Be t we e nQui ne ’ sDi s quot at i onal i s m and Hor wi c h’ sMi ni mal i s m Richard Wei Tzu Hou A thesis submitted in fulfillment of the requirement for the degree of Doctor of Philosophy Department of Philosophy The University of Sydney May 2006 Abstract Many criticisms of the prevalent deflationary theories of truth stem from some mi s unde r s t a ndi ng .Cl a r i f i c a t i onc a nbef oundf r om c ons i de r i ngQui ne ’ sr e a s oni ngon t hedi s quot a t i ona lf e a t ur eoft het r u t hpr e di c a t e .Qui ne ’ sdi s quot a t i ona l i s ma nd Hor wi c h’ smi nimalism are similar theses with respect to the concept of truth, though the difference between the choices of the primary truth bearers and the divergence in their accounts of meaning and reference are striking. Chapter Two is devoted to making plain Quine ’ sr e a s oni ngr e ga r di ngt he disquotational concept of truth, and to constructing a disquotational theory of truth. Also in this chapter, the topic of how to enhance the deductive power of this theory is discussed. The following chapter aims to square Quine ’ st he s e sofi ns c r ut a bi l i t yof reference and ontological relativity, with an account of the disquotational schema of reference. Whether or not a disquotational schema of reference and all its instances can be seen as providing a genuine reference scheme, as claimed by Horwich and most deflationists, is also discussed. In Chapter Four, after an introduction of Hor wi c h’ smi ni ma l i s tc onc e pt i ono ft r ut h,t he r ea r eanumbe rofi s s ue sc ons i de r e d,i n pa r t i c u l a rHor wi c h’ sus e -theoretic account of meaning and compositionality, along wi t ht hedi ve r g e nc ebe t we e nhi sa c c ountofme a ni nga ndQui ne ’ s . Thef i na lc ha pt e r , Chapter Five, provides a thorough analysis of three important factors regarding the disquotational theory and the minimal theory of truth. Among them, the first factor discussed is what sort of equivalence relation occurs within each instance of the disquotational schema or each axiom of the equivalence schema. Following this, there is an analysis of in what way the disquotationalist and the minimalist can explain all general facts involving truth. The last factor involves considering the proper ascription of the disquotational or the minimal truth predicate. Along with the analysis of these three factors, the issue regarding which theory of truth is preferable is elaborated. 1 Table of Contents Acknowledgement…………………………………………………………………2 Chapter I: Introduction…………………………………………………………. . . 3 Chapter II:Qui ne ’ sDi s quot at i onalThe or yofTr ut handSe mant i cAs c e nt 1. Ta r s ki ’ sPa r a di g monTr ut h……………………………………………………10 2. For mul a t i ngQui ne ’ sDi s quot a t i ona lThe or y …………………………………. 25 3. Cons e r va t i ve ne s sa ndSubs t a nt i ve ne s s ………………………………………. . 49 Chapter III: Inscrutability of Reference, Ontological Relativity, and a Disquotational Schema of Reference 1. Qui ne ’ sMa noe uvr e ……………………………………………………………63 2. Fi e l d’ sAr g ume nta g a i ns tOn t o l og i c a lRe l a t i vi t y ……………………………. . . 65 3. OnGi bs on’ sCr i t i c i s mofFi e l d’ sa ndDa vi ds on’ sAr g ume nt s ……………. . . …68 4. Disquotational Schema of Reference and Semantic Ascent of Terms………. . . 75 5. Di s quot a t i ona lSc he maofRe f e r e nc ea ndDi s quot a t i ona lThe or yofTr ut h…. . . 84 Chapter IV:OnHor wi c h’ sMi ni mal i s tThe or i e sofTr ut ha ndMe ani ng 1. TheFr a me wor koft heMi ni ma lThe or yofTr ut h……………………………. . . 95 2. Que l l i ngQui ne ’ sQua l msorQue l l e dbyQui ne ’ sQua l ms ? …………………. . 106 3. OnHor wi c h’ sUs e -The or e t i cAc c ountofWor dMe a ni ng …………………. . . . 125 4. I nf l a t i ngCompos i t i ona l i t y …………………………………………………. . . . 143 Chapter V: Be t we e nQui ne ’ sDi s quot at i onal i s m andHor wi c h’ sMi ni mal i s m 1. The Equiva l e nc eRe l a t i ona ndSe ma nt i cAs c e nt ……………………………. . . 158 2. Exp l a i ni ngGe ne r a lFa c t sI nv ol vi ngTr ut h……………………………………185 3. ThePr ope rAs c r i pt i onoft heTr ut hPr e di c a t e ………………………………. . . 203 4. I sQui neaDi s quot a t i ona l i s t ? ………………………………………………. . . . 21 1 Conclusion……………………………………………………………………. . . . . 220 Bibliography……………………………………………………………………. . 229 2 Acknowledgement I would like to thank my supervisor, Dr. David Braddon-Mitchell, for his support and encouragement, and also to thank my associate supervisor, Dr. Michael McDermott, especially for his supervision during 2003, when Dr. Braddon-Mitchell was away. I am grateful for the one-year research scholarship offered by Academia Sinica, Taiwan, on the period of 2003/2004, and the facilities, the research resources, and the kindly staffs of the Institute of European and American Studies, Academia Sinica. Special thanks to Dr. Patrick Yong and Ms. Kristie Miller, without their proofreading and English correction, this dissertation would have been less readable. I would like to thank my family, and my partner Vivien. Finally, my dog, Shine, thanks to him for keeping my life on track. 3 Chapter I Introduction Deflationism is probably the most noteworthy semantic theory of the past twenty odd years. It is still influential today. Among hundreds, if not thousands, of articles, discussing, criticising, or defending deflationism, some important issues are not as clear as they should be. This is, I argue, because of a pervasive tendency to over-complicate the key ideas. More complex versions of key aspects of the theory are subject to misunderstanding and refutation. The simpler versions are not. This thesis argues that this over-complication has caused problems in dealing with at least eight questions that arise for deflationary theories. These include: 1. What is a proper formulation of a disquotational schema of truth? 2. What is the modal status of disquotation sentences or the axioms of the minimal theory of propositional truth? 3. How sensible are the presuppositions involved in asking these questions about the modal status of sentences or axioms? 4. How should we give the deflationary theory of truth the deductive power it needs so as to adequately explain all general facts involving truth? 5. What is the deflationary conception of reference? Is it feasible to explicate the deflationary conception of truth in terms of the disquotational concepts of reference and satisfaction? 6. What is an appropriate account of propositions for use in a minimalist account of 4 truth? 7. What sort of equivalence relation occurs within disquotational sentences or minimalist axioms of truth? 8. Between disquotationalism, taking sentences as the primary truth bearers, and minimalism, taking propositions as the primary truth bears, which is the more promising approach to a deflationary theory of truth? These questions will be discussed only as they arise for the cases of disquotationalism and minimalism. If I am right the sources of many of the perceived problems of these theories will be exposed. At very least the theories will be clearer, and perhaps the reader will find them more plausible. However, since this dissertation is so confined, other deflationary theories of truth such as the prosentential theory will not be discussed. Hereafter, when I us et het e r m‘ de f l a t i oni s m’Ius ei tt or e f e rt os omeve r s i onsofdi s quot a t i ona l i s mor minimalism. There are of course many versions of deflationism—even versions of disquotationalism or minimalism—which I do not discuss; but my task will be achieved if I clear some of the obstacles to acceptance of the versions I do discuss, and this can be done without a more exhaustive survey. Apart from this agenda, this thesis makes another claim. It claims that one can read Quine as more or less tacitly accepting a disquotationalist or minimalist theory of truth. This is not a view which is widely held. Although the general idea of disquotationalism and some de f l a t i oni s tt he or i e si si ns pi r e dbyQui ne ’ svi e wpoi ntont r ut h,f e ws e r i ous l ys e eQui nea s adhering to a deflationist position regarding the concept of truth in particular, or the semantic theory in general. Among other reasons, the most important one is that Quine 5 specifically expresses his support for a Davidsonian truth conditional approach to explicating the concepts of truth and meaning. However, discussion regarding this issue will not be addressed until the very end of this dissertation. Except in contexts where a deflationist or an anti-deflationist introduces the idea of considering the truth predicate as a device of semantic ascent,Qui ne ’ sr e a s oni ngi sha r dl y mentioned. Put aside for the moment whether or not Quine is really a deflationist, and disregard his viewpoints on truth that is not disquotational. Nevertheless, it is still the case that it is Quine who sets up clear and distinct guidelines on a proper understanding of di s quot a t i ona l i s m.I nCha pt e rTwo,Iwi l le xpl a i nwha tQui ne ’ sr e a s oni ngi s ,a ndhowi t gives us a clear perspective of disquotationalism. Another important task in this chapter is to analyse in what sense a disquotational theory of truth can be seen as providing enough deductive power to explain some general facts involving truth. It will be shown that to explicate truth in terms of the inductive axioms of reference and satisfaction is a dubious move. This leads us to Chapter Three. Along with the deflationary theory of truth, most deflationists accept a disquotational theory of reference (and satisfaction). However, to consider a disquotational theory of reference as providing a genuine reference scheme for ag i ve nl a ng ua ge ,Qui ne ’ st he s i sofont ol og i c a lr e l a t i vi t ys e e mst obef a l l a c i ous .I nt hi s chapter I show that it is not in fact fallacious. After that, I will present my interpretation of the role played by a disquotational schema of reference in Qui ne ’ sdi s quot a t i ona l i s m. Finally I will argue that a disquotational schema of reference cannot be seen as providing us a genuine reference scheme. 