d. m. armstrong belief, truth and knowledge 1973

7/22/2019 D. M. Armstrong Belief, Truth and Knowledge 1973

1/238

Belief, Truth and Knowledge


2/238


3/238

BeliefTruthandKnowledgeD. M. ArmstrongChallis Professor of Philosophy

University of Sydney

C A M B R I D G E U N I V E R S I T Y P R E S SCAMBRIDGE

LOND ON NEW YORK NEW ROCHELLEMELBOURNE SYDNEY


4/238

CAMBRIDGE UNIVERSITY PRESSCambridge, New York, Melbourne, Madrid, Cape Town, Singapore, Sao Paulo, Delhi

Cambridge University PressThe Edinburgh Building, Cambridge CB2 8RU, UK

Published in the United States of America by Cambridge University Press, New Yorkwww.cambridge.org

Information on this t i t le: www.cambridge.org/9780521087063 Cambridge University Press 1973

This publication is in copyright. Subject to statutory exceptionand to the provisions of relevant collective licensing agreements,

no reproduction of any part may take place without the writtenpermission of Cambridge University Press.

First published 1973Reprinted 1974, 1981

Re-issued in this digitally printed version 2008A catalogue record for this publication is available from the British Library

Library of Congress Catalogue Card Num ber: 72-8358 6ISBN 978-0-521-08706-3 hardback

ISBN 978-0-521-09737-6 paperback


5/238

For Madeleine


6/238

AcknowledgementsI am g reatly indebted to Professors M ax Deu tscher, Do uglasGasking, J. J. C. Smart and Mr D. Stove. They read the whole orparts of various drafts of this essay, and assisted me a great dealwith comments and criticisms. Further useful suggestions camefrom R. J. Mynott and the referee for Cambridge University Press.Mr George Molnar helped me with the very difficult topic of thenature of dispositions. I should also acknowledge the help of mygraduate class at Sydney University with whom I worked throughthe first and third part of the book. In particular I am indebted toChristopher Murphy. I should also like to thank Mrs P. Trifonoff,who typed the manuscript.

D .M. A.Sydney UniversityMarch


7/238

ContentsPart I: Belief

The Nature of Belief 3Beliefs as States 7i Three Ways of Conceiving of Beliefs 711 Dispositions are States 11in Differences between Beliefs and Dispositions 16iv Belief and Consciousness 21Belief and Language 2 4I Manifestations and Expressions 24II Belief without Language 25in Sophisticated Belief and Language 31iv Thought and Language 36Propositions 38I The Notion of a Proposition 38II Propositions and Language 42m Tw o Unsatisfactory Accounts of Propositions 43iv An Account of Propositions Sketched 46Concepts and Ideas goI The Distinction between Concepts and Ideas 50II Simple and Complex Concepts and Ideas 54in Th e Self-directedness of Simple Concepts and Ideas 60iv Hum e's Problem 70General Beliefs JJI Having a Reason for Believing Something 77II Efficient and Sustaining Causes 79in Difficulties for the 'Sustaining Cause' Analysis 82iv Principles of Inference 85

vii


8/238

viii ontents

v General Beliefs 89vi Inferring 94VII Reasons and Rationalization 95VIII Good Reasons 96

7 Existential Beliefs 998 Further Considerations about Belief 104

I Contradictory Beliefs 104II Conjunctive Beliefs 106in Degrees of Belief 108

Part II: TruthTruth 113I Correspondence 113II Nominalism 114in Semi-Nominalism 118iv Realisms 119v Predicates and Properties 123vi The Correspondence Relation 130

Part III: KnowledgeI o Knowledge Entails True Belief 137I The Classical Account of Knowledge 137

II The Truth-condition 137in Different Views concerning the Belief-condition 138iv Rejection of the Strong Denial of the Belief-condition 139v Rejection of the Strong Assertion of the Belief-condition

vi Rejection of the Weak Denial of the Belief-condition 143

I 1 TheInfinite Regress of Reasons 10I The Evidence-condition 150II The Infinite Regress of Reasons 152in Different Reactions to the Regress 154


9/238

Contents ix12 Non-Inferential Knowledge ( i ) 162

I W hat are the paradigmsofnon-inferential knowledge?162II The 'Thermometer' viewofnon-inferential knowledge166in Deutscher's objection 171iv A parallel accountofmanifestationsofskill 175v Further objections 178vi Self-fulfilling beliefs 180

13 Non-Inferential Knowledge (2 ) 184I Insupportofour accountofnon-inferential knowledge184II An epistemological objection to our accountofnon-inferential knowledge 190in Restrictions on the scopeof ouraccountof non-inferentialknowledge 192iv Enlargementsofthe scopeofour accountof non-inferentialknowledge 194

14 Inferential and General Knowledge 198I Difficulties about inferential knowledge 198II Non-inferential knowledgeofgeneral propositions 201in Inferential Know ledge 205iv O'Hair 's objection 208v Between inferential andnon-inferential 210

1g Further Considerations about Knowledge 212I Knowledge ofKnowledge 212II KnowledgeofProbabilities 214in Certainty 216iv Metaphysical implicationsofour accountofKnowledge216v Scepticism 217

Conclusion 2 2oWorks referred to illIndex 11g


10/238


11/238

PartI

Belief


12/238


13/238

1The Nature ofBelief

The purpose of this brief chapter is to give a first view of a theoryof the nature of belief, preparatory to working out the theory inlater chapters.In one of his posthumously published 'Last Papers' ('GeneralPropositions and C au sality ', in Ramsey 1931) F . P. Ramsey takesas an example the belief that everybody in Cambridge voted. He saysthat such a belief is 'a map of neighbouring space by which westeer' (p. 238). Here he attributes two characteristics to the belief:it is a map, and it is something by which we steer. Here in minia-ture is the account of belief to be defended in this work.Wittgenstein has little to say directly about belief in the Tractatus.But I suppose that his comparison of sentences to pictures inspiredRamsey's comparison of beliefs to maps.If we think of beliefs as maps, then we can think of the totalityof a man's beliefs at a particular time as a single great map ofwhich the individual beliefs are sub-maps. The great map willembrace all space and all time, past, present and future, togetherwith anything else the believer takes to exist, but it will have asits central reference point the believer's present self. But we mustnot think of the great map as like a modern cartographer's map of

the earth's surface. Such a map is too good a map to be a suitableimage. (It is not just belief, it is knowledge.) The great belief-mapwill be much like the maps of old, containing innumerable errors,fantasies and vast blank spaces. It may even involve contradictoryrepresentations of portions of the world. This great map, which iscontinually being added to and continually being taken away fromas long as the believer lives, is a map within his mind. If the mindcan be (contingently) identified with the brain, as I believe it shouldbe, the map will be literally a map in the believer's head. But thecorrectness or otherwise of this identification will not be at issuein this book.

The belief-map will include a map of the believer's own mind,and even, as a sub-part of this sub-part, a map of the believer's3


14/238

4 Beliefbelief-map (that is, his beliefs that he holds certain beliefs). Butthis entails no vicious infinite regress. If you try to make a completemap of the world and therefore try to include in the map a com-plete map of the map itself, you will be involved in an infiniteseries of maps of maps. But since the belief-map is not a completemap of the world, and since the map of itself that it contains is evenmore incomplete, the situation is no worse than those actual pic-tures which contain, as part of the scene pictured, little picturesof themselves.

In the case of ordinary maps a distinction can be drawn betweenthe map itself, and the map-reader's interpretation of the map. Nosuch distinction can be drawn in the case of beliefs. We do notread off our interpretation of reality from the data supplied by ourbeliefs. Our beliefs are our interpretation of reality. But, despitethis clear difference between beliefs and ordinary maps, the analogy,as I hope will gradually emerge, is still of the greatest value. Beliefsare to be thought of as maps which carry their interpretation ofreality within themselves. Of their own nature, apart from anyconventions of interpretation, they point to the existence of a cer-tain state of affairs (though there may be no such state of affairs).They have an intrinsic power of representation.

We must distinguish between beliefs and mere thoughts: betweenbelieving that the earth is flat and merely entertaining this proposi-tion while either disbelieving it or having no belief one way or theother. Now, if beliefs can be thought of as maps, mere thoughts -the mere entertaining of propositions - seem equally entitled to beconsidered maps (however wild and inaccurate) of the world. Whatmarks off belief-maps from thought-maps? In his Treatise (Book i,Part in, Section 7) Hume asks what marks off believing somethingfrom merely entertaining that thought. He claims to be the firstphilosopher to pose the question. So we may call the problem ofdistinguishing belief-maps from mere thought-maps *H um e's prob-lem' .

