UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Direct reference and belief attributions Bryans, Joan Douglas 1989

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
831-UBC_1990_A1 B79.pdf [ 9.63MB ]
Metadata
JSON: 831-1.0302391.json
JSON-LD: 831-1.0302391-ld.json
RDF/XML (Pretty): 831-1.0302391-rdf.xml
RDF/JSON: 831-1.0302391-rdf.json
Turtle: 831-1.0302391-turtle.txt
N-Triples: 831-1.0302391-rdf-ntriples.txt
Original Record: 831-1.0302391-source.json
Full Text
831-1.0302391-fulltext.txt
Citation
831-1.0302391.ris

Full Text

DIRECT R E F E R E N C E A N D BELIEF ATTRIBUTIONS  by JOAN DOUGLAS BRYANS B.A., University of British Columbia, 1984  A THESIS S U B M I T T E D IN P A R T I A L F U L F I L L M E N T OF T H E R E Q U I R E M E N T S FOR T H E D E G R E E OF DOCTOR OF P H I L O S O P H Y  in T H E F A C U L T Y OF G R A D U A T E STUDIES Department of Philosophy  We accept this thesis as conforming to the required standard  T H E U N I V E R S I T Y OF BRITISH C O L U M B I A October 1989 (c) Joan Douglas Bryans, 1989  In presenting this thesis in partial fulfilment of the  requirements for an advanced  degree at the University of British Columbia, I agree that the Library shall make it freely available for reference and study. I further agree that permission for extensive copying of this thesis for scholarly purposes may be granted by the head of my department  or  by his  or  her  representatives.  It  is  understood  that  copying  or  publication of this thesis for financial gain shall not be allowed without my written permission.  Department The University of British Columbia Vancouver, Canada  DE-6 (2/88)  ii  ABSTRACT  The  aim  of this dissertation is  to provide  a non-Fregean  account  of the  functioning of belief attributions ( B A ' s ) , specifically those of the form ' £ believes that Fa' where ' o ' is a proper name, w h i c h provides a satisfactory account of the  phenomena  associated w i t h the substitution of co-referential names and w i t h the use of vacuous names. A f t e r an intitial study of non-Fregean theories of reference, specifically those of K r i p k e , K a p l a n and Donnellan in w h i c h K a p l a n ' s introduction of content, of the singular proposition, is found to be necessary, an examination of certain proposed solutions for B A ' s , compatable with direct reference, is carried out. These proposals, namely those of Quine, P e r r y and N a t h a n Salmon, are all found w a n t i n g , the latter two due to their being, ultimately, Fregean. A non-Fregean approach is initiated beginning w i t h an examination of our actual practices in using B A ' s . It is found t h a t very different information can be conveyed by the use of the same sentence in the same context. W h i l e this differing information cannot be captured by means of the proposition expressed, it can be captured by treating the B A as an answer to a question. Belnap's logic of questions and answers is then developed to encompass vacuous terms and, w i t h this i n place, two distinct uses of B A ' s emerge. In one, the B A is used to provide a direct answer to the question; in the other it is used in order to modify the c l a i m to t r u t h of the embedded proposition, to provide a tentative answer. Problematic cases of B A ' s are then examined. It is found that substitution in all cases is permissible. Supposed difficulties w i t h this position in the area of belief itself and w i t h the explanation of action are discussed  and resolved, the latter p a r t l y by  means of the development and application of an account of ' w h y ' questions and answers.  iii  The use of vacuous names is then investigated and a difference noted between cases in which the B A is used to provide a tentative answer and those in which it constitutes a direct answer. In the former case, the use of a vacuous name results in no answer being given. However, given the nature of tentative answers, no problems specific to belief attributions are generated in such cases. In order to deal with cases where the vacuous name occurs in a B A asserted as a direct answer, Evans' account of pretend games is invoked, though modified to permit a possible world account of counterfactuals.  iv  TABLE OF CONTENTS  Abstract  11  Acknowledgements  V 1  Preface  I  1  D i r e c t Reference Theories The F r e g e a n V i e w  II  6 7  The Negative A r g u m e n t s  13  The Positive V i e w  23  Theories of Belief A t t r i b u t i o n s  34  The F r e g e a n Solution  38  Proposed Methods of Resolution  48  The Q u i n e a n T r e a t m e n t  48  The Perry/Kaplan Proposal  52  The S a l m o n Proposal  57  III Belief Attributions: W h y Do W e M a k e T h e m ?  68  IV  Questions and A n s w e r s  82  V  B e l i e f A t t r i b u t i o n s as A n s w e r s  102  Direct A n s w e r s  102  Tentative A n s w e r s  112  Summary  120  VI  Substitution  122  Tentative A n s w e r s  128  Direct A n s w e r s  131  V I I Problems w i t h the Account of Substitution Offered  137  Belief.  137  The E x p l a n a t i o n of A c t i o n  142  Cases of Conflation  143  Cases of Duplication  155  Conclusion  161  Names  164  VIII Vacuous  Tentative A n s w e r s Devious U s a g e Direct A n s w e r s  164 168 181  S u m m a r y and Conclusions  187  Bibliography of W o r k s Cited  193  vi  ACKNOWLEDGEMENTS  A b o v e a l l I should like to thank m y supervisor T h o m a s P a t t o n for the unstinting encouragement and attention he has given me, and for his helpful and insightful criticism. T h a n k s also are due to G a r y  Wedeking  and E d L e v y  for their helpful  comments and advice. I should also like to thank the Social Sciences and H u m a n i t i e s and Research Council and the U n i v e r s i t y of B r i t i s h C o l u m b i a for providing the necessary  financial  support in order for me to continue m y studies. O n a different but equally i m p o r t a n t level I would like to thank m y daughter Laura,  whose  unflagging  support  and  completion of this dissertation possible.  encouragement  over  the  years  made  the  1  PREFACE  T h i s work concerns itself w i t h theories of direct reference and, more specifically, w i t h the problems which such theories face i n the a r e a of belief attributions. Direct reference theories have i n common the view t h a t proper names do not refer to the person or object i n the w o r l d which is their referent by means of some intervening medium  such  as  a Fregean  sense,  but do so directly. O n  such  a view  the sole  contribution made by a proper name to the t r u t h conditions of an assertion of a sentence in w h i c h it occurs is the referent itself. F o r example, the sole contribution of the name ' J i m S m i t h ' to whether an assertion of ' J i m S m i t h is at w o r k ' is true is the m a n himself. O n l y i f it is true of that m a n that he is at work is the assertion true. A l t h o u g h intuitively appealing, many problems beset such a simple view. None are more demanding t h a n those associated w i t h belief attributions, that is, statements w h i c h attribute a belief to someone, as, for example, does ' J o h n believes that J i m S m i t h is at work'. The problems associated w i t h such statements fall into two distinct groups: those which concern the possibility of substituting one co-referential name for another w i t h i n the embedded sentence of that used to attribute the belief; and those concerned w i t h the use of a vacuous name, a name w h i c h has no referent, w i t h i n the embedded sentence. Direct reference theories have been seen to be singularly inequipped to deal w i t h such problems. Indeed it would seem to be the capacity to provide a solution to these problems that constitutes m u c h of the appeal of the F r e g e a n view. Are the problems t h r o w n up by belief attributions insurmountable, given a direct reference point of view? Do they spell the death knell of direct reference theories as  Preface  2  some have maintained? Does the Fregean theory in fact provide non-problematic solutions? It was with such questions in mind that I began my investigations. It is important to get as clear as possible on just what the direct reference view is and so, in Chapter 1, various arguments of direct reference theorists are examined. As a direct reference theory is distinguished as much by its negative arguments against the opposing, Fregean, position as by its positive views, it is necessary first to briefly sketch the relevant details of Fregean theory. The negative arguments offered against the Fregean are, I believe, strong and compelling. When the positive views are examined it is seen that there are two very distinct positions on the workings of direct reference: that of Saul Kripke and that of David Kaplan. For example, Kaplan's theory appeals to propositional content whereas Kripke's does not. Kaplan's theory will b e found to b e prefereable, distinguishing as it does between the functioning of proper names and that of definite descriptions, a distinction, it will be argued, which Kripke's theory cannot make. Irrespective of one's choice between Kaplan's and Kripke's views, the problems of belief attributions still loom and these are examined with in Chapter 2. It is seen that, as they stand, the views of Kaplan and Kripke offer no solution whatsoever to the problems generated by the phenomena of substitution and vacuous terms. In the light of this failure, despite the previous arguments against it, the Fregean view is examined once again, this time for the solution which it offers to the problems of belief attributions. Upon examination, the Fregean solution is found to be riddled with problems of its own, so much so that it offers no compelling reason to abandon the direct reference view. However, some solution to the problems must be found. Consequently, other potential solutions, the most promising of the contenders which assume the direct reference view, are then examined. All are seen ultimately to fail to provide the  Preface  3  necessary solutions. The reason why they f a i l , however, is instructive. T h e  proposed  solutions which are examined in C h a p t e r 2, although a s s u m i n g a direct reference view of reference, are nevertheless deeply Fregean. A l t h o u g h rejecting F r e g e ' s notion of sense w i t h respect to the functioning of naming, a l l introduce it, in one f o r m or another, at another level. W h a t is necessary is a non-Fregean approach; a solution w h i c h does not rely on the F r e g e a n notion of sense at a l l . T h i s is w h a t is attempted i n the remainder of this work. I begin b y noting that the problems  at hand  are problems  concerning belief attributions  of language,  not problems  of mind,  problems  not problems about belief. Before proceeding towards  a  solution, an examination of the phenomenon of belief attributions is necessary. Although belief attributions have been the subject of much discussion in recent j'ears, their use is r a r e l y , if ever, looked at. W h y do we use belief attributions? H o w do we use them? W h e n do we use them? Such are the questions addressed in Chapter 3. W h e n this e x a m i n a t i o n is carried out, i t is found t h a t the v e r y same sentence, of the f o r m ' B believes that a is F", can be used to convey v e r y different i n f o r m a t i o n within the same (physical) reference  theories,  context. Propositions, even the context-inclusive propositions of direct are  not  fine-grained  enough  to  capture  such  differences  of  information. W h a t is necessary is a treatment of the assertion of such sentences which is sensitive not only to the physical context but also to the conversational  context.  The differing information which can be i m p a r t e d in different conversational contexts can be captured by treating the belief- attribution as an answer to a  question.  Before c a r r y i n g out a detailed examination of belief attributions as answers, a brief account of the functioning of questions  and answers is necessary.  T h i s is given in  C h a p t e r 4 where N u e l Belnap's treatment of erotetic logic is expanded to include cases of vacuous terms. W i t h an account of the functioning of questions and a n s w e r s in hand, non-problematic cases of belief attributions are examined more closely in C h a p t e r 5.  Preface  4  W h e n treated as answers, two distinct uses of belief attributions emerge: those in which the attribution is made in order to i n f o r m the audience about something w h i c h concerns the believer or the belief itself; and those in w h i c h they are made in order to inform the audience about some quite other matter. In the former case, the attribution constitutes a direct answer to the question asked; i n the latter, it provides w h a t I t e r m a tentative answer. Cases involving tentative answers are p a r t i c u l a r l y interesting and are examined in some depth. In such cases, it w i l l be argued, the belief attribution is made for the sole purpose of m o d i f y i n g the c l a i m to t r u t h of the embedded proposition. The belief attribution offers a means by which the embedded proposition can be asserted, but only tentativelj'. W i t h the treatment of nonproblematic belief attributions in place, attention is then turned to the problems of substitution and vacuous names. In C h a p t e r 6, which is devoted to the problem of substitution, cases of attributions w h i c h constitute tentative answers are examined separately from those in w h i c h the attribution constitutes a direct answer. However, i t will be argued, despite first appearances to the contrary, the a b i l i t y to substitute co-referential names w i t h i n the belief context without affecting the t r u t h conditions of the assertion are identical in both types of case: substitution is possible regardless of w h a t type of answer the belief attributions constitutes. However, due to w h a t is being asserted and the implications w h i c h the assertion does and does not c a r r y w i t h it, the ability to so substitute does not do violence to the position of the, perhaps protesting, believer. The conclusion reached is perfectly compatable w i t h the direct reference theory. However it does invite certain counter-arguments which concern the nature of belief itself, and others w h i c h concern the explanation of action. C h a p t e r 7 is devoted to their rebuttal. In C h a p t e r 8 the problems associated w i t h the possibility of a vacuous name occurring in the embedded sentence of that used to make a belief attribution are treated.  Preface  5  Again cases of tentative answers are dealt with separately from those which constitute a direct answer. Because of the nature of a tentative answer, a belief attribution containing a vacuous name asserted in such cases warrants no different treatment with respect to the functioning of the vacuous term than that which would be given to a straightforward assertion of the embedded sentence. Hence its treatment under the direct reference theory poses no particular problems. Nevertheless, one group of cases is of interest. These are cases in which the attributor is deliberately trying to mislead the audience. Such cases are dealt with at some length as the manner of their treatment is of use when the second major type of case is investigated: those involving vacuous names which are asserted as direct answers. Unlike the case of substitution, the distinction between tentative and direct answers is relevant in the case of vacuous names. Upon examination, it is found that, though he would not do so in supplying a tentative answer, the (sincere) attributor may well use a known vacuous name in making the attribution to supply a direct answer. The treatment evolved earlier to deal with cases of deceiving tentative answers is evoked to deal with such cases. This involves an appeal to Gareth Evans' work on fictional discourse, though modified somewhat to deal with certain inconsistencies generated by combining his theory with that of direct reference. With such modifications in place, I will argue that the theory provides the means by which the problems of vacuous names can successfully be dealt with given the direct reference view.  6  1. D I R E C T R E F E R E N C E T H E O R I E S  Introduction  The idea that one can, using language, refer directly to a n object, person or place without the mediating effect of a cognitive, epistemic or conceptual element forms the foundation of several recent theories of reference. These theories of direct reference are to be found i n , or are inspired b y , the w o r k of S a u l K r i p k e , K e i t h Donnellan,  David  K a p l a n , and H i l a r y P u t n a m , and i t is the views of these theorists, in particular those of the f i r s t three mentioned, which w i l l be examined in this chapter.  1  Direct reference theories have a strong negative as well as a positive basis a n d their appeal lies as m u c h i f not more i n their proponents' opposing  arguments  view as i t does i n their positive claims. Consequently,  against the  in order to fully  appreciate the negative arguments proffered b y such theorists a brief outline of the opposing view w i l l be useful.  The Fregean V i e w .  The view of reference against w h i c h the arguments  of the direct reference  theorists are opposed finds its modern roots i n the work of Frege. Frege argued that  Since P u t n a m is p r i m a r i l y concerned w i t h the functioning of n a t u r a l kind terms, his work is not of central importance to the topic i n hand. Recently the title ' D i r e c t Reference T h e o r y ' h a s come to be associated solely w i t h the w o r k of D a v i d K a p l a n a n d its offspring a n d is not seen to cover the views of K r i p k e or others who hold a non-Fregean view. H o w a r d Wettstein uses the t e r m ' N e w Theory of Reference' to cover the views of all those w h o hold that reference is achieved directly without a n intervening F r e g e a n sense. However, since this title is a t best appropriately applicable for but a few y e a r s and since i t lacks the descriptive power of the original, I shall unfashionably call a l l theories which hold that reference is achieved directly ' D i r e c t Reference Theories'. 1  Direct Reference  7  Theories  reference is never direct but is a l w a y s achieved v i a a sense. H e held t h a t there is a sense attached to all singular terms occurring i n a well-formed sentence, indeed to a l l expressions  occuring  therein w i t h  the possible  exception of the t r u t h functional  connectives. Some of the f o r m a l relations holding between singular terms as they occur in w e l l formed asserted sentences, their senses and their referents are: (i) D i f f e r i n g singular terms m a y have the same sense. Frege gives as a n example the terms 'dog' and ' c u r ' w h i c h , though d i f f e r i n g i n colouring, possess the same sense. 2  (ii) T e r m s w i t h the same sense refer to the same entity. (iii) T e r m s w i t h differing senses m a y refer to the same entity, or, to p u t it another w a y , co-referential terms need not have the same sense. T h e terms ' M o n t B l a n c ' a n d 'the highest m o u n t a i n i n Europe', occuring in the assertions ' M o n t Blanc is over 4,000 meters h i g h ' and 'The highest mountain in Europe is over 4,000 meters high', express distinct senses but these senses determine the same referent, the mountain M o n t Blanc. (iv) the sense and referent of any token singular t e r m are a l w a y s distinct.  3  (v) The sense of a complex singular t e r m is a function of the senses of its parts. (vi) A singular t e r m m a y have a sense but no referent. F o r example, '[the] words " t h e celestial body most distant f r o m the E a r t h " have a sense, but it is doubtful if they also have a reference'. 4  (vii) W h e r e the singular t e r m occurs i n a n indirect context, such as the complement of the belief in a belief attribution statement, it does not refer to its customary referent but to its c u s t o m a r y sense. 5  See ' O n Sense and Reference', i n Translations from the Philosophical Writings of Gottlob Frege, eds. P. Geach and M. Black, Oxford, 1960, pp. 56-78, p. 58; and also Posthumous Writings, eds. H . H e r m e s , F. K a m b a r t e l and F. K a u l b a c h , trans. P. L o n g and R. W h i t e , O x f o r d , 1979, pp. 127-8. 2  It is necessary to distinguish the sense and reference of a singular t e r m i n terms of the t e r m ' s token occurrence since, given F r e g e ' s treatment of indirect contexts, a sentence m a y contain two tokens of a singular t e r m , the sense of one (in the direct context) being the referent of the other (in the indirect context). 3  4  Frege, ' O n Sense and Reference', p. 58.  See, for example, Frege, ' O n Sense and Reference', pp. 66-7; and Philosophical and Mathematical Correspondence, eds. G . G a b r i e l , H . Hermes, F. K a m b e r t e l , C. T h i e l , a n d 5  Direct Reference  8  Theories  F u r t h e r , the sense of a singular t e r m w h i c h occurs i n an asserted sentence contributes to w h a t is expressed by the sentence, i.e. to the T h o u g h t .  6  Frege also holds that the sense of a singular t e r m is a non-subjective entity which is 'grasped by everybody who is sufficiently f a m i l i a r w i t h the language or totality of designations to w h i c h it belongs'.  7  E a c h sense connected w i t h co-referential terms  'serves to illuminate only a single aspect' of the referent; each 'determines the object in 8  different w a y s ' .  9  A n object can be determined i n m a n y different w a y s and e v e r y one of  these w a y s , Frege argues, c a n give rise to a special name - a proper n a m e or description w i t h a unique sense. If one had a comprehensive knowledge of the object one would be able to tell whether any given sense belonged to it. However, given that the properties and relations of any object are infinite, '[to] such knowledge we never a t t a i n ' .  1 0  F r o m such talk it might seem that the sense of a singular t e r m is s i m p l y a specification of one or more actual, mind-independent, properties (or relations) of the referent, that is, properties inherent in the object w h i c h is the referent irrespective of our awareness of them. B u t there is more to the notion of sense t h a n this. Senses are, it will be remembered, 'grasped' by competent language users. W h e n viewed in this light, sense is seen as a  conceptual notion. A singular t e r m expresses a certain concept of, or  a ' w a y of regarding', the object. The terms ' 2 ' , '1 + 1', ' 3 - 1 ' and '6-r3', for example, all A. V e r a a r t ; abridged by B. M c G u i n n e s s , trans, by H . K a a l , O x f o r d , 1980, p. 149 and p. 164. See, for example, Frege, Philosophical and Mathematical Correspondence, pp. 79-80. F o r an excellent discussion of these and other f o r m a l relations of Frege's notion of sense see G.P. B a k e r and P.M.S. H a c k e r , Frege: Logical Excavations, O x f o r d , 1984, esp. pp. 300-332. 6  7  Frege, ' O n Sense and Reference', pp. 57-8.  8  Frege, ' O n Sense and Reference', p. 58.  9  Frege, Philosophical  1 0  and Mathematical  Correspondence,  Frege, ' O n Sense and Reference', p. 58.  p. 152.  Direct Reference  9  Theories  express different 'conceptions and aspects' of the n u m b e r 2 .  11  Frege would seem to be  a s s i m i l a t i n g actual properties of the object to our concept of it. F u r t h e r , the notion of sense is  cognitive and epistemic. The senses of singular  terms are u l t i m a t e l y distinguished f r o m one another b y w h a t G a r e t h E v a n s has called ' T h e Intuitive C r i t e r i o n of D i f f e r e n c e ' .  12  The C r i t e r i o n applies i n i t i a l l y to that w h i c h  sentences express: thoughts. The thought associated with- one sentence S as its sense m u s t be different from the thought associated w i t h another sentence S' as its sense i f it is possible for someone to understand both sentences at a given time while coherently t a k i n g different attitudes towards them, i.e. accepting (rejecting) one while rejecting (accepting), or being agnostic about the o t h e r .  13  If the only difference between S and S' lies in the occurrence, in the same position, of differing but co-referential singular terms then the C r i t e r i o n also supplies the test for whether these singular terms possess the same sense. The terms 'the morning s t a r ' and 'the evening s t a r ' differ i n sense because someone could sincerely assent to ' T h e m o r n i n g s t a r is a body i l l u m i n a t e d by the s u n ' and dissent to, or be agnostic about, ' T h e evening star is a body i l l u m i n a t e d by the s u n ' .  14  T h a t the sense of a singular t e r m  is cognitive and epistemic can also be seen in the fact that, for Frege, in indirect contexts it is the sense which is the referent of the singular term. The sense of the t e r m contributes to the content of our- thoughts, beliefs, and doubts about the object.  Frege, ' F u n c t i o n and Concept', in Translations from the philosophical Gottlob Frege, eds. P. Geach and M . Black, O x f o r d , 1960, pp. 21-41, p. 53. 1 1  1 2  In The Varieties of Reference, O x f o r d , 1982, see esp. pp. 18-22.  1 3  E v a n s , The Varieties of Reference, pp. 18-19.  See Frege, Philosophical and Mathematical Correspondence, for examples of Frege's application of the C r i t e r i o n . 1 4  Writings  of  p. 80, p. 153, and p. 197,  10  Direct Reference Theories  Finally, the sense of a singular term is a semantic entity. The referent is that object which uniquely fits the sense. The sense is the mechanism by which one particular object rather than another is determined to be the referent. As Nathan Salmon has described this aspect of sense: A singular term does not have any intrinsic affinity to denote a particular object independently of that object's meeting certain conditions. Speaking metaphorically, the term lets its sense rummage through the existing things of a possible world at a given time in search of a uniquely suitable denotation, whoever or whatever that turns out to be. 15  We have here the flip side of the earlier view that the actual properties of the object give rise to the various concepts of it. Given this earlier view, the object that the sense eventually 'picks out' is guaranteed to be the right one, i.e. to be that which initially prompted the concept. Hence the sense of a singular term in Frege's account is a multi-faceted entity: (i) it is intimately connected with the actual properties of the object (if that exists); (ii) upon 'grasping' it we acquire a 'way of regarding', a concept of, the object; (iii) it contributes to the content of our thoughts about the object; and finally (iv) it performs the semantic role of referring.  16  How, given all this, is a particular sense of a singular term to be delineated? What, for example, is the sense of the proper name 'Aristotle'? A technical problem arises here. Senses are what names express, one cannot refer to the sense by using the name. Since, in direct contexts, a singular term always refers to its referent, whenever one wishes to talk about the sense of a given term one ends up talking about the referent. As John Searle has so evocatively characterized the problem:  15  Reference and Essence, Oxford, 1982, p. 11.  Not surprisingly, it is just these differing aspects of the notion which generate the counter arguments made by the direct reference theorists. 16  11  Direct Reference Theories  Imagine a game where marbles are dropped into bowls through pipes. This act is called referring. Pipes (senses) lead to bowls (references). It is a rule of the game that anything can be referred to. The difficulty is though that we cannot live up to this rule because we cannot refer to a pipe. Every time a marble drops into a pipe it goes through to the corresponding bowl. 17  In practice, when Frege provides examples of the sense of a singular term he does so by employing a definite description. While it may be the case that a sense cannot be equated with what is expressed by a definite description, it has traditionally been conceded that in order to communicate what a given sense is one has to use such a description, and that, in doing so, the sense in thus adequately represented. Let us now return to the question of what the sense of a proper name such as 'Aristotle' is. While maintaining that there is but one sense associated with most singular terms, Frege conceded that in the case of proper names this may not be the case - the sense attached by one speaker to the name 'Aristotle' may differ from that attached by another : 'With proper names it depends on how whatever it refers to is 18  presented'.  19  Frege has two distinct attitudes to such a possible phenonemon. Viewed as  an occurrence in natural language 'so long as the reference remains the same such variations of sense may be tolerated'.  20  (Frege had a rather low opinion of natural  language.) Strict!}' viewed, however, in terms of an ideal language, when it appears that  'Russell's Objections to Frege's Theory of Sense and Reference', in Essays on Frege, ed. E.D. Klemke, Urbana, 1968, pp. 337-348, p. 341. It may be thought that one could get round this difficulty by mentioning rather than using the name, for example by using the description 'the sense of W , where 'AT is the proper name. For an argument against this possibility see below, pp. 44-5. 17  18  Frege, 'Sense and Reference', p. 58.  'The Thought', trans. A . M . and M . Quinton, in Philosophical Oxford, 1967, pp. 17-38, p. 25. 19  2 0  Frege, 'Sense and Reference', p. 58.  Logic, ed. P.F. Strawson,  Direct Reference Theories  12  two senses are being attached to the same name one must conclude that the two speakers are not speaking the same language.  21  Frege is surely correct in arguing that there is no one concept which we all associate with any given proper name. Indeed it is hard to imagine, were we to attempt to idealize the language, just how such a singular concept could be arrived at. Given that we never reach a complete knowedge of the object, what one sub-set of all the properties and relations of, say, the average guy in the street, Albert Smith, could be employed as the associated sense of his name by his wife, his grade one teacher, his car mechanic and his income tax inspector? However, if the senses of proper names can vary from speaker to speaker then such senses are not reflective of linguistic competence but rather of the speaker"s knowledge and beliefs about the referent. That being the case, the possibility exists, indeed is extremely likely, that one of the speaker's beliefs about the referent is faulty. If this belief is reflected in the sense used by the speaker has he thus failed to refer? If the concept Mrs. Smith has of Jennie Jones includes her having fair hair, blue eyes, living in the white house, etc. and Jennie in fact has brown eyes then, since there is noone who matches the description, it would seem that in this case there is no referent. Faced with these difficulties some theorists have rejected Frege's view of the sense of a proper name as being made up of a fixed set or conjunction of properties and relations of the object. Instead it is argued that the sense of a term should be seen rather as a disjunction of properties and relations. Should it happen that one or more properties occurring in the disjunct do not in fact apply to the referent, reference may still occur so long as a sufficient number of properties occurring in the disjunct do so apply. The decision as to what constitutes a sufficient number of properties occurring in  21  Frege, 'The Thought', p. 25.  Direct Reference Theories  the disjunct is to be settled by common consent.  22  13  So, in response to the question as to  whether or not proper names are logically connected with characteristics of the object to which they refer, a proponent of this view, John Searle, answers 'Yes, in a loose sort of way'.  23  Even such a loose connection is denied by the direct reference theorists.  The Negative A r g u m e n t s The aim of the negative arguments offered by the direct reference theorists is to show that the Fregean view of the functioning of proper names is misconceived. The referent of a proper name, the direct reference theorist maintains, is not reached via a sense whose effect is to survey the domain of individuals for that which correctly fits certain criteria, criteria which are supplied by a definite description. The arguments are aimed at the 'cluster' view that holds that only a sufficient core of the properties individuated by the definite description need be satisfied by the object as well as the standard Fregean view that the object must satisfy all such properties. The arguments, per se, are not concerned with disputing the semantic aspect of sense, where sense is seen solely as a semantic mechanism associated with the name by means of which reference is achieved,  24  but deal with the results of the combination of this aspect with  one of the other aspects of Fregean sense. So we find arguments concerned with various cases in which the referent is reached by means of a description which specifies (i)  See, for example, John Searle, 'Proper Names', Mind, 67 (1958), 166-173, p. 171; and P.F. Strawson, Individuals, London, 1959, pp. 191-2.  2 2  2 3  Searle, 'Proper Names', p. 173.  Indeed Nathan Salmon goes so far as to argue that direct reference theorists accept that names have sense in this understanding of the term (Salmon, 'Reference and Essence, pp. 13-14, esp. n.8). While this may be true for some theorists (Kripke, for example), it is by no means clearly the case with respect to Donnellan who, I suspect, would substitute speaker's intentions for this aspect of sense. No specific arguments are offered by an}' of the theorists against this aspect of sense, however. 24  14  Direct Reference Theories  certain properties of the object, (ii) the speaker's concept of the object, and (iii) the speaker's beliefs about the object.  25  Against sense being a specification of properties of the object. The arguments advanced against the view that the definite description is furnished by some of the properties of the object are the modal arguments given in the main by K r i p k e .  26  For the Fregean who adheres to Frege's formal view, the sense of a  proper name such as 'Galileo' is, given the constraints of the epistemic aspect of sense, usually held to specify some of those aspects of the man which have made him wellknown, perhaps something like - the man who discovered that the earth revolves around the sun and was silenced by the Catholic Church. If the Fregean view is correct then it would seem that 'Galileo' just means the man who discovered that the earth revolves 27  arround the sun and was silenced by the Catholic Church.  28  So the sentence  (1) Galileo, if he existed, discovered that the earth revolves around the sun and was silenced by the Catholic Church. simply means Although his handling of the arguments is somewhat different than mine Salmon, Reference and Essence, pp. 23-31, provides an excellent discussion of the direct reference theorists' position. 2 5  In 'Naming and Necessity', in Semantics of Natural Language, eds. D. Davidson and G. Harman, Dordrecht, 1972, pp. 253-355, and pp. 763-869; reprinted as a book with additional preface, Cambridge, Mass., 1980 (all page references are to this book), see pp. 61-2, and pp. 74-8 among many others. 2 6  For convenience I shall talk here simply of the sense of a name. Strictly, this should be read as the sense of the name when used in a particular context. 2 7  Dagfinn F^llesdal remarks felicitously that since, for Fregeans, names just save us from repeating the whole description, '[they] could be called "names of laziness" just as Geech talked about pronouns of laziness' ('Reference and Sense', in Philosophy and 2 8  Culture,  Proceedings  of the XVIIth  Montreal, 1986, pp. 229-239.  World  Congress  of Philosophy,  Montreal,  1983,  Direct Reference Theories  15  The man who discovered that the earth revolves around the sun and was silenced by the Catholic Church, if he existed, discovered that the earth revolves arround the sun and was silenced by the Catholic Church. (1) therefore will be analytic and since an analytic statement is true by virtue of its meaning and true in all possible worlds in virtue of its meaning it will also be necessary and a priori (that's sort of stipulative)'. 29  (1) therefore will express a necessary truth. But, the argument continues, it is by no means necessary that Galileo got involved in scientific discoveries. He could have decided to become a farmer, settled in the country and concentrated his energies in perfecting a better yield from his olive trees. Such things are possible. Had they happened, Galileo would not have discovered that the earth revolves around the sun, would not have proclaimed it so vehemently, and so would not have been silenced by the Church. Since such a result is possible, (1) does not express a necessary truth. Adherents of the cluster view would agree that a sentence such as (1) does not express a necessary truth. Nevertheless, they would hold that it is necessary that a sufficient number of the properties must hold of the object. Kripke's argument, however, is to the effect that it is not necessary that any of the properties hold: Galileo need not have discovered that the earth revolves arround the sun nor need he have been silenced by the Church.  30  Similarly, if 'Galileo' means the man who discovered that the earth revolves around the sun and was silenced by the Catholic Church, then  2 9  See Kripke, Naming  and Necessity, p. 39.  This is not quite correct. Kripke buried in the description which are and, perhaps, living at the time However, a specification of such individual from all the others. 3 0  is willing to allow that there may be some properties essential to the individual; for example, being a man he did (see Kripke, Naming and Necessity, p. 62). properties would not be sufficient to pick out one  16  Direct Reference Theories  (2) If someone exists (or has existed) who discovered that the earth revolves arround the sun and was silenced by the Catholic Church then that person is Galileo. will also express a necessary truth. However, suppose that Galileo had gone off to tend olive trees, then, given the intellectual and religious climate of the time, the recent invention of telescopes, and so on, it is quite possible that someone else would have made the discovery and that he too would have insisted on telling the world about his discovery, and so he too would have been silenced by the Church. It is not necessary that only Galileo have done such a thing or have such a result befall him. Hence (2) does not express a necessary truth. As with 'Galileo' so with any proper name. Whatever the description given by the Fregean,  31  it is neither necessary that the individual has (or had) the properties  described or that whoever has (or had) those properties be that individual. Kripke's arguments are generated from two intuitively plausible claims which form the core of his position. The first is a metaphysical  claim - that the properties  specified by the definite description are such that the individual could have failed to have had them; that is, that they are not essential properties. This metaphysical claim rests on and is supported by a linguistic  claim that when speaking of the possibility that  Galileo may not have had any of the properties which he actually had the name 'Galileo' continues to refer in that possible world to the same individual that it does in the actual world.  32  It is the combination of these two views that enables Kripke to reach the  It is important to note that the property of being that very individual, the haecceity of Galileo, as Kaplan has dubbed it (see 'How to Russell a Frege-Church', Journal of Philosophy, 72 (1975), 716-727, pp. 722-3) is not the kind of property the Fregean has in mind as being captured by the definite description. For Kripke, a definite description furnished by such a haecceity would be weakly rigid, would pick out the same individual in every possible world in which he existed. 3 1  See Kripke, Naming and Necessity, p. 62. This merging of the two claims leads to certain undesired results in Kripke's positive views. See below, p. 26, n. 55. 3 2  Direct Reference  Theories  17  conclusion that (1) and (2) do not express necessary truths and so the sense of a proper name cannot be thought of as containing a specification of some of the properties of the object.  33  Against sense being a description of the speaker's concept of the object.  The argument given against the conceptual aspect of sense is very similar to the modal argument given above. If 'Galileo' means the man who discovered that the earth 34  revolves arround the sun and who was silenced by the Catholic Church, then it will be  Some neo-Fregeans, in the light of Kripke's criticisms, have moved from the standard Fregean view to that wherein the description which backs the name is held to specify not just some properties of the object but also that these properties are held of the object in the actual world (see, for example, L. Linsky, Names and Descriptions, Chicago, 1977, p. 84). For example, they might hold that 'Galileo' is backed by the description 'the man in the actual world who discovered that the earth revolves arround the sun'. Such a description would seem to be immune to the modal arguments since, given that Galileo did indeed make that discovery, there is no possible world in which Galileo does not satisfy the description. However a term such as 'actual' can itself be seen to function in two ways. As Salmon asks 3 3  Does the sentential actuality operator 'actually' describe the actual world... or does it directly and nondescriptionally indicate the actual world for semantic evaluation? {Reference and Essence, p. 44, italics mine). While Fregeans take the former view, that 'actually' functions as a description, direct reference theorists hold that, if it contributes to the semantics of the sentence at all, it is a directly referring term, (see Kaplan, 'Demonstratives', published with slight emmendations after circulating in draft form for many years in Themes from Kaplan, ed. by J. Almog, J. Perry, and H. Wettstein, New York, 1989, pp. 481-563, passim, esp. pp. 500-505. Hence the neo-Fregean has not resolved the issue by moving to such descriptions but has merely invited the direct reference theorists' criticisms at another level. (Kaplan seems to be moving away from this view somewhat in his latest discussion of the topic. In 'Afterthoughts' [in Themes from Kaplan, pp. 565-614], he argues that although 'actual' is not a purely descriptive term it is not a purely referential one either. Unfortunately, as this new work of Kaplan's has only just become available to me, I cannot do justice to it here and only certain salient points can be incorporated, and then only in the briefest fashion.) The argument is not explicitly given but can be constructed from the separate points made by the direct reference theorists. See, for example, Kripke, Naming and Necessity, pp. 73, 78-87. See also Salmon, Reference and Essence, pp. 27-9. 3 4  18  Direct Reference Theories  inconceivable that Galileo did not make such a discovery or have such a result befall him. It will be inconceivable in the same way as a bachelor being married is inconceivable. (1) will be knowable a priori because it is analytic. Likewise it will be inconceivable, given that the sum of properties involved is unique (it is a definite description), that anyone else did such a thing with such disastrous results. (2) also will be knowable a priori. Yet we can entertain the possibility that history could be wrong, that it could come to light that in fact it was not Galileo but someone else who made that discovery with such results. After all we do sometimes find out that things are in fact not how we think they are or as they appear to be. We learn that it was not Columbus but a Norseman who discovered America, that what is said of Jonah in the Bible is substantially false, that Everest may not in fact be the highest mountain in the world, that Mary's eyes are not blue but brown. Faced with the knowledge that our beliefs about a particular individual are often wrong, we can easily imagine circumstances which would result in our conception of the object being totally false.  35  Inasmuch as we can conceive of the individual not possessing the properties contained in the definite description and conceive of someone else having them, the properties contained within the definite description cannot be derived from our concept of the object.  3 6  Again, perhaps, not quite all. Perhaps it is inconceivable that Galileo should turn out to have been not a man but, say, a dog. (See Donnellan, 'Proper Names and Identifying Descriptions', in Semantics of Natural Language, eds. D. Davidson and G. Harman, Dordrecht, 1972, pp. 356-79, esp. pp. 366-7.) However, as before, such a description as would be left would not be unique. 3 5  Kripke accepts that there may be odd cases in which statements similar to (1) and (2) are, in some sense, known a priori. This would occur when the speaker fixes the referent of the name by means of a description, for example when someone says 'I shall call that heavenly body over there "Hesperus"', or 'Let us call whoever it is who is committing all or most of these murders "Jack the Ripper'". In such cases, statements such as 3 6  Direct Reference Theories  19  Against sense being a specification of the speaker's beliefs about the object. The arguments against the notion thiii the sense of a proper name specifies some (or all) of the speaker's beliefs about the object form the bulk of the arguments against the Fregean notion of sense given by the direct reference theorists. One such argument relies on the point that in many cases what we believe about the individual is not sufficient to yield a unique description. The cases range from individuals remote from us in space and time to those we have actually met. A l l many people know of Aristotle is that he was a Greek philosopher. They know nothing of his life or his philosophy. They cannot provide any description which would distinguish him from Plato (also known only as a Greek philosopher).  