6 The nwet ur nt oHor wi c h’ ss e ma nt i ct he or y .I nCha pt e rFour ,a f t e rabr i e fi nt r oduc t i ont o Hor wi c h’ svi e wson truth, along with the discussion of whether or not minimalism can be finitely formulated, the focus is on one rationale of the minimalist theory of propositional truth—namely, a viable account of proposition. The purpose of this chapter is to show that Hor wi c h’ sr e j e c t i onofQui n e ’ si nde t e r mi na c yt he s i sf a i l s ,t ha tHor wi c h’ s use-theoretic account of meaning is unsatisfactory, and that his account of compositionality not only fails to be deflationary, but is also based on some uncritically accepted assumptions such as that different languages share the same synthetic features. Based on the outcome of the preceding three chapters, I make some important clarifications in Chapter Five; specifically I enunciate three important factors for a successful deflationary theory of truth. The first of these factors concerns the functional role of the truth predicate to do the semantic ascent required to help express infinite conjunctions and disjunctions. The second of these concerns the adequacy of the explanatory role in terms of which all facts involving truth are explained along with other suitable explanatory resources. The last of these concerns the proper ascription of the t r ut hpr e di c a t e .It he ns howt ha t ,a l t houg hHor wi c h’ ss t a ndpoi ntont het he or yofme a n i ng i ss ha r pl yde ma r c a t e df r om Qui ne ’ s , some basic ideas of his minimalism are not very different from those behind my formulation of Qui ne ’ sdi s quot a t i ona l i s m. Thef i r s t section of the chapter, then, concerns the first factor above. In it I analyse what sort of equivalence relation occurs within disquotation sentences and the axioms of the minimalist theory of truth. The second section is devoted to the analysis of how disquotationalism and minimalism can account for all general facts involving truth. Then 7 in the third section, the proper application of the disquotational or minimalist truth predicate will be given. Along with the study of these three factors, I make an assessment ofQui ne ’ sdi s quot a t i ona l i s m ve r s usHor wi c h’ smi ni ma l i s m.Fi na l l y ,i nt hel a s ts e ction I consider some not very coherent attitudes of Quine, toward the matter of truth. There are two additional matters that need to be addressed. In Chapter Three I concluded that a disquotational schema of reference is not supposed to be seen as providing a genuine reference scheme. It should be noted that the consequence of this observation is not pursued in this dissertation. Furthermore, the issue of how to make sense of the disquotational and the minimalist accounts of truth conditions is not discussed, for this issue does not undergird any difference between disquotationalism and minimalism and is therefore tangential to this thesis. The other matter is rather minor. It is about the relation between a deflationary account of truth and any given theory of meaning, insofar as it is not a truth conditional one. In both Chapter Two and Chapter Five, after I draw the conclusion that a proper understanding of the disquotational schema of truth and its instances should not involve any particular way of interpreting the content of the sentence mentioned and used in each instance of the schema, I will not further pursue how to sort out an account of meaning that is suitable for a disquotational theory of truth. For I rather think that this dissertation will not benefit from any conclusion reached by our study of this issue. Accordingly, these two issues are left for future study. 8 Chapter II Qui ne ’ sDi s quot at i onalThe or yofTr ut h and Semantic Ascent1 Deflationism is among the most prominent theories of truth. Two of the most important de f l a t i ona r yt he or i e soft r ut ha r edi s quot a t i ona l i s m,i ns pi r e dbyQui ne ,a ndHor wi c h’ s mi ni ma l i s m.Bot hoft he s ec a nbes e e nt oha vede ve l ope df r omTa r s ki ’ sConve nt i onT, though offering quite distinct theories that diverge in significant ways.2 Qui ne ’ sf a moust he or ya boutt r a ns l a t i on—that translation is indeterminate—is one of the most significant and influential theories of meaning. It is opposed to the more intuitive account of meaning that takes intensional notions such as analyticity and synonymy as playing central roles in theories of meaning and/or content. This approach also has an effect on his account of truth, namely, that his disquotational theory of truth is a revision ofTa r s ki ’ si ng e ni ousT-schema. Al t houg honeoft hema j orc ont r i but i onsofTa r s ki ’ s semantic account of truth is to explicate the semantic conception of truth in terms of the a na l y s i sofs a t i s f a c t i ona ndr e f e r e nc e ,t hi si swha tQui ne ’ sdi s quot a t i ona la c c ountoft r ut h rejects. The importance of this disquotationalist account is not just that it offers a new 1 A draft of this chapter has been presented in a seminar held at the Institute of European and American Studies, Academia Sinica, Taiwan in 2004. 2 The simplest version of Convention T is the set of bi-conditionals with the form (T): (T) X is true if and only if p, whe r e‘ X’i sr e pl a c e dbyt h ena meofa nys e n t e n c eoft h eobj e c tl a n gu a g ea n d‘ p’byt h es e n t e n c ei t s e l f or by a suitable translation within the metalanguage. Conv e n t i onTpl a y sa ni mpor t a n tr ol ei nTa r s k i ’ st heory of truth, that is, in the material adequacy of any given definition of truth. More precisely, any definition of truth (D) is materially adequate if and only if it implies all the T-biconditionals of the object language and implies that only sentences of the object language are true in the same language. 9 way of thinking with regard to the notion of truth, but it also fills in the final touch of Qui ne ’ sge ne r a lc onc e pt i onofs e ma nt i c s . Anot he rc r uc i a lpa r toft hi sa c c ounti sQui ne ’ s inter pr e t a t i onofTa r s ki ’ ss e ma nt i ct he or yoft r ut h.Hi si nt e r pr e t a t i ons howst ha tt het r ut h predicate has a special function both as grammatical device and in philosophical contemplation as semantic ascent (and descent), by which we substitute the talk of sentences for the talk of reality. However, Quine does not offer a precise formulation of his disquotational theory of truth. He also fails to specify to a desirable extent how this theory behaves if it is more desirably strong in deducing some generalisations. This chapter thus consists of three sections. The first section devotes a brief introduction t oTa r s ki ’ sConve nt i onT. Theot he ri mpor t a ntpa r toft hi ss e c t i oni st oe xpl i c a t eQui ne ’ s di s a g r e e me ntwi t hTa r s ki ’ st he or y —the disagreement that set him on a divergent road. Thef o l l owi ngs e c t i ona i msa tc l a r i f y i ngQui ne ’ sa c c ountoft r ut h.Unde rapr ope r characterisation of deflationism, a major task of this section is to roughly provide a basic formulation of disquotationalism. Further, how the truth predicate helps express infinite conjunctions and disjunctions, and how strong the deductive power of the disquotational theory of truth is, are two questions in need of answers. In order to answer these questions, a discussion with regard to the problem of conservativeness of disquotational theories over logic or the base theories forming them, is given in the third section. In addition, in vi e wofbot hQui ne ’ se a r l i e ra ndl a t t e rr e ma r konhi sa c c ountoft r ut h,as pe c i f i c a t i onof whether or not his theory of truth is substantive rather than deflationary will not be discussed until the final section of Chapter Five. Here I must emphasise that I will not discuss disquotationalism in terms of 10 counter-factuality or related notions in this chapter. This is why many accounts of di s quot a t i ona l i s mf ol l owi ngt hi