Ramsey's formula gives us the solution (which was in some de-gree anticipated by Hume). Beliefs are maps by which we steer.Unlike entertained propositions, beliefs are action-guiding. Enter-tained propositions are like fanciful maps, idly scrawled out. Butbeliefs are maps of the world in the light of which we are preparedto act.The task of the remaining chapters of this Part of the book will


15/238

The Nature of Belief 5be to spell out and articulate in detail Ramsey's suggestion. Thesuggestion is bold and simple. But, as might be expected, its work-ing out is laborious and complex.It might be objected straightaway that this Ramseyan account ofbelief can at best give an account of beliefs concerning things atparticular times and places - beliefs of a historical/geographical sortin that widest sense of 'history' and 'geography' which ranges overall time and all space. (Such beliefs will in future be referred to as'beliefs concerning particular matters of fact'.) But what of thebeliefs that arsenic is poisonous or that every even number is thesum of two primes? How can beliefs in the truth of such unre-stricted universally quantified propositions be represented as mapsof reality?I think that the objection is justified, and that a different accountmust be given of such beliefs. Ramsey himself saw the necessityfor a different account. Indeed, his solution to the problem of whatit is to believe that an unrestricted universally quantified proposi-tion is true is much more widely known than his quickly thrown-out remark about beliefs concerning particular matters of fact. Hesuggested that such 'general beliefs' (as we shall in future call them)were 'habits of inference' which dispose us to move from a beliefabout some particular matter of fact to a further belief about someparticular matter of fact. General beliefs are dispositions to extendthe original belief-map according to certain rules.Ramsey was following C. S. Peirce here. (See, for instance,Peirce's essay 'The Fixation of Belief, reprinted in Peirce 1940.)But Douglas Gasking has pointed out to me that Ramsey wasprobably led to this view by reflecting upon the difficulties ofWittengenstein's Tractatarian view of unrestricted universally quan-tified propositions as infinite conjunctions of particular propositions.Any attempt to give an account of general belief along such lineswould clearly face impossible problems.In Chapter Six a version of this Ramseyan doctrine will be deve-loped, and linked with the notion of a man's holding a belief for acertain reason. It will be found to be applicable to general beliefsin the truth of both necessary and contingent propositions.

A note on symbolism. At a number of points, formulae familiarfrom 'epistemic logic' will be employed. Thus, 'A believes that p'will sometimes be written 'Bap' and 'A knows that p' will be


16/238

6 Beliefwritten ' K a p \ Very often, however, when I use such expressions as' B a p ' and 'Kap' reference will not be being made to some proposi-tion but to some state of affairs or situation: A's believing or know-ing that p. Thus, in the course of the argument it may be said that'Baq is the cause of Bap'. This will mean that A's believing thatq brings it about that A believes that p. Max Deutscher has sug-gested that on such occasions it might be better to write ' sBaq is thecause of sBap' to indicate that it is situations, not propositions, thatare in question. I will occasionally adopt his suggestion. But since Ithink that the context normally makes it clear how the formulae areto be taken, aesthetic reasons plead in favour of omitting such super-scripts wherever possible.


17/238

2Beliefs as States

I T h re e W ays of Co nceiv ing of BeliefsHaving sketched a theory of the nature of belief in the broadestoutline, let us begin detailed investigation by asking under whatgeneral category belief falls. I think that there exist in our philo-sophical tradition three different answers to this question, notalways explicitly spelled out. First, there is the view that beliefsare conscious occurrences in the believer's mind. Second, thatbeliefs are dispositions of the believer. Third, that beliefs are statesof the believer's mind. In this section these three views are set out.

i. Beliefs as Conscious Occurrences. The classical instance ofsuch a theory is Hume's account of belief {Treatise, Book i, Part in,Section 7) as a vivid or lively idea associated with a present impres-sion. The *association w ith a present im pre ssion ' will only fit thoseinductively acquired beliefs concerning particular matters of factwhich Hume is especially interested in at that point of the argumentof the Treatise. Hence we may take his view of belief in generalto be that A's believing that p is equivalent to A's having presentto consciousness a vivid or lively idea of p.Such a view, it is notorious, fails to do justice to the way we

talk and think aboutbelief. For it is perfectly intelligible to attributea belief to somebody although there is no relevant vivid idea in hisconsciousness. We can, for instance, intelligibly attribute a currentbelief that the earth is round to a man who is sleeping dreamlesslyor is unconscious. Hume's vivid ideas may or may not occur, andmay or may not have something to do with a man's beliefs, butasserting the presence of such an idea in a man's consciousnesscannot be what it means to assert that he has the correspondingbelief.The difficulty has nothing to do with the particular form whichHume's theory takes. The difficulty faces any theory which equatesa man's current belief with some current content of his conscious-ness, whether it be a vivid idea of p, an inward motion of assent

7


18/238


19/238

Beliefs as States 9thoughts, mental images or inward motions of assent are, primafacie, possible manifestations of A's belief that p. Hume's 'vividideas' are not beliefs, but, provided there are such things, there isno reason why they should not be manifestations of beliefs. It isawkward to use the word 'manifestation' in connection withinner mental occurrences - an awkwardness which the word 'ex-pression' shares - but it is no more than awkward. There may beother arguments to show that such inner occurrences cannot bemanifestations of belief, but the mere demonstration that beliefsare dispositions would do nothing to support any limitation ofpossible manifestations to outward behaviour.3. Beliefs as States. The dispositional view of belief is certainlymore satisfactory than the 'conscious occurrence' view. There is,however, a third and, I believe, a still more satisfactory way ofthinking about belief which is at least implicit in Western philo-sophical thought. According to this view, A's believing that p isa matter of A's being in a certain continuing state, a state whichendures for the whole time that A holds the belief. In the case ofbeliefs which are acquired, this view thinks of A's belief that p asa matter of A's mind being imprinted or stamped in a certainway. Plato's image in the Theaetetus (191 C-E) is that of theimprint made by a seal on a block of wax, an imprint which thenendures for a greater or lesser time. I would add to this by sayingthat there is no reason why this state should be something whichthe believer is conscious of being in. He may or he may not knowthat he holds a certainbelief.

The notion of a state of an object deserves some consideration.To say that an object is in some state is to attribute a property tothe object. But what sort of property? First, it is a non-relationalproperty of the object. We distinguish a thing's state from itscircumstances. Of course, the state may itself involve relation, thatis to say, the state may be a structural property. Indeed, the conceptof the state of an object is naturally associated with the idea of astructure of, or within, the object. (But I doubt whether the idea ofa structure is part of the concept of a state. The particular tempera-ture of an object is naturally said to be a state of the object. Yetthe concept of temperature does not entail that the hot object isstructured in any way.)But not every non-relational property of an object defines a stateof the object. A thing is a horse in virtue of a conjunction of


20/238

io Beliefnon-relational properties, and a conjunction of properties is a pro-perty. But we could not say that last year's Melbourne Cup winner isin the state of being a horse. Consider Proteus, however, who couldtake on the form of any animal he pleased. If he becomes a horse,is he not in the state of being a horse ?

This suggests that when we speak of states of an object wealways have in mind some classification of the object relative towhich the state is an accidental or changeable feature of the object.Perhaps the object always possesses the feature, but it will be anintelligible supposition that the feature should be lost. If beliefs arestates, then they will be accidental and changeable features ofminds (or, if this is objected to, of persons).

But our account of states is still too broad. We should not wantto say that a man is in the state of running, but this is permitted bywhat has been said so far. How is running to be excluded? I thinkthe answer is that the concept of running is of necessity the conceptof a process: meaning by *process' here som ething whose differentphases are different in nature. {Uniform motion would not be anexample of a process.) But a state need not involve a process. Itmay in fact be a process, but it is not entailed that it is a process.Beliefs might in fact be processes. For instance, if physicalism istrue, beliefs might be reverberating circuits. But, unlike the con-cept of running, it is not part of the concept of belief that belief is,or involves, a process.A counter-instance may be proposed to the argument of theprevious paragraph. We speak of a person or a liquid as being in anagitated state. Yet does not agitation involve processes? I think,

however, that what is meant by 'an agitated state' is that state,whatever it may be, which is responsible for the agitation of theperson or thing. We do draw a distinction, even if a fine one,between a person's being agitated and his being in an agitatedstate. The latter seems to refer to some continuing condition, ofunspecified nature, which produces a good deal of agitated beha-viour. So perhaps the counter-instance fails.So much by way of explicating the notion of a state. AlthoughI think that the view that beliefs are states is to be preferred tothe view that beliefs are dispositions, I think also that the disputeis a very confusing one. For it seems that (i) although not all statesare dispositions, dispositions are a species of state, (ii) There is onespecies of belief, viz. general beliefs, which may be plausibly said


21/238

Beliefs as States nto be dispositions, (iii) In respect of other sorts of belief, despite theundoubted important resemblances between disposition-states andbelief-states, there are significant differences between them. Thesedifferences make it very misleading to say that non-general beliefsare a species of disposition. The next two sections try to substantiatethese three propositions.

