sight, see Gombrich (1977), pp. 215-16. 72
Words and Pictures identical to its actual shape. It is better, therefore, to stick to the Occlusion Shape Principle, and to acknowledge the exceptional nature of anamorphoses - somewhat as the theory of meaning can acknowledge irony and hyperbole, without being stymied by them. Anamorphosis is exceptional; for whereas we normally perceive a picture's internal subject by perceiving shapes and colours on its surface, anamorphosis requires us to wmperceive D, in order to perceive / as we are meant to. Anamorphoses are, in effect, pictures that don't mean what they say. What goes for occlusion shape goes equally for relative occlusion size. If I hold out my hands in front of me, and extend one arm further than the other, my hands will not appear to differ in size, but the greater occlusion size of the nearer hand will be evident: the nearer hand will occlude or be occluded by a larger patch on a plane perpendicular to the line of sight. The relative occlusion size of two objects is a function of their relative size and their relative distance from the spectator, and is, no less than occlusion shape, a perfectly objective feature of the visible objects in our environment. And just as the occlusion shape of /, relative to the spectator's line of sight, and the actual shape of D must be identical, so the relative occlusion size of I, and I2 relative to the spectator's point of view, and the relative size of D, and D2 must be identical. I shall call this the Relative Occlusion Size Principle. The two Occlusion Principles disprove Goodman's doctrine that the difference between a picture and a description in written words is wholly attributable to the distinctive pattern of 'syntactic and semantic relationships among [its] symbols'. They also explain the fact on which - as we have seen - Goodman confers a crucial significance, namely, that a small change to a picture may effect a change in what it represents although a similar change to an inscription would not have a similar effect. For if the Occlusion Shape Principle is true, any change to the actual shape of D will, eo ipso, change the occlusion shape of /, relative to the spectator's line of sight; and if the Relative Occlusion Size Principle is true, any change to the relative size of D, and D2 will change the relative occlusion size of I, and I2, relative to the spectator's point of view. The resemblance theory - in its cruder forms at least - implies something more than this, as Beardsley acknowledged: 'if two representational designs differ as designs,' he claimed, 'no matter how little, they must differ in what they depict, even if only trivially' (1981, p. 295). But this is false, for if two paintings differ only in absolute size, and not at all in the relative proportions of their parts, then in the absence of any explicit convention to this effect, there will not be a difference in what they depict. (Although of course it 73
John H y m a n does not follow that they will not differ aesthetically).12 The Occlusion Principles, by contrast, do not imply that a change in absolute size, large or small, will have a corresponding effect on the size of the picture's internal subject. The Occlusion Principles also reveal how much truth there was in Descartes' final objection to the resemblance theory. The objection was this: Indeed the perfection of an image often depends on its not resembling its object as much as it might. You can see this in the case of engravings...although they make us think of countless different qualities in these objects, it is only in respect of shape that there is any real resemblance. And even this resemblance is very imperfect, since engravings represent to us bodies of varying relief and depth on a surface which is entirely flat. Moreover, in accordance with the rules of perspective they often represent circles by ovals better than by other circles, squares by rhombuses better than by other squares, and similarly for other shapes. Thus it often happens that in order to be more perfect as an image and to represent an object better, an engraving ought not to resemble it. Since an engraving represents objects on a flat surface, the actual shape of / and the actual shape of D cannot be identical if the shape of / is three-dimensional; and the example of perspective shows that the actual shape of / and the actual shape of D need not be identical even if the actual shape of / is two-dimensional. This much is undeniable. And we can add two further points, both in the spirit of Descartes' argument: first, as we have seen, the occlusion shape of / and the occlusion shape of D (both relative to the spectator's line of sight) need not be identical; and second, / a n d D need not resemble each other in colour, for the internal subject of a grisaille or an engraving (whatever the colour of the ink) need not and generally will not have a determinate colour. Descartes was therefore right to reject the resemblance theory, if the term 'resemblance' is taken au pied de la lettre. But having said that, if the Occlusion Principles are true and truth can be mixed with falsehood, then the idea that resemblance explains depiction surely has some truth in it - which is why my first remarks about the formula were equivocal. More importantly, Descartes was wrong to conclude that experimental investigation alone can determine what regular correspondences in shape exist between the surface of a picture and its internal subject. Whether the same is true of correspondences in colour is another matter, and one which I shall discuss elsewhere. 12
74
On the aesthetic significance of scale, see Wind (1985), pp. 65 and n. 124.
Words and Pictures I cannot claim that the Occlusion Principles are an original idea, for something very similar appears in many places. In book xxxv of his Natural History, for example, Pliny insists patriotically that the claim of the Egyptians to have discovered the art of painting is spurious, but apart from this he tells us little about the origins of painting - except that he claims that it is generally agreed that painting began 'with the outlining of a man's shadow'. In fact, the analogy between an image and a shadow makes its first appearance in the Odyssey (xi.204-8), when Odysseus encounters his mother's image (eidolon): As my mother spoke there came to me out of the confusion in my heart the one desire, to embrace her spirit, dead though she was. Thrice, in my eagerness to clasp her to me, I started forward with my hands outstretched. Thrice, like a shadow or a dream, she slipped through my arms and left me harrowed by an even sharper pain. Sometimes indeed 'learning philosophy is really recollecting'.13 References Alberti, L. B. 1966. On Painting, revised edition, trans. J. R. Spencer. New Haven: Yale University Press Anscombe, G. E. M. 1965. 'The Intentionality of Sensation: A Grammatical Feature', in R. J. Butler (ed.), Analytical Philosophy: Second Series. Oxford: Basil Blackwell Beardsley, M. C. 1981. Aesthetics: Problems in the Theory of Criticism, 2nd edition. Indianapolis: Hackett Berkeley, G. 1975. An Essay Towards a New Theory of Vision, in Philosophical Works, ed. M. R. Ayers. London: J. M. Dent Budd, M. 1995. Values of Art. London: Allen Lane Descartes, R. 1985. Optics, in The Philosophical Writings of Descartes, Volume I, trans. J. (Nottingham et al. Cambridge University Press Gage, J. 1993. Colour and Culture. London: Thames & Hudson Gombrich, E. H. 1977. Art and Illusion, 5th edition. Oxford: Phaidon Goodman, N. 1972. Problems and Projects. Indianapolis: Bobbs-Merrill 1981. Languages of Art, 2nd edition. Brighton: Harvester Hacker, P. M. S. 1987. Appearance and Reality. Oxford: Basil Blackwell. Hospers, J. 1946. Meaning and Truth in the Arts. Chapel Hill: University of North Carolina Press Hyman, J. 1992. 'Language and Pictorial Art', in D. Cooper (ed.), A Companion to Aesthetics. Oxford: Basil Blackwell Morris, C. 1939-40. 'Aesthetics and the Theory of Signs', Journal of Unified Science (Erkenntnis) 8, 131-50 1946. Signs, Language and Behaviour. New York: Prentice-Hall 13
Wittgenstein (1993), p. 179. 75
John Hyman Ryle, G. 1949. The Concept of Mind. London: Hutchinson 1971. 'Sensation', in his Collected Papers, Volume 2. London: Hutchinson Scruton, R. 1974. Art and Imagination. London: Routledge & Kegan Paul Wind, E. 1985. Art and Anarchy. London: Duckworth Wittgenstein, L. 1993. Philosophical Occasions, ed. J. Klagge and A. Nordmann. Indianapolis: Hackett Wollheim, R. 1973. 'Nelson Goodman's Languages of Art' in his Art and the Mind. London: Allen Lane 1987. Painting as an Art. London: Thames & Hudson
76
Social Externalism and Conceptual Diversity ANDREW WOODFIELD I. Social Externalism and its Ramifications Social externalism is a thesis about the individuation-conditions of thoughts. Actually, the thesis applies only to a special category of 'trained' thoughts, thoughts which issue from trained thinking. It isn't that the thinker of such a thought has to have had special training about the subject-matter. It is rather that he or she needs to have acquired certain basic linguistic skills and values. For trained thoughts are thoughts whose contents are tailored to the demands of communication. Social externalism, as I understand it, says that people who are competent in a public language are equipped to have certain thoughts whose contents are fixed (in part) by the lexical semantic norms of their language.1 This restriction to trained thoughts has far-reaching consequences. It makes the thesis more modest and more compelling. Not all of a language-user's mental states have their contents fixed in this way, and the mental states of a non-language-user'cannot be so fixed. There must be various primitive kinds of thinking which subserve and predate trained thinking. If prelinguistic children were not naturally endowed with the intelligence to construct cognitive maps of their surroundings, to imagine possibilities, to choose between alternatives, and so on, they would be incapable of learning any language and consequently would be unable to acquire the capacity for trained thinking. Tyler Burge has devised some notable arguments that support social externalism (in Burge 1979, 1982 and 1986). Burge does not call his own position 'social externalism', however. His project has been to defend 'anti-individualism', a position that is doubly externalist. He 1
When I say 'fixed' by the norms, I do not mean to deny that most lexical items do not have very precise meanings. There may be indeterminacy in the meaning of every word and sentence of language L, when L is viewed from the perspective of another language. My position is that if a given word of L has an indeterminate meaning, the concept expressed by it will have an indeterminate content, and it will be neither more nor less indeterminate. That mental states cannot transcend the indeterminacy of public meanings by being more fine-grained and precise is a position endorsed by Quine (witness the comment appended to Follesdal 1990). 77
Andrew Woodfield holds that many contents depend individuatively upon the physical environment as well as upon the linguistic environment. I shall say nothing in this paper about physical environment externalism. The original thought-experiment that Burge set out in 1979 has become extremely well known. The 1986 one is perhaps less wellknown. I shan't retell either of them in detail, but I shall, very briefly, compare them. In each story the protagonist is a normal speaker of English, living in an English-speaking community, who says something odd. The oddness hinges upon the character's use of a certain word. Mediumfrequency nouns that most people know but cannot precisely define lend themselves to these thought-experiments; but the experiments also work with medium-frequency verbs and adjectives. Such words belong to the communal lexicon, and they express general concepts. In the 1979 experiment, the protagonist - let's call him 'Alf says 'I have arthritis in the thigh'. This is odd because the word 'arthritis' refers to a disease that occurs only in the joints of the body. Apparently, Alf doesn't fully understand the word. Despite this, everybody accepts that he believes he has arthritis in the thigh. In the 1986 experiment, the protagonist - whom I shall call 'Bert' - understands perfectly the standard meaning of 'sofa'. But he has become critical of the usual definitions. He has developed the eccentric theory that sofas are really religious artefacts, and that other people are under the illusion that sofas are ordinary items of furniture. Although Burge does not quote the actual words in which Bert expresses his scepticism, we may imagine that Bert says 'Sofas are not primarily for sitting on.' Hearers would find the utterance odd, but they would interpret it at face value and would credit Bert with the belief that sofas are not primarily for sitting on. The crucial steps in each experiment are a counterfactual supposition and a comparison. The comparison leads to the conclusion that the protagonist's belief-content is not fixed by his present physical state or physical life-history. The content depends rather upon the public meaning of the key word. Take the 1979 case first. Suppose (counterfactually) that Alf were living in a community where they speak a language just like English except that the word-form 'arthritis' properly denotes a range of bone and joint conditions. The sounds 'I have arthritis in the thigh' have a different meaning there. Suppose that Alf's physical causal history were the same as it is in the actual world, and Alf were to utter those sounds. He would not thereby express the belief that he had arthritis in the thigh, nor would it be the case that he believed that. The content of Alf's belief in the counterfactual situation is different from the content of his belief in the actual situation, and yet 78
Social Externalism and Conceptual Diversity Alf's intrinsic physical make-up and history are exactly the same in both. How is this possible? After all, we must suppose there to be physical differences between the environments. People will have said sentences and written dictionary entries in English which cannot be supposed to have had counterparts in the variant community (and vice versa). How could Alf s internal neural organisation fail to carry traces of the linguistic environment in which he has been brought up ? The answer is: Alf has been causally affected by the linguistic environment in which he has been brought up - but only to the extent that the inner traces have come to be a partial reflection of the ambient language. Alf has picked up the word-sound by interacting with other people, and he has gleaned information from these encounters. But his life-time of experiences with the word and with the topic add up to a limited sample of the totality of utterances that have occurred within the community. No utterance that would have been distinctive to English and alien to the variant language has ever impinged upon Alf (although some language-differentiating utterances may have occurred out of Alf's earshot). There is no difficulty in supposing, then, that all the brain-changes that were induced by speech-encounters in the actual world have exact physical counterparts in the counterfactual world. In the 1986 argument, Burge is not aiming to establish social externalism but to take a step beyond it. He now claims that 'even where social practices are deeply involved in individuating mental states, they are often not the final arbiter. This is because the sort of agreement that fixes a communal meaning and norms for understanding is itself, in principle, open to challenge' (1986, p. 707). Burge may well be right about this. But the 1986 argument runs parallel to the earlier one in the following respect: it shows that if Bert had inhabited a linguistic environment in which 'sofa' did not mean sofa but his physical history had been 'substantially the same', Bert would not have had the belief that sofas were primarily for sitting on. And in most cases standard word-meanings are still the final arbiters and individuators, because most subjects do subscribe wholeheartedly to the standards. I am persuaded that social externalism is true and that it has important implications for psychology. It challenges cognitive science to develop more sophisticated models of the human mind. Surprisingly, its wider implications have hardly been explored. In the next section I assess a criticism of Burge's first argument. Then, in section III, I investigate how it is possible for psychological processes to operate on thoughts that are socio-linguistically individuated. In section IV I try to relate social externalism to the hypothesis of 79
Andrew Woodfield linguistic relativity proposed by the linguists Edward Sapir and Benjamin Lee Whorf. Sapir and Whorf held that languages gradually structure people's mental capacities so that speakers of different languages come to vary in their habitual ways of thinking. The Sapir-Whorf hypothesis is a causal-developmental hypothesis. It is in principle testable. The currently fashionable opinion seems to be that there is very little hard evidence to support it. Social externalism is not a causal-developmental hypothesis; it is a thesis of individuative dependence. Social externalism and the hypothesis of linguistic relativity are quite different theses. Nevertheless, there is a link between them. If social externalism is true, and if my explanation of how it can be true is on the right track, learning language L partly determines the selection of the concepts that L-speakers use in their trained thinking. This form of determination is a peculiar hybrid. It isn't the kind that linguists and psychologists traditionally had in mind when they considered the proposition that language determines thought. Also, the conceptual diversity that is engendered via the new route does not doom members of different cultures to mutual unintelligibility. II. Social Content and Psychological Content No one doubts that some propositional attitudes have contents that coincide exactly with semantic contents. It can happen that a person says exactly what he or she thinks, that one person comes to believe what another person said, and so on. Equally, no one doubts that sometimes there is a failure of match; occasionally the proposition expressed by the subject's utterance is not the proposition that he or she really believes. Burge's arguments rely upon our accepting that the protagonists' words be taken at face value. Many critics refuse to accept this. They attack the arguments by claiming that in a Burgean scenario the character's key utterance does not accurately and literally express his belief.21 find their strong line unconvincing, because our normal interpretative practice, in a range of scenarios similar to Burge's, is to credit the subject with a literal utterance and to infer that he has the belief as expressed, provided that the subject is a 2
Fodor (1982), Bach (1987) and Crane (1991) have taken the strong line that when a subject incompletely understands the meaning of a key word in his utterance, it is plain wrong to attribute a belief whose content is the normal social content of the utterance. Bach (1988) argues that the same goes for Bert, whose weird theory subverts the normal meaning of the word that he utters. Bach suggests that when Bert says 'sofa', he may not intend to be using the word literally. 80
Social Externalism and Conceptual Diversity normal adult speaking in his own language and not making any slip of the tongue. Philosophers who have a prior theoretical antipathy to social externalism are liable to be biased in their assessment of the interpretation-problem. Some critics, however, have adopted a milder position, and I should like to focus upon this. The first statement of it was by Loar (1988). Loar introduced a distinction between social contents, which are individuated in line with the semantics of the subject's public language, and psychological contents, individuated in accordance with the subject's personal psychological profile. In certain cases, the two modes of individuation pull apart. According to Loar, Burge's characters present us with a choice between these two modes of individuation. For ordinary purposes it is satisfactory to attribute social content. But if you are interested in Alf's psychology, you'll prefer to attribute psychological content. And it is not the case that the psychological content of Alf's belief is that he has arthritis in the thigh. So the psychological content is something else. To get at this something else, the psychological interpreter needs to delve deeper within Alf's psyche. The first task is to identify the pure, linguistically untainted thought-content. Then, having identified it, the interpreter faces the problem of characterising it verbally. The problem of capturing the (supposed) 'inner' content is a familiar one to internalists. It has been suggested that an individual's inner contents may be so idiosyncratic as to be inexpressible in English. One way not to solve the problem is to introduce neologisms. It has been suggested, for example, that Alf really believes he has tharthritis in the thigh. The invented term is stipulated to have the sense with which 'arthritis' is associated inside Alf's mind. It is supposed to capture Alf's personal concept. But once this first step has been taken, lots more new terms will bid to be admitted into the psychologist's technical vocabulary. Alf, like all of us, slightly misunderstands many of the words he uses. The theorist will need to pepper his reports with plenty of 'th' prefixes to signal Alf's other personal concepts. When Alf, after returning from the surgery, says 'The specialist was fiddling with her stethoscope', what he believes from a strictly psychological point of view - and here I write with due scientific precision - may be that the thpecialist was thiddling with her thtethoscope. 3 And the introduction of artificial terms has only just begun, because different neologisms will be required to 3
The neologising move is a step in a bad direction. Now is not the place to warn of all the wrong tracks it leads down. In Woodfield (1993 and 1996) I argue that the individualistic notion of what a concept is makes impossible any theory of conceptual development. Considerations later in the present paper will show that it leads also to a wrong account of cognitive self-management. 81
Andrew Woodfield capture the concepts of other imperfect understanders whose idiosyncrasies are different from Alf's. Internalist accounts of thought-content imply that only people who fully understand the public sentences they utter are in a position to have thoughts whose contents coincide exactly with the public meanings. This implausible doctrine has the effect of drastically undercounting the number of fully articulate thinkers in a given language-population. With social externalism, on the other hand, it isn't just the elite who have their thoughts aligned with public sentence-meanings. Burge distinguishes three levels of understanding: first, understanding sufficient to engage in responsible ratiocination with a notion expressed by a term; second, normal competence; third, full understanding (Burge 1986, p. 713). Speakers at all three levels have the right to be construed at face value. Of course, people do not always express themselves literally. Mismatches between thought and speech can be deliberate or unintentional. One may choose to express a thought in a figurative or indirect manner, intending the audience to recognise that one's expression is figurative or indirect. On the other hand, one may have a certain thought which one intends to express literally, and then utter a sentence that fails to express the thought. What seems to me interesting and significant is that in most cases where there is a mismatch the real thought-content is still social content. Take malapropisms. Mrs. Malaprop is Sheridan's invention, so we can never know the precise psychopathology in her case. But mistakes like hers could surely be due to a malfunction in the accessing of phonological representations. The right word (one that matches the thought) is activated in the mental lexicon, it stimulates the trace of another similar-sounding but wrong word, and then the wrong word gets sent to the speech production centre. Imagine Bert, in a discussion about sofas, blurting out the sentence 'Sophisms are religious artefacts'. The hearers rightly take him to mean (and think) that sofas are religious artefacts. This thought has social content because it is fixed by the public meaning of 'sofa', even though Bert did not actually produce the sound 'sofa'. Loar maintains that we need the notion of fine-grained psychological content. He tells a story in which one person allegedly has two co-existing concepts with the same social content. A boy, Paul, is brought up by a French nanny. Paul and the nanny speak English together, but the nanny always calls cats 'chats' (using the French word), and Paul adopts the same practice. Paul learns a lot about cats, but he never learns that in English the normal word is 'cats'. He forms a belief that he would express by saying 'All chats have tails.' Clearly, this is a belief that all cats have tails. 82
Social Externalism and Conceptual Diversity Paul has contact with his English parents. They introduce him to the word 'cat', but they happen to use it in conversations where no cat is present. Through the parents, Paul learns various facts about the animals they call 'cats', but he never realises that these animals are of the very same kind that he calls 'chats'. One sentence his parents say to him is 'All cats have tails'. He thereby acquires another belief that all cats have tails. 'And it seems clear', says Loar (ibid., p. 103), 'that Paul has two beliefs, with distinct psychological contents. For they interact potentially with other beliefs in different ways despite their common univocal ascription.' Loar argues that this boy's complex cognitive structure requires us to make content-distinctions that are finer than the distinctions enshrined in the public lexicon, and therefore Burge was wrong to conclude on the basis of his thought-experiments that belief-contents, in cases like those of Alf, Bert and Paul, must be individuated in the socially externalist way. There is another way, which (arguably) is not socially externalist.4 Loar's example is ingenious, for it is undeniable that in some sense of the phrase 'way of thinking', Paul has two ways of thinking about cats. The question is: does this require us to postulate a new kind of content} I am inclined to deny it, because I think we can account for the case more naturally by appealing to the distinction between concepts and conceptions.
Suppose we take a concept to be a classificatory norm or rule. The concept cat is a norm that determines the correctness or incorrectness of acts of symbolic classification which in English are marked by the symbol 'cat', in French are marked by 'chat', and in other languages are marked by other symbols. A French speaker has the same concept cat as an English-speaker, so long as the rule for classifying things under the general term 'cat' is the same as the rule for classifying things under the term 'chat'. On this view, having the 4
Some critics press a little harder. Patterson (1990), building on Loar's argument, urges that individualistic content-individuation is not merely available as a theoretical possibility for use by scientific psychologists, but is actually attributed by ordinary people in some cases. Patterson says we normally attribute social content to a person's opinions when holding that person intellectually responsible, but we sometimes attribute individualistically individuated content to beliefs when we are engaged in commonsense psychological explanation of the person's actions. I find both of the supposed contrasts puzzling. If S really holds the opinion that p, surely S believes that p? And if S's act is correctly explained (partly) by S's belief that q, don't we normally hold S responsible for believing that q? The practices of holding-responsible and explaining must be coherently integrated. This is especially clear in cases where the act to be explained is S's uttering of a sentence. 83
Andrew Woodfield concept cat does not consist in having an internal representation of cats (though it is compatible with the proposition that every possessor of the concept does in fact have an internal representation of cats). Let us now consider what a conception is. Roughly, an individual's conception of kind K is a package of information and misinformation about K. It may be regarded as a set of linked beliefs about K. This is an oversimplification - the package may contain perceptual information about particular Ks as well. But this won't affect the point I want to make here.5 Conceptions admit of a type token distinction. Token-conceptions are personal possessions; my conception of K is numerically distinct from your conception of K. Our respective token-conceptions of K , being directed upon the same topic, are instances of the same conception-type (in a classification-scheme where conceptiontypes are individuated by their respective topics). Yours and mine may be qualitatively identical in the sense that they contain the same information about K. They will be qualitatively distinct if they contain different bodies of information about K. Suppose you have ten linked beliefs about K, while I have only two. In this case your conception of K is psychologically different from mine, even though they are of the same type. Each token-conception is a mental structure containing a set of propositional (and non-propositional) contents that the owner has built up over time. Let us now return to the treacherous notion of a way of thinking. There is ambiguity in the question: 'How was Paul thinking of that animal, at the time when he called it "cat"?' Is the question asking which concept Paul made use of? Or is it asking which conception Paul had? If it is the latter, an appropriate answer might describe the contents of Paul's personal conception associated with the word 'cat' (with the answerer taking it for granted that this conception was of the type, conception of cat). The interesting fact about Paul is that he has two token conceptions of the cat type, one is linked with his internal representation of the word 'cat', the other is linked with his internal representation of the word 'chat', and the two packages contain different (somewhat overlapping) collections of information. We explain people's actions by citing their conceptions as well as their beliefs and desires. For it sometimes happens that a whole conception becomes causally relevant. Consider this very common scenario: a subject S, already believing that Ks are F, consciously judges, and then asserts, that Ks are F. The belief concerns a topic K, of which S has a prior conception. (This particular belief-content may, or may not, be part of the collection of contents that made up S's prior conception of K.) When S consciously entertains the 5
84
For more on conceptions, see Woodfield (1991).
Social Externalism and Conceptual Diversity thought, the conception is activated, and the other contents in the package exert an influence. The whole conception, and not just the conscious thought, helps to cause S's next actions and thoughts. Another way of putting this is to say that the causal role of the foreground thought depends upon S's background conception, as well as upon S's current motivation and other mental states. So if it occurs to a third party to wonder what made S think that thought, or to wonder what S is going to do or say as a result of thinking it, adverting to the background conception is an obvious help. In everyday interpretation, we try to fill out the mental condition of any subject whose background conceptions are unusual. We set his or her foregrounded thoughts against the background of the conceptions. According to social externalism, the contents we ascribe to (trained) thoughts and beliefs are always social contents. But - and here is the most striking fact - when we spell out the contents of idiosyncratic background conceptions, we find that these contents too are socially individuated. For example, suppose that part of Alf's conception of arthritis is that it is a disease that can afflict knees and elbows. This background conception has social content, because Alf employs, and is answerable to, the public concepts disease, knee and elbow. His conceptions of these things may be impoverished and partly erroneous, just as his conception of arthritis is. Even so, those public concepts are the ones that he subscribes to. It is they that fix the individuation-conditions of all the relevant propositional contents. Loar's notion of psychological content was designed to provide the means to theorise about individual idiosyncrasies. But we can already achieve that goal. We cast our net widely so as to take account of the subject's background conceptions. We attribute sets of social contents to the conceptions. So we don't need a new kind of content. It is a mistake to try to concentrate all the idiosyncrasy into the foreground. The foreground is only the tip of the iceberg.6 Undoubtedly, the motivation behind Loar's move comes from the 6
In denying the existence of psychological content, I do not imply that all contentful mental states have social content. On the contrary, I maintain that social contents are found only at the level of conceptualised thinking in language-users (i.e. trained thinking). Minds contain many representations exhibiting various species of non-social content, and it is, of course, open to anyone to call these 'psychological'. My argument is directed specifically against Loarian psychological content - a kind which high-level thoughts and beliefs are supposed to have, and which is relativised to the individual's idiosyncratic conceptions. Loar's notion differs from notions like natural informational content, teleofunctional content, scenario content, and so on, both in the motivation behind it and in the job it is supposed to perform. 85
Andrew Woodfield theory of functional role semantics. This theory tries to fix content by functional role. It says: if a thought has a peculiar functional role within a given individual, it must have an idiosyncratic content for that individual. Social externalism, on the other hand, says, as far as trained thoughts are concerned: the content of a thought is not fixed by its liaisons within the individual but by the semantic norms to which that individual owes allegiance. That allows the content to be stable and shareable. So the content is not idiosyncratic, even though the conception to which the thought is linked is idiosyncratic. The idiosyncrasy stems from the binding together of particular sets of beliefs into packages. This account allows the liaisons between S's beliefs to change over time without these changes affecting their contents. This is an indispensable advantage, given that conceptions continually change. If the content of one's thoughts were determined by one's shifting conceptions rather than by stable concepts, one couldn't count on being able to entertain the same thought twice. Let us return to the notion of concept that social externalism favours.The members of a community of thinkers who speak the same language exhibit all sorts of psychological differences. Each is psychologically unique. However, they have access to a stock of public concepts. Two people share a given concept - the concept elm, for instance - if and only if they subscribe to the norm for classifying things as elms. Subscribing to a classification-norm is quite different from behaving or functioning in conformity to the norm. Two people possess the same concept just in case they both subscribe to the same correctness-standard for evaluating appropriate performances; it is irrelevant whether they produce the same performances on the same tests, or have the same level of skill. The old empiricist theory, that a concept is a recognitional capacity, is clearly wrong. It is not true that a person must achieve a certain level of skill at recognising elms if he or she is to possess the concept elm. A forester may be more skilled at this than an office-worker, and yet both can entertain elmthoughts equally well. The office-worker may have zero capacity to discriminate elms from beeches in perceptual recognition tasks. Many of our concepts are of substances whose samples we cannot identify without carrying out special tests. I cannot tell the difference visually between pure nicotine in powdered form and powdered borax, but this doesn't stop me having distinct concepts. Similarly, it is false that S's concepts are individuated by their respective functional roles in S's cognitive system. Consider the doctor who figured in Burge's 1979 experiment. The doctor and Alf communicated successfully using the concept arthritis, though it 86
Social Externalism and Conceptual Diversity was clear that they had qualitatively different conceptions of arthritis. The functional role of the concept in Alf's thinking was idiosyncratic and deviant, whereas the doctor's thinking may be presumed to have exemplified a pattern characteristic of a full understander. If functional role were individuative, these two subjects would have distinct concepts. Most philosophers accept that concepts are bound up with norms. They often say that the notion of a concept is a normative notion. However, social externalism suggests to me that a concept is a norm. A concept stands outside the psychological states and processes that it evaluates, outside even the cognitive practices of an ideally competent subject. For no empirical kind-concept is there a definite role in cognitive processing that it should have, or would have in a supposedly 'ideal thinker'. For the role that a concept has - and should have - in an individual depends upon that individual's contingently amassed set of background conceptions. There is no 'right' set of conceptions. And it makes little sense to try to specify which background conceptions a purely idealised thinker would have. Each subject accumulates a stock of knowledge as opportunities arise, and even an expert's stock is bound to be partial and idiosyncratic. III. The Mind of a Trained Thinker If social externalism is true, then 'live' concepts — concepts that are actually employed by human beings - depend (in some sense) upon social practices. Concepts themselves are abstract, they are not located in people's heads or in any other place. Still, something needs to be said about the mental capacities and structures within an individual who thinks with the help of communal concepts. Two theses will be defended in this section, one concerned with conceptpossession, the other concerned with concept-acquisition. We need to rethink the notions of concept-possession and concept-acquisition. For the terms 'possession' and 'acquisition' are not quite right. ('Concept-formation' is even worse.) An individual cannot possess a concept as an item of personal property. Concepts, like traditions, are the intellectual property of a culture. In the UK there is a tradition of Morris dancing, but it makes no sense to say that any individual in the UK possesses the tradition of Morris dancing. What a person can do if he or she so wishes is take part in a tradition. Similarly, one can avail oneself of the concepts that are offered by one's socio-linguistic environment. The thesis that I propose, concerning conceptual ability, is as follows: 87
Andrew Woodfield An individual S who knows L is able to make fruitful use of the concept expressed by a given general term in L if S is committed to the communally recognised standard for symbolically classifying things under the general term, and is able to collect and manage information about things so classified.
After I have elaborated upon this, I shall try to defend the following thesis about the acquisition of conceptual ability: Provided that S is suitably ensconced in a language community and is internally ready and conditions are favourable, then S's learning a general term normally causes S to acquire the ability to use the concept expressed by it.
Let me begin by listing some characteristics that are typical of trained thinkers. Such thinkers hold themselves to the opinions they sincerely avow, and regard themselves as bound by basic principles of rationality. They strive for consistency in their beliefs, bowing to pressure if inconsistency is detected. They relinquish beliefs that are shown to be false, and they defer to authoritative sources of evidence. Most importantly, they defer to authority on matters of word-meaning. They accept correction when it is pointed out to them that they have a belief that rests upon a verbal misunderstanding. Once the misunderstanding has been explained to them, they adapt their use of the word. They acknowledge that they previously had beliefs which were false or ill-founded through being tainted by the earlier misunderstanding. But they are not slavish copiers. They generally defer to authoritative others (such as experts). They defer to experts not because the experts are the definers of meaning, but because experts are the most reliable estimators of meaning in the language. Any thinker-speaker who possesses these characteristics is competent to take part in rational social intercourse, a complex gamelike practice. To compare it with a game is to invite questions like: What are the rules? What are the permissible moves? How does a player make a move of a given type? Is any special apparatus required? As with all analogies, this one has its limits. For example, chess-play is aimed at winning; rational social intercourse, in contrast, does not seem to have a single aim. It has a multiplicity of different aims. However, there is a set of aims that endure across generations and have a special importance, namely, to discover, make public and record truths. Games like chess are defined by various sorts of constitutive rules, including rules that define the basic repertoire of valid move88
Social Externalism and Conceptual Diversity types and rules for the conduct of play. The rules of chess include also a set of conventions that determine the structure of the board, the different types of chessmen and their starting positions. Rational social intercourse is not such a well-defined game as chess. But there seem to be various strata of constitutive norms in rational dialogue which are roughly comparable to the different kinds of constitutive rules of chess. For example, linguistic sounds or shapes are the analogues of pieces of wood or plastic in chess; they are the counters that get moved around. Conventions (or norms) ensure that all tokens of a given type have recognisably the same sound or shape. The sound and the shape are arbitrary. Their role is just to distinguish that type from the other types. Players need to know which type a given counter belongs to, because a 'move' is an operation upon a counter qua token of a type. In rational dialogue, the basic moves are locutionary acts. The simplest locutionary act of all is the act of calling an object something (also known as symbolically categorising). Such an act can be performed by demonstrating the intended object and uttering a general word with the right intention. This is like the basic chess act of moving a piece. By moving a piece the chess player constrains the opponent's range of possible next moves. Depending upon the positions of the other pieces at the time, a basic move may generate a non-basic move (such as checking the opponent's king) which is itself governed by a set of higher-level rules. In rational dialogue, the non-basic moves are illocutionary acts (such as asserting, denying, supposing, suggesting, and a host of oth-
ers). It is on the non-basic level that the players' interest and engagement mainly reside. But illocutionary acts could not exist without locutionary acts of symbolic categorisation. An agent publicly performs a basic move by making a distinctive mark or sound which enables others to identify which move it was. A player publicly symbolically categorises an object as a dog, for example, by producing whichever sound or mark is assigned the role of marking the act-type categorising-as-a-dog, within the language that he happens to be speaking. A player speaking French utters 'chien', a Portuguese speaker may do it by uttering 'cachorro'. The word-sound of the selected language is the external apparatus through which the player publicly registers his performance of the basic move. Making the sound 'chien' when in France is obviously not what the act of categorising-as-a-dog consists in. Nor would a disjunction of all the acts of uttering words for 'dog' in each of the world's languages take us any closer to its essence. What, then, does the act-type of symbolically categorising essen89
Andrew Woodfield tially consist in? I find this question extremely difficult and cannot give a very satisfactory answer, but my hunch is that symbolic categorisation is an essentially game-internal type of act, defined (perhaps partially) by its role in the game of rational social intercourse. Since the game can be played in any language, its moves must be specifiable without essential reference to a particular language. Nevertheless, the game cannot be played fluently and extendedly unless the players share some articulated symbolic medium through which to make their moves mutually identifiable. The medium is bound to contain a repertoire of symbol-types dedicated to the expression of distinct types of categorising-as. To maintain the stability of a symbol-system, normative pressures are applied by the community of players. Symbolic categorising-as is the sort of act which requires a background set-up of symbols, just as chess requires a board and a set of chessmen. Symbolic categorising is intrinsically symbolic, in contrast to the operations of perceptual categorising and behavioural categorising. Animals (and human beings) perform the latter operations in the course of low-level cognitive processing. Although animals can be trained to make perceptual and behavioural categorisations, this is not what I mean by 'trained thinking'. Trained thinking is wedded to symbolic categorising, which is a distinctive type. Players need to know three lots of things before they are qualified to play a worthwhile game. They need to know which sounds or shapes are locally recognised as the basic vocabulary-items. They need to know which are the categories into which things may be symbolically put. And they need to know which sound or shape goes with which category. However, it would be wrong to think that these three are completely independent. For knowing which are the symbolisable categories is the same as knowing the basic moves in the game, and knowing them may amount - for a player - to knowing how to play. This knowledge is practical rather than theoretical. It involves the ability to respond to and be regulated by the normative pressures that make themselves felt in the course of play. Individuals become responsive through training and practice to the norms which regulate other people's shared classifications. They witness their own and other people's acts of calling, then they observe the consequences. They feel normative pressures, and they imagine what it is like for others whom they see pressured. It is through taking part in the game that they learn the proper uses of the symbols. The norms of symbolic categorisation are discerned through the learner's experiences of assessments and criticisms and consequences of actual productions of local sounds and shapes. 90
Social Externalism and Conceptual Diversity Rarely is any player able to state (for a given kind K) the norm for classifying an object as a K. The norm is usually too complex to be stated. It is grounded in a history of performances of symbolic acts by contemporaries and ancestors, accompanied by reflective assessment of the performances and discussion about the proper use of the relevant symbol. Empirical discoveries may feed in to the tradition and influence subsequent practices. An individual receives limited exposure to normative pressures in the course of observing and making moves, hence the average person acquires only a partial knowledge of the classificatory norms. The player subscribes to the norms in advance of fully knowing what they are and without knowing in detail how to apply them, just as a club-member subscribes to the rules of the club without having properly read the rule-book. What goes on inside the head of a trained thinker/conceptuser? I take my cue from the Eleatic visitor: in thinking, the soul converses with itself (Plato's Sophist, 263e). To be a trained thinker is to be so organised internally that one subscribes in one's thinking to the same norms of categorisation that one subscribes to when engaged in rational dialogue. This form of cognitive organisation has some distinctive features, which I shall try to sketch. I am keen to distance myself from cliches like ' T h e English think in English, French people think in French.' (Ryle, in his later work, was concerned to get away from this idea too. See Ryle 1979, p. 41.) When we think trained thoughts, we do not think in a public language. Yet symbol-norms are somehow in the offing, and the problem is to explain how. T h e solution must be compatible with Burge's point that full understanding is not required. I start by supposing that every subject S who speaks language L has, for each general word W of L that he or she knows, an internal representation Rep(W) of the phonological properties and some of the syntactic properties of W. Rep(W) is securely linked, in a special way, to S's conception C of the topic which W names. C's topic may be a thing-kind, a stuff-kind, a property, an act or an event-type. The link between Rep(W) and conception C remains highly stable over time. The stable link is relied upon by a 'librarian' system inside S which manages S's stock of conceptions. House-keeping operations, such as adding a new entry to a mental file and deleting an old entry from a file, are controlled by this system. It uses the links between conceptions and public words, mediated by Reps of the words, as guides in deciding where to file a new piece of verbally communicated information and whether to delete an old piece. Its file-management procedures are topic-led, in the 91
Andrew Woodfield sense that the existence of the structural link between C and Rep(W) serves the key functions of fixing C's topic and keeping C targeted upon that topic over time. For this reason I call it a topicfixing link. The link between C and Rep(W) ensures that C is and continues to be a conception of the category named by W in public language L. With trained thinkers, a lot of the structure of the overt game of rational social intercourse carries over to the mental game of thinking. The existence of parallelism should not be surprising. It is well-known that correlations exist between mental-act verbs and speech-act verbs. Judging is the mental analogue of asserting; entertaining a thought goes with expressing a proposition; thinking of a particular correlates with referring to a particular, and so on. In fact,
many verbs for contentful acts ignore the 'public versus private' distinction; they cover both speech-acts and purely mental acts. For example, accepting, hypothesising, supposing and categorising are acts
that can be performed either by speaking or by silent thinking. Once S has become a committed subscriber to the semantic norms of L, those norms individuate both S's speech-acts of symbolic categorising and S's mental acts of symbolic categorising. For the norms which fix the public meanings of S's words also fix the topics of the conceptions that S activates when thinking trained thoughts. Consider a native French-speaker Hilaire who, on a number of occasions, tries to categorise various objects as beeches. Sometimes he is right, sometimes wrong. On the occasions when he performs these categorisations publicly, he uses the word 'hetre'. A condition for the possibility of evaluating any of his beech-categorisings as correct or incorrect is that the act in question be identifiable as being of the type beech-categorising independently of whether it is correct. The tokening of the symbol 'hetre' provides the independent line for identifying a public act as being of that type. The logical requirement of independent type-identifiability still holds when Hilaire performs private acts of beech-categorising. Neither he nor a third party can evaluate the correctness of any such particular act unless there is an independent method of fixing its type. The content of his act of categorical judgement must not be undermined just because the categorisation happens to be false or unjustified. According to my hypothesis, when Hilaire thinks of something as a beech, he exercises a particular conception C which is his conception of beech, and the independent criterion of C's being a conception of beech is the existence of a topic-fixing link between C and Hilaire's internal representation of the word-form 'hetre' - all this being dependent upon a complex set of back92
Social Externalism and Conceptual Diversity ground conditions concerning Hilaire's competence to play the game in French. 7 Hilaire's conception of beech may contain many false beliefs. Whenever he hears someone else call an object 'hetre', and whenever he himself calls an object 'hetre' and gets a reaction from someone else, Hilaire has the opportunity to glean new information about beeches. Hilaire will intelligently update his conception and will manage it according to rational principles. In encounters like these, where the word is actually uttered and heard, Hilaire's perception of the word-sound will activate Rep('hetre') and this will be responsible for activating C. My cognitive hypothesis does not imply that Rep('hetre') is activated on every occasion that he entertains a beech-thought.Suppose Hilaire is lying in bed, eyes closed. He neither receives nor transmits. He muses that beeches make poor hedges (false, by the way). He has a thought attuned to the public concept beech. What happens at the sub-personal level? His musing is subserved by the endogenous activation of conception C, which is joined to Rep('hetre') by a topic-fixing link. The link is of a kind that ensures that the management of that conception is sensitive to communications which involve the use of 'hetre' and sensitive in particular to the semantic norm for using it. It is not necessary that Rep('hetre') be activated at that/moment. Only the conception is activated. Hence a subject can think a private thought with a social content at time t without having any linguistic imagery at t, and without 'silently saying' any words. If neurophysiologists were to discover that the part of Hilaire's brain that realised Rep('hetre') was inactive at the time, the hypothesis would not be discontinued. How much true information about K must a conception of K contain? Clearly, it is not necessary that the subject should know what it is for something to be a K. On the other hand, it does seem, intuitively, that S should know what Ks are. Yet the criteria we use 7
Suppose that Hilaire is the young Hilary Putnam (who was brought up as a French-speaker - see Putnam 1988, p. 11). Suppose he has another conception, linked to Rep('orme') (French for 'elm'), which happens to contain all the same descriptive information as his conception of beech, and even contains a qualitatively identical perceptual prototype (cf. the situation described in Putnam 1981, pp. 18-19; but where Putnam puts a different gloss on it). Not only is it the case that Hilaire has access to two distinct public concepts, but he also has in his head two distinct conceptions linked to different Reps. The two conceptions stay separate because Hilaire takes 'hetre' and 'orme' to be names of different kinds of trees. He resembles Loar's Paul, who takes 'chat' and 'cat' to name different kinds of animals. But whereas Hilaire is right not to merge the two files, Paul is mistaken in not merging his two files. 93
Andrew Woodfield for deciding whether a person knows what Ks are vary according to the context and purpose of our inquiry. Sometimes we are quite lax. It would be too strict to insist that S must have uniquely individuating knowledge of kind K in order to qualify.8 As a conception becomes linked to a phonological-syntactic representation of a word for K, a modicum of information about K comes free with the word-learning process. For example, syntactic clues (presence or absence of indefinite article, plural endings) can help the learner to guess whether a common noun W refers to a stuff or refers to a thing-kind. If W occurs within a phrase or a sentence, and the other words are already understood, these constrain the learner's hypotheses about the meaning of W. If the word is learned in the context of one or more ostensions (and if the teacher is reliable), the learner gleans a great deal about the kind K by scrutinising the demonstrated particular(s). But it seems that no general rule can be laid down about how much S must know about K. Our test of whether S has an adequate conception of K is flexible and pragmatic. I suggest that we are satisfied that S knows what Ks are so long as S's conception contains enough true information to sustain S across a reasonable number of sensible moves in the game of rational social intercourse. This completes my gloss on the thesis that S is able to avail himself or herself of the concept expressed by a general word W provided that S is committed to the prevailing standard for classifying things with the word and can collect and manage information about things so classified. Let us now take a brief look at the developmental process which leads a child into this fortunate state. Acquiring a concept amounts to acquiring an ability to use something abstract that is offered by the socio-linguistic environment. If the child's mind is suitably prepared and the conditions conspire in favour (e.g. if the word-learning situations feed her with enough true information about the things), then, whenever she appropriates a new general term into her working vocabulary, she 'acquires' the concept that the term expresses. Or so I shall argue. You may not agree with my theory of what is in the head, but you surely agree that some story about what is inside the head is needed, if we are to account for conceptual thinking. And equally, some story is needed to explain the causal-developmental process by which a child gains the needed inner organisation and skills. I believe that normal children start to join the club at a very early age, soon after the onset of the vocabulary explosion (which may come 8 This strict principle concerning kind-conceptions would be an analogue of Russell's principle concerning thoughts about particulars, as interpreted by Evans (1982). 94
Social Externalism and Conceptual Diversity as early as eighteen months after birth). This is the period when they are setting up stable internal representations of the commonest words in the local lexicon. The process of first language learning can be regarded as a series of preparations to become a member of two clubs in tandem. The major club is the association of rational discourse. Its membership is worldwide. The minor club is the neighbourhood of speakers of a particular language L in which the child is brought up. By joining both the rational discourse association and the L-users' club at more or less the same time, the child becomes qualified to take part in rational discourse conducted in L. Think of a population of human beings as comprising a social community if every member of the population is tied by social bonds (of kinship or marriage, of love, affection and trust) to some other members of the population. Infants are members of a social community from birth in virtue of their social relationships to relatives, carers and others with whom they establish emotional ties. Infants are not yet members of either of the two clubs, but they are primed for entrance into both. It is possible for a member of a social community to belong to one club without belonging to the other. Adults who immigrate into the community are new entrants. Such adults will already be capable of rational discourse, and will be fluent in their native language. But if the social community they join has L as its only community language and they do not know L, they are not members of the minor club. Conversely, some mentally insane members of the social community may utter well-formed sentences of L but be unable to think rationally or engage in rational discourse. They are not members of the major club. Preverbal infants, rational non-L speakers, and nonrational L-speakers are unable to play the game with members of the surrounding social community. With normal children the processes of joining the two clubs happen incredibly fast. This is because human infants possess some of the qualifying capacities innately, and they are innately programmed to acquire some of the other capacities. The evidence is now substantial that they have an innately structured module dedicated to the processing of specifically linguistic stimuli (see Pinker 1994 for an overview). Of course, the specific vocabulary and grammar and phonology of L have to be learned; but whichever language L happens to be, it will exhibit its distinctiveness in a highly constrained way within a set of parameters that are common to all human languages. Human infants 'expect' these parameters to be there, and they are predisposed to adjust the settings quickly on the basis of the slightest cues present in the frag95
Andrew Woodfield ments of L to which they are exposed. Thus by the age of two, most children have discovered most of the grammar. They also have areas of memory 'waiting' to be filled by lots of lexical items; during the vocabulary spurt, the child has only to hear a word-sound once to remember it. Children are also predisposed to subscribe to communal concepts. Many of the cognitive capacities that are preconditions for joining the rational thinkers club appear to be innate. For example, the capacity to represent objects as three-dimensional continuants occupying locations in space, the capacity to remember landmarks, to solve spatial problems and so on, all these elements of primitive cognition are manifested from a very early age (see Spelke 1988). Moreover, normal children naturally trust the ones who nurture them, they engage in joint cooperative activities, and develop sensitivity to the approval and disapproval of others. These tendencies, rooted in social instincts, are the origin of the capacity to appreciate norms. Understanding the difference between right and wrong is not just central to morality. It is also needed for symbolic categorisation, and indeed for rationality. Clearly, much more needs to be said about the steps in the process, but the general picture of conceptual development, as viewed from the perspective of social externalism about concepts, looks pretty plausible. Human infants are biologically designed so that they can latch on quickly to accessible public concepts.