sl i ne ,Fi e l d’ squa s i -disquotationalism, for instance, are not discussed until Chapter Five. Counterfactuals and possible world semantics seem to owe their clarity to a decent theory of truth but not the other way round. So it is probably a good idea not to give an extensional analysis of disquotationalism in an intensional atmosphere, at least not now. 1. Tar s k i ’ sPa r a d i g mo nTr ut h Many philosophers have felt the urgent need to deal with semantic paradoxes, especially the Liar Paradox. Tarski tells us that for any semantic closed language, if the usual laws of logic hold then there will be inconsistent applications in the application of some that 3 l a ng ua g e ’ ss e ma nt i ct e r ms . The formulation of a semantic account of the intuitive c onc e ptoft r ut ht ha te l i mi na t e ss uc hs e ma nt i cpa r a doxe si st hepr i ma r yt a s kofTa r s ki ’ s theory.4 Another way to see it is, according to Field (1972), to reduce the concept of truth to other semantic concepts and then to reduce these concepts to, or to explicate them in terms of, physical or non-semantical terms. Since the purpose of my discussion is to a c c ountf orQui ne ’ si nt e r pr e t a t i o nofTa r s ki ’ spa r a di g m,onl yt heg e ne r a li de aofTa r s ki ’ s theory is in need of explication. Following that, my ve r s i onofQui ne ’ si nt e r pr e t a t i onc a n be addressed. As ke t c hofTar s ki ’ ss e mant i cc onc e pt i onoft r ut h In order to establish a semantic theory of truth without the threat of semantic paradoxes, 3 That is, it can account for its own semantics. A semantic account of truth explicates the intuitive concept of truth in terms of semantic terms and first order standard logic plus set theory. 4 11 Tarski restricts the attribution of truthfulness to sentences of a language whose semantics is to be constructed in another richer language. This is the famous distinction between object- and meta-l a ng ua ge s . Ta r s ki ’ si ns i g hti st ha t ,i fal a ng ua g ec ont a i nsi t sown semantic vocabularies and the usual laws of logic hold in it, the semantic theory of this language that is constructed within the language itself, will be inconsistent. Tarski presupposes that the object language is a set of expressions (of the first order arithmetic, say,) with meanings. Thus if a semantic theory of truth for that given language is to be constructed, a list of the object language containing all primitive expressions, as well as the rules of definition by means of which new expressions are introduced in terms of primitive ones, has to be laid down. A set of formation rules determining which sequences of symbols from its alphabets are sentences among all primitive expressions has to be specified. Meanwhile, the deductive apparatus of this language has to be given by setting down (by fiat) how to infer some sentences from other sentences.5 Further, besides the complete devices of first-order logic and set theory, the metalanguage must include expressions referring to those in the object language. The metalanguage is to be capable of describing structural features and relationships of expressions of the object language, t ha ti s ,f ol l owi ngTa r s ki ’ st e r ms ,t hemor phol ogyoft heobj e c tl a ng ua ge .I tha st oi nc l ude expressions of the object language or their translations as well. According to the above characterisation, object languages and metalanguages are expanded languages and metalanguages in particular are with full logical and set-theoretic machinery. 6 Si nc et heobj e c tl a ng ua gei sa boutr e a l i t y ,as e nt e nc e ,s a y ,“ Snowi swhi t e ” of the object 5 If the object- and the meta-languages are both artificial, the deductive apparatus of the object language has to be given by setting down (by fiat), firstly, that certain sentences of the language are to be axioms and, secondly, the rules of inference by means of which the theorems are deduced from the axioms. 6 Throughout this thesis I use double quotation marks around sentences, and single quotation marks around 12 language is true if and only if snow is white. This relation can be captured by the f ol l owi ngme t a l a ng ua g es e nt e nc e :“ Snowi swhi t e ”i st r uei fa ndonl yi fs nowi swhi t e . This is the truth paradigm—a paradigm embraced by Quine. These paradigmatic sentences (T-se nt e nc e s )s uc ha s“ ‘ Snowi swhi t e ’i st r uei fa ndonl yi fs nowi swhi t e ” i ns t a nt i a t eTa r s ki ’ st r ut hs c he ma( Conve nt i onT) , (T) X is true if and only if p, whe r e‘ X’i sr e pl a c e dbyt hena meofa nys e nt e nc eoft heobj e c tl a ng ua gea nd‘ p’by the sentence itself or by its suitable translation within the metalanguage. Further, Tarski claims that any successful definition of the concept of truth is required to entail every instance of (T). In other words, given the successful definition of truth, we can prove a T-sentence for every sentence of the object language. This is the requirement of material adequacy, by which a definition of truth can be shown to get the extension of the truth predicate right. That is, this requirement can guarantee the defined concept to be the concept of truth rather than, say, the concept of falsity. Later Tarski states this in a looser way: Statements of this form [, the instances of (T),] can be regarded as partial definitions of the concept of truth. They explain in precise way, and in conformity with common usage, the sense of all special expressions of the type: the sentence x is true. (Tarski 1956a, p.404) For Tarski, what is defined in (T) is not a general concept of truth but a particular concept of truth relative to a given language. That is, for every given object language L, there is a unique T-schema: (TL) X is trueL if and only if p, byr e pl a c i ng‘ X’wi t ht hena meofa nys e nt e nc eofL a nd‘ p’wi t ht hes e nt e nc ei t s e l f words and phrases. 13 or its suitable translation within the metalanguage. Moreover, it is Tarski who precisely shows in his formulation of the proper way to define the concept of truth, that there is no significant way to provide a satisfactory account of the general concept of truth. However, even for a particular concept of truth relative to a g i ve nl a ng ua g e ,t heme t a l a ng ua g es e nt e nc e ss uc ha s“ ‘ Snowi swhi t e ’i st r uei fa ndon l yi f s nowi swhi t e ”donotj oi nt l yc on s t i t ut eac ompl e t ea na l y s i soft hec onc e ptoft r ut h. Th e complication of expanded languages under consideration, such as the possibility of the infinite number of sentences, makes the enumeration of all partial definitions or, in other words, all T-sentences impractical. Although the number of sentences might be infinite, the number of the constituent parts of sentences, that is, names or predicates, is not. This prompts Tarski to further form an analysis of the semantic structure of sentences in terms of the semantic concepts of satisfaction and reference. By defining these two semantic concepts recursively at first, he provides a recursive definition of the semantic concept of truth. With the full set-theoretic machinery, the recursive definitions of satisfaction and reference can be converted into explicit definitions7. Following this, Tarski successfully forms an explicit definition of truth (for a given object language). It is the general idea behind this procedure that renders it successful. The corresponding 7 Briefly, there are normally at least three ways to define something. First, if the number of the terms or the concepts or the like in the definiendum is finite, then it can be defined enumerativly—that is, by providing a complete list of all the objects defined or all the expressions containing the objects defined. Or, if the number is infinite, the definiendum can be defined recursively—that is, by defining the possession of a property by a term x in terms of its possession by all the terms preceding x in a series, together with a definition of its possession by the first term of the series. Finally, all recursive definitions can be turned into explicit ones in terms of set theory and other suitable explanatory resources. Please cf. some instances in the following. 