II Di spo si tio ns ar e StatesThe argument appears to have seven steps.

1. It seems obvious that for every true contingent propositionthere must be something in the world (in the largest sense of'something') which makes the proposition true. For consider anytrue contingent proposition and imagine that it is false. We mustautomatically imagine some difference in the world. Notice that itis not being argued here that for every different true contingentproposition there is adifferent something in the world which makesthat proposition true. (I think the latter doctrine is in fact demon-strably false. See Part II.)

2. As a corollary of i, where a predicate ' F ' is not applicableto an object, a, up to the time t but is applicable to that objectafter t, it must be the case that a has changed in some way at t.

Dr D. H. Mellor has objected (in discussion) that this corollaryis falsified by a case where a man is not a hundred years old up tot, but becomes a hundred years old at t. Previously the predicate4a hundred years old' was not applicable to him, now it is. Yetthe man himself need not have changed in any way. This objec-tion enables me to make clear how small a claim it is that is beingmade at this stage of the argument. For one of the man'srelationalproperties viz. his relation to his birth date) has changed. And forthe purpose of this argument a change in a relational property is achange in the object.

3. It follows immediately that if a dispositional predicate is notapplicable to an object up to t but is applicable to the object aftert then there must have been a change in the object at t. For instance,if a piece of glass cannot truly be said to be brittle up to t butcan be truly said to be brittle after t then the object must havechanged at t.

4. It will now be argued that the change must be a change inthe non-relational properties of the disposed object. A disposition


22/238

12 Beliefentails the presence (or absence) of non-relational properties of theobject.

Consider an occasion where a disposition is manifested. A brittlepiece of glass is struck and, as a result, it breaks. In any causalsequence the nature of the effect depends upon three things: thenature of the cause; the nature of the circumstances it operates in;the nature of the thing the cause acts upon. The glass breaks becauseit is struck, it is not carefully packed around with protectivematerial, and it is brittle. Cause of a certain natu re + circumstancesof a certain na ture + disposition = effect of a certain na tu re . N owa disposition is something which the disposed thing retains in theabsence both of a suitable initiating cause and of suitable circum-stances for the cause to operate in. A brittle piece of glass is stillbrittle, even although it is not struck and is so packed around withprotective material that striking would not cause breaking. Butthe presence or absence of the initiating cause, and the presence orabsence of suitable circumstances for its operation, are the onlyrelational properties of the piece of glass which are relevant to itsbreaking or not breaking. The possession of the disposition musttherefore depend upon non-relationalproperties of the glass.

This argument might be challenged by pointing to the logicalpossibility of surrounding circumstances of the piece of glass, cir-cumstances which formed a nomically indissoluble unity with theglass, and which also played an essential causal role in the glass'sbreaking when struck. (I would like to give an example, but thesituation is so weird that I can think of no case that has the slightestintuitive appeal.) But if a piece of glass did stand in this extra-ordinary relation to its surrounding circumstances, then I think itcould be said that the two would form a unit which we wouldnaturally treat as a single thing. And then the disposition would benaturally redescribed as a disposition of this larger thing. Hence Ido not think that we would allow such a case as a counter-instance.

5. So a disposition entails the presence (or absence) of non-relational properties of the object. It will now be argued that adisposition entails that the disposed object is in a certain state.(It will be recalled that, altho ug h states o objects are non-relational properties of objects, not all non-relational properties ofobjects are states of these objects. It was argued that the specialmarks of states are these, (i) If an object of a certain sort is in astate it is always intelligible, at least, that it should cease to be in


23/238

Beliefs as States 13that state while remaining an object of that sort. This entails thata state is always a state relative to some prior classification of thething, (ii) Although in fact states may involve a process, such as areverberating circuit, the concept of the state never entails theexistence of such processes.)

It seems clear that the presence (or absence) of non-relationalproperties entailed by possession of a disposition may properly besaid to be a state of the object. For it is always intelligible to supposethat the thing which is brittle or elastic, etc. should cease to bebrittle or elastic and yet still be the same sort of thing (still glass,rubber, a solid, etc). So the first condition for a state is satisfied.And since dispositional concepts leave us in ignorance concerningthe properties of the disposed object which give it that disposition,it follows that attribution of dispositions does not entail that someprocess takes place in the object. So the second condition for a stateis satisfied.

6. The argument so far has purported to establish that attribu-tion of a disposition to an object entails that the object is in acertain state. What is the concrete nature of this state? Noapodeictic answer to this question is possible. But I suggest thatthe most plausible general answer is: whatever state of the objectscientists find to be responsible for manifestation of the dispositionwhen a suitable initiating cause acts upon the object. In the caseof brittleness, for instance, the state will be a certain sort of bond-ing of the m olecules of the brittle object.However, it is important to realize that accepting this step in theargument leaves a deep question about dispositions unanswered,a question which our investigation will leave unanswered. Thequestion may be illustrated by referring again to the state of mole-cular bonding which makes a brittle object brittle. To talk ofmolecular bonding is surely to talk again in terms of dispositions ofthe bonded things. If our argument has been correct, the attribu-tion of this new disposition will entail that the bonded things arein some state. When the description of this state is known, it mayinvolve further dispositions. Where will this process end? Primafacie, it seems that it might end in one of two ways or, perhaps,not end at all. First, we might come in the end to properties of thedisposed thing which involve no element of dispositionality. Theywill be the ultimate properties on which truthful attributions ofdispositions rest. Second, we might reach ultimate potentialities


24/238

14 Beliefof the disposed thing, potentialities which do not depend upon non-dispositional properties. To adopt such a solution would involveaccepting an ultimate ontological division among non-relationalproperties into potentialities and non-potentialities. Finally, thereis the possibility that the process goes to infinity, dispositions restingupon states which involve further dispositions which involve furtherstates . . . The question whether all, or only one or two, of thesealternatives are genuine possibilities, it is our good fortune to beable to ignore in this essay. But it is important to realize that theargument, at least, has not foreclosed any of these options.

7. The argument so far, if successful, only shows that attribu-tion of a disposition entails that the object so disposed is in a cer-tain state. The state is such that a triggering cause of a suitablenature acting upon the object in that state brings it about (in suit-able circumstances, at least) that the disposition is manifested.It will now be argued further that it is linguistically proper toidentify the disposition w ith this state of the disposed object. It islinguistically proper, for instance, to say that brittleness is a certainsort of bonding of the molecules of the brittle object. The ground

for saying this is simply that scientists and others often speak inthis way, and there seems to be no objection to such speech. (It isnot argued that wehave to speak in this way.)The propriety of this identification has been challenged by RogerSquires (Squires 1968). He argues in the following way. Supposethat a disposition is identified with some state of the disposed thing.Suppose also that, at a certain time, the disposition is unmanifested.Must not the currently inactive state have a disposition to bring

about the appropriate manifestation in appropriate conditions? Butthen, by parity of argument, this new disposition will have to beidentified with some further state of the object, and soad infinitum.Thus , Squires argues, the attribution of a disposition will commitus a priori to postulating an infinite number of states in the objectwhich has the disposition. He rightly regards this conclusion asunacceptable.I do not think that there is a vicious regress here. The first point

to notice is that dispositions are ordinarily attributed to things. H owcan a state, which is a species of property, have a disposition? How-ever, it seems that Squires can reconstruct his argument. Suppose,again, that a thing's disposition is a state of the thing, but thedisposition is not manifested. The thing which is in that state will


25/238

Beliefs as States 15have a disposition to display the appropriate manifestation in appro-priate conditions. Squires can argue that this new disposition willhave to be identified with some further state of the object, and soadinfinitum.

But have we really got a new disposition requiring a new state?A piece of glass is brittle because it has molecular bonding M.The piece of glass having the M-type molecular bonding is dis-posed to break if it is hit. But is this not the disposition of brittle-ness all over again? And can we not say that the state of the glasswhich this disposition should be identified w ith is . . . molecularbonding M? We can allow Squires that there is a regress. But itappears to be virtuous, not vicious. It is like the expression: 'if p,then it is true that p, and then it is true that is true that p . . . andso on'. I conclude that the actual identification of dispositions withstates of the disposed thing is perfectly in order.