IV. Lexical Diversity Creates Conceptual Diversity How does this picture relate to the hypothesis of linguistic relativity proposed by Sapir and Whorf? In his paper 'Science and Linguistics', Whorf writes 'We cut nature up, organize it into concepts, and ascribe significances as we do, largely because we are parties to an agreement to organize it in this way - an agreement that holds throughout our speech community and is codified in the patterns of our language' (Whorf 1956, p. 213). He goes on to formulate his 'new principle of relativity': 'all observers are not led by the same physical evidence to the same picture of the universe, unless their linguistic backgrounds are similar, or can in some way be calibrated' (ibid., p. 214). My view of concept-acquisition makes language the independent variable, mind the dependent variable. The direction of determination is the same as that which is proposed by the Sapir—Whorf hypothesis. I also agree with Sapir and Whorf that certain differ96
Social Externalism and Conceptual Diversity ences between languages would give rise to different conceptual repertoires. Is it the case that languages differ from one another in a relevant way? It is an undeniable empirical fact that languages vary in their lexical classifications. There are currently six thousand or so languages spoken in the world. If we select any pair of languages and compare their lexicons, it is likely that we shall find categories that are lexicalised in one language but not lexicalised in the other. Even closely related languages such as English and French do not have equivalent lexicons. In English, for example, the word 'mouth' includes animal mouths and human mouths in its extension; in French, the word 'gueule' is used for animal mouths, while 'bouche' is the word for human mouths. French does not have a single noun whose proper (literal, non-slang) extension coincides with that of 'mouth'. Bororo Indians have names for varieties of Amazonian plants for which English has no names. Greenberg (1954) reports that English-speakers can name a category - parrot — for which the Bororo language has no name. The more distant two languages are, the more taxonomic mismatches there are likely to be. Such linguistic differences causally and individuatively determine different repertoires of concepts. My account of concept-acquisition is developmental-cum-individuative. The individuative component stems from the social externalist assumption that the lexical-semantic norms of L fix the identities of the concepts that S latches on to. The developmental component is a hypothesis of partial causal determination: S's exposure to stimulation in L causes S to subscribe to the concepts that L's words express. Linguistic stimulation and feedback from others is only part of the cause. Other causal factors include S's human biological and psychological propensities. The instincts which underlie the capacity for symbolic communication channel S's development so as to bring it about reliably that S ends up respecting the same taxonomic standards as his or her carers and friends, whatever those standards happen to be. Speakers of LI learn to categorise things in the ways that LI enables them to mark; if language L2 lexicalises a different range of categories, L2-learners get channelled into those instead. So not only do the two language-communities play with differently shaped pieces, but also the games they play are slightly different. As previously noted, there is a worldwide association of players, and the child joins it when she knows how to make moves in the local language. But the local language happens to belong to people who are playing a parochial version of the game. Their version allows certain 97
Andrew Woodfield types of moves that are not played elsewhere. So although it is true that the child joins the universal association, she enters by joining the local branch.9 Is my view a version of the Sapir-Whorf hypothesis? It depends what you take the hypothesis to be. There are at least three reasons to be cautious. Firstly, neither Sapir nor Whorf gave any indication of being content-externalists. They never addressed the issue of conceptindividuation. So they probably did not envisage the developmental-cum-individuative account. It is possible that they would have liked it if they had thought of it. But they say nothing very precise about how language is supposed to determine thought. Secondly, both Sapir and Whorf downplayed lexical semantics.10 Their big claim was that grammar affects thought. Whorf's claim that the Hopi Indians have a different conception of time from Westerners was based upon his analysis of tense and aspect in Hopi verbs, not upon an inquiry into which verbs are in the Hopi lexicon. Since my hypothesis concerns only lexical semantics, it has nothing to say about the more ambitious claim about grammar. Sapir and Whorf certainly also claimed that the lexicon of L constrains the concepts that are expressible in L." But, for them, this was no big deal - a sentiment echoed again and again by commentators.12 I beg to differ. In my view, lexically-driven conceptual diversity is useful to humanity and urgently needs to be conserved. But there is no space here to elaborate on this. The third qualification is about the dependent variable. Sapir and Whorf did not claim that language infects only trained, high-level thoughts. They claimed that language structures the whole of men' This point fits in with Quine's doctrine that meaning is inseparable from theory. As Follesdal, interpreting Quine, notes, 'The special problem with semantics, which makes it different from phonetics/phonology and syntax, is that we learn the semantics of a language together with our teachers' theory of the world' (Follesdal 1990, p. 102). 10 See Sapir (1921), pp. 218-19. Whorf (1956, p. 258) says, 'Because of the systematic, configurative nature of higher mind, the "patternment" aspect of language always overrides and controls the "lexation" (Nama) or name-giving aspect.' 11 Cf. Whorf (1956), p. 240: 'Languages differ not only in how they build their sentences but also in how they break down nature to secure the elements to put in those sentences. This breakdown gives units of the lexicon'. 12 'It goes without saying that the mere content of language [i.e. vocabulary - AW] is intimately related to culture...But this superficial and extraneous kind of parallelism is of no real interest to the linguist' (Sapir 1921, p. 219). See also Black (1969), p. 32, and Davidson (1974). 98
Social Externalism and Conceptual Diversity tal life, including perceptions and primitive cognition.13 Thus their general hypothesis that language structures the mind includes the specific hypothesis that language structures perceptual experience. The latter amounts to saying, in the idiom of contemporary cognitive science, that linguistic input penetrates the perceptual systems and alters the ways in which those systems process sensory information. For the past thirty years the establishment has taken this controversial sub-hypothesis to be the core of Sapir and Whorf 's position. In their attempts to test it experimentally, psychologists have devised ways of probing the subjects' perceptual discriminations (with particular emphasis upon colour). Great efforts have been made to design tasks in which the subjects' responses do not involve language at all. The rationale for this self-denying ordinance has been that if language is the independent variable and perception is the dependent variable, it would be question-begging to measure the dependent variable by looking at the subjects' linguistic responses. Psychologists say: 'Of course the subjects' reports of their experiences will be tainted by language. We want to find out if perceptual processing mechanisms have been affected by language. Therefore, we must cordon the language-system off and tap into the upstream processes by other routes.' If the hypothesis of linguistic relativity is essentially this core hypothesis about perception, then the position that I am defending is not a version of it. I do not know whether language-exposure has any medium- or long-term effect upon people's perceptual systems or upon their low-level cognition. These seem to me to be open empirical questions. My claim is only that the language that S learns affects S's trained conceptual thinking, via the developmental-cumindividuative route that I have sketched. I should like to end by considering a serious objection that a philosopher might urge against the linguistic relativity hypothesis (cf. Davidson 1974). The objection hits directly at the part of the hypothesis that I agree with, namely, the thesis concerning concepts and the lexicon. It runs as follows. Take any language of sufficient size and complexity - that is, any actual human language - which lacks a word for so-and-so. The fact that it lacks a single term doesn't handicap the speakers, if they can construct a phrase that expresses the same idea. Suppose there is no generic term in Bororo that translates the English word 'parrot'. 13 'Language is heuristic..its forms predetermine for us certain modes of observation and interpretation...[Language does not] stand apart from or run parallel to direct experience but completely interpenetrates with it' (Sapir 1957, pp. 7-8). 99
Andrew Woodfield Suppose, though, that the Bororo language has names for all the different species of parrots - which happen to be species A, B, C and D. Bororo also has a word that means 'bird', and it has the combinatorial resources to construct a complex expression - call it 'Com' - which translates the English expression 'genus of birds which comprises A, B, C and D'. This expression is intersubstitutable for 'parrot' and arguably it expresses the same concept. So the Bororo people have an expression in their language which expresses the concept parrot. Therefore, they have the concept even though they don't have one word to express it. The objection rests upon an assumption that the concept expressed by the phrase 'genus of birds which comprises A, B, C and D' is identical with the concept expressed by 'parrot'. Although this assumption could be questioned, I am willing to concede it. For I think I can rebut the objection by taking a different tack. This is just the sort of situation where my cognitive model can be put to good use. Both the objector and I agree that the Bororo language has a compositional semantics. The complex expression 'Com' has a satisfaction-condition derived from the meanings of the component words and their mode of combination. So I agree that the language has the necessary resources to express the concept parrot. I submit, however, that the complex expression is almost certainly not one which any Bororo Indian has ever heard or used. None of them will have formed a Rep(Com), that is, a stable, perceptually derived internal representation of that phonological-syntactic unit. Consequently, none of the conceptions residing in a Bororo Indian's head is structurally linked to an internal representation of that expression. There is no mental file whose management is controlled by the subject's allegiance to the semantic norms governing 'Com'. I submit that such a subject does not possess a conception targeted upon the category parrot, and that he does not avail himself of the concept parrot.
The example shows just how unhelpful it is to keep pressing the question 'Does he possess the concept?' Possession is not the relation that matters, when you take a social externalist view of concepts. There are various relations that can hold between an individual S and a concept that is expressible in a language that S knows. The language can make a certain concept potentially available to a population of speakers, in the sense that fluent speakers, in virtue of their knowledge of the language, have no need to learn anything new in order to start subscribing to the concept. All they need is an incentive to make the construction, to do the mental work. But it is quite another thing for the concept in question to be one that helps 100
Social Externalism and Conceptual Diversity them to organise their conceptions of things. And I think Whorf was probably alluding to conceptions when he said that words determine our habitual modes of thinking.
References Bach, K. 1987. Thought and Reference. Oxford: Clarendon Press 1988. 'Burge's New Thought Experiment: Back to the Drawing Room', Journal of Philosophy 85, 88-97 Black, M. 1969. 'Some Troubles with Whorfianism', in S. Hook (ed.), Language and Philosophy. New York University Press Burge, T. 1979. 'Individualism and the Mental' in P. French, T. Uehling and H. K. Wettstein (eds.), Midwest Studies in Philosophy Vol IV: Studies in Metaphysics. Minneapolis: University of Minnesota Press 1982. 'Two Thought Experiments Reviewed', Notre Dame Journal of Formal Logic 23, 284-93 1986. 'Intellectual Norms and Foundations of Mind', Journal of Philosophy 83, 697-720 Crane, T. 1991. 'All the Difference in the World', The Philosophical Quarterly 41, 1-25 Davidson, D. 1974. 'On The Very Idea of a Conceptual Scheme', in Davidson (1984) 1984. Inquiries into Truth and Interpretation. Oxford: Clarendon Press Evans, G. 1982. The Varieties of Reference. Oxford: Clarendon Press Fodor, J. 1982. 'Cognitive Science and the Twin Earth Problem', Notre Dame Journal of Formal Logic 23, 98—118
Follesdal, D. 1990. 'Indeterminacy and Mental States', in R. B. Barrett and R. F. Gibson (eds.), Perspectives on Quine. Oxford: Blackwell Greenberg, J. 1954. 'Concerning Inferences from Linguistic to Nonlinguistic Data', in H. Hoijer (ed.), Language in Culture. University of Chicago Press Loar, B. 1988. 'Social Content and Psychological Content', in D. D. Merrill and R. H. Grimm (eds.), Contents of Thought. Tucson: University of Arizona Press Patterson, S. 1990. 'The Explanatory Role of Belief Ascriptions', Philosophical Studies 59, 313-32 Pinker, S. 1994. The Language Instinct: How the Mind Creates Language. London: Penguin Putnam, H. 1981. Reason, Truth and History. Cambridge University Press 1988. Representation and Reality. Cambridge, MA: MIT Press Ryle, G. 1979. On Thinking. Oxford: Blackwell Sapir, E. 1921. Language. (Page-numbers from the 1970 British edition, London: Hart-Davis) 1957. Culture, Language and Personality. Berkeley and Los Angeles: University of California Press 101
Andrew Woodfield Spelke, E. S. 1988. 'The Origins of Physical Knowledge', in L. Weiskrantz (ed.) Thought Without Language. Oxford: Clarendon Press Whorf, B. L. 1956. Language, Thought, and Reality. Cambridge, MA: MIT Press Woodfield, A. 1991. 'Conceptions', Mind 100, 547-72 1993. 'Do Your Concepts Develop?' in C. Hookway and D. Peterson (eds.), Philosophy and Cognitive Science. Cambridge University Press 1996. 'Which Theoretical Concepts Do Children Use?' Philosophical Papers 25, 1-20
102
The Explanation of Cognition JOHN R. SEARLE The Problem What sorts of systematic explanations should we and can we seek in cognitive science for perception, language comprehension, rational action and other forms of cognition? In broad outline I think the answer is reasonably clear: We are looking for causal explanations, and our subject matter is certain functions of a biological organ, the human and animal brain. As with any other natural science there are certain assumptions we have to make and certain conditions that our explanations have to meet. Specifically, we have to suppose that there exists a reality totally independent of our representations of it (in a healthier intellectual era it would not be necessary to say that) and we have to suppose that the elements of that reality that we cite in our explanations genuinely function causally. Not all functions of the brain are relevant to cognition, so we have to be careful to restrict the range of brain functions we are discussing. Cognitive science is about the cognitive functioning of the brain and its relation to the rest of the organism and to the rest of the world in the way that nutrition science is about the digestive functioning of the digestive system and its relation to the rest of the organism and the rest of the world. Like other organs, and indeed like other physical systems, the brain has different levels of description and cognitive science is appropriately concerned with any level of description of the brain that is relevant to the causal explanation of cognition. These can range from conscious processes of decisionmaking, at the top level, to the molecular structure of neurotransmitters, at the bottom. Typically, the higher levels will be causally emergent properties of the behaviour and organisation of the elements of the brain at the lower levels. Consider an obvious, commonsense example of an explanation at one of these higher levels. If I explain my driving behaviour in Britain by saying I am following the rule 'Drive on the left', I have given a genuine causal explanation by citing a mental process. The operation of the rule is itself caused by lower-level neuronal events in the brain and is realized in the brain at a level higher than that of individual neurons. In what I hope is an unmysterious sense of 'emergent property' the operation of the rule in 103
John R. Searle producing my behaviour is a causally emergent property of the brain system. Another way to put this same point is to say: we can give genuine causal explanations that are not at the bottom level, not at the level of neurons, etc., because the higher levels of explanation are also real levels. Talk of them is not just a manner of speaking or a metaphor. In order to be a real level, a putative causal level has to be appropriately related to the more fundamental levels, for example by being a causally emergent property of those levels. Let us call this constraint, namely that in explaining cognition we have to cite real features of the real world which function causally, the causal reality constraint.
So, just to summarise these constraints, we are seeking causal explanations of brain functioning at different levels of description. We
allow ourselves complete freedom in talking about different levels of description, but that freedom is constrained by the requirement that the levels be causally real. The claim I want to defend here is that some, though of course not all, of the explanatory models in cognitive science fail to meet the causal reality constraint. I will also suggest some revisions that will enable the explanations to meet that constraint. Marr's Version of the Information-Processing Model My dog, Ludwig, is very good at catching tennis balls. For example, if you bounce a tennis ball off a wall, he is usually able to leap up and put his mouth at precisely the point the ball reaches as he grasps it in his teeth. He doesn't always succeed, but he is pretty good at it. How does he do it? According to the current explanatory models in cognitive science, Ludwig performs an information-processing task of enormous complexity. He takes in information in the form of a 2D pattern on his retina, processes it through the visual system until he produces a 3D representation of the external world, and inputs that representation into the motor output system. The computations he is performing, even for the motor output module, are no trivial matter. Here is a candidate for the first formulation of one of the algorithms. Ludwig is unconsciously following the rule: Jump in such a way that the plane of the angle of reflection of the ball is exactly equal to the plane of the angle of incidence of impact, and put your mouth at a point where the ball is in a parabolic arc, the flatness of whose trajectory and whose velocity is a function of impact velocity times the coefficient of elasticity of the tennis ball, minus a certain loss due to air friction. That is, on the standard computational 104
The Explanation of Cognition model of cognition, Ludwig unconsciously computes a large number of such functions by unconsciously doing a lot of mathematics. In form, the explanation of his behaviour is just like that of the person who follows the rule 'drive on the left' except for the fact that there is no way even in principle that he could become consciously aware of the operation of the rule. The rules are not just not present to consciousness in fact, they are not even the sort of rules he could become aware of following. They are what I have called 'deep unconscious' rules (Searle 1992, ch. 7). I have never been completely satisfied with this mode of explanation. The problem is not just that it attributes an awful lot of unconscious mathematical knowledge to Ludwig's doggy brain, but more importantly, that it leaves out the crucial element that Ludwig is a conscious rational agent trying to do something. The explanatory model seems more appropriate for someone building a machine, a robot canine, that would catch tennis balls. I think in fact that the intuitive appeal of the approach is that it would predict Ludwig's behaviour and it is the sort of information we would put into a robot if we were building a robot to simulate his behaviour. So let us probe a bit deeper into the assumptions behind this approach. The classic statement of this version of the cognitive science explanatory paradigm is due to David Marr (Marr 1982), but there are equivalent views in other authors. On this paradigm cognitive science is a special kind of information-processing science. We are interested in how the brain and other functionally equivalent systems, such as certain kinds of computers, process information. There are three levels of explanation. The highest is the computational level, and this Marr defines in terms of the informational constraints available for mapping input information to output information. In Ludwig's case the computational task for his brain is to take in information about a two-dimensional visual array and output representations of muscle contractions that will get his mouth and the tennis ball at the same place at the same time. Intuitively I think Marr's idea of the computational level is clear. If you were instructing a computer programmer to design a program, the first thing you would tell him is what job you want the program to do. And the statement of that job is a statement of the computational task to be performed at the computational level. How is it done? Well, that leads to the second level, which Marr calls the algorithmic level. The idea is this. Any computational task can be performed in different ways. The intuitive idea is that the algorithmic level determines how the computational task is per105
John R. Searle formed by a specific algorithm. In a computer we would think of the algorithmic level as the level of the program. One puzzling feature of cognitive science versions of this level is the doctrine of recursive decomposition. Complex levels decompose into simpler levels until the bottom level is reached and at that level it is all a matter of zeros and ones, or some other binary symbols. That is, there is not really a single intermediate algorithmic level but rather a series of nested levels that bottom out in primitive processors, and these are binary symbols. And the bottom level is the only one that is real. All the others are reducible to it. But even it has no physical reality. It is implemented in the physics, as we will see, but the algorithmic level makes no reference to physical processes.
I used to think that the computations I gave for Ludwig might be the causally real level on this model, but not so. All Ludwig is really doing is manipulating zeros and ones. All the rest is mere appearance. Any computational claim we make about Ludwig reduces to the claim that he is manipulating zeros and ones. The bottom level for Marr is the level of implementation, how the algorithm is actually implemented in specific hardware. The same program, the same algorithm, can be implemented in an indefinite range of different hardwares, and it is quite possible, for example, that a program implemented in Ludwig's brain might also be implemented on a commercial computer. So, on Marr's tripartite model you get the following picture. Cognitive science is essentially the science of information-processing in a very special sense of that notion, and it is primarily concerned with explaining the top level by the algorithmic level. What matters for cognitive science explanation is the intermediate level. Why? Why should we explain brains at the intermediate level and not at the hardware level? The answer is given by my initial characterisation of the brain as a functional system. Where other functional systems are concerned, such as cars, thermostats, clocks and carburettors, we are interested in how the function is performed at the level of function, not at the level of microstructure. Thus in explaining a car engine we speak of pistons and cylinders and not of the subatomic particles of which the engine is composed; because, roughly speaking, any old subatomic particles will do as long as they implement the pistons and the cylinders. In Ludwig's case we are interested in the unconscious rule he is actually following and not in the neuronal implementation of that rule-following behaviour. And the rule he is actually following must be statable entirely in terms of zeros and ones, because that is all that is really going on. So on this conception my earlier characterisation of cognitive science as a science of brain function at a certain level or levels of description was 106
The Explanation of Cognition misleading. Cognitive science is a science of information-processing, which happens to be implemented in the brain but which could equally well be implemented in an indefinite range of other hardwares. Cognitive science explains the top level in terms of the intermediate level but is not really concerned with the bottom level except in so far as it implements the intermediate level. One problem with Marr's tripartite analysis of cognitive function is that just about any system will admit of this style of analysis. And the point is not just that clocks, carburettors, and thermostats admit of the three-level analysis, (this is welcomed by adherents of the classical model as showing that cognition admits of a functional analysis similar to that of clocks, etc.) The problem is that any system of any complexity at all admits of an information-processing analysis. Consider a stone falling off the top of a cliff. The 'system', if I may so describe it, has three levels of analysis. The computational task for the stone is to figure out a way to get to the ground in a certain amount of time. It has to compute the function S=^gt2. At the intermediate level there is an algorithm that carries out the task. The algorithm instructs the system as to what steps to go through to match time and space in the right way. And there is the familiar hardware implementation in terms of masses of rock, earth and intervening air. So why isn't the falling stone an information-processing system? But if the stone is, then everything is. This is a crucial question for cognitive science and several authors have answered it. According to them, we need to distinguish between a system being describable by a computation and one actually carrying out the computation. The system just mentioned is describable by a computable function, but it does not carry out that computation because (a) there are no representations for the computation to operate over and (b) a fortiori there is no information encoded in the representations. A genuine science of cognition, an information-processing science, requires computations performed over symbols or other syntactical elements, and these are the representations which encode information which is processed by the algorithm. These conditions are not met by a falling rock, even though the rock is computationally describable. If we are going to make this reply stick, we will need a satisfactory definition or account of 'information', 'representation', 'symbol' and 'syntax', not to mention 'computation' and 'algorithm'. And these accounts must enable us to explain how information, representation, etc., gets into the system in such a way as to satisfy the causal reality constraint. The account will have to show how information gets into the system in some intrinsic form in the first place, 107
John R. Searle and then retains its character as information throughout the processing. Furthermore the account would have to show how the real information-processing level is an emergent property of the more fundamental micro-levels. To nail this down to specific cases, it is not going to be enough to say, as Marr did, that there is a twodimensional visual array on the retina as an input to the system, we now have to say what fact about that visual array makes it information, and what exactly the content of the information is. I have looked at a lot of the literature on this issue, and I cannot find a satisfactory definition of representation or information or the other notions that will solve our problems. To their credit, Palmer and Kimchi (cited in Palmer, forthcoming) admit that they have not the faintest idea what information, in their sense, might be. I want to explore the notion of information a little more fully. The basic question of this paper is: can we give any empirical sense to the basic concepts of the information-processing model that would make the information-processing version of cognitive science a legitimate empirical discipline? Following a Rule If we are going to be clear about the claim that the cognitive agent is following unconscious rules we first have to understand what is involved in rule-following behavior. Consider a case where it seems clear and unproblematic that the agent is following a rule. When I drive in England, I follow the rule: Drive on the left hand side of the road. And if I stay in England for any length of time, I find that I get used to driving on the left and I don't have to think consciously of the rule. But it seems natural to say that I am still following the rule even when I am not thinking about it. Such an explanation meets the causal reality constraint. When I say that I am following a rule I am saying that there is an intrinsic intentional content in me, the semantic content of the rule, that is functioning causally to produce my behavior. That intentional content is at an emergent level of brain processing. The rule has the world-to-rule direction of fit and the rule-to-world direction of causation. I want to explore some of the features of this type of explanation to see whether they can be preserved in Marr-style informationprocessing cognitive science. I will simply list what seem to me some important features of rule-following behaviour: 1 . The single most important feature is the one I just mentioned. The intentional content of the rule must function causally in the production of the behaviour in question. To do this it must be at an 108
The Explanation of Cognition emergent level of brain functioning. This is how rule explanations in real life meet the causal reality constraint. Any rule-following explanation in cognitive science also has to meet that constraint. 2. Rule-following is normative from the point of view of the agent. The content of the rule determines for the agent what counts as right and wrong, as succeeding or failing. 3. The next feature is a consequence of the first. The rule must have a certain aspectual shape, what Frege called the 'mode of presentation'. This is why extensionally equivalent rules can differ in their explanatory force. I can be following one rule and not another, even though the observable behaviour is the same for both cases. For this reason, that rule explanations must cite specific aspectual shapes, rule explanations are intensional-with-an-s. For example, the rule 'On two-lane roads drive on the left' is extensionally equivalent to the rule 'Drive so that the steering wheel is nearest to the centre line of the road', given the structure of British cars that I drive. But in Britain, I follow the first rule and not the second, even though each would equally well predict my behaviour. 4. In ordinary rule-governed behaviour the rules are either conscious or accessible to consciousness. Even when I am following the rule unthinkingly, still I could think about it. I am not always conscious of the rule, but I can easily become conscious of it. Even if the rule is so ingrained in my unconscious that I cannot think of it, still it must be the sort of thing that could be conscious. 5. Accessibility to consciousness implies a fifth requirement. The terms in which the rule is stated must be terms that are in the cognitive repertoire of the agent in question. It is a general characteristic of intentionalistic explanations, of which rule explanations are a special case, that the apparatus appealed to by the rule must be one that the agent is in possession of. If I wish to explain why Hitler invaded Russia, I have to use terms that are part of Hitler's conceptual repertoire. If I postulate some mathematical formula that Hitler never heard of and couldn't have mastered, and couldn't have been aware of, then the explanation cannot be an intentionalistic explanation. It is a peculiarity of cognition, often remarked on by people who discuss the special features of historical explanation, that explanations that appeal to cognitive states and processes must employ concepts available to the agent. 6. The next feature is seldom remarked on: rule-following is normally a form of voluntary behaviour. It is up to me whether I follow the rule or break it. The rule does indeed function causally but the rule as cause, even the rule together with a desire to follow the rule, does not give causally sufficient conditions. This is typical of rational explanations of behaviour. It is often 109
John R. Searle said that actions are caused by beliefs and desires, but if we take that to be a claim about causally sufficient conditions, it is false. A test of the rationality of the behaviour is that there is a gap between the intentional contents (beliefs, desires, awareness of rules, etc.) and the actual action. You still have to haul off and do the thing you have decided to do, even in cases where the rule requires you to do it. I am going to call this gap between the rule and other intentional phenomena which are the causes, and the action which is their effect, the 'gap of voluntary action' or simply 'the gap'. 7. A feature, related to the gap, is that rules are always subject to interpretation and to interference by other rational considerations. So, for example, I don't follow the rule, 'drive on the left hand side of the road', blindly. If there is a pothole, or a car parked blocking the road, I will swing around it. Such rules are in this sense ceteris paribus rules. 8. The final feature is that the rule must operate in real time. For actual rule-governed behaviour, the rule explanation requires that the time of the application of the rule and the time of the causal functioning are co-extensive. Just to summarise, then, we have eight features of intentionalistic rule explanations. First, the intentional content of the rule must function causally; second, the rule sets a normative standard for the agent. Third, rules have aspectual shape and so rule explanations are intensional-with-an-s. Fourth, the rule must be either conscious or accessible to consciousness. Fifth, rules must have semantic contents in the cognitive repertoire of the agent. Sixth, rule-governed behaviour is voluntary, and therefore, because of the gap of voluntary behaviour, the rule explanation does not give causally sufficient conditions. Seventh, rules are subject to interpretation and to interference by other considerations. And finally, eighth, the rule must operate in real time. Let's compare this with Marr-style cognitivist forms of explanation. In such explanations only features 1 and 3 are unambiguously present. Now one problem with the causal reality constraint on cognitive science explanations is that it is not clear how you can have those two without any of the other six. It is no accident that these features hang together, because rule-following explanations are typical of intentionalistic explanations of rational behavior. How can it be literally the case that Ludwig is following a rule with a specific semantic content, if that rule is not normative for him, is not accessible to his consciousness even in principle, has concepts totally outside his repertoire, is not voluntarily applicable, is not subject to interpretation and appears to operate instantaneously rather than in real time? 110
The Explanation of Cognition Some Preliminary Distinctions In this section I want to remind you of certain fundamental distinctions. First we need to recall the familiar distinction between rule-governed or rule-guided behaviour, on the one hand, and ruledescribed behaviour, on the other. When I follow a rule, such as the rule of the road in England, drive on the left hand side, the actual semantic content of the rule plays a causal role in my behaviour. The rule does more than predict my behaviour, rather it is part of the cause of my behaviour. In this respect it differs from the laws of nature which describe my behaviour including its causes, but which do not cause the behaviour they describe. The distinction between rule-guided and rule-described can be generalised as a distinction between intentionality-guided and intentionality-described. All descriptions have intentionality, but the peculiarity of intentionalistic explanations of human cognition is that the intentional content of the explanation functions causally in the production of the explanandum. If I say 'Sally drank because she was thirsty' the thirst functions causally in the production of the behaviour. It is important to remind ourselves of this distinction because if an information-processing cognitive science is to meet the constraint, the intentionality of the information must not merely describe but must function causally in the production of the cognition that the information-processing explains. Otherwise there is no causal explanation. To meet the
causal reality constraint, the algorithmic level must function causally. I believe that the standard cognitive science accounts acknowledge this point when they distinguish between being describable by a function and actually computing a function. This a special case of the general distinction between rule-described and rule-guided. The second important distinction is between observer-relative and observer-independent features of reality. Basic to our whole scientific world-view is the distinction between those features that exist independently of any observer, designer or other intentionalistic agent and those that are dependent on observers, users, etc. Often the same object will have both sorts of features. The objects in my pocket have such observer-independent features as a certain mass and a certain chemical composition, but they also have observer-relative features: for example, one is a British £10 note and another is a Swiss Army knife. I want to describe this distinction as that between features of the world that are observer- (or intentionality-) relative, and features that are observer- (or intentionality-) independent. Money, property, marriage, government and correct English pronunciation as well as knives, bathtubs and motor cars are observer-relative; force, mass, and gravitational attraction are observer-independent. Ill
John R. Searle 'Observer-relative' does not mean arbitrary or unreal. The fact that something is a knife or a chair or a nice day for a picnic is observer-relative but it is not arbitrary. You can't use just anything as a knife or a chair or a nice day for a picnic. The point about observer-relativity is that observer-relative features, under those descriptions, only exist relative to human observers. The fact that this object in my hand has a certain mass is not observer-relative but observer-independent. That the same object is a knife is relative to the fact that human agents have designed it, sold it, used it, etc. as a knife. Same object, different features: some features observerindependent, some observer-relative. It is characteristic of the natural sciences that they deal with observer-independent features - such as force, mass, the chemical bond, etc. - and it is characteristic of the social sciences that they deal with observer-relative features, such as money, property, marriage and government. As usual, psychology falls in the middle. Some parts of psychology deal with observer- relative features, but cognitive psychology, the part that is the core of cognitive science, deals with observer-independent features such as perception and memory. Wherever there is an observer-relative feature, such as being a knife or being money, there must be some agents who use or treat the entities in question as a knives or as money. Now, and this is an important point, though money and knives are observer-relative, the fact that observers treat certain objects as money or knives is not observer-relative, it is observer-independent. It is an intrinsic fact about me that I treat this object as a knife, even though the fact that this object is a knife only exists relative to me and other observers. The attitudes of observers relative to which entities satisfy observer-relative descriptions are not themselves observer-relative. This is why social science explanations can satisfy the causal reality constraint even though the features appealed to are observer-relative features. So for example, if I say 'The rise in American interest rates caused a rise in the exchange value of the dollar against the pound' that is a perfectly legitimate causal explanation, even though pounds, dollars and interest rates are all observer-relative. The causal mechanisms work in such an explanation even though they work through the attitudes of investors, bankers, money changers, speculators, etc. In that respect the rise in the value of the dollar is not like the rise in the pressure of a gas when heated. The rise in pressure of a gas is observer-independent, the rise in the value of the dollar is observer-dependent. But the explanation in both cases can be a causal explanation. The difference comes out in the fact that the explanation of the observer-relative phenomena makes implicit reference to human agents. 112
The Explanation of Cognition The third distinction is an application of the second. It is the distinction between intrinsic or original intentionality and derived intentionality. If I am currently in a state of thirst or hunger, the intentionality of my state is intrinsic to those states - both involve desires. If I report these states in the utterances of sentences such as 'I am thirsty' or 'I am hungry', the sentences are also intentional because they have truth-conditions. But the intentionality of the sentences is not intrinsic to them as syntactical sequences. Those sentences derive their meaning from the intentionality of English speakers. Mental states such as beliefs, desires, emotions, perceptions, etc., have intrinsic intentionality; but sente'nces, maps, pictures and books have only derived intentionality. In both cases, the intentionality is real and literally ascribed, but the derived intentionality has to be derived from the original or intrinsic intentionality of actual human or animal agents. I want this distinction to sound obvious, because I believe it is. And I also believe it is a special case of the equally obvious distinction between observer-relativity and observer-independence. Derived intentionality is observer-relative, intrinsic intentionality is observer-independent. There are, furthermore, intentional ascriptions that do not ascribe either of these kinds of intentionality. These are typically metaphorical or as-if ascriptions. We say such things as 'My lawn is thirsty because we are in a drought', or 'My car is thirsty because it consumes so much gasoline'. I take it that these are harmless metaphorical claims of little philosophical interest. They mean, roughly, my lawn or my car is in a situation similar to and behaves in some ways similar to an organism that is literally thirsty. Such as-if intentionality should not be confused with derived intentionality. Derived intentionality is genuine intentionality all right but it is derived from the intrinsic intentionality of actual intentional agents such as speakers of a language. Hence, it is observer-relative. But as-if intentionality is not intentionality at all. When I say of a system that it has as-if intentionality, that does not attribute intentionality to it. It merely says that the system behaves as if it had intentionality, even though it does not in fact. To summarise these distinctions: we need to distinguish between rule-guided and rule-described behaviour. We need to distinguish observer-independent features from observer-relative features. Furthermore, we need to distinguish observer-independent (or intrinsic) intentionality, from both observer-dependent (derived) intentionality and as-if intentionality.