14 satisfaction schema with one free variable is as follows: For all a, a satisfies the sentential function x if and only if p. Thei ns t a nt i a t i onoft hi ss c he mawi t hr e ga r dt ot hes e nt e nt i a lf unc t i on‘ xi swhi t e ’c a nbe put as follows: For all a, a s a t i s f i e st hes e nt e nt i a lf unc t i on‘ xi swhi t e ’i fa ndonl yi fa is white. Utilising the set-theoretic machinery of classes and sequences, the general schema for sentential functions (or open sentences) with more than one free variable can be characterised in a similar fashion as follows: For all infinite sequences of classes f, f satisfies the sentential function x, if and only if p, whe r e‘ x’i st ober e pl a c e dbyt hena meofa nys e nt e nt i a lf unc t i ona nd‘ p’t hef unc t i on i t s e l f( wi t hi t sf r e eva r i a bl e s‘ vi’ ,‘ vj’…r e pl a c e dby‘ fi’ ,‘ fj’…) . Fori ns t a nc e ,t hes e nt e nt i a lf unc t i o nc ons i s t i ngof‘ i st a l l e rt ha n’ ,c ollocated alphabetically with ith and jt hva r i a bl e sonbot hs i de sof‘ i st a l l e rt ha n’ ,i ss a t i s f i e dbyag i ve ns e que nc e if and only if the ith element in the sequence is taller than the jth element in the same sequence. To the extreme extent, any sentence in which all variables are bound by universal quantifier(s) is satisfied by any sequence with no regard to what its elements are; otherwise it is not satisfied by any of them. It is simply all or none. An explicit definition of truth is then given in terms of being satisfied by all sequences: For every close sentence a, a is a true sentence if and only if a is a sentence and for every sequence of objects f, f satisfies a. In terms of the account characterised above, it seems that Tarski successfully reduces the explication of the concept of truth to another semantic concept of satisfaction, specifying 15 the circumstances in which any singly given sentence can be said to be true or false—that is, satisfied by all sequences or else not at all. Whether or not a sentence is satisfied by all sequences is not the task of the semantic accounts of the concepts of truth or satisfaction. I tde pe ndsone mpi r i c a ls c i e nc e .Howe ve r , Ta r s ki ’ st he or yi si nt e r pr e t e di nma nydi f f e r e nt ways. In the following, I will discuss two major criticisms related to proper formulations of disquotational theories of truth respectively. Et c he me ndy’ sc r i t i c i s m andHe c k’ sr e s pons e One of the most enlightening observations comes from Etchemendy (1988). If we follow Ta r s ki ’ spr oc e dur et ode f i nethe concept of truth explicitly, then there is no way our semantic theory has any empirical significance. As a result, the definition does not capture t hei nt u i t i vec onc e ptoft r ut h. Ta r s ki ’ se xpl i c i tde f i ni t i onsofs a t i s f a c t i ona ndr e f e r e nc ea r e enumerative. This does not seem to be an unreasonable strategy by itself since Tarski holds that the number of names and predicates of the object language is finite. However, according to Etchemendy, explicit definitions as such are stipulative. It follows that all T-sentences turn out to be logical truths, or at least necessary truths. This means a T-s e nt e nc e ,s a y ,“ ‘ Snowi swhi t e ’i st r uei fa ndonl yi fs nowi swhi t e ”doe snotha ve empirical content.8 The definition does not give us any idea of the empirical content that the concept of truth in question might have. But explicit definitions of semantic concepts are indispensable for Tarski in order to guarantee the consistency of the semantic theory constructed in the metalanguage. Therefore, doing empirical semantics and providing a consistent semantic theory are not possible to achieve at the same time. 8 Putnam, Soames and many others also hold a similar view. 16 In order to do empirical semantics, we should rather take the concept of truth as a primitive, and formulate an axiomatic theory of truth for a given language. Every clause of the recursive definitions of satisfaction, reference, and truth, are turned into axioms, providing an extra axiom that the truth predicate of the given language, say, L—that is, trueL—holds for all and only true sentences of L.9 This axiomatic theory of truth also satisfies the requirement of material adequacy. The difference is that T-sentences are no longer logical or necessary truths because they are only derivable from those axioms, and these axioms in turn are not without empirical content .Da vi ds ona r g ue di n“ TheSt r uc t ur e a ndCont e ntofTr ut h”( 1990)t ha tTa r s ki ’ sde f i ni t i onsofs a t i s f a c t i ona ndr e f e r e nc ea r enot necessarily stipulative. It depends on our interpretation of his intention. On the other hand, Davidson does agree that we shoul ddr opt hef i na ls t e pofTa r s ki ’ spr oc e dur eoff or mi ng e xpl i c i tde f i ni t i ons ,be c a us eTa r s ki ’ sde f i ni t i onsonl ye ns ur et ha tt hee xt e ns i onoft het r ut h predicates is right, without further explicating what these predicates have in common; that is, his definitions do not help us specify the meaning.10 Nevertheless, the guarantee of consistency of the target semantic theory must be abandoned if the final step of turning all recursive definitions into explicit ones is dropped. Adding to our axiomatic theory the above-mentioned new axiom that the truth predicate of a given language L, namely, trueL, holds for all and only true sentences of it, is not a threat to consistency, for the axiom gives no new properties to the truth predicate.11 For 9 This extra axiom is very likely to be indispensable for every deflationary theory of truth in one way or a n ot h e r .Fori ns t a n c e ,Hor wi c hc on c e de st h a t ,du et oGu pt a ’ sobs e r v a t i on ,t h ee x t r aa xi om“ ( x) (x i st r u e→ x is a proposition) is needed (Horwich 1998b, p. 23, note 7). This is implicitly acknowledged hereafter and will not be brought up again. 10 It is also clear why Davidson prefers to see the concept of truth as explanatorily prior to the concept of r e f e r e n c e ,be c a u s eh eh a sn oobj e c t i on st oQu i n e ’ si ns c r u t a bi l i t yofr e f e r e n c et h e s i s . Th i st he s i s ,ont h eot h e r h a n d,i son eoft h ema j orr e a s on sf orQui n e ’ sde pa r t i n gf r omTa r s k i ’ si na c c ou n t i ngf ort hec on c e ptoft r u t h . 11 However, according to Davidson ,a ddi n gt h i sn e wa xi omi sa l s opoi n t l e s s ,“ Forwec a nj u s ta swe l l r e g a r dTa r s k i ’ st r u t hpr e di c a t e‘ s is trueL’a sh a v i n gt h epr ope r t i e sofou rr e a l -l i f epr e di c a t e‘ s is true in L’ ,a s 17 Heck, on the other hand, there is something more. He further holds that there are parallel axioms of satisfaction and reference. Heck calls these three new axioms the connecting principles. These connecting principles reflect the requirement of material adequacy, and the requirement that all the truth predicates of various successful definitions of the concept of truth relative to a given language, are applicable to the same set of sentences (Heck 1997, p. 541). Etchemendy credits Tarski with the techniques he uses in his definition; that is, Getting from a Tarskian definition of truth to a substantive account of the semantic properties of the object language may involve as little as the reintroduction of a primitive notion of truth. (Etchemendy 1988, pp.59-60) [I]f we define a set TRUE using the standard recursive definition, then the claim that all and only true sentences of the language are members of TRUE is logically equivalent to [the axiomatic theory of truth]. (ibid., p. 59) ForHe c k,t hi sonl ys ugge s t s ,“ a l lt hea xi omsof the axiomatic theory can be proven from t hee xpl i c i tde f i ni t i onoft r ut ha n dt he s ene wa xi oms ,t hec onne c t i ngpr i nc i pl e s ”( He c k 1997, p. 541). For instance, the axiom of truth, true(A) if and only if (A is a sentence and .sat(A, )), can be derived from the explicit definition of truth, ATRUE if and only if ASAT(A, ) and the connecting principles of truth, true(A) if and only if ATRUE and of satisfaction, .sat(A, ) if and only if ASAT(A, ). The consistency of the axiomatic theory is therefore not under threat. Heck points out: l on ga st h os epr ope r t i e sdon otc r e a t ei n c on s i s t e n c i e s ”( Da v i ds on1990, p. 292). 18 Since the connecting principles are little more than restatements of the conditions of material adequacy, it would be at best uncharitable not to credit Tarski with having proved as much. (p. 547). The magic touch is of course the two new axioms added by Heck. He expands the requirement of material adequacy to include the concepts of satisfaction and reference, so that what are defined in the explicit definitions of reference and satisfaction are indeed the intuitive concepts of reference and satisfaction. In addition to this, the distinction between object-and meta-languages allows Heck to guarantee the consistency of the semantic theory brought by the explicit definitions. Meanwhile, the criticism Etchemendy made is r e s ol ve d. Ta r s ki ’ se xpl i c i tde f i ni t i onsa r enots t i pul a t i ve .ForHe c k, What [Tarski] has done is clearly to separate the mathematical from the empirical aspects of semantics: the mathematical part is wholly absorbed into the definition of truth; the empirical, into the claim of material adequacy.12 (p. 543) Butt hi si spr obl e ma t i c .Fi r s tofa l l ,c ons i de rGupt a ’ si ns t a nc e( 2002) ,i fade f i ni t i onof true-in-L a ppe a r st obe“ x is true in L if and only if x is not true in L” ,t he ni ts a t i s f i e st he material adequacy requirement trivially because it also implies all T-sentences due to its inconsistency.13 Or we can give an inconsistent theory of truth with the following biconditional: true(A) if and only if [A is a sentence and (.sat(A, ) & .sat(A, ))]. This theory is also materially adequate. The connecting principle of truth thus can be false. Since the connecting principle of truth is little more than the restatement of material adequacy conditions, it is a material biconditional. This means it can be true when both of its antecedent and consequent are false. Surely that an inconsistent theory plus the 12 Heck thinks what really does the trick of avoiding semantic paradoxes is the distinction between objectand meta-languages. But there is a good reason why appealing to the explicit definitions is necessary, viz., recursive definitions cannot prevent self-referential sentences. 