It must be confessed, however, that while the identification canlegitimately be made, there seems to be no way of arguing that ithas to be made. While allowing that attribution of dispositionsentails attribution of states of a certain sort to the objects which havethe dispositions, nevertheless we can, if we want to, still make averbal distinction between the disposition and the state. (A verbaldistinction that cuts no ontological ice.) But equally there seems tobe no linguistic objection to the identification, and I think that,very often at least, it is the natural way of talking. It is the way oftalking which will be adopted in this work.Our seven-stage argument is now concluded. If dispositions arestates of the disposed object, they are marked off from (many) otherstates by the way they are identified. When we speak of the brittle-ness of an object we are identifying a state of the object by referenceto what the thing which is in that state is capable of bringing about(in conjunction with some active, triggering cause), instead ofidentifying the state by its intrinsic nature. And this in turn is con-nected with the role that dispositional concepts play in our thinking.We introduce such a concept where, for example, it is found that anobject of a certain sort, acted upon in a certain way, behaves in

certain further ways of a relatively unu sua l sort. W e assign respon-sibility for this behaviour to some relatively unusual state of theobject. But since we norm ally do not kn ow , prior to painful andextensive scientific investigation, what the nature of the state is, wenam e it from its effects. W e thus expose ourselves to M oliere's


26/238

16 Beliefridicule, and, if w e did n othing further, w e would deserve it. But wehave set up the formal structure of an explanation which laterresearch may turn into a genuinely helpful one. Dispositions, in fact,are primitive theoretical concepts: concepts of states specified bywhat the thing in that state can effect. (I assume, of course, aRealistic view of theoretical concepts.) Successful scientific investi-gation leads to a contingent identification of the nature of thesestates.

Ill Differences betw ee n Beliefs and D ispositionsIf the argument of the previous section has been correct, it is wrongto oppose the view that beliefs are dispositions to the view that theyare states. A belief might be a disposition, and yet be an imprintingon the mind, or some other state, for all that. But, as will now bedemonstrated, there are some important, even if not overwhelming,differences between dispositions and beliefs (with the exception ofgeneral beliefs).

(i) One point of distinction between dispositions such as brittle-ness, and beliefs, is that the concept of the former involves the notionof an initiating cause of a certain sort which triggers off the mani-festation. The brittle glass is brittle because it breaks when hit. Apiece of sugar is soluble because it dissolves when placed in water.But the concept of beliefs seems to involve no notion of a class ofinitiating causes which in turn bring about the manifestation orexpression of the belief. No doubt initiating causes will always bepresent when the belief is manifested. But they play no special rolein the concept ofbelief.Chomsky has frequently called attention to the stimulus-indepen-dent nature of speech-acts (see, for instance, Chomsky 1968, p. 11).When a speaker produces a grammatical and meaningful sentencethere is in general no characteristic external stimulus which hascaused him to produce this sentence. What speakers say is ratherdetermined by their interests and purposes. The manifestations orexpressions of beliefs are similarly stimulus-independent. But themanifestations of dispositions like brittleness are stimulus-dependent.General beliefs are a plausible exception. Consider A's belief thatarsenic is poisonous. This can plausibly be treated as a disposition ofA's. T he initiating cause of the man ifestation of the disposition isA's coming to believe that some portion of stuff is arsenic. This


27/238

Beliefs as States 17belief about a particular matter of fact triggers off a certain mani-festation: A's acquiring the further belief that this stuff is poison-ous. (A full treatment of general beliefs is reserved for ChapterSix.)

(ii) If brittleness is manifested, it can be manifested in only onesort of way: by the brittle object breaking if struck. But there is noone such way that a belief that the earth is flat must manifest itself,if it does manifest itself. For instance, the manifestations need nottake the form of outer or inner assent.Ryle, at least, was well aware of this difference, and tried to meet

the difficulty by distinguishing between *single-track' and 'm any -track' dispositions (Ryle 1949, pp. 43~5)- Brittleness, he said, is asingle-track disposition, and so is manifested, if it is manifested, inonly one way. But Smith's belief that the earth is flat can haveindefinitely m any m anifestations. It is a 'ma ny -tr ac k' disposition.Ryle even allowed that the set of different sorts of possible mani-festion mig ht be an infinite on e: an 'infinite- track ' disposition,apparently.But Ryle's distinction still has not done justice to the difference.Anyone who understands the term 'brittle' understands what themanifestation of brittleness is: breaking when struck. But, at thesame time, the notion of 'breaking when struck' can be understoodwithout making any reference to the notion of brittleness. Thenotion of brittleness can be introduced as that state of an objectwhich is responsible for the manifestation of breaking when struck.By contrast, the characteristic manifestations of a belief can onlybe identified as manifestations of the belief by reference back to the

belief.To illustrate. If A believes that the earth is flat and is an Anglo-Saxon he may well manifest his belief, on a particular occasion, byuttering the English sentence 'the earth is flat'. Such a manifesta-tion must surely figure in any list of possible manifestations of A'sbelief. But what makes it a manifestation of A's belief? Only thefact that the rules of English are such that uttering these phonemeswo uld be a natural wayof expressing such abelief.

The point may be put thus. If we take the, perhaps infinite, setof possible sorts of manifestation or expression of a belief that p,the only unifying factor we can discover in the set is that theymight all spring from, be manifestations or expressions of, the onebelief. And this, I think, shows that even if we do not think of


28/238

18 Beliefdispositions as actual states of the disposed thing (though I haveargued that we should think of them as states), we must think ofbeliefs as actual states of the believer. For it is only if the believeris in a state which, in suitable circumstances, might give rise to,that is cause, all these manifestations, that we can understand whatholds them together as a class. The situation is something like thatin a house where a window has been opened, a glass of whiskydrunk, cigarette ash dropped and some money taken. In themselvesthere is nothing which unifies this heterogeneous collection of events.But a unifying principle is at once supplied if we suppose that theywere all caused by an unauthorized intruder.

The distinction between the belief which is not being manifested,and the belief which is, then becomes the distinction between acausally quiescent state, and the same state causally active. (A use-ful analogy is the 'information' inactive in the memory-banks of acomputer, and the same 'information' currently playing a causalrole in the computing process, and so in bringing about the com-puter's 'print-out ' .)Once again, general beliefs are a plausible exception to what hasbeen said about this distinction between beliefs and dispositions.It is plausible to say that A believes that arsenic is poisonous if,and only if, acquiring the belief that a certain portion of stuff isarsenic brings it about, in normal circumstances at least, that Aacquires the further belief that that portion of stuff is poisonous.There is a single, logically central, manifestation of A's belief.(iii) In the case of beliefs, as opposed to dispositions like brittle-ness, it seems that the states involved must have a certain internal

structure. Suppose that A believes (i) the cat is on the mat; (ii) thecat is asleep; (iii) the cat is black. These three beliefs, although alldifferent, involve a common element. Now if we take beliefs to bestates of the believer, must we not take it that these states have aninternal structure such that to common elements in the thingsbelieved correspond common elements in the state which is thebelief? W e must, of course, distinguish betw een the believing andwhat is believed: between A's believing that p which is a state ofA, and the proposition ' p ' wh ich is believed. B ut m ust not theinternal structure of the belief-state reflect the structure of the pro-position believed? How otherwise could beliefs with different con-tent give rise to different manifestations or expressions? Comparethis with the case of brittleness. W e are (I have arg ue d) logically


29/238

Beliefs as States 19committed to a state of the brittle thing. But we are not logicallycommitted to any further characterization of that state.

It may be thought that the notion of the internal structure of astate is a confused one. An object, or perhaps an event, may have astructure of a certain sort, but how can a state, which is a speciesof property, have a structure? In answer to this objection, perhapsit is true that the notion of the structure of a state is an unusualone. Nevertheless, it seems a notion that, once explained, can beunderstood and applied without much difficulty. The molecules ofa certain brittle piece of glass may be bond ed in a certain fashion.That is a state the glass is in. Now since this state involves a certainsort of arrangement of the constituent molecules of the glass, it canbe said to have a structure. The state involves elements in a relation.I am arguing that belief-states must have a structure in this sense.T he belief-states involve elements in a relation .