113
John R. Searle Information and Interpretation I now want to apply these distinctions to the information-processing model of cognitive explanation. I will argue that if the Marrstyle model is to have explanatory force, the behaviour to be explained by the information-processing rules must be rule-guided and not just rule-described. It can only meet that condition if the information is intrinsic or observer-independent. To make the distinction between Ludwig and the falling rock, we have to show that Ludwig is actually following a rule and that can only be because he has an appropriate intrinsic intentional content. The difficulty with the classical model can now be stated in a preliminary form. Every key notion in the model is observer-relative: information, representation, syntax, symbol and computation, as typically used in cognitive science, are all observer-relative notions, and this has the consequence that the classical model in its present form cannot meet the causal reality constraint. I will try to state this more precisely in what follows. Let us go through these notions, starting with 'symbols' and 'syntax'. I take it as obvious that a mark or a shape or a sound is a symbol or a sentence or other syntactical device only relative to some agents who assign a syntactical interpretation to it. And indeed, though it is less obvious, I think it is also true that an entity can only have a syntactical interpretation if it also has a semantic interpretation, because the symbols and marks are syntactical elements only relative to some meaning they have. Symbols have to symbolise something and sentences have to mean something. Symbols and sentences are indeed syntactical entities, but the syntactical interpretation requires a semantics. When we get to 'representation' the situation is a bit trickier. A representation can be either observer-relative or observer-independent. Thus maps, diagrams, pictures and sentences are all representations and they are all observer-relative. Beliefs and desires are mental representations and they are observer-independent. Furthermore an animal can have such mental representations as beliefs or desires without having any syntactical or symbolic entities at all. When Ludwig wants to eat or wants to drink, for example, he need not use any symbols or sentences at all to have his canine desires. He just feels hungry or thirsty. The tricky part comes from the fact that sometimes observer-independent beliefs and desires make use of sentences, etc. which are observer-relative. Indeed some philosophers have said that all beliefs and desires are 'propositional attitudes' in the sense of being attitudes towards propositions or sentences or some other form of representation. I 114
The Explanation of Cognition used to think this was a harmless mistake, but it is not. If I believe that Clinton is President of the U.S. I do indeed have an attitude towards Clinton, but not towards a sentence or a proposition. The sentence 'Clinton is President of the U.S.' is used to express my belief and the proposition that Clinton is President of the U.S. is the content of my belief. But I have no attitudes towards the sentence or the proposition. Indeed the proposition, construed as believed, just is identical with my belief. It is not the object of the belief. The doctrine of propositional attitudes is a harmful mistake because it leads people to postulate a set of entities in the head, mental representations, and having a belief or desire is supposed to be having an attitude towards one of these symbolic, sentence-like entities. The point for present purposes is that intrinsic mental representations such as beliefs and desires (intentional states, as I prefer to call them) do not require some representing device, some syntactical device, in order to exist. And where there is a syntactical device, the syntactical device, being observer-dependent, inherits its status as syntactical and semantic from the intrinsic intentional content of the mind and not conversely. The crucial point for the present discussion is that all syntactical entities are observer-relative. This distinction between observer-independent and observerdependent features of the world applies to information. 'Information' is clearly an intentionalistic notion, because information is always information about something and typically the information is: that such and such is the case. Aboutness in this sense is the defining quality of intentionality, and intentional content of this propositional sort is typical of intentionality. So it should not be surprising that the distinctions between the different kinds of intentional ascriptions will apply to information. Thus if I say, 'I know the way to San Jose', I ascribe to myself information which does not depend on any observer. It is intrinsic or observer-independent. If I say 'This book contains information about how to get to San Jose', the book literally contains information, but the interpretation of the inscriptions in the book as information depends on interpreters. The information is observer-dependent. There are also as-if ascriptions of information. If I say 'These tree rings contain information about the age of the tree', that is not a literal ascription of intentionality. There is no propositional content expressed by the wood. What I am actually saying, stated literally, is that a knowledgeable person can infer the age of the tree from the number of rings, because there is an exact covariance between the number of the tree's rings and its age in years. I think that with the widespread use of the notion of 'information', particularly as a result of information theory, many people would now say that the 115
John R. Searle stump literally contains information. I think they think they are speaking literally when they say that DNA contains information. This is perfectly reasonable, but it is a different meaning of 'information', one that separates information and intentionality. There is no psychological reality to the 'information' in the tree ring or the DNA. They have neither propositional content nor intentionality in the sense in which the thoughts in my head have original intentionality and the sentences in the book have derived intentionality. Of these three types of intentional ascription only intrinsic information is observer-independent. Which type of information is appealed to in cognitive science information-processing theories? Well, 'as-if' information won't do. If the explanation is to satisfy the causal reality constraint, some actual informational fact must be appealed to. Why won't derived information satisfy the reality constraint? After all, we can give genuine scientific accounts of the flow of money in the economy; why not scientific accounts of the flow of information in the cognitive system, even though the information, like the money, is an observer-relative phenomenon? The brief answer is that in the case of economics, the agents who treat such and such physical phenomena as money are parts of the subject-matter we are studying. But in cognitive science, if we say we are giving an information-processing explanation of the agent's cognitive processes we cannot accept an explanation in which the agent's information-processing only exists relative to his intentionality, because we then have not explained the intentionality on which all of his cognitive processes depend. We will in short have committed the homunculus fallacy. If, on the other hand, we think of the information as existing relative only to us - the observer - then we have not satisfied the causal reality constraint because we have not identified a fact independent of the theory which explains the data that the theory is supposed to explain. So if cognitive science explanations are going to satisfy the causal reality constraint they are going to have to appeal to information which is intrinsic to the agent, information that is observerindependent. Computation and Interpretation Well, why must the requirement be so strong? Why can't we just say that the brain behaves like any other computer? We give causal explanations of ordinary computers, explanations which meet the causal reality constraint but which do not force us to postulate intrinsic intentionality in the computer. 116
The Explanation of Cognition The answer is that the distinction between observer-independent and observer-relative applies to computation as well. When I add 2 plus 2 and get 4 the arithmetical calculation is intrinsic to me. It is observer-independent. When I punch out '2 + 2' and get '4' on my computer, the computation is observer-relative. The electrical state transitions are just that — electrical state transitions — until an interpreter interprets them as a computation. The computation is not intrinsic to the silicon nor to the electric charges. I and others like me are the computer's homunculi. So if we say that the brain is doing computation we need to say whether it is observer-relative or observer-independent. If it is observer-independent then we have to postulate a homunculus inside the brain who is actually manipulating the symbols so as to carry out the computation, just as I am consciously manipulating arabic numerals when I add 2 plus 2 to get 4. If we say it is observer-relative then we are supposing that some outside observer is assigning a computational interpretation to the neuron-firings. I think this last point is clear if you think about it, but not everyone finds it so and I will therefore explore it a bit further. We are blinded to the observer-relativity of computational ascription because we think that since computation is typically mathematical and we also think that the world satisfies certain mathematical descriptions in an observer-independent fashion, that somehow it must follow that the computation is observer-independent. However, there is a subtle but still important distinction between the observer-independence of certain mathematically described facts and the observer-relativity of computation exploiting those facts. Consider the example I gave earlier of a rock falling off a cliff. The rock satisfies the law 5 = | gt\ and that fact is observer-independent. But notice, we can treat the rock as a computer if we like. Suppose we want to compute the height of the cliff. We know the rule and we know the gravitational constant. All we need is a stopwatch. We can then use the rock as a simple analogue computer to compute the height of the cliff. So what is the difference between the first case, where the rock is just a rock and is rule-described, and the second case where the rock is a computer carrying out a computation implementing exactly the same rule? The answer is: in the second case we have assigned - that is, there is an observer-relative assignment of - a computational interpretation. But what is true of the rock is true of every computer. What is special about the rock is that the law of nature and the implemented algorithm are the same. In a commercial computer we exploit the laws of nature to assign other algorithms to electronic processes: for addition, subtraction, word-processing, etc. 117
John R. Searle But the general principle is this: we cannot appeal to the analogy between the computer and the brain to justify the special character of the tripartite model as applied to the brain, because something is a computer only relative to a computational interpretation. What I have tried to show with the parable of the falling rock is that one and the same mathematical description can be treated both as a description of an observer-independent process, and as an observer-relative computation. It is just a fact about the stone that it falls in accordance with the laws of physics. There is nothing observer-relative about that. But if we treat that fact computationally, if we think of the stone as carrying out a computation, then that computation only exists relative to us. I think that you can see this point if I give you a simpler example. If it is a fact that there are three cows in one field and two cows in the next field, both are observer-independent facts. But if I then decide to use these facts in order to perform a mathematical calculation, and I add three plus two to get five by counting the cows, the computational process of addition is not something that is intrinsic to the cows in the field. The process of addition is a process that I perform using the cows as my adding machine. Now, what is true of the rock and the cows in the field is true of computation, generally. If I am consciously doing arithmetic, that computation is intrinsic. If a pocket adding machine is doing arithmetic, that is observer-relative. It is worth pointing out, by the way, that over the years the word 'computer' has changed its meaning. When Turing wrote his famous 1950 article, the word 'computer' meant 'person who computes'. That is why Turing called the article 'Computing Machinery and Intelligence' not 'Computers and Intelligence'. 'Computer' meant: person who computes. Nowadays, the word 'computer' has changed its meaning from an observerindependent to an observer-relative phenomenon. 'Computer' now refers to a class of artefacts. This shift in the meaning of 'computer', like the shift in the meaning of 'information', has tended to blur the distinction between intrinsic intentionality and other sorts of phenomena, and it has tended to foster the confusions I am trying to expose here. Information-Processing in the Brain The crucial question for the classical model can now be stated with more precision. What fact exactly corresponds to the claim that there is an algorithmic level of information-processing in the brain? And what fact exactly corresponds to the claim that everything 118
The Explanation of Cognition going on at this level reduces to a level of primitive processing which consists entirely in the manipulation of binary symbols? And are these computational information processes observer-independent or observer-relative? As a first step let's ask how the proponents of the model think of it themselves. The answer to that question is not as clear as it ought to be but I think the answer is something like this. At this level, the brain works like an ordinary commercial computer. Just as there are symbols in the computer and they are information-bearing, so there are sentences in the head and they are information-bearing. Just as the commercial computer is an information-processing device so is the brain. This answer is unacceptable. As we have already seen, in the commercial computer the symbols, sentences, representation, information and computation are all observer-relative. Intrinsically speaking the commercial computer is just a complicated electronic circuit. For the commercial computer to meet the causal reality constraint we have to appeal to the outside programmers, designer and users who assign an interpretation to the input, to the processes in between, and to the output. For the commercial computer, we are the homunculi who make sense of the whole operation. This sort of answer can never work for Ludwig because whatever else he is, he is a conscious intentional agent trying to do something, trying to catch a tennis ball; and all of that is intrinsic to him, none of it is observer-relative. We want to know how he works really, intrinsically, not just what sorts of stances we might adopt toward him or what computational interpretations we might impose on him. Well, why can't Ludwig be computing intrinsically; why can't he be carrying out algorithms unconsciously the way I carry out the algorithm for long division consciously? We can say that he does, but if we do we have abandoned the model, because now the explanatory causal mechanism is not the algorithm, but the mental agent inside who is intentionally going through the steps of the algorithm. This answer, in short, commits the homunculus fallacy. We don't explain Ludwig's intentionally-trying-to-catch-the-ball behaviour by an algorithm if we have to appeal to his intentionallycarrying-out-the-parabolic-trajectory-computation behaviour and then explain that in turn by his intentionally-going-through-millions-of-binary-steps behaviour. Because now the explanatory mechanism of his system remains his irreducible intentionality. The idea of the model was that the information in the system is carried along by the computational operations over the syntax. The semantics just goes along for the ride. But on this analysis it is the syntax 119
John R. Searle that is going along for the ride. The intrinsic intentionality of the agent is doing all the work. To see this point notice that the psychological explanation of my doing long division is not the algorithm, but my mastery of the algorithm and my intentionally going through the steps of the algorithm. The upshot can be stated in the form of a dilemma for the classical model: either the crucial notions are taken in an observer-relative or in an observer-independent sense. If observer-relative then the explanation fails because it fails to meet the causal reality constraint. If observer-independent then it fails because of the homunculus fallacy. The homunculus is doing the work. You get a choice between an outside homunculus (observer-relative) or an inside homunculus (observer-independent). Neither option is acceptable. Deep Unconscious Rule-Following I think one way to meet my argument would be to offer a convincing existence proof to the contrary. Are there convincing and unproblematic examples of deep unconscious computational rulefollowing? I have argued elsewhere that a specific aspectual shape requires accessibility to consciousness at least in principle. In many cases, blindsight for example, the content is not accessible to consciousness in fact, but precisely for that reason, we understand such cases as pathological, as due to deficits, repression, etc. I won't repeat that argument here but will try to ask a different question: are there any unproblematic examples of deep unconscious rule-following? If we had some convincing examples, then we would have fewer doubts about the overall principle. If we could agree that there are cases of rule-following in this technical sense, which departs from our ordinary commonsense notion of rule-following, and if we could agree further that these explanations have genuine causal explanatory power, then we would at least have a good beginning of a justification for a general cognitive science strategy of postulating such deep unconscious rule explanations. The two examples that have been presented to me are the operation of modus ponens and other logical rules, and secondly, the operation of the vestibular ocular reflex. (There is a certain irony about the VOR because I presented it as what I thought was an obvious example of a case that superficially seemed to satisfy the causal reality constraint, but where it was obvious that it didn't (Searle 1992, pp. 235-7).) I will consider each of these in turn. People clearly have a capacity for making logical inferences. They do this, so the account goes, 120
The Explanation of Cognition by following rules that they are totally unaware of and that they could not even formulate without professional assistance. So, for example, people are able to make modus ponens inferences, and thus follow the rule of modus ponens, even though they could not formulate the rule of modus ponens, and indeed, do not have the concept of modus ponens. Well, let's try this out and see how it works. Here is a typical inference using modus ponens. Before the 1996 election I believed that if Clinton could carry the state of California, he would win the election. Having looked over the poll results in California, I came to the conclusion that Clinton would carry California, so I inferred that he would win the election. Now, how did I make that inference? Well, the cognitive science explanation would go: when you made the inference you were in fact following an unconscious rule. This is the rule of modus ponens, the rule that says if you have premises of the form 'p', and 'if p then q', then you can validly infer 'q'. It seems to me, however, that in cases like this, the rule plays no explanatory role whatsoever. If I believe that Clinton will carry California, and believe that if he carries California he will win the election, that is already enough to enable me to infer that he will win the election. The rule adds nothing to the explanation of my inference. The explanation of the inference is that I can see that the conclusion follows from the premises. But doesn't the conclusion only follow from the premises because it instantiates the rule of modus ponens - doesn't it derive its validity from modus ponens? I believe the answer to these questions is: No. Modus ponens, construed as a syntactical computational rule, is simply a pattern that we use for describing inferences that are independently valid. We don't follow the rule of modus ponens in order to make the inference. Rather, we make the valid inference, and the logician can formulate the socalled rule of modus ponens to describe an infinite number of such valid inferences. But the inferences do not derive their validity from modus ponens. Rather, modus ponens derives its validity from the independent validity of the inferences. To think otherwise leads to the Lewis Carroll paradox (Carroll 1895). So, it seems to me, modus ponens plays no explanatory role whatever in an inference of the sort I just described. But what about purely formal proof-theoretic inferences? Suppose I just have a bunch of symbols and I infer from 'p' and 'p —> q' to 'q'? Now it seems to me that once we have subtracted the semantic content from the propositions, there actually is a role for the rule of modus ponens. But then precisely because there is such a rule, we are no longer talking about valid inferences as part of human cognitive processes. We are talking about a formal analogue 121
John R. Searle to these valid inferences in some formal proof-theoretic system. That is, if you are given a rule that says whenever you have symbols of the form: 'squiggle blotch squaggle', followed by 'squiggle', you can write down 'squaggle', that is a genuine rule. It tells you what you can do in certain circumstances and it has all of those features that I described as typical of rule-governed behaviour, or rule explanations - every single one. But that is precisely not the operation of the rule of modus ponens in ordinary reasoning. To put this point precisely, if we think of modus ponens as an actual description of the operation of mental contents, then modus ponens plays no explanatory role in valid inferences. If we think of it as a prooftheoretical rule describing operations on meaningless symbols, then it does indeed play a role, but its role is not that of explaining how we actually make inferences in ordinary cognitive processes, but how we can represent the formal or syntactical structure of those inferences in artificially created systems. I now turn to the vestibular ocular reflex. It looks as if we are unconsciously following the rule: 'Move the eyeball equal and opposite to the movement of the head', when in fact we are not following any such rule. There is a complex reflex mechanism in the brain that produces this behaviour. I thought the point was obvious, but not so. Recently, some of my critics have said that there are even subdoxastic computational states intrinsic to the system that are at a more fine-grained level than the rule I just stated. Martin Davies says, Another way to describe the VOR is as a system in which certain information-processing takes place, not just from head-movements of certain velocities to eye-movements of certain velocities, but from representations of head-movement velocities to representations of eye-movement velocities. It is only against the background of this second kind of description that there is any question of crediting the system with tacit knowledge of the rules relating head velocity to eye velocity. (Davies 1995, p. 386) This assumption of 'semantic content' in the input and output states is a necessary but not a sufficient condition for tacit knowledge of rules. The sufficient condition requires that 'the various input-output transitions that are in conformity with the rule should have the same causal explanation' (ibid.). The VOR easily satisfies that condition, so it turns out that the VOR is a case of unconscious tacit knowledge of rules and is a case of rule-governed behaviour. To support this, Davies gives various statements of computational descriptions of the VOR from David Robinson, Patricia Churchland and Terry Sejnowski (Churchland 122
The Explanation of Cognition and Sejnowski 1993). He thinks mistakenly that I am arguing that the computational ascriptions are trivial. But that is not my point. My point is about the psychological reality of the computational ascriptions. I see no reason to treat the computational description of the VOR any differently than the computational description of the stomach or other organs. My question is, is there a causal level distinct from the level of the neurophysiology at which the agent is actually unconsciously carrying out certain computational, information-processing tasks in order to move his eyeball? I see nothing in Davies's account to suppose that the postulation of such a level meets the causal reality constraint. What fact about the vestibular nuclei makes it the case that they are carrying out specifically mental operations at the level of intrinsic intentionality? I do not see an answer to that question. It is not an objection to the usefulness of the computational models of the VOR to point out that they are models of neurophysiology, not examples of actual psychological processes: they are at the level of observer-relative neuronal information-processing not intrinsic intentionality. It is one thing to have a computational description of a process, quite another to actually carry out a mental process of computing.
Conclusion On the account I am proposing, computational descriptions play exactly the same role in cognitive science that they play in any other branch of biology or in other natural sciences. Except for cases where an agent is actually intentionally carrying out a computation, the computational description does not identify a separate causal level distinct from the level of the physical structure of the organism. When you give a causal explanation, always ask yourself what causal fact corresponds to the claim you are making. In the case of computational descriptions of deep unconscious brain processes, the processes are rule-described and not rule-governed. And what goes for computation goes a fortiori for 'informationprocessing'. You can give an information-processing description of the brain as you can of the stomach or of an internal combustion engine. But if this 'information-processing' is to be psychologically real, it must be a form of information that is intrinsically intentionalistic. Cognitive science explanations using the deep unconscious typically fail to meet this condition. I would like to conclude this discussion with a diagnosis of what I think is the mistake. It is very difficult for human beings to accept non-animistic, non-intentionalistic forms of explanation. In our 123
John R. Searle culture we only fully came to accept such explanations in the seventeenth century. Our paradigmatic forms of explanation are intentionalistic: I am eating this food because I am hungry, I am drinking this water because I am thirsty, I am driving on the left because that is the rule of the road. The idea that there are mechanical explanations that cite no intentionality is a very hard idea to grasp. A form of animism still survives in certain research projects in cognitive science. Marr's intermediate level of rule-following at the subdoxastic level in the brain is a form of animism. Now, since these postulated processes are not conscious, are not even accessible to consciousness in principle, we postulate deep unconscious rule-following behaviour. This is the mistake of primitive animism. Now, this is aided by a second mistake: we are misled by the apparent intentionality of computers, thermostats, carburettors and other functional systems that we have designed. It seems apparent to us that these systems have an intentionalistic level of description. Indeed, standard textbooks of cognitive science give Marr's intermediate-level description of the thermostat, as if the algorithmic level explanation obviously satisfied the causal reality constraint. But I think it is clear that it does not. In the case of thermostats we have rigged up physical systems to behave as if they were following computational rules. But the intentional, rule-following computation of the thermostat is entirely observer-relative. It is only because we have designed and used these systems that we can make intentionalistic explanations at all. Now, what goes for the thermostat goes for other functional systems, such as clocks, carburettors and, above all, computers. So, we are making two mistakes. The first is a mistake of preferring animism over naturalistic explanations, and the second is the failure to make the distinction between observer-relativity and observer-independence. In particular, we fail to distinguish the cases where we have genuine intrinsic intentionality as opposed to observer-relative intentionality. The intentionality in thermostats, carburettors, clocks and computers is entirely observer-relative. The situation we are in is exactly like the following: Suppose cars occurred as natural phenomena and we did not know how they worked. We would be tempted to think that much of what they do is computational information-processing. For example we might try to explain the speedometer system by saying that it computes the speed of the car in miles per hour from the input of information about wheel rotation in revolutions per minute. We might even figure out the algorithm by which it maps rpm onto mph. But such an explanation has no causal reality at all. The actual causal mechanism is that a small electric generator is attached to a wheel in such a way 124
The Explanation of Cognition that increases in rpm produce increases in electricity generated. An ammeter with a needle moves higher or lower as the electricity increases or decreases. As far as causation is concerned, that is the entire story. Intrinsically speaking there is no computation and no information. In addition to the straight physics, the computation and the information are all in us. They are observer-relative. There is nothing wrong with observer-relative computation and information. After all, that is what we in fact designed the speedometer system for. The wrong thing is to mistake the observer-relative attribution of computation and information-processing for a causal explanation. Now, the hard thing to see is that many of the intentionalistic descriptions of brain processes are also observer-relative, and consequently do not give us a causal explanation. What then is the correct model for cognitive science explanation? And, indeed, how do we account for much of the apparent rationality of cognition if we do not postulate rule-governed behaviour at Marr's intermediate level? To answer this, it seems to me we have to remind ourselves of how Darwin solved a similar problem by showing how the apparent goal-directedness in the structure of species could be explained without postulating any intentionality. In a Darwinian style of explanation, we substitute two explanatory levels for one. Instead of saying 'the fish has the shape it has in order to survive in water', we say (1) the fish has the shape it has because of its genetic structure, and (2) fish that have that shape are going to survive better than fish that don't. Notice that survival stills functions in the explanation but it is no longer a goal. It is just something that happens. Now, analogously, we should not say 'The eyeball moves because it is following a rule of the vestibular ocular reflex.' We should say that the eyeball moves because of the structure of the visual system - it is just a mechanical process. There is no rule-following at all. The rule, however, does describe the behaviour of the eyeball, and the eyeball satisfies that description for basically Darwinian reasons. Eyeballs that behave that way are going to produce a more stable retinal image, and organisms that have a stable retinal image are more likely to survive than organisms that don't. Analogously, Ludwig does not follow the parabolic trajectory rule, rather he tries to figure where the ball is going to be and jump to put his mouth at that point. After much practice he gets rather good at it. He has paw-eye co-ordination skills that enable him to be described by the parabolic trajectory rule, but he is not following that rule. Dogs that can develop such skills are more likely to survive than dogs that don't or at least they are more likely to catch tennis balls.
125
John R. Searle References Carroll, L. 1895. 'What the Tortoise Said to Achilles', Mind 14, 278-80 Churchland, P. S. and Sejnowski, T. J. 1993. The Computational Brain. Cambridge, MA: MIT Press Davies, M. 1995. 'Reply: Consciousness and the Varieties of Aboutness', in C. MacDonald and G. MacDonald (eds), Philosophy of Psychology: Debates on Psychological Explanation. Oxford: Blackwell Marr, D. 1982. Vision. San Francisco: Freeman and Co. Palmer, S. E. (forthcoming). Vision Science: An Interdisciplinary Approach. Cambridge, MA: MIT Press Searle, J. R. 1992. The Rediscovery of the Mind. Cambridge, MA: MIT Press
126
Thought without Language: Thought without Awareness? L. WEISKRANTZ Some philosophers have laid down rather severe strictures on whether there can be thought without language. Wittgenstein asserted that 'the limits of language...mean the limits of my world' (1922, §5.62). Davidson (1984, p. 157) has argued that 'a creature cannot have thoughts unless it is an interpreter of the speech of another'. Dummett (1978, p. 458) has interpreted some pronouncements as meaning that 'the study of thought is to be sharply distinguished from the study of the psychological processes of thinking and...the only proper method of analysing thought consists in the analysis of language'. And there is also the position that thought has its own language that might exist even prior to or in the absence of natural language. But here I am going to concentrate on what might be possible in the absence of natural language. I do not know what it would mean to consider thinking in the absence of its own intrinsic language, a language of thought, if the two always co-exist. On the other hand, other philosophers have taken quite a tolerant view. Bede Rundle, for example, has written that 'not only is wordless thought possible, as when we think how a room would look with the furniture rearranged; it does not even require attention to the matter in question for us to have thought that something was so, as when, tripping on a stair, we say we thought there was one fewer stair than there in fact was' (Rundle 1995, p. 872). Taking Dummett's lead, as a psychologist I am obviously concerned with thinking. I do not aim to define or analyse thought as a philosophical concept, which I leave to others. I am perhaps not overly sanguine in believing, however, that some facts about thinking in the absence of language might have a bearing on philosophical aspects of the issues, for those who are interested. Indeed, I suppose I would not have been asked to speak on the topic were that not the case. I do not wish to be constrained in how I use the word 'thinking'. In science definitions often come later or even last - after understanding — not first. In its various connotations, I take it to refer to such activities as reasoning, calculating, deliberating, reflecting, pretending, forming representations of things and categories of things, but also manipulating and connecting representations of 'things' and concepts that are separated in space and time. 127
L. Weiskrantz It is this idea of connecting or manipulating which could be a convenient peg on which to hang some of the evidence. While it might be commonsensical to accept that we can think wordlessly, as when we think how a room would look with the furniture rearranged, it could be argued, I suppose, that our ability to shift images around in our heads might nevertheless require an established linguistic endowment as a prerequisite, for example, being able to exploit the ability to label an object as an 'armchair' or a 'carpet', and to use the verb 'to move' appropriately, even if we do not always explicitly attach the labels. And such a case as 'thinking' that there is one less stair on a staircase, while the term is, of course, used that way in the vernacular, is probably no more than a matter of the reeling off of an automatic motor sequence which is imperfectly established or has gone slightly askew. Such visuomotor skills when they become automatic are typically devoid of accompanying thoughts. Or, to take another example, as a bit of slightly inaccurate spatial knowledge when we say we 'thought' the door was safely open when we bumped into it. There is a safer tack than depending on such thoughts about wordless thought or enjoying the generosity of vernacular usage. It is to take organisms who lack language, and consider what cognitive skills they can display. They lack it (a) because they have yet to acquire it (human infants), or (b) because they have lost it (braindamaged adults), or (c) because they cannot acquire it (non-human animals). Some of the evidence is reviewed in the edited volume Thought Without Language (Weiskrantz 1988), and I will give examples from each of these three domains from that source, as well as from others. Each domain, of course, requires its own particular techniques for demonstration and analysis. Human Pre-verbal Infants There is now a veritable avalanche of studies on infants. They almost all depend on careful examinations of where and at what an infant is looking, combined with ingenious variations by experimenters of the displays. This has uncovered well-known results showing, for example, innate preferences by the infant for various shapes, such as the human face. One can, by simply examining the infant's target of gaze, discover that it becomes directed to a distant object to which an adult is pointing, as studied by Butterworth and Grover (1988) and others (see fig. 1). (The same is true, incidentally, of the chimpanzee. See Povinelli and Eddy 1996.) This tells us something about the connection that the infant can make between 128
Thought without Language: Thought without Awareness? Infant fixates target
6
• i n f a n t fixates hand
9
12
Infants age (mts)
Figure 1. Pre-verbal communication: the ability of the pre-verbal infant to direct its gaze to a target at which the adult is pointing. (From Butterworth and Grover 1988, with permission) local and geometrically distant loci. Experiments involving more complex relationships depend upon the 'habituation' technique. A novel visual display is presented to the infant that first grabs its attention, but with repetition the infant becomes bored and looks elsewhere. If the display is changed, the question is whether the infant perks up - because it treats it as a sufficiently novel scene or goes on being bored because it is just more of the same old stuff. An example from Kellman and Spelke (1983) concerns whether the infant treats a partially occluded object as a single unified object, or as separate parts (see fig. 2). A whole host of such experiments have been carried out that reveal how objects are perceived, based on their cohesion, boundaries, substance and spatio-temporal continuity. Spelke (1988) speaks of the 'theory of the physical world' of the infant but - theory or not - the evidence reveals a highly-structured perceptual world from an early age. More complex conceptually is the perception of causal relationships. An ingenious series of studies has been carried out by Leslie and Keeble (1987; cf. Leslie 1988) involving the so-called 'billiardball' illusion in which one object appears to cause another to be launched, a phenomenon thoroughly studied in adults some years ago by Michotte. Young infants who habituate to the 'launch' scene (fig. 3, top row) dishabituate more strongly in switching to a nonlaunch scene (rows 2, 3 or 4) than those who habituate to a nonlaunch scene. Objects that are partially occluded intermittently or continuously are nevertheless treated as whole by the infant. The boundary between perception and memory can be studied by hiding the object whilst the infant is looking on - complete occlusion - and 129
L. Weiskrantz
Figure 2. After being habituated to the lower figure, four-month-old infants show more interest in (i.e. preferential gaze) the upper-left figure than in the upper-right. (Based on Kellman and Spelke 1983, with permission)
only allowing retrieval after an interval, the so-called delayed response problem and the closely similar A-not-B problem (Diamond 1988). At the earliest age at which an infant can reach and retrieve an object, about seven months, it will reach correctly for the position of the hidden object, but only for short delays. As age increases, the delay stretches out for longer intervals. Both the position and presumably the identity of the object are successfully stored. (We know that the rhesus monkey stores both.) Spelke has been one of the pioneers in this area of research. She has taken it a step further by seeing how the child changes as language starts to develop. Not surprisingly, there is a big increment in the flexibility with which, say, remembered locations of objects in a complex environment can be stored and retrieved. The human infant and the primate both start with a highly structured world, but language as it intervenes allows great manipulation of elements within that structure. But that is another story. How much of this activity of pre-verbal infants one wishes to call thinking, rather than perception or the forming of representations, is a moot question. It is towards the later stages of infancy that more conceptual phenomena emerge less ambiguously. Especially relevant are those involving games of pretence by children, which read130
Thought without Language: Thought without Awareness? Direct launching
D
•
a
cm
•
a •
a
•
a •
a
•
Delayed reaction
a
•
a
•
CM
Launching without collision
a
•
a
•
a
•
a
•
a
•
•
a
•
Delayed reaction without collision
a
•
-» a
•
a
•
o
Figure 3. After being habituated to the top display, infants show more dishabituation to any of the displays in the lower three rows than if they are habituated to one of them. (From Leslie and Keeble 1987, with permission) ily occur at about two years of age or so (Leslie 1988). The child readily enters into all kinds of pretend scenarios in which empty cups are spilled over toy animals, who are then dried with pretend towels, and given a bath in a pretend bathtub filled with pretend water from an empty cup, and so forth. It is the sort of game that parents like to play with their young children, and even more that children like to play themselves. The scenarios can become quite elaborate. They involve a conceptual distinction between appearance and reality. Here we would not, I take it, quarrel about designating such pretence scenarios as thinking, but of course they also are more easily studied when the child can understand some instructions using language. Piaget, however, describes some examples of pretence occurring even earlier, at the age of one year or so. For example, he describes a child 'pretending' that she held a pillow, another who 'pretended' to put a napkin-ring in her mouth, 131
L. Weiskrantz and to drink out of empty glasses, making noises with lips and throat. He describes these acts thus: [U]sing schemas which are familiar, and for the most part already ritualized in games...but instead of using them in the presence of the objects to which they are usually applied, [the child] assimilated to them new objectives unrelated to them...These new objects...are used with no other purposes than that of allowing the subject to mime...It is the union of these two conditions application of the scheme to inadequate objects and the evocation for pleasure - which...characterizes the beginning of pretence' (Piaget 1951, pp. 97-100) Piaget acknowledges that these patterns emerge at about the time when the child becomes capable of learning to speak, and raises just the question with which we are concerned in this paper: is language essential for imitation and pretence? He leans in the negative direction. 'First, the formation of such symbolism is not always accompanied by speech or contact with others...Baby L. pretends to be sleeping while smiling broadly, without saying a word and unaware that she is being watched. This by itself would of course prove nothing, since interiorised verbo-social behaviours might already exist. In conjunction with other arguments, however, it has its value.' And here, among other reasons, he turns to examples of chimpanzees playing certain symbolic games, quoting Wolfgang Kohler on an infant chimpanzee treating one of its legs like a doll, rocking it in its arms, and so forth. Piaget is not transparently clear in his arguments, and they are buried in complex discussion bound up with his assumptions about transitions between stages. But I do accept that our decision might well turn on the adequacy of evidence of pretence in young pre-verbal infants, for which it would be good to have more systematic observations. As far as it goes, I am inclined to give both Piaget and the infants the benefit of my doubts. I would add a further argument to his, namely that the child's language may be quite rudimentary at the age at which it engages in sustained pretence sequences. It will be noted that Piaget also links the status of the question of thought without language to the animal domain, and it is to that I now turn. Animal Studies There is a long tradition of research on the cognitive abilities of animals, not only of mammals and birds, but even of invertebrates such as the octopus (the late J. Z. Young, for example, told me of his 132
Thought without Language: Thought without Awareness? study of cross-modal transfer of learning between vision and touch in the octopus). There is also a large body of fascinating natural history anecdotes and highly engaging accounts of casual experiments, starting in the last century (such as those carried out in the postDarwinian period by Romanes and by Hobhouse) and leading up to the more recent era of Kohler, Robert Yerkes, and Henry Nissen. In the 1930s and beyond there was an intense flowering of carefully controlled laboratory work on animal learning, especially but not only in the rat. There is also a long tradition of intense disputation as to whether explanations can be reduced to stimulus-response or (for Skinnerians) to response-reinforcement accounts. It was the era of the experimentum cruets that never was quite crucial, and there has been a gradual switch to the detailed experimental analysis of behaviour. The theoretical alternative in terms of stimulus-stimulus associations opened a wedge in entrenched positions, but there are examples from animal cognitive behaviour that go beyond any explanations based on stimulus and/or response associative chains, no matter what the elements might be. A direct contrast with accounts in terms of stimulus-response chains came from Herbert Terrace (1995), who was himself instrumental in casting serious doubt on the linguistic status of the use of American deaf signs by chimpanzees. He taught pigeons to press coloured keys in a particular sequence, say A, B, C, D, regardless of the position of the coloured keys. Pigeons can learn to do that. He then asked whether pigeons could correctly retain information about isolated sub-parts of the sequence, such as B-D, in which the bird no longer has the advantage of the starting element in the sequence or the immediately adjacent one. He provided positive evidence that they can. There are other examples of sequential learning of food locations in which the performance reveals an ability to create maps of an array of locations independently of chains of rigid sequences, as in the well-known radial maze (Olton 1979), in which the animal is allowed to retrieve a food reward from the end of several alleys radiating from a central position. In performing the radial maze, a rat must, so to speak, subtract each site that it has visited from the remaining range of sites to be visited, and it does so efficiently and with relatively few returns to arms from which the food has already been removed by it, irrespective of the order in which the sites had been visited. Spatial learning introduces another type of example. A range of similar and nicely controlled experiments has been carried out on dogs and other mammals by Thinus-Blanc (1988) and her collaborators. Consider the situation on a large meadow, from a study by 133
L. Weiskrantz
22.5m/''
A •'' 40m
Figure 4. Dogs were first taken on the leash along the route A—D—B, and shown pieces of meat at points A and B but not allowed to eat them. The task consisted of finding the food hidden at A and B when released from D. Dogs went directly from A to B without returning to starting point. (From Chapuis and Varlet 1987, with permission)
Chapuis and Varlet (1987). Each dog is led along the route ADB (see fig. 4). It is shown pieces of meat being covered at points A and B but is not allowed to eat them. The task consists of finding the food hidden at A and B when the animal is released from D. On 96 percent of the occasions, dogs went directly from A to B, on a route never traversed before. A classical experiment by Tolman and Honzik (1930; cf. also Tolman 1948) on laboratory rats provides a compelling example of the ability of animals to manipulate their spatial knowledge. The rat first learns a maze (see fig. 5), without any barriers in place. When 134
Thought without Language: Thought without Awareness?
Food "" Block 3
• — Block A Route 1 \
•
Route 2
Route 3
Start
Figure 5. Rats were first exposed to the maze without any blocks in position. With a block at A, rats selected route 2. With a block at B, they selected route 3. (From Tolman and Honzik 1930, with permission)
given a free choice, it chooses the shortest route, route 1. When a barrier is then introduced, which alternative route is the rational one to take depends on the rat reconstructing its relationship to the stored spatial arrangement of the maze. If the block is at A it should take the route to the left. If the block is at B, it should take the route to the right. Rats are said to behave rationally, using such spatial knowledge appropriately. Another approach consists not in linking spatial locations, but in seeing whether monkeys could extract a relationship of conditionality from a set of associations between objects. The strategy used by Richard Saunders and me was to train the animal on the following task, where each letter stands for a particular object, each '+' means food reward (under the object) and each '-' means no food reward.
135
L. Weiskrantz
A B versus A C
D B versus D C
A B versus D B
D C versus A C
It will be seen that each object is equally often rewarded and nonrewarded. In any particular arrangement, whether it is rewarded is correlated perfectly with which object is its partner. In the top two discriminations, note that B is positive and C is negative when paired with A, and just the reverse when each of these is paired with D. In the final stage of our experiment, we presented just one object (A) or (D) in isolation, and then asked the animal to make an appropriate choice from another pair of objects: D
B or C ?
B or C ?
We asked the animal, in effect, to 'think'. Think: What other object had it been shown with previously when it had been rewarded? If A, then B had been the rewarded object and not C. If D, then C had been the rewarded object and not B. Monkeys could do this, that is, when given the single object in isolation they could extract its rewarded partner. (The control condition was the presentation of the last stage without experience of the previous pairings.) In any case, if you doubt that monkeys think in this situation, humans in the same situation certainly say they do! One might say, nevertheless, that this performance — impressive though it seems to me - is based on associative learning. But rule learning by animals cannot be readily handled in such terms. Monkeys show 'learning set', that is, they learn to learn, so that on each set of problems that can be solved by the same rule the animals get better and better, until they can even solve a new problem with just a single exposure. Such a rule might be 'win stay, lose shift,' or the opposite 'win shift, lose stay', rules that Harlow - who first opened up this area of research - deemed to be 'hypothesis learning'. The rule could also be 'matching', e.g. of a sample colour with one of two choice colours, or 'pick the odd one out'. This kind of capacity is not restricted to primates. Birds such as pigeons and jays also show learning set (Mackintosh et al. 1985). The kinds of errors that monkeys make in a learning set procedure also suggested to 136
Thought without Language: Thought without Awareness? Harlow that they spontaneously liked to try out variations in the rules as they become well habituated to and bored with a particular rule. Chimpanzees have been at the heart of much of the controversy about whether their ability to learn signs constitutes language, but this has tended to mask some other demonstrations of quite impressive cognitive skill independent of that issue. Premack, for example, who was a pioneer in training chimpanzees with tokens which signified different objects or actions, concluded that the chimpanzee does not acquire syntactical language - which he considers to be species-specific to humans. But, nevertheless, especially instructive is the importance of a particular type of relation that emerges from training in his experiments, namely learning the meaning of 'same/different', which leads to a strong advantage in the successful solving of analogies. Remarkable are some of his findings (Premack 1988) with juvenile chimpanzees when set the problem of adding ratios, which they could manage even when the samples consisted of different types of items, e.g., 1 apple plus 1 bottle = | disc? or, 1 disc? Various claims of an ability to engage in pretence have been made for primates. Also, Savage-Rumbaugh stated in a recent symposium {Toward a Science of Consciousness, Tucson, 1996) that she has positive evidence from 'theory of other minds'-type of experimentation in pygmy chimpanzees, such a theory being a prerequisite for pretence. I think the balance of evidence for the chimpanzee would support the claim that thinking is possible in these creatures without language. Evidence in this domain would be more compelling, perhaps, than that derived from the classical and seminal experiments on apes by Kohler, especially as it is possible from these more recent experiments to infer what the details are about the content of other minds. Recently, birds have come back into prominence on this scene, especially the parrot, which has been said to be capable of counting up to six or eight, to answer quite complex questions directed to it about the categories of shapes and colour of items in visual displays. Also, pigeons have been shown by Herrnstein (1984, 1985) to be remarkably able to form consistent categorisations, from pictures and exemplars, of classes such as trees, or fish, or bodies of water, or even of a particular person, even though the exemplars are very numerous and change from trial to trial. The relationship is similar to learning 'same/different' in Premack's chimpanzee work. Learning categories is not the same as thinking about them, although it does not preclude this as a concomitant of such learning. 137
L. Weiskrantz I cannot resist the temptation to quote a remark by Herrnstein in relation to his work (1984): 'to categorize, which is to detect recurrences in the environment despite variations in local stimulus energy, must be an enormous evolutionary advantage. Seen in this light, categorization is just object constancy, which is perhaps the fundamental constancy towards which all other perceptual constancies converge. Rather than psycholinguistics, it is the psychology of perception and stimulus discrimination that impinge most directly on categorization.' In our work (Weiskrantz and Saunders 1984) we have been able to show that monkeys will treat an object that is transformed in size or viewing angle or texture as equivalent to the original, even on the very first occasion in the animal's experience with such a task, i.e. when the retina has never been exposed to that pattern of stimulus in the animal's entire life. Humans Who Have Lost Language Typically, patients who have become aphasic as a result of brain damage are impaired in a host of ways — the lesion that causes global aphasia is large - and it is not surprising that the patients may have a cocktail of cognitive deficits (Kertesz 1988), which can appear as isolated selective deficits with more restricted brain damage. But there are aphasic patients whose language impairment is disproportionately severe in relation to their other cognitive capacities, who nevertheless one would normally describe as being able to 'think' in a demanding way. A striking example has been described by Newcombe (1987) of an academic scientist who sustained head injury. Although unable to perform better than a four- or five-year-old child on verbal tests of comprehension, and with speech that was the gibberish of severe jargon dysphasia, he performed normally or above average on nonverbal tests, viz. Raven's Advanced Progressive Matrices, as well as on tests of short-term memory, long-term maze learning, and visual completion. In particular, Raven's Matrices task is deliberately designed so as not to require verbal mediation. Each set introduces a particular principle to be inferred from a series of drawings, followed by a multiple choice of drawings from which the correct one must be selected. For example, the principle might be that the drawings in a row must be added, or that overlapping parts of successive drawings must be deleted, or that they follow a numbering principle, or a systematic re-arrangement of symbols, and so forth. The problems can be quite tough, especially for one achieving the level reached by this particular subject. I have seen colleagues spend 138
Thought without Language: Thought without Awareness? a considerable amount of time and much anguish on some of the problems, and they would certainly insist that they were engaged in thinking about them. Thought Without Awareness? The evidence from aphasia raises an interesting and surprising issue. Such patients may be unable to discriminate a meaningful sentence from gibberish, and yet there are some with whom it can be demonstrated that they nevertheless have intact processing of syntax, semantics and pragmatics. The patient him- or herself does not know it, but the capacity lurks somewhere in the remaining brain tissue, and can be extracted. The demonstration depends upon using reaction times of responses to target words. It is known that reaction times are longer if the target word is embedded in a degraded sentence than if it falls within a linguistically correct sentence. Tyler (1988, 1992) presented an aphasic subject with sentences that were degraded semantically, syntactically or pragmatically. The subject succeeded in being able to follow the instruction to respond as quickly as possible whenever a particular target word was uttered, e.g., the word 'guitar' in the following examples: A. The crowd was waiting eagerly. The young man grabbed guitar and... B. The crowd was waiting eagerly. The young man drank guitar and... C. The crowd was waiting eagerly. The young man slept guitar and... D. The crowd was waiting eagerly. The young man buried guitar and...
the the the the
Sentence A is a normal and un-degraded. Sentence B is semantically anomalous because it violates the meaning of 'guitar'. Sentence C is syntactically anomalous because 'slept' is an intransitive verb. Finally, D is what Tyler calls 'pragmatically' anomalous, burying a guitar is a possible act, but a silly and certainly an unusual one. Related studies have been published by Frederici (1982) by Blumstein et al. (1982) and by Milberg and Blumstein (1981). Even though Tyler's aphasic patient was severely impaired in his ability to judge overtly when a sentence was anomalous or normal, his pattern of reaction times to target words in a degraded context showed the same slowing that is characteristic of normal control subjects. Overall the patient was slower (not surprising in a braindamaged subject), but the same pattern emerged. In other words, it 139
L. Weiskrantz was demonstrated that the patient retained an intact capacity to respond both to the semantic and grammatical structure of the sentences. But he could not use such a capacity, either in his comprehending of speech or his use of it and, indeed, when this patient was explicitly asked whether a sentence was correct or not, he could not discriminate. Tyler distinguishes between the 'on-line' use of linguistic information, which she posits was preserved in her patient, and its exploitation 'off-line', which was deficient. She argues that a capacity can be intact when used 'on-line', but is not available for 'off-line' processing. The subject could 'develop the appropriate representations of an utterance; his problem lies in being unable to gain access to them for the purpose of making explicit decisions about them' (Tyler 1992, p. 170). We simply do not know whether the kind of patient who can master the Ravens Matrices problems is also intact for 'on-line' linguistic structure - it is an interesting question, but no-one has examined this, as far as I am aware. But I am inclined to consider that it is irrelevant because presumably it requires 'off-line' use of language to use in any particular application, even if that application itself is non-verbal. To return: the 'intact possession of linguistic structure is not available for normal use by the patient, and this is an example of 'covert' or 'implicit' processing. The experimenter knows it, and somewhere in the remaining brain structures of the patient there is a system that 'knows' it. The patient does not know it in the sense of acknowledging awareness of it or being able to use it in communication. Of course, the possession even of a full-blown, intact and normal linguistic skill does not mean that there is thinking whenever it is used. One only need consider the mindlessness of the typical ritualistic verbal display of the British in discussing the weather, and the inappropriateness of the verbal commentaries of the anosagnosic patient, well described by Bisiach (1988). An anosagnosic patient denies the existence of a patently obvious deficit, such as unilateral paralysis, and gives intricate explanations of why he is normal in the face of evidence to the contrary. There can be language without thought. The possession of an intact linguistic structure of which the patient is unaware does not tell us anything about whether it could be or ever is used for thought. Language can even interfere. Thus, Gazzaniga (1988) describes a split-brain patient who carried out a 'same/different' task with hard-to-name colours: when a pair of colours was presented to either hemisphere with a brief (180 msec) interval between them, performance was equally good no matter to which hemisphere the visual information was 140
Thought without Language: Thought without Awareness? directed. But when the interval was longer - a full second - the left hemisphere, the one normally in control of language, was much slower. Both hemispheres are equally proficient when there is a demand for a quick response based on perceptual appearance. But given the licence of a longer stretch of time the verbal left hemisphere cannot help trying to name the hard-to-name colours, and so gets held up. We all see the unwieldy dominance of the left hemisphere being expressed often enough in everyday life, especially by those with a tendency towards logorrhoea. But, as a general matter, could there be thought without awareness? The man in the street would probably think not. But at least some philosophers, including those whose positions derive from folk psychology, for example, David Rosenthal (1993), accept the existence on theoretical grounds of unconscious mental activity, including unconscious thoughts. Rosenthal posits the need for a higher-order thought to render the lower-order thought conscious. One can appeal, in seeking relevant empirical evidence, to the phenomenon which psychologists used to refer to, in the old days, as 'reminiscence', that is, to improvement in retention after the cessation of practice - in fact, some days afterwards. There is quite a bit of positive evidence with human verbal learning, although rather stringent controls would be needed to demonstrate that it is not due to involuntary rehearsal, testing fatigue and the like, or even to the benefits of forgetting interfering material (cf. Hovland 1951). I believe the evidence needs to be taken seriously, even if it is not cast-iron. There seems to be intriguing anecdotal evidence along the same lines; for example, the well-known case of the solution of a difficult mathematical problem suddenly occurring to Poincare as he stepped onto a bus, having put the problem aside after a period of unsuccessful mental effort (Hadamard 1945). The covert status of linguistic structure in aphasic patients is but one example of such processing that can be found in virtually every type of cognitive impairment caused by brain damage (cf. Weiskrantz 1991, 1997). Patients who cannot remember consciously can be shown to store information; patients who neglect part of their visual world nevertheless can be shown to process neglected information; patients rendered blind in part of their visual fields can still process information of which they are unaware; and patients who have lost the meaning of visual objects or faces can still be shown to retain the meaning in a form of which they have no awareness. Here are some brief examples. Amnesic patients do not 'remember' any new event beyond a minute or so. It is a severely incapacitating condition caused by damage to certain medially placed brain structures, on which a great 141
L. Weiskrantz deal of research has been conducted. Nevertheless, such patients can be shown to retain information, including new information, over intervals of weeks or even months, but they do not know they are 'remembering' it. One can demonstrate normal priming for words, combined with complete failure to recognise the words for which they have just been primed (Warrington and Weiskrantz 1968, 1978). That is, a list of words seen a few minutes before will not be recognised or recalled. But the subject will show enhanced correct 'guessing' of the completions of the words if just shown the first three letters (relative, of course, to the words not seen). It is not only priming, however, that is intact. The patients can acquire new meanings without knowing that they have. Here are two examples from McAndrews et al. (1987). The subjects are asked to produce a single word that would help to give sense to a strange sentence. If, as is usually the case, they cannot do so, they are told the answer. For example, 'The haystack was important because the cloth ripped.' Answer: 'Parachute'. Or, 'The notes were sour because the seams split'. Answer: 'Bagpipe'. The point is that once one knows the answer, it is retained for a long time, even a matter of weeks. And so it is for amnesic patients, except that they deny ever having seen the sentences before, let alone having been told the answers. And yet they give the correct answers. Unilateral neglect, a condition associated with damage to the (typically) right posterior part of the brain, provides another example of retained implicit capacities: the ability to make correct judgements about information which is not 'seen'. Such patients tend to ignore visual material in the left half of the visual field. A striking example was reported by Marshall and Halligan (1988). Two views of a house were shown in which both right halves are identical, but the left halves are quite different (see fig. 6). The neglect patient consistently judged the two pictures to be identical, ostensively ignoring the left half of each picture. When asked to say which house she would prefer to live in, she retorted that it was a silly question because they were the same, but nevertheless reliably chose the house not in flames. When there is damage to primary visual cortex, phenomenal 'blindness' results in the hemifield of vision contralateral to the damaged hemisphere (older subcortical pathways from the eye still remain intact, connecting with subcortical targets in the brain), a condition known as 'blindsight' (Weiskrantz et al. 1974, Weiskrantz 1986). Such patients can discriminate wavelength (colour) by forced-choice guessing in their blind fields even though they have absolutely no experience of colour (Stoerig and Cowey 1992, Barbur et al. 1994, Brent et al. 1994). Similarly, they can guess 142
Thought without Language: Thought without Awareness?