13 Gu pt a ’ si n s t a n c ei sme a n tt or e v e a lt h ef a c tt h a ts omer e s t r i c t i on sh a v et obel a i ddownf orCon v e n t i onT. But it seems to also be relevant to the present discussion. 19 connecting principle, or really, anything, imply another inconsistent theory does not teach us any valuable lesson. The connecting principle of truth is too weak to de r i veHe c k’ s conclusion. Further, Tarski sets up two conditions for a definition of truth to be acceptable, that is, that it is formally correct and materially adequate. To make good sense of the motive for setting up two conditions instead of just the material adequacy condition, implies that there are definitions satisfying only the material adequacy condition. Otherwise, any definition satisfying the condition of material adequacy turns out to be formally correct. OnTa r s ki ’ sde f i ni t i on, however, there may be formally incorrect but materially adequate definitions of truth and they are implied by the explicit definition of truth and the c onne c t i ngpr i nc i pl eoft r ut ha sde s c r i be dbyHe c k.Fori ns t a nc e ,“ x is true if and only if x is true or x is not true, and x”i saf or ma l l yi nc or r e c tde f i ni t i onbuti ti mpl i e sa l l T-s e nt e nc e swi t ht hef or m“ x is true if and only if x” ,whe r e‘ x’r e pr e s e nt st hena meofa g i ve ns e nt e nc ea nd‘ x’t hes e nt e nc ei t s e l f . For Gupta (2002), it turns out that Convention T is not even a satisfactory standard for evaluating the extensional adequacy of the definitions of truth, if the extensional a de qu a c yoft hede f i ni t i onsoft r u t ha mount st ot he i rma t e r i a la de qua c y .Gupt a ’ sa r g ume nt rests on a very minimal assumption that, for any truth definition D for a language L, there is at least a true sentence p of L, s a y ,“ J ohnHowa r di st hepr i memi ni s t e rofAus t r a l i ai n 2004, ”t ha ti snotl og i c a l l yi mpl i e dbyt hene c e s s a r ya ppa r a t us ,c a l l e dAD, needed by D to derive consequenc e s .Mor e ove r ,‘ t r uei nL’doe snotoc c uri nbot hp and AD. Suppose D, “ Fora l lobj e c t sz, z is true in L if and only if (z) ”( Whe r e(z) means that z is true of ), 20 is an extensionally adequate definition of truth. Then D’ ,“ Fora l lobj e c t sz, z is true in L if and only if (if p then (z) ) ”i sa l s oa ne xt e ns i ona l l ya de qua t ede f i ni t i on.I tf ol l ows , however, that D’in conjunction with AD fails to imply all T-sentences of L. Suppose, for the sake of reductio, D’and AD implied all T-sentences, we can infer that they imply the following: not every sentence of L is true. But definition D’is particular in the way that it, AD, and not-p logically imply that every sentence of L is true. This suggests that they are logically inconsistent. Hence, D’and AD logically imply p. Since AD does not imply p, there is at least one interpretation on which all members of AD are true but p is false. If wea s s i g nt hes e tofa l lobj e c t st o‘ t r uei nL’ ,t he nwee s t a bl i s ha ni nt e r pr e t a t i ononwhi c h D’and all the members of AD are true but p is false. This means D’and AD do not logically imply p. But this contradicts the previous conclusion that D’and AD logically imply p. So the reductio completes. Material adequacy is not sufficient for an account of truth.14 I agree with Davi ds on,howe ve r ,t ha tTa r s ki ’ st he or ydoe snote nl i g ht e nusa boutt he meaning of the concept of truth, so that the final step that converts the recursive de f i ni t i onsi nt oe xpl i c i to ne sha st obedr oppe d.Mor e ove r ,i ti snott ha tEt c he me ndy ’ s analysis is pref e r a bl et oHe c k’ s , b utr a t he rt ha ta c c or di ngt oQui ne ’ st he s i sof inscrutability of reference, the concept of truth is more basic than the concepts of reference and satisfaction. The thesis of inscrutable reference teaches us that the reference 14 The current discussion is also relevant in connection with an acceptable formulation of disquotationalism. Disquotationalism in its simplest form takes all the T-sentences as its axioms. The undesirability of this formulation is its lack of deductive power. Some disquotationalists such as Halbach, propose to include all those axioms that are formed by the original clauses of the recursive definitions of satisfaction and reference in the disquotational theory, so as to help derive some important generalisations. If Heck is right, then this sort of axiomatic theory is, absurdly, deducible from the explicit definition of truth with the help of the corresponding connecting principle. 21 scheme of any given language or theory is inscrutable.15 That is to say, there are no empirical pieces of evidence that confirm any particular reference scheme over the others. Among others, a reference scheme is equally good for a given language or a given theory if it can preserve the distribution of the truth-values of all their expressions. If truth set limits on the accounts of reference and satisfaction, it does not seem right to hold them explanatorily more basic. A similar point will be made in the following discussion of Fi e l d’ si mpor t a ntc r i t i c i s mofTa r s ki ’ st he or y . Moreover, in spite of all the clarity and efficiency, Quine disagrees with the approach of explicating the concept of truth in terms of the analysis of the sentential structure of any given sentence. The immediate reason for Tarski to define the concept of truth in accordance with the recursive definition of the concept of satisfaction is the impossibility off or mi ngar e c ur s i vede f i ni t i onofs e nt e nt i a lt r ut hout r i g ht .I n“ Tr ut ha ndDi s quot a t i on” (Quine 1976), Quine demonstrates that constructing the truth definition in the light of the notion of infinite sequence and the recursive definition of satisfaction, is incidental to the apparatus of quantifications and variables employed by Tarski in the meta-language. In a different kind of logic in which the work of quantifiers and variables are managed by constants, it can be shown, Quine argues, that the complication resulting from other semantic concepts is dispensable (p. 309). Fi e l d’ sphys i c al i s tc riticism and the explanatory priority Asaphy s i c a l i s t ,Fi e l dr e g a r dsTa r s ki ’ st he or ya sapr oj e c tt or e duc es e ma nt i c a lt e r mst o non-s e ma nt i c a lorphy s i c a lone s . Thef i r s tr e duc t i ona ppe a r i ngi nTa r s ki ’ st he or yi sf r o m 15 One of the consequences of abandoning the analytic/synthetic sentences distinction is the continuum of language and theory. 22 the semantic concept of truth to the semantic concepts of satisfaction and reference. The second step is to reduce these semantic concepts to non-semantical or physical concepts.16 And it seems that Tarski successfully employs non-s e ma nt i ct e r mss uc ha s‘ s e que nc e ’t o prepare the second reduction.Fi e l d’ sc r i t i c i s mi st ha tTa r s ki ’ st he or yf a i l st opr ovi dea non-t r i vi a ls e c ondr e duc t i on.Hea r g ue si n“ Ta r s ki ’ sThe or yofTr ut h”( Fi e l d1972)t ha t Ta r s ki ’ sa c c ountoft r ut hha snos i g ni f i c a ntdi f f e r e nc ei ne xpl i c a t i onf r om as i mi l a r account in which the reference scheme of the object language is in terms of primitive denotation. Tha ti s ,“ e ve r yna meprimitively denotes what it denotes; every predicate and every function symbol primitively denotes what it applies to or is fulfilled by; and no complex expr e s s i onpr i mi t i ve l yde not e sa ny t hi ng ”( p.6) .Fi e l d’ spoi nti st ha tt hi s indifference makes the second reduction trivial.17 As a physicalist, it seems natural for Field to criticise on the ground that a non-trivial reduction is necessary for a successful physicalist account of truth. (Notice though, that the same criticism could be made by a non-physicalist who disagrees with the physicalist account.) This criticism helps to make sense of why Field concludes in part that a failure of this second reduction suggests the abandonment of the concepts of truth and reference. For such concepts are not, on this account, qualified according to the criteria set up by physicalism (p. 25). In the “ Pos t s c r i pt ”( 2001)of“ Ta r s ki ’ sThe or yofTr ut h” ,Fi e l ds t i l lma i nt a i nsa refined version of his criticism. He does, however, concede at the end that given his interpretation of Qui ne ’ sa c c ount ,s houl di tt ur noutt obec or r e c t ,t he nFi e l d’ sownphy s i c a l i s ta c c ountof t r ut hi smi s t a ke n.Fori fQui ne ’ sa c c ounti sc or r e c t ,t he nthe truth predicate is just a device for expressing generalisations—that is, the truth predicate has its special merit as 16 Field makes it clear in his paper that it is not necessary to require a reduction in a strict sense. An approximate reduction will do. 17 Many philosophers arrive at different conclusion. Devitt, for instance, holds that the list-like definitions are defla t i on a r y .