I think it is fair to say that there is no more than a difference inthe degree of theoretical commitment involved in the concept of abelief as opposed to dispositions like brittleness. It has been arguedthat to call something brittle involves the claim that there is somestate of the brittle thing which is responsible for the thing breakingif struck. With belief we are carried a step further, and are com-mitted to states having a structure, a structure which corresponds tothe structure of the proposition believed. (This point of distinctionbetween beliefs and dispositions would seem to hold as much forgeneral as for other sorts of belief.)Concerning this structure of the belief-state a great deal more mustbe said in subsequent chapters of this Part.(iv) Finally, there is a distinction between dispositions such asbrittleness on the one hand and beliefs on the other, which hasseemed to many philosophers to be a far deeper and more funda-mental distinction than any of the three already mentioned. How-ever, I believe th at in fact it is a relatively superficial distin ction.If we ask why we attribute particular sorts of disposition to parti-cular objects it seems that our knowledge or belief is always basedupon evidence. It is indirect or inferential. We do not observe the

disposition, but rather its manifestation; or else, even more in-directly, we observe that things of this sort have, in the past, givenrise to manifestations of this sort. Now in the attribution of beliefsto other people there is, we ordinarily assume, a similar restrictionto indirect evidence. But if we take ourselves, then, at least in many


30/238

20 Beliefcases, we can know what our beliefs are independent of any mani-festation or expression of the belief. I know directly, without evi-dence, that I believe that acquired characteristics are not inherited.A lot turns on the question of how we conceive of this knowledgeof our own beliefs. If this knowledge is thought of as logicallyindubitable or incorrigible, or a matter of logically privileged access,then no doubt this would set a great logical gulf between beliefsand ordinary dispositions. For presumably there could be no suchlogically privileged knowledge of the brittleness of a solid object.But I have argued against the notion that we have any such privi-leged knowledge of our own mental states elsewhere (Armstrong1968, Chapter Six, Section X). I will not repeat these arguments.But once such a special knowledge of our mental states is dis-avowed, it seems to be simply a contingent fact that we have director non-inferential knowledge of (some) of our mental states, includ-ing (some) of our belief-states, but that we lack such knowledge ofthe brittleness of certain pieces of glass.

Not all knowledge can be based on evidence, because that evi-dence itself must be something which we know, and so, on painof vicious infinite regress, there must be some things which we knowdirectly, that is, not on the basis of evidence. (An argument which,together with the whole topic of non-inferential knowledge, we willdiscuss more fully in Part III of this book.) It so happens that someof the things we know without evidence are our own current statesof mind, although these are not the only particular matters of factwhich we know directly. We do not in general have such directknowledge of the dispositions of the material things in our environ-ment (although we do seem to have direct knowfedge of such statesof affairs as physical pressure on our body). But it seems a perfectlyintelligible concept that we should have direct knowledge of disposi-tions. We could imagine, for instance, that, as a causal result ofbringing his fingers into contact with pieces of glass, a personshould come to know whether the pieces were brittle or not, but noton the basis of any evidence. Th e state which is the brittleness wouldbe acting upon his mind to produce unevidenced knowledge thatthe object is brittle, either via some established sense-organ or insome other way. Such knowledge would seem to be in no differentlogical position from, say, knowledge gained by touch of the thing'sjaggedness or smoothness. That we have no such capacity to gainunevidenced knowledge of brittleness seems to cast no light on the


31/238

BeliefsasStates 21concept of brittleness, and so sets up no logical distinction betweendispositions like brittleness and beliefs.

But we have found three points of logical distinction betweendispositions and beliefs, (i) Manifestations of dispositions are' stim ulu s-d ep endent'; manifestations of belief a re not. (ii) Disposi-tions, if manifested, are manifested in only one sort of way; beliefs,if manifested, are manifested in indefinitely many ways, (iii) Whileattribution of a disposition to an object attributes a state to the dis-posed thing, the state need be credited with no particular structure;but belief-states must be credited with a structure which correspondsto the proposition believed.When the beliefs are restricted to beliefs about particular mattersof fact, it is the collection of structured belief-states in a particularmind at a particular time which is to be thought of as that mind'smap of the world at that time.It will be argued subsequently that the first two of these distinc-tions between beliefs and dispositions do not hold for generalbeliefs.

IV Belief and Co nsciousnessWe have rejected the view that a current belief is necessarily some-thing we are currently conscious of. Beliefs are states of mind which,so far from us being currently conscious of, we need not even knowthat we possess. (Other people, or we ourselves at a later date, maypostulate their existence in order to explain some feature of ourobserved conduct or our mental life.) Nevertheless, a belief can bea content of consciousness. It can be 'before our mind'. And so itmay be demanded of a theory of belief that it explain what it is fora belief to be before our mind.

We have already distinguished between a belief which is currentlycausally active in our mental workings and one which is not. It isimportant to see, therefore, that currently causal activity is com-patible with the belief not being a content of consciousness. Manyof the beliefs which guide our actions never enter consciousnesswhile the action is being preformed, yet the belief must be causallyactive at that time. Sometimes a confidently held belief turns out tobe false, and as a result the action based on it is unsuccessful, yetonly with failure do we become conscious that we had been allalong assuming the truth of thatbelief.It may be suggested that having a belief currently before our


32/238


33/238

Beliefs as States 23perception is on the right lines it can then be applied to introspection.Introspection will be the acquiring of knowledge or beliefs (inmany cases quickly lost again) concerning our own current mentalstate.

This brief sketch of a doctrine that is developed in detail else-where is simply meant to indicate how difficult it would be to solveall the problems which can be raised about consciousness of beliefswithin the present work. But, as indicated, I think that the problemswhich arise are not peculiar to the topic ofbelief.


34/238

Belief and LanguageI M anifestations and Expressions

In the previous chapter the notorious multiform na tur e of the pos-sible manifestations or expressions of beliefs was mentioned. Thereis no single sort of manifestation or expression which stands to aparticular belief as breaking on being struck stands to brittleness.Nevertheless, the question arises whether there may not be a classof manifestations or expressions which stand in a peculiarly intimaterelation to a belief viz. expressing it verbally.

It is convenient to introduce a terminological convention here, andm ake 'expression' a narrower term than 'man ifestation'. Some ofthe things which occur (if anything occurs) because somebody has acertain belief will be actions of that person: things which springfrom his will. Other effects of abelief, while still something whichthe believer does in the widest sense of 'does', will not be thingswhich spring from his will. Let us restrict the term 'expressions' tothose manifestations of beliefs'which spring from the believer's will.T h u s , my belief that p may bring it about that on a certain occasionI blush. In the terminology now to be adopted this will be a mani-festation of the belief that p, but it will not be an expression of thatbelief because blushing is no t som ething we do at w ill.

The manifestations of mental states that are important in explica-ting the concept of the mental state in question are, in general,expressions of that mental state, not mere manifestations. It wouldseem that beliefs are no exception to this rule. We can further distin-guish between linguistic and non-linguistic expressions of belief:between saying, and doing which is not saying. Now many philoso-phers, particularly in recent years, have been inclined to think thatthe linguistic expressions of belief are their logically primary mani-festations. The thought has been vague, yet it has been pervasive.In opposition to this, I will argue that there is no special logical orconceptual connection between beliefs and their linguistic expres-sion.

24


35/238

Belief and L anguage 25II Belief w ith o u t Language

The first, and obvious, point is that we constantly attribute beliefsto beings such as animals and very small children who lack anycapacity to speak (and, it may be added, have small capacity tounderstand what is said to them). The dog digs frantically at theplace where he buried a bone, or rushes to the door on hearing hismaster's voice. It is natural to think of his actions as expressions of abelief that he has a bone buried there, or that his master is at thedoor. If the bone has been secretly rem oved , or the * voice ' is simplya tape-recording, then the dog's belief is false.It is entirely natural to explain the dog's actions by attributingcertain beliefs to him. And if the explanation is so natural, that isalready some argument for thinking it an intelligible explanation(whether or not it is a true explanation). Against this, however, it isargued that this 'natural' explanation of the observed facts becomessuspect when it is asked exactly what it is that the dog believes.Has the dog got concepts of 'burying', of 'bone', of 'his master', of'the door'? It is sufficiently obvious that he does not have ourconcepts of these things. But if he lacks our concepts, what can itmean to say that 'he believes that he has a bone buried there', orthat 'he believes that his master is at the door'?W e w ant to say thatthe dog believes something - but we do not seem able to say whatIs our attribution of beliefs to the dog really intelligible after all?Perhaps it is concealed nonsense.

Here, however, we can take advantage of a distinction whichQuine has made familiar between 'referentially opaque' and 'referen-tially tra ns pa re nt ' propositions about beliefs. (See his essay 'Ref-erence and Modality' in Quine 1961.) Suppose that I point to aman across the room and say 'Smith believes that the chap overthere is the villain of the piece.' Suppose also that Smith believesthat Robinson is the villain of the piece and believes that Robinsonis at present in Madagascar. Suppose, finally, that Robinson is notin Madagascar but is the chap over there. Did I say something false?It depends on how I meant the sentence to be taken. If I meant it tobe taken in a 'referentially opaque' way then what I asserted wasfalse. In this way of taking my sentence, Smith does not believe thatthe chap over there is the villain of the piece. But if I meant mysentence to be taken in a referentially transparent way we can sub-stitute co-designating expressions ('that chap over there' for


36/238

26 Belief'Robinson') without loss of truth. If I happen to know that Smithbelieves that Robinson is in Madagascar I may even make it quiteexplicit that my sentence is to be taken in the 'transparent' way byadding 'of course, Smith believes the man is in Madagascar'.