37  For many, all that is believed of Van Dyck and Van Eyck is  (i) Hesperus is over there. and (ii) Jack the Ripper is committing these murders. will, in some sense, be knowable a priori. (See Kripke, Naming and Necessity, p. 63, n.26, and p. 79.) The position is highly contentious. Kaplan (the post-'Quantifying-In' Kaplan, that is) embraces such contingent a priori truths without a qualm, casting aside even Kripke's qualification that such truths are knowable a priori in some sense. (See Kaplan, 'Dthat' [1970], published, after circulating in manuscript form for several years, in Syntax and Semantics 9: Pragmatics, ed. P. Cole, New York, 1978, pp. 221-243; reprinted in Contemporary  Perspectives in the Philosophy  of Language,  eds. P.A. French, T.E. Uehling  and H.K. Wettstein, Minneapolis (1979), pp. 383-400, page references are to this latter volume, see p. 397.) Kaplan gives as an example the name 'Newman 1', introduced to name the first child born in the 21st century (if there is one). For Kaplan the sentence 'Newman 1 will be the first child born in the 21st century' expresses a contingent truth yet it is knowable a priori. Donnellan however argues vigorously against such "exotic" a priori truths as he calls them. His position is that when the referent of a name, n, is fixed by means of a description, '<j>', then, while we know that the sentence 'n is <j)' expresses a (contingent) truth, we do not know the truth of what is expressed by that sentence. For example, although we know that 'Newman 1 is the first child to be born in the 21st century' expresses a truth, we do not know, of the child who is the first to be born in the 21st century, that he is the first to be born in the 21st century. (See 'The Contingent A Priori and Rigid Designators', in Midwest Studies in Philosophy  II: Studies in the Philosophy  of  Language, eds. P.A. French, T.E. Uehling, and H.K. Wettstein, Minneapolis, 1977, pp. 12-27, esp. pp. 20-21.) See also Kaplan, 'Demonstratives', p. 560, n. 76. See Kripke, Naming  and Necessity, pp. 80-1.  Direct Reference Theories  20  that they are famous painters, of Neils Bohr and Teller that they are physicists, of Cyndi Lauper and Whitney Huston they they are present day pop singers. Some may even have been in the presence of the individual and still not be able to provide a unique description. Donnellan gives as an example a child who is gotten up from sleep at a party and introduced to someone as 'Tom', who then says a few words to the child. Later the child says to his parents, 'Tom is a nice man'. The only thing he can say about 'Tom' is that Tom was at the party. 38  The child can offer no description of Tom that is unique  3 9  nor can he subsequently  recognize and point out Tom. Nevertheless, Donnellan argues, the child is referring to Tom. Often cases in which the name appears to have a backing of a definite description will, upon examination, fail. Suppose all that is known of Cicero besides his being a Roman orator is that he denounced Catiline. Here we would seem to have a good candidate for the definite description. However, the description contains the name 'Catiline'. Given the restraints of the conceptual facet of sense, what is expressed by the definite description must ultimately consist of properties. However, perhaps all that is believed of Catiline is that he was denounced by Cicero. But  If we do this, we will not be picking out anything uniquely, we will simply be picking out a pair of objects A and B, such that A denounced B. We do not think that this was the only pair where such denunciations ever occurred. 40  So, despite appearances, the description 'the man who denounced Catiline' is not unique.  3 8  Donnellan, 'Proper Names and Identifying Descriptions, p. 364.  Of course, others would have had to have talked to the boy as well else 'the man who talked to me at the party' would be unique. 3 9  4 0  Kripke, Naming  and Necessity, p. 81.  Direct Reference Theories  21  What are we to say of such cases? Both the standard Fregean and the cluster theorist would have it that in such cases reference did not occur. The direct reference theorist, pointing to the widespread occurence of such cases, appeals to our intuitions that reference is in fact occurring here. We can and do refer to Van Eyck, Teller and Cicero even though we lack a definite description of those persons. The Fregean picture does not present an accurate picture of how reference is achieved. In order to strengthen his case, the direct reference theorist appeals to cases in which our beliefs about the individual do generate a unique description, that description, however, being false of the individual. Kripke  4 1  cites the example of Jonah who,  according to Biblical scholars, really existed. None of the things attributed to him in the Bible, however, are true - he was never swallowed by a whale, nor did he go to Ninevah to preach and so on. On the Fregean view we have never succeeded in referring to Jonah. On such a view we could not even make the point that Jonah did none of the things attributed to him in the Bible.  42  Strengthening his case yet further, the direct reference theorist appeals to cases where not only can the speaker provide a unique description which he believes holds of the individual, a description in fact false, but also the description is true of some other  4 1  See Naming  and Necessity, pp. 86-7.  If 'Jonah' does not refer because of there being no object which fits the background description then, on the ordinary language Fregean treatment of vacuous terms as stressed by Strawson, 4 2  (i) Jonah did not get swallowed by a whale nor did he preach in Ninevah. would lack a truth value. On the formal Fregean treatment of such terms as developed by Rudolf Carnap (see Meaning and Necessity, Chicago, 1967, esp. pp. 35-39) wherein an aribitrary object is assigned as referent, (i) would always be true regardless of which object was assigned. (For example, it is true that the null set did not get swallowed by a whale or preach in Ninevah). On either view we cannot make the point that Jonah did none of these things.  22  Direct Reference Theories  individual. Donnellan provides us with an example concerning the name Thales'. We 43  know very little about Thales; what we do stems from the work of Aristotle and Herodotus. Given the paucity of our knowledge, the most likely candidate for the definite description would seem to be 'the Greek philosopher who held that all is water'.  44  Suppose, now, that Aristotle and Herodotus made a mistake about Thales or were setting up a straw man - Thales never in fact either said or believed that all is water. Suppose further, that there was someone else, an unknown dabbler in philosophy living at the time, who did hold such a view. The Fregean would have it that the referent of the name 'Thales' is whoever (or whatever) fits the definite description. The sense functions in such a way as to seek out that individual which fits the criteria given uniquely. So the Fregean would hold that if Donnellan's scenario were in fact the case, all along, in using the name 'Thales', we have in fact been referring to that unknown dabbler in philosophy. Such a result is proposterous. That there is another individual who in fact satisfies the description is purely accidental. Such a person, even if he does satisfy the description, 'is not "our" Thales'.  45  Given recent events, another example is easily drawn. It would seem that 'the tallest mountain in the world' would be a prime candidate for the description backing the name 'Everest'. Yet it has recently been contended that K2 may in fact be taller than Everest. What if that is the case? Under the Fregean view, every time we have used the name 'Everest' in order to talk about Everest, we have in fact been talking about K2. If 4 3  See Donnellan, 'Proper Names and Identifying Descriptions', p. 374.  Salmon (Reference and Essence, p. 30) mischieviously points out that the description which Linsky himself gives - 'the Eleatic philosopher who believed that all is water' is in fact false since Thales was not from Elea but Miletus. On Linsky's own view, therefore, he himself has never succeeded in referring to Thales! 4 4  4 5  Donnellan, 'Proper Names and Identifying Descriptions', p. 374.  Direct Reference  Theories  23  this were not bad enough, note that, since K2 was thought to be the second highest mountain in the world, everytime we have used 'K2' we have in fact been talking about Everest. Faced with such results the direct reference theorist concludes that it cannot be that the referent of a proper name is achieved by means of a desciption furnished by certain of our beliefs. We do manage to refer to the object even if all our beliefs about 46  the object are faulty. So the move to taking the sense of the name as a specification of the beliefs of the speaker about the object fares no better than when one of the other facets of sense was emphasized. Some other explanation of the phenomena of reference is demanded. The negative arguments offered by the direct reference theorists are not simply aimed at blocking the various candidates put forward by the Fregean as the source of the description. What the} attempt to show is that the picture of reference being 7  achieved by means of a search of a domain for an individual which fits certain criteria, specified from whatever source, is incorrect. A new picture of reference is called for. It is this which they attempt to provide in their positive arguments.  The Positive View.  Reference, we have seen, is held to be achieved not by means of a definite description but directly. The name 'Galileo' refers to the man Galileo regardless of any It could be objected that, faced with such results, one could conclude not that the Fregean view of language is at fault but that communication often goes awry: we may think we're talking about something when in fact we are not. Were it just the odd isolated instance I would agree. But given the scope and frequency of possible error it would be lucky if we ever succeeded in saying what we wanted to about an object. There must be a means by which, if nothing else, we can, by talking about someone, correct our false beliefs about him. We must be able to wonder what someone, whom we know nothing about, is like. If communication is only possible in situations in which we know all there is to know about what we are talking about then such learning would be impossible. 4 6  24  Direct Reference Theories  contingent facts, or believed facts, about him. Reference is not at the mercy of the vicissitudes of fortune. How then does a name function?  47  Kaplan, in 'Quantifying In',  introduces the notion of a standard name. Such a name necessarily denotes its object, 4S  the denotation being fixed 'on logical, or perhaps I should say linguistic grounds alone'.  49  So long as the conventions of language are held constant, '9' will denote the number 9 'under all possible circumstances'. Hence there cannot be a possible world in which '9' 50  does not denote the number 9. However, as introduced by Kaplan, only the names of certain abstract objects, principally numbers and expressions, were held to be standard names. It was Kripke who supplied the initial account of the functioning of the names of both physical and abstract objects. Names, for Kripke, are rigid designators: a designator rigidly designates a certain object if it designates that object wherever the object exists; if, in addition, the object is a necessary existent, the designator can be called strongly rigid. 51  So, since numbers are necessary existents, their names are strongly rigid: they will designate the same number in all possible circumstances, in all possible worlds. The names of objects whose existence is not necessary, e.g. Galileo, are simply rigid: they designate the same object in all those possible worlds in which the object exists. Other  For convenience I shall follow Kripke and talk simply of names where strictly what is meant are proper names. 4 7  4 8  In Words and Objections: Essays on the Work of W.V. Quine, eds. D. Davidson and  J . Hintikka, Dordrecht, 1969, pp. 206-242. 4 9  Kaplan, 'Quantifying In', p. 154.  5 0  Kaplan, 'Quantifying In', p. 195.  Kripke, Naming and Necessity, pp. 48-9. See also 'Identity and Necessity', in Identity and Individuation, ed. M.K. Munitz, New York, 1971, pp. 135-164, esp. pp. 145-6. 5 1  25  Direct Reference Theories  designators, those which do not designate the same object in other possible worlds are termed nonrigid or accidental designators. Intimately connected with Kripke's view of the functioning of names is his view on possible worlds.  52  He argues against the view that possible worlds are discovered and  then observed like 'a distant country that we are coming across, or viewing through a telescope'. Rather a possible world is stipulated, given by the descriptive conditions we 53  associate with it. Further, a possible world need not be given purely qualitatively. We can start with the furniture of the actual world and modify it counterfactually. So, if what we start with is the man Galileo, and if we then go on to ask whether he could have failed to have made the famous discovery, it is stipulated that the possible world contains the man Galileo. We then describe how it could come about that he did not make his discovery (for example, Galileo decides to become a farmer, settles in the country and so on). In so doing we are describing a portion of a possible world, a world which contains the man Galileo. Should we be unable to describe a possible world in which Galileo fails to possess a certain property then that property is an essential property of Galileo. With the distinction between rigid and nonrigid designators in place, how is it determined whether  a singular  term is a rigid  designator  or  not? The  usual  interpretation of Kripke's position here is that the following intuitive test should be applied. Start with the designation of the term in the actual world and ask of that object or person whether it could be such as to fail to be as designated by the term. For example, with respect to 'the man who won the election in 1968' (in the U.S.), ask of the man who won the election in 1968, Nixon, if it is possible for him not to have won the  See Kripke, Naming and Necessity, pp. 39-53; and 'Identity and Necessity', pp. 147-8. Kripke, Naming  and Necessity, p. 42.  Direct Reference Theories  26  election in 1968. Clearly this is possible - Nixon might never have gone into politics. So 'the man who won the election in 1968' is not a rigid designator. Is it possible for Nixon not to be Nixon? Of course there will be possible worlds in which Nixon does not exist. 'Nixon' therefore is not strongly rigid. But what of those worlds in which he does exist? Could Nixon there fail to be Nixon? N o . designator.  54  'Nixon' therefore is a (weak) rigid  55  There are problems, however, with Kripke's treatment of names. Not only is a distinction drawn within the realm of names itself so that we have two differing kinds of names (those which are weakly rigid and those which are strongly rigid) but on Kripke's account, many designators which we would normally consider to be definite descriptions pass the intuitive test. Take, for example, 'the offspring of gametes G' which, let us say, designates Nixon. Could the offspring of gametes G fail to be the offspring of gametes  Note that the question is not whether it is possible for Nixon not to be called 'Nixon' (clearly that is possible), but whether the man Nixon might not have been Nixon.  5 4  In fact, as Joseph Almog has pointed out (in 'Naming without Necessity', Journal of Philosophy, 83 [1986], 210-243, p. 222, n.9) there are two intuitive tests proposed by Kripke. (Almog is here developing a distinction first noticed by Kaplan, see Kaplan, 'Demonstratives', p. 4 9 3 , n.16.) The other test is indicated in the following passage from 5 5  Naming  and  Necessity:  One of the intuitive theses I will maintain in these talks is that names are rigid designators. Certainly they seem to satisfy the intuitive test mentioned above: although someone other than the U.S. president in 1970 might have been the U.S. president in 1970 (e.g. Humphrey might have), no one other than Nixon might have been Nixon, (p. 48) (See also Kripke, 'Identity and Necessity', pp. 148-9.) Whereas the first test concerns whether a could have failed to be a, here the test is could anyone else other than a have been a? As Almog points out, the two tests are not identical. Take as an example 'Kripke's father'. Could Kripke's father fail to have been Kripke's father. Yes - Kripke might never have been born. So on the first test 'Kripke's father' is not a rigid designator. Could someone other than the man who is Kripke's father be Kripke's father? On Kripke's view, given his essentialism, no. So on the second test 'Kripke's father' would be a rigid designator. The second test would seem to involve finding sufficient conditions for someone being a certain person; asking, for example, if x were F would he be Nixon? As such it is at odds with Kripke's overall position.  27  Direct Reference Theories  G? Given Kripke's essentialism, no.  5 6  So, on Kripke's account, 'the offspring of gametes  G' is a rigid designator on a par with 'Nixon'. In the Preface added to the 1980 edition of Naming  and Necessity  Kripke  introduces a further distinction, that between de jure and de facto rigidity: the distinction between 'de jure rigidity, where the reference of a designator is stipulated to be a single object, whether we are speaking of the actual world or of a counterfactual situation, and a mere 'de facto' rigidity, where a description 'the x such that Fx' happens to use a predicate 'F" that in each possible world is true of one and the same unique object. 1  57  If a designator is rigid de jure it will not just turn out to be the case that the referent of the term remains the same in all possible worlds. Cases in which, upon consideration along essentialist lines, we decide that it is not possible for the object which is a not to be a, the designator will not be rigid de jure. Hence 'the offspring of gametes G' will not be so rigid.  5 6  If you balk at such essentialism consider the description 'the smallest prime number'.  5 7  p. 21, n. 21. Kripke continues, Since names are rigid de jure I say that a proper name rigidly designates its referent even when we speak of counterfactual situations where the referent would not-have existed.  Note that Kripke is not here denying that the distinction between strong and weak rigidity applies to names. Rather, he is pointing out another way of thinking about our talk of possible worlds. In speaking of a possible world in which the referent of a designator does not exist, we use our language, 'with our meanings and our references' ([1972a], pp. 77-8). Hence with respect to that world the name still designates the same referent for us, here. If you say 'suppose Hitler had never been born' then 'Hitler' refers here, still rigidly to something that would not exist in the counterfactual situation described. ([1980], p. 78.) Nevertheless, in that possible world, the name does not designate the referent. The name is still weakly rigid.  28  Direct Reference Theories  Is Kripke's new distinction sufficient for distinguishing between names and definite descriptions? Unfortunately, it would not seem so. Of the distinction between singular terms which designate by means of a definite desciption and those that do not Kaplan remarks of the latter: the intuitive idea is not that of an expression which turns out to designate the same object in all possible circumstances, but an expression whose semantical rules provide directly that the referent in all possible circumstances is fixed to be the actual referent. 58  If Kripke's 'stipulation' is to be read as 'being given by semantic rules' then his de facto I de jure distinction will not be sufficient to capture the name / description distinction. Consider the description 'the offspring of gametes G in the actual world'. Such a description will be rigid. Suppose the offspring to be Kripke: then there is no possible world in which that description does not designate Kripke. It will, though, be weakly rigid: it is possible for the actual offspring of gametes G not to have existed, Kripke might never have been born. But will it be de facto rigid? No. It will be rigid, as Almog puts it  in virtue of the semantic rules of the language, in virtue of the truth conditions stipulated for all formulas containing 'actual'. 59  The situation is perhaps more easily seen with respect to 'the smallest prime'. To see that 'the smallest prime' is modally stable it is necessary to rely on arithmetic essentialism. However the modal stability of 'the actual smallest prime' is settled by semantics alone.  5 8  'Demonstratives', p. 493.  Almog, 'Naming Without Necessity', p. 223. See also Kaplan, 'Demonstratives', pp. 502-3, and 'Afterthoughts', p. 577, esp. n. 25. 5 9  29  Direct Reference Theories  What is needed in order to distinguish names from definite descriptions is not an account of the differing results of their being used to designate but an account of the way those results are obtained. This was, it will be remembered, the whole thrust of the Negative Arguments offered against the Fregean. We need an account of content - of the way in which a name occurring in an asserted sentence makes a differing contribution to what is said than does a definite description; of the way in which 'Nixon is a Republican' says something different than does 'The (actual) offspring of gametes G is a Republican'. Such an account has been developed by Kaplan. Firstly what is said by the use of a term in an asserted sentence may differ on different occasions of use. If I say 'The woman next door is spraying her roses' what I say will differ if I am standing in mj' own back yard or miles away at a friend's home. The context in which the assertion is made is relevant in determining what is said. The context of utterance must not be confused with the circumstances of evaluation. What is said, a proposition, is expressed by an assertive utterance in a given context. The proposition is then evaluated with respect to a certain circumstance. The circumstance involved may be that obtaining at the context of utterance, i.e. the actual world, or it may be a counterfactual situation. Propositions are structured entities looking something like the sentences which express them. For each occurrence of a singular term in a sentence there will be a corresponding constituent in the proposition expressed. 60  Kaplan, 'Demonstratives', p. 494.  30  Direct Reference Theories  For definite descriptions, which for the purposes at hand I will take to be descriptional terms,  61  the content will be 'some sort of complex constructed from various attributes by  logical composition'. The proposition expressed, for example, by an assertion of 62  The U.S. President in 1970 is a Republican. and by that of The actual offspring of gametes G is a Republican. will contain differing constituents corresponding to the singular term; constituents which determine 'for each circumstance of evaluation, the object relevant to evaluating the proposition in that circumstance'.  63  These propositions  will contain, if you  will,  blueprints for determining the designation. It may be that a definite description need not function as a descriptional term. Donnellan has highlighted two very different uses which he terms the referential use and the attributive use of definite descriptions, 'the fa ' can be used to talk about a particular object or person the speaker has in mind, the description being but one means among many capable of achieving that end. This is the referential use. Alternatively, the description may be used attributively - used in order to talk about whoever or whatever is the <j>. Here, unlike in the referential use, the description occurs essentially. Donnellan argues that the truth values of the assertion may differ depending upon how the description is being used. (See Donnellan, 'Reference and Definite Descriptions', Philosophical Review, 75 (1966), 281-304, esp. p. 283; and 'Putting Humpty Dumpty Together Again', Philosophical Review, 77 (1968), 203-215.) Though no-one would dispute that Donnellan has highlighted a distinction in usage, whether or not the truth conditions of the utterance are affected is a matter of much dispute. (For Kripke's views on the matter see 'Speaker's Reference and Semantic 61  Reference', in Contemporary  Perspectives  in the Philosophy  of Language,  eds. P. French,  T. Uehling, and H. Wettstein. Minneapolis, 1979, pp. 6-27; for Kaplan's view, see 'Demonstratives', passim.) Fortunately, for the purposes at hand, the matter need not be decided here. While I shall draw the distinction as between names and definite descriptions, those sympathetic to Donnellan's distinction can take this only to be a distinction between names and the attributive use of definite descriptions. 6 2  Kaplan, 'Demonstratives', p. 494.  Ibid. Note that since propositions are expressed by an asserted sentence in a context, and since the context will be in the actual world, the word 'actual' in the second example is redundant. 6 3  Direct Reference Theories  31  However, in cases where the singular term is a name, the corresponding constituent in the proposition is the object itself. Hence in the case of an assertive utterance of 'Nixon is a Republican' the proposition expressed contains the flesh and blood man, Nixon. Such a proposition is of the form  ^ is a Republican,  structured entity made up of the object which is the referent,  64  that is, a  j£ being Nixon himself,  together with the content expressed by the predicate. Such propositions are called s i n g u l a r propositions. The identity of singular propositions, given that the predicate is the same, is a function of which individual is referred to.  65  Hence the distinction is drawn between those terms which refer directly and those which designate by means of a description. As we have seen, it is not drawn i n terms of whether or not the designator always designates the same object but i n terms of the differing types of propositions expressed by sentences containing such terms. Such a view captures well the original intuition that whereas definite descriptions may designate by  means of providing attributes which enable the designation to be  determined, names do not function in that way but refer directly. Kaplan's introduction of content has consequences with respect to modal statements which distinguish his position from that of Kripke. Since what is evaluated, in cases where the sentence asserted contains a proper name, is a proposition containing the object which is the referent, in cases where the sentence is embedded in a modal context, where the evaluation is to be carried out in another possible world, then that same singular  proposition  will be evaluated there. This is the case even where the world  concerned is such as to lack the object which is the referent. That is, the name will refer  In the discussion which follows, propositions will be distinguished by being underlined. Singular propositions will sometimes be delineated by means of an underlined sentence containing a proper name but often, where it is useful to remember that the constituent of the proposition is the object itself, a stick figure will be employed. 6 4  6 5  See Donnellan, 'Speaking of Nothing', Philosophical  Review, 83 (1974), 3-31, p. 12.  Direct Reference Theories  32  to that object no matter what world is subsequently involved in the evaluation, even those where the object does not exist. Hence, on Kaplan's view, a name designates the same object with respect to all possible worlds - even those where the object does not exist.  6 6  So Kaplan reaches a different conclusion on this point than Kripke, who, it will be remembered, holds that names of contingent objects are only weakly rigid, they only refer to the object in worlds where the object exists. However both theorists reach the same conclusion on the epistemological status of identity statements. Statements such as 'Hesperus = Phosphorus' are not known a priori. It was an empirical discovery that Hesperus is the same planet as Phosphorus. It is a surprise to find out that Laura Jones, the marketing consultant whom one only meets professionally, is none other than Mrs. Robinson who lives in the corner house. Nevertheless, despite the fact that the truth of such statements is only known a posteriori, they are, it is commonly agreed, necessary. On Kripke's view, with respect to an identity statement 'a = fe', the name 'a' designates the same object in all those worlds where the object exists, likewise with 'fe'.  67  If 'a = <y is true then 'a' and 'fe' both designate the same object in the actual world. Hence 'a' and 'fe' must designate the same object in all those worlds in which that object exists. Hence 'a = fe' is necessary, but only in the weak sense of being true in those worlds where the object exists.  68  Nevertheless if 'a = 6' is true then the statement,'If~a exists  Hence this view of names is a direct descendant of those earlier standard names. See also Kaplan, 'Bob and Carol and Ted and Alice', in Approaches to Natural Language, eds. K . J . Hintikka, J . M . Moravsik and P. Suppes, Dortrecht, 1973, pp. 490-519, esp. Appendix IX, pp. 501-3, for an earlier discussion of this point. 6 6  I am assuming that 'a' (and 'fe') do refer in the actual world, else the statement could not possibly be necessary. 6 7  As to whether 'a = fe', if necessary in the weak sense, is true, false, or lacking a truth value in those possible worlds in which the object does not exist, Kripke notes 'The same three options exist for "Hesperus is Hesperus" and the answer must be the same as in 6 8  33  Direct Reference Theories  then a=b' is true in every possible world, that is, the statement is necessary though only knowable a posteriori. For Kaplan since ' a ' refers to the same object in every possible world, 'a = 6' if true is necessarily true in the strong sense. Hence a statement can be necessarily true yet knowable only o posteriori. So, despite a difference in the treatment of modal statements due to Kaplan's introduction of content, despite a somewhat different treatment of necessity in particular, both theorists sharply distinguish between the status of the truth conditions of a statement and that of our knowledge and beliefs about it. Such a distinction, which appears not only necessary but intuitive in the treatment of the functioning of proper names, gives rise to certain problems, problems so great that, as one commentator has put it, it would seem that the direct reference view 'founders on these rocks'.  69  It is to  such problems that we now turn.  the case of "Hesperus is Phosphorus'" (Kripke, Naming and Necessity, p. 110, n.52). See also Kaplan, 'Afterthoughts', where he quotes from a letter from Kripke as follows: 'a designator cf of an object x is rigid, if it designates x with respect to all possible worlds where x exists, and never designates an object other than x with respect to any possible world, '(p. 569)  Kaplan, speaking for Kripke, adds This definition is designed to be neutral with regard to the question whether a designator can designate an object at a world in which the object doesn't exist (pp. 569-570). Howard Wettstein ('Has Semantics Rested on a Mistake?', Journal of Philosophy, 83 (1986), 185-209, p. 186) quoting Alvin Plantinga ('The Boethean Compromise', 6 9  American  Philosophical  Quarterly, 15 (1978), 129-138, p. 130.  34  2. THEORIES OF BELIEF ATTRIBUTIONS  Belief attributions present grave problems for the direct reference theorist. Believing is a relationship which holds between the believer and that which is believed. The philosophically intuitive position as to the nature of that which is believed, the object of belief, is reached by reflecting along the following lines. If someone, B, believes that p, he can communicate that belief by asserting 'p'. For example, if John believes that Laura Jones has gone to Toronto he can tell someone what he believes by asserting (1) Laura Jones has gone to Toronto. So what John believes, it would seem reasonable to conclude, is that which is expressed by his assertion, (1). So for B to have a belief that p is for B to stand in a certain relationship with the proposition that would be expressed were he to assert 'p' (in the same context). Correspondingly, if someone hears A (seriously) assert 'p', he is justified in concluding (what better evidence ultimately could he have) that ^4 believes what he has just said - that is, that he believes what has been expressed by his assertion of 'p'. For example, if John's father hears John (seriously) assert (1) he can conclude  (2) John believes that Laura Jones has gone to Toronto. So it would seem that, in asserting (2), John's father is attributing to John a belief characterized by the proposition expressed by (1); more precisely he is asserting that John stands in the belief relation to that proposition. So, the conclusion is reached, an  Theories of Belief  Attribution  35  assertion of the form 'A believes that p' asserts that A stands in the belief relationship with the proposition that would be expressed by an assertion of 'p' in the same context. One of the notorious problems which face the direct reference theorist with respect to such assertions concerns the substitution of co-referential names. On the direct reference view, that which is attributed to John in (2) is a singular proposition containing the flesh and blood individual Laura Jones. However, suppose that Laura Jones has another name. While for professional purposes she has kept her maiden name, 'Laura Jones', at home she uses her married name, 'Mrs. Laura Robinson'. Both names are names of that one individual, both names refer to her. Hence an assertion of (3) John believes that Mrs. Laura Robinson has gone to Toronto. must also be true. The problem arises that it is entirely possible that the result of substituting one co-referential term for another in the attribution could be met with a robust denial of the resulting attribution by the believer. For example, perhaps John knows Laura Robinson by sight, often seeing her puttering about her garden. However, he has never talked to her and does not realize that she is none other than Laura Jones whom he has only dealt with by telephone (by which means he has learned of her impending trip to Toronto). Perhaps he saw Mrs. Robinson in her garden this very morning. He might then well deny the truth of (3) while continuing to assent to (1). Not only does the direct reference view seem to run roughshod over the believer's denials, a further question emerges. If the direct reference view is correct, if what John believes is simply the proposition expressed by (1), and if (3) asserts that very same proposition, then how can John deny the truth of (3)? Obviously he denies it because he does not realize that Laura Jones and Mrs. Laura Robinson are the same  Theories of Belief  36  Attribution  person. But this is no answer. How can he fail to realize that they are the same person? A further problem for the direct reference view concerns vacuous terms. The assertion of the sentence 'Fa', where 'a' is a proper name having no referent, results in there being no proposition expressed. As there is no object referred to by 'a', then there is no corresponding constituent which could occur in a proposition. Such a view has repercussions for belief attributions. As an example, suppose that John has just crashed his car but, in order to prevent his parents learning the worrying truth, he has invented a buyer by the name of Martin O'Reilly who, says John, gave him $400 for it. Suppose John's father remarks (sincerely) to his wife (4) Martin O'Reilly got a good deal when he bought John's car for $400. It would seem that, trusting soul that he is, John's father does indeed have a (relevant) belief. But since 'Martin O'Reilly' is a vacuous name, no proposition is expressed by (4) and there is therefore no proposition which is the object of John's father's belief. So the direct reference view would seem to lead to the highly unintuitive conclusion that John's father has no (relevant) belief: that (5) John's father believes that Martin O'Reilly got a good deal when he bought John's car for $400. is not true.  1  Kripke has argued that his position, as set forth in Naming  and Necessity does  not lead to the conclusions reached above. Rather, he argues, his position leaves open 2  It is not always noted that (5) is not false either but rather, since there is no proposition expressed by (4), lacks a truth value. 1  See the Preface, Kripke, Naming and Necessity, pp. 20-21, and ' A Puzzle about Belief, in Meaning and Use, ed. A . Margalit, Dordrecht, 1979, pp. 239-283, esp. p. 273, n.10. 2  Theories of Belief  37  Attribution  the question as to whether or not names function in the same way in belief contexts as they do in direct contexts. Kripke himself seems to have entertained the possibility that names in such contexts might function in a Fregean manner.  3  Certainly Kripke's  positive view does not lead to the conclusions reached above. His arguments concern the functioning of names in direct and modal contexts only, and his position does not depend upon the introduction of propositions. Further, distinguishes between necessity and a prioricity:  as has been seen, he  sharply  statements such as 'Hesperus  =  Phosphorus' are metaphysically necessary but a posteriori. However it is hard to see how the strongest of the negative  arguments against the Fregean, the epistemic, could  survive intact were some extra cognitive element to be introduced with respect to the functioning of names in belief contexts. In particular, recall the argument that, with respect to many names, 'Manet' and " V a n Dyck' for example, most of us do not possess sufficient beliefs about the individuals named to furnish a description which would pick out the referent uniquely. If names were to function in a Fregean way in belief contexts, much of the intuitive impetus for the theory of direct reference would have to be abandoned. Yet the argument by the direct reference theorists that the user of a name need not believe anything about the object named which would enable him to characterize it uniquely, while appearing intuitively correct with respect to reference in direct contexts, denies that extra cognitive element, the harnassing of which in some way would seem the only means by which to block the unintuitive conclusions with respect to belief contexts. Faced with such conclusions, a common initial reaction of many adherents of J . Almog has argued that the direct reference theory can, and should, be separated from questions of content in both modal and epistemic contexts. (See 'Form and Content', Nous, 19 (1985), 603-616. That is, as names function for the Fregean in direct contexts. Kripke, however, abandons such flirting with the Fregean view, see Kripke, 'A Puzzle About Belief, p. 244, and p. 273, n.10. 3  Theories of Belief  38  Attribution  the direct reference view was apparent indifference. Some explicitly bit the bullet, 4  arguing that if B believes that o is F and if ' a ' and '6' are co-referential then, despite his protestations, B does indeed believe that b is F, and that if the characterization of what purports to be what B believes contains a vacuous name then, despite B's assurances that he does indeed have a (relevant) belief, he has no such belief.  5  Many have seen such results of the application of the theory as its death knell. How, it is wondered, can the direct reference theorist be so insensitive to these issues? Are these not the very problems which prompted Frege's introduction of the notion of sense in the first place? How can the direct reference theorist argue so vigorously 6  against the Fregean and ignore his own failure to cope with such fundamental problems, problems which, it is argued, can be handled well under the Fregean theory? In order to see just how much of a case the Fregean has here let us, for the sake of argument, set the Negative Arguments aside for the moment and examine the Fregean treatment of belief contexts.  The Fregean Solution. When a singular term occurs in an indirect as opposed to a direct context then, for Frege, it refers not to its customary referent but rather to its customary sense. Substitution of one name, '6', for another, 'a', in a belief context will, therefore, only preserve the truth of the assertion if 'b' has the same indirect referent as ' a ' , that is, the  Some perhaps were in some way relying on Quine's distinction between relational and notional readings of belief attributions. See below, pp.49-52. 4  5  See, for example, Donnellan, 'Speaking of Nothing', passim.  Strictly speaking, no, they are not. Frege first introduced the distinction between sense and reference for mathematical equations, then sentences - its application to proper names happily falling out from that. Frege's interest, at the time of the introduction, was primarily mathematical. See Frege, 'Function and Concept'; and Baker and Hacker, Frege: Logical Excavations, pp. esp. p. 6  278-283,  282.  Theories of Belief  Attribution  39  same customary sense. As was noted above, two names possess the same customary 7  sense if and only if it is not possible for someone who understands two assertions, whose only difference is the substitution of one name for the other in the sentences used, to accept one assertion but reject the other. Since John accepts (2) but not (3), the names 'Laura Jones' and 'Mrs. Laura Robinson' must possess different senses, and so different indirect referents. Hence one cannot be substituted for the other in a belief context if truth is always to be preserved. The problem of vacuous names is solved by the fact that, for Frege, a name which has no referent may nevertheless possess a sense. When a vacuous name occurs in a belief context therefore, it may well have a referent, its customary sense. So, although there is no referent of the name 'Martin O'Reilly' (in a direct context), the name does have a sense and it is this sense which occurs within the object of belief. Hence there is a belief which John's father has. Frege's use of the notion of sense to dispell the problems of belief contexts appears at first glance breathtaking in its simplicity and power. However, upon closer examination serious problems, apart from those outlined in the Negative Arguments, emerge. With respect to vacuous terms it is not at all clear, given Frege's overall theory, how a name which lacks a referent can have a sense. Perhaps the aspect of sense most amenable to the problem of vacuous names is the epistemic. When this aspect of sense is stressed Frege is seen at his most Cartesian. We are separated from objects in the 8  world, can only get at them when using language, via the medium of sense. Of course Frege insists that senses are not psychological: they are sharply distinguished from  7  See p. 9.  Howard Wettstein, for example, interprets Frege in this way. See 'Cognitive Significance without Cognitive Content', Mind, 97 (1988), 1-28, esp. pp. 5-8. 8  Theories of Belief Attribution  40  ideas in that they do not belong to the individual but are public. Nevertheless, as Wettstein argues Consider...the Fregean semantic perspective vis-a-vis the Cartesian 'mirror-of-nature' tradition in the philosophy of mind. That tradition, it is often noted, sees the mind as set against nature, as the repository of images and conceptual representations of things. Pieces of language become meaningful by being associated with the conceptual representation. Leaving aside the question of images, and the probably related Fregean 'ideas', isn't this picture at least a very close relative of Frege's? 9  If we are separated from the world by the realm of sense then it would seem possible that, just as we could possess a concept to which no object corresponded, so there could be a name with a determinate, unique sense which lacked a referent. However, while a stressing of the epistemic aspect of sense would seem to accomodate the position that a name may have a sense but no referent, a difficulty emerges on the level of Thought. Given the identification of sense with the object believed, e.g. if we believe that John is tall then what occurs in the object of belief is the sense of 'John is tall', the position that a vacuous name may possess a sense results in their being determinate and wellformed  10  thoughts which lack a truth value. As Evans argues  What can it mean on Frege's, or on anyone's principles, for there to be a perfectly determinate thought which simply has no truth value? Remember that the notion of thought that Frege was intending to use had strong links with notions embedded in ordinary propositional-attitude psychology - the notions of belief, knowledge, memory, information, judgement and so on. If someone understands and accepts a sentence containing an empty name, then, according to Frege, he thereby forms a belief; not a belief about language, but a belief about the world. But what  9  Wettstein, 'Cognitive Significance Without Cognitive Content', p. 7.  They do not, for example, contain properties which are not defined for the subject as perhaps in The tree smiled. 10  Theories of Belief  Attribution  41  sense can be made of a belief which literally has no truth value - which is neither correct nor incorrect. 11  A further problem which arises when the epistemic aspect of sense is given priority concerns the connection of language to the world. Thoughts, for Frege, exist in the third r e a l m ; some are true, some false, and others are neither true nor false. To 12  posit Thoughts which lack a truth value, while permitting the tie to language to be maintained (they can all be 'clothed' in sentences), would seem to sunder any logical connection between the third realm and the world: that a Thought possess a truth value becomes a purely contingent matter. Yet, for Frege, we use language to talk about the world. How, if Thoughts need not be true nor false, can this connection be secured? As Michael Dummett remarks Granted that, in general, the possession of a sense does not guarantee the existence of a referent, and granted that an object can be given to us only in some particular way, namely by our grasp of a suitable sense, how can we ever establish that, by means of an expression with a certain sense, we have in fact succeeded in referring to an object, that there really is an object for which that expression stands? 13  To lay undue stress on the epistemic aspect of sense is to ignore much of Frege's discussion of the notion. Consider his position that the sense of a name stems from some of the actual properties of the object. When the sense is grasped we acquire a way of  1 1  Evans, The Varieties of Reference, p. 24.  That is, not the realm of ideas (which is subjective and psychological) nor that of objects in the world but another. Members of the third realm are like ideas in that they cannot be perceived by the senses but are like objects in the world in not being subjective. See Frege, 'The Thought', p. 29. 1 2  The Interpretation of'Frege's Philosophy, and Oratio Obliqua', p. 142. 