“ Ta r s ki ’ sde f i n i t i ont e l l su sal ota bou t‘ t r u e -in-L’ , ”De vi t ts a y s ,“ I tt e l l susn ot hi nga bou t truth-in-L be c a u s ei ti si mpl i c i t l yc ommi t t e dt ot h ev i e wt h a tt h e r ei sn ot h i n gt ot e l l . ”( 2001) 23 abbreviating infinite conjunctions or disjunctions of sentences.18 According to Romanos (1983), however, this is not the only problem t ha tFi e l d’ st he or y f a c e s . Ta r s ki ’ ss e ma nt i ca c c ountoft r ut hr e s ul t sf r oma t t r i but i ng ,orf a i l i ngt oa t t r i but e , truthfulness to sentences, in terms of the analysis of the sentential structure of the object language sentences as well as the list-like defini t i o nsofr e f e r e nc ea nds a t i s f a c t i on.Fi e l d’ s criticism rests on the point that Tarski fails to provide a satisfactory explanation of the second aspect of this process. Romanos correctly observes that the consequence of this failure—that is, the abandonment of the concepts of truth and reference because of their not being qualified according to the criteria set up by physicalism—is acceptable only if the original assumption that the analysis of the referential relation plays a decisive role in explicating the concept of truth, is itself without a shred of doubt (p. 170). But if one contests that the concept of truth should not be explicated in terms of the referential r e l a t i on,t he nFi e l d’ sc r i t i c i s mi snol ong e rde c i s i ve . Romanos continues his criticism by citing Field in order to show that Field employs Qui ne ’ st he s e sofi ns c r ut a bi l i t yofr e f e r e nc e ,unde r -determination of theories, and ontological relativity, in order to support the view that gluing our conceptual scheme onto reality from the outside is impossible (Romanos 1983, p. 168; Field 1972, p. 24). For Quine, these theses seem to further entail the implausibility of asking for causal theories of reference to nail language to reality (Quine 1960, p. 24). But according to Field the causal connection between language and (extralinguistic) reality is still required in order to avoid our conceptual scheme breaking down from the inside (Field 1972, p. 24). It 18 Please see the following section for a more detailed elaboration. 24 seems to Romanos that Field is in danger of being inconsistent. However, Field in fact disagrees wi t hQui ne .I ti snotc l e a rwi t hwhi c ht he s i sofQui ne ’ sFi e l ddi s a g r e e swi t hi n Field 1972. Is it the implausibility of asking for causal theories of reference or is it one of the theses mentioned above? He does agree that our reflection of the accounts of primitive reference and truth is not explanatorily prior to our scientific investigation. I n“ Qu i nea ndt heCor r e s ponde n c eThe or y ”( Fi e l d1974) ,wr i t t e nt woy e a r sl a t e r ,Fi e l d de ni e sQui ne ’ st he s i sofont ol og i c a lr e l a t i vi t ya ndpr opos e sa na c c ountofpa rtial signification to remove the obstacle set up by the thesis of inscrutability of reference.19 If t hi spos i t i oni se s t a bl i s he d,Fi e l d’ sa t t e mptt of i ndac a us a lt he or yofr e f e r e nc ei snott ha t puz z l i ng .I nt hene xtc ha pt e r ,i twi l lbes hownt ha tFi e l d’ scriticism of the thesis of ont ol og i c a lr e l a t i vi t yi sac ons e q u e nc eofhi smi s i nt e r pr e t i ngQui ne ’ sont ol og i c a l relativity thesis. This discussion centres around which concept is explanatorily more basic, reference or truth. It is clear why Quine takes the concept of truth to be more basic to the concepts of satisfaction and reference. Davidson pushes this point slightly further. Normally our understanding of sentences of a given language, presumably infinite in number, is dependent on our prior understanding of the words, finite in number, that form these sentences. Thus the semantic properties of words seem to be conceptually prior to those of sentences because the latter owes their comprehensibility to the former. According to Davidson, it is a mistake to draw such a conclusion about the conceptual priority of the 19 Although there is no empirical ground to choose one translation manual over another and thus relativise the reference scheme of a given theory or language to another, for Field, it can be held that a general term x, for instance, of a given language partially signifies the sets of objects of all competing manuals. A simplified example can be put like this: the term gavagai partially signifies the set of rabbits (relative to a translation manual M1) and the set of undetached rabbit parts (relative to M2) but fully signifies neither. 25 semantic properties of words over sentences, from a platitude about how we can possibly understand all sentences. That is, This mistake is to confuse the order of explanation of why the theory is correct. The theory is correct because it yields correct T-sentences; its correctness is tested against our grasp of the concept of truth as applied to sentences. Since T-sentences say nothing whatever about reference, satisfaction, or expressions that are not sentences, the test of the correctness of the theory is independent of intuitions concerning these concepts. Once we have the theory, though we can explain the truth of sentences on the basis of their structure and the semantic properties of the parts. (Davidson 1990, p. 300) Quine and Davidson, of course, do not preclude there being a role played by a theory of reference in the explication of the truth of sentences. Still, the difference between the reference schemes suggests a difference between the translation manuals, and one of the reasons for selecting one reference scheme over the other is its capacity for preserving the distribution of true and false sentences in the interpreted language. Truth preserving sets limits on the selection of the domains of discourse. To regard the theory of the referential relation between the constituents of sentences and the reality as explanatorily basic is nothing but a misconception. Addi t i ona l l y , Ta r s ki ’ ss e ma nt i ca na l y s i si se i t he ri nc ons i s t e ntwi t h,ora tleast, is a peculiar c ompa ni onf orQui ne ’ st he s i sofont ol og i c a lr e l a t i vi t y . Thet he s i si s ,puts i mpl y : What makes sense is to say not what the objects of a theory are, absolutely speaking, but how one theory of objects is interpretable or reinterpretable in another. (Quine 1969, p. 50) To construct an account of sentence truth in terms of satisfaction conditions is to base the analysis of the truth of a sentence on a function of the referents of its component terms (Romanos 1983, p.145). This view raises concerns about the relativity of ontology in a sense that a primitive concept of truth does not (ibid.). If the explicit definitions of 26 satisfaction and reference are stipulative, some concerns with regard to the relativity of ontology have to be raised. For these stipulative definitions do offer a way to talk about what the objects are absolutely.20 Romanos puts this well: I ft r u t hde f i n e dbyTa r s k i ’ spr oc e du r ei se s s e n t i a l l yaf un c t i onoft h ef e a t u r e sofag i v e n t h e or y ’ sun i v e r s eofdi s c ou r s e —the objects comprising its ontology—then how can our u n de r s t a n di ngof‘ t r u e ’a sa ppl i e dt os e n t e n c e soft h et h e or ybea nyl e s sr e l a t i v et h a nou r de s c r i pt i onoft h et h e or y ’ sun i v e r s e .( i bi d. ,p. 159) 2. Fo r mul a t i ngQui ne ’ sDi s quot at i ona lThe o r y Intuitively speaking, to say that a sentence is true is to say that it has a certain relation to reality, and that it is in virtue of this relation that it is true. This is the correspondence theory of truth in a very loose sense. What seems to be widely accepted by many philosophers is that the correspondence theory of truth is about the correspondence relation between the meaning of a given sentence, or the proposition expressed by it, and some fact of the matters or state of affairs. Still, for the correspondence theory to be more philosophically interesting it must be supported by a theory of meaning strong enough to allow two sentences, whether belonging to the same language or not, to bear some synonymous or equivalent relation. It is absurd to claim that the basic unit of meaning is not sentences and at the same time to hold that the uncritically accepted synonymous relations between sentences from different languages make sense.21 It is clear then that the advantage of the correspondence theory of truth, under the normal understanding of it, 20 Wor s e ,t h e r ei si n de e dar e f e r e n c es c h e mef ore v e r yl a ng u a g ei na na bs ol u t es e ns ei fwea c c e ptHe c k’ s interpretation of the materially adequate conditions that includes the consideration of reference scheme in the requirement of material adequacy. 