Figure 6. Pictures shown to a neglect patient, who judged the two to be identical because she neglected the left half of each. But she preferred the non-burning house as one in which to live. (In the original experiment, the flames were coloured red.) (From Marshall and Halligan 1988, with permission) whether a grating is orientated horizontally or non-horizontally, without any experience of the gratings (Weiskrantz 1986). They also, according to Marcel (forthcoming), can be semantically primed by words flashed into their blind field even though they do not have any experience of the words. For example (see fig. 7), depending on whether the word 'river' or 'money' was flashed into their blind field, the definition of 'bank' presented to the intact field was biased accordingly. The well-known split-brain patients can process pictures of words flashed into their right non-verbal hemisphere, which is disconnected from their left 'verbal' hemisphere, and guide their action accordingly. And yet they deny seeing the words (Sperry 1974). Prosopagnosic patients can show evidence of linking correct 143
L. Weiskrantz
Figure 7. Example of stimuli used to study the semantic biasing effect of words projected into the 'blind' hemifield on the meaning of a word subsequently shown in the intact hemifield in 'blindsight'. (Based on Marcel, forthcoming)
names with faces faster than they can for incorrect names, although they are at chance in recognizing familiar from unfamiliar faces (Young and De Haan (1992)). Interestingly, in this condition as in some of the others, there are autonomic responses that demonstrate intact knowledge or discrimination. Thus, the prosopagnosic patient shows a larger galvanic skin response to familiar faces than to unfamiliar faces (Tranel and Damasio (1985)). There is nothing wrong with that part of their nervous system. In all of these examples there is typically a disconnexion between the existence of an intact capacity and the awareness of the subject of its existence. But do any of these examples qualify as thinking without awareness? Here we are again at the mercy of definitions, but if we consider a psychological approach, involving the mental manipulation of images, symbols, words, memories and the like, the covert processes of these patients would be thinkingless. They are unable to manipulate the knowledge or the perception that they possess but do not know they possess. The blindsight subject cannot form an image of the stimuli he is guessing to compare with the stimuli he guessed about yesterday or even a minute ago. The amnesic subject cannot compare his acquired memories with any other memories — they are not memories for him. The patients are 144
Thought without Language: Thought without Awareness? unable to connect items in time or space, depending on their particular deficit. In this sense the patients have been deprived of their ability to think about their intact content. The loss that they suffer is severe within the mode in which their deficits lie. The blindsight patient (fortunately he or she has an intact half field of vision) bumps into objects in his blind field. The amnesic patient requires more or less constant custodial care. His implicit memory ability is powerful but useless as a tool for thinking or relating; it is interesting, for example, that the amnesic patient fails to benefit from the use of imagery to form associations between unrelated words, which are of great help to normal subjects. Similarly, the prosopagnosic patient continues to be deeply embarrassed by the inability to recognise the faces of his family or friends, even though his autonomic nervous system is fully appraised of the faces being those of familiar folk. With global aphasia, the population is far from homogeneous because many lose non-verbal cognitive skills as well as their verbal skills, and such cases can be assumed to be thinkingless even if it were shown that they might retain their so-called 'on-line' sensitivity to syntactical and semantic structure. Less unfortunate aphasics, like the academic scientist referred to earlier, can retain good non-verbal skills with which to 'think', as we have seen, but this does not imply that they can use their 'implicit' on-line verbal capacity for thinking. We do not deny, of course, that patients can have conscious thoughts even if they are reduced to the level of covert processing in particular domains. The amnesic patient makes correct assertions and inferences about things in the world based on acquired information, but the origin of the assertions is missing; nor can he connect them with other pieces of acquired information. The blindsight patient has a conscious thought that he is guessing, and with forced-choice guessing even of what he has been told to guess. The prosopagnosic patient knows that he is looking at a face, and not his mother-in-law's hat, but cannot connect that with any other knowledge about that particular face. Efforts to rehabilitate any of these patients have yielded some encouraging results, but they involve a very hard slog and never achieve anything like a return to normality. It may be that the only solution in the end will be to transplant tissue into the damaged parts of their brains. One cannot live by implicit processing alone. Whether one can live without it is another matter, but not one for here. I deliberately leave untouched any reference to assumptions of unconscious desires and defensive thinking in Freudian theory. I am content to leave such matters to others. To try to draw some of these threads together, I have argued that some quite impressive cognitive processes can be found in creatures 145
L. Weiskrantz who lack language, either because they have not yet acquired it or have acquired only it only rudimentarily (infants), or because they cannot acquire it (animals), or who have lost it (brain-damaged adults). How much of this evidence, in detail, demonstrates thinking turns on matters of definition, but to the psychologist it is interesting and instructive in its own right, whether or not it arbitrates in philosophical controversy. In terms of everyday parlance, or in terms of a psychological emphasis on the ability to manipulate images or representations or knowledge, all three domains include at least some positive examples: in infants, the appearance of pretence and fantasy (although we need more evidence here); in animals, the ability to calculate or to count or to form and manipulate cognitive maps; and in aphasic patients, the ability to solve quite tough non-verbal problems. If anyone wishes to deprive these languageless creatures of the capacity to think, he must do it by fiat. I take the evidence to mean that such a severe conclusion would be misplaced. There is another way of slicing the cake: in everyday parlance, the word 'think' arises in such phrases as 'think of, 'think through' and 'think in'. That there can be mental content, the ability to 'think of, in the absence of language, would not seem to be in doubt. The infant can understand the unity of a partially occluded object, and can know what the location of a hidden object is. The right hemisphere isolated from its 'linguistic' partner can process information categorically. The pigeon can form complex categories involving trees or lakes and so forth. The aphasic patient can 'think through' a difficult brain-teaser. The chimpanzee can solve the ratio problem. To 'think in' is another matter. Clearly many of us think in words, but others in images — for example, Einstein was insistent that this was his mode of thinking (cf. Wertheimer 1945, p. 184). No one doubts the enormous increment in power that language adds to thinking, but one can ask whether this is quantitative only — although quantitatively huge — and actually exists in evolutionary continuity with the non-verbal skills alluded to in this paper. Evidence has also been reviewed showing that cognitive deficits seen in the clinic can mask the existence of quite powerful capacities of which the patients (and often the clinician) are unaware. I raised the question of whether there can be thinking in the absence of awareness. There is some evidence, not terribly secure, from the phenomenon of reminiscence that this might be possible, and anecdotally we probably all have such examples - for example, the crossword addict who returns to a puzzle to find the answer he or she struggled unsuccessfully to achieve a few hours earlier. But the more 146
Thought without Language: Thought without Awareness? solidly established evidence from neuropsychology, of intact implicit processing of which a patient is unaware, I consider to be poor and unsatisfactory evidence for thinking based on such processing. In none of the syndromes are subjects able to manipulate images or memories in their particular defective universe by using their implicit capacities, as robust as these may be. That is, the subjects do not have the capacity to think within the domain in which their deficit lies. And so: on grounds of empirical evidence I would take issue with a position that would deny thinking in the absence of language, but I would also be reluctant to accept a position for thinking without awareness, in the sense of connecting images across space or items of acquired knowledge across time. But, given that we are talking about evidence and not about positions etched in rock, it is worth the thought that conclusions can change with further evidence. If you disagree about the role of evidence, then we are thinking along different lines.
References Barbur, J. L., Harlow, J. A., Sahraie, A., Stoerig, P. and Weiskrantz, L. 1994. 'Responses to Chromatic Stimuli in the Absence of VI: Pupillometric and Psychophysical Studies', Optical Society of America Technical Digest 2, 312-15 Bisiach, E. 1988. 'Language Without Thought', in Weiskrantz (ed.) (1988) Blumstein, S., Milberg, W. and Shrier, R. 1982. 'Semantic Processing in Aphasia: Evidence from an Auditory Lexical Decision Task', Brain and Language 17, 301-15 Brent, P. J., Kennard, C. and Ruddock, K. H. 1994. 'Residual Colour Vision in a Human Hemianope: Spectral Responses and ColourDiscrimination', Proceedings of the Royal Society of London B 256, 219-25 Butterworth, G. and Grover, L. 1988. 'The Origins of Referential Communication in Human Infancy', in Weiskrantz (ed.) (1988) Chapuis, N. and Varlet, C. 1987. 'Shortcut by Dogs in Natural Surroundings', Quarterly Journal of Experimental Psychology 39B, 49-64 Davidson, D. 1984. Inquiries into Truth and Interpretation. Oxford: Clarendon Press Diamond, A. 1988. 'Differences between Adult and Infant Cognition: Is the Crucial Variable Presence or Absence of Language'? in Weiskrantz (ed.)(1988) Dummett, M. A. E. 1978. Truth and Other Enigmas. London: Duckworth Frederici, A. D. 1982. 'Syntactic and Semantic Processes in Aphasic 147
L. Weiskrantz Deficits: The Availability of Prepositions', Brain and Language 15, 245-58 Gazzaniga, M. S. 1988. 'The Dynamics of Cerebral Specialization and Modular Interactions', in Weiskrantz (ed.) (1988) Hadamard, J. 1945. An Essay on the Psychology of Invention in the Mathematical Field. Princeton University Press Herrnstein, R. J. 1984. 'Objects, Categories, and Discriminative Stimuli', in H. L. Roitblat, T. G. Bever and H. S. Terrace (eds.), Animal Cognition. Hillsdale, NJ: Erlbaum 1985. 'Riddles of Natural Categorization', in Weiskrantz (ed.) (1985) Hovland, C. I. 1951. 'Human Learning and Retention', in S. S. Stevens (ed.), Handbook of Experimental Psychology. New York: Wiley Kertesz, A. 1988. 'Cognitive Function in Severe Aphasia', in Weiskrantz (ed.) (1988) Kellman, P. J. and Spelke, E. S. 1983. 'Perception of Partly-Occluded Objects in Infancy', Cognitive Psychology 15, 483-524 Leslie, A. L. 1988. 'The Necessity of Illusion: Perception and Thought in Infancy', in Weiskrantz (ed.) (1988) Leslie, A. L. and Keeble, S. 1987. 'Do Six-month-old Infants Perceive Causality?' Cognition 25, 265-88 McAndrews, M. P., Glisky, E. L. and Schacter, D. L. 1987. 'When Priming Persists: Long-lasting Implicit Memory for a Single Episode in Amnesic Patients', Neuropsychologia 25, 497-506 Mackintosh, N. J., Wilson, B. and Boakes, R. A. 1985. 'Differences in Mechanisms of Intelligence among Vertebrates', in Weiskrantz (ed.) (1985) Marcel, A. J. (forthcoming). 'Blindsight and Shape Perception: Deficit of Visual Consciousness or of Visual Function?', Brain Marshall, J. and Halligan, P. 1988. 'Blindsight and Insight in Visuo-spatial Neglect', Nature 336, 766-77 Milberg, W. and Blumstein, S. E. 1981. 'Lexical Decision and Aphasia: Evidence for Semantic Processing', Brain and Language 14, 371-85 Newcombe, F. 1987. 'Psychometric and Behavioural Evidence: Scope, Limitations, and Ecological Validity', in H. S. Levin, J. Grafman and H. M. Eisenberg (eds.), Neurobehavioral Recovery from Head Injury. Oxford University Press Olton, D. S. 1979. 'Mazes, Maps, and Memory', American Psychologist 34, 583-96 Piaget, J. 1951. Play, Dreams and Imitation in Childhood. London: William Heinemann Premack, D. 1988. 'Minds With and Without Language', in Weiskrantz (ed.) (1985) Povinelli, D. J. and Eddy, T. J. 1996. 'Chimpanzees. Joint Visual Attention', Psychological Science 7, 129-35 Rosenthal, D. 1993. 'Thinking that one Thinks', in M. Davies and G. W. Humphreys (eds.), Consciousness. Psychological and Philosophical Essays. Oxford: Blackwell 148
Thought without Language: Thought without Awareness? Rundle, B. 1995. 'Thinking', in T. Honderich (ed.), The Oxford Companion to Philosophy. Oxford University Press Spelke, E. S. 1988. 'The Origins of Physical Knowledge', in Weiskrantz (ed.) (1988) Sperry, R. W. 1974. 'Lateral Specialization in the Surgically-Separated Hemispheres', in F. O. Schmitt and F. G. Worden (eds.), The Neurosciences: Third Study Program. Cambridge, MA: MIT Press Stoerig, P. and Cowey, A. 1992. 'Wavelength-Sensitivity in Blindsight', Brain 115, 425-44 Terrace, H. S. 1985. 'Animal Cognition: Thinking Without Language', in Weiskrantz (ed.) (1985) Thinus-Blanc, C. 1988. 'Animal Spatial Cognition', in Weiskrantz (ed.) (1988) Tolman, E. C. 1948. 'Cognitive Maps in Rats and Men', Psychological Review 55, 189-208 Tolman, E. C. and Honzik, C. H. 1930. 'Degrees of Hunger-Reward and Non-Reward in Maze-Learning by Rats', University of California Publications in Psychology 4, 241-56 Tranel, D. and Damasio, A. R. 1985. 'Knowledge Without Awareness: An Autonomic Index of Facial Recognition by Prosopagnosics', Science 228, 1453-55 Tyler, L. K. 1988. 'Spoken Language Comprehension in a Fluent Aphasic Patient', Cognitive Neuropsychology 5, 375-400 1992. 'The Distinction between Implicit and Explicit Language Function: Evidence from Aphasia', in A. D. Milner and M. D. Rugg (eds.), The Neuropsychology of Consciousness. London: Academic Press Warrington, E. K. and Weiskrantz, L. 1968. 'New Method of Testing Long-term Retention with Special Reference to Amnesic Patients', Nature 217, 972-74 1978. 'Further Analysis of the Prior Learning Effect in Amnesic Patients', Neuropsychologia 26, 169-76 Weiskrantz, L. 1985. 'Introduction: Categorization, Cleverness and Consciousness', in Weiskrantz (ed.) (1985) 1986. Blindsight: A Case Study and Implications. Oxford University Press 1991. 'Disconnected Awareness for Detecting, Processing, and Remembering in Neurological Patients', The Hughlings Jackson Lecture. Journal of the Royal Society of Medicine 84, 466-70 1997. Consciousness Lost and Found: A Neuropsychological Exploration. Oxford University Press Weiskrantz, L. (ed.). 1985. Animal Intelligence. Oxford: Clarendon Press 1988. Thought without Language. Oxford University Press Weiskrantz, L. and Saunders, R. C. 1984. 'Impairments of Visual Object Transforms in Monkeys', Brain 107, 1033-72 Weiskrantz, L., Warrington, E. K., Sanders, M. D. and Marshall, J. 1974. 'Visual Capacity in the Hemianopic Field following a Restricted Occipital Ablation', Brain 97, 709-28 149
L. Weiskrantz Wertheimer, M. 1945. Productive Thinking. New York: Harper and Row Wittgenstein, L. 1922. Tractatus Logico-Philosophicus. London: Routledge and Kegan Paul Young, A. W. and De Haan, E. H. F. 1992. 'Face Recognition and Awareness after Brain Injury', in A. D. Milner and M. D. Rugg (eds.), The Neuropsychology of Consciousness. London: Academic Press
150
Philosophy, Thought and Language HANS-JOHANN GLOCK One of the most striking features of twentieth-century philosophy has been its obsession with language. For the most part, this phenomenon is greeted with hostile incredulity by external observers. Surely, they say, if philosophy is the profound and fundamental discipline which it has purported to be for more than two millennia, it must deal with something more serious than mere words, namely the things they stand for, and ultimately the essence of reality or of the human mind (Gellner 1959 provides an amusing, if unsophisticated, example.) This reaction is not confined to lay people, but is shared by many philosophers who are far removed from common sense. Indeed, Michael Dummett has claimed that the concern with language is the elusive factor X, long sought for in vain by Anglo-European conferences, which separates the phenomenological tradition on the continent founded by Husserl from Anglophone analytic philosophy, which he traces back to Frege. Dummett defines analytic philosophy as based on the idea that a philosophical understanding of thought can and must be given by an account of language. This contrasts with the philosophy of thought, which retains the idea that philosophy should investigate thought, but claims that this investigation is independent of and antecedent to an understanding of language (1993, chs. 2, 12, 13). According to Dummett, this view informs not just phenomenology, but also recent work within the Anglophone philosophical community; he mentions the Oxford philosophers Gareth Evans and Christopher Peacocke, but John Searle, Thomas Nagel and Colin McGinn also come to mind. Strictly speaking, therefore, these philosophers no longer count as analytic. Dummett deserves credit for reviving the discussion about the origins of analytical philosophy. As I hope to show, he is also onto an important point in emphasising the role of thought. But even if one takes into account the scope of the canvas on which he paints, his brush-strokes are inaccurate. A concern with language does not mark off analytic from continental philosophy. Paradigmatic analytic philosophers like Moore, Russell, Hart and Rawls would not accept that the basic task of philosophy is to analyse language. Equally, a work by Heidegger, Husserl's most famous pupil, bears the title On the Way to Language. And, for better or worse, 151
Hans-Johann Glock Heidegger's followers have reached that destination. The idea that human experience is essentially linguistic is a commonplace among hermeneutical philosophers. And the jargon of much current philosophy on the continent (notably of French post-structuralism) is taken not from metaphysics or psychology, but from linguistics and semiotics (see Gadamer 1967, p. 19, Derrida 1967). In fact, it might well be argued that the dominant empiricist strand within analytic philosophy, forever obsessed with the raw given (impressions, sense data, neural stimulations), is less equipped to do justice to a complex phenomenon like language than the hermeneutic tradition. In 1918, when German philosophers like Hamann, Herder, von Humboldt, Dilthey and Schleiermacher had been exploring the social and historical nature of linguistic understanding for over a hundred years, an analytic genius like Bertrand Russell remained so mesmerized by the idea that the meanings of words are private sense-data that he was capable of claiming that people 'would not be able to talk to each other unless they attached quite different meanings to their words' (1956, p. 195). Dummett's distinction between analytic philosophy and the philosophy of thought is not just inaccurate as a philosophical taxonomy, it also suffers from another shortcoming. He neither justifies nor explains the assumption behind it, namely that the fundamental task of philosophy is at any rate to analyse thought. We can readily grant that thought is an important topic in the philosophy of mind, but why should it be the topic of philosophy as a whole? I suggest that this prima jade startling idea can be understood as the result of two historical trends. The first trend is the nineteenthcentury debate about the nature of logic, which was a debate about the status of the so-called 'laws of thought'; the second trend is the Kantian idea that philosophy, instead of itself thinking about reality, thinks about the way we think about reality. I shall discuss these two roots of modern philosophy's concern with thought in section I. Section II shows how these two trends come together in the early Wittgenstein. The Tractatus Logico-Philosophicus initiated the linguistic turn of twentieth-century philosophy by maintaining that logic is concerned with quasi-Kantian preconditions of representation, and that this representation is linguistic rather than mental. The final two sections turn from historical to substantive issues. Both mentalist and linguistic conceptions of thought take for granted that thought requires a medium, that we always think in something. I shall try to cast doubt on that premise with the help of Ryle and the later Wittgenstein (section III). The paper ends by indicating two links between thought and language that do not involve the questionable idea of a medium of thought (section IV). 152
Philosophy, Thought and Language I. The Laws of Thought and the Preconditions of Representation The nineteenth-century controversy about the status of logic took place amidst the development of modern logic by Frege and Russell. But in spite of the revolutionary progress in the formal or technical aspects of logic, the philosophical debate about the nature of logic proceeded on the traditional assumption that logic studies the laws of thought, laws of correct thinking and reasoning, as in the title of Boole's major work - An Investigation of the Laws of Thought (1854). There were various suggestions as to the nature of these laws (see Glock 1992). According to psychologistic logicians like Boole and Erdmann, they describe how human beings (by and large) think, their basic mental operations, and are determined by the nature of the human mind. Against this, Platonists like Frege protested that logical laws do not describe how we actually think, but prescribe how we ought to think. They are strictly necessary and objective laws of 'truth', not contingent laws of 'holdings-tobe-true' (1893, pp. XVIff.).1 And it seemed that this objectivity can only be secured if their subject matter - thoughts and their structure — is not private ideas in the minds of individuals, but abstract entities in a realm beyond space and time. Russell's conception of logic differs from Frege's Platonism in its Aristotelian outlook. For Russell, logical propositions are truths about the most general aspects of reality, to which philosophy has access by abstraction from non-logical propositions (1903, pp. v-vi, 106; 1912, ch. 14; 1984, pp. 97ff.). Thus 'Socrates is married to Xanthippe' yields 'There are binary relations', or 'Something is somehow related to something else.' The symbols which survive this process are 'logical indefinables' or 'logical constants' ('relation', 'set', propositional connectives, quantifiers, etc.). They name 'logical objects' to which we have access by 'logical intuition'. The latter is a 'logical experience' analogous to the 'acquaintance' we 1
Dummett (1991, p. 225), has recently maintained that Frege did not condone the idea that logic is concerned with normative laws of thought ('Denkgesetze', 'Grundsatze des Denkens'). But Frege clearly does so in the writings through which he participated in our debate. Indeed, he emphasises that the laws of logic, being laws of truth rather than laws of holding something to be true, are concerned 'not with the question of how men think, but with the question of how they must think if they are not to miss the truth' (1969, p. 149). Dummett is right, however, in pointing out that these normative laws are based on a correct description of a domain of objects, which in the case of logic are thoughts understood as abstract entities (see Frege 1969, pp. 145-6). 153
Hans-Johann Glock have with sense-data like redness or the taste of pineapple. For Russell, logic is the most abstract science and differs from empirical disciplines only in degree of generality. It describes more abstract features of the same world. Frege, by contrast, not only rejects psychologism, but also denies that logic and mathematics deal with the empirical world in any way. According to his Platonist position, they instead deal with abstract entities inhabiting a 'third realm'. Unlike physical things, these entities have no spatial or temporal location, but unlike mental ideas they are objective, independent of any individual psyche. Logical and mathematical symbols are names of such entities (thoughts, concepts, truth-values, numbers); logical and mathematical propositions are self-evident truths which unfold the essence of these entities and are certified by a 'logical source of knowledge' (1969, pp. 269-73, 278). In spite of their differences, Frege and Russell both agree on one point: the normative laws of thought follow from the correct descriptions of logical relations which hold between thoughts or propositions, just as technical rules follow from correct descriptions of physical reality. The second root of the idea that philosophy investigates thought is Kant's 'Copernican Revolution'. The Aristotelian tradition had viewed metaphysics as a discipline which differs from the special sciences by describing not specific features of particular objects, but 'being qua being' {Metaphysics, IV. 1, VI.1, XI.3). This was understood to mean that it scrutinises the most general features of reality by abstracting from concrete features of particular objects (an idea we have already encountered in Russell's conception of logic). Kant brought about a fundamental reorientation by insisting that transcendental philosophy is 'occupied not so much with objects as with the mode of our knowledge of objects' {Critique of Pure Reason, B25).2 While science and common sense think or talk about reality and its objects, philosophy is not concerned with objects of any kind, not even the abstract entities postulated by Platonism. Instead it reflects on our way of knowing or experiencing the objects of the material world. Forebodings of the idea that philosophy is a reflective enterprise 2
There is an often-ignored complication (see Glock 1997). According to this passage, transcendental philosophy is concerned not with experience or knowledge in general, but exclusively with synthetic ('of objects') knowledge a priori. Hence transcendental philosophy is initially defined as the study of the possibility of metaphysics. However, because of the explanation Kant gives of that possibility (see below), transcendental philosophy can also be defined as concerned with the necessary preconditions of the possibility of experience {Prolegomena, Appendix). 154
Philosophy, Thought and Language can be found elsewhere. Augustine intimated that philosophical issues like the nature of time confront us with the task of reminding ourselves of something we already know outside philosophical contemplation {Confessions, XI. 14). Plato insisted that philosophising involves a 'periagoge', a change of the perspective we occupy in everyday life. However, according to Plato's doctrine of anamnesis the result is a recollection of our previous acquaintances with abstract objects {Republic, 514ff.; Phaedo, 249c). Thus Plato's reorientation merely leads from one kind of objects - those of sensible experience - to another - abstract ideas. Kant's reorientation is of a more radical kind. Kant accepts the Aristotelian slogan that metaphysics is the study of being qua being, in his terms, of 'whatever is insofar as it is' (A845/B873). But for him this does not amount to a process of abstraction from empirical concepts (from 'Texan' to 'human being' to 'material object') or of generalisation from empirical facts (from 'Socrates loves Xanthippe' to 'There are binary relations'). Instead, the investigation of being qua being is understood as an investigation into the essential features of objects, and hence into what it is to be an object. In a crucial passage Kant maintains that the conditions of the possibility of experience in general are likewise conditions of the possibility of the objects of experience (A158/B196, see A155-7/B194-5) This means that the essential features which define what it is to be an object are determined by the necessary preconditions for the possibility of experiencing objects. For example, we experience objects as located in space and time, and as centres of qualitative changes which are subject to causal laws. According to Kant, these are not contingent facts about human cognition to be studied by empirical psychology, but necessary features of experience, which at the same time define what it is to be an object of experience. There is a difference between experiences and their objects, and the content of experience is contingent - a matter of brute fact. But there are also necessary or structural features of experience, and these determine the necessary or essential features of the objects of experience. What looked like ontological essences, necessary features of the world itself, are in fact reflections of the conceptual scheme we use in experiencing the world. Consequently, the metaphysical search for the essence of reality is transformed into a second-order reflection on the way in which objects are given to us in experience. 'Metaphysics deals not with objects, but with cognitions {Erkenntnissen)' {Reflektionen §4853, see §4457). Philosophy is a non-empirical reflection on the way we represent objects. 155
Hans-Johann Glock Kant's idea was taken up and generalised by later German philosophers. First, Schopenhauer shifted the emphasis from Kant's notion of experience to that of mental representation (Vorstellung). Next, Husserl shifted it to the notion of consciousness, by maintaining that philosophy reflects on 'modes of consciousness'. He also went beyond Kant. Instead of reflecting exclusively on what it is to be an object of sensory experience, he distinguished various types of object - physical, mental, abstract according to their different 'modes of givenness' to the subject (1913, vol. I, §13, vol. Ill, §7). Finally, Heidegger went beyond the consciousness of objects - and perhaps the limits of intelligibility when he reflected on the 'disclosedness of world' (Erschlossenheit der Welt), a kind of awareness which is not directed at objects, but includes undirected moods and the often unconscious way we are immersed in our cultural traditions (1927, §§18, 32). Without the explicit invocation of the mind, this 'reflective turn' is also at work in the neo-Kantian philosopher-scientists of the turn of the century. Thus Heinrich Hertz accounted for the possibility of scientific explanation by reference to the nature of representation. Science forms pictures (Bilder) of reality. But we must sharply distinguish the empirical and the a priori elements of science. Scientific theories are not only determined by experience, but are actively constructed according to formal constraints - the 'laws of thought' (as well as pragmatic constraints like simplicity and clarity). These laws impose restrictions on our models of reality: they lay down preconditions for the scientific representation of reality which are analogous to Kant's transcendental preconditions of experience (1894, pp.1, 8). II. The Tractatus and the Linguistic Limits of Thought These two strands, the debate about the nature of logic and the Post-Kantian debate about the nature of philosophy, merge in the early Wittgenstein. The point of contact is the notion of thought. As we have seen, both the discussion about logic and the Kantian reflection on the nature of representation make reference to the laws of thought. And it is to thought that Wittgenstein turns in the famous Preface to the Tractatus Logico-Philosophicus. Thus the aim of the book is to draw a limit to thought, or rather - not to thought but to the expression of thoughts: for in order to be able to draw the limits of thought, we should have to find both sides of the limit thinkable (i.e. we should have to be able to think 156
Philosophy, Thought and Language what cannot be thought). It will therefore only be in language that the limit can be drawn, and what lies on the other side of the limit will simply be nonsense. Wittgenstein did not accept that thoughts constitute the subjectmatter of logic, a realm of entities which logic analyses just as chemistry analyses molecules. Partly inspired by Schopenhauer and Hertz, he draws a Kantian contrast between science, which pictures or represents the world, and philosophy, which reflects on the nature of this representation. Philosophy has the task of setting limits to thought. It is not a doctrine which extends our knowledge, but a critical activity which clarifies non-philosophical thoughts, namely those of science, and demarcates them from metaphysical confusions (4.1 Iff., 2.22-1, 4.021, 6.53 - incidentally, Kant anticipated some of these 'subversive' methodological ideas when he maintained that philosophy is more concerned with preventing error than with extending our knowledge: see A851/B879). Consequently, philosophy is concerned with thought by virtue of reflecting on the nature and limits of representation. For it is in thought that we represent reality. A sequence of signs is a meaningful proposition which truly or falsely depicts reality only if it expresses a thought. At the same time Wittgenstein gives a linguistic twist to the Kantian tale. Pace Platonism and psychologism, thoughts are neither abstract nor mental entities, but simply unasserted propositions with a sense. And a proposition (Satz) is not the name of an entity (a truth-value, as in Frege, or a complex object, as in Russell) but simply a sentence-in-use, a propositional sign which has been projected onto the world (3.Iff.; 3.32ff., 4). Consequently, thoughts can be completely expressed in language, and philosophy can draw limits to thought by establishing the limits of the linguistic expression of thought.3 Equally, these limits must be drawn in language. They are set by 'sign-rules' (Zeichenregeln), the 'rules of logical syntax', which determine whether a combination of signs is a meaningful symbol. A combination of signs like 'This square is round' is not just false, but simply nonsensical. However, any attempt to describe such a logical impossibility, if only for the purpose of excluding it as illegitimate (as in Russell's theory of types), is equally nonsensical, since what it tries to exclude cannot be referred to by a meaningful expression (3.03, 3.3ff., 6.02, 6.124-6). Therefore the limits of 3
My account leaves out the dark, psychologistic side of the Tractatus, according to which thoughts are not identical with meaningful propositions, but things which stand between meaningful propositions and reality (see Glock 1996, pp. 357-8). 157
Hans-Johann Glock thought cannot be drawn by propositions talking about both sides, but only from the inside, by delineating the rules which exclude certain sign-combinations as meaningless. What lies beyond these limits is not unknowable things-in-themselves, as in Kant, but only nonsensical combinations of signs. These linguistic rules also explain the nature of logical necessity. Wittgenstein's conception of philosophy and of logic are intimately linked. He had an abiding conviction, inspired by Russell, that philosophical problems are logical in nature. However, whilst he took over (and transformed) important elements of Frege's and Russell's logical systems, his understanding of the character of logic (and hence of philosophy) departed radically from that of his predecessors. The Tractatus incorporates Frege's rejection of psychologism (4.1121, 6.3631, 6.423). But it also challenges the assumption which Frege's Platonism shares with psychologism and Russell's empiricism. Although they disagree about the nature of these entities, they agree that logic is a science which makes statements about objects of some kind, just as empirical sciences make statements about their objects. The 'fundamental thought' (4.0312) of the Tractatus is to undermine this conception of logic as a science dealing with objects. In the first instance, Wittgenstein attacks Frege's and Russell's idea that 'logical constants' are names of entities: the role of prepositional connectives and quantifiers is not to name objects of any kind, but to express truth-functional operations. His ultimate target, however, is the resulting view that necessary propositions are statements about entities. According to the Picture Theory, the only genuine propositions are pictures of possible states of affairs. Such pictures are bipolar - capable of being true, but also capable of being false - and hence cannot be necessarily true (2.225, 3.04ff.). By contrast, necessary propositions are not statements at all. They do not represent a special kind of object, but reflect the rules of logical syntax. These rules antecede questions of truth and falsity; they cannot be overturned by empirical propositions, since nothing which contravenes them could count as a meaningful proposition. As a result they promise an account of necessary propositions which eschews psychologism, empiricism and Platonism. The special status of necessary propositions is due not to the abstract and sempiternal character of their alleged referents, but to the fact that they are linked to the pre-empirical preconditions of representation. The propositions of logic are tautologies (6.Iff., 6.126ff.)- They say nothing about the world, since they combine empirical propositions in such a way - according to rules governing truth-functional 158
Philosophy, Thought and Language operations - that all factual information cancels out (6.121). A statement like '~"(p A ~>p)' is neither a statement about how people think, nor about the most pervasive features of reality, nor about abstract entities in a Platonic hinterworld. Instead, it reflects a rule for the employment of signs like 'A' and '""', a rule which excludes a combination like 'both p and not p' as senseless. Like Kant's 'transcendental logic' (A55-7/B79-82), Wittgenstein's logic is a 'logic of representation (Abbildung)' (4.015). While contingent propositions are rendered true or false by reality - how things are - necessary propositions in general, and philosophical propositions in particular, reflect the necessary preconditions for representing reality. In contrast to Kant, these preconditions no longer reside in a mental machinery which constructs the phenomenal world out of intuitions, but in a system of rules for the employment of signs. Moreover, the principle of bipolarity stipulates that only empirical propositions can be meaningful, and thereby puts on the index all necessary propositions apart from logical tautologies, which are degenerate empirical propositions. This excludes synthetic judgements a priori, which Kant considered to be the metaphysical harvest of his transcendental labours. For Wittgenstein, the only expressible necessity is logical necessity, which is tautologous and hence analytic (6.Iff., 6.126ff.). This last claim stimulated the logical positivists of the Vienna Circle. Their aim was to develop a form of empiricism that could account for logical necessity without reducing it to empirical generality or admitting synthetic a priori truth. Necessary propositions, they argued, are a priori, but they do not amount to knowledge about the world. For, with the help of the Tractatus, it seemed that all necessary propositions could be seen as analytic, i.e. true solely in virtue of the meanings of their constituent words. Far from mirroring super-empirical essences, necessary propositions are true by virtue of the conventions governing our use of words. In fact, however, Wittgenstein's early conception of logic is closer to Kant than to the conventionalism of the logical positivists. The rules of logical syntax, though linguistic, are not arbitrary conventions. Just as Kant's categories must be present in any conceptual scheme, the rules of logical syntax must be present in any 'symbolism', any sign-system capable of depicting reality. Nevertheless, there are three respects in which the early Wittgenstein initiated the linguistic turn. Firstly, although the Tractatus did not treat them as arbitrary conventions, it made rules of symbolic representation central to logic, giving it a linguistic orientation that it has retained ever since. Secondly, although it was left to logical positivism and the later Wittgenstein to implement the 159
Hans-Johann Glock programme, it was the Tractatus that first insisted that 'all philosophy is a critique of language' to be pursued by logico-linguistic analysis (4.003ff.). Finally, since logic comprises the most general preconditions for the possibility of symbolic representation, there is no such thing as a logically defective language. The systems of Begriffsschrift and Principia Mathematica are not ideal languages, as Frege and Russell held (see 4.002, 3.323-3.325). They are not in better logical order than ordinary language, but are better at revealing that order, an order which must be present - if only under the grammatical surface - in any natural language (for this reason the Tractatus, rather than Frege, is the real source of the immensely influential programme of constructing a theory of meaning for natural languages). All three points start out from the Kantian idea that philosophy reflects on the preconditions of representing reality, an idea familiar to the Kantian tradition, but alien to Frege's Platonism and Russell's empiricism. A crude summary of this reflective turn is provided by the following scheme: SCIENCE / COMMON SENSE
Dep iction
LP)
F
onsciiousnes
3
(Hu sserl)
epresentatio
•-t
r
n
JO
chopenhaue
p H Y
Expexience
o
reflects on
ant)
P H I L O S
H X O
c o X
F
REALITY / WORLD
III. Thinking in' - a Rethink Thus Wittgenstein's transformation of Kant's reflective turn is the root of the linguistic turn of twentieth-century analytic philosophy. It also sets the stage for the present conflict between mentalist and 'lingualist' conceptions of thought to which Dummett and others (e.g. Hacking 1975; Tugendhat 1976) have alluded. Both sides agree that we represent the world through thoughts. But whereas mental160
Philosophy, Thought and Language ists claim that the medium of thought is the mind, and that philosophy should hence investigate the nature of representation by investigating mental structures, their lingualist opponents retort that the medium of thought is language and that philosophy should reflect on the way we speak about the world. In many ways, therefore, this conflict is a continuation of the conflict between two versions of Kant's reflective turn: the original mentalist version reflects on the way the world is given to us in pre-linguistic experience or consciousness, while the linguistic version reflects on the symbolic representation of reality. This provides a historical explanation of how the contrast between thought and language came to seem crucial to the very nature of philosophy, or so I hope. But the terms in which that mind v. language debate has been conducted remain unclear. Expressions like 'representation' or 'thought' have been used in an extremely general way which skates over the differences between phenomena such as thinking of or meaning an object, reflecting on a problem, believing or opining that p, having an occurrent thought at a particular moment in time, perceiving or imagining something, etc.4 T h e idea is that in all these cases our thoughts represent or depict reality, and that they can do so either in the medium of the mind or the medium of language. But what could that mean? A straightforward mentalist answer is given by Locke's representationalism: we are aware of external objects only via mental representatives which constitute the interface between the knowing subject and objective reality. To represent reality is simply to have such 'ideas'. However, this model is no longer fashionable, even among mentalists. It implies that we never know anything immediately other than representations, and for this reason leads to idealism or scepticism. In any event, it is implausible to claim that a judgement like 'The table is made of wood' is not about a table, but about the mental representative of a table. Most mentalists (in both the phenomenological and the empiricist traditions) have therefore conceded that we usually think about mind-independent objects, but maintained that we do so in mental representations, by having mental representations (Husserl 1900, II. 1; Mackie 1976, p. 43). However, this leaves open the question of what representations are. Once more, the British 4 I shall not give prominence to these differences in what follows, since it would clutter the presentation intolerably. My main concern is with two phenomena: that of reflecting on a problem, and that of occurrent thoughts, and it should be obvious from wording and context which of these is at issue in any specific passage. 161
Hans-Johann Glock empiricists provide a straightforward answer: representations are mental images. To think about objects is to have mental images of them which resemble the objects thought about. Now, it would be foolish to deny, as some behaviourists have done, that when one is thinking mental images may cross one's mind. However, such inner goings-on are neither sufficient nor necessary for me to think. In a delirium I may have mental images but do not think, and I may think about a problem without any images crossing my mind. Not all of our mental life can be characterised as having mental images, a point Berkeley and Kant made vis-a-vis 'general ideas' or concepts. Indeed, not even imagining is always a case of picturing something to oneself. It makes sense to imagine things which it makes little sense to picture to oneself, e.g. Queen Victoria's last thoughts, or that there are prime numbers greater than 1010. Mental images are accompaniments of thinking. They may give rise to thoughts and serve as heuristic or mnemonic devices. But they do not determine what I think, and having them is not logically necessary for me to think. In short, they do not constitute a universal medium of thought. This objection has no bite against mentalist positions like that of Husserl, which try to dissociate the idea of representation from its pictorial connotations. But such positions once more face the task of explaining what having a representation amounts to. In Husserl's case, for example, we seem to be left with the idea that it is 'just like' mental picturing, only without mental images. But that simply boils down to saying that to have a representation of X is to think about X, which means that the explanation of thought has moved in a circle. A logical relation, thinking about X or believing that£, has first been construed as a pictorial one, and has then been robbed of the pictorial aspects which alone can give it any substance (Tugendhat 1976, pp. 62-3, 276-7). At this point the lingualist might claim victory and maintain that talking about objects is an indisputable and clear way of representing reality. Furthermore, whereas our access to the alleged mental representations (ideas, thoughts, mental states or acts) involves the contested idea of introspection, linguistic behaviour is intersubjective and straightforwardly perceptible. It is interesting to note that Wittgenstein adopted such a position during the early thirties. He suggested that philosophy is a 'descriptive science...of thought'; but thought and its internal relations 'must be examined through the expressions conveying them'. A little later, he drew the conclusion that philosophy 'could be called the descriptive science of speaking, in contrast to that of thinking'. But the conflict between these two characterisations of 162
Philosophy, Thought and Language philosophy is only apparent, since at this stage Wittgenstein identified thinking with speaking; thought with its linguistic expression. He called thought a 'symbolic process' and thinking 'the activity of operating with signs', an activity which is performed by the hand in writing - or mouth and larynx - in speaking (1980b, pp. 4, 25; 1958, p. 6; 1993, p. 163). This is obviously wrong. While we write with our hands, we only think with them in the very special sense that we may be prone to accompany our speech by exuberant gestures. Wittgenstein himself soon changed his mind and admitted that thinking and speaking are 'categorially distinct' (1980a, §7; my account of Wittgenstein's discussion of thinking is indebted to Hacker 1990). He was right to do so. T h e lingualist claim that language is the universal medium of thought fares no better than its mentalist counterpart. Saying that/) and thinking that/) are obviously not the same. We do not express all of our thoughts in words - and thank goodness for that. Equally, we sometimes say that/) when we think that q. A lingualist might reply that in such cases we talk to ourselves inforo interno, and that thinking is a kind of internal monologue (as Plato had suggested in Theaetetus 189e). But speaking to oneself in the imagination is no more sufficient or necessary for thinking than having mental images. When I count sheep in order to induce sleep I talk inwardly, but I don't think; and one can perform even the most complex intellectual tasks without talking to oneself in the imagination. Moreover, it is implausible to suppose that mechanical activities like driving have to be accompanied rapidly by the words we would use in subsequently expressing our thoughts (e.g. 'You fool; there is a radar control behind the bridge, you had better slow down to fifty!'). In such situations rapid physiological processes occur, but these are arguably incapable of having symbolic content, and certainly do not amount to words crossing the subject's mind. Thus the debate between mentalist and lingualist conceptions of thought seems to end in a stalemate. We might try to overcome this deadlock by striking a compromise. 5 But a compromise between two misconceptions is not a promising idea, outside of practical politics, that is. I suggest that instead we should take a critical look at two premises which both sides of the debate accept. For one thing, both mentalists and lingualists presuppose that 5 In this vein, Dummett (1993, ch. 12) has suggested that while genuine thought is essentially linguistic, animal thoughts or mechanical activities like driving involve 'proto-thoughts' which cannot be adequately explained in language. 163