There are these two ways of talking about beliefs, but it is clearthat the referentially opaque construction is the more fundamental.For it tells us the actual content of the belief which is in thebeliever's mind. The referentially transparent mode of speech is away of talking about beliefs without actually saying what their con-tent is. Such a glancing or indirect style of reference has obviousutility in discourse. It is particularly useful where the exact contentof the belief is not known. And this, I suggest, is what is happen-ing in our reference to animals' beliefs. We do not know the exactcontent of their beliefs and our attributions of belief to them haveno more than a referentially transparent force.

Let us return to the dog. To say that the actual content of theanimal's belief is that he has a bone buried in that place or that hismaster is at the door is almost certainly incorrect for the reasonsalready given. Yet on the evidence of the dog's behaviour it is anatural hypothesis to attribute some such belief to him. The exis-tence of a loose or referentially transparent way of talking aboutbeliefs enables us to resolve the dilemma. In saying that the dogbelieves that his master is at the door we are, or we should be,attributing to the dog a belief whose exact content we do not knowbut which can be obtained by substituting salva veritate in the pro-position 'th at his ma ster is at the do or '.Our assertion about the dog makes a claim of the following sort.

The dog has a belief of the form R(a, b). 'a' is a canine 'individualconcept' which picks out the very same individual as our individualconcept that we express by the words 'the dog's master', 'b' is acanine individual concept which picks out the very same individualas our individual concept that we express by the words 'the door'.' R ' is a canine concept tha t is applicable to the sam e class of orde redpairs (or much the same class) as our spatial concept that is expressedby the word 'at' in sentences like 'the dog's master is at the door'.Our characterization of the dog's belief does not render the exactcontent of his belief. But, then, neither does the remark 'Smithbelieves that chap over there is the villain of the piece' in ourprevious example. Both characterizations are referentially trans-parent.


37/238

Belief and Language TjWe may happen to know the exact content of Smith's belief. Inthe case of the dog we do not. Generations of work by animal

psychologists may be necessary before the exact content is known.(Piaget, of course, has already begun the work in the case of chil-dren who have not yet learnt to talk.) Other species of animal arebeings who both lack language and are very different from our-selves. It is not surprising that, although we can recognize that theyhold a belief of a certain general sort on a certain occasion, yet wecannot delineate the exact contours of thebelief.The argument, of course, does not show that dogs or other ani-

mals have beliefs, or even that it is intelligible to say that they havebeliefs. It simply shows that we need not give up our natural inclina-tion to attribute beliefs to animals just because the descriptions wegive of the beliefs almost certainly do not fit the beliefs' actual con-tent. So the question still remains whether there is any further argu-ment for animal-belief beyond what we are naturally disposed tosay w hen we observe their beha viour.The following argument seems to have weight. It is obvious thatanimals perceive. Now there is some very close connection betweenperceiving, on the one hand, and acquiring knowledge or beliefsabout the environment, on the other. The connection is differentlyconstrued by different theorists, but acknowledged by all. Thesenses inform us about what is going on in our environment (that isthe evolutionary reason for their existence) and, sometimes, deceive

us. But to be informed and to be deceived is to acquire knowledgeand/or belief.We distinguish between perceiving things, events and occurrences

on the one hand, and perceiving that something is the case on theother. The logical connections between A's perceiving an X andthe knowledge and/or belief that A acquires are somewhat complex.But if A perceives that something is the case (sees that the cube islarger, although more distant, than the sphere), then it is entailedthat A knows that that thing is the case. And 'A knows that p'entails 'A believes that p.' (This entailment has been questioned. Itwill be argued for in Chapter Ten.) Now it seems obvious that thedog perceives that a cat is streaking across the lawn in front of him,or that his dinner has been put on his plate. (Allowing, of course,that what he is here said to perceive may not be the exact content ofhis perception.) So the dog acquires knowledge and, if he acquiresknowledge, acquires beliefs.


38/238

28 BeliefBut if animals have beliefs, then there is no necessary connectionbetween having beliefs and having the capacity to express them

linguistically.The absence of this necessary connection is often admitted, but itis admitted in a very grudging fashion.It is often said, or hinted, that those who lack linguistic com-petence have beliefs, but have them in some * logically secondary'sense only. The question then arises what this phrase 'logicallysecondary' means. I can think of three reasonably precise meaningsfor the phrase. (I would rather have been able to quote some otherphilosopher's account(s) of the phrase 'logically secondary'. But,to use F. H. Bradley's phrase, here I have been forced to do myscepticism for myself.) In what follows I try to show that animalbeliefs are certainly not logically secondary cases of belief in thefirst of these meanings; that some animal beliefs may be logicallysecondary cases in the second sense; and that it is quite unproventhat they are logically secondary in the third sense.

The first meaning that might be attached to 'logically secondary'is this. Let it be given that there is a class of things of the sort X,and a sub-class of X: the class of things of the sort Y. Y's are thenlogically secondary instances of X if, and only if, (i) it is logicallypossible that there should be no Y's and yet there still be X's; but(ii) it is logically impossible that Y's should be the only X's whichexist. For instance, a distan t relative m igh t be j>aid to be a m em berof a certain family, but it might also be said that the relative wasa member of the family in a logically secondary sense only. For ifthere were no members of the inner family circle, logically therecould be no distant relatives. Yet there could be no distant relativesand still the inner family circle would exist.

But it is completely implausible to say that if there were no beingswith beliefs which they could express linguistically, then it would beimpossible for a dog to believe that he had a bone buried at a cer-tain place. (That there would be nobody toassertthe belief is neitherhere nor there.) If the dog believes, then what makes his belief abelief is something which pertains to the dog, and has nothing to dowith hu m an beings and the ir speech. Th is is utterly different fromthe case of the distant relative, where what makes him or her adistant relative is (distant) relationship to members of the innerfamily circle.

There is a second, related, but more subtle, sense which can be


39/238

Belief and La nguage 29given to the phrase * logically secondary'. Y 's m ight be said to belogically secondary cases of X's without it being entailed that otherX's, of a logically primary sort, exist. It is necessary only that thedescription ' X ' m akes essential reference to (possible) X 's w hichare not Y's, but that, in applying to X's which are not Y's, thedescription ' X ' m akes no essential reference to X 's which are Y's.(It will be seen that our first sense of 'logically secondary' is aspecial, stronger, case of this second sense.)

For instance, it is ordinarily said of brittle things that, if they arestruck, they shatter. Yet, as we have had occasion to notice, a brittlething may be struck and fail to shatter, because of the nature of thecircumstances in which the brittle thing is placed. (The glass waspacked around with protective material.) Further, it is conceivablethat the only occasions when brittle things are struck are occasionswhere the circumstances are unpropitious for the manifestation ofthe disposition. If this was the history of the world, it is unlikelythat anyone would ever form the concept of brittleness. But lack ofthe concept would not mean that there were no brittle things.Faced with this rather complicated situation, it seems possibleto say that there are certain 'logically primary' cases of brittlethings: those things that are so circumstanced that, if hit, theyshatter. In calling something brittle we make essential reference tocases of this sort. Brittle things that are so circumstanced that, whenhit, they fail to shatter may therefore be said to be 'logicallysecondary' cases of brittleness. Yet there might be no logically pri-mary cases.

It is clear that this sense of 'logically secondary' is weaker thanthe first sense. Does it provide a sense in which animals' beliefs aresecondary to verbally expressible beliefs?Beliefs may or may not be expressed, and among those which arenot expressed there may be those which the believer is incapable ofexpressing. The capacity to express a belief is similar in many waysto the capacity to manifest a disposition. So it might be argued thatbeliefs that cannot be expressed are logically secondary cases ofbelief.But this argument would only give a certain logical primacy toexpressible beliefs. There is nothing here, short of begging thequestion, to give linguistic expressibility any special place. The bestthing that can be said in favour of a special place for linguisticexpressibility is this. There may be certain beliefs which, if they areexpressed, can be given nothing but linguistic expression. (An issue


40/238

30 Beliefto be discussed in Section Three.) If there are such beliefs, and ifthey can be attributed to animals, then such animal beliefs will belogically secondary cases of belief. But there is nothing in the argu-ment to show that animal beliefs generally are logically secondary inthis sense.