21 Eventually the consequent will strike back to undermine this very assumption. To hold that there is a genuine discrepancy between the synonymy relations for a language and for different languages or between sentences of mine and sentences belonging to a language that is other than my mother tongue, is indeed to object to the public feature of languages—that is, the promotion of the private language account. Worse, this assumption has to offer some sort of explanation to demarcate the learning processes of my mother tongue and the second language, which is definitely not an easy job. 27 lies in its capacity to provide a satisfactory explication of how sentences from different languages which share the same truth value, can also share the same meaning. This explication depends heavily on the foggy distinction between facts or the like, and how sentences articulate them. According to this account what correspond to reality are of course not sentences themselves but their meanings. Thus, for instance, to say that a s e nt e nc e“ Snowi swhi t e ”i st r uei st os a yt ha ti t sme a ni ng that snow is white corresponds to the fact that snow is white. Quine thinks that it is here that we encounter the inexplicable part of the correspondence theory. To what extent can the notion of sentential meaning, the resultant notion of a relation of synonymy between sentences, and the notion of a fact of the matter, be clarified? One might propose to use the notion of a proposition to explicate the notion of s e nt e nt i a lme a ni ng . Thes e nt e nt i a lme a ni ngof“ Snowi swhi t e , ”f ori ns t a nc e ,i st he propositiont ha ts nowi swhi t e .Si mi l a r l y ,t heme a ni ngof“ De rSc hne ei s twe i s s ”i sa l s o this very same proposition. Truthfulness should be attributed to propositions instead of sentences. Therefore, sentences themselves become strings of physical signs of different notations in which propositions, their meanings, are couched. The intuitive account of the correspondence theory in the last paragraph suggests a two-step analysis of the notion of t r ut h. Tha ti s ,t os a yt ha tas e nt e nc e“ Snowi swhi t e ”i st r ue ,i st os a ythat its meaning that snow is white corresponds to the fact that snow is white. The first step is from sentences to their meanings and the second step is from their meanings to the facts to which they correspond. Utilising the notion of a proposition, we no longer seem to recourse to the first step in this process. The problem with this account seems to be that it is open to Qui ne ’ sc r i t i c i s moft hemy t hoft hemus e um.Fori ts e e mst ha tpr opos i t i onsa r e 28 uncritically accepted as exhibits, and sentences from different languages are their labels. More importantly, accepting propositions as abstract entities introduces the problem of how to individuate such propositions.22 Thes e nt e nc e“ Snowi swhi t e ”doe snot individuate the proposition that snow is white because it is just one of (infinitely) many notations of this proposition. Each notation of the proposition is nothing but a string of physical signs. The problem of individuation leads us back to the original consideration of equivalent or synonymous relations between sentences.23 Qui ne ’ sma i nc r i t i c i s m ont hi s a ppr oa c hi sa sf ol l ows :“ [ T] hea p p r opr i a t ee qui va l e nc er e l a t i onma ke snoobj e c t i ves e ns e a tt hel e ve lofs e nt e nc e s ”( Qui ne1 970,p.3) . Thi sc r i t i c i s m wi l lbedi s c us s e di nmor e detail later. As described above, an adequate account of the notions of sentential meaning and synonymous or equivalence relations between sentences is vital to a successful e xpl i c a t i onoft hec or r e s ponde nc et he or yoft r ut h.Qui ne ’ sc r i t i c i s m,i fe s t a bl i s he d,ha sa severe impact upon any such explication. It does not just show that the correspondence theory fails to provide an acceptable account of equivalence or synonymous relations. It indicates the dim chances of getting a plausible account of sentential meaning. It is not compelling to both hold that the equivalence or synonymous relations do not make objective sense at the sentential level, whilst also holding that the notion of meaning does make objective sense at the level of sentences. Consequently, the correspondence theory of truth loses ground. 22 Hor wi c h’ smi ni ma lt h e or yoft r u t ht a k e spr opos i t i on sa st h epr i ma r yt r u t hbe a r e r s .I nCh a pt e rFi v e , where the evaluation of this theory i sg i v e n ,h owe v e r ,Qu i n e ’ sobj e c t i ont ot h en ot i onofapr op os i t i ondoe s not play any essential role. 23 This, on the other hand, seems to shed some doubt on the rationale of introducing the notion of a proposition to play such an essential role in the explication of the notions of truth and sentential meaning. 29 It is rather puzzling why Tarski would declare that his theory is a kind of correspondence theory of truth, despite the fact that for Tarski there is no correspondence relation between the meaning of sentences on the one hand, and facts or states of affairs on the other. Da vi ds onma dei tc l e a ri n“ TheFol l yofTr y i ngt oDe f i neTr ut h”( Da vi ds on1996) : But facts or states of affairs have never been shown to play a useful role in semantics, and on eoft h es t r on g e s ta r g ume n t sf orTa r s k i ’ s definitions is that in them nothing plays the r ol eoff a c t sors t a t e sofa f f a i r s …Ta r s k i ’ st r ut hde f i n i t i onsma k en ous eoft h ei de at h a ta s e n t e n c e‘ c or r e s pon ds ’t oa ny t h i nga ta l l .( p.267) Indeed, the material adequacy condition ensures that every definition of truth satisfying this requirement is successful. But no part of this analysis has anything to do with corresponding to facts or states of affairs. The analysis of the sentential structure of any given sentence does not even appear in the formulation of this requirement. Moreover, if onec o n s t r ue sTa r s ki ’ spos i t i ona st a ki ngt hes a t i s f a c t i onf unc t i ont opl a yt her ol eoft h e so-called correspondence relation, just as Davidson admits that he once held this position24, then the correspondence theory is, as Davidson later puts it25, thoroughly uni nt e l l i g i bl e . Ac c or di ngt oTa r s ki ’ sde f i ni t i on,e ve r yc l os e ds e nt e nc ei st r uei fa ndon l yi f it satisfies every sequence. It follows that every true sentence corresponds to exactly the same thing, the big fact. This makes the Tarskian conception of the correspondence theory useless. Ani nt r oduc t i onofQui ne ’ sr e as oni ngondi s quot at i onalc onc e pt i onoft r ut h The intuitive correspondence theory does seem right in the sense that truth should hinge on reality rather than language. The employment of the notion of a proposition is one prominent way to account for this point. Quine agrees that the truthfulness of sentences 24 25 Davidson 1990, p. 302, foot note 36. Ibid., p. 298. Also Davidson, 1996, p. 268. 30 depends on reality to make them so (Quine 1970, p. 10). For the truth of singly given sentence he states: In speaking of the truth of a given sentence there is only indirection; we do better simply to say the sentence and so speak not about language but about the world. So long as we are speaking only of the truth of singly given sentence, the perfect theory of truth is what Wilfrid Sellars has called a disappearance theory of truth. 26 (ibid., p. 11) According to the correspondence theory of truth, in speaking of the truthfulness of a given sentence we are saying that the meaning of the sentence corresponds to a fact, a state of affairs, or the like. Even putting aside the fact that the concept of a correspondence relation is unclear, it will be difficult to specify the notions of sentential meanings and facts or the like without utilising any linguistic framework. The only refuge for the correspondence theory is the direct relation between sentences and experience, at l e a s ti nt hebr oa ds e ns e .Qui ne ’ sc r i t i c i s mont hi sma t t e rs howst ha tane wwa yof thinking about the notion of truth is required. And Quine ’ spr opos a lont het r ut hf ul ne s sof a nys i ng l yg i ve ns e nt e nc ei ss i mi l a rt oSe l l a r s ’di s a ppe a r a nc et he or yoft r ut h. Thepa s s a ge c i t e da b ovei spr oba bl yt hemos ti mpor t a ntr e a s onwhyma nyi nt e r pr e tQui ne ’ s disquotationalism as well as most disquotational theories of truth, as no different to the redundancy theory of truth in the cases of singly given sentences. This will be discussed in Chapter Five. However, this passage is a bit misleading anyway. This is not the whole story though. Quine is not an advocate of the redundancy theory of truth. His position, on the contrary, reveals the unsatisfactoriness of this theory. For Quine, the truth predicate is not redundant and thus has its special function in languages. He 26 Th et r u t hpr e di c a t ei sn e e de df ors e n t e n c e st h a ta r en otgi v e n ,s u c ha s‘ Wh a tSmi t hs a i di st r u e ’( Qui n e 1990, p. 80). In the cases similar to this, the truth predicate is not applied to a quotation but a pronoun or bound variable (ibid.). However, the disquotational schema, which will be formulated in a latter discussion, is still applicable to these contexts through some technical paraphrasing. For instance, Paul Horwich in Truth (p. 21). 