Hans-Johann Glock there must be certain occurrent events which determine (not just in a causal sense) what we think and whether we think. But our previous discussion suggests that internal goings-on, whether mental imagery or words, are irrelevant. What we think is evident from what we say and do (if we are sincere), not from what images or words may flit across our minds. A motorist can be credited with the aforementioned thought if she sincerely avows it - either then or later. Similarly, whether or not I thought about a problem on a given occasion is not determined by the presence or absence of internal accompaniments, but by what I am capable of doing, by the way I speak and act, and it may well depend on what went on before or after. When I studied mathematics as an undergraduate, we asked one of our professors to have a look at a problem which had withstood our sustained efforts. He glanced at the problem, briefly stared out of the window, walked to the blackboard, and sketched the solution. In reply to our question 'How did you do that?' he said 'Well, it's simple. I glanced at the question, stared out of the window, walked to the blackboard and wrote down the solution.' Of course, in some circumstances we would deny that the professor had been thinking. If he just reproduced an answer that he had arrived at previously, we would certainly do so. We might also hesitate if the solution occurred to him in a flash, for example on the grounds that this is less a case of thinking than intuition. But what if at successive stages of this short tale, he would have given different answers to questions like 'What are you thinking just now?' or 'Where have you got to with this problem?' In that case there is no denying that he was thinking through the problem, and yet we would not be entitled to conclude that he must have been lying when he denied that specific mental images or symbols crossed his mind. If these arguments are on the right lines, what we think is not determined by anything going on 'in the head'. What people think is evident from what they avow to be thinking - if they are sincere - and from how they explain their thoughts when challenged. None of this is to deny that there is a host of mental or physiological accompaniments of thinking. But these goings-on do not constitute our thoughts. They are causal preconditions for my having thoughts, but in a logical or conceptual sense they are neither necessary nor sufficient for my having a particular thought. To put it in Aristotelian terms, even those types of thinking which most suggest an actual going-on, namely, reflecting on a problem and having an occurrent thought like that of our motorist, are more akin to a potentiality than an actuality. This may sound paradoxical. But all it amounts to is this: when we ascribe thoughts to others, we draw a 164
Philosophy, Thought and Language cheque not on what goes through their minds, but on what thoughts they would sincerely avow, and on how they would explain and justify their actions and utterances.6 The second assumption of the debate between mentalism and lingualism which I want to question is that there must be a medium of thought. Both sides presuppose that thought requires a 'vehicle', and that we must always think 'in' something, either words or images. But after all, thoughts are not passengers and language is not a means of transportation. What seems to give substance to these metaphors is the fact that one can talk inwardly in a particular language. But this is not the same as to think in a particular language. For, as we have seen, interior monologue is neither necessary nor sufficient for thinking. Arguably, the question of what language I think in arises only with respect to a foreign language. And there it boils down to questions such as these: Do I speak that language hesitantly? Do I have to decide first what I want to say and then try to remember the equivalent in the foreign tongue, or can I simply say it? But there is no need to suppose that in general I must first think in some symbolism, linguistic or mental, and then transpose my thoughts into utterances of a different symbolism. That picture — enshrined in Fodor's idea of a language of thought - has the absurd consequence that I might always be mistaken about even the most simple of my own thoughts. For I might first read them off incorrectly from my internal display of words or images, and then mistranslate them into the public language. It might be objected that this picture is supported by certain indisputable phenomena. For example, people often say that a certain sentence or phrase is on the tip of their tongue. People also sometimes say that they know what they think, but are incapable of expressing it. The language of thought picture would indeed explain these phenomena. But it is not the only explanation, or even the most plausible one. When a word is on the tip of my tongue, or when I am groping for the right word to express my thought, I am facing a 6 This is not to say that thinking is a disposition. For one has to find out what one's dispositions are through noting how one is prone to react in certain circumstances. But one does not normally find out what problem one is reflecting on or what one's occurrent thoughts are. Moreover, dispositions are inevitably realised given certain circumstances. By contrast, human beings need neither avow nor explain their thoughts. It is nevertheless tempting to provide a counterfactual analysis of thinking, e.g. A thinks that/) at t,:= if A had been asked at t, what she thinks, she would have said that p. For a discussion of the problems facing this kind of approach see Glock and Preston (1995). 165
Hans-Johann Glock problem that is different from a problem of translation. I am not asking myself which English phrase is the correct translation to a mentalese phrase that has come up on my inner display. Rather, in the first case I am trying to remember a sentence or phrase that I think adequately captures the situation I am trying to describe or the emotion I am trying to express. And in the second case I am trying to come up with such an appropriate sentence or phrase. So even these seemingly favourable cases do not force us to adopt the idea of a language of thought, and that idea remains wholly implausible in the vast majority of cases, in which we express our thoughts unhesitatingly. IV. Human Beings as Language-using Animals Both mentalism and lingualism are diverse and resourceful paradigms. The foregoing considerations do not purport to refute them conclusively. But they do provide arguments against the idea that unexpressed thoughts must always be in something (words or images). If these arguments are sound, the whole controversy between mentalism and lingualism about the medium of thought, from Kant through Husserl and the Tractatus to Dummett, may turn out to be misguided. Language should not matter to philosophy because it is the universal medium of representation. This does not mean, however, that language plays no special role in philosophy. There are, after all, essential links between thought and language, but they do not require any actual inner vocalisation. Both mentalism (psychologism) and Platonism (anti-psychologism) believe that the thoughts expressed in words - the meanings of what is said - must be entities beyond language. Wittgenstein (1967, §§316-17; 1958, pp. 147, 161; 1980a, I/§§579-80) and Ryle (1979) questioned that assumption. Although thoughts are not identical with their linguistic expression, they are not entities beyond language either. The answer to the question 'What are you thinking?' is neither a description of an inner process, nor a description of an elusive third realm, but an expression of my thought in words, e.g. 'I think that it will rain.' If I am challenged to express the thought behind that utterance, I do not re-examine some inner process or abstract entity to see whether I can describe it better. Rather, I paraphrase my utterance into other words. Similarly, two people think the same thought not by sharing a qualitatively identical mental entity, or relating to a numerically identical abstract entity, but if the expressions of their thoughts say the same thing. Language is not just the only expression of thought, i.e. a distorting expression of something beyond language, as Frege had it. It is the only genuine expression. 166
Philosophy, Thought and Language The first essential link between thought and language, then, is that we identify thoughts by identifying their linguistic expressions. The second essential link is that the capacity for thought requires the capacity to manipulate symbols, not because unexpressed thoughts must be in a language, but because the expression of thoughts must be. The reason is that ascribing thoughts makes sense only in cases where we have criteria for identifying thoughts. In Quine's memorable phrase, 'No entity without identity' (1969, p. 23). Something must count as thinking that p rather than that q, otherwise ascriptions of thoughts are vacuous. This means that thoughts, although they need not actually be expressed, must be capable of being expressed. And only a very restricted range of thoughts can be expressed in non-linguistic behaviour. To use Wittgenstein's famous example, a dog can think that its master is at the door, but not that its master will return the day after tomorrow (1967, p. 174, §650, see also §§342, 376-8). The reason is that its behavioural repertoire is much too restricted to display such a thought. These reflections suggest that philosophy's concern with language is not just a quirk of twentieth-century philosophy dons or hermeneutics professors. They do not support Dummett's assumption that the basic task of philosophy is to investigate thought. But they do imply that in so far as philosophy is concerned with thought, it is concerned not with mental or physiological accompaniments of thought, but with the meaningful expression of thought. Philosophical problems concerning the various areas of human thought (common sense, science, morals, art) should be clarified by attending to the way we express such thoughts in language. Of course, as it stands, this statement is far too vague. But it can be given more specific content. To mention just one example, philosophers and psychologists have recently taken a keen interest in what Kant called the unity of consciousness. One of the things that has baffled them is the apparent fact that it is impossible to have more than one thought at the same time. And many have sought to explain this 'linear nature of consciousness' by reference to certain mental mechanisms. But a solution to at least one aspect of this riddle can be sketched by reference to the linguistic expression of thought. It is at the very least an oversimplification to hold that we cannot have more than one thought at the same time. Obviously, it is not impossible to hold more than one belief at any given time. But it also problematic to suggest that we cannot think about two objects at the same time. The fact that some people manage to avoid letting the milk boil over in the morning shows that it is not impossible to pay attention to two things at once, the milk on 167
Hans-Johann Glock the stove and the news on the radio, for example - although, alas, this gift evades me personally.7 In my view, one of the things that the alleged impossibility of having more than one thought at the same time boils down to is the following. It is impossible verbally to specify two thoughts simultaneously. But the reason for that does not lie in the 'linear structure' of the mind, but in essential features of the linguistic expression of thoughts: there is no such thing as the concurrent expression of two thoughts by a single speaker. Take my two thoughts that torture cannot be justified and that the Gulf War is over. Any attempt to express these two thoughts concurrently would amount to something like 'torture the cannot Gulf War be is justified over'. However, that string of signs is not a meaningful proposition, but garbled nonsense, which implies that the idea of concurrent thoughts in this sense is incoherent. Finally, the above reflections do not just have methodological implications for the philosophical approach to thought. They also support and radicalise the venerable Aristotelian idea that human beings are essentially language-using animals. Those features that have, at various times, been thought to distinguish human beings from all other creatures — e.g. a capacity for knowing necessary truths, the possession of a moral sense, self-consciousness or a sense of history — are all derivative from our distinctive language-using abilities.8
References Boole, G. 1854. An Investigation of the Laws of Thought. Reprinted, New York: Dover, 1958 Derrida, J. 1967. Of Grammatology. English translation, Baltimore: Johns Hopkins University Press, 1976 Dummett, M. A. E. 1991. Frege and other Philosophers. Oxford University Press 1993. The Origins of Analytical Philosophy. London: Duckworth Frege, G. 1893. Grundgesetze der Arithmetik, Volume I. Hildesheim: 01ms. English translation, Berkeley: University of California Press, 1964 1969. Posthumous Writings. English translation, Oxford: Blackwell, 1979 7
For a different and much more elaborate view, see White (1964). For comments on previous drafts of this paper, I should like to thank Bob Arrington, John Cottingham, Peter Hacker, John Hyman and John Preston. I am also grateful for feedback from audiences at Oxford, Atlanta, Reading, Lampeter and Swansea. 8
168
Philosophy, Thought and Language Gadamer, H. G. 1967. Philosophical Hermeneutics. English translation, Berkeley: University of California Press, 1976 Gellner, E. 1959. Words and Things. London: Gollancz Glock, H. J. 1992. 'Cambridge, Jena or Vienna? The Roots of the Tmctatus', Ratio 5, 1-23 1996. A Wittgenstein Dictionary. Oxford: Blackwell 1997. 'Kant and Wittgenstein: Philosophy, Necessity and Representation', International Journal of Philosophical Studies 5, 139159 Glock, H. J. and Preston, J. M. 1995. 'Externalism and First-Person Authority'. The Monistic, 515-33 Hacker, P. M. S. 1990. Wittgenstein: Meaning and Mind. Oxford: Blackwell Hacking, I. 1975. Why Does Language Matter to Philosophy? Cambridge University Press Heidegger, M. 1927. Being and Time. English translation, Oxford: Blackwell, 1962 Hertz, H. 1894. The Principles of Mechanics. English translation, London: Macmillan, 1899 Husserl, E. 1900. Logische Untersuchungen. Tubingen: Niemeyer 1913. Ideen zu einer reinen Phdnomenologie und phdnomenologischen Philosophie, in jfahrbuch fiir Philosophie und phdnomenologische Forschung
1. Reprinted in Husserliana, Volumes III-V. The Hague: Nijhoff, 1980 Mackie, J. L. 1976. Problems from Locke. Oxford University Press Quine, W. V. 1969. Ontological Relativity and Other Essays. New York: Columbia University Press Russell, B. A. W. 1903. The Principles of Mathematics. London: George Allen & Unwin 1912. The Problems of Philosophy. Oxford University Press 1956. 'The Philosophy of Logical Atomism', in Logic and Knowledge. London: Allen and Unwin 1984. Theory of Knowledge. The 1913 Manuscript. London: Routledge Ryle, G. 1979. On Thinking. Oxford: Blackwell Tugendhat, E. 1976. Traditional and Analytic Philosophy. English translation, Cambridge University Press, 1982 White, A. R. 1964. Attention. Oxford: Blackwell Wittgenstein, L. 1922. Tractatus Logico-Philosophicus. London: Routledge 1953. Philosophical Investigations. Oxford: Blackwell 1958. The Blue and Brown Books. Oxford: Blackwell 1967. Zettel. Oxford: Blackwell 1980a. Remarks on the Philosophy of Psychology, Volume I. Oxford: Blackwell 1980b. Wittgenstein's Lectures, Cambridge 1930-32. Oxford: Blackwell 1993. Philosophical Occasions. Indianapolis: Hackett
169
The Flowering of Thought in Language W. V. QUINE Our first mental endowment was instinct. Then came thought, and later language. Thanks to language, thought then proceeded to flourish. Such was our phylogeny. Ontogeny, then, true to form, recapitulates the sequence in the development of each child. By instinct a child a few days old will show anxiety when an object moving steadily to a screen and behind it fails to emerge on schedule at the far edge. Also new-borns are said to respond emotionally to emotional facial expressions and even to imitate them. A further innate endowment, and an elaborate one, is manifest in one's standards of perceptual similarity. This relation is basic to all learning, so some of it has to be innate. We are born with standards of perceptual similarity, which then develop and change in the course of experience. I treat perceptual similarity as a relation between stimulations, or sensory intakes. The similarity is subjective, each individual having his own standards, but it is objectively testable for each individual by reinforcement and extinction of responses. It does not match up with receptual similarity, that is, mere identity or proximity of the nerve endings that were triggered on the two occasions. Perceptual similarity is illustrated rather by our readiness to equate the various perspectives of an object, prior even to probing it. It is a striking case of innate knowledge, however inarticulate. We have a further instinct that parlays perceptual similarity into an ever widening fund of knowledge: the instinct of induction. This instinct is an innate disposition to expect perceptually similar stimulations to have sequels that are perceptually similar to one another. Flashes of lightning are similar to one another, and their sequels, thunderbolts, are likewise similar to one another. This inductive instinct is the basis of expectation and hence of habit formation and learning. So induction engenders expectation and hence belief, but how knowledge? how truth? Here Darwin lends a hand, the creative hand of natural selection. Successful expectation has had survival value down the ages in helping our forebears avoid their predators and capture their prey. Those who lived long enough to be forebears did so through predominantly correct expectation. Their genes car171
W. V. Quine ried standards of perceptual similarity that meshed pretty well with environmental trends. Thus it is that induction has tended to engender not just expectation but successful expectation, hence knowledge, or anyway true belief. Such is our innate head-start, or part of it, in the business of thinking. It is a head-start that we share with other higher animals, whose standards of perceptual similarity can be elicited experimentally in the same way as ours. Up to this point, we and they were perhaps about equally endowed. Then, with the development of language, human thought proceeded to soar. In respect of cause and effect, language and thought have been intimately intertwined: language as vehicle of thought, thought as innovator of language. I despair of disentangling them. But I shall delineate a succession of advances in language and thought that may have brought us to the level of language and thought that we enjoy today. Birds and apes do have audible signals that are precursors of speech. Chimpanzees are said to have distinctive cries by which to warn one another of predators: one cry for lions and leopards, another for eagles. A third cry may report fruit at some treetop. The cries serve the basic function of language, namely a social sharing of information. The individual gets the benefit of his neighbour's perceptions. The human infant's first acquisition in the way of cognitive language is comparable to the apes' cries: 'Mama' perhaps, or 'Milk' or 'Ball'. Whether the apes' cries are innate or learned is as may be, but the child's words are learned; English has not made its way into the genes. The child learns the word by ostension: the ball is in focus when he is told the word 'Ball', and he makes the association. An imitative instinct is probably at work, imitation of the mother's word; for such an instinct is proverbial in man and ape. At recurrence of the ball, the child is prepared to volunteer the word or assent to it. He has learned to associate the word with a range of mutually fairly similar neural intakes, similar in featuring the ball. The similarity is perceptual similarity by the child's standards. There is a richness here that escapes the casual eye. The recurrences that continue to elicit the word, or to elicit assent to it, from both the child and the mother, are similar for the mother and for the child, but similar according to their separate, private scales of perceptual similarity; no inter subjective perceptual similarity is posited or even defined. Yet the two scales keep pace. Failing such harmony, the second ball-sighting might have resembled the first for the child but not for the mother, or vice versa. This does happen, but not usually. The harmony is this: two shared occasions that are perceptually similar for one witness tend to be so for another. 172
The Flowering of Thought in Language Without this harmony there could be no appreciable communication. The harmony is required not only in the sightings of the ball, but also in the sound of the word. It must sound similar to the child and similar to the mother when spoken by the child and by the mother, and on one occasion and another, though no similarity is assumed between how it sounds to the child and to the mother. The cause of the harmony becomes evident when we recall the cause of the perceptual similarity standards themselves; namely, natural selection, moulding them to suit the trends of the environment. Shared ancestry in shared environment tuned us alike, nearly enough, to our continuing intake. We have the child learning 'Mama' or 'Milk' or 'Ball'. This much is on a par with the apes' cries. Each such word is associated with some range of perceptually more or less similar neural intakes. The child goes on learning further ones hand over fist, promptly outstripping the apes' meagre repertoire. Presently the child masters the use of prepositions, conjunctions and the copula to combine simple expressions, and thus moves into mass production. My examples have been nouns: 'Mama', 'Milk', 'Ball'. But thesentence 'It's cold' or It's raining' would do as well, and the three nouns can be thought of as sentences too: 'That's Mama', 'That's milk', 'That's a ball.' But they all are occasion sentences: true on some occasions, false on others. Sometimes it is raining, sometimes not. Besides this general trait, their distinctive trait is that each has become associated, for the speaker in question, with some range of perceptually more or less similar neural intakes. Any one of them would prompt his unhesitating assent. I call such an expression an observation sentence for the speaker in question. We can imagine a stage in the prehistory of man when his cognitive language was limited to such observation sentences. It was language only by courtesy, for it lacked the recursive power of true grammar that generates essentially new sentences without limit. The conjunctions, prepositions and copula are productive of new observation sentences up to a point, but the added requirement of unhesitating assent in my definition of 'observation sentence' limits their length. Thus far the advantage over the apes' meagre endowment is purely quantitative, but enormous. My fancied Urmensch can not only broadcast a fine-tuned roster of species that he sees approaching or ripening. He can also talk to himself about sealing up his shelter. 'A bit of clay here', he ponders. 'Water on clay. Flat stone in here'. Like the modern engineer, he also invokes arm and finger muscles in planning the moves, and then his vocabulary enables him to tell his son to help him find what he needs. Gestures fill in for a few 173
W. V. Quine verbs. Even a bit of logic can seep in. In learning the conjunction 'and', he learns simply as a matter of usage not to affirm a sentence of the form 'p and q' without being prepared to affirm 'q'\ and similar points of usage regarding 'or', 'not', and 'if. Violations of logic would be like calling a lizard a snake: a misuse of words. At some later point in the dim wastes of antediluvian time man achieved a more resounding linguistic breakthrough; and at some point our child recapitulates it, picking it up from us. The new device puts the yield of our instinct of induction into words. It expresses generalised expectation by combining observation sentences in pairs: thus 'When it's snowing, it's cold'; 'When lightning, thunder'; 'When the sun rises, birds sing.' I call the compounds observation categoricals.
The survival value of our old observation sentences and the apes' cries, as I remarked, is vicarious observation. The value of the observation categoricals, which generalise on those observation sentences, is vicarious induction. They convey to us the inductive generalisations already achieved by our elders, and they can be handed down the generations. This linguistic innovation, putting the output of our instinct of induction into words, marks the advent of standing sentences: sentences which, unlike our old observation sentences and other occasion sentences, are true or false once for all rather than on occasion. And not only that; they are generalisations stating, when true, scientific laws. The observation categorical is a universal conditional of a sort, but its two clauses depend on no sharing of subject matter. To us in our sophistication, generality connotes quantification, and the 'When' in my examples suggests quantification over times; but all this is anachronistic in the present context. The intent of the primitive observation categorical is just to counsel the hearer to expect fulfilment of the antecedent clause to be accompanied or promptly followed by fulfilment of the consequent. But what I picture as the next great prehistoric development is indeed the genuine universally quantified conditional, the classical categorical 'All ravens are black' or, phrased as an observation categorical, 'When anything is a raven, it is black.' In symbolic logic it is the universally quantified conditional 'Vx(if x is a raven then x is black)'. I see this as the dawn of reification. It reifies whatever ravens there may be, and it is false if there are any white ones. The humble pronoun 'it' in the second of the above three renderings is an essential pronoun: it cannot be replaced by its grammatical antecedent 'anything'. The essential pronoun, or the 'x' that 174
The Flowering of Thought in Language is its mathematical version, is vital for thought and language because of the logical link that its cross-reference creates between its clause and that of its antecedent. The pure conditional 'If p then q' is only a weak link, contributing little to the structure of scientific theory. The universal conditional 'Vx(if Fx then Gx)' is what structures our theory with its texture of causal laws. I have pinned the idea of an object down to the essential pronoun or bound variable. Other idioms, intertranslatable with these, accomplish the same. An equivalent criterion applicable to other idioms is fulfilment of a predicate. I have favoured the former rendering because it rests on a more explicitly regimented, less mistakable notation. The medieval nominalists took a wrong turning when they sought to de-reify universals by dismissing their names as flatus vocis or vocal breeze. You could likewise de-^reify particulars if it were just a matter of dismissing their names as vocal breeze. Caesar could go the way of Pegasus. Where reification sets in, whether of universals or particulars, is where we treat of them by predicate or general term rather than individually by name. Our quantification over ravens reified them without their even bearing any names, and our quantification over numbers and other classes reifies the unspecifiable real numbers and other unspecifiable classes no less than the specifiable ones. The full benefits of reification depended still on another vital achievement by early man, whether earlier or later than reification itself. It was the transcendence of the specious present: our ability to speak of the more remote past and future. In our familiar language this is managed by the tenses of verbs. In a language with the structure of the predicate calculus it is managed by treating physical objects as spatiotemporally four-dimensional and then explicating the predicates 'earlier than' and 'part of or predicates to that effect. This linguistic access to remote times extends the horizons of thought from the specious present to plus and minus infinity. It is only after that much development of the child and the race that it becomes meaningful to speak of identity of a recurrent object after an appreciable absence. This is an achievement that we tend not to appreciate. The dog eagerly welcomes his master after a long absence, but this is just a qualitative identification like the child's continuing recognition of milk, ball, rain, thunder and the mother. In the dog's case it is largely based on odour. We humans, and surely only we, learn to make sense of the question whether a red ball is the same red ball next week or another like it. Our decision or conjecture will depend on mastery of the sophisticated scheme of time, space and the unseen trajectories of objects between appearances. 175
W. V. Quine This imposing development reflects a substantial theory of nature and a subtle language to embody it. It is only after all this that man can be said to have fully reified physical objects; for it is only then that identity and distinctness of physical objects over time makes sense. This development was vitally important in tightening the structure of the four-dimensional world as we conceive it; for the continuing identity of a changing object is what integrates the time dimension, relating one time to another. Physical objects are time-binding, in Korzybski's happy phrase. It was with the advent of the universal conditional that science could begin to develop its texture of causal laws, but only with the transcendence of the specious present that these laws could proliferate. From then on the continuity of bodies over time and the causal connections between the times have been explained or conjectured hand in hand. The reification of abstract objects - numbers, notably, and other classes - ensued on the analogy of physical objects, and took the same linguistic form: they are values of essential pronouns, or bound variables. This move has been indispensable in mathematics and hence in the hard sciences generally. The special service of time-binding, however, is not one to which the abstract objects contribute. Being changeless, they are above all that. With science a going concern, the old observation categorical gets pressed into vital service as empirical check-point for scientific theory. The primordial adverb 'When' with which I apologetically formulated it can now be taken frankly as quantifying over times, making the observation categorical a categorical in the classical sense of a universal conditional. But its two clauses are still, in this new use, observation sentences. However sophisticated in content, as I define them they must still command the qualified observer's unhesitating assent when the stimulation is clearly within the associated range. The assent is fallible and corrigible, but unhesitating. The checking of a theory then proceeds by logically deducing an observation categorical from the theory, and testing the categorical. The experimenter arranges fulfilment of its antecedent clause to his satisfaction and awaits fulfilment of the consequent.
176
Talking to Cats, Rats and Bats K. V. WILKES
This paper tries to argue that at least some alluring, trendy or fashionable problems to do with thought and language - several of which are discussed in this volume - are in fact alluring, trendy or fashionable red herrings or cul-de-sacs. I shall primarily be concerned with the ascription of thought and intelligence to non-language-users; but, en route to that, will need to brood over our ascriptions of such terms quite generally. Allowing myself what Weiskrantz has nicely called 'the generosity of vernacular usage' (this volume, p. 128), I shall use 'intelligence' in its wide everyday sense, with no scientific or philosophical technicalities imposed upon it; an umbrella term under which we allow, or refuse to allow, some communicative or expressive or problem-solving behaviours to fall. 'Communication', too, will be construed very broadly: as I shall be using it here, 'communication' can be purposive or not, recorded or recognised or understood or not, verbal or not, conscious or not, interpreted or misinterpreted, sincere or deceptive, simple or sophisticated. In such a liberal - but also harmlessly everyday reading of 'communicate', computers communicate; a scream communicates pain by expressing it; the road sign that says 'Diversion Ahead' may communicate very different things to a Canadian and a Briton. Language is undeniably the most striking and clear-cut form of expressive and communicative behaviour, heavily implicated in many kinds of sophisticated planning and problem-solving. It is usually overt; generally sensori-motor, involving the mouth, the hands, or other peripheral organs. (Granted, one can talk - communicate or express things - to oneself, in words, silently; but this is a highly sophisticated enterprise, coming well after the ability to utter or produce words and sentences; I shall not, in this paper, be concerned with it at all.) It is not, then, surprising that linguistic activity is a form of behaviour that is generally taken to be the most decisive indicator of communicative intelligence; but it should be considered along with other forms of behaviour that compete as indicators. Not only painting, sculpting, composing, humming, whistling; not only the 177
K. V. Wilkes artisan-like activities of making and using tools; but also pointing, gesturing, hiding, choosing, imitating, pretending, rejecting, selecting, facial expressions, posture, direction of gaze, eye contact - all these may display intelligence, as viewed in the 'generous' vernacular sense. As, of course, may many behaviours other than those listed above. I hope and believe that this much is uncontentious, once I am allowed the generous everyday readings of 'intelligent' and 'communication'. Although I shall come before long to non-human animals, I want for the moment to consider the implications that putting language in its place like this - as just one of the signs that typically indicate the presence of some degree of intelligence - may have for some familiar issues. We want to know whether, and to what extent, language or concept-mastery are essential for ascriptions of believing, of intelligence, intentionality, purposiveness, consciousness; in other words, wherever we find what seems indubitably to be thought in adult human activity. For this we need to seek out contexts where we are - to put it mildly - uncertain of what role (if any) language plays. A related further question is what role is or is not played by the ability to wield concepts in mental utterances that have, or are based upon, a specific propositional content. Familiar issues all. But just what are the problems? Notoriously, artisans, musicians, sculptors (etc.) are often poor at verbal communication: describing what they are doing or about to do. Describing a well-known face is a difficult technique which few can master; yet some talented people can draw faces well, and most of us, defeated by the demand to describe a familiar face, might nonetheless claim truthfully that we would at once recognise that person were we to see him. There are very many beliefs, desires, attitudes, feelings that skilled language-users try to ascribe (both to others and to themselves) where there is considerable uncertainty about what content to give that mental state; in the case of beliefs, what propositional content. Dennett (1981; reprinted 1987, pp. 84—7), uses a nice example of a boy selling lemonade who (unintentionally) gives a customer the wrong change from a quarter for a 12c drink; this makes the point well. What beliefs could the boy be said to hold that might explain his error? Another sort of example: how would we specify the content of my beliefs about neutral bosons, leptons or any partially understood, or partially forgotten, entity or event? 'Daddy has lost his bank statement': how can we cash-out the meaning such a sentence holds for a four-year-old? What do we say that the child believes? Compare: just where does the photographer want me to go, when he says 'stand roughly there' (pointing)? 178
Talking to Cats, Rats and Bats I have not so far, it should be noted, talked of subliminal, preconscious, subconscious, blindsighted, thoughts, feelings, attitudes, or beliefs; nor of such states as seem to show themselves only under hypnosis. I have not tried to look at what we are to say about the beliefs (etc.) of schizophrenics, nor of those with, say, hemisphere neglect, nor of those with one of the agnosias; nor of the beliefs of the right hemisphere of commissurotomy patients; nor (yet) of the thoughts of those with autism. For even if we leave these and other abnormalities aside for the moment, there is a vast array of thoughts, feelings or beliefs ascribed to normal adult humans in favourable circumstances, the content of which - propositional or not - cannot always be specified with precision. Even where there is propositional content, we may be uncertain whether we should call the ascription 'true' or 'false'. Compare a 'true likeness'; there is no one thing that fits this bill. A police- or passport-type photograph? An impressionistic painting? Certainly we are often justified in complaining about our passport photographs; an impressionistic painting may often be more revealing of the individual than the clearest and crispest studio study. That we find it difficult to specify the thoughts or beliefs of the dog that chases a cat up a tree, and continues barking excitedly up it even when the cat has escaped to another tree - this fact should not, then, attract nearly as much alarm and quibbling discussion as it has so far. After all, we find the same over and over again when communicating with each other. That the dog has no concept of a bone qua 'one of the parts making up a vertebrate animal's skeleton' need not stop us from ascribing to it a belief that a bone is buried just here. Vis-a-vis bones, the dog is not much worse off than the adult human who has never been told of the differences between vertebrates and invertebrates; nor than the pre-school child. Neither should the fact that the dog does not behave in the same way to a fishbone as it does to a bullock's shinbone, but that it might behave in the same way to a plastic or rubber one - i.e. that its 'concept' (note the scare-quotes; 'proto-concept', would do as well or as badly) has a different extension from that of the anatomist - disturb us. In the routine to-ing and fro-ing of everyday discourse we are simply not much bothered when ascribing millions of hazily specifiable beliefs to other adult humans. The professional anatomist can link up his knowledge of bones to a multitude of further backing beliefs, the consistency and coherence of which can be examined. It is these inferential relations that some think to be crucial for belief-ascription in general, and for the precision of the propositional content ascribed in particular. The more we can assume that our companions-in-discourse share a 179
K. V. Wilkes broad network of bone-related beliefs largely isomorphic to our own, the more straightforward and specific is the content we can ascribe to their thoughts. Too much inconsistency or oddity can cause us to modify or withdraw an earlier judgement. This is easy to illustrate. Even the professional anatomist might, for example, have bizarre beliefs about the composition of human bones (perhaps that they include some non-physical, 'spiritual', substance); beliefs that do not fit together with the rest of his or our chemical and biochemical understanding, but fit to some extent with some inchoate religious or creationist assumptions of his. Is he indeed thinking about human bones} Whatever we say, his bonethoughts in the normal case would still fit a richer and wider pattern of related thoughts and beliefs than would those of the layman, the dog or the child. This 'fitting' is a matter of degree, and certainly the smaller the network of inference available to an agent, the harder it may well be to pin down the content of a thought; we may want to withdraw the ascription altogether. For instance, what some non-scientist says about an electron may be largely false of electrons, but largely true — albeit still very skimpily so — about positrons. He believes something about something, surely; but there may be no such thing as 'the' right answer to what it is that he believes. It will clearly be a difficult and problematic enterprise to single out somehow 'the' belief that we want to ascribe on this occasion from the background of supporting knowledge or presuppositions to which we are also tacitly committing ourselves; and this background will be more like a cloth of many colours, light and shade, bits clear and bits obscure, some of it highlighted, some distorted and vague - not a list of propositions like beads on a string. I am thus untroubled by the difficulty of specifying (with our linguistic categories) the contents of a cat's, rat's or bat's mind. The difficulty is no worse, or better, than that of pinning down the content of the thoughts and anticipations of a composer or an artist, or of a non-scientist about abstruse or difficult theoretical entities. There will be a huge spectrum between vagueness and precision amongst our ascriptions of what is said to be thought, believed or felt. Some of the problems suspected here are pseudo-problems. II This may seem to be dismissing too high-handedly issues that have vexed so many. To build a more convincing argument I must go further back. I need first a distinction between common sense psy180
Talking to Cats, Rats and Bats chology1 (hence: 'CSP', and so 'CS' for common sense generally) and scientific psychology ('SP'). My arguments for this are given at greater length elsewhere;2 although I cannot assume that these have already been seen, I am reluctant to repeat myself. So here I shall simply summarise the main strands. First: CSP and SP are on a continuum; the difference is one of degree not of kind. That is unproblematic; so are the hills of Hampshire, and the Himalayas. Thus there will be a swathe of grey areas (when does a hill become a mountain?) where it is unclear, and does not matter much, whether something is put down as belonging to CSP rather than to SP, or vice versa. The same holds outside CSP, extending to CS vis-a-vis science generally. There will be a lot of blurred overlap between CS 'theories' of gravity - what will happen when we set a booby-trap - or of health - what happens if we eat no protein - for example, and the theories of physicists or medical dieticians.3 But just as the Himalayas are very different from the rolling hills of Hampshire, so is most of CSP from SP, and CS generally from science. ('Most', not 'all'; I repeat that there is substantial overlap.) Second: the objectives are different. We are encouraged to believe that one central ambition of science is to explain and predict systematically. This is broadly true. CSP does, quite often, explain and predict too. But it rarely does so systematically, for it infrequently seeks, uses or propounds laws, nor does it often conduct experiments to test its predictions. It is at its best, as far as its concern with the explanatory-predictive domain goes, when it is used to discuss what some specific, known agent might do or have done in a particular context at a given time. Even more to the point, though: it does thousands of other things other than predicting and explaining. We use it to console and congratulate, to warn and encourage, to express pride or shame, to joke and jeer, to hint and lie, to denounce, to punish, to deplore, to sympathise, to express pride or horror...the list is endless. CSP seeks to understand others, to form expectations, to imagine or conjecture what another wants or feels or thinks; it involves far, far more than the artificial construct of the decision' Along with most psychologists (but not philosophers) I will not call it 'folk psychology'. This makes the everyday vernacular psychology sound too twee and 'folksy', whereas it is in fact incredibly rich, subtle and sophisticated; moreover, Wundt has first claim on the use of the expression 'Folk Psychology' in his Volkerpsychologie (1900). In, for example, Bogdan (ed.) (1991), pp. 144-60; and Christensen and Turner (eds.) (1993), pp. 144-87. 3 Incidentally: the 'difference of degree' that holds between CSP and SP holds just as much for the term 'theory' itself; there is a huge difference of degree - between what passes as a 'theory' in CS/CSP, and in scientific theories. 181
K. V. Wilkes theorist trying to show what idealised perfectly rational creatures would do vis-a-vis each other in fixed circumstances. So it is used for indefinitely more than belief-desire prediction and explanation. Third: the objectives being so different, so is the terminology. Science seeks for, even if it does not always find, clear and unambiguous terms, with the minimum of overlap and ambiguity, the minimum of context-dependence and hearer-dependence. Such are the sources of its precision. (Much of SP is quite a long way away from this; but over and over again you find scientists deploring the fact, or putting terms like 'memory', 'intelligence' inside protective scare-quotes to indicate that they are unhappy with the looseness or inconsistency of their colleagues' uses or misuses.)4 CSP, by contrast, revels in ambiguity, vagueness, the lack of sharp definition; in terms that are almost but not quite synonymous, in contextdependence, audience-dependence; it exploits tone of voice, the stress given to words or phrases, accompanying facial or gestural expressions; and all of these are the sources of its incredible accuracy and fine focus. To overstate matters a little, definitions in CSP are generally either trivially boring, or unwisely restrictive; quite the reverse of SP. Put another way, a science that needed a Rogetstyle Thesaurus is a science with a long way to go; whereas the fuller and richer the CS Thesaurus, the fuller and richer the CS framework.5 Fourth: I have suggested already that CSP does not, unlike SP, resort to 'laws', or only rarely; and such 'laws' as it uses are generally proverbial ('out of sight, out of mind' / 'absence makes the heart grow fonder') and/or are hedged around with so many ceteris paribus, or ceteris absentibus, conditions that they can do little work. There is much generality, certainly, in CSP; but the explanation of that generality lies in our mastery of the language, our 'know-how' ability to use terms like 'generous', 'responsible', 'intelligent', 'mean', 'open-minded', 'canny', 'ambitious', 'courageous', 'sophisticated', 'a jerk', 'a nerd' - in context, to a particular hearer and 4
As Newton-Smith (1981) pointed out, the terms 'force' and 'spacetime' were vigorously attacked for obscurity and vagueness when they were first introduced. (These terms did, though, make possible just those experiments and hypotheses by means of which they were refined and more tightly characterised.) 5 Evidently this difference between SP and CSP will depend upon the area(s) of SP in question. The terminology of social psychology, for instance, has much greater overlap with CSP than does physiological psychology. Whether this is in part due to the relative immaturity of the social sciences, or something else, is a question I will not discuss here. 182
Talking to Cats, Rats and Bats about a specific individual. There is little 'know that' about ascribing a term such as 'responsibility'. Fifth: SP must transcend societies and cultures. Theories developed in Tokyo must be open to examination and test in Toronto; SP terminology must be common, or readily accessible, to psychologists everywhere. The terms and assumptions of CSP, by contrast, are culturally variable. Needham (1972) surveyed many societies and examined a range of states that seemed more or less like our 'belief, and concluded that often there was nothing that matched our notion; nor did those societies have equivalents to the related notions of stating, affirming, defending, accepting, etc. I have tried to show (1988) how the terms 'mind' and 'consciousness' have little echo in some of the CSPs of other times, and/or of other cultures. Different CSP networks, then, each internally consistent enough, and suited well enough to the society in question; and no reason to assume that one must be 'closer to the truth', or 'better', than that of another culture. Sixth: as we have seen in the previous section, CSP is as much a set of strategies as it is a set of assertions - even in the linguistically adept human animal. These six will do, I think. CSP is 'bound to be baroque, improvisational, reconstructional, and heavily ceteris paribus' (Bogdan 1991, p. 176). The distinction I have drawn so crudely - while repeatedly insisting that it is a distinction 'of degree' - has at least two important implications. (1)A term in CSP may or may not have the same extension as a term in SP. This goes for CS and science more generally. Consider: (a) Sometimes we may talk of things that are in fact natural kinds,6 and then our terms have extensions very close to those of science: 'gold', 'elephant', 'water'. 'Pain' might be an example from CSP. (b) But then also there is often partial overlap: in CS generally, 'wolf, 'lily' or 'tree' have no simple relation to any scientific counterpart (for instance, science recognises two kinds of wolf). So it is with CSP and SP. Consider the term 'intelligence', with which we began - using it at the start of this paper with its wide CS range. Examination of writings in SP discloses some scientists deciding in frustration not to use the term at all; others defending and trying to pin down a notion of 'general intelligence'; others insisting that at 6 To reassure those suspicious of the notion of 'natural kinds': I am not presupposing any full-blown doctrine'of them here. By 'a natural kind term' I need no more than 'term which is thought to pick out a systematically fruitful explanandum'. 183
K. V. Wilkes best we can talk of problem-, niche- and species- intelligences, or saying such things as '[MJarmoset intelligence...is whatever marmosets do', and 'judgments regarding intelligence might best be viewed as folk taxonomy rather than scientific taxonomy' (Menzel and Juno 1985, pp. 155, 157). Other words, common to both CSP and SP, show the same feature - of scientists trying to recast, subdivide, refine the notions to make them amenable to scientific study: cf. 'memory', 'learning', 'phobia', 'representation'. (c) Then there is the lack in SP of terms like 'integrity', 'responsible', 'weird', 'obstinate' or 'prejudiced' (etc.), and the lack in CSP of such locutions as 'comparator', 'metamemory', 'anosagnosia', 'covariant assessment' (etc., again). It is just the same in CS and science generally; consider the CS term 'fence' - this has no echo in physics - and the scientists' term 'boson' - this is scarcely everyday linguistic currency. There is much here, quite apart from the understanding of the term 'intelligence' itself, that is of particular relevance to our present purposes. For example, the extent of the overlap - if there is one between the pairs 'belief'/'information', and 'thinking'/'informationprocessing'. Then there is the overlap in use and scope between terms used in both CSP and SP, such as 'theory', 'memory', 'learning' or 'consciousness'. We cannot take a strong similarity for granted. (2) A second implication. If, for the purposes of argument, the difference of degree between CSP and SP which I have outlined is broadly accepted, then there are no less than three-and-a-half (associated and overlapping) endeavours. The first is the study of the CSP framework. That, I take it, is what traditionally goes by the name 'philosophy of mind'. It need not detain us; it is legitimate and worthwhile. Unfortunately, however, it does have a bastard offspring, the 'half of my 'three-and-ahalf, which is alarmingly fashionable. I shall call it 'philosophical psychology'.7 This should have been aborted; it is, crudely, the philosophers' attempt to force the flux of CSP into the wire grid of a 'wannabe' scientific theory, or to treat it as if it were one. Because of its undoubted trendiness, I shall need to return to pour further scorn on it later. The second legitimate endeavour, though, is the study of SP. That is one part of the philosophy of science - the 'philosophy of psychology', or better, 'the philosophy of the brain and behavioural sciences'. I have no quibbles with this excellent enterprise. 7 Nothing hinges on this label, chosen ad hoc for convenience. Many of those who describe what they do as 'philosophical psychology' are not necessarily attacked.