13  Cambridge, Mass., 1981, Ch. 6, 'Indexicality  Theories of Belief  42  Attribution  regarding the object. When this aspect of sense is stressed Frege is seen to be a realist. Dummett is the prime example of this way of interpreting Frege.  14  He argues  For Frege, at least when we attend to the healthier of the divergent strands within his notion of sense, the sense of a proper name is the way we arrive at the object, but not conceived as a means to a separable end: the apprehension of the object is not an outcome that may be detached from the process that led to it. From this standpoint sense is better understood as the manner in which we pick out the object than as the route we take to it. We are never given an object, complete in itself; we can think about it, speak of it or apprehend it only as presented to us in some particular way... the thesis does not threaten realism, any more than to say that we can at any time see a building only from a particular angle casts doubts on the proposition that we can see buildings. 15  However, if sense is seen as a specification of some properties of the object, if what we grasp is a way of regarding the object, then how can a name which lacks a referent possess a sense? There is no object for the sense to specify the properties of, there is nothing to be regarded. This aspect of sense would seem incompatible with the position that vacuous names may possess a sense. The Fregean would seem to be caught in a bind. That a name could have a sense but no referent would seem more explicable given the epistemic aspect of sense as it is this aspect of sense which stresses the connection between the name and its sense. However such an explanation is only forthcoming if the logical connection between the  See Frege, Philosophy of Language, London, 1973, esp. Ch. 5, and Dummett, The Interpretation of Frege's Philosophy, esp. Ch. 6. 1 4  Dummett, The Interpretation of Frege's Philosophy, p. 132. Intimately connected with such divergent interpretations of the notion of sense is the varying emphases given to Frege's views on the prioricity of a Thought and its constituents. On the one hand Frege argues that that which is grasped is a Thought which is then decomposed to reach the senses of its component parts (see Frege, 'Function and Concept', pp. 24-5; Postumous Writings, p. 253). This is opposed to the constructivist view that the Thought is created by combining the senses corresponding to the elements of the sentence (see Frege, Postumous Writings, p. 243). For a discussion of this point vis a vis the realism / antirealism debate see Hans Sluga, 'Frege and the Rise of Analytic Philosophy', Inquiry, 18 (1975), 471-87, esp. pp. 479-80; Dummett, 'Frege as a Realist', Inquiry, 19 (1976), 455468; and Sluga, 'Frege's Alleged Realism', Inquiry, 20 (1977), 227-242. 15  Theories of Belief  43  Attribution  sense and the referent is severed. Such a move not only introduces difficulties at the level of thought, it violates Frege's conviction that we use language to talk about the world. On the other hand, if the connection to the world is secured, if the relationship between sense and reference is made tight by arguing that sense is a specification of some of the properties of the object, then how a name can be vacuous yet possess a sense becomes inexplicable.  16  A further problem for the Fregean position on indirect contexts stems from the fact that, from the truth of, for example, (i) Aristotle taught philosophy and  (ii) Jones believes that Aristotle taught philosophy we can correctly conclude that Jones believes something true. But if the referent of 'Aristotle' in (ii) differs from that of 'Aristotle' in (i) then the conclusion can only be reached if a logical relation holds between the referent of the name in the two premises, i.e. between the sense and the referent of a name. As Russell demands,  17  what the  Fregean must provide, if he holds that the referent of a name differs when the name  Fregeans have, of course, been hard at work to amend Frege's overall theory in such a way as to accomodate vacuous names which possess a sense. See, for example, John McDowell, 'On the Sense and Reference of a Proper Name', Mind, 86 (1977), 159-85, 16  esp. pp. 172-5; Evans, The Varieties of Reference; Dummett, The Interpretation  of Frege's  Philosophy, esp. Ch. 6. The various reformulations are in much dispute. The emendations would seem to discard vital elements of Frege's overall position so that the resulting position lacks cohesion. Clearly the matter needs a closer examination than can be given here. M y point, however, is not that Frege's overall theory cannot in principle be emended to accommodate vacuous terms but rather that the question as to whether it can or not remains unresolved. In 'On Denoting', in Logic and Knowledge, ed. R.C. Marsh, London, 1956, pp. 41-56. See also, for the interpretation of the passages concerned upon which I am relying, S. Blackburn and A . Code, 'The Power of Russell's Criticisms of Frege: "On Denoting pp. 48-50"', Analysis, 38 (1978), 65-77. 17  Theories of Belief  Attribution  44  occurs in an indirect context, is an account of that logical relation, i.e. that which holds between a name, its sense(s) and an object in the world. In order to provide such an account the Fregean cannot be content simply to distinguish, by means of the Intuitive Criterion of Difference, when two names differ in sense: the sense of the name must feature in the account of the logical relation. However, as was noted above, we cannot talk about the sense of a name by using it: 18  whenever we use it we end up talking about the referent. Nor can we adopt the other stategy and simply mention it, for example, using the description 'the sense of "Aristotle"'. To posit that the logical relation is that which holds between the man Aristotle and the sense of 'Aristotle' is, as Russell maintains, to posit a relationship which is 'merely linguistic through the phrase'.  19  A particular lexical item, e.g.  'Aristotle', may be used to name many different people and in so doing may have many different senses attached to it. .To posit a relationship between Aristotle the man and the sense of the lexical item 'Aristotle' is, as Blackburn and Code argue, to posit a relationship such as that which exists between the referent of 'Aristotle' in (i) and that of the name in (iii) Aristotle, the magnate, married Mrs. Kennedy. It is to a particular usage of the name that a Fregean sense belongs. Mentioning  the  name in a description will not yield the required sense. Could not the sense be denoted by mentioning not the name itself but the usage of the name, by means of the description 'the sense of "Aristotle" as used to refer to Aristotle'. This may seem to do the trick for the example at hand. But if a sense is to be  1 8  See above, pp. 10-11.  19  See Russell, 'On Denoting', p. 49.  Theories of Belief  Attribution  45  denoted by means of mentioning the usage of the name then, since difference in usage is distinguished  by difference in referent, and as there is no backward route from referent to  sense, how are the differing senses of co-referential names to be distinguished? Since the Fregean cannot appeal to either the use or the mention of the name in order to discuss what the sense of a name is, it would seem that he cannot provide the required account of the logical connection between the sense of the name and the object which is the referent. Yet without such an account the conclusion that someone has a true or false belief cannot be reached. It is not just the failure of providing the necessary account of the logical relation between the sense of a name and its referent that is generated by the difficulty in saying just what the sense of a name is. In indirect contexts we are supposed to refer to the sense of the name; but how do we know just what to refer to, or what another has referred to? Leonard Linsky's position is that the sense of a name is such that in principle it can never be stated the only manner in which a sense can be presented, so to speak, is by the use of words to express this sense. 20  However, if, in indirect contexts, we refer to the sense then clearly the sense must be such as to be capable of being 'presented' in a manner other than merely being shown. Dummett's position is that, although we cannot state what the sense of an expression is but only show it, we can state what someone can do when he grasps the sense of a name:  To grasp the sense of a word is to master a certain ability to determine the truth conditions of sentences containing it; and there is no reason to impute an ineffable character to such an ability. Even if we cannot say Oblique Contexts, Chicago, 1983, p. 51. Curiously, but a few pages later, Linsky sees nothing wrong in giving this unstatable entity a name (see p. 54), and below, n. 25. 2 0  Theories of Belief  Attribution  46  what a sense is, there is no obstacle to our saying what it is that someone can do when he grasps that sense; and this is all that we need the notion of sense for. 21  This might have been sufficient if the sole employment of the notion of sense was to determine reference in direct contexts. But again, if we refer to the sense of a name in indirect contexts, more is needed. We do not refer in those contexts to what someone can do. Hence not only has the Fregean not provided us with a satisfactory account of the indirect referent of a name but it would appear that there are logical restrictions on his ability to provide such an account. Problems arise also with respect to the sense of a name when it occurs in an indirect context. The referent of the name in such a context is its customary sense (call this sense ). However, since, for Frege, sense and reference are always distinct, the name as it occurs in an indirect context must have another, indirect, sense (sense ). 2  Moreover, indirect contexts may be the result of multiple embeddings. As Rudolf Carnap has pointed out,  22  while the sense of 'a' in  C believes that a is F. is its indirect sense (sense ), the referent being its customary sense (sense ), in 2  B believes that C believes that a is F.  2 1  2 2  Dummett, Frege, Philosophy Meaning  of Language, p. 227.  and Necessity, pp. 266-269.  Theories of Belief  the referent of 'a' is not sense but sense , 2  23  47  Attribution  and a further sense, sense  3>  has to be  introduced to refer to that entity. Hence not only is there no such thing as the sense of a name, Frege seems committed to an infinite hierarchy of senses for any given name. Dummett has argued that the solution to this problem lies in amending Frege's theory so that, while the indirect referent differs from the direct referent in that it is its customary sense, the indirect sense of the name is held to be the same as its customary sense.  2 4  So, in all indirect contexts, no matter how deeply embedded, the referent of the  name will be the same, its customary sense. But such a move, while it blocks an infinite series of senses being generated, violates the basic Fregean principle that the sense of a name is always distinct from its referent. To violate this principle is to permit instances of the identity puzzle to occur which lack a Fregean solution. Linsky gives the following example Consider an instance of the puzzle ' A = B ' where 'A' and ' S ' denote objects of the lowest logical type. This, we will suppose, is informative, because, although'A' and 'J3' agree in reference, they differ in sense. Now nothing prevents our language from containing another name (with another sense) for the sense of A; say ' C . And now we can raise Frege's question again, How can 'C = the sense of A ' differ in cognitive value, if true, from 'C = C ? The Fregean answer is that the reference of ' C is the sense of A\ and the sense of ' C is a concept of the sense of 'A'; thus it is a sense of the sense of ' A ' . l  25  See Frege, Philosophical as much. 2 3  2 4  and Mathematical  See Dummett, Frege, Philosophy  Correspondence,  p. 154, where he admits  of Language, pp. 266-9.  Linsky, Oblique Contexts, p. 54. Linsky is here presenting a form of an argument first proffered by A. Church. (See Church, 'The Need for Abstract Entities in Semantic Analysis', in The Structure of Language, eds. J . J . Katz and J . A . Fodor, Englewood Cliffs, N.J., 1964, pp. 437-445.) Note that, in appealing to this argument, Linsky is contradicting his earlier claim that we can never state what the sense of any expression is but can only show it by using it. 2 5  Theories of Belief  Attribution  48  Dummett cannot give such a reply. If the sense of a name is always the same then the sense and the referent of ' C are identical, likewise with 'the sense of "A"'. Hence there is no difference in sense between ' C and 'the sense of "A"  7  and the puzzle for which the  notion of sense was first introduced is generated once again. If such puzzles are to be completely resolved (leaving aside all other problems) , it would seem that it would be at the price of postulating an infinite hierarchy of senses. What now of the claim that the Fregean theory can handle well the problems connected with substitution and vacuous names in belief contexts? Of the latter we have seen that it may well be that they cannot be resolved in a manner consistent with Fregean theory. Of the problem of substitution w e have seen that it would seem to be only completely solved at the expense of accepting an infinite hierarchy of senses, a price that man}' would see as being too high to pay. We have seen that the mechanism employed for the resolution of the problems, the sense of a name, may in principle be incapable of precise formulation, yet without it not only can the required account of the relation between a name, its sense(s), and its referent not be given, but the whole infinite hierarchy of senses is incomprehensible. What appeared at first glance to be a powerful solution to the problems of indirect contexts is, when studied further, riddled with problems of its own. When the points raised in the Negative Arguments are added to those of immediate concern to indirect contexts, Frege's solution appears to be no solution at all. It may be that some, though not all, of the difficulties that beset the Fregean theory may be overcome, but at present, given the state of the Fregean enterprise, it offers no strong argument in favour of abandoning the direct reference view.  Theories of Belief  49  Attribution  Proposed Methods of Resolution. If the Fregean theory of sense, and with it the Fregean 'solution', can be abandoned with a clear conscience, the problems posed by belief attributions are still pressing. Not all direct reference theorists, however, have been indifferent to these problems and, of late, attention has increasingly been paid to them, so much so that the literature is now filled with proposed strategies for their resolution, given the direct reference standpoint. Here I shall deal with what appear to be the most promising of these new proposals. But first a not-so-new response to the problems of belief attributions which may be seen to offer a solution. The Quinean Treatment. For Quine, sentences of the form (6) B believes that a is F. are ambiguous, capable of being given two distinct readings. On one reading, the notional, belief is seen as a dyadic relation between a believer and a proposition. (Quine dislikes such entities but let us follow his example and leave this issue aside for now.) Hence, on the notional reading, (6) states that the belief relation holds between B and the proposition a is F. On a relational reading, however, belief is treated as a triadic predicate holding between a believer, an object and an attribute.  26  For (6) the triple  consists of B, the referent of 'a', and the attribute F-ness, which Quine expresses, by  Where there is more than one singular term in the embedded sentence then the possibility exists of positing a tetradic or greater belief relation, as in 2 6  Tom believes yz(y denounced z) of Cicero and Catiline. (See 'Quantifiers and Propositional Attitudes', Journal of Philosophy, 53 [1956], 177187, p.180.)  Theories of Belief  50  Attribution  means of prefixing a variable to a sentence in which it occurs free, as z(z is ah F). So on a relational reading (6) can be written as 'B believes z(z is an F) of a ' .  2 7  Quine argues that on the notional but not the relational reading of (6) 'a' occurs in a position which is opaque, i.e. any singular term which occupies that position will function in a way that is not purely referential.  28  When a singular term occurs in an  opaque position, if truth is always to be preserved, substitution of a co-referential term is not possible nor is quantification into the opaque context.  29  It would seem, in Quine's distinction, that we have the means to hand to deal with the problems that beset the treatment of belief attributions. In the notion of a relational reading, the necessary connection to the object is established.  30  Given B's  sincere assertion of 'Aristotle was wealthy' then, on the relational reading of 'B believes that Aristotle was wealthy', we have the means to relate B's belief, or part of it at least,  2 7  See Quine, ibid.  Quine at times vacillates between the view that it is the position which is not purely referential and the view that it is the occurrance of the term that is so. Such vacillation gives rise to an inconsistency in his argument (see Kaplan, 'Opacity', in The Philosophy of W.V.Quine, eds. L.E. Hahn and P.A. Schilpp, L a Salle, Illinois, 1986, pp. 229-289, esp. pp. 231-235). Quine later acknowledges his lack of clarity on the point and argues that what should be treated as not purely referential is the position concerned (see Quine, 'Reply to David Kaplan', op. cit., pp. 290-4). 2 8  The relationship between these three notions, opacity, failure of substitutivity, and failure to quantify in, is not always easy to establish from Quine's work-as he himself more or less admits (see Quine, 'Reply to Kaplan', p. 290). He sometimes takes failure of substitutivity as the criterion for opacity, from which in turn follows an inability to quantify in (see Quine, 'Notes on Existence and Necessity', Journal of Philosophy, 40 (1943), 113-237, esp. pp. 113-8; and Quine, 'Quantifiers and Propositional Attitudes', pp. 178-9). At other times he takes failure to quantify in as the criterion for opacity (see 'Three Grades of Modal Involvement', Proceedings of XI International Congress of Philosophy, Brussells, 1953, vol.14; reprinted in The Ways of Paradox, New York, 1966, pp. 156-174.) 2 9  I do not wish to imply, by my use of 'necessary connection', that the tie between language and the world is a necessary one, something with which Quine would disagree. Rather what is necessary is that a (logical) connection exist between the name as it occurs within the embedded sentence of one used to attribute a belief, its referent, and its normal referent. 3 0  Theories of Belief  51  Attribution  to Aristotle himself. Alternatively, should we wish to concentrate on what it is that B believes, on characterizing the content of B's belief, then this can be achieved on the notional reading. One time we would want to do so, for example, would be when, unlike B, we believed that there was no referent of the name. One drawback to Quine's account lies in the characterization of the notional reading of an attribution. If, on this reading, B is related to a proposition then, since substitution is barred, this cannot be a singular proposition. But in that case what, if not the object which is the referent, is the entity in the proposition corresponding to the singular term? Whatever it is, that entity must bear some semantic relation to the name. But what can this entity be? Fregean senses having been rejected, there would seem to be nothing available to fill the role. Perhaps we could, with Quine, retreat from talk of propositions, at least on the notional reading, and hold that on that reading the relationship posited is between B and the sentence 'a is F\ While this move removes the problem of invoking strange entities to be the constituents of the belief attributed, it generates specific problems of its own. Perhaps the most immediate worry in following Quine in this matter is that it is never entirely clear from Quine's work just what one is committing oneself to when one gives a notional reading to a belief attribution sentence. Sometimes Quine talks as if the reading given is reflective solely of the attributor's interest in making the attribution, depending, for example upon whether we wish to perform a valid deduction of not.  31  At  other times he talks as if a notional reading correctly depicts what is in fact believed by the believer, that the object of belief is indeed a sentence.  31  32  If the former is the case  See, for example, Quine, 'Quantifiers and Propositional Attitudes', p. 182.  See Quine, 'Quantifiers and Propositional Attitudes', p. 178, where, of the two readings of 'Ralph believes that someone is a spy' viz:  3 2  (i)  x) (Ralph believes that x is a spy)  Theories of Belief  52  Attribution  could not the attributor want to attribute precisely what it is that B believes? If the latter is the case, if a notional reading commits one to the view that what is believed is literally a sentence, then not only would such a position be unwelcome to those of a nonnominalist persuasion but problems would arise from the fact that a name may be used on different occasions to refer to different people. B may assent to (or assert) 'a is F" when used in one way but deny it on another. For example he may assent to 'Aristotle was wealthy' when used to talk of the shipping magnate (in his youth) but deny it when used to talk of the philosopher. If, on the notional reading, what is attributed to the believer is a belief in a sentence, then one cannot distinguish between the various uses of a proper name. It would seem that in this case we can only conclude that B believes l  that Aristotle was wealthy' is both true and false on a notional reading.  33  Perhaps the major drawback of an appeal to the Quinean account of belief attributions, however, is that, regardless of whether the notional reading is treated as attributing belief in a- proposition or in a sentence, it is not possible to attribute the complete belief  34  and forge the link to the world in the same attribution. In relating B  to an object in the world we use the relational reading. But in so doing we can only relate part of the belief held, for example, F-ness, to the object. Should one wish to attribute the complete belief held then one cannot relate B to an object in the world, and (ii) Ralph believes that  (Hx){x is a  spy)  Quine remarks that the difference between the two is vast 'indeed, if Ralph is like most of us, [ii] is true and [i] false'. Someone who presses this interpretation of Quine's work is Kaplan, 'Quantifying In'. The situation is even more desperate in the case of indexicals. For any predicate F in S's lexicon, an assertion of 'He is F' (or 'she', 'that' etc.) may be both assented to and denied depending upon the referent of the indexical. 3 3  By 'complete belief I mean that entity, whatever it may be - sentence, proposition or what-have-you, whose structure in its entirety mirrors (in some loose appropriate way) the sentence used to express it. 3 4  Theories of Belief  Attribution  53  Quine's account fails, as Frege's does, in failing to account for the fact that B can believe something true (or false) about the actual flesh and blood man, Aristotle.  The P e r r y / K a p l a n Proposal.  Of the more recent propositions for resolving the problems associated with belief attributions, that of John P e r r y ,  35  is perhaps one of the more promising. Perry's  strategy, also endorsed to some extent by David Kaplan, is based on Kaplan's work on indexicals.  36  Indexicals are, for Kaplan, directly referential. As we have seen, the  contribution made by a directly referential term to what is said in a given context, to the proposition expressed, is the object referred to. If you say today  I was insulted today. the proposition expressed contains you, the person. In order for me to express that same proposition, using an indexical, I would have to assert You were insulted today. Tomorrow I would have to assert You were insulted yesterday.  See Perry, 'Frege on Demonstratives', Philosophical Review, 86 (1977), 474-97; and 'A Problem about Continued Belief, Pacific Quarterly, 61 (1980), 317-332. For a much more complex version of the theory see Jon Barwise and John Perry, Situations and Attitudes, Cambridge (Mass.), esp. pp. 253-264. 3 5  3 6  Kaplan, 'Demonstratives', esp. pp. 529-537.  Theories of Belief Attribution  54  Here then we have one meaning of indexicals - their content, the object or place, day or person referred to. However, as Kaplan argues, there is another meaning associated with expressions: The second kind of meaning, most prominent in the case of indexicals, is that which determines the content in varying contexts. The rule 'I' refers to the speaker or writer is a meaning rule of the second kind. 37  This second meaning Kaplan calls the character of an expression. The character of an expression is set by linguistic conventions and, in turn, determines the content of an expression in every context. Because character is what is set by linguistic conventions, it is natural to think of it as meaning in the sense of what is known by the competent language user. 38  The character of indexicals is context-sensitive, the content varying with the context; the character of non-indexicals is not so context-sensitive but yields the same content in all contexts. The character of an expression pertains to the expression as a type. It is, if you like, the dictionary meaning (if we but had ideal dictionaries). It is (ambiguities apart) what leads to two occurrences of the same word having the same meaning, making language learning possible. If I today say 'I am tired today', and you tomorrow say 'I am tired today' there is a sense in which our words have the same meaning. It is this meaning which is captured in Kaplan's notion of character. From this is may appear that Kaplan has simply smuggled in, in the guise of the character of an expression, the Fregean notion of sense. Not so. While the character of an expression is like a Fregean sense in that, being a linguistic rule, it determines the content for any particular occurrence of the word, it tells us what the referent would be  3 7  3 8  Kaplan, 'Demonstratives', p. 505. Ibid.  Theories of Belief  Attribution  55  for any possible occurance, the character of an expression differs from a Fregean sense in one vital respect: it is not a part of the content, constitutes no part of the proposition expressed. Only the object referred to forms part of the proposition expressed, not the rule by means of which that object is obtained. Further, Kaplan insists, the character of an expression cannot be equated with a higher order of Fregean sense, with an indirect sense.  39  Although the character of an expression is not to be equated with a Fregean sense, both Kaplan and Perry see it as sharing a further potentially powerful feature with that notion. If one knows the character of an expression this does not entail that one knows, when the expression is used on two different occasions, that the contents are or are not identical. I may twice... hear T and not know if the content is the same. What I do know is this: if it was the same person speaking, then the content was the same. 40  A distinction is drawn between the object of thought and (Kaplan) the cognitive significance of, or (Perry) the believer's cognitive perspective towards, that object.  3 9  Kaplan's argument for this point is somewhat obscure. He remarks In Frege's theory, a given manner of presentation presents the same object to all mankind...But for us, a given manner of presentation - a character...will, in general, present different objects (of thought) to different persons (and even different Thoughts to the same person at different times). ('Demonstratives', p. 530)  What Kaplan is not saying here is that, for Frege, indexicals are not context sensitive. Rather his point is that, for Frege, the sense of an expression is completed by features from the context. Hence the sense of any given utterance of an indexical will contain features of the context and that completed sense will always present the same object to everyone. A character, on the other hand, being a rule or function, does not depend upon features of the context for completion but is already complete. Hence it will present a different object in different contexts. Kaplan, 'Demonstratives', p. 505, n. 30.  Theories of Belief  Attribution  56  Whereas the object of thought is given by the content, the cognitive significance of or perspective towards the object of thought is yielded by the character of the sentence used. With the depiction of the notion of character as having a different cognitive status to that of content it would appear that a vital third element has been introduced which could perhaps be exploited in the resolution of the problems of belief attributions. This is precisely what Perry, albeit sketchily and tentatively, proposes for the problem of substitution.  41  It is not, he argues, the prime function of belief attributions to  represent the way the believer believes a certain proposition but rather to attribute to the believer a belief in that proposition. Hence, for example, since (7) John believes that Laura Jones has gone to Toronto, is true and since Laura Jones is Mrs. Laura Robinson, then (8) John believes that Mrs. Laura Robinson has gone to Toronto, is also true. However, unlike (7), (8) fails to show the way John believes the proposition. It fails to show how the proposition has cognitive signficance for John - by means of the character of the sentence 'Laura Jones has gone to Toronto.' It is this fact which explains John's vehement denial of (8). If one wishes to capture the cognitive significance to the believer of the proposition believed then this can only be  See Perry, 'Frege on Demonstratives', esp. pp.  494-6.  Theories of Belief  Attribution  57  done by using as the embedded sentence that which has the character by means of which the proposition is accepted. Perry later argues that attributions of belief carry a Gricean implicature to the effect that the believer's cognitive perspective of the proposition is represented by means of the embedded sentence.  42  Hence (8) is infelicitous in this regard since it carries the,  false, implicature that John would accept the proposition as presented by the character of 'Mrs. Laura Robinson has gone to Toronto'. One problem with this solution is that even in cases involving indexicals an appeal to character sometimes fails to explain the speaker's differing reactions to the same proposition. As Wettstein points out, while it seems to work well for cases which involve two (or more) different indexicals, for example 'he' and T (or 'his' and 'my', as in 'His pants are on fire' and 'My pants are on fire'),  43  it fails in cases involving two  separate instances of the same indexical, say 'he' and 'he', where the characters involved are identical.  44  There is a yet more fatal flaw in this proposal, however, which becomes clear when we reflect on Perry's position that (3) is infelicitous with respect to presenting the believer's cognitive perspective of the proposition since it carries the false implicature that John would accept the proposition as presented by the character of 'Mrs. Laura  See Barwise and Perry, Situations and Attitudes, pp. 258-9. The exception to this state of affairs is when indexicals are involved. I simply cannot represent the character by means of which Dan understands the proposition which he believes when he utters 'I am too fat'. I cannot say 'Dan believes that / am too fat', but rather will say 'Dan believes that he is too fat'. Obviously in such circumstances there can be no Gricean implicature to the effect that I am accurately representing how Dan would understand the proposition. 4 2  Though even here there is some doubt as to whether the notion of character will do what is required. Kaplan later voices the worry that competent speakers may not in fact know what the character of an indexical is. See Kaplan, 'Afterthoughts', pp. 577-8, n.26. 4 3  44  See Wettstein, 'Has Semantics Rested on a Mistake?', pp. 195-6.  Theories of Belief  58  Attribution  Robinson has gone to Toronto'. Just what does the character of the embedded sentence consist of? More precisely, what is the character of the name 'Mrs. Laura Robinson'? While there would certainly seem to be linguistic rules connected with indexicals such as T , 'here', and 'yesterday', it is by no means clear that names are the kind of expression which could have linguistic rules connected with them. Even if one sees names as being part of the language, in the sense of being suitable for dictionary entries, what other linguistic rule could there be connected with a name beyond that it refers to a particular individual? As Kaplan himself argues,  45  it would seem that in the case of names,  character, content and referent collapse into one. Yet without a character which is distinct from the content, no solution to the problems of belief attributions can be forthcoming.  The Salmon Proposal. Another recent proposal is that of Nathan Salmon. In Frege's Puzzle  46  Salmon  also holds to a strong line concerning the truth conditions of belief attributions statements where substitution is involved. Despite the believer's protestations and the contrary view of ordinary language speakers, if B believes that Fa and if a = b, then B does indeed believe that Fb. Salmon holds that such a view is sharply at odds with ordinary usage but maintains that those who take the opposing view are simply mistaken. Nevertheless, Salmon feels, in taking such a strong stance, it is incumbent upon him to explain how such a gross and persistent error could occur among ordinary speakers.  Kaplan, 'Demonstratives', pp. 558-563, esp. p. 562. Cambridge, Mass., 1986. See especially pp. 103-118.  Theories of Belief  Attribution  59  He begins by drawing an analogy between an attitude towards a proposition and that towards some other object: believing a proposition is similar to liking ice cream, finding a certain piece of music pleasant, loving a certain person. In the case of loving or liking something, should the subject fail to recognize the object when encountered again then it may well be that the subject will not adopt the same attitude towards it, indeed may dislike or even hate it. Failure to recognize objects comes about when the object is presented under an unfamiliar guise. As with liking or loving an object, so too with believing a proposition. One may fail to recognize a proposition when it is reencountered. What one believes may not be believed when encountered again. Should it seem strange to talk of 'failing to recognize' a proposition, Salmon reminds us that, for direct reference theorists, a proposition expressed by means of a sentence containing a proper name contains an object, place or person, and  If an individual has a certain appearance, either objective or subjective, and through perceiving the individual one comes to have some thought directly about that individual - say, a thought that would be verbalized as 'Gee, is he tall' - then there is a sense in which the cognitive content of the thought may be said to have a certain appearance for the thinker since one of its major components does. 47  Such objects can change their appearance and so we might fail to recognize them when encountered again. The question arises as to just how acquainted we need to be with an object in order to entertain singular propositions about it. Salmon does not give an answer to this thorny question but argues:  Salmon, Frege's Puzzle, p. 109.  Theories of Belief  Attribution  60  Whatever mode of acquaintance with an object is involved in a particular case of someone's entertaining a singular proposition about that object, that mode of acquaintance is part of the means by which one apprehends the singular proposition. 48  This mode of acquaintance, whatever it may be, of the object and hence of singular propositions, is analogous to an appearance or guise. It is by means of a guise that we grasp a singular proposition. John grasps the proposition expressed by 'Laura Jones has gone to Toronto' in a different way than when he grasps the very same proposition when he hears 'Mrs. Laura Robinson has gone to Toronto'. Since he grasps the proposition in two different ways he takes the single proposition to be two different ones. Here then, in the guise of the proposition, we again have a third element which can be put t o use in resolving the problem of substitution. Salmon stands firm on his hard line with respect to the truth conditions of belief attributions: the sole semantic  contribution made by the 'that' clause in such assertions  is the proposition attributed as being believed. Nevertheless, Salmon argues, we have two purposes in attributing a belief. Not only do we want to specify the proposition believed, we also want to specify something about the way the believer grasps the proposition when he believes it. 'Believe', however, is a dyadic predicate. Were we t o have a triadic predicate in the language we could represent belief as a triadic relation between a believer, a proposition and a guise. How would the guise be represented?  Of  the sentences 'Laura Jones has gone to Toronto' and 'Mrs. Laura Robinson has gone to Toronto', Salmon would argue that although they are materially and modally equivalent only the first manages to convey how John agrees to the proposition. The second sentence, when embedded in a belief attribution, while presenting the same content as does the first, presents it 'in the wrong way'. Hence for Salmon,  48  Ibid.  Theories of Belief  Attribution  61  the guise or appearance by means of which the believer would be familiar with a proposition at a particular time t were it presented to him or her through a particular sentence is a function of the believer and the sentence. 49  For example, using a triadic B E L predicate and where 'f stands for the function which is the guise, John's belief could be attributed as B E L [John, that Laura Jones has gone to Toronto, f (John, 'Laura Jones has gone to Toronto')] or B E L [John, that Mrs. Laura Robinson has gone to Toronto, f (John, 'Laura Jones has gone to Toronto')] 50  In ordinary language, given the diadic predicate 'believe', although one cannot explicitly mention the third relatum one can, as Salmon says, 'fake it' by using as a 'that'-clause a sentence that determines the third relatum in question. 51  Thus while the 'that' clause has the sole semantic function of specifying the proposition believed, it also has a pragmatic function involving 'an autonomous mention-use of the clause'.  52  While, in using 'believes', we say what proposition is believed, we also can  4 9  Salmon, Frege's Puzzle, p. 117.  5 0  But not B E L [John, that Laura Jones has gone to Toronto, f (John 'Mrs. Laura Robinson has gone to Toronto')]  nor B E L [John, that Mrs. Laura Robinson has gone to Toronto, f (John, 'Mrs. Laura Robinson has gone to Toronto')] 5 1  Salmon, Frege's Puzzle, p. 117.  5 2  Salmon, Frege's Puzzle, pp. 117-8.  Theories of Belief  show how it is believed.  53  Attribution  62  It is the failure to distinguish between information that is  semantically encoded and that which is pragmatically imparted which has led to the confusion of ordinary language users regarding the truth conditions of belief attributions. Salmon's schema is intriguing and ingenious. It suffers, however, from a serious and, I believe, fatal difficulty - one, yet again, which concerns the third element of the belief relation. Recall the question, which Salmon ducks, as to just what mode of acquaintance with the object is necessary in order to entertain a singular proposition about it. Before ducking the issue Salmon questions just what this mode of acquaintance might be. Must the mode of acquaintance be causal? Is any causal relation enough? (Consider the case of numbers and mathematical knowledge.) Is it enough simply to have heard the individual mentioned by name? Is it enough simply to be able to refer to the object? (Consider the shortest spy.) Is it enough simply to point at the object, without even looking to see what one is pointing at? Must one have some conception of what kind of thing the object is (a person, an abstract entity etc.)? Can one have mistaken opinions about the object? How many? Does one have to know who the individual is, or which object the object is, in some more or less ordinary sense of 'know who' or 'know which'? Must one know some feature or characteristic of the object or individual that distinguishes it (or him or her) from all others? Is it sufficient simply to know some distinguishing feature or characteristic (i.e. is what Russell called 'knowledge by description' always enough)? 54  From the nature of the questions asked, it is clear that what is at issue is how reference is achieved. But has not that question already been dealt with by the direct reference theorist? How can Salmon, adherent of the direct reference view that he is, still be asking whether a definite description is what is required? To question Salmon thus is to misunderstand the direct reference position. What is held in common by all direct reference theorists is that the speaker's beliefs about the  The exception again is in the case of indexicals. See above, p. 57, n. 42. Salmon, Frege's Puzzle, p. 108.  Theories of Belief Attribution  63  object cannot be what determines which object is the referent. That is done, according to the picture usually suggested, by the historical or causal chain connected with the use of the name. The question Salmon is addressing is: under what conditions is a speaker 55  in a position to use that name to refer to the object? It is quite consistent with the direct reference view to hold that, quite apart from how reference is determined, certain conditions must be met, for example that the speaker must have certain beliefs about the object, in order that he be able to use that name to refer to i t .  56  The referent, while  not being determined by the descriptions furnished by those beliefs, would be recognized (in the proposition expressed) under the guise of these descriptions. However, while there is certainly no contradiction involved in Salmon permitting that the necessary mode of acquaintance may be any of the options he outlines, a difficulty emerges when we consider how he fits the mode of acquaintance into his overall schema. In a belief attribution (those involving indexicals aside) we say what proposition is believed but we show how it is believed. But how in (1) John believes that Laura Jones has gone to Toronto. do we show the operative mode of acquaintance by means of which John grasps the proposition attributed? More specifically, how can the  name  'Laura Jones' show us the  Of late the view has been growing that this position is not strictly accurate. The causal or historical chain attached to a name is now seen not to form part of the semantics of the language but rather to function pre-semantically. It provides the object which is then semantically assigned to the name. Strictly speaking, therefore, the referent is not determined by means of the chain but is given by the assignment made to the name. See Kaplan, 'Demonstratives', pp. 558-566, esp. p. 563, n. 78; Joseph Almog, 'Semantic Anthropology', in Midwest Studies in Philosophy, vol. IX, eds. P. French, T. Ueling and H. Wettstein, Minneapolis, 1984, pp. 479-490; and Wettstein, 'Has Semantics Rested on a Mistake?', pp. 194-5. 5 5  The subject has long interested Kaplan. See Kaplan, 'Quantifying In', and 'Demonstratives', esp. pp. 555-6, where he sketchily puts forward a view very similar to that developed by Salmon. 5 6  Theories of Belief  Attribution  64  mode of acquaintance John has with this woman, how he recognizes her in the proposition? If we take the requisite mode of acquaintance to be literally that the speaker must be able to recognize the object in the proposition then how the referent is recognized will vary from speaker to speaker: some may be able to recognize the person by sight, some only if they hear the person's voice. There would seem no reason to prefer one sense over another as the means of recognition. However, where two means of recognition are possible then the potential exists for a speaker to think, in cases where there are two (or more) co-referential names, that they refer to different people. John recognizes Laura Jones by voice, having onlj' talked to her on the telephone when she uses her maiden name. Mrs. Laura Robinson, on the other hand he recognizes by sight. The name alone, however, is not capable of showing which means of recognition is operative. The same situation would prevail if actual recognition is deemed not to be necessary but merely that the speaker be able to pick out the object by means of some description furnished by his beliefs about the object.  57  The name cannot tell us which  beliefs are the operative mode of acquaintance. In order to overcome the problem of speakers possessing differing sets of beliefs it could be argued that the necessary mode of acquaintance was simply a belief in the kind of thing the object is - being a man, for example, or a dog, or a star. In such a case all who (successfully) used a name to refer would possess the same belief about the object and, presumably, this could be shown by the name. But if the necessary mode of acquaintance were deemed to be the possession simply of a sortal this would not solve the problem of substitution: no different mode of  Kaplan introduces the term recognition by glossing it as 'knowing-(or believing) who', Kaplan, 'Demonstratives', pp. 555-6. 5 7  Theories of Belief Attribution  65  acquaintance would be shown by the name 'Laura Jones' and 'Mrs. L a u r a Robinson' John believes that they are both women. One thing we could show, by using a certain name in the belief attribution, is the name itself. Could we not, by using one name rather than another, show which one the speaker would use to refer to the object when expressing the attibuted belief? Certainly. But if the only difference between names, apart from their linguistic form (they are different names), is that they may be used to refer to different objects, then how, in the case of two co-referential names, is the fact that the speaker used one rather than another  informative? It is not quite correct to say that the only difference between names is their form  and what the}' maj' be used to refer to. If the causal / historical theory is correct then there is, associated with each name (perhaps pre-semantically), a particular chain which links the use of the name to the referent. The chain connected with the name 'Laura Jones', for example, although stretching back to the same woman, would differ from that of 'Mrs. Laura Robinson'. Could not this chain, while forming no part of the proposition expressed, be what is shown whenever a belief is attributed which contains the name? Using the name 'Laura Jones' in the attribution of belief to John, could we not be showing that the way he believes the proposition is by means of that particular causal or historical chain? This certainly would seem a plausible argument to maintain. However, there is one major drawback: ordinary speakers of the language generally do not know anything about the chain which links their use of the name to the referent. Salmon's theory was to capture the cognitive  58  The impetus behind  aspect of the way the believer regards the  The major exception to this state of affairs is when the speaker initiates the name in a dubbing ceremony. See Kaplan, 'Bob and Carol and Ted and Alice', Appendix IX, pp. 501-3; and 'Demonstratives', pp. 560-561. 