T h e th ird m ean ing for * logically seco ndary' finds application insituations where there is a scale of some quantity or a scale of com-plexity. In such cases the term ' X ' m ay have clear application toinstances at one end of the scale, may clearly lack application toinstances at the other end, but may be said to apply to instancesfalling into a twilight zone between in a logically secondary wayonly. (This third meaning may be another special case of thesecond.)

It is clear that men with language, on the one hand, and animals,on the other, lie along a certain scale or continuum. Consider thedeclension: men, apes, dogs, lizards, ants, earthworms and amoebae.It seems quite clear that amoebae lack beliefs, and only a few wouldwant to argue a case for earthworms. But above that point it isnot so clear. We then climb up a scale until we come to the oneclear or paradigm set of cases: men with language. Now, it may besuggested, must we not treat the intermediate cases as, at best,logically secondary cases of belief?Such a conclusion does not follow. Such a line of argument wouldlicence us to say that, although a man with a somewhat sparse headof hair was not bald, nevertheless he was, of necessity, a logicallysecondary case of a man with a head of hair. But this is false. Hehas a head of hair and so he is not bald. He would be a logically

secondary case of a man with a head of hair if it was so sparse thathe had entered the zone where we had begun to be undecidedwhether it might not be appropriate to use the word 'bald'. But ifhe is clear of that zone, then he is just as much a man with a headof hair as a m an w ith the bushiest mo p im aginab le.By parity, in order to show that animals have beliefs in thissecondary sense only, it would be necessary to show that animalsstand in a borderline zone between cases w here the term 'be lief is

definitely applicable and cases where it is definitely not. It is notenough to point out that animals are nearer on a scale of complexityand sophistication to amoebae (who do not have beliefs) than menare (who have beliefs). For animals might still be clear-cut cases ofbelievers for all that. Now how is it to be shown that animals' beliefs


41/238

Belief and La nguage 31are of this twilight sort? By comparison with men, animals lack oneimportant means of expressing their beliefs: language. But it wouldbe begging the question to argue that lack of this important meansof expression entails that we are dealing with logically secondarycases ofbelief.

Admittedly, it is not clear how it is to be proved that animalbeliefs are not, in this sense, logically secondary cases. But is notthe onus of proof on the other side? As already emphasized, ordin-ary thought and language quite readily attribute beliefs to those whoare incapable of using, or even understanding, language. Whyshould we not take such attributions at face value? Future scientificinformation about non-human animals and the springs of theirbehaviour might just possibly be forthcoming, information whichwould persuade us that animals do not have beliefs or have themonly in some secondary or dubious sense. Their inability to talkshould not persuade us of this.

Notice that whatever we say of beliefs must also be said of theconcepts which are involved in those beliefs. So animals have con-cepts as well as beliefs.Ill Soph isticated Belief and Language

But even if it is admitted that animals have beliefs in a quitestraightforward sense, it may still be pointed out that their beliefsare simple and unsophisticated things. If we consider beliefs witha more complex content, it may be said, these are logically linkedto the requisite linguistic competence because without languagethe beliefs could not be expressed. Wittgenstein's remark that a dogcannot believe his master will come the day after tomorrow(Wittgenstein 1953, p. 174) was, I suppose, intended to support somesuch view. And even if the example is incorrect, we can substitutea belief of a much more abstract sort. Could a dog believe in thetruth of Goldbach's conjecture (that every even number is the sumof two primes)? How could we suppose this without endowing thedog with a sophisticated linguistic competence?

It is important to realize that there are two questions to bedecided here. First, are there beliefs which can be given none butlinguistic expression? Second, if there are such beliefs, is it possiblefor a being without the necessary linguistic competence to holdsuch a belief despite the fact that he cannot express it?


42/238

32 BeliefIn answer to the first question: I think that there are some beliefswhich can only be expressed linguistically, by bringing into existence

some sort of symbolic entity either in the believer's mind or in theworld. A class of such beliefs will be mentioned in Chapter Seven(unrestricted existentially quantified beliefs about contingent exis-tences). But the class of beliefs which can only be expressed linguisti-cally is very much smaller than is often assumed. A little ingenuitycan often produce descriptions of behaviour which are plausiblecandidates for non-linguistic expressions of quite sophisticatedbeliefs. Suppose one dog's master is accustomed to leave the dog fora full day, while another regularly leaves his dog for two days. Ifthe first dog appears alert for his master's return the very next day,but the second dog lets a day pass before appearing restless andexpectant, it would not be unreasonable to suggest, on the basis oftheir behaviour, that the first dog expected his master to returntomorrow, but that the second dog expected there to be a day'sinterval before his master got back. Admittedly, the belief involvedis still quite simple. It is hardly comparable with a dog's believingthe truth of Goldbach's conjecture. But it may be that one couldwork up through a series of ever more complex cases culminatingin the providing of behavioural but non-linguistic expressions ofbeliefs in the truth of abstruse mathematical hypotheses.

For the sake of argument, however, let it be admitted that thereare beliefs which can be expressed in a linguistic way only. It is afurther step to argue that a being which lacks linguistic competencecannot hold such abelief. And in fact it seems to be an incorrectstep.A t this point we should recall the following facts. W e speak ofgroping successfully or unsuccessfully for the words which expressour belief. We say that the words which we do utter quite fail tocapture what it is that we believe, yet we may not know whatbetter words to utter. We speak of our never having put certainbeliefs into words, even although they belong to that class of beliefswhich have no natural expression except a linguistic one. Thereseems no reason to discount such ways of talk ing . But if they 'are

not discounted, then there is a gap between (i) holding a belief and(ii) knowing how to express the belief linguistically. So why shouldit not be possible to hold sophisticated beliefs in the complete absenceof the linguistic capacity to express them ?It may be replied that the sort of talk about beliefs just mentioned


43/238

Belief and L anguage 33cannot be taken to support any radical distinction between sophis-ticated belief and the competence to express it. When I struggle toexpress my belief in words, I do not lack linguistic competenceentirely. I can indicate in words the general area with which mybelief is concerned. I can recognize that some verbal formulationsare more appropriate than others. Perhaps there exists a verbalformulation which, if it occurred to me, I would recognize asrendering my belief exactly. So, it might be argued, even in thedifficult cases, belief and its linguistic expression are linked in someloose, admittedly hard to formulate, but nevertheless logical rela-tion.At this point in the argument it seems appropriate to appeal toempirical data of a relatively recondite sort. Certain persons haveclaimed to hold beliefs of a quite abstract and sophisticated sortwhen quite unable to express such beliefs verbally or understand averbal formulation of them. A famous case is reported by WilliamJames (James 1950, Vol. 1, pp . 266-9).A deaf-mute, Mr Ballard, lost his hearing as a child. He said ofhimself:

I could convey my thoughts and feelings to my parents and brothersby natural signs or pantomime and I could understand what they saidto me by the same medium; our intercourse being, however, confined tothe daily routine of home affairs and hardly going beyond the circle ofmy own observation . . .Nevertheless, Ballard records that during that time hegained ideas of the descent from parent to child, of the propagation ofanimals, and the production of plants from seeds.H e says that at that timeI believed that man would be annihilated and there was no resurrectionbeyond the grave ...He also entertained various views about the earth, sun and moon.Besides holding such beliefs, he records that he struggled with suchquestions as 'the source from wh ich the universe cam e'.The case is discussed by Wittgenstein in the Philosophical In-vestigations (Section 342). Wittgenstein's direct concern is withwordless thought rather than wordless belief, but it seems clear thatwhat he says about thought he would also have said about belief.


44/238

34 BeliefWittgenstein is rather tentative, but he raises the sceptical questionhow Ballard can be sure that he has correctly translated his thoughtinto lan gu age, an d suggests that these recollections may be a * queermemory phenomenon'. This latter remark seems to link up withwhat he says elsewhere in the book about dreaming (p. 184) and'calculating in the head' (Section 364), where it is hinted that allthat occurs is that the waking man has the impression of certainthings having happened which never did happen or the *calculator'has the (incorrect) impression of having (overtly) calculated.

But why should we say that Ballard's recollections are a queermemory phenomenon? Why might they not be a correct reportof what he believed and thought? (Allowing, of course, for theempirical fallibility of memories of childhood.) The chief obstacleseems to be thea prioriconviction that there is a logical link betweenbelief and thought on the one hand, and linguistic competence onthe other. But to use this as an argument is to beg the question atissue.One problem which obviously worries many philosophers here,including, I think, Wittgenstein, is the problem of verification.How could we ever have any independent check on reports suchas that of Ballard? But I think that these philosophers have failedto see (despite their frequent references to such possible evidence)that neurophysiological evidence could, in principle at least, providethe independent check required.