184
Talking to Cats, Rats and Bats Then there is the third; the most important for my present purposes, since it concerns the study of CSP by SP: the attempt to discover what the cognitive pattern of the normal (and abnormal) human is like, and how he comes to have it, or to lack parts of it; and how the non-human animal comes to think (if it does) to the extent that it does. This enterprise gives rise to such discussions as are beautifully illustrated in the notorious Premack and Woodruff paper 'Does the Chimpanzee Have a Theory of Mind?' (Premack and Woodruff 1978) and the commentary it provoked and still provokes. This article introduced ideas which can be and have been extended to ask the same of autistic children, the mentally retarded, idiots savants, etc. It is no wonder that this third endeavour has been so interdisciplinary and vexed. Straddling the distinction between CSP and SP, it can lure us into what I have called 'philosophical psychology': to conflate the theory of part of SP with what is studied, the alleged 'theory' that we are said to use in CSP; so also to conflate the CSP and SP deployments of such notions as 'memory', 'learning', 'intelligence', 'motivation'; so also to conflate the beliefs we examine with the 'information', 'representation', 'metarepresentation' or 'protorepresentation' typically discussed by the scientists. These conflations may sometimes be justified and defensible, as I have insisted above when showing that there are terms with extensions that do indeed coincide almost exactly between CS and science; very often, though, they are not. To illustrate the importance of the possible confusions or conflations I have in mind, consider that it is in terms of the SP notions - however rough at present their definitions - of information acquisition, processing, storage and retrieval that we hope to explain how the man in the street comes by his CS beliefs. And, incidentally, how Searle's dog Ludwig catches his ball (see Searle, this volume). The familiar response to many attempts to explain CSP phenomena in terms of SP theory we all know, particularly when consciousness is at issue: the scientist has 'left something out'. (Just as the physical scientist, in his theories, has left out examination and explanation of most of the physical-object terms to be found in a medium-sized dictionary: 'fence', 'sunshade'?) The knee-jerk reaction of philosophers confronted with fluffy, hard-to-define, CSP phenomena like 'belief, 'understanding', 'intelligence' or 'consciousness' is (implicitly) to reify them, (implicitly) to adopt a naively simple causal theory of reference — and then to claim that 'the referent' has not been 'found'. This is all the more easily done when we think of CSP as philosophical psychology does; as constituted by sets of specific emotions and propositional attitudes. There 185
K. V. Wilkes is far more to it than the oversimplified and impoverished picture of philosophical psychology can admit, though.8 Although it is tempting at this point to jeer at the use of 'theory' in CS (as in: 'the folk theory of mind') - to ask where the axioms, the principles, the laws, the articulation of CSP 'theory' may be found, to mock talk of 'theories' that denote 'all sorts of inchoate, implicit, or imputed beliefs' (Hacking 1983, p. 175), as 'theory-shmeory' - I shall resist the temptation. It remains a fact that whatever I say, people will continue to talk of theories of mind in CSP. Nonetheless the differences (of degree, remember) between CSP and SP theories are there, and for my present purposes essential. In particular, the question whether the framework of CSP should be construed realistically is crucial, and demands specific attention. Ill CSP, whether we think of human or non-human animals, is, I have claimed, a set of strategies as much as, if not more than, it is a set of assertions. If for this and other reasons CSP does not constitute a 'theory' in the sense that scientists in laboratories and white coats would give the term, then 'it' is not something of which we can always ask sensibly 'realist or instrumentalist?' It will prove to be a right rich royal mixture of the true and the false, of terms that refer and of those that do not. So it is with all CS. 'The sun is rising.' 'I believe it with all my heart.' 'The car refuses to start.' 'I am in two minds.' These CS locutions are all, I take it, false. But: 'My foot hurts.' 'The butter weighs 250 g.' 'He died of E.coli 0157.' 'I have just remembered.' These could all be unequivocally true in some contexts. Then there are locutions which we do not - or rather, I would prefer to say, need not - classify as true or false, referring or non-referring: 'The dog thinks that the cat is up the tree.' 'The plant seeks the sun.' 'Macbeth saw Banquo at the feast.' 'The cat is pretending to play with a mouse.' 'The computer is trying to capture my knight.' What is common to all the CSP phrases above is that they are all useful; they get their message across. To lump a diverse set of phenomena under a single heading, as we do with umbrella-like terms 8
Hence the Churchlands (see, e.g., P. M. Churchland 1979, and P. S. Churchland 1986 - but it is a note that sounds throughout their works) are wrong to indict CSP as a poor, feeble, wrong, shallow, stagnating, theory. If it were the starving skeleton offered by belief-desire theorists, the philosophical psychologists, and if it did aim to meet the conditions of a science, then their accusations could not be avoided. But it isn't. 186
Talking to Cats, Rats and Bats such as 'intending', 'thinking', 'understanding' (etc.), doesn't matter most of the time, and is more convenient for standard communicative purposes. Usually the locutions we might call 'false' or 'neither clearly true nor false' are briefer than any literally true expression; they are almost always more informative to the non-expert; frequently, too, we do not know how to redescribe some phenomena (the plant seeking the sun, being struck by a thought) in terms of others (chlorophyll, photosynthesis, synaptic discharges in neural nets). That is validation enough for their use in CS, and CSP. Am I saying 'if it ain't broke, don't fix it'? Not quite. I am saying that in CS we generally get by with this - it suits our diverse purposes - very well. CSP is brilliant in its own domain, with a few exceptions that mainly concern the irrational and non-rational.' SP, however, must reject such a 'getting by beautifully' attitude because part of its remit is to discover why some of the things that work, do work. CSP, by and large, gets more excited by things that attract its fickle attention: things that don't work, or that are unexpected, novel, surprising; what routinely and regularly happens rarely attracts CS interest. Science wonders why and how the plant seeks the sun (or, if you prefer - it matters not at all - acts as if it sought the sun). It tries to explain what happens in habituation. It asks how Ludwig can catch his ball. It may ask why referring, pointing and showing are rarely seen in non-human animals. (This, incidentally, is a very interesting question; see Gomez et al. 1993.) Such kinds of research may of course feed back into the educated layman's language - perhaps he now describes a murderer as a paranoid schizophrenic, rather than as evil. But even enlightenment may not wholly displace what it corrects: even the astronomer may complain 'the sun rose half an hour ago; get up!' In particular, scientific discoveries will probably never replace or displace CSP descriptions and explanations if they are too complicated, too technical, too longwinded — as they would inevitably be in the example of the ballcatching abilities of the dog Ludwig. CSP is a framework which we come to acquire, and its terminology is shot through and through with bits of science and pseudoscience, or with metaphors that may trace back to former scientific theories (consider 'Oedipal', 'ill-starred', 'psychotic', 'a gut feeling'); all this over time, and as a part-product of massive social learning and conditioning. Humans become adept at handling a certain picture or framework, and almost all use language to help this 9
Why CSP is so foxed by such everyday phenomena as akrasia, selfdeception, dreaming, and by most kinds of mental abnormality, is a most intriguing question. It can be answered; but to provide one would require a separate treatment. 187
K. V. Wilkes mastery; but the fact that we let our language be guided by this picture, and vice versa, says little about how much we believe true about it, or how much of it we take to refer. We may, of course (and the philosophical psychologists do) get misled into reification; of thoughts, beliefs, sense-data, desires, consciousness.10 Thus to emphasise or assume literal truth and secure reference in CSP is a treacherous and unnecessary path to follow." But on the other hand how, and to what extent, we acquire such a framework where 'we' includes abnormal humans and non-human animals is an important, fruitful and interesting question for SP. And then again: leaving CSP concepts alone and studying how they fit together is a fascinating enterprise for the philosophy of mind. There is no room for anything else in the middle: CSP must be taken as it is. Thus: 'Do non-human animals have beliefs?' 'Could a computer be conscious?' 'Does Searle's Chinese Room understand Chinese?' In CSP, maybe they do, maybe they don't; the question is whether it suits our multifarious purposes to speak as if they do. It is misleading to call this 'instrumentalism', given the vexed and technical examinations of realism and instrumentalism in discussions of the sciences, but it has this much of instrumentalism about it: the apparatus is a useful tool for getting around, most of the time, and its truth or falsity may not be the most interesting feature of it. In CSP, 'believe', 'conscious' and 'understand' get their utility - I have claimed - in part because we resist, or are uninterested in, attempts to pin them down. There may, or may not, be a unitary referendum for 'consciousness', a thing which is 'being conscious'; or it may rather be a term like 'the economy' (an analogy which I heard Dennett use on the radio recently, when he was asked whether we could say where or what consciousness is). Or consider 'fascism'. This line of argument has all the advantages of theft over honest toil. But if, where and when the honest toil is, as I believe, honestly misplaced, the thief deserves his prize. Now it is time to turn to non-human animals. 10
And what beliefs they need! 'Philosophers' belief includes...the belief that if I drink this cup of coffee I will not turn into a dinosaur...' The same goes for desires. See Morton (1991). " Of course CS assertions, even if they use expressions whose literal truth or reference we doubt, may yet be false even by CS standards: 'the plant seeks the sun' - no, that plant just doesn't; 'he's conscious now!' - no, that was a reflex spasm; 'he wants to buy some cigarettes' - no, he's gone into the shop to ask directions. 188
Talking to Cats, Rats and Bats IV And, of course, to turn to abnormal people; those who are mentally retarded or brain-damaged, or who suffer from Down's syndrome, autism, neglect phenomena and the like. Time, that is, to turn to the theory of mind (where 'theory' is subject to the qualifications mentioned above: rather a 'framework', or a 'theory-shmeory' when compared with scientific theories) which normal people enjoy, and which is absent, or only partially present, in other organisms. What I shall not argue, because I think it is very widely accepted these days, is that any framework-theory of mind presupposes that the entity to which it is ascribed is a social animal. The demands of social life are major sources of pressure for cognitive advance (see, e.g., Humphrey 1983). Which comes first is a non-question, a chicken-and-egg question (although many, I admit, seem to find it fascinating). To thrive in a social ambience of any degree of complexity requires a way of coping and communicating with your peers, whether they be allies or enemies; and to do that requires acting and reacting towards them in certain ways - whether 'hardwired' or 'learned' we can forget for the moment. Species of all kinds inhabit niches, ranging from the relatively secure and unchanging to the constantly changing and challenging. Food sources may be hard or easy to find; dangers scarce or omnipresent. If a niche is changed dramatically, as for Washoe or Kanzi, laboratory rats and pigeons, then strategies for survival can be expected to alter too, at least among the most successful. If a niche, or more generally an environment, changes too fast and dramatically, then whole species may become extinct. Pressures of conspecifics and of the environmental niche shape strategies for success and survival as much among humans as among non-humans. The problem-'space' invites or encourages certain problem solutions, which can take the form of developing wings, prehensile tails, camouflage, sharper teeth; running faster, moving towards an hierarchical structure in the group, which involves the recognition of siblings, parents, friends and associates; ritualised gestures, learned gestures, expressions or attitudes for submission, threat, warning, surprise; cries and songs, co-operation and competition, deceiving, teaching, protecting the young, making and using tools. Communication skills. And, of course, with some animals there is language. Success in a niche can be more or less full. Many non-human animals, like dogs and cats, grow up in a human environment and come to chase sticks, play with balls of wool, fetch the slippers or the 189
K. V. Wilkes paper, herd sheep, come when called. But note that wolves, tigercubs, even foxhounds and sheepdogs generally fail as household pets; their inbuilt tendencies to hunt and herd sit poorly with walks on a lead through the local park. Wild rats will give up and die very soon when translated to a laboratory and dropped into a tankful of circulating water; they dive to explore the tank fast and thoroughly, find no way of escape, come to the surface, and die within seconds. Seasoned laboratory rats swim stolidly around for hours: 'oh no, another one of those damn experiments; I expect I'll be fished out sooner or later'. The laboratory rat has adapted to its part-natural, part-artificial, environment.12 The social niche is particularly difficult for autistic people. They can, or at least the 'high-performing' ones13 can, learn to interact with other people in ways that pass most superficial surveys; but it is a difficult and often painful process, and their lives remain substantially maimed in many respects. Sometimes, for example, they confuse the pronouns 'I' and 'you', which hints that much of the social world may be well-nigh incomprehensible to them. The 'personal' world seems to be seen impersonally: not ordered and illuminated by CSP.14 Success or survival, whether with conspecifics or with those of another species (as sheepdogs with their masters, rats with their experimenting scientists) may thus take many forms, and reveal a degree of cognitive sophistication that is perhaps unexpected. Rats, for example, can be trained to discriminate their own acts, to indicate - by pressing one of a number of levers - what they had just done (see Beninger et al. 1974); an ability to monitor one's own actions is clearly something that has substantial consequences for success. With highly socialised species we find recognition of dominancerelations, rank orderings, sometimes matrilineal kinship; we often find experiments that require the subject to know, to anticipate or to guess what another is likely to do, or to chase; or what it thinks, sees, is feeling or wants. Chimps are well aware that another chimp wants the food it has cornered, and that it will try to get it; several species react to injured members by strategies of protection, grooming and 12
One of the many interesting features of this experiment is that the wild rats, having been once rescued just before death, will on a second immersion swim around for even longer than the laboratory rats; they learn fast. 13 Between 75 per cent and 80 per cent of autistic individuals are mentally subnormal. 14 Interestingly enough, we could say that the autistic 'theory of mind' is much more properly a theory than is that of the normal human. Autistic people need often to work out fairly explicitly for themselves what other people may be doing, wanting or thinking. 190
Talking to Cats, Rats and Bats aiding. Deceitful behaviour between apes is observed frequently both in the wild and under controlled laboratory conditions; we even find (in controlled experiments) one chimp's recognition of another chimp's attempt to deceive (see Whiten and Byrne 1988). To describe all this necessarily, unavoidably, requires talk of what animals want, think, expect, hope, recognise. With animals as with humans we may often find no important difference between saying that an animal 'sees', or that it 'knows', something, on the basis of which it acts; whether it 'sees the Fa' or 'sees that (= 'knows') this a is F. The recurrent question 'but is it really thought}' is, or should be, a non-starter. If you cannot be said to believe without having the concept of belief, and cannot have that without grasp of the notions of true and false beliefs, then of course most non-human animals, and children under the age of three, don't have beliefs. But why should we want to accept either premise - why should SP be tied to the explanation of this notion of belief, that put forward by the philosophical psychologists? 'Belief in CSP is a more slippery term than this suggests. Putting, in these cases, 'belief into scare-quotes, or calling it 'proto-belief, or a 'subdoxastic state' (as philosophical psychology does) may soothe the philosophical itches of those yearning for an armchair, but ^wasf-scientific, theory for CSP: disciplining it into nice niches. But I have suggested that we need and want no such armchair enterprise.
As things stand now, in a fast-moving and still terminologicallyfluid SP, notions of 'information' (acquisition, processing, hierarchical processing, storage, retrieval), 'representation' and their ilk purport to explain (inter, of course, much alia) the extent to which organisms can be described in CSP as having thoughts, beliefs, feelings, wants, plans, memories, consciousness; as learning, trying, imagining, playing, pretending. But as we have seen, SP is not committed to explain everything that may be apparently reified in CSP. To recap: the reification, or the appearance of reification given by the use of a noun ('memory', 'consciousness'), however understandable on grounds of parsimony or ease of use, may not in fact pick out a natural kind explanandum with which SP should concern itself. That CSP very often provides accurate and precise descriptions, explanations, and predictions of behaviour is not in doubt. That it does so because its terms refer to real phenomena amenable to systematic study per se is however highly dubitable: sometimes they will, sometimes they won't. Yet nothing need be 'left out', any 191
K. V. Wilkes more than physics leaves out analysis of tables per se, ornaments per se\ or, closer to home, why the physiologist has failed to understand breathing as such after he has described the techniques that allow animals to breathe via lungs, the skin, the gut and the rectum. It is no accident that much of this debate has centred around second-order beliefs, 'metarepresentation', thoughts about thoughts; this is because what is required for notions of the self, for the 'what it is like to be' questions, for the 'I/you' distinction, for our attitudes to others, for moral responsibility in general - hinges on this. The use of language, some kind of consciousness, seem to be (and are) irresistibly bound up in thinking about thinking. With normal human adults, language-use is usually a decisive indicator for ascriptions of thoughts about thoughts, beliefs about self and others, beliefs about others' intentions. When we lack language as a test, or when we doubt whether chimpanzee achievements really count as language, or if we speculate that animal cries may form a 'protolanguage', standing to language as tadpoles stand to frogs or caterpillars to butterflies...still we can reasonably wonder whether command of language is a precondition for the abilities, capacities, and opportunities that distinguish the human animal from others. This is a perfectly legitimate research area, and the water is hot. For every experiment designed to show that non-human animals have the capabilities that presuppose thoughts about thoughts, there are debunking experiments and arguments claiming that all documented behaviours can be explained more parsimoniously by simpler hypotheses such as paired-associate learning. ('Simpler' in the sense of being theoretically more economical; some paired-associate learning is formidably difficult, for humans as for animals.) It is hard science, both on the theoretical and on the experimental levels. Difficult as second-level ascriptions may prove to be, though, the central fact remains: the way CSP describes the behaviours, capacities, and strategies of human and non-human animals cannot be abandoned. It can be accused of being misleading and imprecise if we follow the philosophical psychologists and expect of it the systematic grid of a scientific theory; but that route leads nowhere. CSP as it is is precisely what SP is trying to understand and explain. Hence, since we do indeed ascribe communicative and cognitive abilities to non-human animals, SP has the task of showing what it is about the animal that explains why these ascriptions are natural and appropriate; by citing, for example, parallel and serial hierarchies and interactions of information-processing mechanisms in the brain. Nothing need be 'left out'. Whether it is beliefs, or consciousness, or intentionality ('genuine' intentionality is genuinely a creation of philosophical psychology; what could be more unprob192
Talking to Cats, Rats and Bats lematically intentional than the very notion of 'information'?), the way is open for psychology and neurophysiology to explore all of it. VI Finally, what of the objection that beliefs and desires cause action, that explanation is causal, and so that any theory of behaviour whether in SP or not - must take them on board per sel This has become a rather easy objection to refute. First, an intuition owing most to Davidson: even if CSP says or implies that A causes B, and even if causality is nomological, the laws underlying the event may not link A qua A with B qua B (Davidson 1970, pp. 79-101). 'A' may well drop completely out of the picture; its role in a causal explanation may require a radical recasting of what lies beneath. If 'A' is a CS belief, for instance, such causality as there may be between it and some B (itself perhaps recast) would lie with indefinitely complex sets and series of levels of information-processing transactions, all swept up under the generous umbrella of 'belief, or 'thought'. Second, who said that explaining had to be a causal matter anyway, even in science? Very many of the simple examples used to introduce beginners to D-N explanation are not causal at all. ('Why is this bird black?' 'This bird is a raven; all ravens are black'.) Whatever we are to say about SP, where explanations may usually although not always - be causal ones, in CSP I would guess that the bulk of what counts as satisfactory explanation is certainly not overtly causal. There are profound theoretical difficulties of the sort highlighted by Libet in proposing conscious mental events as causes either in CSP or SP (see, e.g., Libet 1985; and for one of the many discussions, see Velmans 1991). But more generally than this: to someone explaining behaviour to another, any bit of the mental and environmental surround may be picked out as salient, informationally pregnant: 'because it's raining', 'because his bicycle has broken', 'because he's absent-minded', 'because he's upset'. 'It's a long story: here's what happened...' Much depends on what the questioner wants to know. In sum, explanation may have a logical form, but it has pragmatic matter. Sometimes, indeed, we cite a belief to explain; then we 'isolate and lift the belief thatp out of the many data-structures among which it is implicatively buried in different brains busy doing different things in different environments' (Bogdan 1991, p. 179); this implicatively buried belief is scarcely 'a cause'. Scientific terminology and hypotheses criss-cross the taxonomies of CSP; but there is no fight between them. 193
K. V. Wilkes Another result from the discussion above: causality cannot be brought in to aid armchair belief-desire theorising, 'philosophical psychology'; and, worse: this - with its 'wannabe' pseudo-science does have a battle on its hands, which it is bound to lose, against the superior enterprise of SP. Thus cats, rats and bats can think, have desires and feelings, can communicate. So can chimps, dolphins, the autistic. So can we. What typical members of each category can and cannot do in what circumstances is a matter for empirical research: designing or discovering problem-spaces, observing what strategies are and are not deployed to find solutions, the extent of novelty, flexibility, co-operation. Different degrees and depths of processing can be hypothesised; these might well prove to explain what in CSP is captured by talk of thoughts about thoughts, consciousness, intelligence. No room, however, for talk of identity, parallelism, dualism, emergentism; the 'reference' of CSP terms, such as it is, is too fluffy. That's the way that things are, when we try to line up CS with science; so much for one large part of 'the mind-body problem'. We can sometimes communicate with cats, rats and bats; and they with each other, and sometimes with us. With all animals, we can leave talking to the chattering classes. References Beninger, R. J., Kendall, S. B. and Vanderwolf, C. H. 1974. 'The Ability of Rats to Discriminate Their Own Behaviours', Canadian Journal of Psychology 28, 79-91 Bogdan, R. J. 1991. 'Common Sense Naturalized: the Practical Stance', in Bogdan (ed.) (1991), Mind and Common Sense. Cambridge University Press Christensen, S. M. and Turner, D. R. (eds.) (1993). Folk Psychology and the Philosophy of Mind. Hillsdale, NJ: Lawrence Erlbaum Churchland, P. M. 1979. Scientific Realism and the Plasticity of Mind. Cambridge University Press Churchland, P. S. 1986. Neurophilosophy. Cambridge, MA: MIT Press Davidson, D. 1970. 'Mental Events', in L. Foster and J. W. Swanson (eds.), Experience and Theory. Amherst: University of Massachusetts Press Dennett, D. C. 1981. 'Making Sense of Ourselves', Philosophical Topics 12. Reprinted in Dennett, D. C. 1987. The Intentional Stance. Cambridge, MA: MIT Press Gomez, J. C, Sarria, E. and Tamarit, T. 1993. 'The Comparative Study of Early Communication and Theories of Mind: Ontogeny, Phylogeny, and Pathology', in S. Baron-Cohen, H. Tager-Flusberg, and D. J. Cohen (eds.), Understanding Other Minds. Oxford University Press 194
Talking to Cats, Rats and Bats Hacking. I. 1983. Representing and Intervening: Introductory Topics in the Philosophy of Natural Science. Cambridge University Press Humphrey, N. 1983. Consciousness Regained. Oxford University Press Libet, B. 1985. 'Unconscious Cerebral Initiative and the Role of Conscious Will in Voluntary Action', The Behavioral and Brain Sciences 8,529-66 Menzel Jr., E. W. and Juno, C. 1985. 'Social Foraging in Marmoset Monkeys and the Question of Intelligence', in L. Weiskrantz (ed.), Animal Intelligence. Oxford: Clarendon Press Morton, A. 1991. 'The Inevitability of Folk Psychology', in Bogdan (ed.) (1991) Needham, R. 1972. Belief, Language and Experience. University of Chicago Press Newton-Smith, W. H. 1981. The Rationality of Science. London: Routledge and Kegan Paul Premack, D., and Woodruff, G. 1978. 'Does the Chimpanzee Have a Theory of Mind?' The Behavioral and Brain Sciences 4, 515-26 Velmans, M. 1991. 'Is Human Information-Processing Conscious?' The Behavioral and Brain Sciences 14, 651-69 Whiten, A. and Byrne, R. W. 1988. 'Tactical Deception in Primates', The Behavioral and Brain Sciences 11, 233-73 Wilkes, K. V. 1988. ' , yishi, duh, um, and Consciousness', in A. J. Marcel and E. Bisiach (eds.), Consciousness in Contemporary Science. Oxford: Clarendon Press 1991. 'The Long Past and the Short History', in Bogdan (ed.) (1991) 1993. 'The Relationship Between Scientific Psychology and Common Sense Psychology', in Christensen and Turner (eds.) (1993) Wundt, W. M. 1900-9. Volkerpsychologie: Eine Untersuchung der Entwicklungsgesetze von Sprache, Mythus und Sitte, 2 vols. Leipzig: Engelmann
195
Analyticity, Linguistic Rules and Epistemic Evaluation CHRISTOPHER HOOKWAY Two Ways to Think about Thought We can characterise thought in two different ways. Which is preferred can have implications for important issues about reasoning and the norms that govern cognition. The first, which owes much to the picture of the mind encountered in Descartes' Meditations, observes that paradigmatic examples of thoughts and inferences are events and processes whose special characteristics stem from their being 'mental' occurrences. For example they are conscious or, if unconscious, they stand in some special relation to thought processes that are conscious. They typically involve attitudes towards contents or propositions. In general, thoughts have a distinctive ontological status and this status depends upon their being mental and typically conscious. The second emphasises that thought is a kind of activity with a definite function. It involves the use of intelligence to solve problems, answer questions, make plans and so on. Thought should be studied as a kind of goal-directed activity. Those interested in the norms that govern thought should attend to the role of responsible disciplined reflection in carrying out this activity. The first ('Cartesian') view is typical of much recent philosophy of mind. It assumes that an examination of the varieties of mental representations provides us with a vocabulary for describing and explaining thought processes. The second ('problem-solving') approach is more common in epistemology and the philosophy of science. Popperians, for example, emphasise the role of public critical discussion in the advance of knowledge and refuse to place those processes of reasoning that are 'internal' to a single consciousness at the centre of their epistemologies; and others may stress the role of public activities such as experimentation in the process of thought. The pragmatist approach to cognitive norms found in John Dewey's Logic also reflects this approach. The two ways of thinking about thought encourage differences of emphasis and approach in examining some important issues. One of these issues concerns the relations of language and thought. When wondering about the role of language in thought, 'Cartesians' may consider whether there are any kinds of thoughts 197
Christopher Hookway (mental states) which are available only to language-users or whose possession can only be manifested in linguistic behaviour. Or they might ask how far a thought's being conscious depends upon its being available for linguistic expression. The alternative picture invites us to examine ways in which language helps us to solve problems which would otherwise be beyond our reach. Linguistic expressions and performances are viewed as means to effective problem-solving, as artefacts or tools which increase our reach or our efficiency. The testimony of other people and reference books extends the reach of the senses, making information available which we could not otherwise make use of. The understanding, too, is enlarged when we exploit others' greater mastery of abstract or technical concepts or their expertise in using statistical or other refined techniques. Thus Karl Popper emphasised two important functions of language (Popper 1972, ch. 3, especially pp. 119-20). Much of our knowledge is present in no individual thinker's mind, but is rather 'available' in reference books and data-bases. And the role of language in facilitating critical discussion is indispensable in sharpening views, eliminating prejudices and avoiding errors. Finally, there are many cases where examining linguistic representations of elements of our problems can be absolutely indispensable for solving them. For example, little progress was made in settling questions about the computability of deductive arguments until first-order logic had received an explicit public formulation which could be made the subject of rigorous meta-logical analysis. Cognitive Evaluations If thought is a goal-directed activity, aiming to solve problems and plan conduct, then it can be carried out more or less well: responsible disciplined thought involves reflection, self-criticism and the application of normative standards. We evaluate our formulations of our problems and our strategies for solving them. And, in the course of executing these strategies, we reflect upon the plausibility or coherence of proposed solutions and upon inferences that we employ on the way to a solution. Given what we have said about thought in general, this will involve raising questions and considering problems concerning our formulation of our problem, the adequacy of the answers we have considered and the acceptability of the standards we employ in evaluating those answers. Reflective thought involves second-order problem-solving, addressing problems raised by our first-order thinking. In that case, our rationality 198
Analyticity, Linguistic Rules and Epistemic Evaluation and intelligence will be manifested in the questions we raise, the strategies we adopt for answering them, the questions we consider about the acceptability of those strategies and also in the questions that we see no need to raise. A major concern of epistemology is the task of describing the normative standards we employ, understanding their role, and deciding whether they do indeed have the normative force we take them to possess. When we ask about the role of linguistic or semantic rules in guiding epistemic evaluations, we should focus on whether they have any role within this sort of evaluative practice. What sort of role might this be? Three related observations seem relevant here. First, reflective thought requires us to understand—and be clear about—the problems or questions that we are concerned with. Second, we need to identify whether a given proposition or hypothesis is a possible answer to our question. And third, we have to be able to tell whether a given answer is actually correct. Each of these calls for information which is involved in our understanding the questions and answers we are concerned with. Since semantic knowledge has an obvious role in understanding, it is plausible that semantic knowledge is involved in our ability to understand questions, to establish whether a given proposition answers the questions that concerns us, and to see how to tell whether an answer is correct. These are not obviously distinct. We might suppose that we do not understand a question until we know which propositions serve as possible answers to it; and that we do not understand a possible answer unless we know how to evaluate it. So a clear role for semantic knowledge in epistemic evaluation is becoming clear; it seems to have an identifiable role in the evaluations we need to make in the course of our problem-solving activities. Part of the motivation for the traditional analytic/synthetic distinction was the belief that the standards employed in these evaluations, if they are explicitly formulated, are propositional reflections of rules of language. These standards are grounded in facts of a semantic or linguistic nature. If that were so, then linguistic competence would be an indispensable necessary condition for responsible disciplined thought. Of course that distinction promised other insights as well: explaining the difference between a priori and a posteriori knowledge; accounting for the distinction between necessary and contingent truths; and (important for the first way of thinking about thought and language) yielding criteria of propositional identity and the identity of beliefs and thoughts. These further applications of the distinction are not relevant to this paper. My main concern is the relation between questions and their answers. What determines whether a particular proposition counts 199
Christopher Hookway as a possible answer to a question that concerns us? We naturally assume that this is settled clearly and immediately by the meaning of the sentence that expresses the question: a fairly straightforward account of this relationship falls out of the semantics of questions. An alternative view would hold that the relation is heavily constrained by facts about the background knowledge of the person who is asking the question or by the purposes of the discussion or conversation in which it is posed. Nobody would deny that these background or contextual matters rule out some answers as implausible or as already rejected, or that they influence how accurate or vague an answer should be. The suggestion put by this alternative view is stronger than that: what counts as a possible answer to the question at all is affected by these contextual and background matters. The later sections of this paper argue that once the problemsolving picture of thought is taken seriously, this alternative picture should be adopted. Semantics does not constrain the question/answer relation as it should if the notion of analyticity is to have a fundamental role in grounding epistemic norms. Before turning to this topic, I shall suggest that if we adopt the 'problemsolving' picture of thought, issues about analyticity are revealed to be rather more complex than they commonly appear. In order to illustrate this, I shall devote the next two sections to a sketch of the debate between Quine and Carnap about the analytic/synthetic distinction during the 1950s. Two Ways to Think about Analyticity As we noted above, the philosophical interest of the traditional analytic/synthetic distinction lay in its promise to provide non-mysterious explanations of a range of phenomena involving the difference between a priori and empirical knowledge, the difference in modal status between necessary and contingent truths and the epistemological and metaphysical differences between disciplines such as mathematics and the natural sciences. The distinction also offered to explain the surprising degree to which different people will agree upon the bearing of new evidence or arguments upon the acceptability of their beliefs. If these epistemic norms reflect shared meanings, we can explain the inter-subjective objective character of our patterns of acceptance. In this section I shall argue that adopting the problem-solving picture of thought offers a richer perspective upon the debates about the role of the analytic/synthetic distinction. 'Analyticity' is a theoretical notion which is internal to a variety of distinctive philosophical projects. It is common to assume that its 200
Analyticity, Linguistic Rules and Epistemic Evaluation role is to clarify and explain some distinctions that are present in our everyday common-sense ways of thinking about our beliefs and ideas. Those who use this distinction, we assume, are committed to the view that we all treat some propositions as analytic and others as synthetic. This then explains various features of our practice of thinking, features that are available from a 'Cartesian' perspective. If that is right, then the theoretical distinction is no more secure than the common-sense 'facts' that it is supposed to explain. As Grice and Strawson pointed out (1956), with the aid of a few slogans and some unproblematic examples,.most people can catch on to the distinction and will generally draw it in much the same place. This rests on 'obvious' examples of analytic truths: anyone who doubted whether vixens were female or questioned that all bachelors were unmarried would thereby reveal that she did not possess full understanding of 'vixen' or 'bachelor'. These common-sense facts do not settle the matter, since they are compatible with rejecting the philosophical interest of the analytic/synthetic distinction. Thus Quine himself happily endorses the everyday notion, applying it to propositions about the marital status of bachelors, the gender of vixens and the fundamental laws of logic. Since The Roots of Reference (1973, see also - for a particularly clear formulation - Quine 1994a, pp. 503-4), he has called sentences analytic if their truth is learned in the process of learning some of their constituent words. First, the philosophical distinction offers a distinctive theoretical explanation of these facts, whereas the common-sense one does not address the explanatory question. Second, the philosophical distinction applies this theoretical apparatus to phenomena which go beyond these everyday banalities, most notably to our understanding of the theoretical vocabularies used in the sciences. Quine argues that once we move beyond these familiar examples, we cannot find unproblematic analytic entailments between propositions and the patterns of experience which would warrant us in accepting or rejecting them. 'Moderate holism' entails that the bearing of evidence upon theory is always contextbound, depending upon the believer's views of related topics (Quine 1994, pp. 503—4). In that case, there is no reason to expect that the best explanation of our mastery of language will appeal to the battery of links between sentences and experience which constitute the meanings of the sentences in question. Once we adopt the problem-solving picture of thought, the viability of the analytic/synthetic distinction is not simply a matter of how we should explain distinctions that can be drawn in everyday experience. Defence of the distinction is compatible with a denial that it has much of a role in explaining facts about ordinary lan201
Christopher Hookway guage. We might argue that the lack of such facts in ordinary language was an impediment to effective problem-solving. Perhaps our thinking would be improved by introducing such a distinction, by employing linguistic tools for which it could be drawn. For Carnap, the analytic/synthetic distinction is drawn clearly only for constructional systems, proposed as rational reconstructions of areas of ordinary discourse. This means that, for certain purposes, they were seen as preferable to ordinary language, as artefacts that would better serve our cognitive goals. For example, it could be argued that co-operative inquiry about abstract theoretical or philosophical matters cannot be sustained by customary patterns of agreement in applying terms or evaluating inferences. We can only establish a shared practice of applying terms and evaluating inferences by undertaking to treat certain sentences as irrefutable unless an explicit decision is made to revise this decision. We should not ignore the possibility that the presence of an analytic/synthetic distinction may be one respect in which these constructional systems are preferable to ordinary language as a way of representing systems of knowledge and making explicit our practice of epistemic evaluation. In that case the fundamental issue would not concern whether the analytic/synthetic distinction is implicit in our ordinary practice. Instead we should focus on a question of practical rationality: should we construct linguistic tools of thought which incorporate such a distinction? Two questions can be asked about the analytic/synthetic distinction: (1) Is it necessary to use this distinction in order to explain facts about our ordinary practice of evaluating beliefs and ordering our inquiries? (2) Will it facilitate problem-solving if we develop linguistic artefacts for which an analytic/synthetic distinction is explicitly formulated? Adopting the problem-solving picture of thought helps us to see that these questions are independent. In particular, a negative answer to (1) does not entail a negative answer to (2). Conceivably, a negative answer to (1) could provide a premise in an argument for a positive answer to (2). Carnap and Quine: A Digression We can illustrate the merits of the problem-solving picture of thought by looking briefly at some features of the debates over ana202
Analyticity, Linguistic Rules and Epistemic Evaluation lyticity in the middle years of this century. Quine's challenge to the analytic/synthetic distinction was most clearly directed at the views defended by Carnap in books such as The Logical Syntax of Language (1937). Carnap's position is puzzling, for he defended a form of epistemological holism as strong as any defended in 'Two Dogmas of Empiricism' (Quine 1953, ch. 2).1 Rather than arguing that the analytic/synthetic distinction should be abandoned he simply claimed that, in certain circumstances, it is rational to respond to surprising experience by revising our analytic truths. Carnap endorsed the epistemology of 'Two Dogmas' without seeing this as an obstacle to retaining an analytic/synthetic distinction. However he was not simply blind to the difficulties. As Quine has recently noted, Carnap's diaries reveal that as early as 1933 he wondered whether the distinction was a matter of degree (see Quine 1994b, pp. 227-8). I shall suggest that this admission need not immediately threaten Carnap's philosophical views. As the previous section suggests, arguments for the analytic/synthetic distinction can be of two kinds. A 'theoretical-explanatory' justification would show that we need to appeal to the distinction in order to account for facts concerning our cognitive experience. Such a defence could be challenged by showing that the distinction is not implicit in familiar patterns of thought and inquiry, in which case there would be nothing for it to explain. If confirmation has an holistic character, or if the only distinction implicit in our cognitive habits were a vague one, then the prima facie case for such a challenge would be strong. If the facts which make the distinction initially appealing could be explained without appeal to rules of language or the concept of meaning, then there may be little work for the distinction to do. It is natural to read Quine's attack as taking this form: the traditional distinction has no real explanatory bite. A defence of the distinction grounded in practical rationality might point to the benefits to be obtained from reconstructing important areas of our knowledge in ways that explicitly employ the distinction. It is not unusual, when discussing a difficult issue, to lay down explicit definitions of key words, announcing that (for the time being) these will be taken as fixing the identity of the question 1
In a famous passage, Carnap wrote that 'No rule of the physical language is definitive; all rules are laid down with the reservation that they may be altered as soon as it seems expedient to do so. This applies not only to [rules which lay down basic physical principles] but also to the L-rules (those which correspond to analytic truths) including those of mathematics. In this respect there are only differences of degree; certain rules are more difficult to renounce than others' (1937, p. 318). 203
Christopher Hookway that we are discussing. Practical advantages attach to doing this; we can discipline our investigations and keep track of progress, even if we might subsequently adjust these 'definitions'. Formulating our intention to regard some proposition as fixed may improve the chances of our doing so, by adding to our motivation to resolve tensions in our beliefs in ways that respect this basic 'principle'. When trying to solve a problem co-operatively, we need to know that those on whose opinions we rely will resolve difficulties in ways of which we would approve. If the everyday distinction is vague, so that we apply concepts on the basis of a cluster of criteria not all of which may apply in a particular case, progress may be held up if others resolve clashes in ways that we do not anticipate. Hence laying down 'definitions' makes it clear that we understand our question in the same way and have a shared grasp of the relevance of new information to its answer. This can be a useful tool for problem-solving. We reduce the degree to which we trust to luck that others will react in the same way as ourselves by making our shared standards explicit. Since Carnap's distinction was explicitly drawn only for constructional systems, it is reasonable to expect his defence of it to be a practical one — even if the practical considerations that influenced him were not precisely those suggested here. And in that case, the fact that the distinction is not implicit in our everyday practice (like the fact that the distinction is vague) might add to the practical benefits to be obtained from regimentation. Showing that the philosophers' distinction is not implicit in our everyday practice is thus not sufficient for undermining it. Two kinds of supplementation may be employed. First, the critic should address the question of practical rationality and show that reconstructions of our knowledge which embody the distinction are not valuable tools for thought. Hence in the first chapter of Word and Object, Quine describes our epistemic norms as 'more easily felt than stated' and as employing shared instinctive standards of simplicity and inductive cogency which may well resist formulation as explicit principles (Quine 1960, p. 19; see Hookway 1994, sec. V). This could be used as a premise for an argument which suggests that the practical benefits of making our standards explicit may involve significant losses. They only obtain if it is possible to carry out inquiries under the guidance of standards of rationality which are fully explicit. Quine himself probably rejects this: he is certainly sceptical about the possibilities of constructing a formal system of inductive logic. Indeed, one might even be able to argue that the fact that there is no analytic/synthetic distinction implicit in our everyday cognitive practices would itself provide reason for suspecting that the benefits of introducing one would not be unalloyed. 204
Analyticity, Linguistic Rules and Epistemic Evaluation At least such arguments would not show that it is obviously rational to introduce the distinction across the board. A second line of criticism might suggest that the distinction supported by the practical argument is a poor relation of the traditional one. The latter was supposed to account for the distinctive modal status of analytic propositions and to explain their a priority. At best we get a strongly context-bound form of a priority which emphasises that 'analytic' propositions may readily be discarded if it is convenient to do so. It could be argued that such propositions can be thought of as 'necessary' or a priori in something like the traditional sense only if 'analyticity' can be shown to correspond to a term of a distinction that is already implicit (at least to a degree) in ordinary language. I sympathise with the force of this objection although it may not disturb a philosopher who is strongly unsympathetic to metaphysical distinctions. My concern here has not been to defend the distinction in general. Rather I have wanted to show that some features of the Quine-Carnap debate are placed in a somewhat different light once we adopt the problem-solving conception of thought and raise questions about the rationality of introducing explicit analytic truths as a tool for efficient problemsolving. Questions If the norms that govern reflective thought are designed to enable us to engage in problem-solving activities in a responsible, disciplined manner then, as we have seen, those norms will help to fix what is involved in understanding questions, recognising possible answers to them and identifying which answers should be accepted. Much work on the logic and semantics of questions adopts what David Harrah (1963, 1984) has called the 'sets of answers methodology'. Since understanding a question requires knowing what would count as an answer to it, it is plausible that the meaning of a non-ambiguous interrogative sentence determines which propositions would provide answers to it. If I am to respond to a question intended to elicit information that I possess, my understanding of the question must enable me to identify which piece of information it is intended to elicit. And if I am to use a question to identify the goal of an inquiry or process or deliberation, my understanding of the question must enable me to identify when the goal has been achieved. Since it is plausible that we search and access information stored in our memories or 'belief boxes' by seeking answers to questions that we ask ourselves, it again seems plausible that the content 205