5 8  Theories of Belief  Attribution  66  proposition. If the speaker knows nothing about the chain connected with his use of the name then that chain cannot be used to explain the cognitive perspective the speaker has of the proposition. So, of the proffered candidates for the requisite mode of acquaintance, those that can be shown by the use of a name either fail to capture the necessary cognitive aspect or are not fine grained enough to distinguish modes of acquaintance in the manner necessary to solve the problem of substitution, and those which would seem capable of solving that problem cannot be shown by the name.  Beyond the specific difficulties of the Perry / Kaplan and Salmon proposals mentioned above, there is a more general shortcoming which concerns the overall strategy of their approach. The approach used towards the resolution of the problem of substitution in both of these proposals is deeply Fregean. While the third element which is appealed to is not identical with a Fregean sense, it shares many important features with that notion. Indeed some of the differences that one finds between the element used in the Perry / Kaplan proposal and that in Salmon's reflect the differing interpretations given of Fregean sense which were discussed above.  59  Consider the third element of the Perry / Kaplan proposal, character, in the light of the Cartesian interpretation of sense. Like a sense under that interpretation, a character is the means by which reference is determined, it is the route by which we reach the referent, a referent considered as an end separable from the route taken to it. Dummett's objection to the Cartesian interpretation of sense - that it would seem impossible to establish if we have ever succeeded in referring to an object, impossible to  5 9  See above, pp. 39-42.  Theories of Belief  67  Attribution  determine whether or not the expression stands for anything - would seem equally applicable to the Perry / Kaplan proposal. Salmon's third element, on the other hand, shares many  features  with  Dummett's own interpretation of sense. As in that interpretation, the referent cannot be detached from the mode of acquaintance; the mode of acquaintance is the way in which we pick out an object rather than the route we take to it, it is the way the object is presented to us. That being the case, as with Dummett's treatment, an appeal to such a mode of acquaintance is of no use in the case of vacuous terms. Where the term is vacuous there is no proposition expressed and hence no 'way of regarding' a proposition which can be appealed to. While rejecting Fregean theory at the level of semantics both proposals incorporate many of its elements at the level of pragmatics. But, as the. arguments given above against the two proposals show, versions of the direct reference theorists' own arguments against the Fregean position can be turned against them when they make such a move. In order to avoid being hoisted with their own petard, some means other than a reliance on the Fregean 'solution' must be found by direct reference theorists if the problem of substitution is to be resolved.  68  3. BELIEF ATTRIBUTIONS: WHY DO WE MAKE THEM?  If the Fregean approach towards the problem of belief attribution statements is rejected, it is difficult to see what an alternative approach might be. How can an appeal to the believer's cognitive relationship with the proposition, whatever that relationship may be, be ruled out? The feeling is pervasive that that is just what belief is all about. It is important to keep in mind, however, that the subject under discussion is not belief but belief attributions. We can resist the Fregean pull towards examining what is going on in the mind of the believer by keeping a firm grasp on that distinction. While it may turn out that an examination of belief is ultimately necessary (though even that event should not be presumed), it is with an examination of such statements that a beginning should be made. Asserting a belief attribution statement is something we all do regularly and with ease, even quite young children can handle and understand them perfectly well. Implicitly, it would seem, we understand how they function. It is interesting, in light of this, how little attention has been paid to our actual practices attributions.  Belief  attributions  have  been  much  thought  of making such  about  but,  despite  Wittgenstein's remonstrations, little looked at. One reason for such neglect is the low esteem in which ordinary language is held by some theorists. For others it is, I think, a result of too high a. concentration on the puzzles. While Russell was certainly correct in pointing out that  Belief Attributions: Why Do We Make Them?  69  A logical theory may be tested by its capacity for dealing with puzzles, and it is a wholesome plan, in thinking about logic, to stock the mind with as many puzzles as possible... 1  nevertheless one can become so mesmerized by the puzzles, often illustrated by timeworn examples, that one loses a grip on the phenomena of belief attributions as a whole. Our rich and varied use of such statements is ignored and attention narrowed to a small and seemingly lifeless arena. In order to bring to light what we implicitly understand about the functioning of belief attributions, an examination of the practices of ordinary language users in everday situations would seem vital. Before formulating a theory about the functioning of belief attributions we would do well to look at why, when and how we make them. :|:  :[:  *  :|;  :|:  :|:  *  ;|:  *  *  *  *  *  :|:  :|:  :|:  Why do we make belief attributions? Let us take an everj'day example of a belief attribution and consider some contexts in which it could be made. (At this stage of the inquiry we will ignore the problems: no substitution is involved, no term is vacuous.) Talking to his mother (Ma), the attributor (A) asserts: (1) Mary believes that Dick is bringing Jane to dinner tonight.  2  Suppose (1) is asserted in the following situation:  1  Russell, 'On Denoting', p. 47.  Because a study of self-attributions encourages a blurring of the distinction between belief and belief attributions I shall concentrate on looking at attributions of belief to others. 2  Belief Attributions:  Why Do We Make Them?  70  (a) Ma has remarked how reserved Mary is with her, so much so that. Ma feels she hardly knows her any more.'Well, let me tell you', A replies, 'Mary is getting weird. She has the funniest notions - especially about Dick. Do you know she thinks that he has secretly been seeing Jane. More than that.' A then asserts (1).  Why does A assert (1)? He is trying to illustrate what strange ideas Mary has about her brother. Let us suppose that Dick can't stand to be around Jane and that she would be the last person on earth he'd ever invite to dinner. Anyone who knew Dick at all well would know that. In asserting (1), A is giving an example of a belief which Mary has which is typical of the state, so A argues, that Mary is in vis-a-vis her brother. In asserting (1), A is passing on information which he has about Mary which will, he believes, help Ma understand Mail's state of mind. Now let us consider the following scenario:  (b) Ma has noticed that Mary is setting an extra place at table. 'Well', she says, 'I don't know why she's doing that'. A then asserts (1).  Here A is telling Ma why Mary is setting an extra'place. He is explaining Mary's actions - she is setting an extra place because she believes that Dick is bringing Jane to dinner. Now A would not be likely to assert (1) if he himself believed that Jane were coming. Were he to believe that she was, he would simply assert  (2) Dick is bringing Jane to dinner tonight.  However, his lack of belief in (2) is irrelevant to explaining why Mary is setting an extra place - all that matters is that Mary believe it. Because he knows of her belief, A 3  Note that were A to believe that Dick was bringing Jane and simply assert (2), his assertion on its own would not explain Mary's action since (2) could be true and Mary not believe it. However, since (2) would be offered as an explanation of Mary's action and understood as such, the truth of (1) would be taken as understood; or, to put it in Gricean terms, when (2) is asserted in (b), it conversationally implicates (1). (See 3  Belief Attributions:  Why Do We Make Them?  71  understands why Mary is acting as she is. He passes on that knowledge of her belief so that M a too will come to understand Mary's action. (c) Aware that he has asked some girl friend or other to dinner, M a asks A whom Dick is bringing. A asserts (1). Now if A is being open and sincere, he would not reply to M a by asserting (1) if he knew that it is Jane whom Dick is bringing. M a has not asked whom Mary believes Dick is 4  bringing - she asked whom Dick is bringing. Why then does A reply by asserting (1)? Not knowing the answer to her question himself he turns to the next best thing - the view of someone who thinks she knows. Unlike the previous two cases, here it is precisely because A doesn't know the 'desired information that he makes the attribution of belief. Nevertheless, A is still trying to supply the desired information. Not possessing it himself he does the best he can. From the above cases it would seem that the general purpose behind the making of a belief attribution is to supply, or attempt to supply, information which the audience lacks and desires to know. Not all cases of belief attributions, however, are in reponse to  H.P. Grice, 'Logic and Conversation', in The Logic of Grammar, G. Harman, Encino, Calif., 1975, pp. 64-75.)  eds. D. Davidson and  The qualification that A is open and sincere is necessary in order to rule out the following sort of case. 4  Dick is aware that his mother never trusts him in matters such as these. Why she bothers to ask him in the first place he'll never know. Still, he knows, she believes Mary to be very reliable and completely au courant with Dick's affairs. Hence, although A knows full well who is coming to dinner, if he wants his mother to know, he had better not assert (2); she would never believe him. If he asserted (1) however, given her trust in M a r y , she would conclude that it was Jane who was coming. So, although knowing full well the answer, A, being devious, asserts (1). Such cases need not concern us unduly since A here would seem to be pretending know.  not to  Belief Attributions: Why Do We Make Them?  72  such a need. Sometimes a belief attribution is made in response to an asserttion which the attributor considers false. For example, (d) Talking to A, M a asserts (3) Mary believes that Dick is bringing Suzie to dinner tonight. A responds with (1). Here A is correcting Ma's assertion with a view to correcting her false belief: it is not Suzie but Jane, A is saying, whom Mary believes that Dick is bringing. (It is normal, in such circumstances, to lay stress on the word, or words, instrumental in the correction.) Again, a correction can  be  made to a belief attribution which has not  been  asserted but merely presupposed. Consider: (e) M a asks A, 'Why on earth does Mary believe that Dick is bringing Suzie to dinner tonight? Suzie's in New York'. A responds with (1), accentuating 'Jane'. A question 'Why P?' presupposes the truth of P. Ma's question as to why Mary believes that Dick is bringing Suzie presupposes that Mary believes that Dick is bringing Suzie.  5  In asserting (1), A is correcting that presupposition. Not all occurrences of belief attributions made for the purpose of correction, however, are in response to a belief attribution which is considered faulty. Consider the following case.  (f) M a says to A, 'Dick is bringing Suzie to dinner tonight', or she asks, 'Why is Dick bringing Suzie to dinner tonight?'. A asserts (1) with 'Jane' stressed.  5  For an account of the presuppositions of questions, see below, pp. 88-98.  Belief Attributions:  Why Do We Make Them?  73  Here A is not directly contradicting Ma's assertion or presupposition. To do so he would simply reply Dick is bringing Jane to dinner tonight. In relying on what Mary believes, A, if he is being open and sincere, cannot himself know for sure who Dick is bringing. Nevertheless, in asserting (1) in such circumstances, he must think that it is at least possible that it is Jane and not Suzie - that there is a good chance that Mary is right in her belief. Here A is not in a position to correct M a himself. Still he thinks it at least possible that she is wrong. It is precisely because he cannot correct Ma's assertion outright that he relies on the attribution of belief. The effect of the attribution is to soften the force of ^4's assertion from that of an outright correction to one casting doubt on Ma's assertion. Let us take stock of what we have seen so far. We have observed two main purposes for which a belief attribution is made. The first is to supply (or attempt to supply) information to the attributor's audience, information which the audience professed or implied she lacked and which she desired to know. The second purpose observed was to correct (or to attempt to correct) a previous assertion or presupposition made by the audience. Not all belief attributions are made in the middle of a conversation, however. Sometimes they initiate the conversation. This is how assertions, in particular belief attributions, are usually considered. They are viewed, even by those who insist on the importance of context in determining the proposition expressed, in isolation from any conversational context.  6  They are seen as if made in a conversational vacuum.  There are, however, some notable exceptions to this common view. One thinks not only of Grice's work on conversational implicature (Grice, 'Logic and Conversation') but also of Austin's treatment of the distinctions direction of fit and onus of match (in 'How to Talk', Proceedings of the Aristotelean Society, [1952-3], 227-246); and of Strawson's 6  Belief Attributions:  74  Why Do We Make Them?  However, even if the assertor initiates the conversation there is a conversational purpose behind his assertion. Normally we want to supply  the audience  with  information, information which, we believe, the audience lacks and would like to know.  7  Sometimes, however, we want to correct her opinions. If we believe the audience to have a certain belief, a belief which she would express in making what we would consider to be a false assertion, F, then we can make an assertion with the purpose of correcting F even if the audience has never in fact asserted F. These two general purposes, I would argue, involving as they do the passing on of information (or the attempt to do so) and the dispelling of misinformation (or its attempt), form the core of much of our language use. If philosophers could ever get clear on the functioning of belief attributions made for these two reasons we would have come a long way indeed. Consequently, I will now turn from an examination of our general purposes in making such attributions to look more closely at the specific purposes involved. *  *  *  *  *  *  *  *  *  *  *  *  *  *  *  We have so far seen, for each of the general purposes involved, different uses for which a particular sentence can be employed. Given the general purpose of passing on desired information, it can be used to tell M a about a belief M a r y has in order to insistance on the importance of topic in treating statements, a notion which is treated as an antecedently introduced class (see 'Identifying Reference and Truth Values', Theoria, 30 [1964], 96-119). Indeed Strawson appeals to question and answer pairs in order to illustrate the notion of topic. John Searle has argued that sometimes we make a statement without caring whether the audience is open to what we say or not. For example, we may tell someone something simply because we feel it our duty to tell them, not caring whether they already know or want to know. (See Searle, Speech Acts, Cambridge, 1969, pp. 45-7; and Intentionality, Cambridge, 1983, pp. 1.65-6.) While Searle is certainly correct in that such cases no doubt occur, nevertheless, I would argue, in such cases the assertor acts as if his audience lacks and desires to know the information he is attempting to provide. Such a use is parasitic on that of communication. 7  Belief Attributions:  75  Why Do We Make Them?  illustrate her state of mind or explain her action, and it can be used in an attempt to tell M a about who Dick is bringing to dinner.  8  However, there are many more uses for  which it can be employed. Consider (1) in the light of the following batch of situations: (g) M a asks 'Who is bringing Jane to dinner tonight? (h) M a and A are discussing Dick and Jane's coming over tonight. M a wonders what time they are coming over. (i) M a and A are discussing Dick's bringing Jane round to dinner. M a has forgotten what day that was to take place. (j) M a and A are wondering whether it's true that Dick is bringing Jane to dinner tonight. In each of these situations the information which A is providing, or attempting to provide, in asserting (1) differs. The information supplied has a differing focus in each case; each has, as Dretske puts it, 'a featured exclusion of certain possibilities' which 9  differs from case to case. Yet the differing information which a sentence may be used to supply is not captured by the proposition expressed. On the straightforward direct  So put, readers may believe they recognize the seemingly familiar de re I de dicto distinction. However, firstly, these terms have of late been used in so many different ways that they now fail to mark any particular distinction. The situation is so bad that some theorists argue that these terms should be banished from the philosophical vocabulary - a position I tend to agree with. Secondly, and more importantly, to stop here is to draw too quick a conclusion, to miss all that is most interesting in our use of belief attributions. 8  In his insightful article 'Contrastive Statements', Philosophical Review, 81 (1972), 411437, esp. pp. 411-2. Dretske was perhaps the first to stress the significance of the differences involved when a sentence is asserted in different conversational contexts. He does not, however, treat sentences used to attribute a belief, doubting whether such 'featured exclusions' occur there. (This view is perhaps implicitly modified in his later 9  article 'Referring to Events', in Contemporary  Perspectives in the Philosophy  of Language,  eds. P.A. French, T.E. Uehling and H.K. Wettstein, Minneapolis, 1979, pp. 369-378, esp. pp. 376-7.)  Belief Attributions:  Why Do We Make Them?  76  reference view, the proposition expressed by A's assertion of (1) in each of the above scenarios is the same.  10  How then are we to capture the differing information being supplied in each case? Recall the general purpose for which the attribution is made: to supply information which, the attributor believes, the audience lacks and desires to have. Now where there is a lack of information there is a question which can be formulated which captures that lack.  11  The question might not be actually asked or implied (implicated) but nevertheless  if the lack is thought to exist then a question can be generated to capture it. Since the assertion is made in order to supply information which the audience lacks, and since that lack can be formulated in terms of a question, then the information supplj'ing aspect of A's assertion can be captured by viewing that assertion as an answer, an answer to the question generated by the perceived lack of information. Let us return to our original three situations (a) - (c) and view A's assertion as an answer. The three relevant questions are those expressed by: (a?) What does Mary believe about Dick? (b?) Why is Mary setting an extra place at table? (c?) Who is Dick bringing to dinner tonight?  I.e. that which consists of Mary, the belief relation and the singular proposition which contains Dick (and Jane). 10  For example, in (i) the question would be When is Dick bringing Jane to dinner? in CDIs it true that Dick is bringing Jane to dinner tonight? 11  Belief Attributions:  Why Do We Make Them?  77  One way of viewing a question such as those above is as an incomplete proposition together with a request to fill in the blank.  12  (a?), together with such a request, can be  seen as the incomplete proposition M a r y believes about Dick that he Of course not any belief which Mary has about Dick will constitute a satisfactory reply. A reply of Mary believes that Dick has two eyes. would be of no help at all. The belief ascribed has to be relevant to the purpose behind the question. Here the context determines that the relevant belief is one which illustrates the strange beliefs Mary has about Dick. In asserting (2), A complies with the request to fill in the blank in the relevant way. Note that in order to directly answer (a?) A must attribute a belief to Mary: the attribution of belief is essential to his reply: 13  if A is to answer Ma's question directly then he must do so by attributing a belief. Question (b?) is a request for an explanation. The incomplete proposition in this case is Mary is setting an extra place at table because  Note that this schema, as given, cannot accomodate all types of questions. It cannot, for example, handle questions which demand a 'yes' or 'no' answer. However, I use this schema for its simple heuristic value. For a more detailed treatment of questions and their answers see below, Ch. 4. The points made here could equally, though more laboriously, be made using the analysis given there. 1 2  The notion of a direct answer is a notoriously difficult one to capture. For our purposes here it is sufficient to note that it can be either true or false but, in either case, as N.D. Belnap puts it, 'it is the sort of thing which if true would tell anyone asking the question exactly what he wants to know, neither more nor less' ('Questions and Presuppositions', in The Logical Way of Doing Things, ed. K. Lambert, New Haven, 1969, pp. 23-37, p. 27. See also below, pp. 83-4). 13  Belief Attributions: Why Do We Make Them?  78  In this case it is not essential that the blank be filled in with an attribution of belief nothing about the question demands it. It would be quite correct to reply, were the circumstances different: You gave her five place settings and told her to set the table. She's just following your instructions. While such an answer would be a correct means of reply, it is not the reply A believes is the right one. In order to communicate to M a what he knows - that it is a certain belief Mary has which is the relevant cause of her action, A has to ascribe a belief to her. Hence, unlike (a) where the attribution of belief was essential to directly answering the question at all, here it is only essential in that it must be made in order to provide the true (or believed to be true) answer. What now of question (c?)? Unlike (a?) it is not essential that one answer it by means of a belief attribution. More importantly, unlike (b?) it is not possible to answer (c?) directly by means of one. The incomplete proposition is Dick is bringing  to dinner tonight.  M a wants to know who, in some ordinary sense of knowing who, Dick is bringing. Such a request can only be filled by means of a name or a description of someone. It is because he is unable to fill in the blank himself, because he cannot directly answer Ma's question, that A resorts to a belief attribution. Only by not directly answering the question can the attribution of belief be made. Unable to provide a definite answer himself, A does the best he can. He would seem to be providing a possible answer: the thrust of his assertion being: it could be Jane who's coming; Mary believes so. So we see that, not only does the treatment of attributions as answers bring out the differing information being supplied in each case, it also captures, indeed sheds light  Belief Attributions: Why Do We Make Them?  79  on, the distinction between cases in which the attribution of belief is made to supply the desired information and those in which it is used because the speaker does not know the desired information. What now of assertions made from the other general purpose observed - that of correcting a false assertion? In such cases there is no lack of information which the audience desires to know. How then are they to be treated? One thing of note in such cases is that, in responding to the false assertion, the attributor does not simply deny it. For example, he does not, when faced with  (3) Mary believes that Dick is bringing Suzie to dinner tonight.  reply  Mary does not believe that Dick is bringing Suzie to dinner tonight.  He corrects the false assertion, a specific part of it. The correction has the same structure as his assertion would have, were he answering a particular question specifically the question generated by the lack created when the disputed element of the audience's assertion is removed, i.e.  Who does Mary believe Dick is bringing to dinner tonight?  In the case where the audience asserts something of the form  (i) a is  F.  and A corrects that assertion with  (ii) a is G.  Belief Attributions:  80  Why Do We Make Them?  the structure of (ii) is identical to that of a direct answer to the question generated by requesting the gap to be filled in 'a is  ', i.e. 'What (or who) is aV.  Consequently, although the speech act differs, the structure of the correction can be captured, as in cases of supplying information which the audience desires to know, by treating the correction as an answer to a question. It is an answer to the question that, in the attributor's view, the audience, did she but know it, should ask. Hence for cases (d) - (f) the relevant questions (given, for convenience, with the original assertion or question) are: (d?) Who  does Mary believe Dick is bringing to dinner tonight? (d) Mary believes that Dick is bringing Suzie to dinner tonight.  (e?) Who  does Mary believe Dick is bringing to dinner tonight?' (e) Why on earth does Mary believe that Dick is bringing Suzie to dinner tonight?  (f?) Who  is Dick bringing to dinner tonight? (f) Dick is bringing Suzie to dinner tonight.  or Why  is Dick bringing Suzie to dinner tonight?  Note that, in correcting the presupposition of Ma's question in (e) and (0, A is answering a different question than that in fact asked by Ma. Here my  treatment of correction  differs from that of Belnap and Steel who see a correction made to any presupposition of a question Q as an answer to Q.  14  However, in asserting (1) in such situations, A, while  responding to Ma's question, is not answering it. Rather he is blocking it, denying the  See N.D. Belnap and T.B. Steel, The Logic of Questions and Answers, 1976, esp. p. 129. See also below, pp. 99-101.  14  New  Haven,  Belief Attributions:  Why Do We Make Them?  81  foundation upon which it rests. He does this by addressing a question which has to be settled before Ma's question can arise. Again, in the case of correction, we see the dichotomy between those cases in which an attribution is essential in order to give a direct answer to the question, cases (d) and (e), and those in which it is not, (f), but where the attribution of belief is made because the attributor himself does not know the answer. We are now in a position to answer, if as yet somewhat sketchily, the title question of this chapter: Why do we make belief attributions? We do so in order to supply information or to correct misinformation. The lack of information perceived in our audience can be captured by means of a question. The question may, given the conversational context, be explicit or implicit or, in cases where the attribution initiates the conversation, it may be attributed to the audience as one she would like (or be willing) to ask. In the case of correction the question being answered is one which, in the attributor's view, the audience should ask but, given her error, won't. The belief attribution can then in all cases be treated as providing an answer to a question. In some cases the attribution is made because the attributor knows (or thinks he knows) the answer to the question and is attempting to provide it. In others it is because he does not know the answer to the question that the attribution is made. In-order to explore in more detail the relationship between the belief attribution and the question asked, in order to understand more completely the attribution as an answer, a more detailed account of the question and answer relationship is required. It is to such an account that we will now turn.  82  4. QUESTIONS AND ANSWERS  A question, 'what is asked', is usually asked by means of the use of an interrogative sentence. Questions stand to interrogatives much as propositions do to declarative sentences: the same question can be asked by using differing interrogatives, and the same interrogative (e.g. 'Is that it?') can, in differing contexts, be used to ask differing questions. The account of questions and answers given by Nuel Belnap, upon which I shall 1  mainly rely, is grounded upon C L . Hamblin's dictum 'knowing what counts as an answer is equivalent to knowing the question'. A s it stands, Hamblin's dictum may be 2  somewhat misleading. There are many possible responses to a given question and, even when one has understood the question, it is not always obvious whether a certain response is or is not an answer to it. To take an example, let us suppose that Q (the questioner) is writing an article on the standard of living of various university folk and asks R (the respondant)  (1?) How much do you make a year?  Consider the following responses which could be made by R.  (a) I make $20,000 a year. (b) $50,000.  1  See Belnap, 'Questions and Presuppositions', and Belnap and Steel, The Logic of  Questions and Answers. 2  'Questions', Australian  Journal of Philosophy,  36 (1958), 159-168.  Questions and  Answers  83  (c) I make less than the rest of my colleagues - $15,000, and I can't live on it. (d) I mjike $10,000 for teaching one of my courses. (e) Look it up in the Annual Review - it's a matter of record. (f) I don't know. (g) None of your business. (h) Get out of here. (i) Isn't it a beautiful day! Clearly (a), were it to be asserted in this conversational context, is an answer to (1), 3  while a response of (i) would seem to involve changing the subject and therefore not constitute an answer to the question. It is not at all obvious, however, whether, for example, (f) or (g) would count as an answer or not. Nevertheless, we do understand the question. Although we may not know, when we understand a question, whether a certain response constitutes an answer or not, we do know what kind of answer is being sought. In understanding the question, we understand what sort of response would, directly and without beating about the bush, give precisely the sort of information requested, no more and no less. In understanding (1), for example, we understand that an answer is being sought which attributes to the respondant a certain sum of money, specifically that sum claimed to be earned each year, (a) and (b) (which is an abbreviation for 'I make $50,000 a year') are just such answers. Of the other responses given above, though some provide a sort of answer, none answer the question in the desired way. (c) provides more information than is requested, and (d), though it gives some of the desired  For convenience, I shall talk of the assertion of (a), or even sometimes of (a) itself, in the conversational context, as being (or not being) an answer. This is to be read as short for what (a) expresses in the context were it to be asserted. 3  Questions and  Answers  84  information, does not give it all. (e) provides a way by which the answer can be found but does not give it directly. Answers which provide in a direct way exactly the sort of information which is requested, neither more nor less, are termed direct answers. Only (a) and (b) of the responses listed above are direct answers. Of course, there are many other possible responses which would be direct answers to (1?): I make $25,000 a year. I make $30,000 a year and so on. Note that a direct answer need not be true: (a) and (b) are both direct answers but at least one must be false. To take another example, if one understands the question (2?) What colour is your car? one understands that the questioner is seeking an answer such as (2) My car is blue. or (2') My car is black. or (2")  My car is red.  and so on, for all the direct answers, that is, those which attribute a certain colour to your car. Correspondingly, if you understand that the questioner is seeking such an answer you have understood the question. Hamblin's dictum is therefore to be read as: knowing what counts as a direct answer is equivalent to knowing the question. It would seem that we could identify a question with the set of its direct answers. But now we face the task of specifying more precisely what a direct answer is. To begin we shall take it that at least one aspect of a direct answer is that it is a  Questions and  85  Answers  proposition. Hence if questions are to be identified with their direct answers, one aspect of a question is that it is a set of propositions. The interrogative can be seen as presenting  to the person questioned a range  of alternative  propositions. The range of  alternatives presented by (2?), for example, are the propositions expressed by the assertion in a context of 'My car is blue', 'My car is black', and so on. The range of alternatives is limited in various ways. Sometimes it is limited by the form of the question asked. Consider the question: (3?) Which colour is your car: blue or black? Here the questioner wants to know whether your car is blue or whether it is black. Hence the question has only two possible direct answers: (3) M y car is blue. and (3') M y car is black.  4  Similarly the question (4?) Is your car black? has only two direct answers: (4) M y car is black. or 'Yes', which is a coded answer for 'My car is black', and 5  Note that 'My car is yellow' is not a direct answer to (3?). It is rather a corrective reply, correcting the presupposition carried by the question that your car is either blue or black. For a more detailed treatment of the presuppositions of questions see below, pp.88-98. 4  See Hamblin, 'Questions', p. 1 6 2 ; and Belnap and Steel, The Logic of Questions and pp. 1 4 - 1 5 . Another possible response to (4?), which is also a coded answer for 'My car is black,' is a simple nod. 5  Answers,  Questions and  Answers  86  (4') M y car is not black, (or 'No'). In other cases the range is limited not only by the form of the interrogative used but also by the meaning of certain words which occur in the interrogative. In (2), for example, the range consists of those propositions in which a certain colour is attributed to your car. How many direct answers there are is dependent upon the number of colours that can be specified. Again, as B. van Fraassen has argued, the range of direct answers can be 6  limited by the context in which the question is asked. For example, if (5?) Which car is yours? were asked at the entrance to a parking lot downtown, a reply such as (5*) M y car is the one parked outside my house. (when the respondent lived miles away) would not be a direct answer. Here the context limits the range of direct answers to those identifying as your car one of those in or near the parking lot."  6  In The Scientific Image, Oxford, 1980.  Similarly, were (2?) to be asked at the entrance to a parking lot which happened to contain only blue cars, the context would determine that an answer such as 7  M y car is royal blue. or M y car is light blue, is required. Before leaving the discussion of the limitation of the range of alternatives, it should be noted that the range can be infinite. The question What is an example of a prime greater than ten?  Questions and  Answers  87  We have noted that a question presents a range of alternatives. Any direct answer will then be a member of that range. However, as Belnap argues, in order to fully understand an answer to a question, one must do more than understand barely that it is an answer, one must also understand how it is an answer. 8  The question, besides presenting the range of alternatives, includes a request.  One  aspect of the request is that the audience is requested to make a selection from amongst the range of alternatives. Usually the request is to select one from the range - the one considered true. However, there are certain questions which have more than one true direct answer. For example, the question (6?) What is an example of a woman who became the prime minister of her country? has several. Hence another aspect of the request element of a question is that the 9  degree of completeness  of the selection, measured against the whole set of true  alternatives, be given. A direct answer will consist of a proposition selected from the range of alternatives together with an (implicit) claim that it is true and an (again implicit) claim with respect to the degree of completeness of the answer. To take an example, consider the question  presents a range which is limited in that the example given has to be a prime, but is infinite. 8  Belnap and Steel, The Logic of Questions and Answers,  p. 34.  This form of question would seem to share many features with certain commands. (6?), for example, is similar to 9  (6*) Give me an example of a woman who became the prime minister of her country. Nevertheless, (6?) should not be cast as a command. Not only is an interrogative used in (6?) but not in (6*), the speech act differs.  Questions and  88  Answers  (7?) Which boy wrote this on the blackboard? and the response (7) John Brown did, Sir. (7?) is to be seen as presenting a range of alternatives, X, consisting of a class of propositions ( P .... P^) where the propositions in the class attribute writing a particular Q  word or group of words on the blackboard to the boys determined by the context, those in the classroom perhaps, or those in the school, together with a request that a selection be made from X and the status of that selection be specified. Where the proposition expressed in (7) = P^,, one can cast (7) as Pfr and for all i \  k, not P-, Sir.  We shall later see that this schema has to be expanded for more complicated forms of question but for the moment it will suffice. Let  us  now  turn  to  another  important  feature  of  questions  -  their  presuppositions. The notion of presupposition in the case of assertions is generally associated with problems concerning vacuous terms, with statements which lack a truth value. However, as Belnap argues,  10  that questions carry presuppositions can be seen  without any appeal to examples which involve singular terms which lack a referent.  11  The question asked by  See Belnap and Steel, The Logic of Questions and Answers, pp. 108-118, esp. pp. 110111.  1 0  This is not to say that a vacuous singular term could not be constructed which would be relevant to the circumstances. With respect to (8?), for example, the term 'the period of time during which you loved your wife' might be vacuous. This term does not, however, appear in the question asked. 11  Questions and  Answers  89  (8?) When did you stop loving your wife? presupposes that you used to love your wife and that you have since stopped. (2?) presupposes that you have a car and that colour is a sort of thing that a car has. The question asked by (9?) Where did you put the car keys? presupposes that you had the car keys and that you put them somewhere. The presuppositions that I have mentioned in the above examples are perhaps the most obvious. However there are others. (8?)  12  presupposes not only that you used to love  your wife and that 3'ou have since stopped but also, amongst others, that you have a wife, that loving is the sort of relationship one can have with a wife, and that loving is the sort of thing that can stop. Strictly, these latter presuppositions are perhaps not immediate presuppositions of the question but rather are presuppositions of presuppositions of the question. For example, the presupposition that you used to love your wife has itself the presupposition that you have a wife. Indeed there are cases which can be seen as involving a whole hierarchy of presuppositions. For example  (10?) Is that the girl who's engaged to your wife's sister's boy? presupposes that some girl is engaged to your wife's sister's boy, which presupposes that your wife's sister has a boy, which presupposes that your wife has a sister which presupposes that you have a wife. For certain purposes it may be necessary to treat the various levels of presuppositions separately. For our purposes here, however, it is  For convenience. I shall sometimes talk of (8?), for example, presupposing something or having such and such answers. Strictly, it is the question expressed by the use of the interrogative which has such properties. 12  Questions and  90  Answers  sufficient, and more illuminating, to treat all presuppositions in a unified way, as presuppositions of the question. How are-we to capture the notion of a presupposition of a question? Addressing this question, Belnap begins with Leonard's definition of a presupposition of a question as 'any proposition whose truth is necessary to the validity of the question'.  13  The  problem that now has to be faced is to explicate the notion of the validity.of a question. Belnap argues: Once we begin to think that a fruitful way of ascribing properties to questions is via the answers thereto, we are almost driven to Leonard's own account of 'validity': a valid question is one 'that has a correct answer'. 14  Belnap then combines these two definitions and arrives at the following definition of presupposition: Dl:  A question, q, presupposes a statement, P, if and only if the truth of P is a (semantically) necessary condition for their being some true answer to q.  15  As it stands this definition would not seem to capture what Belnap intended it to capture. As we have seen, there are many possible types of response to a question. One  13  H. Leonard, Principles  of Reasoning, New York, 1959, p. 35.  1 4  Belnap and Steel, The Logic of Questions and Answers,  p. 113.  Belnap uses the term 'logically necessary condition' which may confusingly suggest that the generation of presuppositions is purely a matter of form. Yet 'necessary condition' on its own is insufficient in the definition since it may be a necessary condition of, for example, Mary's having fair hair, that she has the genes she has. Yet it is not envisaged that 'Mary has X Y Z genes' could be a presupposition of 'Mary's fair hair is long'. We are dealing here with neither logical nor metaphysical necessity but rather with a necessity which is connected to the content of a true answer qua content. 15  91  Questions and Answers  type of response is that which involves correcting a presupposition of the question. For example, a reply to (9?) of (9) I never had the car keys. would be a correction of the presupposition that you had the car keys. Now Belnap holds that such corrective responses are answers. Suppose it is true that you never had the car keys. Then the presupposition is false and (9) is true; that is, there is a true answer to the question although a presupposition of that question is false. The matter is easily resolved by amending Belnap's definition so that it reads: D2: A question, q, presupposes a statement, P, if and only if the truth of P is a semantically necessary condition for there being some true direct answer to q. As was noted, the account given of the presuppositions of questions does not rest upon issues concerning possible reference failure. Indeed, Belnap is specifically working with an interpretation in which there are no vacuous terms, in which each statement has a truth value. Since each direct answer will have a truth value, D2 is equivalent  16  to  = D2 : P is a presupposition of q if and only if every direct answer to q semantically implies P .  1 7  To take an example, suppose, in answer to (8?) you were to assert  Van Fraassen, while acknowledging that he is 'following Belnap, who clarified this subject completely', ignores D I and, without comment, employs the following equivalent to D2. 16  Semantic implication could be explicated in possible world terms as follows: for every answer, a, to question q, P is a presupposition of q if and only if in every possible world w in which a is true P is true in w. 17  Questions and  92  Answers  (8) I stopped loving my wife a year ago. Suppose that (8) is false. Then the presupposition, P, that you once loved your wife, may be either true or false - (8) could be false because you stopped loving her at some other time, because you never loved her or because you still do love her. Suppose that (8) is true. If (8) is true, then it must be true that you used to love your wife, that is, P must be true. Hence (8) semantically implies P. A question with a false presupposition, while it will not have a true direct answer, nevertheless is still a question,  18  a question which can be answered directly,  though falsely. Suppose your car is neither blue nor black. Then the presupposition carried by (3), that your car is either blue or black, is false. Nevertheless, (3) is still a question, a question which has the false direct answers M y car is black and M y car is blue. As Belnap argues: 'To ask a question with a false presupposition is not like saying something "meaningless". Rather it is very much like making a false statement. One can do it knowlingly and maliciously, and be exactly on a par with a liar, or one can do it innocently and be subject to exactly the same kind of benevolent correction as a maker of false statements. 1 9  Before broaching the problems stemming from the possibility of vacuous terms, let us continue to discuss Belnap's treatment of presuppositions. Questions, as we have  Remember, Belnap has eliminated any possible reference failure from his system. Belnap and Steel, The Logic of Questions and Answers, pp. 115-6.  