It may be helpful to consider first the parallel question concerningme ntal images. W e kn ow that others besides ourselves have m entalimages because they tell us so. Nothing else in their behaviour givesus any clue. Do animals and small children have mental images?W e can at least conceive of the issue be ing settled fairly conclusivelyby physiological evidence. Suppose that we discovered, by interrogat-ing human subjects, that in the case of those interrogated, certainidiosyncratic neurophysiological processes were necessary and suf-ficient for the occurrence of mental images. This correlation wouldhave to be established by reference to introspective reports. But weregularly accept as true, generalizations whose scope is far widerthan the evidential base on which they rest. So there would be nobar to extending the scope of the generalization to include animalsand small children. If the idiosyncratic processes occurred in theircentral nervous systems, this would be compelling evidence (not, ofcourse, logically conclusive evidence) that they too had mentalimages.


45/238

Belief and L anguage 35Now, in just the same fashion, it seems perfectly possible thatneurophysiological evidence should bear on the question whether a

deaf-mute did or did not have beliefs or thoughts which he was un-able to express. The identification of the nature of these beliefs frommere neurophysiological evidence is, no doubt, a piece of science-fiction in the present state of neurophysiology. But it does show thelogical possibility of an independent check upon claims like that ofBallard. This should be sufficient to remove the suspicion that suchclaims are a legitimate targe t for philosophicalscepticism.Returning to the particular case of the dog who believes Gold-bach's conjecture to be true, it might be objected that it is impossibleto hold such a belief by itself.A belief of this sort presupposes thepossession of a whole set of other mathematical beliefs and concepts.Such a belief can only exist as a mem ber of a system of beliefs.Whether or not this contention is true, it clearly has a good dealof plausibility. Let us grant it for the sake of argument. But itseems to have no force against the supposition of our mathematicalbut inarticulate dog. For, although the objection forces us to endowthe dog with a whole complex of beliefs and concepts all of which

he is unable to express, there seems to be no greater difficulty inthis than in the original supposition.Of course, none of what has been said rules out the view thatdevelopment of a sophisticated linguistic competence is empiricallyessential for the development of very abstract beliefs. It seems to bea matter of fact that systems of abstract belief (and thought) andthe attempt to express such beliefs (and thoughts) linguistically,develop hand in hand. (A development which is, perhaps, a matter

of mutual causal interaction, belief leading to expression of belief,and the expression leading in turn to the forming of more sophis-ticated beliefs. The process is one of 'positive feedback'.) Theremay be deep psychological reasons why it is impossible for a beliefin the correctness of Goldbach's conjecture to develop except inassociation with the development of appropriate linguistic skills. Allthat has been argued for is that there is no logical link betweenpossessing such beliefs and having the corresponding linguisticcompetence.It has, however, been admitted that in one rather weak sense of'logically secondary' such beliefs might be said to be logicallysecondary cases of beliefs.


46/238

36 BeliefIV Th ou gh t and Language

It seems clear, and it has already been implied in passing, that ifbelief is not logically dependent upon the competence to express thebelief verbally, then the same will apply to thou gh ts.For each possible belief whose content is p, there is a thoughtwith the same content. Someone who does not believe that p maynevertheless have the thought that p. We will restrict discussion of'thoughts' in this essay to this sub-class of thoughts, thoughts in theabsence of the correspondingbelief, since our concern is with belief,and it is thoughts in this restricted sense which it is natural to con-trast with beliefs. (It is such thoughts which bring up 'Hume'sproblem' of the distinction between belief and thought.)

Now it seems obvious that there are such thoughts, and that theyare mental occurrences that can be dated. It would seem furtherthat they endure for a certain time (often a very short time), andthat they are entire for the whole time that they endure. The argu-ment for the latter is that it seems to make no sense to speak ofbeing, say, half-way through having a certain thought. One can behalf-way through a certain train of thought, or half-way throughexpressing the thought in words, but the thought itself is just there,or it is not. In this, thoughts resemble beliefs. Unlike beliefs,thoughts may be acts, may be things which the speaker does, whichspring from his will, although they may also be 'u nb id de n'.Like beliefs, thoughts may or may not be expressed in outwardaction. In the case of thoughts, this outward action would seem tobe simply a matter of putting the thought into words (or other

symbols). Just as in the case of mental images, there seems to be noother form of outward behaviour which constitutes a natural mani-festation of a thought.But it cannot be concluded that thought is impossible without thecorresponding capacity to express it, or without any capacity toexpress it at all. Nothing in the behaviour of dogs points to theirever having thoughts, in the way that it does point to their havingbeliefs. Since dogs have no langu age it is difficult to see how any-thing in their behaviour should so point. But it is not logically im-possible that they should have thoughts. Furthermore, physiologicalevidence can be easily conceived which would convince us that theydid have thoughts.

We have already noted that we can be under the impression that


47/238

Belief and La nguage 37certain beliefs are temporarily or permanently beyond our powers oflinguistic expression. Exactly the same is true of thoughts. Andwhile in most cases language can at least skirt around our thought,Ballard reports that he had thoughts, as well as beliefs, of consider-able sophistication while quite unable to use or understand languagebeyond the most elementary gestures. He may have been deceivinghimself when he made these reports in later life. But it is mere dog-matism to say that he wasnecessarilydeceivinghimself.Belief and thought appear to introspection as somewhat impal-pable things, things hard to grasp. The positivistic and/or beha-viouristic spirit that is even now still abroad in philosophers' think-ing about the mind, seizes with relief on the concrete, observable,outward phenomenon of linguistic utterance. So philosopherswrongly try to give an account of belief and thought in terms of their(possible) linguistic expression.

A final observation.Under the influence of Chomsky, modern theoretical linguisticsis taking a less behaviouristic view of language. Behind the *surfacestructure' of the sentence we are asked to postulate a 'deep struc-tu re ' which carries, or is paired with, the semantic content of thesentence. This 'deep structure' is seen as, or as corresponding to, apsychological reality which the speaker transforms into a surfacestructure and the hearer recovers from the surface structure.One hypothesis which might be considered, therefore, is that the'deep structure' is not a specifically linguistic entity at all, but issimply the belief or thought (or other mental state such as intentionor desire) which is to be expressed in words. Such a view wouldhave to allow that there could be 'deep structures' within an indi-vidual mind which were quite beyond the linguistic competence ofthat mind to express perhaps because, as in the case of Ballard, thatperson lacked any linguistic competence. But I do not see whylinguistics should not accept such a view of 'deep structure'. Inparticular, it seems to be at least consilient with the positions ofthose wh o now call themselves 'gen erative sem anticists'.


48/238

4Propositions

I The Notion ofaPropositionIt was argued in the third section of Chapter Two that belief-statesmust be assumed to have an internal complexity, a complexitycorresponding to the content of the proposition believed. Thoughts,at least in the narrow sense which we have given to the term, alsoinvolve propositions, and so must also be credited with an internalcomplexity. Hence, if we wish to cast light on the nature of belief(and thought), it seems important to understand what propositionsare,and in what way they enter beliefs and thoughts.

The notion of a proposition seems forced upon us when weconsider beliefs, thoug hts and also assertions.It is clear that different people may all believe the same thing.Suppose, for instance, that nine men believe that the earth is flat.We have nine different beliefs. There is A's belief, B's belief, C 'sbelief... If what we have said about belief already is correct, thenthere are nine numerically different states. (Whether these statesare to be conceived of as purely physical states of the brain, or asstates of a spiritual substance, or in some other way, is not at issuehere.)

However, we distinguish between a man's belief-state and thething he believes: between the believing and what is believed:between the state of affairs Bap and p. In the case of the nine men,what is believed is the same thing in each case: that the earth is flat.It is just such a case that philosophers, at any rate, describe bysaying that the nine men all believe the same proposition. So it issometimes true that people believe the same proposition. Whether,or in what sense, we are thereby committed to the existence ofentities called propositions, is a further question that will have to beconsidered. But, to repeat, it is sometimes true that people believethe same proposition.

Different people may all think the same thought. Suppose that ninepersons entertain the possibility that the earth is flat. (If we place


49/238

Propositions 39them in the past, they might plausibly have done this without havingany belief about the matter one way or the other.) There is A'sentertaining . . ., B's entertain ing . . . If beliefs are properly con-strued as belief-states, then the corresponding thought (the thoughtthat the earth is flat as opposed to the belief that it is flat) will pre-sumably be a state which endures for as long as the thought ispresent in the mind. At any rate, there will be nine numericallydifferent things, thought-states or thought-episodes, however theyare best conceived.

But we distinguish

d. m. armstrong belief, truth and knowledge 1973

Documents