Christopher Hookway of a question must somehow fix which belief-contents will serve as answers to it. If this question-answer relation is indeed a semantic matter, this seems a good starting point for an account of a role for semantic information in grounding cognitive or epistemic norms. My aim in the remainder of this paper is to cast doubt on the assumption that this is indeed a straightforwardly semantic matter. Since the question-answer relation is obviously important for cognition, this will, I hope, yield insight into the structure of the evaluations we use in problem-solving. Hence one might explain an eliciting use of a question as a pair composed of an instruction ('Tell me truly') and a set of propositions (the possible answers to the question). And when the question is used to set a goal for an inquiry, the same set of propositions may be paired with a different instruction. Thus we might introduce a question operator '?', which expresses a function from 'instructions' or 'requests' and sets of answers (or 'subjects') to questions. Study of the formal logic of questions would then break down into two tasks: first - the part which most easily lends itself to formal treatment - the analysis of how sets of answers are or can be represented; and second, the treatment of the variety of and representation of requests or instructions. This is close to the strategy adopted in the most influential recent work on the logic of questions, due to Nuel Belnap and Thomas Steel (1976). This approach to questions cannot be extended to all the problems and challenges posed by the use of interrogative sentences in deliberation and discussion. Its applicability seems most evident for cases of questions used to elicit information, although even there it is not wholly unproblematic. 2 Belnap and Steel consciously limit their logic of questions to cases where 'what counts as an answer to a given question is well defined in advance' (1976, p. 11). This is understandable because, although interests in the semantics and psychology of questions lie in the background of their work, their chief aim is to construct a system applicable in situations in which questioner and respondent are motivated to help each other and the respondent has access to a well-structured information source. Where the respondent is a machine and the information source is a 2
Moreover, most interest in the structure of questions in the writings of philosophers has been focused on questions which do not easily fit this pattern. A central philosophical motivation for work in erotetic logic has been the desire to make progress with the analysis of (scientific) explanation by exploiting the insight that explanations are typically answers to 'why' questions (see Lipton 1991, Bromberger 1992, Van Fraassen 1981). Bromberger is unusual in attaching a wider significance to the study of questions through the development of a theory of 'Rational Ignorance'. 206
Analyticity, Linguistic Rules and Epistemic Evaluation data-base, it seems clear that direct answers should take a predetermined canonical form. This restriction does not prevent the system providing a valuable ideal-type against which the vagaries of other contexts of questioning can be measured. I shall start by contrasting three sets of questions which meet this condition of clarity. The simplest, 'whether questions' present a finite set of alternative possible answers and these are explicitly presented in the question. These include simple 'yes-no' questions and questions such as: (1) Are you an undergraduate, a graduate student or a staff member? Approximating such clarity are 'which questions', such as: (2) Which even numbers are prime? (3) At which time (when) will you arrive home? (4) At which place (where) are you currently living? (5) Which students are attending your class this term? In these cases, the propositions providing possible answers are not explicitly given in the question. Instead, we are presented with a propositional function or matrix, the answers being propositions identifying items that satisfy the function or matrix. Thus answers to questions (2)-(5) will involve completions of 'x is prime', 'I will arrive home at x', 'I am currently living at x', 'x is attending my class this term'. Moreover, the questions impose category-restrictions on the objects whose satisfaction of the propositional function is at issue: they concern even numbers, times, places and students respectively. We could represent these questions thus: (2') (Which x such that x is an even number)(x is prime), (3') (Which x such that x is a time)(You will arrive home at x), (4') (Which x such that x is a place)(You live at x), (5') (Which x such that x is a student)(x is attending your course this term). Although someone could understand one of these questions while unable to formulate all of the possible answers, they would still know how to recognise whether a statement was a possible answer. More generally, a 'which' question will involve a matrix involving a number of free variables ('queriables' (Belnap and Steel); 'traces' (Bromberger)), and a function which assigns some or all of those variables to categories. Correct answers to the question will be true propositions which replace all of those variables by appropriate expressions referring to objects of the specified categories. Thus, the question 'Which boys are brothers of which girls?' requires us 207
Christopher Hookway to identify pairs of objects which satisfy the matrix 'x is a brother of y'. And, once again, 'category-restrictions' are imposed: it is stipulated that x is a boy and y is a girl. Two respects in which the questions we have considered are particularly clear or straightforward can now be explained. A question is sound if it has at least one correct answer; and those propositions which must be true if the question is to be sound are the presuppositions of the question. The existential presuppositions determine what sorts of objects must exist for the question to be sound, and the attributive presuppositions determine what properties must be instantiated for the questions to be sound. Borrowing an example from Bromberger, (6) What is the height of the Empire State Building? carries the existential presupposition that the Empire State Building exists and the attributive presupposition that it has a height. So long as these presuppositions are met, the question has a true answer and so it is sound. Moreover, reflection on (6) suggests that anybody who understands the question must acknowledge what is required for soundness, and this carries with it information relevant to establishing what we must do in order to discover what that answer is. In the case of 'which' questions (and 'whether' questions), we only need access to their logical forms and a grasp of the predicates they employ - we only need to understand them - to work out their existential and attributive presuppositions. As Bromberger puts it: 'In short, unassisted wits and cogitation on meaning alone will tell [someone] what he has to find out to know whether the question is sound, and thus to know whether it is accessible to him' (1992, p. 166). When we search a data-base for an answer to a question, the category-restrictions control the search by fixing the existential presuppositions of the soundness of the question, while the matrix fixes the attributive presuppositions. 3 Thus 'which' questions seem to provide the information that is required for searching for a correct answer. 3
Also clear (although not conforming to the analysis so far) are things like 'How many' questions. These have a different logical form from the questions we have considered. The set of possible answers is easily extracted from the form of the question, but the question is not inviting us to complete a matrix with a term picking out an object: it is, rather, looking for a numerical quantifier. It would be possible to assimilate them to those already considered through offering an artificial paraphrase: 'How may students are attending the lecture?' could be paraphrased as 'Which number numbers the class of students in the lecture?' I will not consider the logic of such questions. 208
Analyticity, Linguistic Rules and Epistemic Evaluation The most important distinction between 'whether' and 'which' questions is that the latter do not actually determine what counts as an answer to the question - and indeed reflection upon them shows that the question-answer relation is not fully fixed by logical or semantic considerations. An answer must take the form of a sentence or proposition, and the subjects of 'whether questions' enable us to identify appropriate sentences or propositions. The subject of a 'which' question specifies a character that answers must have, but it does not exclude lots of propositions which, even if true, would not be acceptable as answers. 'Understanding' a 'which' interrogative may provide us with only a necessary condition for a sentence or proposition functioning as an answer. The reason, of course, is that while the subject of the question determines that an adequate answer must complete an expression of the attribute with expressions referring to objects of appropriate categories, it does not determine what sort of referring expression should be employed, how the object should be picked out. This is hidden in example (2) because numbers have canonical names (numerals); but (example 5) students can be referred to by name, registration number, descriptions based on appearances and so on. An answer will be satisfying to the questioner only if the object is picked out appropriately. When we are concerned with provision of testimony, this is an issue for pragmatics. But it arises in the case of conscious deliberation too: when I use a question to extract information from my 'belief box', success or failure may depend upon searching for the right kind of representation of the object in question, upon raising the question in the right way. Limits to the Semantic Treatment: Problems and 'Why' Questions In sketching a 'semantic' treatment of the logic of questions and answers in the last section, I suggested that Belnap and Steel were primarily interested in its application to issues in artificial intelligence - to the role of requests in eliciting information from databases. I now want to suggest that, suggestive as it is as a kind of ideal-type, this model does not address the most important issues that arise when we consider the use of questions in discussion and deliberation. When we use questions to formulate problems and to question the credentials of conclusions, and when we use them to elicit information from other people or indeed from the information stored in our own 'belief boxes', our understanding is heavily overlaid with what we might think of as pragmatic considerations. In 209
Christopher Hookway this section, I will describe some questions which Belnap and Steel deliberately exclude from their analysis, and suggest that such questions have a fundamental role in deliberation and inquiry. Then, in the following section, I shall argue that even those questions that appear to fit the Belnap-Steel model share important features with these 'excluded' ones. Stressing their commitment to the 'sets of answers methodology' Belnap and Steel announce that 'the scope of our formalization is confined to the situation in which what counts as an answer to a given question is well-defined in advance' (1976, p. 11); 'expressly excluded from formalization are the problem-solving situation and the "Please relieve my vague puzzlement" situation'. This excludes a wide range of questions, such as: 'What is the relation between thought and language?'; 'What is a number?'; 'Why are there so many kinds of subatomic particles?'; 'How do children learn language?'; and 'What is a good account of the question-answer relation?'. Their exclusion does not involve a denial of the importance of interrogatives which articulate problems: they can be exceedingly important in formulating the targets of inquiries. However, in such cases we often do feel that the chief obstacle to progress is unclarity about what our question amounts to. A range of phenomena have given rise to puzzlement: our first task is to discover exactly what the questions are that are raised. The Belnap proposal is that once we move beyond the ambit of clear or elementary questions, the scope for (and benefits of) formal analysis and clarification is limited. I doubt that much can be said about the structure of 'problems'. Dewey may have such questions in mind when he claims that inquiry starts out from an 'indeterminacy' in our cognitive situation; our first task in dealing with such an indeterminacy is to constitute it as a question whose resolution, we hope, will eliminate the indeterminacy (Dewey 1938, pp. 101-19).4 Considerable epistemological importance attaches to the fact that we frequently have to get started on trying to solve a problem before we have a clear understanding of the question. There is, however, a potential difficulty here. For if 'problems' have a fundamental role in setting targets for inquiry and in expressing uncertainties about the progress of inquiry, and if we conclude that little can be said about the representation of problems 4
Dewey talks of responding to the indeterminacy by instituting a problem, not by instituting a question. We can treat this as a minor verbal difference. But not quite: for a vague sense of 'not knowing our way around' may tentatively be transformed into an 'unclear' problem which passes through a number of somewhat clearer 'problems' before we arrive at anything deserving to be called a question. 210
Analyticity, Linguistic Rules and Epistemic Evaluation or about their formal structure, we might be caused to have doubts about representational theories of cognition. We must hope that even if they lack the clear structure which makes it easy to provide a formal analysis of elementary questions, we can still say something systematic about the representation of problems. More interrogatives are excluded from our class of central cases than might at first appear. Belnap acknowledges that his analysis fails to say much about such 'deceptively simple questions' as 'Who is that man living next door?': Although this looks superficially like a definite information-seeking question, reflection leads us to see that it is very hard to predict in advance just what kind of answer would satisfy the questioner: does he want a proper name or, perhaps, a description? And if a description, is it one involving occupation, family connection, previous acquaintance, or what? Our thought is that he doesn't really know himself what he wants, but after your reply, he will tell you if your answer relieves his anxiety and, perhaps, ask a clearer question. (Belnap and Steel, 1976, p. 11). This example is important because it frees us from the temptation to view questions as essentially linguistic phenomena. A question is an abstract object: different interrogative sentences can express the same question and the same sentence can express different questions in different contexts. It is easy to imagine contexts in which this sentence will express a wholly clear question; it can be used to seek an identification of the person next door from a previously identified class. Thus 'understanding the question expressed by interrogative sentence S' is distinct from 'understanding the interrogative sentence S'. Consider some issues about 'why questions'. 'Why' interrogatives can pose 'problems', obscure questions and relatively clear questions. I might use 'Why was Jones late to the lecture today?' simply to elicit which from a limited range of possibilities was operative on this occasion: did he oversleep, miss the bus, and so on. 'Why' questions are also relatively clear when, for example, we are puzzled by some phenomenon that we take to be an exception to a general rule, and we want to know the cause of this exception to the established pattern. I may ask why kettles emit a humming noise just before the water begins to boil (Bromberger), expecting a simple physical explanation or confident that there is such an explanation, even if all the candidates I can think of are easily dismissed.5 Or I may express 5
It is a familiar point which I will not discuss that 'why-questions' are often contrastive: there is a difference between asking why kettles begin to hum then rather than after boiling and asking why kettles hum rather than (say) whistling the national anthem. 211
Christopher Hookway a general puzzlement by asking why whatever can be thought can be said with no clear sense of the sort of explanation that is in order or, indeed, if any is required at all. Since 'why' questions do not exhibit the quantificational structure of the clear questions we discussed above, we cannot establish that they are sound (that they have correct answers) by reference wholly to the existential and attributive presuppositions that are recoverable from a semantic or logical analysis of the question itself. And an understanding of the question need provide no clues about the direction our inquiries should take if we wish to answer it. At best we are rational to hope that we will recognise the correct answer when we encounter it. Our understanding of the question need provide us with no idea about where to start looking. We suggested above that 'problems' are commonly used to formulate or identify the goal or target of an inquiry. We would expect a formulation of the target of an inquiry or deliberation to have three properties. First, it must contain information enabling us to determine whether the inquiry has been completed successfully. Since inquiry is a goal-directed activity, responsible conduct of inquiry requires criteria of success. Second, it must provide information relevant to planning and executing the inquiry. My understanding of the problem should enable me to tell where to start. Third, it must be motivating: we must be able to understand why this problem (unlike other possible problems) provokes us into investigation: the problem must engage with our goals. While it is easy to see how inquiries directed at solving clear questions are guided by representations containing information relevant to establishing criteria of success for the inquiry and for planning how to arrive at an answer, it is less easy to see how problems can provide this information. The obscurity of problems makes it difficult to see what a clear representation of their content would be. As we saw, a problem may be experienced as a need to have answers to questions which cannot at present be clearly formulated: the first step towards responding to a problem appears to be a search for questions (or clearer formulations) which can be used to guide an inquiry which will, we hope, meet the demands of the problem. But if problems cannot receive a clear representation, it is hard to see how they serve as targets for inquiries and deliberations. When considering what his Rational Ignoramus should do when faced with 'why questions' which fail to provide such strategic guidance and clear criteria of success, Bromberger writes: My guess is that the rational thing for him to do is to forget about the why-question and to turn to other questions instead, remem212
Analyticity, Linguistic Rules and Epistemic Evaluation bering that answers to why-questions usually emerge from work on questions with more reliable credentials. That same guess probably looms behind the distrust of why-questions often felt by experimentalists... (1992, p. 169) A fortiori with problems: if a problem fixes a subject-matter (thought, language and various relations between them), it may offer little advice on the routes to be taken in solving the problem. It motivates posing a variety of other questions related to that subject matter - usually concerning the beliefs which have prompted the puzzle. But it is too unclear to provide strategic advice or criteria for success. The problem is solved when the sources of puzzlement have been removed; but this does not involve an answer to some question that was antecedently in play. How then is a problem represented? Following Dewey's remarks about indeterminate situations, we might suppose: as a set of propositions and questions which don't fit together. But we are aware of many sets of propositions and questions whose failure to cohere does not motivate us to inquiry. So we appear to require a non-cohering set of propositions and questions with a motivational force: we desire to remove this incoherence, and we are aware of it as a problem because it motivates us to find questions and answer them with the aim of distilling the problem into a question or set of questions and then answering it (them). The problem has the force of what is sometimes called a real doubt.1' A problem is a disturbing set of propositions and questions, and the problem will be resolved only when distilled into clear questions whose resolution renders our cognitive position undisturbing. While we should be able to say something about the 'structure' of particular problems, and perhaps even make a start on a search for a taxonomy of them, there is, I suspect, little ground for optimism that an account of the semantics of the sentences or representations that articulate problems will provide much of a clue to the construction of such a taxonomy. T o conclude: our understanding of a problem is largely determined by the way in which it emerges out of a complex structure of background knowledge. Its formulation contains clues that will direct us towards a grasp of its key elements and the features of our situation which are problematic, but it does not provide a semantic representation of a particular target for our inquiries and it has no clear semantic content. 6
It is likely that we will not understand the motivational force of such obscure representations without acknowledging that emotions, sentiments and other affective phenomena have a more substantial role in sustaining deliberation and rational cognition than many philosophers have supposed. See Hookway (1993). 213
Christopher Hookway Nominal Answers and the Question-Answer Relation As we noted, wherever we have no system of canonical designators for the values of the queriable, the logic of questions offers no guidance on how we should describe them in giving answers to questions. Consider, as an example, 'who' questions. Even if we think of proper names as, in some sense, canonical designators for people, an answer to the question 'Who is coming to dinner this evening?' which employs a proper name need not be satisfying. First, I can understand the question without knowing proper names for all possible satisfiers of the propositional function: the question can be posed and responded to by people with no knowledge of the guest's proper name. Second, we do not possess an a priori ordering of proper names which will guide searches through our body of information: even if we can order names alphabetically, there is no guarantee that some name we know has not been omitted from the ordering we employ. And third, since the 'who' question is an attempt to elicit information which can then be used to remove further problems or puzzlements, the required answer must be cast in a form that will enable us to link new information to old in a way that is relevant to our concerns. Thus whether something is a suitable nominal answer will depend upon the expression or sense of the expression used to refer to the person who satisfied the matrix. Even if I can 'understand' the question by grasping the abstract question, I will not be able to answer it successfully until I possess further information about the kind of mode of presentation to be used in the answer. The required mode of presentation will be a function of the questioner's purposes and current state of information. Understandably, given their interests, Belnap and Steel say little about these issues. In the case of discussion, how we understand what question is posed by an interrogative sentence is a matter for pragmatics. Understanding the sentence may serve to identify the matrix to be filled by a satisfactory direct answer and it may also suffice to determine the category of object a designation of which should replace the queriable. But we do not understand the question until we have learned from context what sort of mode of presentation is to be employed in identifying this object; and this is not always determined by the semantic properties of the sentence. We shall only learn what sort of mode of presentation is appropriate by taking account of the questioner's needs and background knowledge: an answer is required which is relevant to his purposes. Context guides the understanding of interrogatives in at least two ways. First, as just noted, it contributes to determining the kind of 214
Analyticity, Linguistic Rules and Epistemic Evaluation proposition that will serve as a possible answer. But second, it can also contribute to identifying the abstract question-structure involved. Context may make clear that 'who' does not mean 'which person' but rather means 'which of the people present at the moment', 'which of the members of the team', 'which of the relations of the accused', and so on. The category-restriction, and hence the identity of the question, may not be fully explicit from the interrogative sentence. These references to contextual determinants of how to answer a question may suggest that these are all clues enabling the respondent to identify the questioner's intention. However there is another phenomenon we need to take account of. We noted in the first section that questions have a role in eliciting information from those with more knowledge in the course of cooperative inquiry and discussion. Expertise may have a role in determining the content of a question as well as in providing information: an unclear question may be posed by someone who has little idea what precise category-restrictions are appropriate and little information about the sort of mode of presentation that should occur in an acceptable answer. The expert is trusted, not only to provide a reliable answer to the question, but also to re-articulate the problem disturbing the questioner: the unclear question invites its respondent to pose a clear question and answer it. Until the questioner receives an answer, she may not understand what question she was asking. This applies especially in the case of 'why' questions: in collaborative discussion and inquiry, such questions can be used to elicit information, but only because we can trust expert informants to know which questions we ought to have asked. I shall now turn towards some implications of these issues for our understanding of processes of (conscious) reasoning and to some speculations about thought and language. Since I mentioned three functions of questioning in deliberation (elicitation, setting targets and monitoring), I shall say something about each. The topics to be discussed will all stem from the role of questions which fall short of the highest standard of 'clarity'. They will also exploit the fact that once we take seriously the facts that reasoning is a goal-directed activity, and that questioning presents one locus of the active dimension of deliberation, we must acknowledge that many of the norms governing questioning (and hence deliberation) will be norms of practical rationality. It is important to note that pragmatic matters (issues of practical rationality) enter at two distinct stages. First, since monitoring and reflection are tied up with the activity of questioning, we need an account of the norms and mechanisms which determine whether a question should be raised at a particular stage of a deliberation. 215
Christopher Hookway Obviously, logic and semantics offer no such account: whether a question should be raised in the course of an inquiry depends upon the expected benefits of doing so. Habits and strategies for questioning can be vindicated by their efficiency in obtaining rapid answers to target questions and lemmas, to their fruitfulness in guiding us to reformulate questions or raise new important issues, and so on. In general, such strategies are vindicated if they are means to satisfying cognitive aims efficiently and reliably. Whether a question is appropriately raised is a function of the record of the inquiry so far, the background, and contextual information: I have no general picture of these pragmatic determinants to offer beyond noting that they are likely to take the form of flexible strategies and heuristics, to be highly context-sensitive and, quite possibly, to be specific to particular kinds of subject-matters or styles of inquiry. It would be unrealistic to expect a neat formal treatment of the rational determinants of the raising of questions; but since it seems plausible that the appropriateness of a question is sensitive to the expected utility of posing it, we would expect it to be determined by the full set of beliefs and expectations of the poser. Second, answering a question is not always (or indeed usually) a matter of selecting the true member of a class of possible answers which is determined by the content of the question. This picture would fit only those questions which explicitly or implicitly take the form of 'whether' questions. Whereas some 'unclear' questions may be rough-and-ready expressions of clearer underlying expressions, this is not always the case in conversation and there is no reason to think that it is always the case in deliberation either. The content of the question provides us with some necessary features of any proposition that can serve as an answer, but it need not fully determine the class of possible answers. A possible answer to the question posed will be a proposition meeting those necessary conditions which is 'good' from the point of view of the deliberation in process, which can serve as a premise in inferences that need to be carried out, or which will suggest further profitable routes of inquiry. Although the processing to be carried out in responding to such questions is strongly analogous to forms of processing studied within linguistic pragmatics, it is important to note that it will often resemble cases where one defers to an informant the role both of providing an answer and of determining what sort of proposition can actually serve as an answer. In that case, the individual inquirer occupies the double roles of questioner and 'better-informed respondent'. Pragmatic considerations — considerations about the instrumental role of the act of questioning within the process of inquiry - have a role in determining which propositions will serve as possible nomi216
Analyticity, Linguistic Rules and Epistemic Evaluation nal answers. No fully semantic account of the question-answer relation need be available. Conclusion Once we adopt the second picture of thought - as a problem-solving activity - then issues of practical rationality intrude: many of the evaluations we have to make concern means to ends and the desirability of various cognitive ends. This was emphasised in our discussion of the analytic/synthetic distinction where we suggested that some of Carnap's claims might be understood within this perspective. The later sections explored the suggestion that once this picture is adopted, issues about how we represent the structure of questions and problems become important. I argued that the key 'logical' relation - that between questions and their answers - is largely determined by pragmatic considerations, by means-ends reasoning which took note of the goals of the inquiries to which we are contributing. References Belnap, N. and Steel, T. 1976. The Logic of Questions and Answers. New Haven: Yale University Press Bromberger, S. 1992. On What We Know We Don't Know. University of Chicago Press Carnap, R. 1937. The Logical Syntax of Language. London: Routledge and Kegan Paul Dewey, J. 1938. Logic: the Theory of Inquiry. New York: Henry Holt Grice, H. P. and Strawson, P. F. 1956. 'In Defense of a Dogma', The Philosophical Review 65, 141-58 Harrah, D. 1963. Communication: A Logical Model. Cambridge, MA: MIT Press 1984. 'The Logic of Questions', in D. Gabbay and F. Guenther (eds.), Handbook of Philosophical Logic, Volume II. Dordrecht: D. Reidel Hookway, C. 1993. 'Mimicking Foundationalism: On Sentiment and SelfControl', The European Journal of Philosophy 1, 156-74 1994. 'Naturalised Epistemology and Epistemic Evaluation', Inquiry 37,465-85 Lipton, P. 1991. Inference to the Best Explanation. London: Routledge and Kegan Paul Popper, K. R. 1972. Objective Knowledge: An Evolutionary Approach. Oxford University Press Quine, W. V. 1953. From a Logical Point of View. Cambridge, MA: Harvard University Press 217
Christopher Hookway 1960. Word and Object. Cambridge, MA: MIT Press 1973. The Roots of Reference. La Salle, IL: Open Court 1994a. 'Responses', Inquiry 37, 495-505 1994b. 'Exchange between Donald Davidson and W. V. Quine following Davidson's lecture', Theoria 40, 226-31 Van Fraassen, B. C. 1980. The Scientific Image. Oxford University Press
218
How to do Other Things with Words DANIEL C. DENNETT Nomina si nescis, perit et cognitio rerum.1 (Lucretius, quoted by B. F. Skinner in Verbal Behavior, 1957, p. 441). John Austin's masterpiece, How to Do Things with Words, was not just a contribution to philosophy; it has proven to be a major contribution to linguistics, one of the founding documents of pragmatics, the investigation of how we use words to accomplish various ends in the social world. Strangely, not much attention has been paid by philosophers - or by psychologists and linguists - to how we use words in private, you might say, to think. As Wittgenstein (1967, p. 17e) once noted, 'It is very noteworthy that what goes on in thinking practically never interests us.' John Maynard Keynes was once asked if he thought in words or pictures. His reply was 'I think in thoughts.' This is in one way an admirable answer. It abruptly dismisses the two leading mistakes about what thinking might be. But it also encourages that lack of interest that Wittgenstein noted. What, then, are thoughts, if not words or pictures? And more particularly, what is the role of words in thinking? To this overdue question, I want to explore, very tentatively, a few neglected avenues.
Wordless Thought: A Phenomenological Fact First, I want to sound an alert about an all too familiar line of thought: the idea that thinking goes on in a special functional area of the brain. One of its many guises is what I have called the Cartesian Theatre, but it also appears in the roles of Central Processing, or the Central Executive, or Jerry Fodor's (1983) nonmodular central arena of belief fixation. What is wrong with this idea is not (just) that it (apparently) postulates an anatomically discernible central region of the brain — a maximally non-peripheral region, one might say - but that it supposes that there is a functionally identifiable subsystem (however located or distributed in the brain) that has some all too remarkable competences achieved by some all too remarkable means. There are many routes to it. Here is 1
If you are ignorant of names, knowledge of things perishes. 219
Daniel C. Dennett one that starts off in an excellent direction but then veers off. The mistaken fork is not explicitly endorsed by anybody that I can name, but I daresay it has covertly influenced a lot of thinking on the topic. (a) One of the things we human beings do is talk to others. (b) Another is that we talk to ourselves - out loud. (c) A refinement of (b) is to talk silently to oneself, but still in the words of a natural language, often with tone of voice and timing still intact. (For instance, one can often answer such questions as this: 'Are you thinking in English or French as you work on this problem?') So far, so good, an oft-told tale (going back at least to G. H. Mead 1930, Vygotsky 1934, and Skinner 1957), but watch out for the next step: (d) A further refinement of (c) is to drop the auditory/phonemic features (indeed all the features that would tie one's act to a specific natural language) and just think to oneself in bare propositions.
Introspection declares that something like this does occur but we must be circumspect in how we describe this phenomenon, since the temptation is to use it to anchor a remarkable set of dubious implications, to wit: (e) Since propositions, like numbers, are abstract objects, they would need some vehicles of embodiment in the brain. Moreover, when one thinks in bare propositions, one's thoughts still have one feature of sentences: logical form. (They must have logical form if we are going to explain the phenomenon of reliable deductive inference as the manipulation of these items.) (f) So there must be a medium of representation distinct from any natural language (call it the language of thought or Mentalese) that has this feature. and finally: (g) This activity of manipulating formulae of Mentalese is the fundamental variety of thinking - 'thinking proper' or 'real thinking' or even (as somebody recently put it) 'where the understanding happens'. It has seemed to some that this idea has been properly licensed by an analogy to the central processing unit or CPU of a von Neumann 220
How to do Other Things with Words machine, with its requirement that all instructions be in its proprietary machine language - the only language it 'understands'. But this is a broken-backed analogy, a fantasy that has little or nothing to do with how computers do their work. The 'machine language' understood by the CPU of a traditional computer is entirely composed of imperatives - commands or orders to execute basic tasks which have no external world reference at all. You can't say 'Snow is white' in machine language. You can't even say 'Look for something white' or 'Get some snow' in machine language. What you can say is things like 'Get whatever is is memory location X and add it to the number in the accumulator', or 'Output the result in the accumulator' or 'Jump to the instruction in memory location Y'. So machine language won't do as a model for Mentalese, even if we want to be computationalists (as I do) about thinking. We do introspect ourselves thinking, and sometimes it does seem that our thinking is wordless but 'prepositional'. This, then, is an indisputable (hetero-)phenomenological fact, bolstered by such widely acknowledged experiences as the tip-of-the-tongue phenomenon (in which we surely do have a particular content 'in mind' and are frustrated in our attempts to find the word that normally clothes it or accompanies it). Publishers have brought out humorous dictionaries of neologisms to fill in the gaps in English: 'sniglets' is proposed, I seem to recall, as a word for those handy little metal sleeves on the ends of shoelaces, and 'yinks' would name the pathetic strands of hair that some men drape carefully but ineffectively over their bald spots. We have all thought of (noticed, reflected upon) these items before - that's why these books are able to strike a humorous chord in us - so it seems obvious that we can have bare or wordless concepts in our consciousness. And so, it seems, we can think bare propositions composed of them. This kind of thinking is a personal level activity, an intentional activity, something we do. It is not just something that happens in our bodies. (Think of the sign that admonishes 'THINK!'). When we think thoughts of this sort, we do, it seems, manipulate our thoughts, and it can be difficult or easy work. This too is part of the heterophenomenology of consciousness, and hence forms part of the explicandum of any theory of thinking. But it is not given to introspection that these processes, or our perceptual processes, or any of the other cognitive transitions that are not personal actions but do just happen in our bodies, involve the transformation or manipulation of formulae of Mentalese. Even if the (d) phenomenon occurs quite frequently in people like us, professional thinkers, (e) and (f) do not follow, and there are good reasons to resist them. Here is one that is often overlooked. If we 221
Daniel C. Dennett view Mentalese as the lingua franca of all cognition, and view it as 'automatically understood', this apparently secures a 'solution' to the problem of understanding. But in spite of declarations by Fodor and others that Mentalese does not itself require interpretation by the brain that speaks it, this solution-by-fiat is both unsupported (except by an unsound analogy to machine language) and costly: it creates an artefactual problem about the 'access' of consciousness. How? If Mentalese is the lingua franca of all cognition, it must be the lingua franca of all the unconscious cognition in addition to the conscious thinking. Unconscious cognitive processes are granted on all sides, and if it is conducted in Mentalese (as is commonly asserted or assumed by theorists of the language of thought persuasion), getting some content translated into Mentalese cannot be sufficient for getting it into consciousness, even if it is sufficient for getting it understood. There must then be some further translation or transduction, into an even more central arena than central processing, into some extra system - for instance, Ned Block's (1992) postulated consciousness module. Beyond understanding lies conscious appreciation, according to this image, and it needs a place to happen in.2 This is what I have called the myth of double transduction, and I have criticised it elsewhere (Dennett 1996), so will not pursue it further here. There are other ways of modelling human thinking, even 'wordless' human thinking, that are much more plausible. Indeed, it is retrospectively astonishing that an idea as bizarrely unbiological as the language of thought has had as robust a career as it has. (What about the genetic code found in DNA? Isn't it a well-attested biological language? Yes, but it too, like machine language, consists of imperatives - orders to create proteins - combined into recipes. The descriptions of body parts and behaviours that emerge from the genetic code's recipes are amazingly indirect; you can't say 'Give this body the belief that snow is white' in DNA, even if something like this belief is innate in polar bears. So the genetic code is also a poor model for the language of thought.) Contrast the idea of a central processing unit in which orderly transactions are conducted in a common code with the following: a competitive arena in which many different sorts of things happen Grand Central Station, in which groups of visitors speaking many tongues try to find like-minded cohorts by calling out to each other, 2 If we follow Jackendoff's idea (1987), this would not be the central summit but a sort of ring surrounding that summit - a tempting idea, but not, I think, one to run with. (I'd rather get the benefits of Jackendoff's vision by other routes, other images, but that's a topic for another occasion.) 222
How to do Other Things with Words sweeping across the floor in growing crowds, waving their hands, pushing and shoving and gesturing - and occasionally managing to make contact and co-ordinate activity. How could such an alternative vision work? I have a few hints and hunches to offer. Making Tools to Think with Plato, in the Theaetetus (196c-d), compares human memory to a huge cage of birds: SOCRATES: Now consider whether knowledge is a thing you can possess in that way without having it about you, like a man who has caught some wild birds - pigeons or what-not - and keeps them in an aviary for them at home. In a sense, of course, we might say that he 'has' them all the time inasmuch as he possesses them, mightn't we?
THEAETETUS: Yes. SOCRATES: But in another sense he 'has' none of them, though he has got control of them, now that he has made them captive in an enclosure of his own; he can take and have hold of them whenever he likes by catching any bird he chooses, and let them go again; and it is open to him to do that as often as he pleases. The trick is: getting the right bird to come when you need it. How do we do it? By means of technology. We build elaborate systems of mnemonic association — pointers, labels, chutes and ladders, hooks and chains. We refine our resources by incessant rehearsal and tinkering, turning our brains (and all the associated peripheral gear we acquire) into a huge structured network of competences. (No evidence yet unearthed shows that any other animal does anything like that.) Getting the information into usable form, usable position - that is the task of resource management or resource refinement that faces us and, I want to suggest, we need words for that task. As Goethe once said, 'When ideas fail, words come in very handy.'3 Noam Chomsky has often said that birds don't have to learn their feathers and babies don't have to learn their language. I think there is a better parallel between birds and language: a child acquiring language is like a bird building a nest; it is a matter of 'instinctual' 3 I cited this in Dennett 1995 (p. 370), while doubting the attribution to Goethe, but as several readers informed me, he did indeed have Mephistopheles say it, in Faust: 'Denn eben wo Begriffe fehlen, Da stellt ein Wort zur rechten Zeit sich ein'. 223
Daniel C. Dennett or 'automatic' resource-enhancement, taking found objects and constructing something of great biological value - part of what Dawkins (1982) calls the extended phenotype - which blurs the boundary between an organism (or agent) and the environment in which it must act. Many years ago, Bertrand Russell (1927, pp. 32-3) made a wry observation: Animals studied by Americans rush about frantically, with an incredible display of hustle and pep, and at last achieve the desired result by chance. Animals observed by Germans sit still and think, and at last evolve the situation out of their inner consciousness. Wolfgang Kohler's (1925) early experiments with chimpanzees were the inspiration for Russell's witticism, which helps to perpetuate a common misunderstanding. Kohler's apes did not just sit and think up the solutions. They had to have many hours of exposure to the relevant props - the boxes and sticks, for instance - and they engaged in much manipulation of these items. Those apes that discovered the solutions — some never did — accomplished it with the aid of many hours of trial and error manipulating. Now were they thinking when they were fussing about in their cages? What were they manipulating? Boxes and sticks. It is all too tempting to suppose that their external, visible manipulations were accompanied by, and driven by, internal, covert manipulations - of thoughts about or representations of these objects, but succumbing to this temptation is losing the main chance. What they were attending to, manipulating and turning over and rearranging were boxes and sticks, not thoughts. They were familiarising themselves with objects in their environments. What does that mean? It means that they were building up some sort of perceptuo-locomotor structures tuned to the specific objects, discovering the affordances of those objects, getting used to them, making them salient, etc. So their behaviour was not all that different from the incessant trial and error scrambling of the behaviourists' cats, rats and pigeons. They were acting in the world, rearranging things in the world - without any apparent plan or insight or goal, at least at the outset. Animals at all levels are designed to tidy up their immediate environments, which are initially messy, confusing, intractable, dangerous, inscrutable, hard to move around in. They build nests, caches, escape tunnels, ambush sites, scent trails, territorial boundaries. They familiarise themselves with landmarks. They do all this to help them keep better track of the things that matter - predators 224
How to do Other Things with Words and prey, mates, etc. These are done by 'instinct': automated routines for improving the environment of action, making a better fit between agent and world. This wise husbandry of one's own behavioural territory is economically focused on the most important and ubiquitous features. Pre-eminent among these portions of the environment is the agent's own body, of course, always present, its condition always intensely relevant. Animals instinctively groom themselves and engage in behaviours that are apparently designed (although they need not realise this) to repair flaws, to maintain and improve their co-ordination and muscle tone, and, in effect, to familarise themselves with their own bodies. A part of the body that must not be overlooked in this maintenance and improvement schedule is the brain. It, too, can become messy, confusing, inscrutable, an overwhelmingly complex arena of action. So we should expect animals to be instinctually equipped to engage in mindless, automatic routines that tidy up their own brains. We should especially expect it in Homo sapiens, whose huge brains are so plastic, so inundatable, so at the mercy of invading memes and memories. Resource management for a young human brain is, I think, a major task, and we should expect it to be accomplished by activities that are rooted, at the outset, in our biology, in our 'instincts', but which also get enhanced in major ways by techniques that are themselves part of the influx of new resources. Human infants engage in exploratory activities much like those exhibited by the chimpanzees, handling every object within reach, bringing it to their mouths, looking at it, over and over. One of the fruits of this repetitive activity is a strong capacity to recognise and track individual things. Tracking and recognition are strictly different phenomena - you can track something you cannot identify at all, except as the whatever-it-is you're tracking. Recognition depends heavily on this familiarisation process. Consider the potent experience of seeing, for the first time since you were, say, five years old, a favourite toy or article of clothing from your childhood; the warm rush of recognition suggests that there has been very little diminution in the associational structures built in your brain so many years before. This strong capacity to recognise does not cover items or individuals or places of mere passing acquaintance with which you have not actively familiarised yourself. And yet active as the process is, it need not be deliberate or directed or thoughtful - it can be unthinking, automatic. We more mature Homo sapiens also engage in a more directed version of this behaviour, purposefully and attentively moving things around in the world in order to solve problems. For instance, most 225
Daniel C. Dennett Scrabble players would be seriously handicapped if they were prevented from sliding the little tiles around on their little shelf. We write words on index cards and slips of paper, doodle and diagram and sketch. These concrete activities are crutches for thinking, and once we get practised at these activities, we can to some degree internalise these manipulations. Can you alphabetise the words in this sentence in your head?