Questions and  seen,  carry  many  presuppositions.  However,  93  Answers  the  presupposition,  the  basic  presupposition, of a question is defined by Belnap as follows: The basic presupposition of q is the proposition which is true if and onty some direct answer to q is true. Hence, where the truth of any presupposition is a necessary condition for q's having a true direct answer, the truth of the basic presupposition is both necessary and sufficient for that to be the case. The basic presupposition is, to talk in the metalanguage, that at least one of the direct answers to the question is true. (3?), for example, carries the presupposition that you have a car and that colour is the sort of thing that a car has, but the basic presupposition is that j'our car is either blue or black. The basic presupposition in (4?) is that your car either is or is not black; in (5?) that one of the (relevant) cars is yours. As was noted, Belnap is specifically assuming an interpretation in which every statement will have a truth value. However, if we wish to apply Belnap's insights on the functioning of questions to ordinary language then we must admit the possibility of vacuous terms, and hence of statements which lack a truth value. For assertions, treatment of possible reference failure by means of presuppositions is to be found in the work of Strawson. For Strawson, the use of a sentence A presupposes B if the truth of B is a necessary condition for A's being successfully used to make a statement which has a truth value. The sentence 'Wilfred wears glasses' can only be used to make a true or false statement if there is such a person as Wilfred. Where 'Wilfred' is vacuous then, the existential presupposition being false, that sentence cannot be used to make a statement at all. As Belnap points out,  20  if a standard use of sentences is assumed, and truth,  Belnap and Steel, The Logic of Questions Referring, New York, 1967, pp. 915. 2 0  and Answers,  p. 109. See also Linsky,  Questions and Answers  94  derivatively perhaps, is seen as attaching to sentences, then the difference between Strawsonian presupposition and standard logical implication can be brought out as follows: A  logically implies B if B is a necessary condition for A's being true.  A  presupposes  B if B is a necessary condition for A's having a truth  value.  Now, as Belnap remarks,  21  his treatment of erotetic presupposition  22  is an analogue of  standard logical implication. Consequently, when possible reference failure is admitted, it might seem that we should employ an analogue of Strawsonian presupposition as outlined above, viz: D3 : q presupposes P if and only if the truth of P is necessary for each and everj' one of q's direct answers to have a truth value. The trouble with D3, as indeed is the case with D2, is that it defines the presupposition / direct answer relationship in terms of the truth  of the presupposition.  But what of cases in which the presupposition is not true? Where the presupposition which is not true is an existential presupposition, what then?  2 3  D2 would seem to be  2 1  Belnap and Steel, The Logic of Questions and Answers,  2 2  That is, presupposition as it pertains to the logic of questions and answers.  p. 117.  In the case where 'a exists' does not express anything that is true (where ' a ' is a directly referring term), whether one wants to say that a proposition has been expressed which is false, or that one has been expressed which is neither true nor false, or that no proposition has been expressed at all, will depend upon how one wants to treat existential statements. M y formulation is intended to remain neutral as to these various alternatives. 2 3  Questions and  silent on such a situation.  24  Answers  95  The trouble in the case of D3 is more acute. To see the  problem let us begin by noting that D3 would seem to imply that q's direct answers would lack a truth value. To take an example, consider (11?) Did Martin O'Reilly go home? where 'Martin O'Reilly' is vacuous (that is, the presupposition 'Martin O'Reilly exists' is not true) and the responses (11) Martin O'Reilly has gone home. and (11') Martin O'Reilly has not gone home. Now (11) and (11') are neither true nor false - there is no-one of whom it is either true or false that he has gone (or not gone) home. So far, D3 would seem to be correct. But D3 would seem to admit (11) and (11') as  answers and (11?) as a question, a question  which carries the presupposition P. But where 'Martin O'Reilly' is vacuous how ean the interrogative 'Did Martin O'Reilly' go home?' be used to ask a question - there is no-one of whom it can be asked whether he went home or not. How can the responses to the use of that interrogative, responses which lack a truth value, be answers?  Just as a  failure of the existential presupposition to be true in the case of sentences used to make assertions results in a failure to assert anything that is true or false, to express a proposition at all, so too in the case where an interrogative is used to ask a question. What we need is an account which will be sensitive to the fact that if an existential presupposition is false, no question has been asked which can be answered.  Note that, when vacuous terms, and hence statements which lack a truth value are introduced, the equivalence between D2 and =D2 may not be assumed to still hold. 2 4  Questions and  96  Answers  One conclusion is obvious: we must give up the position that questions carry presuppositions in favour of the more Strawsonian position that presuppositions are carried by the use of an interrogative  (11?),  to ask a question. The use of the interrogative in  although it does not result in a question being asked, does carry  the  presupposition that Martin O'Reilly exists. Further, the definition of presupposition must capture the fact that the status of a presupposition can affect whether or not the use of an interrogative results in a question being asked. The foregoing would seem to suggest the following definition: D4 : The use of an interrogative presupposes P if the truth of P is necessary in order that the interrogative be used to ask a question. As we have seen, D3 is too weak in that it permits questions to be asked whose direct answers all lack a truth value. D4, however, is too strong. We have seen that (3?)  Which colour is your car, blue or black?  presupposes that your car is either blue or black. Suppose that presupposition false your car is in fact red. Now (3?) cannot be (directly) answered truly.  25  Nevertheless (3?)  does ask a question, a question which, though falsely, can be answered. But we cannot 26  simply amend D4 so that it says that all that is required in order that an interrogative The direct answers being M y car is black and M y car is blue. 2 5  The question of whether an interrogative can be used to ask a question is different from that of whether or not the question arises - as in 'When defending your loved ones from those barbarians, the question of fair play does not arise'. The first deals with whether or not a question can be asked; the second with whether its asking is pertinent in the circumstances. 26  Questions and  97  Answers  be used to ask a question is that the presupposition have a truth value - as we saw above, existential presuppositions generated by the use of proper names have to be true in order that a question be asked. The same intuition which guides us with respect to the truth conditions of an assertion of a well-formed declarative sentence containing a directly referring term in subject position can help us here. In the case of such an assertion, if the directly referring term which it contains does in fact refer, the existential presupposition that that thing exists is satisfied, then, assuming all else is well formed, a proposition will have been expressed which is either true or false. However if the directly referring term lacks a referent, the existential presupposition is not true, then no proposition will have been expressed and, consequently, nothing said which is either true or false. Similarly, in the case of the use of an interrogative to ask a question, if the existential presupposition generated by the use of a directly referring term in the interrogative is true then a question has been asked, a question to which an answer possessing a truth value can be given. If the existential presupposition is false, if there is no referent to the singular term, then no question has been asked and hence no answer possible which is either true or false. Consequently, the following definition arises of erotetic presupposition for cases where the singular terms involved are directly referring ones : 27  The qualification is necessary in that the case is not the same for presuppositions generated by the use of a definite description (used attributively), in particular for existential presuppositions so generated. For example the interrogative 'Did Fred climb the tallest mountain in Tasmania' presupposes that there is a tallest mountain in Tasmania. Suppose that presupposition not to be true - that there are two mountains of equal height sharing the honour, 'the tallest mountain in Tasmania' would then lack a referent. Nevertheless, although the existential presupposition would not be true, a question would have been asked. Remember that under Kaplan's treatment of definite descriptions the contribution made by such a term to the proposition is not the object which is the referent but a blueprint for finding the referent. Consequently, although the description does not denote, although the presupposition is not true, a question has been asked. 2 7  Questions and  Answers  98  D5 : The use of an interrogative presupposes P if and only if (1)  P's having a truth value or, where P is an existential  statement generated by the use of a directly referring term, P's being true, is necessary in order that the use of the interrogative result in a question, q, being asked, and (2) P is implied by all direct answers to q. Hence, by possessing the correct (relevant) truth value, presuppositions perform two functions: they permit a question to be asked by the use of the interrogative, and they permit the question to have a true direct answer. Note that, where any singular term involved is a directly referential one,  28  since all existential premises must be true in  order that the use of an interrogative result in a question being asked, and since a question is identified with the set of its direct answers, all direct answers must have a truth value. Hence the second conjunct can be incorporated without a qualm. # * # * • *  *  *  *  *  *  *  *  *  *  *  *  *  Given the notion of presupposition and that of direct answer, certain types of question and answer can be delineated. A n answer is uninformative if it is implied by the basic presupposition of the question. A p a r t i a l answer is one implied by a direct answer (but not conversely), an answer which implies a direct answer is complete. A question is empty if all its direct answers are necessarily true, and foolish if none of its direct answers are even possibly true. Van Fraassen gives, as an example of the latter,  Did you wear a hat which is both black and not black, or did you wear one which is both white and not white?  But not in cases where definite descriptions (used attributively) are concerned. See n. 27 above. 2 8  Questions and  Answers  99  A direct answer which is necessarily false is also called foolish. Any direct answer 29  which is not foolish is possible. A further type of answer is illustrated by the following question and answer: cj : Did you wear a black hat yesterday or did you wear a white one? a : I didn't wear a white one. Note that A's reply is not a partial answer - it is not implied by either direct answer. The respondant could have worn both hats that day, one in the morning and one later. However, given the basic presupposition of q, that she at least wore one of the two, a, together with that presupposition, imply the direct answer that she wore a black hat. Consequently, we can define a relatively complete answer as follows A relatively complete answer to q is an answer which, together with the basic presupposition of q, implies some direct answer to q. This can be further generalized. For example, a complete answer to q, relative to a theory T, is something which, together with T, implies some direct answer to q. The point, as Van Fraassen remarks, is  that we should regard the topology of answers as open ended, to be extended if needs be when specific sorts of questions are studied. 30  There remains to discuss a further type of response to an interrogative used to ask a question - those which deny the truth of a presupposition. The lack of truth of a 31  2 9  3 0  Hence Belnap proves the theorem: Ask a foolish question and get a foolish  answerl  Van Fraassen, The Scientific Image, p. 141.  A distinction must be kept in mind here between using an interrogative to ask a question and a question being in fact asked by a given interrogative used with that intent. As with assertions, sometimes we try to ask a question but don't succeed. 31  100  Questions and Answers  presupposition can have two effects, depending upon which type of presupposition is involved: if it is an existential presupposition it can result in no question being asked, and, where the presupposition is not an existential one, it can result in a question having no true direct answers. Consider the following responses to (3?):  (3 ) My car is neither blue nor black. a  (or the abbreviated 'Neither') and (3 ) I don't have a car. b  (3 ) denies the truth of the basic presupposition (that your car is either blue or black), a  (3^) denies the truth of the existential presupposition (3x)(x is a car and x is yours). (3 ) a  results in the claim that the question cannot be (directly) answered truthfully; (3^) that no question has been asked. Hence (3 ) cannot be seen as an answer to (3?) since, in effect, it denies the existence of the question. With respect to (3 ), since (3 ) merely a  a  denies the validity of the question (that is, it claims that there are no true direct answers) but not that a question has been asked, we could see (3 ) as a sort of answer a  the best the respondant, trying to be truthful, can give.  32  K  The responses (3 ) and (3 ) deny the truth of certain presuppositions. That is all they do. Consider now:  (3 ) My car is red. C  This is the type of example which Belnap concentrates on. Since he is assuming that all existential presuppositions are true, he concludes that all cases of correction are answers, answers which concern themselves with the possibility of there being a true direct answer. However, as I go on to argue, this is not always the case even for those cases which concern a non-existential presupposition. 3 2  Questions and  Answers  101  as a response. (3 ) involves a denial of the same presupposition as ( 3 ) , the basic C  a  presupposition. But it does more. (3 ) communicates new information, information not C  requested by (3?). Hence, although an effect of (3 ) is to deny a presupposition of (3?), it C  is not primarily an answer questioner should  to (3?). It addresses the more basic question which the  have asked -'What is the colour of your car?'. Consider now  (3 ) I have a truck. d  This cannot be dealt with quite so straighforwardly. It could be true that A has a truck and a car - that is, (3 ) need not necessarily be denying the truth of the presupposition that A has a car. A could be changing the subject. However, if A is not changing the subject, if (3 ) is given as a response to Q's use of the interrogative, then (3 ) has the d  d  effect of denying the existential presupposition, hence claiming that no question was asked. Nevertheless, (3 ) is not an answer to (3?) - how could it be if no question is d  asked - but is an answer to the question 'What kind of vehicle do you have?'. So we see that the intial treatment of correction given in the previous chapter is born out. A corrective statement which contains new information, that is, information not contained within the range of alternatives, is best dealt with as an answer to a question other than that in fact asked. That completes our examination of the functioning of questions and answers for the present. The treatment given being sufficient for present purposes we can now return to the examination of belief attributions as answers.  102  5. B E L I E F A T T R I B U T I O N S A S A N S W E R S .  Having examined the question / answer relationship in more detail, we are now in a position to look more closely at the belief attribution considered as an answer. In the brief examination of belief attributions as answers carried out in Chapter 3, a distinction was noted between cases in which the attribution is essential in order to answer the question directly and those in which it is not. In the latter case the attribution is made because the attributor is not capable of answering the question directly. Such differences are reflective of the differing constraints placed upon the attributor by the nature of the question to be answered. As we turn now to a more detailed examination of the belief 1  attribution as an answer such cases will be treated separately.  B e l i e f A t t r i b u t i o n s as D i r e c t A n s w e r s .  We have seen that a sentence can be asserted to convej' very different information. Consider a sentence of the form 'B believes that P' asserted as follows: (1) B believes that P. (2) B b e l i e v e s that P. (3) B believes that P.  Or by the aspect of the assertion which is to be contradicted. For convenience I shall concentrate on cases in which the general purpose is the imparting of information. Since a belief attribution made for the purpose of correction is itself to be treated as answering a question (though a different one than that perhaps asked), my discussion of attributions made in order to pass on information as being an answer to a question will cover such cases too. 1  Belief Attributions  as  Answers  103  where the bold print represents the stress carried by that part of the asserted sentence. Corresponding questions to (1) - (3) are: (1?) Who believes that P? (2?) What is B's attitude towards PI (3?) What does B believe? As was argued, a question presents certain propositions from which the addressee is requested to chose. The differences between the three questions can be seen in the differing sets of propositions they present: (1?):  (A believes that P) (B believes that P) (C believes that P) (for however many people are delineated by the context)  (2?):  (B believes that P) (B doubts that P) (B does not believe (disbelieves) that P)  (3?):  (B believes that P) (B believes that P') (B believes that P") (  )  (1?) requests that a person be chosen who stands in the belief relation to a given proposition, (2?), that the relation be specified between a given person and a certain proposition, while (3?) requests that a proposition be chosen from among those offered as being the one with which a given person stands in the belief relation to. When the assertions (1) - (3) are treated as answers their differences become clear. If we call the different sets of propositions presented by (1?), (2?) and (3?), '< 1>'. ' < 2 > ' and ' < 3 > ' respectively, then as answers, (1), (2) and (3) can be cast as follows:  Belief Attributions as Answers  104  (1) B believes that P, in contrast to the rest of < 1>. (2) B believes that P, in contrast to the rest of <2>. (3) B believes that P, in contrast to the rest of < 3 > . Just as questions (1?), (2?) and (3?) differ in that the set of propositions offered in each case differ, so the answers (1), (2) and (3) differ with respect to that which S's believing that P is contrasted with. Concerning as they do questions as to what constitutes believing that P as opposed to doubting that P, I shall not here be concerned with attributions of type (2). Nor shall I be concerned with those of type (1) since, as we shall see, in relevant respects they resemble those of type (3) sufficiently to be covered by my discussion of the latter type. Let us then examine attributions of type (3) more closely. Immediately it becomes clear that the formalization '£? believes that P ' conceals a distinction. If 'P' is replaced by, for example, 'Alan smokes', then one can see that even within attributions of the third type very different information can be conveyed, depending upon whether A asserts (4) B believes that Alan smokes. or (5) B believes that Alan smokes. Similarly, were 'P' to be replaced by a sentence which included reference to time and place, were 'P' to be, for example, 'Alan flew to Toronto yesterday', then the sentence could be used in many ways to assert different information. A could assert not only (6) B believes that Alan flew to Toronto yesterday. and (7) B believes that Alan flew to Toronto yesterday.  Belief Attributions as Answers  105  but also (8)  B believes that Alan flew to Toronto yesterday.  (9) B believes that Alan flew to Toronto yesterday and (10) B believes that Alanflewto Toronto yesterday. Clearly, given the different information which can be conveyed by attributions of the type 'B believes that P' we cannot simply categorize such attributions, as we have done, as all answering the question 'What does B believe?' (8), for example, answers the question (8?) When does B believe Alan flew to Toronto? and (9) (9?) Where does B believe Alan flew 3'esterday? while the questions to which (10) and (6) are answers are, respectively, (10?) How does B believe Alan got to Toronto yesterday? and (6?)  Who does B believe flew to Toronto yesterday?  (7) would seem to be the best candidate for an attribution that does answer the question 'What does B believe'. But even here the information being given is not that B believes  Belief Attributions  that  106  as Answers  Alan flew to Toronto yesterday (as opposed to what?) but information about 2  what B believes about Alan.  3  So the question to which (7) is a direct answer is  (7?) What does B believes about Alan? Hence we see that the types of question to which a given belief attribution can be asserted in direct answer are many and varied and hardly ever, if at all, is the type that of 'What does B believe?'. The next point of interest concerns the relationship of the desired type of answer to a particular question. Consider the following example: The family are discussing Jane's new boyfriend, Phil. In particular, they are wondering what he does for a living. No-one present seems to know. 'What about Betty?', M a asks of A. (Betty is not present.) 'She's met him, hasn't she. Doesn't she know?' A replies: (11) Betty believes that Phil is a lawyer.  2  I suppose someone could ask the question What does B believe?  when he was just wanting an example of one of the beliefs (as opposed to doubts) which B had, regarding no subject in particular. For the life of me I cannot imagine why anyone would ever do so. 3  This is not to say that B believes that  Alan flew to Toronto yesterday.  could not be given in direct answer to a question that begins 'What does B believe...?'. It might, for example, be a direct answer to What does B believe about the goings on next door? or What does B believe about the crisis at work? Where the question concerns S ' s beliefs about a state of affairs then a direct answer to that question will have the form 'B believes that o is F (or a F's)'. M y point is simply that often, in conversation, we do not completely specify the question by means of the interrogative used but rely on contextual elements to complete it.  Belief Attributions as Answers  107  Note that, despite Ma's question, A is not answering (12?) What does Betty know about Phil? but rather (11?) What does Betty believe about Phil?  4  Now such a question presents a vast array of propositions in the set offered. Among the set will be propositions such as (Betty believes that Phil is a lawyer) (Betty believes that Phil is a bank clerk) (Betty believes that'Phil is a bus driver) and so on. The set will also, however, contain such propositions as (Betty believes that Phil has blue eyes ) (Betty believes that Phil has brown eyes) and (Betty believes that Phil has two legs) (Betty believes that Phil has one leg) (  :  )  So were A to assert (11*) Betty believes that Phil has two legs. in reply, he would have given a direct answer to (11?). Yet clearly (11*) is not the sort of answer requested given the conversational circumstances. M a is not interested in Betty's beliefs about the number of legs Phil has. She is interested in Betty's beliefs If B is being open and sincere then he cannot know what Betty knows else he would have volunteered that information in the first place. 4  Belief Attributions  as  108  Answers  about Phil only insofar as they are related to the main interest of the conversation - the topic  question of what Phil does for a living. It is Ma's intent to answer that question which prompted her to ask (11?) of A. More is demanded of an answer given in a conversational context than that it be a direct answer to the question asked. The answer has to be, in some way, relevant. The notion of relevance has proved to be a particularly difficult one to define. So far as I have seen, a satisfactory account of just what makes one particular assertion relevant to another in a given conversational context has yet to be given.  5  However,  given the analysis of question and answer outlined above, we can, at least as far as is sufficient for our present purposes, get some way towards resolving the issue. Consider A's answer of (11) Betty believes that Phil is a lawyer, not in the light of Ma's question (11?) What does Betty believe about Phil? but with respect to the topic question (12?) What does Phil do for a living?  The reason why (11) is relevant and (11*) is not is that (11), but not (11*), provides some sort of an answer to the topic question. Note that whereas the embedded  This is not for want of trying. See, for example, T. van Dijt, 'Relevance Assignment in Discourse Comprehension', Discourse Processes, 2 (1979), 113-126; M. Dascal, 'Conversational Relevance', in Meaning and Use, ed. A. Margalit, Dordrecht, 1979, pp. 153-174; P.N. Werth, 'The Concept of Relevance in Conversational Analysis', in Conversation and Discourse, ed. P . N . Werth, London, 1979, pp. 129-54; and D. Wilson and D. Sperber, 'On Defining "Relevance"', in Philosophical Grounds of Rationality, eds. R.E. Grandy and R. Warner, Oxford, 1986, pp. 243-258. 5  Belief Attributions  as  109  Answers  proposition Phil is a lawyer occurs in the set offered by the topic question, Phil has two legs does not. Upon receiving A's answer, if she trusts Betty's judgement, Ma herself has the means to answer the topic question directly. Let us call a question asked within the framework of a topic question a subsidiary  question. Only if a direct answer to a  subsidiary question bears a certain relation to a member of the set offered by the topic question can it be relevant. In cases where the direct answer to the subsidiary question is of the form 'B  believes that P\  it would seem that we could conclude that such an  answer will only be relevant if_P is among the set offered by the topic question.  6  This, however, would be too quick. Consider the following example. A  : Mary is nuts. Do you know what she's dreamed up now. She thinks she can tell what someone is like, all the details of their life, by the kind of shoes the}' wear. She thinks that Fred is very rich. I didn't like to disillusion her and tell her he's constantly struggling to pay his mortgage.  Ma : What does she believe about Phil? A : She believes that Phil is a lawyer. Here the topic of conversation is Mary herself and the topic question roughly (13?) What does Mary believe about her acquaintances on the basis of her views on shoes?  In assuming that there is but one topic question I am, for the purposes at hand, simplifying the situation. Often there is not one but a whole hierarchy of topic questions, their inter-relationshipships being exceedingly difficult to define. For example, a topic question, TI ('What is Alan doing these days?') can itself be asked within the framework of another topic question, T2 ('Is Alan available to come to dinner tonight?'). Now although the embedded proposition of a belief attribution ('Betty believes that Alan flew to Toronto yesterday') given in answer to a subsidiary question to TI will, if it is a relevant answer, be contained within the set offered by TI, it need not be contained within the set offered by T2. ('Is Alan available to come to dinner tonight?' does not contain, within the set offered, Alan flew to Toronto yesterday.) Such an answer, while relevant to TI, will only be as relevant to T2 as a direct answer to T l would be relevant to T2. This example illustrates that there is much more to the notion of relevance than is being treated in the text. 6  110  Belief Attributions as Answers  Ma's question to A is a request to provide yet another example of those beliefs. Hence A's answer is not only a direct answer to Ma's question, being relevant (assuming that Mary has come to her belief about Phil on the basis of his shoes), it also provides a direct answer to the topic question. Unlike the previous case, here, where the topic question concerns what the believer believes, it is not that the attribution is relevant because _P is contained within the set offered by the topic question but because the proposition B believes that P is contained within that set. Only where the topic question concerns what B believes will a direct answer to a subsidiary question of the form 'B believes that P' constitute a direct answer to the topic question. Hence, given the two relationships which may hold between a subsidiary question and the topic question, we can conclude that a belief attribution of the form B L  believes that P' which constitutes a direct answer to a subsidiary 'what' question is relevant if and only if, in cases where the topic question does not concern B's beliefs,_P is contained within the set offered by the topic question or, in cases where the topic question does not concern the believer's beliefs, B believes that P is among that set.  7  The differing relationships which may hold between a subsidiary direct answer to a 'what' question and the topic question also hold in cases where the attribution is a direct answer to a 'where', 'when', 'who' and 'how' subsidiary question. For example, 8  9  suppose that M a , looking round the members of her club, asks  The distinction between the two cases is usually marked by differing stress given to the question asked. 'What does Betty believe about Phil?' signals that the subject is being changed temporarily in order to gain information about the topic question. 'What does Betty believe about PhilT signals that information about a particular instance of the topic question is being sought. 7  8  As in 'Who does B believe is Fl'.  Note that 'why' questions have not been dealt with. They require separate treatment and will be discussed below, see pp. 145-151. 9  Belief Attributions  as  111  Answers  (14?) Who's a good person to ask about legal problems? No-one ventures an opinion. The conversation continues as they leave the club and hop on a bus. Remembering that Betty is fairly well connected, M a asks (15?) Who does Betty believe is a good person to ask about legal problems? Now Betty may believe of many people that they would be good to ask about legal problems, people in Ottawa, Toronto, New York... Nor is the set offered by (15?) limited by the context in which it is asked (no-one is interested in the people on the bus.) Rather, the relevant direct answers to (15?) are determined by the context of the topic question. Also, if A's answer to (15?) (15) Betty believes that Dan is a good person to ask about legal problems. is relevant, then the embedded proposition is contained in the set offered by (14?). Alternatively, M a could have asked (15?) in the context where A had been telling her of Betty's weird notion about people's shoes. In which case A's answer, if relevant, would be contained, completely, in the set offered by the topic question. Hence any attribution of the form B believes that P' which constitutes a direct l  answer to a subsidiary question, irrespective of the form of that question will, if relevant, possess one of two distinct relations to the topic question: either the (entire) attribution will express a proposition contained in the set offered by the topic question or the embedded proposition expressed by the attribution will be offered by that set.  10  The case is different for attributions of the form 'B believes that P' which answer the question 'Who believes that PI'. There, the topic question always concerns something other than B's beliefs - usualty some aspect of P itself. Hence such attributions will be covered by my treatment of attributions of the form 'B believes that P' whose topic question is something other than S's beliefs. 1 0  Belief Attributions  as  Answers  112  B e l i e f A t t r i b u t i o n s as Tentative A n s w e r s . The attributions to w^'-.V* we now turn do not constitute a direct answer to the question asked. Consider the following question and answer pairs: (16?) Who else is coming to the meeting? (16) Bill believes that S a m is coming to the meeting. (17?) What is Sam doing? (17) Bill believes that Sam is c o m i n g to the meeting. (18?) When is Sam coming? (18) Bill believes that Sam is coming to the meeting. That the answers given in such cases are not direct answers can easily be seen by examining the range of propositions offered by the questions concerned. For example, the questions above could offer: (16?) :  (Joe is coming to the meeting) (Bob is coming to the meeting) (Sam is coming to the meeting)  (17?) :  (Sam is coming to the meeting) (Sam is going for a walk) (Sam is working in his office)  (18?) :  (Sam is coming just before the meeting) (Sam is coming to the meeting) (Sam is coming long after the meeting)  Since the proposition expressed by 'Bill believes that Sam is coming to the meeting' is not one of those offered by (16?), (17?), or (18?), (16), (17) and (18) are not direct answers to those questions.  Belief Attributions  as  Answers  113  Yet if not direct answers, (16), (17) and (18) are, nevertheless, some kind of answer to the respective questions. Reflections on such examples in the light of the theory of questions and answers yields, if not an obvious solution to their correct treatment, at least a means of characterizing such answers. A l l such cases involve a question which concerns part of the content of the proposition which occurs in the attribution; that is, all of the questions present, among the set of propositions offered, that proposition  which is  embedded in the attribution. A selection from the set offered is  made, but it is not asserted as it stands but is embedded within the belief attribution. Note that the selection is logically prior to the embedding - one first selects the proposition and then says of it that someone believes it. Such answers, which I propose to call tentative answers, can then be distinguished from amongst the many other possible forms of answer in just this respect. A tentative answer is one which selects a proposition from among the set offered and says of it that some individual (or individuals), who may or may not be specified, believes (believe) it. To so define a tentative answer is not, however, to explain how it functions as an answer. How, for example, does A's assertion of (a) Bill believes that M i k e y hit Joey. function as an answer to Ma's question  (a?) Who hit Joey? An initial response could be that (a) provides M a with a possible answer to her question: that (a) in effect says - it could be Mikey who hit Joey, Bill believes that it was. The function of the attribution in that case would seem to be to present the reason for declaring the possibility. However, to see (a) as presenting a possible answer cannot  Belief Attributions as Answers  114  be right. None of (a?)'s direct answers are foolish, that is, necessarily false. Suppose £he set of propositions offered by (a?), given the context, is (Jamie hit Joey) (Mikey hit Joey) (Bobby hit Joey) None of the propositions offered, if expressed by an assertion, would result in an answer that is necessarily false. In particular, an assertion of (b) M i k e y hit Joey. would at least be possibly true. So to assert (b) as a possibility, that is to assert (c) <>(Mikey hit Joey) would be simply to state the obvious. Of course it is possible that it was Mikey - M a already knows that much.  11  A's claim as to the truth of (b) is stronger than that it is  merely possible. Yet (a) is not a direct answer. If the claim of truth for the embedded proposition in (c) is too weak, that in (b) is too strong. Yet if (a) is not a possible answer nor a direct answer, what is the relationship between (a) and (a?)? Help can, I think be found by examining assertions of a type not 3'et discussed: those which involve a parenthetical use of a parenthetical verb. In an insightful paper, J.O. Urmson  distinguishes parenthetical verbs  as those which  have no  12  present  continuous tense. Some examples are 'suppose', 'guess', 'conclude;, 'suspect', and  Of course, (a) also gives M a the information that Bill believes it. However, while this must contribute in some way to the answer, it does not explain the relationship between a tentative answer and the question. 11  12  'Parenthetical Verbs', Mind, 61 (1952), 480-496.  Belief Attributions as Answers  'believe'.  13  115  They are further distinguished grammatically in that, in the First person  present indicative, they can be used in a sentence in three different ways: they can be followed by 'that' and an indicative sentence - 'I suppose that Ronald is very rich.' they can be inserted in the middle of the sentence 'Ronald is, I suppose, very rich.' • or they can occur at the end of it 'Ronald is very rich. I suppose.' While parenthetical verbs may have non-parenthetical uses, their principle parenthetical use is in the first person present indicative. So used, Urmson argues, parenthetical verbs do not provide psychological descriptions. When such a use is contrasted with the use of a descriptive verb, the difference becomes clear: (i) He is, I regret, unwell. (ii) I am miserable because he is unwell.  While (ii) reports a psychological state, (i) does not. Note that we cannot say (iii) He is, I am miserable, unwell. nor (iv) I regret because he is unwell.  Although, for example, 'I am running' or 'I am typing' is perfectly good English, 'I am suspecting' or 'I am concluding' is not. 13  Belief Attributions as Answers  116  In (ii), 'because he is unwell' gives the cause of the assertor's psychological state. The absurdity of (iv) lies in there being no psychological state put forward for which such a cause can be given. In order to see how parenthetical verbs function, we must first examine the claim to truth associated with assertions of sentences in which they occur. If one asserts a sentence in normal circumstances (one is not, for example, performing on a stage) there is an implicit claim as to the truth of what is being expressed and we will be taken by our audience to be so claiming. However, as Urmson argues,  14  in the case of  sentences which contain a parenthetical verb occuring in the first person singular together with an embedded indicative sentence, S, there is a claim not only that what is expressed by the whole sentence is true but also that what S itself expresses is also true. When one asserts (v) He is, I regret, too old. or (vi) I conclude that he is too old. one claims not only that (v) and (vi) are true but also that He is too old is also true. Similarly with 'believes'. In I believe it will rain there is the claim that It will rain is true, in  Ted will come, I believe, that Ted will come is true.  Urmson's point is not quite as given as he talks in terms of sentences being true.  Belief Attributions as Answers  117  What then is the function of the parenthetical verb? It is to point out to the 15  audience some aspect of the proposition expressed by the embedded sentence. We may prime our audience to see the emotional significance attached to the embedded proposition, as in He is, I regret, unable to come to the meeting. or its logical relevance I conclude that Friday is the best day for the meeting. or, the important case, the reliability of the assertion She is, I suppose, at home. I guess that the penny will come down heads. He is, I firmly believe, thefirstin line for the job. Parenthetical verbs of the latter type, perhaps modified by adverbs, concern themselves with the degree of truth claimed for the embedded proposition. They may uphold, modify or weaken that claim. When such verbs are used it is not stated that the claim is, in the appropriate way, being affected; rather it is shown. As Urmson concludes 'I guess that this is the right road to take' is a way of saying that this is the right road, while indicating that one is just plumping and has no information, so that the statement will be received with the right amount of caution; 'I know' shows that there is all the evidence one could need, and so on... We are, in fact, in a position where we can either make our statements 'neat', and leave it to the context and the general probabilities to show to the hearer how much credence he should give to the statement; or, in addition to making a statement we can actually describe the evidential situation in more or less detail; or give a warning  For convenience I will sometimes speak of the functioning of parenthetical verbs as a shorthand for the parenthetical use of a parenthetical verb. 15  Belief Attributions  as  118  Answers  such as 'Don't rely on this too implicitly, but...'; or I can employ the warning device of a parenthetical verb 'I believe it will r a i n ' . 16  How then does 'believe' modify the claim to truth? It would seem to depend upon how it is pronounced - straightforwardly, the claim to truth is modified but slightly; with the second syllable heavily accentuated and drawn out, quite a bit. That 'believe' is used to modify, if normally only slightly, the claim to truth can be seen in the way its use is exploited for politeness' sake. The spluttering colonel who asserts Madam, you are, I believe, standing on my foot. is in no doubt as to the fact of the matter, nor is he taken to be in any doubt. The convention is being appealed to, with the result that the assertion is softened - the colonel is being polite. Many of our uses of 'I believe' are like this. We often use the construction in order not to appear too aggressive or 'know-it-all' in the eyes of our superiors: I believe I wrote the answer at the bottom of the page, sir. Yet, whether we are actually indicating a small doubt or simply utilizing a convention, the resulting force of the assertion, its claim to truth, is modified. Here, in the device of a parenthetical verb used parenthetically we have the means by which to span the gulf between direct and possible answers. When viewed as an answer, an attribution involving a parenthetical use is never direct. Although the claim to truth may be modified hardly at all, as in the cases which exploit the convention, it nevertheless is modified. Depending upon the verb used and how it is modified, we can affect the claim to truth of the embedded proposition. So doing, we can  Urmson, 'Parenthetical Verbs', pp. 485-6.  Belief Attributions  as  Answers  119  affect whether the answer given is almost direct or closer to being merely possible. Since such answers reflect a degree of uncertainty as to the truth of the embedded proposition, the term 'tentative answer' is apt.  17  However, the initial examples of tentative answers given above do not contain parenthetical uses of a parenthetical verb. The verbs in those examples are not in the first but the third person singular. When, in answer to the question (a?) Who hit Joey? A asserts (a) Bill believes that Mikey hit Joey. A is not indicating that the claim to truth Bill would make with respect to the proposition Mikey hit Joey is in any way modified: he is not attributing a parenthetical use to Bill. There is no reason to suppose that Bill would modify his assertion of 'Tommy is coming' in any way whatsoever.  1 8  I do not wish to argue that such attributions involve an extension of the parenthetical usage. I do not equate a tentative  answer with one involving a  parenthetical use of 'believe'. While all answers which involve a parenthetical use of 'believe' are tentative answers, not all tentative answers involve a parenthetical use. Rather, what I wish to retain from cases involving parenthetical usage is the fact that the claim to truth can be shown to be modified.  While such answers are normally given when the speaker is in an uncertain frame of mind with respect to the truth of the proposition expressed, the term 'tentative answer' should not be taken as a psychological term indicating the speaker's frame of mind but rather as characterizing a certain type of answer, viz. one in which the degree of truth claimed is shown to be modified. 1 7  Urmson would seem to take the opposing view on this point, see Urmson, 'Parenthetical Verbs', pp. 488-9. 18  Belief Attributions  120  as Answers  Have we not, in answers given to questions concerning an element of the embedded proposition which involve attributions of belief to another, just such a modificiation of the truth claim? That there is a claim to the truth of the embedded proposition, however that claim be modified, is given by the fact that the attribution is given in answer to that particular  question. A proposition has been selected from among  those offered. However, unlike a parenthetical use of the verb, the modification of the claim to truth is not shown by some element in the assertion made. Rather, the claim to truth is shown to be modified by the fact that the assertion is offered in answer to that particular question. While the attribution would not be made in answer to the question if A thought it merely possible that Tommy was coming, neither would it be made if A himself believed that Tommy was coming. It is the very fact that A makes the attribution in answer to the question rather than answering directly which shows that the claim to truth of the embedded proposition is modified. *  *  *  *  *  :|: *  *  *  *  *  *  :|: *  *  *  *  Non-Problematic Uses of Belief Attributions; Summary. This  completes our present  study  of the non-problematic  use of belief  attributions. It will perhaps be useful at this stage to review our findings so far. The treatment of a belief attribution as an answer was initiated, it will be remembered, when it was noted that very different information could be conveyed by the assertion of a sentence of the form 'B believes that a is F" in the same (physical) context. While the differing information communicated in the various cases could not be captured by means of the proposition expressed by the asserted sentence, it could be captured by treating the belief attribution as an answer to a question. When viewed thus, a sharp dichotomy emerged between belief attributions asserted in direct answer to the question and those asserted because a direct answer could not be given.  Belief Attributions  as  Answers  121  In the case of direct answers we have noted the many differing types of questions to which the belief attribution could be given in answer, hence providing, in the various cases, very different information. The same holds true in the case of tentative answers. For example, 'B believes that Mary will come here tonight' can be asserted, with differing stress, as a tentative answer to When is M a r y coming here? Who is coming here tonight? Will Mary come here tonight? and so on. In the previous chapter the relationship between a direct answer and the question was studied in general and it was seen that in order to constitute the kind of answer being sought by the questioner, in order to be a direct answer, the proposition expressed had to be among that group offered by the question. In this chapter, consideration of the use of belief attribution as a direct answer revealed that in such circumstances more is demanded of an answer if it is to constitute the kind of answer being sought by the questioner than that it be simply direct - the answer must also be relevant. When and if an answer is relevant depends upon certain relations holding between the propositions expressed by the belief attribution and those expressed by the topic question. Tentative answers were then examined more closely and an analogy was drawn to cases of paranthetical verbs used parenthetically. When a belief attribution is made in order to supply a tentative answer, it was argued, the attribution is made in order to modify the claim to truth accompanying the embedded proposition.  122  6. SUBSTITUTION.  The phenomenon of the substitution of one proper name for another within a sentence embedded within another used to attribute a belief presents grave problems for the direct reference view. Ultimately the best evidence that we have that someone, B, believes something lies in his willingness to assert it and in his concurrence with the belief attribution. However, if B asserts 'Fa' and concurs with the attribution (1) B believes that Fa. where a = b, B may well refuse to assert Fb' and deny the truth of l  (2) B believes that Fb. Given 5's denial it would seem that (2) is false. But, given the truth of (1) and the coreferentiality of the names, on the direct reference view (2) would be true. Not only does it seem that the direct reference theorist goes against one's natural inclinations in insisting that (2) is true, he also faces the more insistent problem: given the truth of (1) and the co-referentiality of names, how, on the direct-reference view can B deny the truth of (2)? Let us first look at how, when and why a name, any proper name, comes to be used in an answer in thefirstplace. In a direct answer, one way a name can arise is in 1  Unless otherwise specified, for ease of handling I shall assume that the attributor, A, is correct, in cases where two names are involved, in believing that, where there is a referent, the two names co-refer. I shall further assume that A is also right in that the believer does have a belief (or, not to beg the question, thinks he has a belief) which he would express (or attempt to express) partly by using the name used by A or a coreferential one. 1  Substitution  123  response to a question posed by means of an interrogative which itself contains a name. For example the questions (3?) Is Miss Jones smoking? (4?) Which office is Miss Jones's? (5?) When is the present being given to Miss Jones? (6?) Did Miss Jones get the job? all offer a set of propositions each of whose members contain Miss Jones herself. The set offered by (6?), for example, is ( £ got the job) ( ^ did not get the job) (where the stick figure is doing duty for Miss Jones). What must A do in order to answer the question directly? He must select one of the propositions in the set and, by making an assertion, so express it; that is, he must chose a sentence which, when asserted, expresses just that proposition. In order to do so, since Miss Jones herself occurs in the proposition, he must use a directly referring term for Miss Jones. Again, a name can occur in the sentence used to provide a direct answer when the interrogative used to ask the question does not contain either that name or a coreferential one, for example,  (7?) Who got the job? (8?) Who is on the committee? (9?) What is the highest mountain in the world?  Substitution  124  The members of the sets of propositions offered by such questions are all singular propositions, there being one singular proposition for each person or object delineated by the question or by the question and the context. Suppose the context yields the following set for  (7?): ( & got the job) ( -£ got the job)  where ^  and ^ are doing duty for Miss Jones and Charles Wright respectively. As  before, A, in order to answer the question directly, must chose one of the propositions in the set and, by using a sentence which when asserted expresses just that proposition, so assert it. Suppose now that the lad}' involved has two names. She is none other than Miss Laura Jones of a previous example who, though known to her family and friends as 'Mrs. Laura Robinson' (ever since she married Herbert Robinson), has decided to retain her maiden name for professional purposes. All her co-workers at the office know her by the name of 'Miss (Laura) Jones'. Hence 'Miss Laura Jones' and 'Mrs. Laura Robinson' (and even perhaps 'Mrs. Herbert Robinson' ) refer to the same woman. 2  So if, with respect to a question asked by means of an interrogative which contains one of those names, A wishes to provide a direct answer, he could do so by using either one of the co-referential names. If, for example, in answer to (6?) A wants to express  The last name is an interesting one and not, so far as I know, ever commented upon in the literature (due to feminist feelings perhaps?). Is 'Mrs. Herbert Robinson' to be treated as a proper name of Laura or as the name of a position or role which Laura happens to fulfill? If Laura had not married Herbert she would not be called 'Mrs. Herbert Robinson'; if someone else had, that woman would take on the name. So it would seem that the name functions as a disguised description, picking out whoever it is who satisfies being married to Herbert. However, given that Laura married Herbert and so has (if she is willing) been given the name, does not this constitute a form of dubbing? 2  Substitution  125  got the job he could do so by using either 'Miss Laura Jones' or 'Mrs. (Laura) Robinson'. However, if A uses a name other, than that used in the interrogative he runs the risk that the questioner may not understand his answer as a direct answer. Suppose the questioner is Laura's co-worker, Caruthers. Caruthers only knows her by the name 'Jones'. Were A to reply (6) Mrs. Robinson got the job.  3  how would Caruthers understand A's reply? He might figure out that 'Mrs. Robinson' was another name for'Laura Jones, a name preferred by A. But he is far more likely to conclude that something has gone wrong with the communicative process. He might think that A had misheard him, mistakenly thinking he had asked if someone called 'Mrs. Robinson' had got the job. Or Caruthers might think that, despite the stress given to his reply, A is in fact answering (7?) Who got the job? and, so answering, is indirectly implying that Miss Jones had not got it. If A is to 4  guarantee against such a misunderstanding, he had better use the name the questioner used. So although it is not necessary to use the same name as that used in the interrogative in order to answer the question directly, it may well be necessary in order to safeguard that the answer is communicated.  Note that A does not reply (6') M r s . R o b i n s o n got the job. He is answering (6?) not (7?) Who got the job? 3  This situation, of course, begs the question as to how Caruthers could fail to know that Miss Jones = Mrs. Robinson. This question will be treated below, pp. 137-142. 4  Substitution  126  The situation is not so easy with respect to the second type of question, those in which no (relevant) name occurs in the interrogative used. Suppose that the question asked was indeed (7?), and that A wishes to express a  ~& got the job As before, he can express just that proposition, and so provide a direct answer to the question, regardless of which name for Laura he uses, though again, if he wishes to communicate his answer, he had better use a name which his audience is familiar with. In this case however, he has been given no linguistic clue as to which name to use. The situation is further complicated by the fact that, where the set of propositions offered by the question is determined in part by the context, the questioner need not be aware of all the propositions in the set offered by the question he asks. Suppose the questioner again is Caruthers. Caruthers need not be aware of all the people who are in the running for the job in order to ask (7?). Were A to reply (7) Mrs. Robinson got the job. Caruthers in this case  would  probably take A to be answering his question, he would  take A to be telling him that some woman whom he did not know had got the job thus, perhaps, dashing his hopes that Laura had got it. Again, though A can provide a direct answer whichever name he used, the use of one particular name may be necessary in order to ensure the questioner understand his answer. The situation with respect to tentative answers would seem to be exactly similar. A will use the name already used in the interrogative or, if the question is of the second type, will try to use one with which the questioner is familiar. If A is not certain that Laura got the job but thinks it quite likely that it was she who got it, since B (one of  Substitution  127  Laura's family friends who knows her only by the name of 'Robinson') believes that to be the case, A will reply ((7') B believes that M i s s J o n e s got the job. not (7") B believes that M r s . R o b i n s o n got the job. for exactly the same reasons as those given above. A t this point a critic might respond Look, while I might agree with you that A would tend to substitute one name for another in circumstances of the sort described, this is not the issue. The issue is what gives him the right to do so. After all, a belief attribution has been made. You are simply forcing on 5 a belief which he would vehemently deny. (B believing that Mrs. Robinson got the job and not knowing that Miss Jones is Mrs. Robinson would hardly concur with (7').) Even leaving aside the vital question of how, under the direct reference view, B could fail to know that Mrs. Robinson is Miss Jones, the attribution in the light of S's denial would seem to be simply false.  Here we have the heart of the critic's position against the direct reference theorist. The point is not that we do in fact often substitute one co-referring name for another irrespective of the believer's concurrance, rather it is a demand to show what sanctions such usage. In order to deal with the critic's point it will be as well to treat separately 5  See Kaplan, 'Demonstratives', n. 71), on a related point involving the substitution of one definite description for another, where he remarks 5  The reporter has simply substituted his description for John's [the believer]. What justifies this shocking falsification of John's speech?  Substitution  128  those cases in which the belief attribution is made as a direct answer and those in which it constitutes only a tentative answer. Tentative A n s w e r s . One cannot deny, with respect to belief attributions made as tentative answers, that a belief attribution has been made. Insofar is the critic correct. But remember that it has been made for the sole purpose of modifying the claim to truth. B, or his beliefs, are not under discussion in such cases but some quite different matter - for example, who got the job. The belief attribution is principally performing the purely functional role of modifj'ing the truth claim. Nevertheless, to foresee the objection, the belief attribution cannot simply be treated as a purely functional operator modifying the claim to truth, if for no other reason than that the name of the person to whom the belief is ascribed contributes to the semantic content of what is expressed when the attribution is made. The attributions B believes that Fa. and C believes that Fa. (where B f C) say different things. While the primary function of what is expressed by 'B believes that' in such cases is to modify the truth claim of the embedded proposition, the embedded proposition is just that - embedded, in another proposition expressed by the complete sentence. How then, are we to handle the truth conditions of the complete proposition? Surely here we have to face the fact of B's denial of the proposition as so expressed? But Nothing! But we do it, and often recognize--or don't care-when it is being done.  Substitution  129  why should we? The focus of attention, I repeat, is not on B or his state of mind but on, say, who got the job. That the belief attribution performs the secondary  role of  attributing a belief to B does not force us into expressing the embedded proposition from B's point of view. A l l that matters is that he have the belief, not that he have it as characterized in a particular way. In such cases, because the focus of attention is not on 6  In fact, though it is not immediately germane to the particular topic in hand, Uhe substitution of co-referential names) I would argue a much stronger thesis; that the believer need not even believe the proposition attributed to him. Suppose that all that B believes is that the woman in the corner house, whoever she is (B has never met the lady and does not know her name), got the job being offered at the X Y Z firm. The woman in the corner house is, of course, Miss Jones. In such a situation, I would argue, A could still assert.(7'). Why? Suppose Caruthers were to ask (7?) of B. B would give the answer 6  (7"') The woman in the corner house got the job. (7"') does not provide a direct answer to (7?) Nevertheless, it provides a means for Caruthers himself to find the correct direct answer: it is an indirect answer. By finding out who it is that satisfies the description Caruthers would be in a position to say, of that person,'She got the job.' Caruthers, however, has not asked B, he asked A . A could, of course, reply with ( 7 " " ) B believes that the w o m a n i n the c o r n e r house got the job. thus supplying Caruthers with a tentative indirect answer. But A knows that the woman in the corner house is Miss Jones. This, plus his knowledge of B's belief, is what led him to select the proposition got the job as the most likely in the first place. Though B cannot give a direct answer, A can. A, though only tentatively, can give the answer B would give were he suitably informed. Moreover, for communicative purposes, it is better, so long as one will be understood, to give a direct answer than to give an indirect one. Since the belief attribution is employed solely to modify the truth claim of the embedded proposition, it is sufficient, I would claim, that B have a belief which would provide an indirect answer, an answer which could provide the questioner with just that direct answer. Further, although I cannot go into it in detail here, I would argue that A can substitute another definite description for that used by B. That such substitution is often an aid to communication can be seen from the fact that B and/or the questioner may not know the name of the person denoted by the description and, while the substituted description will enable the questioner to immediately provide himself with the direct answer, that used by B will not. In such cases, although A does not provide the same indirect answer as that which would be provided by B, it nevertheless enables the  Substitution  130  B or his beliefs, the belief attribution does not carry the implication that B would accept the proposition as so expressed. The position that, when using a belief attribution, we must always correctly characterize the belief from the believer's point of view would seem to be a remarkably insidious inheritance of the Fregean view. It is a point advocated by most direct reference theorists. Salmon, Perry and Kaplan, for example, in their work towards a 7  solution of the substitution problem by appeal to a third entity (the believer's cognitive perspective associated with the referential term) espouse it. However; to view the matter thus is to ignore the many uses for which we employ a belief attribution. Sometimes, certainly, we wish to talk about a person and the beliefs he has. However in many cases, attributing a belief is but a means of getting one's point across on some quite different matter. It enables us to answer a question, to convey some information, unrelated to the believer or his beliefs, in a tentative way. So used, how the believerwould characterize his belief is irrelevant to the truth of the attribution. So, with respect to belief attributions made as tentative answers, the problems of substitution simply do not arise since no claim is made that the believer would characterize the proposition in the way in which it occurs in the belief attribution.  questioner to reach exactly the same direct answer as would be reached were B to assert his belief in the presence of a knowledgeable audience. Clearly the matter is more complicated than can be presented here. Questions involving modalities, for example, would need special attention. I hope to develop the argument in more detail elsewhere. A notable exception is Howard Wettstein (see 'Has Semantics Rested on a Mistake?', esp. pp. 205-7) who argues that, dependent upon the purpose for which the attribution was made, one can substitute one singular term for another no matter of what type they are - proper name can be substituted for proper name, proper name for definite description, and vice versa, and definite description for definite description. With this I would wholeheartedly agree (see above note). Wettstein's reasons for so arguing would seem different to mine (he does not develop his views on this matter in any detail) though there is much in common with our views. 7  Substitution  131  Direct Answers. Let us begin by looking at substitution with respect to attributions whose topic question concerns something other than the believer's belief. A n example will be useful. Suppose the chairman of the department, looking for a replacement teacher at short notice, wonders whether James Armstrong is available. He asks the committee (8?) What is James Armstrong doing this summer? Shrugs all round. The chairman presses on: Your daughter knows him, doesn't she, Francis? Does she know what he's doing? Suppose now that James Armstrong, in his non-professorial moments, plays jazz with a local band under the name of 'Miles Abbo'. It is through his jazz playing that Maggie, Francis's daughter, has met the man and she knows him only by that name. When visiting the jazz club she has heard Miles say that he is going to be away all summer. However Maggie has also heard her father talk disparagingly of James Armstrong who, she knows, occasionally teaches for the department. If the chairman, using the name 'James Armstrong', had directly asked her if she knew what James Armstrong was doing in the summer, Maggie would reply that she didn't know. Let us suppose that Francis knows all about his daughter's confused state of mind. Faced with the chairman's question, which, being co-operative, Francis will take to be  (9?) What does Maggie believe that James Armstrong is doing? how should Francis answer? Both  Substitution  132  (9) Maggie believes that James Armstrong is going to be away all summer. and (9') Maggie believes that Miles Abbo is going to be away all summer. constitute direct answers to (9?). Further, both (9) and (9') are relevant to the topic 8  question (8?). However, if the chairman does not know that James Armstrong sometimes goes by the name of 'Miles Abbo' then (9'), while erotetically relevant, is not relevant for the chairman. He would not regard (9') as constituting an answer to his question. So if Francis were to answer with (9') he would not have succeeded in conveying the requested information to the chairman. In order to do so he must answer with (9). But should Francis answer by using (9)? What of the fact that Maggie would deny it? How can he attribute a belief to Maggie which she would deny? But why should we take Francis to be capturing, in the belief attribution, all that is necessary for Maggie's concurrance? Why should we take his assertion of (9) as asserting, showing or implying that the way the embedded proposition is expressed is the way in which Maggie would express that proposition? It is true that, unlike a tentative answer, the attribution is given as a direct answer to a question concerning what Maggie believes. Yet the situation shares many features with one in which a tentative answer is given. The main topic of conversation is James and what he is doing this summer. Attention is only focused on Maggie and her beliefs in passing, in order to settle that question. Maggie's beliefs, while being sought, are only temporarily of interest and of interest only insofar as they can shed light on the matter of where James Armstrong is  Being co-operative, Francis will not mention the fact that, if Maggie herself were asked the question by means of the interrogative used in (9?) she would reply that she had no opinion on the matter. 8  Substitution  133  spending the summer. It matters little whether Maggie would express her belief in that manner or not, all that matters is that she have a belief which could  be expressed in  that manner. So, where the topic of conversation is something other than the believer and his beliefs, an attribution made as a direct answer to a subsidiary question need not present the belief in the manner acceptable to the believer since in such circumstances there is again no implication carried by the attribution that the belief is being so presented. What now of cases in which the topic question does concern the believer and his beliefs? Our interest in another's belief for its own sake can stem from several reasons. Sometimes it is because there is something unusual or odd or interesting about the belief itself. We have seen an example of this in the case of Mary, whose view of the significance of people's shoes was thought to be weird in A's eyes and to lead to all sorts of crazy beliefs. In such cases the topic question presents in the set offered only propositions whose assertion would constitute a belief attribution. Again, we may attribute a belief to another because the belief is typical in some way of the kind of person they are. If A were asked to describe B to another, as well as giving such information as  He is quite tall, athletic. Plays rugby. He's a good lawyer. Conservative, rather old fashioned. he might include certain of S's beliefs which are typical of the type of person B is: He believes that Mulroney is the best thing to have happened to Canada in a long time. He believes that his daughter should have a two year engagement before getting married.  Substitution  134  Or again, with someone we know well, hearing of a recently acquired belief can inform us about how they are getting on: Bill believes that he's found just the house he was looking for. In such cases the topic question presents in the set offered, both propositions which, when asserted, attribute a belief and those which attribute some other property to B. Let us examine an example of the latter kind of case. Suppose that Francis is talking to another member of the department who asked him 'How is Maggie doing these days?' 'Oh. pretty good', Francis replies. 'She's been going arround a lot with James Armstrong recently. She believes that James might 'pop the question' any day now.' Here, although the focus is on Maggie, substitution has taken place. Is Francis wrong to do so? Should he not say She's been going arround with James Armstrong a lot recently. She believes that Miles Abbo might 'pop the question' any day now. Surely not. The questioner, ignorant that James is also called 'Miles', would take those two sentences to concern two quite different people. But why should Francis say She's been going arround with Miles Abbo a lot recently...  since the questioner has never heard of Miles Abbo and, anyway, the name does not occur in a belief context? Perhaps we need a purer type of example. One less concerned with the believer and more directly with her beliefs. One whose topic question only presents belief  Substitution  135  attributions in the set offered. Suppose Maggie is a firm believer in the predictive ability of tarot cards and Francis is telling his colleague about her predictions. Francis: She believes that the chairman will resign at the end of the year. Anne  : Oh my! And what does she beUeve about James Armstrong?  Francis: She believes that James will be very famous one day. (Of course, Maggie would deny this while she would agree to 'Miles Abbo will be very famous one day'.) Here, if anywhere, it would seem, with the focus firmly on what Maggie believes, will Francis be incorrect in attributing a belief to Maggie which she would den.y. But yet again, the interest of Anne and Francis is not on how Maggie believes the belief but is on the belief itself. The question concerns what Maggie believes about James Armstrong not how she believes it. Again Francis's assertion does not state, show or imply that the belief is expressed in a way in which Maggie would accept it. When the focus is on the belief itself, all that matters is that the believer would accept the belief in some way or other. Do we, in ordinary life,  ever  take an interest in the manner in which someone  believes a belief? Of course we are often interested in how someone could come to believe something - in the grounds that give rise to the belief  How can Maggie believe that that jerk is going to very famous one day. I thought she had never met him? or How can she think he's going to be famous? She believes he's hopeless as an academic.  136  Substitution  But to ask how someone believes a certain belief is a philosophical question (and a pretty specialized one at that) and certainly not the kind of question being answered in everyday life when a belief attribution is being made. More importantly, it is not the kind of question which can be answered directly by attributing a belief. In attributing a 9  belief when answering a question directly, as we saw, we inform our audience of the believer's belief because it is odd or of interest in itself or sheds light on the character of the believer, or because it provides potential information about some aspect of the world. How the believer believes the belief or how she would express it is irrelevant to that enterprise. Having now concluded the survey of all the types of questions to which a belief attribution is a direct answer, we have seen that nowhere is substitution  not  permissable. So the hard-line theorists who maintain that, despite the believer's protestations, if 'B believes that F a ' is true and if a = b then 'S believes that F6' is also true, are vindicated. They are vindicated not because they are not guilty of the charge of ignoring the believer's protestations, of ramming something down the throat of the believer irrespective of her denials. It is not that they ignore the believer's protestations and carry on willy nilly; it is that, in making a belief attribution, the believer's protestations are irrelevant. How the believer would represent the belief is not germane to the belief attribution. Because what is of interest is the belief itself, what matters is that the believer would concur with the embedded proposition when expressed in some manner or other.  9  It remains an open question whether it can be answered in language at all.  137  7. PROBLEMS WITH THE ACCOUNT OF SUBSTITUTION GIVEN.  In the previous chapter it was argued that the substitution of co-referential terms is possible at all times in belief contexts. Such a view is seen to generate problems in two distinct areas: that of the explanation of action, and that of belief itself. Let us begin by examining objections in the latter case.  Belief: On the account of substitution offered above, if Maggie believes that Miles Abbo is going to pop the question any day now then it is quite correct to attribute that belief to her by means of the sentence 'James Armstrong is going to pop the question any day now'. But what of poor Maggie protesting that she has no such belief? Alas, poor Maggie is confused. She believes that Miles may soon ask her to marry him but has no such belief about James; that is,  (i) Maggie believes that  Q  and (ii) Maggie believes that  (where  A" is going to pop the question. A is not going to pop the question  ^ = Miles / James) are both true. Indeed it is even true that Maggie believes  the following propositions: % is going to pop the question and A is not going to pop the question. She is not irrational though: she does not believe the proposition ^ is and is not going to pop the question since she does not realize that the object referred to by 'Miles Abbo' is identical to the object referred to by 'James Armstrong'. Aha, says the Fregean, perking up. How can this be? How can Maggie assent to the proposition  Problems with the Account of Substitution  Given  138  is going to pop the question. when it is expressed by means of the name 'Miles Abbo' but deny it when expressed using 'James Armstrong'? How can she fail to realize that the same proposition is involved. If an examination of the way we use belief attributions does not reveal it, surely here, in the fact of Maggie's varying response to the same proposition, we have the proof that there must be more to belief than the proposition believed. Note that the focus of enquiry has shifted dramatically. In concentrating on Maggie and her confused state we have moved from a discussion of belief attributions to that of the nature of belief itself. Now it seems perfectly permissable for the language theorist at this point to recognize her error in venturing into foreign territorj' and withdraw from such a discussion. All that is incumbent on the language theorist is a coherent, and hopefully true, account of the functioning of certain bits of language. She need not be drawn into the debate as to the true nature of belief itself. The question as to the relationship between belief attribution statements and belief itself can remain open. However I shall continue the investigation not only because the subject is of interest for its own sake but also because consideration of the problem can shed further light on certain theorists' positions regarding belief attributions. The critic's question is: How can Maggie fail to realize that the same proposition is expressed by means of 'James Armstrong is going to pop the question' as by 'Miles Abbo is going to pop the question' when she believes that proposition when it is expressed by the latter sentence. Behind the critic's question, as Wettstein has argued,  1  is the view that the believer, in order to have a belief about a person or object, must have a substantial cognitive fix on the object, that something in one's thought must correctly distinguish the referent from everything else in the universe. Though placing  1  In Wettstein, 'Cognitive Significance without Cognitive Content'.  Problems with the Account of Substitution  Given  139  differing demands on the believer, both Frege and Russell subscribe to such a view. As Wettstein puts it: Russell requires, and this seems closely connected with the fact that his motivation was epistemic (and his epistemology Cartesian), that the speaker or thinker stand in an extremely strong epistemic relation to the referent, namely, that of direct acquaintance. To be really thinking about something, the thing need be, as it were, smack up against one's mind. Otherwise, held Russell, the thinker would not really know which thing was in question, even if he possessed a concept that uniquely applied to it. Frege, motivated not so much (perhaps not at all) by epistemic considerations, but rather by what we might call his 'referential dualism', maintains more modestly that one need merely (merely?) possess a uniquely denoting concept. 2  Some direct reference theorists, while rejecting the Fregean view of reference, are loath to give up his view of belief. They hold that while we can refer directly to an object we do not believe directly of an object. Rather, in thinking of an object, of necessitj' conceptual features are involved. This leads to the, to me highly unintuitive, position 3  that we do not assert what we believe. Others, accepting that one can have a singular belief, that is, a belief whose connection with the world is not completely mediated by concepts but which has as one of its components the object itself, are at pains to determine how much one needs to know about the object before that can occur, how much one needs to know before, to use Wettstein's happy phrase, the object is smack up against one's mind. On such a view we can only use a name to refer directly when we have a singular belief about the object. This leads to the untenable position that there are many, many names which we simply cannot use. It cannot be that in such cases we use the name in a Fregean fashion as referring to whoever or whatever satisfies the  2  Ibid., p. 3.  Evans would seem to be tending towards this view, at least for the vast majority of speakers, (see The Varieties of Reference, Ch. 11, pp. 373-404). Simon Blackburn, in 'The Identity of Propositions', (in Meaning, Reference and Necessity, ed. S. Blackburn, Cambridge, 1975, pp. 182-205) holds a version of this view. 3  Problems  with the Account of Substitution  140  Given  concepts we do have of the object. As was pointed out above, in many cases what we believe about the individual is not sufficient to yield a unique description. For many, all that is believed of Van Dyck and Van Eyck is that they are famous painters, of Neils Bohr and Teller that they are physicists, of Cyndi Lauper and Whitney Huston that they are present day pop singers. 4  Yet, if one cannot use such names as 'Van Eyck' and 'Cyndi Lauper', how can one gain information about them? How can I find out, for example, if Van Eyck came from Amsterdam? There is, however, another alternative to the two positions outlined above: one which accepts that we do indeed assert what we believe, that we do have beliefs which are singular in form, beliefs in which the object itself is a constituent, yet one which rejects the Russellian view that to have such a belief the object must be 'smack up against one's mind'. To ask when one has such a singular belief, given the identification of what is believed with what is asserted, is to ask when one can refer directly by means of a certain name. Kripke and Donnellan answer this question by invoking a causal or historical chain which links present day users of 'Van Eyck' say, to the man himself. Whatever the merits of this particular proposal - I prefer, following Patton, to see it 5  more in terms of a communal tradition  to use the name to refer to the man, what is  important is that reference to the man is achieved by means of being a member of a linguistic community. Being part of a language-using community, joining the traditional  4  See above, pp. 19-20.  See T.E. Patton, 'Referential / Attributive: A n Appraisal of an Account', unpublished manuscript, 1983, pp. 13-14. 5  Problems with the Account of Substitution  141  Given  practice to use the name to refer to a particular individual, enables us to refer directly to that individual. As Wettstein argues Names are thus not to be thought of as externalizations of inner gazings, mediated or not, but as social instruments, as tags that allow us to make into subjects of discourse those things with which the tags are conventionally associated. And it is crucial to the utility of these linguistic devices that they function even in the face of a referent's epistemic remoteness. Far from there being an epistemic requirement of any sort traditionally supposed, names allow speakers to bridge great cognitive gulfs. The mere possession of a name for an item, that is, provides a crucial kind of contact with it. 6  To be sure, such a contact does not result in a belief in which the object is 'smack up against one's mind', to be the object of an unmediated intellectual gazing. But such is not necessary for singular belief. To believe  was born in Amsterdam. one need not have ^ before one's inner intellectual eye. If one takes seriously the view that belief is a cognitive relation to (rather than, say, an intellectual possession of) a proposition, then for there to be an object in that belief one will have to be in a cognitive relation to that object but that relation need not be epistemically by means of entering into a community  practice  fixed. It can be fixed  with respect to that individual's  name. What now of the critics' question as to how Maggie could fail to realize that A  (where ^  IS  gOl! going to pop the question  once again stands for James/Miles) is the self-same proposition when  expressed by means of 'Miles Abbo' as when expressed by means of 'James Armstrong'?  6  Wettstein, 'Cognitive Significance without Cognitive Content', pp. 23-4.  Problems with the Account of Substitution  142  Given  Perhaps now it can be seen how the question thrives on the view that in order to have such a belief Maggie must have  directly before her intellectual gaze and so could not  fail to recognize that the same object is involved. But as Wettstein argues for a similar case, given how little knowledge is required in order to be able to refer to James/Miles 7  and so have a singular belief involving that man, is it any wonder that Maggie fails to realize that the same person is involved, that .'Miles Abbo' and 'James Armstrong' corefer? Indeed, given that most people have but one name and that there is a general presumption to that effect, it would be exceedingly suprising, had she not been told, if, upon entering the differing practices of using 'James Armstrong' and 'Miles Abbo', she did realize that they were co-referential.  The Explanation of Action.  The  second area in which problems are seen to arise for the account of  substitution offered above wherein substitution of co-referential names is at all times possible is that of the explanation of action. Often we attribute a belief in order to explain someone's actions. Such attributions (B believes that a is F) do not, in themselves, constitute a direct answer to the-question asked ('Why  is (was) B ^'ing?')  but may form part of a direct answer to such a question . That direct reference and the intersubstitutivity of co-referential terms within the embedded sentence of the belief attribution generate a problem with respect to action explanations was perhaps first noticed by Perry. As outlined by Perry the problem 8  would seem to be intimately concerned with the nature of self-locating beliefs. Lynne  7  8  In 'Cognitive Significance without Cognitive Content', pp. 21-6. See 'The Problem of the Essential Indexical', Nous, 13 (1979), 3-21.  Problems with the Account of Substitution  143  Given  Rudder Baker has argued, however, that the problem is much more widespread than Perry's account would show and can be generated in cases where self-locating beliefs are not involved. I shall concentrate on such latter cases not only because the involvement 9  of self-locating beliefs may introduce extra elements not specifically germane to the question of whether ordinary singular belief attributions can explain behaviour, but also because I wish to focus more on the problem as it pertains to the use of proper names. The problem cases fall into two general categories: those involving conflation, where there are two (relevant) people (or objects) involved but the believer thinks there is but one; and those in which there is only one (relevant) person involved but the believer thinks there are two, which, for want of a better term, I shall call cases of duplication. Let us examine the difficulty raised by cases of the first type first.  Cases of Conflation. This type of case was first illustrated, I believe, by Richard Feldman. Feldman's argument is directed against what Roderick Chisholm calls  10  latidudinarian  theories of de re belief but which more accurately should be called latitudinarian theories of de re belief ascriptions.  11  On a latitudinariarian theory, a de re belief ascription of  the form 'B believes of a that it is F' is true if B assents to an assertion 'n is F' where 'n' is a proper name or a definite description. ; 12  9  10  In 'De Re Belief in Action', Philosophical In 'Actions and De Re Beliefs', Canadian  that is, it is not necessary that B possess  Review, 91 (1982), 363-387, esp. p. 381. Journal of Philosophy,  8 (1978), 577-82.  See Chisholm, 'Knowledge and Belief: De Dicto and De Re', Philosophical (1976), 1-20. 11  Studies, 29  Recall that we are assuming throughout that definite descriptions are not being used referentially. 12  Problems with the Account of Substitution  144  Given  a directly referring term for a in order to attribute to B a belief of a .  1 3  Feldman's  examples concern cases in which all that B possesses are two descriptions which he takes to denote the same object but which in fact denote two different objects. However, as Feldman fails to see and Baker points out,  14  Feldman's argument that such cases  present a problem for action explanations can be applied equally well to cases in which the believer stands in an unmediated relation to the object, to cases in which, when the believer is asked what he believes, what he expresses is a singular proposition. Let us examine such an example.  15  Joe, who is fairly new to the neighbourhood, has left his bike outside on the sidewalk for a moment while he goes indoors to get a snack. Upon his return he discovers that it is no longer there. 'My bike's gone', he wails. The girl next door informs him that Sam has just taken it and has ridden off towards his house. Although Joe has never met Sam and does not know what he looks like, he does know where he lives and so sets off in hot pursuit. Arriving at Sam's house Joe sees a boy whom he takes to be Sam and lays into him. He's right - it is Sam. Now, why did Joe hit Sam? Well, (1) Joe believes that Sam stole his bike.  See Ernest Sosa, 'Propositional Attitudes, De Dicto and De Re', Journal of Philosophy, 67 (1970), 883-96; and Mark Pastin, 'About De Re Belief, Philosophy and Phenomenological Research , 34 (1975), 569-75. 13  14  Baker, 'De Re Belief in Action', pp. 366-370.  The example is based on one given by Eric Stiffler in 'De Re Belief Ascriptions and Action Explanations', Canadian Journal of Philosophy, 13 (1983), 513-525. Stiffler uses his example in an argument which concerns the second type of problem case, see below, pp. 155-161. 15  Problems  with the Account of Substitution  145  Given  is true and it would seem most natural to explain Joe's behaviour in terms of that belief - 'He's hitting Sam because he believes that Sam stole his bike'. But wait. Joe also has 16  another belief, one which concerns Weasel McKenzie. Weasel is notorious in the neighbourhood for cheating, telling tales and sneaking arround swiping other kid's stuff. Joe has overheard a conversation which involved Weasel and Sam, and has concluded that they are the same person. Hence every belief Joe has about Sam he also has about Weasel; that is, not only is (1) true,  (1') Joe believes that Weasel McKenzie stole his bike. is also true. Joe, however, is mistaken: Weasel is a different, boy entirely. How then can (1) explain Joe's hitting Sam? Had Weasel happened to be near Sam's house, would not Joe have hit him, acting on the same set of beliefs?  17  How then can (1) be an  explanation of the act that actually occurred? *  *  *  *  *  *  *  *  *  *  *  *  *  *  *  *  In order to examine such cases more closely, it is necessary first to provide an account of 'why' questions and answers. I shall, broadly, follow the account given by Van Fraassen  18  which is based on Belnap and Steel's work discussed above. Firstly, to  Clearly there must be more to explanation than the proposition explicitly expressed. That Joe believes that Sam stole his bike does not, in itself, explain Joe's behaviour. More is needed. That Joe was mad at Sam, say, or wanted his bike back, and small boys who are mad at another or want something from another usually / often / always / it is reasonable to expect / understandably... beat up the other. Just what else is being implicitly appealed to (desires, intentions, covering laws, rules of thumb and so on) is the subject of much debate. Since the objections raised do not concern such questions directly such issues will not be dealt with here in any detail. 1 6  Of course, given the example, Joe might well have hit any boy of the right age who happened by. However, given that Joe did not hit the boy by mistake, he might, given his beliefs, just as easily have hit Weasel as Sam. 17  1 8  See Van Fraassen, The Scientific Image, pp. 141-61.  Problems with the Account of Substitution  Given  146  ask why a is F (or why a is doing ep) is to presuppose that a is F (or that a is doing ^ ). The proposition a is F is the topic of the 'why' question and that question asks for an explanation of this state of affairs. As before, a 'why' question presents a range of alternatives. The need for this is illustrated by Alan GarfinkePs legendary example. A bank robber was being pressed by a priest as to why he robbed banks. 'Because', the robber replies, 'that is where the money i s ' ,  19  a response which would hardly satisfy the priest. The interrogative 'Why  does B rob banks?' may present the set (B robs banks) (iB robs stores) (B robs offices) or (B robs banks) {B writes nasty letters to banks) (B pickets banks) or (B robs banks) (B works for a living) (B makes do on unemployment) or (B robs banks) (C robs banks) (D robs banks) and so on. The robber in the example is answering the question as to why he robs banks as opposed to houses, offices etc., whereas the priest was interested in why he robs banks as opposed to earning a decent living, making do on unemployment and so on. The differing range of alternatives which may be offered by identical interrogatives is often exploited in children's humour as in  See Garfinkel, Forms of Explanation, New Haven, 1981, p. 21. In a recent edition of a local paper (Vancouver Courier, 22nd April, 1988), it was reported that, when asked the same question, New York city bank robber Kenneth Stevens gave that very reply. 