Yes, probably, but what you do is not easy. It is not easy because you have to keep track of things. To perform this stunt, you need to use visual imagery and auditory rehearsal. You work to make each component as vivid as possible, to help you do the required tracking. If we were to expand the task just a little bit by making the words hard to distinguish visually, or by lengthening the sentence, you would find it impossible to alphabetise in your head - and this very sentence is such an instance. Try it. Your capacity to alphabetise a list of words in your head, or do mental arithmetic, is an advanced trick, not a novice trick, but it doesn't differ in kind in any important way from the capacity to do the same thing with the aid of paper and pencil. You have made a somewhat fancier nest out of your resources - that's all. What Can We Do with Our Tools? Even very simple animals are capable of associative learning - what I have called ABC learning (for Associative, Behaviourist, Connectionist). We human beings have the capacity for another kind of learning: swift, insightful learning that does not depend on laborious training but is ours as soon as we contemplate a suitable symbolic representation of the knowledge. When psychologists devise a new experimental setup or paradigm in which to test such non-human subjects as rats or cats or monkeys or dolphins, they often have to devote dozens or even hundreds of hours to training each subject on the new tasks. Human subjects, however, can usually just be told what is desired of them. After a brief question-andanswer session and a few minutes of practice, we human subjects will typically be as competent in the new environment as any agent ever could be. Of course, we do have to understand the representations presented to us in these briefings, and that's where the transition from ABC learning to our kind of learning is still shrouded in fog. An insight that may help clear the fog is a familiar maxim of artefact making: if you 'make it yourself, you understand it. Although ABC learning can yield remarkably subtle and power226
How to do Other Things with Words ful discriminatory competences, capable of teasing out the patterns lurking in voluminous arrays of data, these competences tend to be anchored in the specific tissues that are modified by training. They are 'embedded' competences, in the sense that they are incapable of being 'transported' readily to be brought to bear on other problems faced by the individual, or shared with other individuals. The philosopher Andy Clark and the psychologist Annette Karmiloff Smith have recently been exploring the transition from a brain that has only such embedded knowledge to a brain that, as they say, 'enriches itself from within by re-representing the knowledge that it has already represented'. Clark and Karmiloff-Smith note that while there are clear benefits to a design policy that 'intricately interweave[s] the various aspects of our knowledge about a domain in a single knowledge structure', there are costs as well: 'The interweaving makes it practically impossible to operate on or otherwise exploit the various dimensions of our knowledge independently of one another' (Clark and Karmiloff-Smith 1993, pp. 494-5). So opaquely is such knowledge hidden in the mesh of the connections that 'it is knowledge in the system, but it is not yet knowledge to the system' (ibid., p. 495) - like the wisdom revealed in the precocious single-mindedness with which the newly hatched cuckoo shoulders the competing eggs out of the nest. What would have to be added to the cuckoo's computational architecture for it to be able to appreciate, understand and exploit the wisdom interwoven in its neural nets? (Cf. Dretske 1991, Dennett 1987, pp. 3O6ff, and Dennett 1992.) A popular answer to this question, in its many guises, is 'symbols!' The answer is well-nigh tautological, and hence is bound to be right in some interpretation. How could it not be the case that implicit or tacit knowledge becomes explicit by being expressed or rendered in some medium of 'explicit' representation? Symbols, unlike the nodes woven into connectionist networks, are movable; they can be manipulated; they can be composed into larger structures in which their contribution to the meaning of the whole can be a definite and generatable function of the structure - the syntactic structure - of the parts. There is surely something right about this, but we must proceed cautiously, since many pioneers have posed these questions in ways that have turned out to be misleading. To anchor a free-floating rationale (such as the cuckoo's) to an agent in the strong way, so that it is the agent's own reason, the agent must 'make' something. A representation of the reason must be composed, designed, edited, revised, manipulated, endorsed. How does any agent come to be able to do such a wonderful thing? Does it have to grow a new organ in its brain? Or could it build this com227
Daniel C. Dennett petence out of the sorts of external-world manipulations it has already mastered? Children enjoy talking to themselves. What might this be doing to their minds?4 I cannot answer that question yet, but I have some speculative suggestions for further research. Consider what happens early in the linguistic life of any child. 'Hot!' says Mother. 'Don't touch the stove!' At this point, the child doesn't have to know what 'hot' or 'touch' or 'stove' mean - these words are primarily just sounds, auditory event-types that have a certain redolence, a certain familiarity, a certain echoing memorability to the child. They come to conjure up a situation-type - stove-approach-and-avoidance which is not just a situation in which a specific prohibition is typically heard but also a situation in which a mimicking auditory rehearsal is encountered. Crudely simplifying, let's suppose that the child acquires the habit of saying to itself (aloud) 'Hot!' 'Don't touch!' without much of an idea what these words means, voicing them merely as an associated part of the drill that goes with approaching and then avoiding the stove — and also as a sort of mantra, which might be uttered at any other time. After all, children are taken with the habit of rehearsing words they have just heard — rehearsing them in and out of context — building up recognition links and association paths between the auditory properties and concurrent sensory properties and other internal states. They are familiarising themselves with these newfound tools. That's a rough sketch of the sort of process that must go on. This process could have the effect of initiating a habit of what we might call semi-understood self-commentary. The child, prompted initially by some insistent auditory associations provoked by its parents' admonitions, acquires the habit of adding a soundtrack to its activities - 'commenting' on them. The actual utterances would consist at the outset of large measures of what psychologists and linguists have called 'scribble' - nonsense talk composed of wordlike sounds - mixed with real words mouthed with much feeling but little or no appreciation of their meaning, and a few understood words. There would be mock exhortation, mock prohibition, mock praise, mock description, and all these would eventually mature into real exhortation, prohibition, praise and description. But the habit of adding the 'labels' would thus be driven into place before the labels themselves were understood, or even partially understood. I'm suggesting that it's such initially 'stupid' practices - the mere mouthing of labels in circumstances appropriate and inappropriate - that could eventually be turned into the habit of representing 4 For recent overviews of research in this area, see Diaz and Berk (1992), and Fernyhough (1996). 228
How to do Other Things with Words one's own states and activities to oneself in a new way. As the child lays down more associations between the auditory and articulatory processes on the one hand, and patterns of concurrent activity on the other, this would create nodes of saliency in memory. A word can become familiar even without being understood. And it is these anchors of familiarity that could give a label an independent identity within the system. Without such independence, labels are invisible. For a word to serve as a useful, manipulable label in the refinement of the resources of a brain, it must be a ready enhancer of sought-for associations that are already to some extent laid down in the system. Beyond that, words can be arbitrary, and their arbitrariness is actually part of what makes them distinctive. There is little risk of failing to notice the presence of the label; it doesn't just blend into its surroundings. It wears its artificial status on its sleeve. The habit of semi-understood self-commentary could, I am suggesting, be the origin of the practice of deliberate labelling, in words (or scribble words or other private neologisms), which in turn could lead to a still more efficient practice: dropping all or most of the auditory and articulatory associations and just relying on the rest of the associations (and association-possibilities) to do the anchoring. The child, I suggest, can abandon out-loud mouthings and create private, unvoiced neologisms as labels for features of its own activities. We can take a linguistic object as a found object (even if we have somehow blundered into making it ourselves rather than hearing it from someone else) and store it away for further consideration, offline. Our ability to do this depends on our ability to re-identify or recognise and track such a label on different occasions, and this in turn depends on the label having some feature or features by which to remember it - some guise independent of its meaning. Once we have created labels and acquired the habit of attaching them to experienced circumstances, we have created a new class of objects that can themselves become the objects of all the patternrecognition machinery, association-building machinery, and so forth. Like the scientists lingering retrospectively over an unhurried examination of photographs they took in the heat of experimental battle, we can reflect on whatever patterns there are to be discerned in the various labelled exhibits we dredge out of memory. As we improve, our labels become ever more refined, more perspicuous, ever better articulated, and the point is finally reached when we approximate the near-magical human prowess we began with: the mere contemplation of a representation is sufficient to call to mind the appropriate lessons. At this point we are well on our way to becoming understanders of the objects we have created. We might 229
Daniel C. Dennett call these artefactual nodes in our memories, these pale shadows of articulated and heard words, concepts. A concept, then, is a proprietary label which may or may not include among its many associations the auditory and articulatory features of a word (public or private). But words, I am suggesting, are the prototypes or forebears of concepts. Bickerton (1995, p. I l l ) speaks of 'linguistic concepts (undressed words)', and I am suggesting that the first concepts one can manipulate are 'voiced' concepts or well-dressed words. Only concepts that can be manipulated can become objects of scrutiny for us. If you have ever tried panning for gold, you know that the right technique of swirling and swishing and draining is not transparently obvious. There is a simple trick, however, that can swiftly lead you to competence or even expertise. Into each panful of muddy sand and gravel you take up, sprinkle a few - half a dozen - tiny lead shots. They are almost as dense as any gold particles present, and will behave almost the same. As you near the end of your swirling, most of the material having been sluiced away, you should be able to see the little group of lead shot emerge once more from hiding. (If they do not appear, you have been too vigorous, and thrown out the baby with the bathwater! Try again, more gently.) Once you get the hang of it, the cluster of lead spheres will chaperone any bits of gold in the pan, which seem to be drawn to their neighborhood as if by magnetism. This activity of seeding a confused mess with something known and reidentifiable, something familiar, and then leaning on the familiar to help track the novel, has a striking parallel in the use of symbols. This comes out clearly in a recent experiment by Thompson, Oden and Boyson (forthcoming). As Andy Clark explains, in another forthcoming paper (this is fast-breaking news in a new research area), The idea...is that learning to associate concepts with discrete arbitrary labels (words) somehow makes it easier to use those concepts to constrain computational search and hence enables the acquisition of a cascade of more complex and increasingly abstract ideas. The claim...is thus that associating a perceptually simple, stable, external item (such as a word) with an idea, concept or piece of knowledge effectively freezes the concept into a sort of cognitive building-block - an item that can then be treated as a simple baseline feature for future episodes of thought, learning and search. Consider a simple matching task - such as matching one cup to another cup in an array of different objects. Compare that with the 230
How to do Other Things with Words task of matching a relation-between-relations - such as matching a pair of identical items (two shoes) to a different pair of identical items (two cups) amidst a collection of odd pairs (cup and ball, shoe and fork, etc). Or even harder, matching a pair of different items with the only other pair of items that were different. It has been shown that chimps can discriminate the basic relations of similarity and difference, but only language-trained chimps (Premack 1986, Premack and Premack 1983) can master the higher-order judgement required to act on the fact that this pair, like that pair, consists of a brace of different items. (Notice that I did not impute any language of thought to these chimps. They pick up on a higher-order similarity but need not be able to express this competence in any verbal way, even to themselves.) As Thompson et al. show, this is due not to the minimal syntactic competence such linguistic training has imparted (to Premack's chimps), but to 'simply the experience of associating abstract relations with arbitrary tokens'. The chimps were given familiarity with an arbitrary symbol or label associated with 'SAME' or ' D I F F E R E N T ' ; the token, then, is a cognitive crutch, a simplifier of the perceptual world that helps the animal track a pattern that is otherwise just beyond its ken. So this tracking of higher-order categories is a stunt, not a primitive building-block, and as we learn to throw away the sounds and other associations, we make the stunt ever more sophisticated. In this arena, wordless thought should be seen as a rather exotic specimen instead of the foundation of all cognition. The image I want to explore has it that these rare, hypersophisticated transitions occur when very sketchy images of linguistic representations serve as mnemonic triggers for 'inferential' processes that generate further sketchy images of linguistic representations and so forth. (This sort of short-cut transition can take place in any modality, any system of associations. For a trained musician, the circle of fifths imposes itself automatically and involuntarily on a fragment of heard or imagined music, just the way modus ponens intrudes on the perceptions of a trained logician, or the offside line does on a trained soccer player). This way of trying to imagine word-free 'logical inference' makes it look rather like barefoot waterskiing - a stunt that professionals can make look easy, but that is hardly the basic building-block of successful transportation/cognition in the everyday concrete world. And even when we are very, very good, we sometimes fall back on the old crutches. If your mother tongue is English, you may find that long after you have achieved something approaching fluency in French or German, you have to convert 'vierundzwanzig' or 'quatre-vingt douze' into your more familiar mode of numerical repre231
Daniel C. Dennett sentation. Why is this so? Because you have turned your mothertongue words into highly familiarised objects, anchored by a host of associations. They are birds that not only come when you call, but call forth flocks of further birds. Consider a familiar human activity that we rely on in many problem-solving circumstances: we ask ourselves explicit questions. This practice has no readily imaginable counterpart in non-linguistic animals, but what does it gain us, if anything? We are calling the birds. To get the right bird to come when you call is a task calling for technical skills if your birds are stored somewhere in your library or on your hard disk, but also if they are stored in your head. You need to develop retrieval strategies, ways of manipulating your own brain. These methods are themselves tools. You don't use the tools; they jump into action on their own, bidden by circumstances, and beneath your conscious control. They are tools that have to call themselves into action when needed, beating out the competition. Have you ever danced with a movie star? Do you know where to buy live eels? Could you climb five flights of stairs carrying a bicycle and a cello? These are questions the answers to which were probably not already formulated and handily stored in your brain (as sentences of Mentalese in your belief box!), and yet these questions are readily — even 'unthinkingly' — answered reliably by most people. How do they do it? By engaging in relatively effortless and automatic 'reasoning'. In the first case, if no recollection of the presumably memorable event is provoked by considering the question, you conclude that the answer is No. The second question initiates a swift survey (pet stores? fancy restaurants? fish markets or live bait dealers?), and the third provokes some mental imagery which reminds you of relevant facts about your own body and its competences. These are facts about oneself that wouldn't 'automatically' come to mind when needed without a little extra provocation, a little calling. The way to get this sort of information into useful position for current purposes is to ask ourselves questions and see what comes back! Do we need to ask ourselves grammatical questions? Not necessarily. If we look at the history of artificial intelligence, we can detect a series of clashes over the years between the 'neats' and the 'scruffies'. The neats, epitomised by GOFAI (Good Old Fashioned AI - Haugeland 1985), have worked out formal methods for 'querying data-bases', and in these schemes, the logical or grammatical structure of the probes is indeed all-important. The systems work by mechanised formal deduction. This has proven feasible for many real-world applications in which the information stored is itself highly systematic and bounded — airline routes and timetables, stock 232
How to do Other Things with Words market prices and other information, medical records. But such systems are often 'brittle' and hard to revise or expand. They are not at all biological in their shape, at any level, and do not appear remotely applicable to the sorts of open-ended, time-pressured, freewheeling demands of controlling real world engagements. The scruffies, meanwhile, have favoured various less disciplined, less formal ways of teasing the relevant data out of their pigeonholes. At some level of sophistication, and for certain advanced tasks, syntax is surely obligatory. Turning mere 'labels' into names and predicates, nouns and verbs and auxiliaries, is necessary to get the compositionality that permits us - and apparently only us - to extend our comprehension to all topics. But does this structure have to be in the brain from the outset, part of the underlying computational architecture, or can it be brought in, piecemeal, from the cultural environment, to discipline our brains at a higher level? This is an empirical question, and now that we can dimly see the scruffy alternative, it would be wise to push it as far as we can, knowing that there is plenty of structure in the artefactual world to avail ourselves of, for special purposes. The elementary way to get the birds to come is the way real birds do it, by just calling - yoohoo, in hundreds of varieties, but without syntax. Once the birds have been trained to fly in formation, this will provide further structures out of which to build further competences. And where does consciousness come into the picture? It is already there, unnoticed, in the activity just described. Mental contents become conscious not by entering some special chamber in the brain, not by being transduced into some privileged and mysterious medium, but by winning the competitions against other mental contents for domination in the control of behaviour, and hence for achieving long-lasting effects - or as we misleadingly say, 'entering into memory'. And since we are talkers, and since talking to ourselves is one of our most influential activities, one of the most effective ways - not the only way - for a mental content to become influential is for it to get into position to drive the language-using parts of the controls. All this has to happen in the arena of the brain, in 'central processing', but not under the direction of anything. The Person is the virtual governor, not a real governor; it is the effect of the processes, not their cause. A common reaction to this suggestion about human consciousness is frank bewilderment, expressed more or less as follows: 'Suppose all these strange competitive processes are going on in my brain, and suppose that, as you say, the conscious processes are simply those that win the competitions. How does that make them con233
Daniel C. Dennett scious? What happens next to them that makes it true that / know about them? For after all, it is my consciousness, as I know it from the first-person point of view, that needs explaining!' Such questions betray a deep confusion, for they presuppose that what you are is something else, some Cartesian res cogitans in addition to all this brain-and-body activity. What you are, however, just is this organisation of all the competitive activity between a host of competences that your body has developed. You 'automatically' know about these things going on in your body, because if you didn't, it wouldn't be your body! The acts and events you can tell us about, and the reasons for them, are yours because you made them - and because they made you. What you are is that agent whose life you can tell about. You can tell us, and you can tell yourself. The process of self-description begins in earliest childhood, and includes a good deal of fantasy from the outset. (Think of Snoopy in the Peanuts cartoon, sitting on his doghouse and thinking 'Here's the World War I ace, flying into battle...') It continues through life. (Think of the cafe waiter in Jean Paul Sartre's discussion of 'bad faith' in Being and Nothingness (1943), who is all wrapped up in learning how to live up to his selfdescription as a waiter.) It is what we do. It is what we are.5
References Bickerton, D. 1995. Language and Human Behavior. University of Washington Press Block, N. J. 1992. 'Begging the Question against Phenomenal Consciousness', (commentary on Kinsbourne and Dennett), The Behavioral and Brain Sciences 15, 205-6
Clark, A. and Karmiloff-Smith, A. 1993. 'The Cognizer's Innards: A Psychological and Philosophical Perspective on the Development of Thought', Mind and Language 8, 487-519 Clark, A. (forthcoming). 'Magic Words: How Language Augments Human Computation' in P. Carruthers and J. Boucher (eds.), Language and Thought: Interdisciplinary Themes. Cambridge University Press Dawkins, R. 1982. The Extended Phenotype: The Gene as the Unit of Selection. Oxford and San Francisco: Freeman Dennett, D. C. 1987. The Intentional Stance. Cambridge, MA: MIT Press 1992. 'La Comprehension Artisanale', in Daniel C. Dennett et les Strategies Intentionnelles, in Lekton 2, Universite de Quebec a Montreal 1993. 'Learning and Labeling' (commentary on Clark and Karmiloff-Smith 1993), Mind and Language 8, 540-47 5
Parts of this paper are drawn, with slight revisions, from Dennett (1993) and (1996). 234
How to do Other Things with Words — 1995. Darwin's Dangerous Idea: Evolution Evaluation and the Meanings of Life. New York: Simon & Schuster, and London: Allen Lane 1996. Kinds of Minds. New York: Basic Books, and London: Weidenfeld & Nicholson Diaz, R. M. and Berk, L. E. 1992. Private Speech: From Social Interaction to Self-Regulation. Hillsdale, NJ: Erlbaum Dretske, F. I. 1991. 'Replies' in B. McLaughlin (ed.), Dretske and his Critics. Oxford: Blackwell Fernyhough, C. 1996. 'The Dialogic Mind: A Dialogic Approach to the Higher Mental Functions', New Ideas in Psychology 14, 47—62 Fodor, J. A. 1983. The Modularity of Mind. Cambridge, MA: MIT Press Haugeland, J. 1985. Artificial Intelligence: The Very Idea. Cambridge, MA: MIT Press Jackendoff, R. 1987. Consciousness and the Computational Mind. Cambridge, MA: MIT Press Kohler, W. 1925. The Mentality of Apes. New York: Harcourt Brace and World Mead, G. H. 1930. Mind, Self, and Society. University of Chicago Press Premack, D. 1986. Gavagai! Or, the Future History of the Animal Language Controversy. Cambridge, MA: MIT Press Premack D. and Premack, A. J. 1983. The Mind of an Ape. New York and London: Norton Russell, B. A. W. 1927. An Outline of Philosophy. London: Allen & Unwin Sartre, J.-P. 1943/1953. Being and Nothingness, translated by H. Barnes. New York: Philosophical Library Skinner, B. F. 1957. Verbal Behavior. New York: Appleton-CenturyCrofts Thompson, R., Oden, D. and Boyson, S. (forthcoming). 'Language-Nai've Chimpanzees Judge Relations-between-relations in an Abstract Matching Task', Journal of Experimental Psychology: Animal Behavior Processes Vygotsky, L. S. 1934/1962. Thought and Language, edited and translated by E. Hanfmann and G. Vakar. Cambridge, MA: MIT Press 1934/1986. Thought and Language, translated by A. Kozulin. Cambridge, MA: MIT Press Wittgenstein, L. 1967. Zettel, edited by G. E. M. Anscombe, G. H. von Wright, and translated by G. E. M. Anscombe. Oxford: Blackwell
235
Bibliography: Twentieth-Century Philosophical Texts on Thought and Language Alexander, H. G. 1967. Language and Thinking: A Philosophical Introduction. Princeton: Van Nostrand Anscombe, G. E. M. 1981. Metaphysics and the Philosophy of Mind: Collected Philosophical Papers, Volume 2. Oxford: Blackwell Apel, K.-O. 1967. Analytic Philosophy of Language and the Geisteswissenschaften. Dordrecht: D. Reidel Armstrong, D. M. 1968. A Materialist Theory of the Mind. London: Routledge and Kegan Paul Austin, J. L. 1962. How to Do Things with Words. Oxford: Clarendon Press Avramides, A. 1989. Meaning and Mind: An Examination of a Gricean Account of Language. Cambridge, MA: MIT Press Ayer, A. J. 1936. Language, Truth and Logic. London: Gollancz 1947. Thinking and Meaning. London: H. K. Lewis Bach, K. 1987. Thought and Reference. Oxford: Clarendon Press Baker, G. P. and Hacker, P. M. S. 1984. Language, Sense and Nonsense. Oxford: Blackwell Baker, L. R. 1987. Saving Belief: A Critique of Physicalism. Princeton University Press Bartlett, F. 1958. Thinking: An Experimental and Social Study. London: Allen & Unwin Bennett, J. 1964. Rationality: An Essay Towards Analysis. London: Routledge and Kegan Paul 1976. Linguistic Behaviour. Cambridge University Press Bever, T., Carroll, J. M. and Miller, L. A. (eds.). 1984. Talking Minds: The Study of Language in Cognitive Science. Cambridge, MA: MIT Press Blackburn, S. 1984. Spreading the Word: Groundings in the Philosophy of Language. Oxford: Clarendon Press Blanshard, B. 1939. The Nature of Thought. London: Allen & Unwin Block, N. J. (ed.). 1981. Readings in the Philosophy of Psychology, Volume II. London: Methuen Bruner, J., Goodnow, J. J. and Austin, G. A. 1956. A Study of Thinking. New York: John Wiley Carnap, R. 1927. Der Logische Aufbau der Welt. English translation: The Logical Structure of the World. Berkeley: University of California Press, 1967 1937. The Logical Syntax of Language. London: Routledge and Kegan Paul Carroll, J. B. 1964. Language and Thought. Englewood Cliffs, NJ: PrenticeHall Carruthers, P. 1996. Language, Thought and Consciousness: An Essay in Philosophical Psychology. Cambridge University Press 237
Bibliography Carruthers, P. and Boucher, J. (eds.). (forthcoming). Language and Thought: Interdisciplinary Themes. Cambridge University Press Chisholm, R. M. 1981. The First Person: An Essay on Reference and Intentionality. Sussex: Harvester Press Chomsky, N. 1972. Language and Mind. Enlarged Edition. New York: Harcourt Brace Jovanovich 1980. Rules and Representations. Oxford: Blackwell 1986. Knowledge of Language. New York: Praeger 1988. Language and Problems of Knowledge - The Managua Lectures. Cambridge, MA: MIT Press 1993. Language and Thought. London: Moyer Bell Coulter, J. 1979. The Social Construction of Mind. London: Macmillan Crane, T. 1995. The Mechanical Mind. London: Penguin Davidson, D. 1984. Inquiries into Truth and Interpretation. Oxford: Clarendon Press Davidson, D. and Hintikka, K. J. J. (eds.). 1969. Words and Objections: Essays on the Work of W. V. Quine. Dordrecht: D. Reidel Dennett, D. C. 1969. Content and Consciousness. London: Routledge and Kegan Paul 1978. Brainstorms: Philosophical Essays on Mind and Psychology. Cambridge, MA: MIT Press 1987. The Intentional Stance. Cambridge, MA: MIT Press Derrida, J. 1967. La Voix et le Phenomene. Paris: Presses Universitaires de France. English translation: Speech and Phenomena. Evanston, IL: Northwestern University Press, 1973 Devitt, M. and Sterelny, K. 1987. Language and Reality: An Introduction to the Philosophy of Language. Oxford: Blackwell Douglas, M. (ed.). 1973. Rules and Meanings. Harmondsworth: Penguin Dummett, M. A. E. 1981. Frege: Philosophy of Language, 2nd edition. London: Duckworth 1991. Frege and Other Philosophers. Oxford: Clarendon Press 1993a. Origins of Analytical Philosophy. London: Duckworth 1993b. The Seas of Language. Oxford: Clarendon Press Eco, U, Santambrogio, M. and Violi, P. (eds.). 1988. Meaning and Mental Representations. Bloomington: Indiana University Press Egidi, R. (ed.). 1995. Wittgenstein: Mind and Language, Dordrecht: Kluwer Evans, G. 1982. The Varieties of Reference. Oxford: Clarendon Press Farber, M. 1943. The Foundation of Phenomenology. New York: SUNY Press Fodor, J. A. 1975. The Language of Thought. Sussex: Harvester Press 1981. Representations: Philosophical Essays on the Foundations of Cognitive Science. Sussex: Harvester Press 1987. Psychosemantics: The Problem of Meaning in the Philosophy of Mind. Cambridge, MA: MIT Press 1994. The Elm and the Expert: Mentalese and its Semantics. Cambridge, MA: MIT Press Fodor, J.A., Bever, T. and Garrett, M. 1974. The Psychology of Language: 238
Bibliography An Introduction to Psycholinguistics and Generative Grammar. New York: McGraw-Hill Frege, G. 1980. Translations from the Philosophical Writings of Gottlob Frege. 3rd edition, ed. P. T. Geach and M. Black. Oxford: Blackwell 1984. Collected Papers on Mathematics, Logic and Philosophy, ed. B. McGuinness. Oxford: Blackwell French, P., Uehling, T. E. and Wettstein, H. K. (eds.). 1979. Contemporary Perspectives in the Philosophy of Language. Minneapolis: University of Minnesota Press Gadamer, H.-G. 1960. Wahrheit und Methode. English translation: Truth and Method. London: Sheed and Ward, 1975 1967. Philosophische Hermeneutik. English translation: Philosophical Hermeneutics, Baltimore: Johns Hopkins University Press, 1976 Geach, P. T. 1957. Mental Acts. London: Routledge and Kegan Paul 1972. Logic Matters. Oxford: Blackwell Geirsson, H. and Losonsky, M. (eds.). 1996. Readings in Language and Mind. Oxford: Blackwell George, A. (ed.). 1989. Reflections on Chomsky. Oxford: Blackwell Gillett, G. 1992. Representation, Meaning, and Thought. Oxford University Press Glock, H.-J. 1996. A Wittgenstein Dictionary. Oxford: Blackwell Goodman, N. 1969. Languages of Art. Oxford University Press 1984. Of Mind, and Other Matters. Cambridge, MA: Harvard University Press Gregory, R. L. (ed.). 1987. The Oxford Companion to the Mind. Oxford University Press Grice, H. P. 1989. Studies in the Way of Words. Cambridge, MA: Harvard University Press Guttenplan, S. (ed.). 1975. Mind and Language. Oxford: Clarendon Press 1994. A Companion to the Philosophy of Mind. Oxford: Blackwell Hacker, P. M. S. 1990. Wittgenstein: Meaning and Mind. Oxford: Blackwell 1996. Wittgenstein: Mind and Will. Oxford: Blackwell Hacking, I. 1975. Why Does Language Matter to Philosophy?. Cambridge University Press Hale, R. and Wright, C. (eds.). 1997. A Companion to the Philosophy of Language. Oxford: Blackwell Harman, G. 1973. Thought. Princeton University Press Harris, R. 1981. The Language Myth. London: Duckworth 1996. The Language Connection: Philosophy and Linguistics. Bristol: Thoemmes Harris, R. (ed.). 1996. The Origin of Language. Bristol: Thoemmes Hattiangadi, J. N. 1987. How is Language Possible?: Philosophical Reflections on the Evolution of Language and Knowledge. LaSalle, IL: Open Court Hayakawa, S. I. 1949. Language in Thought and Action. New York: Harcourt Brace Jovanovich Hickmann, M. (ed.). 1987. Social and Functional Approaches to Language and Thought. Orlando: Academic Press 239
Bibliography Hoffman, R. 1970. Language, Minds and Knowledge. London: Allen & Unwin Hook, S. (ed.)- 1969. Language and Philosophy. New York University Press Hookway, C. 1988. Quine: Language, Experience and Reality. Cambridge: Polity Press Husserl, E. 1900—1. Logische Untersuchungen. English translation: Logical Investigations. London: Routledge and Kegan Paul, 1970 1929. Formale und transzendentale Logik. English translation: Formal and Transcendental Logic. The Hague: Martinus Nijhoff, 1969 1973. Ehrfahrung und Urteil. English translation: Experience and Judgement. London: Routledge and Kegan Paul Ilyenkov, E. V. 1977. Dialectical Logic: Essays on its History and Theory. English translation, Moscow: Progress Publishers Jackendoff, R. 1985. Semantics and Cognition. Cambridge, MA: MIT Press James, W. 1890. The Principles of Psychology. Reprinted, Cambridge, MA: Harvard University Press, 1981 Katz, J. J. 1966. The Philosophy of Language. New York: Harper & Row 1981. Language and Other Abstract Objects. Oxford: Blackwell 1988. Cogitations. Oxford University Press Katz, J. J. (ed.). 1985. The Philosophy of Linguistics. Oxford University Press Kempson, R. (ed.), 1988. Mental Representations: The Interface Between Language and Reality. Cambridge University Press Kenny, A. J. P. 1989. The Metaphysics of Mind. Oxford University Press Korner, S. 1959. Conceptual Thinking: A Logical Enquiry. New York: Dover Kripke, S. 1980. Naming and Necessity. Oxford: Blackwell. Kiing, G. 1967. Ontology and the Logistic Analysis of Language. Dordrecht: D. Reidel Lakoff, G. 1987. Women, Fire, and Dangerous Things: What Categories Reveal About the Mind. University of Chicago Press Lakoff, G. and Johnson, M. 1980. Metaphors We Live By. University of Chicago Press Lektorsky, V. A. 1980. Subject, Object, Cognition. English translation, Moscow: Progress Publishers, 1984 Loar, B. 1981. Mind and Meaning. Cambridge University Press Luntley, M. 1998. Contemporary Philosophy of Thought. Oxford: Blackwell Luria, A. R. 1979. The Making of Mind: A Personal Account of Soviet Psychology. London: Harvard University Press Lycan, W. G. 1986. Logical Form in Natural Language. Cambridge, MA: MIT Press McCulloch, G. 1995. The Mind and its World. London: Routledge McDowell, J. 1994. Mind and World. Cambridge, MA: Harvard University Press McGinn, C. 1982. The Character of Mind. Oxford University Press 1989. Mental Content. Oxford: Blackwell Malcolm, N. 1977. Thought and Knowledge. Ithaca, NY: Cornell University Press 1986. Wittgenstein: Nothing is Hidden. Oxford: Blackwell 240
Bibliography 1995. Wittgensteinian Themes: Essays 1978-1989. Ithaca, NY: Cornell University Press Martin, R. M. 1987. The Meaning of Language. Cambridge, MA: MIT Press Mead, G. H. 1934. Mind, Self and Society. University of Chicago Press Meiland, J. W. 1970. Talking about Particulars. London: Routledge and Kegan Paul Merleau-Ponty, M. 1948. Sense and Nonsense. English translation: Evanston, IL: Northwestern University Press, 1974 Merleau-Ponty, M. 1969. The Prose of the World. English translation: Evanston, IL: Northwestern University Press, 1973 Merleau-Ponty, M. 1973. Consciousness and the Acquistion of Language. Evanston, IL: Northwestern University Press Merrill, D. D. and Grimm, R. H. (eds.). 1988. Contents of Thought. Tucson: University of Arizona Press Miller, G. A. 1951. Language and Communication. New York: McGrawHill Millikan, R. G. 1984. Language, Thought, and Other Biological Categories. Cambridge, MA: MIT Press 1993. White Queen Psychology, and other essays for Alice. Cambridge, MA: MIT Press Moravcsik, J. M. 1990. Thought and Language. London: Routledge Morris, C. 1946. Signs, Language and Behavior. New York: Prentice-Hall Nagel, T. 1979. Mortal Questions. Cambridge University Press Needham, R. 1972. Belief, Language and Experience. University of Chicago Press Neisser, U. 1967. Cognitive Psychology. New York: Appleton-CenturyCrofts 1976. Cognition and Reality. San Francisco: Freeman Ogden, C. K. and Richards, I. A. 1930. The Meaning of Meaning, 3rd edition. New York: Harcourt Brace Jovanovich Ortony, A. (ed.). 1994. Metaphor and Thought, 2nd edition. Cambridge University Press Osherson, D. N. and Lasnik, H. (eds.). 1990. An Invitation to Cognitive Science, Volume 1: Language. Cambridge, MA: MIT Press Osherson, D. N. and Smith, E. E. (eds.). 1990. An Invitation to Cognitive Science, Volume 3: Thinking. Cambridge, MA: MIT Press Papineau, D. 1987. Reality and Representation. Oxford: Blackwell Pateman, T. 1987. Language in Mind and Language in Society. Oxford University Press Peacocke, C. A. B. 1986. Thoughts: An Essay on Content. Oxford: Blackwell 1992. A Study of Concepts. Cambridge, MA: MIT Press Peirce, C. S. 1931-5. Collected Papers. Cambridge, MA: Harvard University Press Pessin, A. and Goldberg, S. (eds.). 1996. The Twin Earth Chronicles: Twenty Years of Reflection on Hilary Putnam's 'The Meaning of Meaning'. New York: M. E. Sharpe 241
Bibliography Pettit, P. and McDowell, J. (eds.). 1986. Subject, Thought, and Context. Oxford: Clarendon Press Piaget, J. 1926. The Language and Thought of the Child. London: Routledge and Kegan Paul Pinker, S. 1994. The Language Instinct: How the Mind Creates Language. London: Penguin Polanyi, M. 1958. Personal Knowledge: Towards a Post-Critical Philosophy. London: Routledge and Kegan Paul Popper, K. R. 1972. Objective Knowledge, An Evolutionary Approach. Oxford University Press Premack, D. 1986. Gavagai! Or, the Future History of the Animal Language Controversy. Cambridge, MA: MIT Press Price, H. H. 1953. Thinking and Experience. London: Hutchinson Prior, A. N. 1971. Objects of Thought. Oxford: Clarendon Press Putnam, H. 1975. Mind, Language and Reality: Philosophical Papers, Volume 2. Cambridge University Press 1988. Representation and Reality. Cambridge, MA: MIT Press Quine, W. V. 1960. Word and Object. Cambridge, MA: MIT Press 1969. Ontological Relativity and Other Essays. New York: Columbia University Press 1974. The Roots of Reference. LaSalle, IL: Open Court 1992. Pursuit of Truth, revised edition. Cambridge, MA: Harvard University Press 1995. From Stimulus to Science. Cambridge, MA: Harvard University Press Rorty, R. (ed.). 1967. The Linguistic Turn: Recent Essays in Philosophic Method. University of Chicago Press Rosenberg, J. 1974. Linguistic Representation. Dordrecht: D. Reidel Russell, B. A. W. 1912. The Problems of Philosophy. Oxford University Press 1921. The Analysis of Mind. London: Allen & Unwin 1956. Logic and Knowledge. London: Allen & Unwin Ryle, G. 1949. The Concept of Mind. London: Hutchinson 1971. Collected Papers, Volume 2. London: Hutchinson 1979. On Thinking. Oxford: Blackwell Salmon, N. U. and Soames, S. (eds.). 1989. Propositions and Attitudes. Oxford University Press Schiffer, S. 1972. Meaning. Oxford: Clarendon Press 1987. Remnants of Meaning. Cambridge, MA: MIT Press Searle, J. R. 1979. Expression and Meaning. Cambridge University Press 1983. Intentionality: An Essay in the Philosophy of Mind. Cambridge University Press 1992. The Rediscovery of the Mind. Cambridge, MA: MIT Press Searle, J. R. (ed.). 1971. The Philosophy of Language. Oxford University Press Sheets-Johnstone, M. 1990. The Roots of Thinking. Philadelphia: Temple University Press Skinner, B. F. 1957. Verbal Behavior. New York: Appleton-CenturyCrofts 242
Bibliography Stalnaker, R. C. 1984. Inquiry. Cambridge, MA: MIT Press Sterelny, K. 1990. The Representational Theory of Mind: An Introduction. Oxford: Blackwell Stich, S. P. 1983. From Folk Psychology to Cognitive Science: The Case Against Belief. Cambridge, MA: MIT Press Strawson, G. 1994. Mental Reality. Cambridge, MA: MIT Press Strawson, P. F. 1959. Individuals: An Essay in Descriptive Metaphysics. London: Methuen 1971. hogico-Linguistic Papers. London: Methuen Tarski, A. 1956. Logic, Semantics, Metamathematics. Oxford: Clarendon Press Tugendhat, E. 1982. Traditional and Analytical Philosophy. Cambridge University Press Vendler, Z. 1972. Res Cogitans. Ithaca, NY: Cornell University Press Vygotsky, L. 1934. Thought and Language. English translation: Cambridge, MA: MIT Press, 1986 Watson, J. B. 1914. Behavior: An Introduction to Comparative Psychology. New York: Henry Holt 1919. Psychology from the Standpoint of a Behaviorist. Philadelphia, PA: J. B. Lippincott 1930. Behaviorism. New York: Norton Weiskrantz, L. (ed.). 1985. Animal Intelligence. Oxford: Clarendon Press 1988. Thought Without Language. Oxford University Press White, A. R. 1967. The Philosophy of Mind. Oxford University Press Whorf, B. L. 1956. Language, Thought, and Reality. Cambridge, MA: MIT Press Wittgenstein, L. 1922. Tractatus Logico-Philosophicus. London: Routledge and Kegan Paul 1953. Philosophical Investigations. Oxford: Blackwell 1958. The Blue and Brown Books. Oxford: Blackwell 1967. Zettel. Oxford: Blackwell 1980a. Remarks on the Philosophy of Psychology, Volume I. Oxford: Blackwell 1980b. Remarks on the Philosophy of Psychology, Volume II. Oxford: Blackwell 1982. Last Writings on the Philosophy of Psychology, Volume I. Oxford: Blackwell 1992. Last Writings on the Philosophy of Psychology, Volume II. Oxford: Blackwell Woodfield, A. (ed.). 1982. Thought and Object: Essays on Intentionality. Oxford: Clarendon Press Wright, C. 1991. Realism, Meaning and Truth, 2nd edition. Oxford: Blackwell
243
Index accusative(s), intentional- 4 object-, 4 act(s) of thinking, 3, 4, 37-8, 47, 92 agnosia, 179 Alberti, L. B. 65, 75 amnesia, 142-3, 145, 146 analytic/synthetic distinction, 6, 9, 199-205,217 anamorphoses, 71-3 animism, 124 Anscombe, G. E. M. 55n, 75 anthropology, 6 aphasia, 138-40, 142, 146, 147 argument from the poverty of the stimulus, 21 Aristotle, Aristotelianism, 10, 153,154, 155, 164, 168 Arnauld, A. 37 ars combinatoria, 2 artificial intelligence, 6 , 1 1 , 2067, 209, 232
Augustine, St. 51, 155 Austin, J. L. 51,219 autism, 179, 190 awareness, see consciousness Bach,K. 80n, 101 Bacon, F. 43&n, 45, 49 Beardsley, M.C. 52-3, 69, 73, 75 behaviourism, 10, 162, 224, 226 belief(s), 7, 8, 11-12, 22-3, 25, 26,78,80-6,93, 110, 113, 114-5, 167,178-80, 183, 184, 185, 188,191,192,193,199, 209 Belnap, N. D. 206, 207, 209, 210,211,214,217 Bergson,H.15 Berkeley, G. 58-9, 75, 162 biology, 8, 96, 97, 103, 123, 222, 225,233 Black, M. 98n, 101 blindsight, 10, 120, 142, 143, 145,146, 179
Block, N. J. 222, 234 Bogdan, R. J. 181n, 183, 193, 194 Boole, G. 153, 168 brain, the, 2, 8, 19,20,34-5,41, 44,79,93, 103-6, 108, 117-8, 122-3, 125, 138, 139, 141, 142, 146,147,192,219,220,222, 223, 225,227,232-4 Bromberger, S. 206n, 207, 208, 211,212-3,217 Budd, M. 56n, 75 Burge, T. 77-82, 86, 91,101 Carnap,R. 9, 200-205, 217 Carroll, L. 121, 126 causal reality constraint, 104, 108, 110-2, 114, 116, 120, 123, 124 Chomsky, N. 6, 19-20, 27, 42&n, 49, 223 Churchland, P. M. 186n, 194 Churchland, P. S. 122-3, 126, 186n,194 Clark, A. 227, 230, 234 classification, 25, 83-4, 91, 94 cogito, the, 36&n cognitive science; cognitive theorising, 6-7, 9, 13,20,79, 103126 commissurotomy (split-brain) patients, 141, 143, 179 communication, 5, 16, 26, 29, 173,177-9, 187, 189, 194 computers; computer science, 2, 7,8, 105-6, 116-9, 124, 177, 220-2 computationalism, 2, 7, 11, 13, 104-7, 111, 114, 116-25, 2201,230,233
conceptions, 83-7, 91-4, 100-1 concepts, 12, 15-16,21,25,38, 78-100,109,162,178,179,230 conceptual schemes, 10, 15-16, 20,155 245
Index
consciousness; awareness, 7, 9, 12,35,38,45,46, 105, 109, 120, 124, 141-7, 156, 167-8, 178,183,184,185,188,191, 192,193,194,197,198,215, 221,222,233-4 'linear' nature of, 167-8 'content' of thoughts and utterances, 3, 12-13, 16, 21-5, 27, 38,77-86,92-3,98,108-110, 114,115,178,180,197,206, 233 Cottingham, J. 2, 32n, 33n, 37n, 40n, 41n, 42n, 46n,49 Crane, T. 80n, 101 Darwin, C. 125,171 Davidson, D. 8, 10, 11, 12, 13, 22, 27, 30, 98n, 99, 101, 127, 148, 193, 194 Davies, M. 122-3, 126 Dawkins, R. 224, 234 Dennett, D. C. 8, 9, 10, 35&n, 49, 178,188, 194, 222, 223n, 227, 234&n Derrida, J. 152, 168 Descartes, R. 2, 7, 13, 29-50, 53, 54&n, 56-8, 61,69, 74, 75, 197 Dewey, J. 197, 210&n, 213, 217 dispositions, 17, 24-5, 165n Dretske, F.I. 227, 235 dualism, 33,44,45,46, 194 Dummett, M. A. E. 3, 127, 148, 151-2, 153n, 160, 163n, 166, 167,168 Einstein, A. 147 Eliot, T. S. 29&n, 50 error, 26-7 Evans, G. 94n, 101, 151 evolution, 21, 147 experiments, 9, 128-41 explanation, 7, 32-3, 44-5, 46, 103-5, 108-112,114, 116, 1205, 164-5, 191,193,201 externalism, 12, 77-102 Fodor, J. A. 6-8, 13, 19, 80n, 101,165,219,222,235 246
Fellesdal, D. 77n, 98n, 101 Frege, G. 3, 6, 13,109, 151, 153, 154, 157,158, 160, 166, 168 Freud, S. 45n Gadamer, H-G. 152, 169 Gage, J. 60n, 75 gap of voluntary action, the, 110 Gardner, S. 45n, 50 Gellner, E. 151,169 generalisation, 16, 26 Glock, H-J. 3,8,9, 12, 13, 153, 154n,157n, 165n, 169 Goethe, J. W. 223&n Gombrich, E. 53, 58, 60n, 68, 72n,75 Goodman, N. 5, 52-4, 56, 58-68, 70,73,75 Greenberg, J. 97, 101 Grice, H. P. 201,217 Hacker, P.M. S. 9, 13, 58n, 75, 163,169 Hacking, I. 160, 169, 186, 195 Harrah, D. 205, 217 Hart, H. L. A. 151 Haugeland, J. 232, 235 Heidegger, M. 151-2, 156, 169 hermeneutics, 152, 167 Hertz, H. 156, 157, 169 Hobbes, T. 38 holism, epistemological, 201, 203 homunculi, 35, 116-7, 119-20 Hookway, C. 9,12, 204, 213n, 217 Hospers, J. 52-3, 60-2, 69, 75 Humphrey, N. 189, 195 Husserl, E., 3, 151, 156, 160, 161, 162, 166,169 Hyman, J. 4, 33n, 50, 55n, 75 immaterialism, 2, 44 incommensurability, 16 indeterminacy thesis, 12, 77n ineffability, 1 innocent eye, the myth of, 58-9, 60n,68 intelligence, 177-8, 183-4, 185, 194,197, 199
Index
intensionality, 25, 109 intentionality, 4, 7-8, 108, 110-1, 113,115-20,124,125,178, 192-3 intrinsic/derived/as-if, 113, 11520, 123, 124 internalism, 12, 13, 81-2, 163-4 Jackendoff, R. 222n, 235 James, W. 10, 13 Jolley, N. 39n, 50 judgement, theories of, 2-6, 25, 92 Kant, 1.4, 15,20, 152, 154-62, 166, 167 Karmiloff-Smith, A. 227, 234 Keynes, J.M. 219 kinds, natural, 183&n, 191 Kohler, W. 132, 133, 137,224, 235 Korzybski, A. 176
Locke, J. 5, 39, 161 Locke, J. L. 18,27 logical positivism, 159 Lucretius, 219 McDowell, J. 22n, 27 McGinn, C. 7n, 13, 151 Mackie, J. L. 161, 169 Malebranche, N. 35, 39, 46, 50 Marr, D. 105-8, 110, 114, 124, 125,126 materialism, 2, 45 Mead, G. H. 10, 220, 235 mentalese, 7, 19-20, 166, 220-2, 232 mental states, 5, 79, 113, 178, 197-8 Mersenne, M. 39, 43n mind, representational theory of, 8, 13 modules, mental, 7 modus ponens, 120-2
Langer, S. K. 60 language, as a medium for thought, 1,9, 15-20, 152, 161, 165 as clothing thought, 3, 6, 20, 29 inner, 1, 19 -learning, 10, 18-26, 77, 80, 945,141,201,223,228 of thought, 6-8, 10, 13, 19-20, 127, 165-6,220,222,231 philosophy of, 4 without thought, 141 lawsofthought,152, 153, 156 Leibniz, G. W. von 2, 13,35 Libet, B. 193, 195 lingualism; linguisticism, 1-13, 30-1, 160-6 linguistic relativity, see SapirWhorf hypothesis linguistic universals, 19 linguistics, 1, 6, 18-20, 80, 96101, 152,209,214,216,219, 228 linguistic turn, 4, 152, 159-60 Lipton, P. 206n, 217 Loar, B. 81-3, 85, 101
Montaigne, M. de 31, 40&n, 41, 50 Moore, G. E. 151 More, H. 41-2 Morris, C. 60-1, 61n, 76 Morton, A. 188n, 195 Nagel, T. 151 naturalism, 2, 124 Needham, R. 183, 195 Newcastle, Marquis of, 40&n, 42n Newton-Smith, W. H. 182n, 195 nominalism, 175 non-language-users, 11, 12, 13, 25,26-7, 30-1,40-2,48,77, 104-6, 114, 119, 125,128-48, 163n,177-95,223-6 norms, 9, 12,24-5,77,79,83, 86-7, 89-93, 96-7, 100, 109, 153n, 154, 197-200,204-6, 215 objects of thought, 3, 4 observation categoricals, 174, 176 247
Index
observation sentences, 173-4, 176 observer-relativity/-independence,7, 111-20,123, 124, 125 occasion sentences, 173 ostensive teaching, 24, 25-6, 94 other minds, problem of, 36 Palmer, S. E. 108,126 Patterson, S. 83n, 101 Peacocke, C. 151 Peirce, C. S. 60 perception, 5, 10, 18-20, 22-3, 27, 33-5, 54n, 57, 58-9, 68, 99,
103,112,129,130,143,171-3, 221 'perceptual sentences', 23-4 phenomenology, 3, 151, 161, 219,221 philosophy, nature of, 3, 6 Piaget, J. 131-2, 149 pictures, 4-6, 18, 38, 51-76, 219 internal and external subjects of, 55-6, 56n, 70-1 occlusion shape, 69-75 occlusion size, 73 Pinker, S. 19-20, 27, 95, 101 Plato, 1,2,37,57,61,91, 155, 163,223 Platonism, 3, 39, 41, 153-60, 166 Pliny, 75 Popper, K. R. 197, 198, 217 pragmatism, 10, 197 Premack, D. 137, 149, 185, 195, 231,235 Preston, J. 165n, 169 presuppositions, of questions, 208,212 privacy, of thought, 29-32, 3540,42,45,46,47,49, 153 problem-solving, 8-11, 13, 164, 177, 194,197-206,210-13, 217,232 propositions, 4, 5, 7, 8, 80, 92, 114-5, 121, 153, 154,157,158, 159, 179, 197,199,205-9,213, 214-6, 220, 221 propositional attitudes, 4, 7, 8, 248
11, 12-13,25,80, 114-5, 185, 197 propositional verbs, 4, 8 prosopagnosia, 143-4, 145, 146 psychologism, 3, 39n, 154, 157, 158,166 psychology, 1, 5,6,7,8,9, 10, 12, 13,45,69, 79, 80, 81, 83n, 86,97,99, 112, 116, 120, 123, 127-50, 152,167, 185, 193, 194,206, 219, 226, 228 cognitive, 2, 6, 112 common sense, 180-8, 190-4 'philosophical', 184&n, 185-6, 188,191, 192, 194 philosophy of, 184 Putnam, H. 93n, 101 quantification, 174-5, 176 questions, 9, 193, 199-200, 20517,232 Quine, W.V. 6, 9, 10,12,20, 77n, 98n, 167, 169,200-205, 217-8 Rawls, J. 151 realism, 186-8 reductionism, 32&n reification, 175-6, 188, 191 relativism, 12, 16-17 representation(s), 4, 107-8, 114, 119, 122,127, 130, 146, 15262,166,184,185,191,206, 212,213,220,224,226,227, 228-9 linguistic, 4, 8, 17-18, 91-5, 100, 152, 198,231 mental (psychological), 5, 7-8, 9, 11,38,84,91-5, 104-5, 114, 115, 152, 156,197,211,220 meta-, 192, 194, 198 pictorial, 4, 51-76 picture or model theory of, 5, 158 re-, 227 resemblance theory of, 4-5, 5264, 73, 74 semiotic theory of, 5
Index
Robinson, D. 122 Rosenthal.D. 141-2, 149 rule-following, 7, 83, 103, 105-6, 108-114, 120-5 Rundle, B. 127, 149 Ruskin, J. 60n Russell, B. A. W. 4, 5, 7, 13,58, 94n, 151,152,153-4, 157, 158, 160, 169,224,235 Ryle, G. 29, 54&n, 59n, 76, 91, 101, 152, 166,169 Sapir, E. 80, 96-9, 101 Sapir-Whorf hypothesis, 12, 7980, 96-9 Sartre, J-P. 234, 235 Savage-Rumbaugh, S. 137 scepticism, 18, 36, 161 Schiffer, S. 7 Schopenhauer, A. 156, 157, 160 Scruton, R. 62n, 76 Searle, J. R. 4, 7, 8, 12, 105, 120, 126,151, 185, 188 Sejnowski, T. 122-3, 126 self, the, 10, 37,47, 192 sensations, 22, 32n, 33, 41, 45 'sense' of sentences, 3 Skinner, B.F. 10, 133,219, 220, 235 sociology, 6 Socrates, 1, 57-8, 223 soliloquy, silent, 8, 10, 93, 163, 165, 177,220 soul(s), 2, 30, 32-3, 35,41,45, 46,53 Spelke, E. 96, 102, 129-30, 148, 149 standing sentences, 174 Steel, T. 206, 207, 209, 210, 211, 214,217 Strawson, P.F. 201,217 subjective/objective contrast, 11 'third realm', (Frege) 3, 154, 166 'tip-of-the-tongue' phenomenon, 8n,20,165,221
thought, parallels with language, 1 philosophy of, 4, 151-2, 167 without language, 9, 11-12, 12748,219-23,231 thoughts, constituents of, 5-6 Tractatus Logico-Philosophicus (Wittgenstein), 5, 6, 8, 152, 156-60, 166 translation, 10, 16, 20, 165-6, 222 transparent access to one's own mind, 45-6 triangle, of learning, 26-7 truth, 17, 21,24, 26-7, 88, 153, 158,179, 188 Tugendhat, E. 160, 162, 169 Turing, A. M. 118 unilateral neglect, 143 universa mathesis, 39 Van Fraassen, B. C. 206n, 218 vestibular ocular reflex, 120, 122-3 Vienna Circle, 159 Vygotsky, L. 13,220,235 Watson, J. B. 10, 14 Weiskrantz, L. 9-10, 11, 128, 135, 138,142,143,150,177 White, A. R. 4, 14, 168, 169 Whorf, B. L. 80, 96-102 Wilkes, K. V. 4, 11, 12, 195 Wilson, M. D. 45n, 50 Wind, E. 74n, 76 Wittgenstein, L. 4, 5-6, 8-9, 12, 14, 24, 29, 37, 40, 75n, 76, 127, 150,152,156-60,162-3, 166, 167, 169,219,235 Wollheim, R. 62n, 66-7, 76 Woodfield, A. 12, 81n, 84n, 102 Wundt, W. 181n, 195 Young, J.Z. 132
249