1 9  Problems with the Account of Substitution  147  Given  Why do birds fly south in the winter? Because it is too far to walk. Why do humming birds hum? Because they don't know the words. Note that, unlike the questions discussed above, the range of alternatives does not contain the set of possible direct answers but provides only part of any possible direct answer: it serves to delineate precisely what it is that is to be (is being) explained. Hence the identification of a question with the set of its direct answers has, in the case of 'why' questions, to be abandoned. A n explanation which may be true of one topic and range of alternatives need not be true with respect to another. Suppose, to use an example of Dretske's,  20  the  interrogative used is 'Why did Clyde marry Bertha?'. The context is that Clyde has been left a large sum of money on condition that he be a married man. The answer Clyde married Bertha in order to qualify for the inheritance, may then well be true if the question is (2?) Why did Clyde marry Bertha? as opposed to, say, just going out with her, but may be false if the question is (3?) Why did Clyde marry Bertha? (as opposed to Jane or Mary). Clyde could have married in order to qualify the inheritance, who he married being irrelevant to that end. On the other hand he could have married Bertha because he loved her. Finally, with respect to a 'why' question there is, as Van Fraassen puts it,  2 0  See Dretske, 'Contrastive Statements', pp. 417-8.  Problems with the Account of Substitution  Given  148  the respect-in-which a reason is requested, which determines what shall count as a possible explanatory factor, the relation of explanatory relevance. 21  Explanatory relevance is not at all the same notion as that of erotetic relevance discussed above,  22  and is an even more difficult notion to define. Clearly, a 'why'  question requests not only that the topic in the light of the range of alternatives be addressed, it requests that an explanation  be given. What will count as an explanation in  the context will be whether or not the reason given is, in the appropriate way, relevant. To answer the question as to why Joe is hitting Sam with (1") Joe is hitting Sam because Joe has two pencils in his top pocket. is clearly not to answer the question even though it is true that Joe hit Sam and that he has two pencils in his top pocket. It is insisted upon by some that that offered in explanation should provide sufficient conditions for that which is being explained. Others would hold that it provide necessary conditions. Still others that it supply both necessary and sufficient conditions. Numerous problems arise with such positions, not the least of which concerns the method of evaluating whether  the explanation given is indeed necessary  and/or  sufficient. If A explains B then, for A to constitute a sufficient condition for B, some sort of conditional of the form A  B must be true. For A to constitute a necessary condition  for B then some sort of conditional of the form B-^A  must be true. Such conditionals  can be tested in those possible worlds in which the consequent does not hold to see  2 1  Van Fraassen, The Scientific Image, p. 142.  To the question 'Why is Jimmy so hot?' there are many possible answers which are explanatorily relevant - one having to do with his running all the way home perhaps, another with the way in which his heart has increased its rate of pumping, another with how his blood vessels are behaving and so on. Which explanation is erotetically relevant will depend upon the context of the topic question. 2 2  Problems with the Account of Substitution  Given  149  whether the antecedent holds. But are they to be tested in all worlds in which the consequent does not hold? Is (1") to be tested, for example, in worlds where Joe does not have pencils in his pocket because they fell out when he was tying his shoelaces, or where he had to wear his raincoat instead (which had no top pocket)? Are we to be concerned, with respect to (1) with those worlds in which Joe is visiting a nearby town at the relevant moment (from which he may hear by telephone that Sam has stolen his bike - thus coming to have the important belief while at the same time failing to act), or where Joe is lying on his sickbed (from where he can see Sam take it), too ill to move? Let in too many worlds as being relevant for the test and you threaten the possibility of explaining a human action at all. Restrict the worlds for testing too severely and you may end up with cases such as (1") constituting adequate explanations. We shall return to this problem later.  23  In the meantime it is sufficient to note  that, though there are many problems thrown up by the demand placed on an explanation that it must explain, however these problems are ultimately to be resolved it is clear, as was said above, that some kind of relation of relevance must hold between the explanans and the explanandum. Hence we can categorize a 'why' question Q (e.g. 'Why does B rob banks?'), asked by means of an interrogative in a given context, as containing three elements: a topic, P  (B robs banks); a range of alternatives  X ( = {P^...P^...}) ({B robs banks, B robs stores, B robs offices}); and the relevance relation  R. That is:  Q = < P , X, R k  >  To say that a given proposition is relevant to a question is to say that is bears the relevance relation to the topic considered in the light of the range of alternatives; that is,  See below, pp. 153-4.  Problems with the Account of Substitution  150  Given  that is bears the relation R to <P , X>. A response to Q of the form 'P because A ' is a k  ft  direct answer to Q if and only if A is relevant to < P., X >. Hence (with illustration to k  follow) we reach the definition D is a direct answer to Q — < P., X, R > if and only if there is some k  proposition A such that A bears relation R to <P,, X > and D is the k  proposition which is true exactly if (P, ; and for all i p k, not P. ; and A) is true.  24  To take an example let Q be (2?) Why did Clyde marry Bertha? and D be (2) Clyde married Bertha because he wished inheritance.  to qualify  for  the  In order for (2) to be a direct answer to (2?) the following has to be the case. Firstly, the proposition Clyde wished to qualify for the inheritance (A) is relevant to the set containing the topic, Clyde married Bertha, and the range of alternatives (Clyde married Bertha) (Clyde continued to go out with Bertha) (Clyde stopped seeing Bertha) (  :  )  Secondly, Clyde married Bertha in order to qualify for the inheritance is the proposition which is true exactly if  See Van Fraassen, The Scientific Image, p. 144.  Problems with the Account of Substitution  151  Given  (i) Clyde married Bertha (P ) is true and no other proposition in the range of alternatives is true, and (ii) C?yde wished to qualify for the inheritance (A) is true. And so a 'why' question presupposes, amongst other things that  (a) its topic is true. (b) in its range of alternatives only its topic is true and (c) at least one of the propositions that bears its relevance relation to the topic and range of alternatives is also true. *  *  *  :!:  :|:  *  *  *  *  *  :|:  :|:  :|:  *  :|:  *  Let us now return to the problem case outlined above of a belief attribution being used in partial answer to a 'why' question. The objection to  (1) Joe believes that Sam stole his bike.  being used in answer to the question  (1?) Why did Joe hit Sam?  is that, since Joe also believes that Weasel stole his bike, (1) does not explain why Joe did what he did rather than hitting Weasel. We can now see that the critic's objection to (10) is that it fails to answer the question  (4?) Why did Joe hit Sam (as opposed to Weasel, Mike, Johnny, etc.)?  But (1) was not used in answer to (4?) but rather to  Problems  with the Account of Substitution.  152  Given  (1?) Why did Joe hit Sam (as opposed to going indoors, walking to the store, watching TV, etc.) 25  As we saw in the case of Clyde's marriage to Bertha from love or from desire to safeguard the inheritance, what is true with respect to one question asked by means of a given interrogative may be false with regard to another. Similarly, what counts as an explanation  with regards to one question asked by means of a given interrogative may  not count with respect to another. To explain why Stevens robs banks (as opposed to shops etc.)  it maj' indeed be  sufficient to answer that that is where the money is. However that answer would not constitute an explanation as to why Stevens robs banks (as opposed to earning an honest living), an explanation which may involve delving into the robber's past. Similarly, (1) certainly does not provide an explanation of why Joe kicked Sam (as opposed to Weasel). To explain that one would, for example, have to mention the fact that Sam and not Weasel happened to be 'in the wrong place at the wrong time'. However, the fact that a proposition does not constitute a direct answer to one question posed by means of a given interrogative does not prevent it from being a direct answer to another question so posed. Despite the fact that  (1) Joe hit Sam because he believes that Sam stole his bike.  2 5  Or perhaps Why did Joe hit Sam (as opposed to ignore him, him, etc.)?  play with him,  smile at  Problems with the Account of Substitution  153  Given  does not constitute a direct answer to (4?), it nevertheless provides a direct answer to (1?). It explains why Joe hit Sam. The fact that he might just as easily have hit another boy is irrelevant.  26  Hence with respect to cases which involve the believer thinking of two distinct individuals that they are the same person, the claim that theories of direct reference fail to provide an adequate account of action is answered once it is recognized that an explanation is an answer, an answer to a 'why' question, and once the nature of 'why' questions and answers is exposed. Postscript.  The critic's objection to cases in which two relevant people are involved but  the believer believes that there is but one is that attributing a belief to the believer which concerns the person who in fact was acted upon does not constitute an explanation since it fails the sufficiency condition. Joe, given that very belief, could have acted otherwise - he could have hit Weasel. Hence, argues the critic, it is not sufficient for the act to occur that he have that belief. As was pointed out above,  27  one of the notorious  difficulties of enforcing the sufficiency condition is in providing a means of determining the range of the conditional ('If B believes that P, then B  <^'s'). One must determine  which possibilities are to be counted as relevant.  Given that (1) does not constitute a direct answer to (4?), does not the critic still have a point that belief attributions fail to explain such actions given the direct reference view? I do not think that the critic would welcome such meagre crumbs. She was relying on our natural inclination to say, of an utterance like (1), that it does explain Joe's action and so argue that it must be given a different treatment than that given by the direct reference theorist. But when what is being considered is (1) as an answer to (4?), there is no inclination to say, given our knowledge of the facts of the case, that (1), however it is to be interpreted, does provide a satisfactory answer to (4?). 2 6  2 7  See above, pp. 148-9.  Problems with the Account of Substitution  Given  154  The above treatment of belief attributions as answers to 'why' questions given in response to the critic would seem to suggest a method of testing the truth of the relevant conditional. Assess the truth of the conditional B believes that P  B  offered in answer to a certain question q, in all those worlds in which, all else being equal, it is not the case that Bej's but in which  by q'& range of alternatives hold true. For  one of the other propositions offered  example, with respect to (1) the relevant  worlds would be those in which one of the propositions offered by (1?) hold true; those in which, at the relevant time, Joe goes indoors, those in which he watches TV, those in which he goes to the store, and so on. Look to see whether there is such a world in which Joe believes that Sam stole his bike. Since all else is being held equal - Joe, for example, does not lose his ability or inclination to fight, then there will be no such world. With respect to (4?) one sees, with respect to those worlds in which Joe does not hit Sam but hits Weasel, or those in which he hits Mike, and so on, whether there is one in which he believes that Sam took his bike. There is - one in which Weasel is standing where Sam actually was. Hence it would seem that we can delineate the set of relevant possible worlds in which the conditional is to be tested by means of the range of alternatives expressed by the question. To put it another way, when asked why B did this (as opposed to that) the answer, in order to constitute an explanation, should be sufficient to explain why B did this and not that, it need not be sufficient to explain why B did this as opposed to any other possible action.  *  28  *  *  *  *  *  *  *  However, as we shall see, the problems surrounding the sufficiency condition are not quite so easily dismissed. 2 8  Problems with the Account of Substitution  Given  155  Cases of Duplication Let us now examine tbe second type of case which appears to create problems for direct reference theories: that of duplication, in which there is only one (relevant) person but the believer mistakenly thinks there are two. The example is much as before: Joe has discovered that his bike has gone, is told 'Sam took it', and, after running him down, hits Sam. The difference is in this case is that Weasel does not enter the picture. Instead Joe has another belief, one which concerns Slugger. Joe has heard of Slugger, the neighbourhood bully. The stories were so dreadful that Joe vowed never to get mixed up with that boy. Indeed when he first saw that his bike was gone he was afraid that it might have been Slugger who had taken it, in which case he wouldn't have known what to do to get it back. So, in a way, Joe was quite relieved to hear that it was Sam who had done it. Hence, not only is (1) Joe believes that Sam stole his bike. again true, the following is also true: (5) Joe believes that Slugger did not steal his bike. Alas, unbeknownst to Joe, Slugger is none other than Sam. Since 'Slugger' and 'Sam' are co-referential, and if the substitution of co-referential names in belief attributions is permissable then  (6) Joe believes that Slugger stole his bike, is also true, as of course is (7) Joe believes that Sam did not steal his bike.  Problems with the Account of Substitution  156  Given  How then can (1) be an explanation of Joe's act in such circumstances? Intuitively one feels that (1) does explain his act; but how can it, given that (1) and (6) express the same proposition? If Joe had come to his belief,  A~ stole my bike, by other means, if,  for example, the little girl next door had used 'Slugger' instead of 'Sam', then Joe would not have acted as he did. Again the critic's objection is aimed at the failure to provide sufficient conditions for the act. How can (1) constitute an explanation since (1) could be true but the act not occur? One cannot rule out, without begging the question, those worlds in which the girl next door uses 'Slugger' instead of 'Sam'-and in such worlds, while Joe does indeed have the same belief, he does not hit Sam but, perhaps, starts yelling for his father, or goes indoors to think it out. Considerations such as these have prompted many critics to revert to an appeal to some element or other of the Fregean sense. We have seen that some who would deny that sense plays any role in the functioning of names in non-belief contexts, in the face of perceived general difficulties regarding substitution, turn to some aspect or other of it when treating belief contexts. Others, who would accept that in certain cases reference even in belief contexts is direct, would deny it in the case of action explanations. Peter Caruthers, for example, argues that we take two different interests in what he calls 'thought-content': the communicative interest and the explanatory. With respect to the first, the communicative interest, he rightly observes  Often reports of the thoughts of other people will play a role in our lives similar to reports of their overt assertions, providing us with reason to make additions to, or abstractions from, our own stock of beliefs. 29  Peter Caruthers, 'Russellian Thoughts', Mind, 96 (1987), 18-35, p. 27.  Problems with the Account of Substitution  Given  157  A t other times, he argues, we are interested in the thought only insofar as it will explain the action and as such the believer's point of view is paramount. How then to represent such thought contents in a sufficiently fine-grained fashion as to capture the differing reactions of the believer to the self-same proposition? How else but by a Fregean sense.  30  But an appeal to some aspect or other of a Fregean sense is no solution to the problem for all the same old reasons. If the sense is seen to be conventional (an aspect that the Kaplan / Perry proposal relies on) then not only is it extremely difficult to see just what would be the conventional sense of a proper name but, in cases where two names have the same sense (or two different uses of the same name with different pronunciations have the same sense ) then the possibility exists for the believer to think 31  that they have different senses. In which case a similar example could be constructed using (unknown) sameness of sense rather than (unknown) sameness of reference. Alternatively, if sense is seen to be user-determined then firstly the attributor, in many cases, would not know what sense is given to the name by the believer, and secondly it may not be possible to give an explanation which would cover the actions of more than one person: one could not, for example, explain several students' remaining in the student lounge after the time of their class by 'They believe that Professor Green is sick today' since, given the differing senses each of them attach to the name 'Professor Green' what each believed would differ. Let us look at the problem from a slightly different angle. Here are Stiffler's comments on (his equivalent to) (1) and (7)  Or its equivalent. Caruthers in fact depicts the contents in terms of possible worlds which are intensional. 3 0  3 1  Cf. Kripke, 'A Puzzle About Beliefs', pp. 266.  Problems  with the Account of Substitution  Given  158  If this is the only information available to us it is not obvious how we should expect Joe to behave towards Sam. In particular we do not know whether being in this state would prompt Joe to kick Sam. Only one of the ascriptions has any bearing on the action, we might assume, but in the absense of further information it is impossible to decide which is relevant or whether either i s . 32  Stiffler is in effect demanding that an explanation provide a means of predicting what Joe would do. This is, of course, the demand raised by the fact that in order to be an explanation the facts cited must be sufficient for the occurrence of the act. The facts cited must be such that their occurrence guarantees the occurrence of the act; given such circumstances the act must occur.  33  But this is much too strong. As H.L.A. Hart  and A . M . Honore argue It is vital ... to see that logically the demands of the situation in which we ask for the cause of what has happened, and that in which we are concerned to predict are very different. In the first case it is an inquest that we are conducting. The 'effect' has happened. 34  It may seem that all that has to be done in such cases is to modify the relationship seen to hold between the cause cited and the effect which has already  3 2  Stiffler, 'De Re Belief Ascriptions and Action Explanations', p. 517.  Note that the position does not depend upon an appeal to a Hempelian covering-law model of explanation. (See Carl Hempel and P. Oppenheim, 'Studies in the Logic of Explanation', reprinted in Readings in the Philosophy of Science, eds. H. Feigl and M . Brodbek, New York, 1953, pp. 327-8; and Carl Hempel, 'The Function of General Laws in History', Journal of Philosophy, 39 [1942], 35-48.) William Dray, for example, vigorously denies that an appeal to a covering law is always made in explanation. He argues that a satisfactory explanation is one cast in terms of the agent's view of the circumstances he is in, plus his motives and goals and so on. Nevertheless, although no universal or statistical laws are invoked, when the agent's beliefs and goals are so laid out, one should be able to predict the action. (See Dray, Laws and Explanation in History, Oxford, 1957; and 'The Historical Explanation of Actions Reconsidered', in Philosophy and History, ed. Sidney Hook, New York, 1963, pp. 105-135.) 3 3  3 4  Causation and the Law, Oxford, 1959, p. 43.  Problems with the Account of Substitution  Given  159  happened from that of holding inevitably to one which holds in the majority of cases. However, as Hart and Honore remark: 'The truth unfortunately is not thus simple'.  35  What is necessary is not that the occurrence being explained had to happen as it did given the cause, nor that, in the majority of cases, it would happen as at did, but that, as Strawson argues, what it is necessary to show is 'that it is entirely natural that it should have happened as it did'.  36  In the case of explanations of effects which have  already occurred, what is being asked for is an explanation of a particular puzzling or unusual occurrence, or divergence from the standard state or performance of something with whose ordinary states or modes of functioning we are familiar; and when we look for the cause of this we are looking for something, usually earlier in time, which is abnormal or an interference in the sense that it is not present when things are as usual. As Hart and Honore continue In effect, in the typical case with which the law is concerned, when we ask for the cause, we are asking that some abnormal lapse from routine (some accident, injury, or loss) be rendered intelligible by being exhibited as an instance of certain other normalities, namely, those general connexions which characterize experience and are formulated in broad and general terms. 37  Hart and Honore are specifically talking about explanations in law but the same holds, with modifications as to what is considered out of the ordinary, for explanations of actions in everyday life. The fact that it would have been just as natural, just as reasonable, for Joe not to hit Sam, given the belief attributed in (1) (and (6)), is neither  3 5  Causation  and the Law, p. 41.  In his review of Dray's Laws and Explanation p. 267. 3 6  3 7  Causation  and the Law, p. 43.  in History, Mind, 68 (1959), 265-68,  Problems with the Account of Substitution  Given  160  here nor there. Joe did do it and the explanation given by means of (1) shows that it is natural and reasonable that he should have done so. In situations such as that described by Stiffler that something is going wrong is not due to the way proper names are held to function in belief contexts on the direct reference view but rather is due to the excessive demands which are being placed upon the nature of the explanation. Once it is seen that explanations in such situations need not be such as to be capable of predicting that the action take place but merely that they show that is it natural that what occured did occur, then no problem arises from the direct reference view of the functioning of proper names in belief contexts. Perhaps this is too quick. What of Stiffler's point that, given Joe's confused state, we would not know which belief to chose as the one relevant to the explanation of his action, the one attributed in (1) Joe believes that Sam stole his bike. or that in (7) Joe believes that Sam did not steal his bike. Explanations are given within a shared body of information. Explanations only function as explanations if certain facts about the situation, certain general laws, rules of thumb, and so on are presumed. For example  Why is the vase broken?  is only explained by Marcie dropped it.  Problems with the Account of Substitution  Given  161  if it is assumed that things dropped land somewhere - usually on the ground; that the ground usually is hard; that vases (or this particular vase) are (is) brittle, not, say, made of plastic; and that brittle things landing on the ground after being dropped (or this vase if dropped) often (would) break. Given the general background appealed to in the case of Joe's hitting Sam, how could (7) ever explain the act? Only in a world where people were set upon for not doing something nasty does (7) make any sense as an explanation of Joe's act. Since our society is not such a world but one in which small boys (or Joe in particular) tend (tends) to lash out when mad, having one's bike stolen is enough to make anyone mad, and so on, there is no difficulty in deciding which of Joe's two beliefs he was acting from. Hence while the treatment of proper names proffered by the direct reference theorist may be seen to pose problems in the area of the explanation of actions in cases of duplication, it does so only if a certain model of explanation is embraced. However if the sufficiency condition is dropped, as I believe it should be for the explanation of actions, if it not be held that the fact cited in explanation be such as to enable one to predict the action, then no problem is posed by the fact that proper names may be substituted within belief contexts. *  *  :|:  *  *  *  *  *  :|:  *  *  *  *  :|:  *  *  Conclusion.  We have now completed our investigation (with respect to substitution) into two types of troublesome cases in the area of action explanation: those of conflation, where there are two (relevant) objects involved but the believer thinks there is but one, and those of duplication, where there is only one (relevant) object but the believer thinks that there are two. In cases of conflation we have seen that, once the logic of 'why' questions  Problems with the Account of Substitution  Given  162  is understood and once one knows which 'why' question is being answered by the belief attribution, no problem is generated by the fact that substitution of co-referential names is possible within the belief attribution. In cases involving duplication, we have seen that once the demand that the explanation be such as to be capable of predicting the effect is dropped in favour of the demand that, in order to count as an explanation, it has to be shown that what has happened is an entirely natural outcome of the situation cited, then the fact that one can substitute one co-referential name for another within the belief context generates no special problems. To review the treatment of substitution given above more generally, cases of substitution were first examined for belief attributions asserted as tentative answers. There it was noted that, since the focus of attention was not on the believer or his beliefs but on some quite different matter, the use of the belief attribution does not carry the implication that B would accept the (embedded) proposition as so expressed. From this one might be tempted to conclude that, in cases where attention is on the believer or his beliefs, the situation would be otherwise; that in expressing the proposition believed one would somehow have to present it in a manner acceptable to the believer. However when cases of belief attributions asserted as direct answers were examined it was found that, although the focus of belief is on the believer and his beliefs, how the believer would express the belief is a matter of indifference. Even in such cases all that matters is that the believer have the belief ascribed. Hence the conclusion was reached that in both cases substitution of one co-referential name for another is possible, The main objections to such a position are concerned with the explanation of action, and with the problem of how a believer could deny he had a belief when it was presented using a different (though co-referential) name. The objections with respect to action explanations have, I believe, been answered. With respect to the question of how the believer could deny having a certain belief when it is not presented 'in the right  Problems with the Account of Substitution  Given  163  way', it was noted that this is, in fact, a demand for an explanation not in the philosophy of language but in the philosophy of mind. Nevertheless, a brief outline of a view on the nature of belief was given which, while explaining how someone could deny having a certain belief when it was not presented in a certain way, is consistent with the direct reference view. As these constitute the major areas of objection to the direct reference view of substitution I shall now end my investigation of such cases and move on to that other problematic area - that of cases of belief attributions involving the use of a vacuous name within the embedded sentence.  164  8. VACUOUS NAMES.  In discussing vacuous names with respect to belief attributions, for ease of handling I shall, in the main, assume (i) that the questioner does not realize that the name is vacuous, and (ii) that the attributor does. However my remarks on the standard case are generalizable to cover the alternative possibilities. For example, the treatment of the questioner's utterance in the light of his lack of knowledge of the name being vacuous can be generalized to cases in which the attributor was the one (or was also) in such a position. Tentative A n s w e r s . Let us begin by looking at cases in which the attributor, A , is responding to an interrogative which contains a vacuous name. Suppose the following interrogatives are used, at different times, by M a : (1?) Where is Vulcan? (2?) Does Rumpole defend at the Old Bailey? (3?) Who is bringing Frank? and 'Vulcan', 'Rumpole', and 'Frank' are vacuous. As was seen when the logic of questions and answers was examined, the use of an interrogative carries certain presuppositions. In the case of (1?) there is the presupposition, amongst others, that Vulcan exists, in (2?) that Rumpole does, and in (3?) that Frank does. In each case the presupposition is false and, it will be remembered,  165  Vacuous Names  when the use of an interrogative has a false existential presupposition, no question has been asked. Therefore in each case, though she may think she has, Ma has not asked a question. There is nothing of which to ask where it is, no-one of whom to ask who is bringing him, or whether he defends at the Old Bailey. If no question has been asked, no answer can be given. Knowing that the name is vacuous, A, if he is being open and sincere, will not attempt to provide an answer: he is fully aware that no answer can be given. This is not to deny that A could respond to Ma's utterance. He could, for example, give the corrective reponse (4) Vulcan does not exist.  1  thus answering the prior question (4?) Does Vulcan exist? But he will not, if he is being open and sincere, give a response such as (1) Vulcan is over there. or  (1') Betty believes that Vulcan is over there.  2  (1') no less than (1) commits A to the existence of Vulcan. In (1') no less than (1) no proposition has been expressed: there is nothing of which one can say, even tentatively, that it is over there, and this A well knows. What of cases where the vacuous name is not contained in the interrogative used but the questioner believes that the set of propositions offered by the question contains Such assertions in themselves, of course, present a problem for the direct reference theorist. Not being directly concerned with belief attributions, it is not a problem which I shall address here, though some of my remarks on vacuous names may be germane. 1  That is, given that he has not changed the subject. He could, of course, respond with (1') wishing to talk about what Betty believes. 2  166  Vacuous Names  one which is correctly expressed by means of a sentence which contains the vacuous name; that is, she thinks that among the individuals determined by the context is someone whose name is 'a\ but 'a' is vacuous? For example, suppose M a asks (5?) Who is going to carve? wondering who, of those who will be at dinner, is going to do it. Since she thinks that Frank will be present at dinner, she thinks that it is possible that he might carve, that is, she sees her use of the interrogative to present the following set: (Ma will carve) (Frank will carve) (Jane will carve ) (Betty will carve) 'Frank', however, is vacuous: Jane has invented a boyfriend called 'Frank' in selfdefense against her very social sister. There might even be someone coming to dinner 3  whom M a knew nothing about (Michael). Despite the fact that M a has not got the set of alternatives offered right, she nevertheless has  asked a question, a question which  presents the set (Ma will carve) (Jane will carve) (Betty will carve) (Michael will carve)  Note that, in order to set up the problem, we had to assume that there was some kind of solution to it: we had to assume that it made sense to say of M a that she thinks that Frank will be present at dinner, that she thinks her use of the interrogative present, among the set, the proposition Frank will carve. Those who take a hard line on vacuous names occurring in belief contexts and argue that in such cases the believer has no belief, end of story, often fail to realize that they thus deny themselves the very means to depict the problem. 3  Vacuous  Names  167  In such cases, therefore, A can provide an answer. However, if he is being open and sincere, even if he knows what Ma would consider as a possible answer, he will not consider answering  or  (5) Frank will carve. (5') Betty believes that Frank will carve.  Knowing that 'Frank' is vacuous, he will not take the question to offer (5) as a possible answer. So if he is being open and sincere, where A knows that the name is vacuous, no belief attribution containing that name will be made as a tentative answer. Again, as in cases where the vacuous name occurs in the interrogative, were A to reply using the vacuous name, in (5') as in (5), no proposition would be expressed and no answer given, as A well knows. Because, in tentative answers, the belief attribution is made solely in order to modify the assertive force attached to the embedded proposition, attributions made to this end are on a par with direct answers with respect to cases which involve vacuous names. Just as, where 'a' is vacuous and A is being open and sincere, A would not assert 'Fa' as a direct answer, so he would not assert 'B believes that Fa' in an attempt to provide a tentative answer. Because, in the case of tentative answers, the belief 4  attribution is secondary to the original intention to assert (qualifiedly) what is expressed  Similarly, were A to be unaware that 'a' is vacuous, his assertion of 'B believes that Fa' would not, just as the simple assertion of 'Fa' would not, result in a proposition being expressed. This failure of the attribution to express a proposition in such circumstances need not be of concern since, in the case of tentative answers, the subject under discussion is not what B does or does not believe but some quite different matter, and since the attribution is made solely in order to modify the assertive force of what is taken to be one of the propositions offered in the set of alternatives. It is not primarily that A has failed to attribute a belief to B but rather has failed to express a proposition which could be asserted either directly or tentatively. 4  Vacuous Names  168  by the embedded sentence, no problems particular to belief attributions are  generated by vacuous names in such cases. *  *  *  *  *  *  *  *  *  *  *  *  *  *  *  Devious Usage In the above discussion I have assumed that A, when faced with Ma's interrogative, is being open and sincere. When this condition is dropped (assuming once again that A knows that 'a' is vacuous) the situation changes dramatically. In the case where A wishes to be devious, or deceitful, or simply playful, he might well respond to (1?), for example, with (1) or (1'). In this case A would be going along with, indeed encouraging, Ma's mistaken belief about Vulcan. But if no question is asked in (1?), (1) cannot be an answer, nor (1') a tentative answer. How then, knowing that 'Vulcan' is vacuous, can A, even if he is being devious, assert (1')? It is important to note that the problems posed by such assertions are not primarily concerned with belief attributions: the problems principally concern the devious reponse of 'Fa' and only derivatively apply to 'B believes that Fa'. However, while such usage does not present a problem to the direct reference theory for such belief attributions per se, since a discussion of such usage may be of help when we consider certain cases of belief attributions in which the topic of conversation is the belief of the believer, I shall, to some extent, address the problems concerned. I shall not offer a complete account of such usage, such an account demands separate treatment, but rather will present the outline of a view. Firstly, note that in such cases A is not just lying, saying that A is F when in fact it is G, or saying that a is F when it is 6 that is F. He is pretending, pretending that a exists. He is engaged in furthering the fiction which M a believes to be reality. He  Vacuous  169  Names  can do so because, if A knows that a name is vacuous he also knows something else. One cannot tell that a name is vacuous simply by hearing it used. On the contrary, if A did not know, prior to Ma's utterance, that ' a ' is vacuous, he would take her interrogative to be about someone whom he had never heard of. But if he knows that it is vacuous then he knows something about how the name got introduced. Vacuous names fall into two categories, those which are introduced to name some person or object (falsely) believed to exist, and, by far the most common case, those which are introduced to name some imaginary or fictional character or object. Examples of the first kind occur when the referent of the name is fixed by means of a definite description: 'Let us call whatever is the F, " a " ' .  5  A classic case is that of  Another way in which the referent of a name of an actual existent can be fixed is by means of a demonstrative dubbing - 'Let us call that, " a " ' (see Kaplan, 'Bob and Carol and Ted and Alice', pp. 499-500). Are all cases of vacuous names of supposedly existent objects necessarily the result of having their referent fixed by definite description? No. As Kaplan argues with respect to demonstrative dubbings: 5  Only on a view such as Russell's is it at all reasonable to make it a prerequisite for a dubbing that the dubbor know, or stand in some other special epistemological relationship to, the dubbee. Though most pointings are teleological (the finger is aimed at a preconceived individual) blind demonstrations (as in spin-the-bottle) are also possible and provide an equally satisfactory basis for dubbing. ibid., p. 500. If demonstrative dubbings can occur when the object pointed at is unknown, then names could be introduced by such a means which turned out to be vacuous. However, while it is possible that a name be introduced by means of a blind demonstration, as a matter of fact, I would argue, we do not do so. Names are introduced for a purpose. A s Strawson argues (in Subject and Predicate in Logic and Grammar, London, 1974, pp. 42-6), a name is introduced into the language essentially as a means of keeping track of a particular individual over time in a way that is independent of individually known or temporally existing facts about the individual. Which individuals are named depends upon the interests of the community; in particular the individuals named are, amongst other things, those for whom 'there is a frequent need or occasion to make identifying reference' and for whom 'there is an interest in the continuing identity of the particular from occasion to occasion of reference' (Strawson, p. 42). See also F^llesdal, 'Reference and Sense', p. 233.) But what possible interest could the community have in keeping track of some totally arbitrary object such as that which the bottle will point to when it finishes spinning?  Vacuous  Names  170  'Vulcan' which was the name given to what was thought to be an as yet undiscovered planet which was causing certain perturbations in the orbit of Mercury. The scientists, however, were wrong in their theorizing, the perturbations were not caused by an unknown planet. The description did not denote and no referent for 'Vulcan' was Fixed. Now, as direct reference theorists argue, the description which is used to Fix the referent of a name does not contribute to the semantics of the assertion made when the name is used. Such descriptions are, rather, presemantic and establish the name - object 6  semantic relationship. Further, ordinarily it is not necessary that one know how the referent of the name came to be fixed: one need not know whether the name was introduced by means of a description or, as is by far the most common case, if it were introduced by a demonstrative dubbing. Far from being necessaiy, we normally have no idea at all how a name was First introduced. However when we know that a name which has been Fixed by description is vacuous, we do know that a description was used to Fix the referent, and know, in a rough approximate way, what that description was, and know that the description does not describe anything. For example, we know that Blind demonstrative dubbings, while possible as a means of Fixing the referent of a name, are never used and hence no vacuous names could result from that source. Could a vacuous name occur as a result of a non-blind demonstrative dubbing? What of cases of tricks of light, of hallucination? Cases such as the latter are particularly interesting. I suppose someone, hallucinating, could go through a dubbing ceremony with respect to what he saw in front of him, or could hallucinate that the person introduced herself, giving her name. Would that constitute a naming ceremony? Perhaps. However, with the shade of Wittgenstein breathing over our shoulder, note that in principle there could be no check that the name would ever again be used correctly, should further hallucinations occur. (See Wittgenstein, Philosophical Investigations, Oxford, 1978, # 258.) Is this sufficient to rule it out as constituting a naming ceremony? Given Strawson's social criteria for the introduction of names, it would seem so. The situation, however is not quite so straightforward. Group hallucinations are possible. Continual sightings of what is said to be the same ghost are reported, and so on. Clearly the matter demands a more detailed examination than can be given here. See, for example, Kripke, Naming and Necessity, pp. 53-60; Kaplan, 'Demonstratives', pp. 558-563, esp. pp. 559-60; and Wettstein, 'Turning the Tables on Frege, or How Is It that "Hesperus is Hesperus" is Trivial?', forthcoming, esp. pp. 8-12 of the manuscript. 6  Vacuous  Names  171  'Vulcan' does not name anything because we know that it was the name given to what was thought to be a planet, a planet thought to have certain properties (normally we could not specify which ones exactly), and we know that there is no such planet. It is because A  knows that 'Vulcan' was supposed to name a planet with certain properties  that he can pretend that it does.  So doing, he is treating the name as naming some  fictional planet, treating it in effect as a fictional name. The situation with respect to the knowledge necessary in order to know that a name introduced as a fictional name is vacuous is very different. There, one need not know anything about how the name was first introduced, which novel the fictional character appeared in, for example, or who wrote it, or anything about what the fictional character is like. Were you not acquainted with the marvellous fictional character Rumpole, the hero of many wonderful stories, you would, upon reading this (provided you believe me), know all you need in order to know, when it occurs in a sincerely meant assertion or interrogative, that the name is vacuous. Fictional names are not just names that turn out to be vacuous, fixed in ordinary language by some nondenoting description. They are essentially 7  names of characters from the realm of  Consider the possible candidates for such a description for the name of the fictional character Jake Whitfield from the novel Fullsome Creek by Hank Jones. The description cannot be 'the hero of Fullsome Creek". Such a description is not vacuous - there is a hero of the novel, not a real man certainly but a real fictional character. Nor, as Kripke argues (in an unpublished [and untitled] transcript of a seminar given on fictional entites, University of California at Berkeley, 1972, pp. 8-10, henceforth known as "Fictional Entities"), can we appeal to the descriptions of the man which are generated from the novel, 'The first man to battle the railroad in Cheyenne and who lost his left leg in the process', for example. For suppose that there was a real man, let us call him 'Fred', who in fact was the first man to battle the railway, etc. If 'Jake Whitfield' was introduced by means of that description, the name would name Fred. But, given that the flame was introduced as the name of a fictional character, given that the author had no intention of talking about Fred, the name does not name Fred. 7  Vacuous  fiction, of imaginary 8  Names  172  characters. A l l you need to know in order to know that a fictional  name is vacuous is that it is a fictional name. However, while it is not necessary to know anything about the imaginary character named by a fictional name in order to know that the name is vacuous, it may be helpful if one is to continue the fiction successfully. It will help, for example, if A not only knows that 'Frank' is the name of a character conjured up by Mary but also that it is the name given to her imaginary boyfriend, or in the case of 'Rumpole'. it will help to have read the stories of John Mortimer and know that Rumpole is a crusty old London barrister. Sometimes, especially in cases where the question is not of a 'yes' or 'no' tj'pe, A ma}' feel that the deception will be more successful if he only tentatively aserts his response. Faced with (6?) Where does Frank work? A , in order not to jeopardize the deception by asserting something outright which M a would doubt A would be in a position to know, may well respond with (6) Betty believes that Frank works in Seattle. In such as case A is not just confirming the fiction as created by Jane, he is embellishing  it, just as he would be going beyond the story of Rumpole as given by  Mortimer if he were to respond to (7?) Does Rumpole ever eat at Claridges? with  This realm, as Kripke argues, cannot be treated as merely another possible world. Fictional names do not refer to some possible though non-actual individual (see Kripke, "Fictional Entities", pp. 10-13; and Naming and Necessity, pp. 156-8). While names introduced into ordinary language may become fictional names (as with A's use of 'Vulcan' above), the reverse is not possible: 'Jake Whitfield' could never name Fred. 8  Vacuous Names  173  (7) Sometimes. A is not concerned with accurals!*- reporting, as if they were real, the fictional facts. He is continuing  the fiction, enlarging where necessary. He has, as it were, taken over the  story (insofar as he knows it), made it his own, and is now engaged in making it up as he goes along. He is playing a game of make-believe. This point has been made and developed upon by both Kendall L. Walton and Gareth Evans. Let us, following Evans introduce a pretense operator and write 'It is 9  make-believedly the case that P' as '*p*'. So, to take an example both theorists use, when children play a mud pie game, they say things like *That is my pie.  :i:  when all that is there is, in fact, a dollop of mud, and *The pie is in the oven.* when the dollop of mud is in a wooden crate. Now, as Evans argues, the actions of the participants in the mud pie game form a crucial element of the game itself: one child can :;:  eat a pie* while certainly not eating mud; another can *steal a pie . Such make-believe ::  actions can be performed only if the actor has the appropriate beliefs and intentions. In order for Olivia to *steal a pie* she must *know it wasn't hers* and *mean to take it*. In order to incorporate such make-believe actions and propositional attitudes Evans proposes two principles:  See K . L . Walton, 'Pictures and Make-believe', Philosophical Review, 82 (1973), 283319; and 'Fearing Fictions', Journal of Philosophy, 75 (1978), 5-27; and Evans, (1982), Ch. 10, 'Existential Statements', esp. pp. 353-368. 9  174  Vacuous Names  (i) (x)(If * believes that *P* then *x believes that P*) (ii) (*)(If x intends that *P* then *x intends that P*) Once such propositional attitudes are introduced within the game then one can *refer to P*. If one takes the Gricean view of speaker's reference, for example, A can utter a sentence * intending to get his audience to realize that his utterance is true if and only if P*. This will be because he really intends that *his audience realize that his utterance is true if and only if P*. And this state of affairs which he aims to' produce is the audience really realizing (believing) that his utterance is true if and only if P*, i.e. realizing that *his utterance is true* if and only if *p=. 10 :|:  a  as Evans argues when, in the course of the game, I make pretended assertions, I am not to be taken as making real assertions about the game. My utterances are not up for assessment as really true or false (not even as really true of false in virtue of certain facts about the game); they are only up for assessment within the game. 11  •  Now,  I  although the mud pie game as described is existentially conservative (for  each mud pie there is a blob of mud), it can easily be expanded to encompass * referring to something* when there is nothing there to be referred to. Perhaps the children needed *a sink* and no suitable object was available,. The children can incorporate *a sink* into the game by stipulation: *And over there there's a sink.*  10  Evans,  The Varieties  of Reference,  p. 357.  11  Evans,  The Varieties  of Reference,  p. 358.  Vacuous Names  175  said while pointing to an area under the tree. Or again, they may decide that another character is necessary. Jenny might say: *I'll have a sister called Tanya who lives in Toronto and she'll come to., visit and have tea. * 1 2  The referent of 'Tanya' is thus fixed by description but is so fixed within the pretense operator. The name does not refer to what is described but to *what is described*. The main problem of the above account is to provide an account of make-believe truth. Is it *true* that *Tanya is Jenny's sister*?; that *Jenny has two legs*?, that "Toronto* is the same distance from '"Vancouver* as Toronto is from Vancouver? In Evans' account, make-believe truths are generated from three different kinds of principles: (i) basic principles which stipulate outright a (possibly infinite) set of make-believe truths. These are given in creating the make-believe world. An example of a make-believe truth so stipulated is *Tanya is Jenny's sister*.  (ii)  a general incorporation  principle  by means of which any truth not  ruled out by the initial pretense is incorporated into the make-believe world.  If one listens to children playing such make-believe games, particularly those who are good at it, one is immediately struck with how carefully they construct the make-believe world. By far the greatest amount of time is spent in setting up the world, deciding, for example, what kinds of characters they are and who is going to do what and why. So much time is taken up with this task that 'playing' the game seems almost unimportant. On the view advocated here, all of the children's activities would constitute playing the game. 1 2  Vacuous Names  176  This would permit such truths as *the sky being blue* without the creators having to specify that it was (assuming, of course, that they did not specify some other truth that ruled out the sky being blue); and (iii) a recursive  principle  which permits the derivation of make-believe  truths from other make-believe truths. From *Tanya is Jenny's sister* and, let us suppose, the stipulated make-believe truth of *Johnny is Jenny's husband* the recursive principle enables us to reach the makebelieve truth *Johnny is Tanya's brother-in-law*. One problem with Evans account lies in the treatment of the encorporation principle and the recursive principle. These are treated in terms of counterfactuals. The encorporation principle is formulated as follows: If B is true and there is no set A ... A 1  the counterfactual 'If A... A I  n  n  of make-believe truths such that  were true, B would not be true' is true,  then B is make-believedly true. and the recursive princiuple as If A ... A I  A,... A 1  n n  is a set of make-believe truths, and the counterfactual 'If  were true, then B would be true' is true, and there is no set of  make-believe truths A' ... A' I  n  such that the counterfactual 'If A' ... A' I  n  were true, then B would not be true' is true, then B is make-believedly true.  1 3  13  See Evans, The Varieties of Reference, p. 355.  Vacuous  111  Names  Note that these counterfactuals concern the actual world. The problem with Evans' position is that, as he himself acknowledges, a make-believe world is not a possible world.  That being so, if one wishes to retain a possible world  treatment  of  counterfactuals, one cannot appeal to counterfactuals which involve what would be the case were some make-believe truth actually true, since make-believe worlds cannot possibly be true.  14  There is yet another problem which Evans seems only dimly aware of  15  and that  is that while the counterfactuals are given in terms of what would or would not be actually true, make-believe worlds are created.  They are created not from within what  is actually true but what is believed to be true by its creators. Suppose it was decided that T a n y a lives in Toronto and is late for tea because of the traffic*. Suppose that the ::  children are playing in Vancouver and think that Toronto is not too far away but is one of the outlying suburbs of their city. In that case, even though the children have not specified its location, in their make-believe world, * Toronto* is situated not where Toronto actually  is but where the children believe it is. So, even if a possible world  treatment of counterfactuals were abandoned, what would matter is not what would or would not be true, were a make-believe world actually true, but what the creator(s) of the make-believe world believe would be true. We cannot simply accomodate this point, however, by changing the counterfactuals from concerning what is actually the case to what is (in the actual world) believed to be the case since there may well be make-believe truths which are expressed using proper names which, in the real world, are vacuous and so could not be used to specify something actually believed. For example, it might be a make-believe truth that Frank is Betty's new boyfriend but this truth cannot be  Evans himself seems willing to give up a possible world treatment of counterfactuals but does not indicate what treatment he would give them. 14  15  See Evans, The Varieties of Reference, pp. 354-6, esp. n. 23, p. 354.  Vacuous Names  178  generated from a counterfactual which concerns B's believing that Frank is Betty's new boyfriend since, 'Frank' being vacuous, B can have no such belief. How then, if the possible world treatment of counterfactuals is retained, are we to delineate make-believe truth? Perhaps, if makebelieve truth cannot be specified by means of bringing * propositions* to the actual world, it can be specified going the other way - by bringing actual truths into the make-believe world? What is important to notice about fictional worlds is that, once the make-believe world is created, their creators enter into the make-believe world to play their game. One result of this is that once a makebelieve truth is stipulated then, when actually playing the game, the player makebelievedly believes it. For example if it is stipulated by Jenm' and Sally that  Tanya  will come for tea* then, in playing the game Jenny and Sally *believe that Tanya will come for tea*. The fact that one enters into the make-believe world when playing the game can be put to use in encorporating other beliefs into the make-believe world besides those generated by stipulation in the following manner:  (A) If B believes that P and there is no set of make-believe truths A ... A^ which will prevent B *believing that P* then P is a make-believe truth.  16  (B) If A ... A 1  is a set of make-believe truths and 'If B *believes that n  A ... A * then, were he to think about it, B * would believe that P*' is 1  n  true, then P is a make-believe truth.  17  The notion of a make-believe truth T 'preventing' B *believing that P* is perhaps best spelled out in terms of B's inability to *believe that P* if he believed that T* given that 16  :|:  B is being rational. Note the use of this counterfactual is permissable since there is no problem in specifying a possible world in which A thinks about a make-believe world in a different way than in fact he does. 17  Vacuous  179  Names  So, by means of stipulation, we can create make-believe truths which incorporate entities which are purely make-believe. By means of (A), all believed facts about the world are incorporated into the make-believe world without the need for their being given by stipulation (provided, that is, no make-believe truth rules them out). Thus truths such as the sky being blue are incorporated into the make-believe world. By means of (B) it can be *discovered* that *Johnny is Tanya's brother-in-law*, that * Johnny lives in this house*. Note that make-believe truths which are expressed by means of proper names which are, in the actual world, vacuous can only be generated by stipulation or by means of the recursive principle, not by means of the encorporation principle. This account of make-believe truth does, I believe, help to illuminate the phenomena casually referred to as 'pseudo-referring', and as 'pseudo-' or 'mock' assertions by many philosophers, including those who do not hold the direct reference view: Frege and Strawson to name but two.  18  The account, of course, is not complete but  whatever the problems of make-believe truth, they are not specific to direct reference theories far less to belief attributions per se. Consequently, while acknowledging the sketchiness of the theory, I shall conclude by applying Evans' insights to A's attempts to respond to Ma's interrogatives in the examples given earlier. When faced with an interrogative which contains a vacuous name, as in (11?), (12?) and (13?), A can treat it in two different ways. Being open and sincere he will treat it as containing a vacuous name, as not asking a question, and hence will not provide an answer. Alternatively, if he wishes to be devious, he can treat the interrogative as one made within the game of make-believe. He can treat (1?), for example, as  1 8  See  Frege,  Correspondence,  Posthumous  Writings,  p.  130;  Philosophical  p. 152; and Strawson, 'On Referring', p. 182.  and  Mathematical  Vacuous Names  180  (1?F) * Where is Vulcan?* (1?F) does *ask a questio<r , a question* which *can be answered*, viz. 1:  (IF)  * Vulcan is over there.*  The situation is similar with the second type of example where M a believes her use of the interrogative to offer in the set of alternatives a proposition correctly expressed by means of a sentence containing a vacuous name. If A is being open and sincere, although he can provide an answer - a question having been asked, he will not do so be means of a sentence containing the vacuous name. However, if he is being devious he will treat the interrogative as being used to ask a make-believe question. For example, he will treat (5?) Who is going to carve? as (5?F) *Who is going to carve?* (5?F) does present, in the set of alternatives ' F r a n k is going to carve* and so A can * answer* (5?F) with (5F) *Frank is going to carve.* M a of course, did not ask (5?F) she asked (5?). It is the mark of A's deviousness that he treats her use of the interrrogative in this fashion. However way he treats Ma's interrogatives, be it of either type, A will not answer it by means of a sentence containing a vacuous name though he may so  Vacuous  Names  181  * answer* it. He will not attribute a belief to another in order to provide a tentative answer by means of a sentence containing a vacuous name though he may so provide a * tentative answer*.  * * * * * * * * * * * * * * * * * Direct Answers.  In the previous section we saw how, if the attributor was being open and sincere, he would not use a vacuous term in a belief attribution asserted as a tentative answer. For example, if he is being open and sincere, A would not respond to Ma's question Who's coming to dinner tonight? with Mary believes that Frank is. if he knows that there is no individual named by 'Frank'. (Jane, it will be remembered, has invented a boyfriend of that name.) Given this knowledge, A knows that such a response would be of no help to M a at all. It would not provide her with even a tentative answer to her question: there is no-one of whom it is even possibly the case that he is coming to dinner. The situation is very different, however, in cases where, were it (pei impossibile)  to be successful, the attribution would constitute a direct answer. In such  circumstances we often appear to be ascribing a belief to someone using a name we know to be vacuous. For example, A is asked what his nephew Billy is like. After answering,  Oh, he's a great little guy. Full of spunk... a bit cheeky. Very gullable though. he adds  ~  Vacuous  182  Names  (6) He believes that Santa won't come unless he brushes his teeth for three minutes every night. Now A does not believe in Santa. Yet he would seem to be attributing a belief to Billy by means of the name 'Santa'. Such cases need not concern well-known legendary characters: Ma, for instance, interested in whether Betty still does not know that 'Frank' is afigmentof her sister's imagination, utters the interrogative (7?) 'Does Betty believe that Frank is coming to dinner tonight?' to which B replies 'Yes.' or (7) 'Betty believes that Frank is coming to dinner tonight.' despite the fact that he himself knows that Frank does not exist. Such cases would seem to constitute good evidence that something is very wrong with a view of belief attributions which holds that if a vacuous name occurs within the embedded sentence used to attribute the belief then no proposition has been expressed and no belief attributed. B would seem to be answering some information  the question, to be giving Ma  about Betty and Billy. But how can this be if no proposition is  expressed? Further, unlike certain cases discussed above, there is no question of A's being devious here. He is not out to deceive Ma. She realizes that Frank does not exist. She does not believe in Santa. It would seem that in both cases A is treating Ma's utterance as a straight question to which he is straightforwardly attempting to provide a direct answer, to tell Ma what he knows about Billy and Betty. Or so it would appear. But of course on the theory under discussion, if 'Frank' is vacuous then Ma has not even succeeded in asking a question to which A could supply an answer, direct or otherwise.  Vacuous Names  183  Further, with respect to what A knows about Betty, if (7) is neither true nor false then (8) A knows that Betty believes that Frank will come to dinner tonight. is also neither true nor false. Where a vacuous term occurs in a sentence it denies the possibility of truth or falsity to an utterance no matter how deeply the sentence is embedded. But surely A does know something about Betty and little Billy, something which he is attempting to pass on to Ma. Let us approach the problem from the point of view of what B knows about Betty. B is aware that 'Frank' is the name given to a ficticious boyfriend by Betty's sister Mary, and also knows that Betty has been completely taken in by Mary's story; that is, he knows that Betty believes that this ficticious character actually exists. That is, A knows that (9) Betty believes that (3x)(x is Mary's boyfriend and x is called Frank). is true. This Betty would accept. But Betty would also assert that she had a belief which she would express by means of (10) Frank is coming to dinner tonight. A knows that Betty would act this way. Now, as we have seen, (10) does not express a proposition. Yet, because A knows that (9) is true and because he knows the history of the invention of 'Frank', he understands from Betty's actions that she has mistaken a fictional character for a real person. Or, to put it another way, A comes to understand that  Betty believes, of the fictional character Frank, that he is real.  Vacuous  Names  184  This, of course, Betty would deny. This need not concern us - Betty is simply mistaken. What is of concern, though, is that we cannot say that in uttering (10) Betty is, though she does not realize it, in fact talking about a fictional character. If a necessary condition for successful reference is held to be that the speaker intend to refer to the entity that is conventionally picked out by the name, then there is no intention on Betty's part to refer to a fictional entity and so no reference to that entity can be said to be made by her use of 'Frank'. Nor can we say that she is acting within a fiction, and treat her utterance as (10F) * Frank is coming to dinner tonight.* for much the same reasons. The value of the assertive force attached to Betty's utterance is that of all other everyday assertions: she is not talking fiction, not playing the pretend game. So, Betty is not talking about a fictional entity, nor is she talking fiction. For this reason, despite her intentions, Betty has not succeeded in saying anything  by means of  her utterance. No proposition has been expressed whatsoever; there is nothing to understand. Yet, there is a way for A to come to some sort of understanding on the basis of Betty's utterance. It is the utterance of a person who believes that Frank is real. We have all played a make-believe game in which we try hard to believe that a fiction is real. The closest A can get to understanding Betty is by analogy to a game wherein not only is the fictional world set up and is entered into, but the pretense is so successful that the participants sometimes forget that they are in a pretend world and think it real. Children, at times, reach this state; good actors, in a certain conditional way, always do. So A can understand Betty only insofar as her behaviour is analogous to that made within a fictional world - she is playing the pretend game to its ultimate, having included  185  Vacuous Names  herself totally within the pretence. Consequently, A can come to a sort of understanding of Betty's utterance by treating it, not as (10) but as (10F). Similarly, when little Billy tells him (11) Santa will only come if I brush my teeth for 3 minutes every night. Billy has not expressed a proposition; there is nothing to know. But by understanding Billy to be in the equivalent of a fictional world, to be making a serious assertion from so deeply with the make-believe game that he does not realize it is a make-believe, A can understand Billy's utterance as (11F) Santa will only come if I brush my teeth for 3 minutes every night. * :i:  When faced with Ma's interrogative, he takes it not as expressing Does Betty believe that Frank is coming to dinner tonight? but as expressing "Does Betty believe that Frank is coming to dinner tonight?* Hence though not asking a question, it nevertheless asks a question*, a * question* to :|:  which he provides an *answer*. In doing so A, though he does not provide information, there being nothing expressed by his utterance, nevertheless does *express something* by his utterance. M a has sought for and received * information*. *  *  *  *  *  *  *  *  *  *  *  *  *  *  *  *  *  This completes the investigation of belief attributions made by means of a sentence which contains a vacuous name within the embedded sentence. Cases in which  Vacuous  Names  186  the attribution was made in order to provide a tentative answer and those in which it constituted a direct answer were examined. Where the assertor is being open and sincere a difference was noted between the two types of case. Where the attribution was made in order to provide a tentative answer A would not (being open and sincere) use a name which he knows to be vacuous in attributing the belief since he is attempting to provide his audience with information about some aspect of the world other than the believer or his beliefs and he knows full well that this cannot be achieved by such means. Should he, unwittingly, use a vacuous name in the embedded sentence then he has failed to attribute a belief to B. In such cases the situation with respect to belief attributions is on a par with straightforward assertions of the embedded proposition. Since, in the case of tentative answers, the attribution is made for the sole purpose of modifying the claim to truth of the embedded proposition no problems specific to belief attributions are to be encountered. The situation was found to be very different with respect to cases in which the belief attribution constitutes a direct answer. There, A could well attribute a belief to B using, in the embedded sentence, a name known to be vacuous. The problem arose that although, under the direct reference view, no proposition has been expressed in such cases, something nevertheless would seem to have been understood. In order to deal with such cases Evans' account of pretend games, developed for cases of devious tentative answers, was appealed to. There is indeed a form of understanding in such cases. It is an understanding not based on standard semantic content but by analogy to a fictional game.  187  SUMMARY AND CONCLUSIONS  The investigation into the various uses of belief attribution statements is now completed. It was started, it will be recalled, in an effort to come to grips with problems thrown up by an application of the theory of direct reference to such statements. The standard criticism of the direct reference theory is that, while it may initially seem to be the more intuitively appealing when atomic sentences are examined, the theory fails miserably when applied to belief attributions and, for this reason, must be rejected. It is not always noted that the emphasis placed on a theory's capabilities in solving the problems of belief attributions is itself a reflection of the Fregean point of view. Fregeans see it as being the first requirement of a semantic theory that it address the problems of belief attributions. Direct reference theorists need not, however, accept 1  this demand. Nevertheless, while they can in good conscience reject the view that the primary area of concern is that of belief contexts, direct reference theorists must ensure that, at the very least, their theory does not rule out the very possibility  of a solution to  the problems of belief attributions. The examination of the use of belief attribution 2  statements given above which presupposes a direct reference point of view shows that this is not the case. Thus to argue against the Fregean view, as direct reference theorists tend to do, solely in terms of the unintuitive results obtained by an application of Fregean theory to atomic sentences is not strictly to come to grips with the core of the Fregean view. However, as argued in Chapter 2 above, that theory also fails to offer an unproblematic account in the area which it itself holds to be the vital one, that of belief attribution statements. 1  That is, given that it is accepted that there are problems. One could, I suppose, be hard-nosed about it and deny that there are any problems to be solved with respect to belief attributions. Few, however, would be willing to take this position and fly in the face of the widespread intuition that such problems exist. 2  Summary  and  Conclusions  188  The examination started by rejecting the Fregean approach to the problems of belief attributions which relies on an appeal to the speaker's cognitive relationship with the proposition. Faced with the task of developing an alternative method of approach, help was gained by concentrating on the fact that the subject under discussion is not belief but belief attributions. Consequently, an examination of the use of such statements in non-problematic cases was conducted. Two general purposes for their use were identified, that of passing on information and that of correcting mis-information. It was noted that the analysis in terms of singular propositions is not finegrained enough to capture the differing information which could be conveyed by the use of the same belief attribution. What was needed was a treatment of attributions which is sensitive to conversational context. This was achieved by taking the attribution to be an answer to a certain question. With the logic of questions and answers in place, an account of the functioning of the various particular uses of the belief attribution began to emerge. Two main uses were distinguished: that in which the belief attribution- is used in order to supply a direct answer to the question and that in which it is used to offer a tentative answer. While in cases of the former type of use the subject under discussion is the believer or the belief itself, this is not the case in cases involving the attributions being used as a tentative answer. There the subject of discussion is some quite different matter. While the basis of the distinction forms, I believe, the impetus for many other distinctions drawn with respect to belief attributions, no-one to my knowledge has thought it strange that, in a conversation about some quite different matter, the assertion of a belief attribution somehow forwards  the conversation. That a belief  attribution is often asserted in such circumstances should, on the standard Fregean or Kaplanesque treatment of belief attributions for example, appear to be exceedingly odd. Treating belief attributions as answers, however, opens up the way for an alternative  Summary  and  Conclusions  189  treatment of such a usage. Examination of cases shows that belief attributions are used as tentative answer? v.-hen the assertor is uncertain as to the truth of the embedded proposition, though thinking that, of all possibilities, it is the -most likely. Although reluctant simply to assert the proposition itself, the attributor can provide a tentative answer to the question by attributing a belief in that proposition to another and in so doing tentatively endorse it himself. With the examination of non-problematic cases of belief attribution statements in place, attention was then given to the problematic cases. Lest the two types of usage yield differing results with respect to the problems of substitution and vacuous names, cases involving the two type of uses were examined separately. With substitution, although it at first appeared as if there would be a difference between the two types of case, upon further examination this proved not to be so. When cases of attributions asserted as tentative answers were examined, where the subject under discussion is not the belief or the believer but some quite other matter, it was found that substitution was possible in all cases. From this it might be tempting to conclude that when attention is on the believer or, more specifically, is on the belief itself, substitution would not be possible. This, however, would be too quick, for when cases of belief attributions asserted as direct answers were examined, it was discovered that substitution is indeed possible in all cases since, in cases of direct answers, although the focus is on the belief itself, it is not on how that belief is believed. Hence, those theorists who take a hard-line with respect to the substitution of co-referential names are vindicated: substitution is possible in all cases of belief attribution because there is no implication that the proposition is expressed by the belief attribution in a manner acceptable to the believer. This conclusion is seen to throw up problems in the area of the explanation of action. However when the cases were examined more closely this proved not to be the  Summary  and  Conclusions  190  case. Given the position that the determination of the question being answered is vital to the treatment of the belief attribution, and given a more natural account of what constitutes an explanation with respect to human actions, the problems associated with substituion and action explanations disappear. As we saw, the distinction between direct and tentative answers, although illuminating with respect to the different uses a belief attribution can be put to, has no role to play with respect to the possibility of substituting co-referential names within the embedded sentence: substitution being possible in both types of case. It does, however, have a role to play in cases involving the use of a vacuous name within the embededded sentence. An examination of cases of belief attributions asserted as tentative answers revealed that, in such circumstances, the assertor would never knowingly so use a vacuous name. Because of the nature of tentative answers, because the belief is attributed for the sole purpose of modifying the claim to truth of the embedded proposition, the treatment of attributions asserted as tentative answers, so far as cases of vacuous terms are concerned, is on a par with that of a direct assertion of the embedded propostion. No problems, specifically germane to belief attributions, pertain in such cases. Just as A has not asserted a proposition were he to assert Fa, where 'a' is vacuous, so he has not were he to assert 'B believes that Fa' as a tentative answer. So A , knowing the name is vacuous, would never use it in an attribution asserted as a tentative answer (if he was being sincere). The situation was found to be very different, however, were the belief attribution to be made as a direct answer. In such a situation A would indeed be willing to assert 'B believes that Fa' knowing full well that 'o' is vacuous. So the problem was raised as to what is it that A understands about B that he is trying to communicate in such circumstances; what is it that A understands when he understands, as presumably he does, his own utterance. Knowing that 'a' is vacuous, it cannot be a belief in a singular  Summary  and  191  Conclusions  proposition which he attributes to B. Help in this matter was gained by an appeal to Evans work on make-believe utterances, though somewhat altered in order to retain a possible worlds treatment of counterfactuals. What A understands, what he is trying to communicate in such cases, cannot be reached by means of an examination of the standard semantics of the language. Rather his understanding is by analogy to cases of a pretend game in which one has forgotten that one is playing a pretend game. This would seem to be as close as B can get in such circumstances to understanding and talking about A's state of mind. It is important to bear in mind that all of the above findings concern themselves with the nature of belief attributions and their use, not with the nature of belief. The direct reference theorist differs from the Fregean (perhaps this is the main difference) in that, when discussing belief attributions, it is not assumed that one is also talking about belief. Nevertheless, just as it was incumbant upon the direct reference theorist to ensure that the account offered does not deny the possibility of a solution to the problems of belief attributions,  so it is incumbent upon the theorist to ensure that the  possibility of a coherent account of belief is not so ruled out. Not that a complete 3  account of belief has to be provided: simply that it is ensured that an account of belief is possible. That this is the case was shown by the sketch of belief given by Wettstein. To have a belief one need not have the object 'smack up against one's mind', far less 'before one's inner eye'. One may have a belief which concerns an object in the world without  Assuming, that is, that one grants that there is more to belief. One could, of course follow the Fregean and deny any substance to the notion of belief beyond that yielded by the treatment of belief attributions. This is the route Dennett takes (see for example The Intentional Stance, Cambridge, Mass., 1987). Belief, for Dennett, is generated solely by what others wish to attribute to the believer. There is no fact of the matter other than this. Hence Dennett is not only an instrumentalist with respect to belief attributions, he is also an instrumentalist about belief. 3  Summary and Conclusions  192  the relationship between the believer and that object being a cognitive one. This view of belief renders it entirely understandable how one may have a belief, % is F, which one would express by asserting 'a is F' while denying the truth of b is F' even though 'a' l  and ' 6 ' are coreferential. Similarly, since the relationship between the believer and the object need not be cognitive, it is understandable how one could believe that there was an object involved when in fact there was none. Since the object need not be 'smack up against one's mind', it is entirely possible to be mistaken about its existence. Given that the tie to the object need not be cognitive, the 'internal feel' would, in such cases, be the same whether the object existed or not. So the direct reference view proves not only to be the more intuitively satisfying with respect to atomic statements, it satisfies the demands placed on it with respect to belief attribution statements and with respect to belief: a solution to the problems of belief attributions, given the direct reference view, is possible; a coherent account of belief is not ruled out by the theory. I believe that the sketch of belief offered above and, more substantially, the account given of the functioning of belief attributions shows just what form that solution takes.  193  BIBLIOGRAPHY OF WORKS CITED  Almog J., 'Semantic Anthropology', in Midwest Studies in Philosophy, vol. IX, eds. P. French, T. Uehling, and H., Wettstein, Minneapolis, 1984, pp. 479-490. Almog, J., 'Form and Content', Nous, 19 (1985), 603-616. Almog, J., 'Naming without Necessity', Journal of Philosophy, 83 (1986), 210-243. Austin, J.L., 'How to Talk - Some Simple Ways', Proceedings of the Aristotelian  Society,  1922-3, 227-246.  Baker, G.P. and Hacker, P.M.S., Frege: Logical Excavations, Oxford, 1984. Baker, L.R., 'De Re Belief in Action', Philosophical  Review, 91 (1982), 363-387.  Barwise, J. and Perry, J., Situations and Attitudes, Cambridge (Mass.), 1983. Belnap, N.D., 'Questions and Presuppositions', in The Logical Way of Doing Things, ed. K. Lambert, New Haven, 1969, pp. 23-37. Belnap, N.D. and Steel, T.B., The Logic of Questions and Answers, New Haven, 1976. Blackburn, S., 'The Identity of Propositions', in Meaning, Reference and Necessity, ed. Simon Blackburn, Cambridge, 1975, pp. 182-205. Blackburn, S. and Code, A., 'The Power of Russell's Criticism of Frege: "On Denoting", pp. 48-50', Analysis,  38 (1978), 65-77.  Carnap, R., Meaning and Necessity, Chicago, 1967. Caruthers, P., 'Russellian Thoughts', Mind, 96 (1987), 18-35. Chisholm, R., 'Knowledge and Belief: De Dicto and De Re', Philosophical  Studies,  29  (1976), 1-20.  Church, A., 'The Need for Abstract Entities in Semantic Analysis', Proceedings of the American Academy of Arts and Sciences, 60 (1951), 100-112; also in The Structure of Language, eds. J.A. Fodor and J.J. Katz, Englewood Cliffs, N.J., 1964, pp. 437-445.  Dascal, M., 'Conversational Relevance', in Meaning Dordrecht, 1979., pp. 153-74.  and  Dennett, D., The Intentional Stance, Cambridge, Mass., 1987.  Use,  ed. A. Margalit,  194  Bibliography  Donnellan, K., 'Reference and Definite Descriptions', Philosophical 281304.  Review, 75 (1966),  Donnellan, K., 'Putting Humpty Dumpty Together Agaiu', Philosophical (1968), 203-215.  Review, 77  Donnellan, K., 'Proper Names and Identifying Descriptions', in Semantics of Natural Language, eds. D. Davidson and G. Harman, Dordrecht, 1972, pp. 356-379. Donnellan, K., 'Speaking of Nothing', Philosophical Donnellan, K., Philosophy  Review, 83 (1974), 3-31.  'The Contingent a Priori and Rigid Designators', in Midwest Studies in II:  Studies  in  the Philosophy  of Language,  eds.  P.A. French.  T.E. Uehling and H.K. Wettstein, Minneapolis, 1977, pp. 12-27. Dray, W., Laws and Explanation  in History, Oxford, 1957.  Dray, W., 'The Historical Explanation of Actions Reconsidered' in Philosophy History, ed. S. Hook, New York, 1963, pp. 105-135. Dretske, F., 'Contrastive Statements', Philosophical  and  Review, 81 (1972), 411-437.  Dretske, F., 'Referring to Events', in Contemporary Perspectives in the Philosophy of Language, eds. P.A. French, T.E. Ueling and H.K. Wettstein, Minneapolis, 1979, pp. 369-378. Dummett, M . , Frege, Philosophy  of Language, London, 1973.  Dummett, M . , 'Frege as a Realist', Inquiry, 19 (1976), 455-468. Dummett, M . , The Interpretation of Frege's Philosophy,  Cambridge, Mass., 1981.  Evans, G., The Varieties of Reference, Oxford, 1982. Feldman, R., 'Actions and De Re Beliefs', Canadian 577-82.  Journal  of Philosophy,  8 (1978),  F^llesdal, D., 'Reference and Sense', in Philosophy and Culture, Proceedings of the XVII World Congress of Philosophy, Montreal, 1983, Montreal, 1986, pp. 229-239. Frege, G.,  'On Sense and Reference', in Translations  from the Philosophical  Writings  of  Writings  of  Gottlob Frege, eds. P. Geach and M . Black, Oxford, 1960, pp. 56-78. Frege, G.,  'Function and Concept', in Translations  from the Philosophical  Gottlob Frege, eds. P. Geach and M . Black, Oxford, 1960, pp. 21-41. Frege, G., 'The Thought', trans. A . M . and M . Quinton, Mind, 65 (1956), 289-311; also in Philosophical Logic, ed. P.F. Strawson, Oxford, 1967, pp. 17-38. (Page references are to this latter volume.)  195  Bibliography  Frege, G., Posthumous writings, eds. H. Hermes, F. Kambartel and F. Kaulbach, trans. P. Long and R. White, Oxford, 1979. Frege, G., Philosophical and Mathematical Correspondence, eds. G. Gabriel, H. Hermes, F. Kambartel, C. Thiel and A. Veraart; abridged by B. McGuinness; trans. H. K a a l ; Oxford, 1980. Garfmkel, A . , Forms of Explanation, Grice, H.P.,  New Haven, 1981.  'Logic and Conversation', in The Logic of Grammar,  eds. D. Davidson and  G. Harman, Encino, California, 1975, pp. 64-75. Hamblin, C.L., 'Questions', Australian  Journal of Philosophy,  36 (1958), 159-168.  Hart, H . L . A . & Honore, A . M . , Causation and the Law, Oxford, 1959. Hempel, C , 'The Function of General Laws in History', Journal of Philosophy, (1942), 35-48.  39  Hempel, C. and Oppenheim, P., 'Studies in the Logic of Explanation', in Readings in the Philosophy of Science, eds. H. Feigl and M. Brodbek, New York, 1953, pp. 327-8. Kaplan, D., 'Trans-World Heir Lines', Address presented to a joint synposium of the Association for Symbolic Logic and the American Philosophical Association, May, 1967; published, with additional footnotes in The Possible and the Actual, ed. M . J . Loux, 1979, pp. 88-109. Kaplan, D.,  'Quantifying In', Synthese, 19 (1968), 178-214; reprinted in Words and  Objections: Essays on the Work ofW.V. Quine, eds. D. Davidson and J . Hintikka,  Dordrecht, 1969, pp.206-242. A l l page references are to the latter publication. Kaplan, D., 'Bob and Carol and Ted and Alice', in Approaches to Natural Language, eds. K . J . Hintikka, J . M . Moravcsik, and P. Suppes, Dordrecht, 1973, pp. 490-518. Kaplan, D., 729.  'How to Russell a Frege-Church', Journal  Kaplan, D.,  'Dthat', in Syntax and Semantics 9: Pragmatics,  1978, pp.  221-243.  Also in Contemporary  of Philosophy,  Perspectives  72 (1975), 716-  ed. P. Cole, New York, in the Philoasophy  of  Language, eds. P. French, T.E. Uehling and H.K. Wettstein, Minneapolis, 1979, pp. 383-400. Page references are to this latter volume. Kaplan, D., 'Demonstratives', in Themes from Kaplan, ed. by J . Almog, J . Perry and H Wettstein, New York, 1989, pp. 481-563. Kaplan, D., 'Afterthoughts', in Themes from Kaplan, ed. by J . Almog, J . Perry and H. Wettstein, New York, 1989, pp. 565-614. Kripke, S., 'Identity and Necessity', in Identity and Individuation, New York, 1971, pp. 135-164.  ed. M.K. Munitz,  Bibliography  196  Kripke, S., 'Naming and Necessity', in Semantics of Natural Language, eds. D. Davidson and G. Harman, Dordrecht, 1972, pp. 253-355, and pp. 763-769; also published as a book with additional preface, Cambridge, Mass, 1980. All page references are to the latter publication. Kripke, S., (untitled) Transcript of a seminar given onfictionalentities, University of California at Berkeley, 1972 (unpublished). (Referred to as "Fictional Entities".) Kripke, S., 'Speaker's Reference and Semantic Reference', in Contemporary Perspectives in the Philosophy of Language, eds. P. French, T. Uehling, and H. Wettstein, Minneapolis, 1979, pp. 6-27. Kripke, S., 'A Puzzle about Belief, in Meaning 1979, pp. 239-283. Leonard, H., Principles  of Reasoning,  and  Use, ed. A. Margalit, Dordrecht,  New York, 1959.  Linsky, L., Referring, New York, 1967. Linsky, L., Names and Descriptions,  Chicago, 1977.  Linsky, L., Oblique Contexts, Chicago, 1983. McDowell, J., 85.  'On the Sense and Reference of a Proper Name', Mind, 86 (1977), 159-  McGinn, C, The Character of Mind, Oxford, 1982. Pastin, M., 'About De Re Belief, Philosophy 569-75.  and Phenomenological  Research,  34 (1975),  Patton, T.E., 'Referential / Attributive: An Appraisal of an Account', unpublished ms., 1983. Perry, J., 'Frege on Demonstratives', Philosophical  Review, 86 (1977), 474-97.  Perry, J., 'The Problem of the Essential IndexicaP, Nous, 13 (1979), 3-21. Perry, J., 'A Problem about Continued Belief, Pacific (1980), 317-332.  Philosophical  Quarterly,  61  Plantinga, A., 'The Boethean Compromise', American (1978), 129-138.  Philosophical  Quarterly,  15  Quine, W.V., 'Notes on Existence and Necessity', Journal 113-237.  of Philosophy,  40 (1943),  Quine, W.V., 'Three Grades of Modal Involvement', Proceedings of XI International Congress of Philosophy, Brussells, 1953, vol. 14; reprinted in The Ways of Paradox, New York, 1966, pp. 156-174. Page references are to this latter volume.  Bibliography  197  Quine, W.V., 'Quantifiers and Propositional Attitudes', Journal 53 (1956), 177-187.  of  Philosophy,  Quine, W.V. 'Reply to David Kaplan', in The Philosophy of W.V. Quine, eds. L.E. Hahn and P.A. Schilpp, La Salle, Illinois, 1986, pp., 290-4. Russell, B., 'On Denoting', Mind, 14 (1905), 479-493; also in Logic and Knowledge, ed. R.C. Marsh, London, 1956, pp. 41-56. All page references are to the latter volume. Salmon, N.U., Reference and Essence, Oxford, 1982. Salmon, N.U., Frege's Puzzle, Cambridge (Mass.), 1986. Searle, J., 'Russell's Objections to Frege's Theory of Sense and Reference', Analysis, 17 (1957), 137-143; also in Essays on Frege, ed. E.D. Klemke, Urbana, 1968, pp. 337-348. Searle, J., 'Proper Names', Mind, 67 (1958). 166-173. Searle, J., Speech Acts, Cambridge, 1969. Searle, J., Intentionality, Cambridge, 1983. Sluga, H., 'Frege and the Rise of Analytic Philosophy', Inquiry,  18 (1975), 471-487.  Sluga, H., 'Frege's Alleged Realism', Inquiry', 20 (1977), 227-242. Sosa, E., 'Propositional Attitudes, De Dicto and De Re', Journal (1970), 883-896.  of Philosophy,  Stiffler, E., 'De Re Belief Ascriptions and Action Explanations', Canadian Philosophy, 13 (1983), 513-525.  67  Journal of  Strawson, P.F., 'On Referring', Mind, 59 (1950), 320-344. Strawson, P.F., Individuals  , London, 1959.  Strawson, P.F., 'Review of Laws and Explanation (1959), 265-68.  in History by William Dray', Mind, 68  Strawson, P.F., 'Identifying Reference and Truth-values', Theoria, 30 (1964), 96-119. Strawson, P.F., Subject and Predicate in Logic and Grammar,  London, 1974.  Urmson, J.O., 'Paranthetical Verbs', Mind, 61 (1952), 480-496. van Dijt, T., 'Relevance Assignment in Discourse Comprehension', Discourse 2 (1979), 113-126. Van Fraassen, B., The Scientific Image, Oxford, 1980.  Processes,  198  Bibliography  Walton, K.L., 'Pictures and Make-believe', Philosophical Walton, K.L. 'Fearing Fictions', Journal of Philosophy,  Review, 82 (1973), 283-319. 75 (1978), 5-27.  Werth, P.N., 'The Concept of Relevance in Conversational Analysis', in and Discourse, ed. P.N. Werth, London, 1979.  Conversation  Wettstein, H., 'Has Semantics Rested on a Mistake?', Journal of Philosophy, 185-209.  83 (1986),  Wettstein, H., 28.  'Cognitive Significance without Cognitive Content', Mind, 97 (1988), 1-  Wettstein, H., 'Turning the Tables on Frege, or How Is It that "Hesperus is Hesperus" is Trivial?', forthcoming. Wilson, D. and Sperber,D., 'On Defining "Relevance"', in Philosophical Rationality, eds. R.E. Grandy and R. Warner, Oxford, 1986. Wittgenstein, L., Philosophical  Investigations,  Oxford, 1978.  Grounds  of  

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
http://iiif.library.ubc.ca/presentation/dsp.831.1-0302391/manifest

Comment

Related Items