31 expresses his point as follows: Truth hin g e sonr e a l i t y … Wh e r et h et r u t hpr e di c a t eh a si t su t i l i t yi sj u s tt h os epl a c e s where, though still concerned with reality, we are impelled by certain technical complications to mention sentences. Here the truth predicate serves, as it were, to point through the sentence to reality; it serves as a reminder that though sentences are mentioned, reality is still the whole point. (Quine 1970, p. 11) Ce r t a i n‘ t e c hni c a lc ompl i c a t i ons ’a r ea boutt hene c e s s a r yg r a mma t i c a lde vi c e sbywhi c h we are allowed to generalise on sentences with the same feature or to generalise on s e nt e nc e si nt hec a s eofbl i nda s c r i pt i ons uc ha s“ Wha tTa r s kis a i donApr i lf i r s t ,1953a t 1 1a . m.i st r ue ” .I ti sa l wa y se a s yt og e ne r a l i s eons e nt e nc e sl i ke“ Qui nei smor t a l , ” “ Ra ms e yi smor t a l , ”a nds oon,t of or m ag e ne r a l i s a t i on“ Al lme na r emor t a l ”wi t houta ny mention of truth. It is in contrast not that simple to obtain a generalisation on sentences s uc ha s“ Snowi swhi t eors nowi snotwhi t e ”a nd“ Gr a s si sg r e e norgr a s si snotg r e e n” without employing some dummy symbol and the truth predicate in the generalising s e nt e nc e“ Eve r ys e nt e nc eoft hef or m‘ p or not p’i st r ue ”( i bi d. ) .I not he rwor ds ,t het r ut h predicate might be a device for abbreviating or expressing the (possible) infinite conjunctions or disjunctions.27 The truth predicate is thus the grammatical device for semantic ascent— we can talk about reality by talking about sentences instead. The r ei sadi f f e r e nc ebe t we e ng e ne r a l i s i ng“ Al lme na r emor t a l ”f r om “ Qui nei smor t a l ” and gen e r a l i s i ng“ Eve r ys e nt e nc eoft hef or m‘ p or not p’i st r ue ”f r om “ Snowi swhi t eor Snowi snotwhi t e ” .I nt hef i r s tc a s ewha ta r ec ha ng e df r om i ns t a nc et oi ns t a nc ea r ena me s . I tc a nt husber e a da s“ x is mortal for all men x”i nwhi c hx is the sort of things that ‘ Qui ne ’i sana meof .I nt hes e c ondc a s ei twoul d,s i mi l a r l y ,ber e a da s“ p or not p for all 27 Although Quine does not say this explicitly, it is compatible with his account. Moreover, not every infinite conjunction is expressible in terms of the truth predicate. This will be briefly discussed in the third section and more completely in the first section of Chapter Five. 32 things p oft hes or tt ha ts e nt e nc e sa r ena me sof ” .Buts e nt e nc e sa r enotna me s .Soi ti s inevitable that we will need to utilise the truth predicate to allow semantic ascent to talk about sentences instead of names. There seems to be another option though. It can be claimed that sentences are the names of propositions. So the target generalising sentence c a nber e a da s“ p or not p for all propositions p” . Apart from all those problems discussed above, propositions in this particular context do not bring any benefit that the truth predicate cannot provide (Quine 1970, pp. 11-12). The special function played by the truth predicate in semantic ascent is unmistakable in the following message: This ascent to a linguistic plane of reference is only a momentary retreat from the world, for the utility of the truth predicate is precisely the cancellation of linguistic reference. The truth predicate is a reminder that, despite a technical ascent to talk of sentences, our eye is on the world. (ibid., p.12) ForQui ne , Ta r s ki ’ spa r a di g ma t i cs c he maona l lT-sentences explicates this characteristic explicitly. There is a major difference between Tarskian and disquotational T-sentences. For Tarski, T-s e nt e nc e si nc l udei n s t a nc e ss uc ha s“ Mor ni ngs t a ri st hes t a rl a s ts e e ni nt he mor ni ng ”i st r uei fa ndonl yi fVe nusi st hes t a rl a s ts e e ni nt hemor ni ng .Fort her i g ht hand side sentence is an extensional equivalent of the quoted sentence. But disquotationalism does not allow instances like this. Henceforth in this chapter and the following, I will discuss T-sentences, or better, disquotation sentences, in this straight disquotational sense. For a disquotation sentence—“ Snowi swhi t e ”i st r uei fa ndonl yi f snow is white—t hequot a t i oni sana meofas e nt e nc ec ont a i ni ngana me .Ca l l i ng“ Sno w i swhi t e ”t r uei sme a ntt oc a l ls nowwhi t e . Thet r ut hpr e di c a t ei sade vi s eofdisquotation. I n“ No t e sont heThe or yofRe f e r e nc e ”( Qui ne1953a) Quine proposes that the theory of truth should be an extra part parasitic to the language for which the theory is constructed. 33 Although Quine does not say it explicitly, the content of the quoted sentence and the sentence itself has to be without difference. It is not necessary for us to specify the (empirical) content of the given sentence as long as we restrict the interpretation or, in other words, the translation manual of the target language and the disquotational theory, such that it is the same. On the other hand, in a case when we are affirming an infinite set of sentences, then talking about reality outright is out of the question. Indeed the truth predicate has its characteristic as a function of semantic ascent to restore objective reference. It is because of this characteristic we can affirm an infinite conjunction by affirming the corresponding generalisation. A preliminary set-up of a proper formulation The first question to consider is whether or not the disquotational theory respects Tarski ’ s distinction between object- and meta-languages. And the answer is no. This, however, seems to be an advantage rather than a flaw. Many disquotationalists, Halbach, for instance, take the aforementioned distinction as a problem rather than a requirement of a s uc c e s s f ult he or yoft r ut h. Ta r s ki ’ spr ooft ha tt hes e ma nt i ct he or yofa nys e ma nt i c a l l y closed language within which the theory is constructed is destined to be inconsistent, only shows that the theory of truth of a target language cannot be constructed within the same language. It does not show that the truth theory has to be constructed in a much stronger meta-language. To resort to a stronger meta-language is not always achievable—it depends on how strong the object language is. In some cases, this might cause a problem. If the disquotational theory of truth is not constructed in a meta-language—that is, a meta-language relative to an object language, then the relation between the target 34 language, call it the base language, and its theory of truth has to be specified.28 Call the language that forms the truth theory the truth language. Then the main difference between the truth language and the base language is that the former has the additional truth predicate.29 There is of course another constraint that we must impose. The base language does not contain any vague, ambiguous, indexical, or self-referential sentences.30 As long as this is a general problem for every deflationary theory, it is not a pa r t i c ul a rpr obl e mf ormyf or mu l a t i onofQui ne ’ sdi s quotational theory.31 Since I do not a doptTa r s ki ’ st he or yofe xpl i c i td e f i ni t i onsoft r ut h,Is houl df ur t he rc ha r a c t e r i s et he standard di s quot a t i ona lt he or ya se mbr a c i ngt hedi s quot a t i ona ls c he ma“ ‘ p’i st r uei fa nd only if p”a sa na xi oms c he ma ,a n da l li t sinstances as axioms. Whether or not there should be some additional axioms is left to the discussion in the following section. It is clear then, that the truth predicate has as its syntactical function cancelling the quotation marks and cashing out exactly the quoted sentence itself.32 Although Quine does not specifically carry out his account in this way, I think there should be no significant disputes over this procedure. However, there is at least one other way to formulate the disquotational theory—namely, a disquotational definition of true sentences. 28 The terminology ‘ base language’i sbor r owe df r omHa l ba c h . Th i si sc on s i s t e n twi t hQu i n e ’ sownwor ds .I nt h es u bs e c t i ondi s c us s i ngQui n e ’ sr e a s on i ngr e garding the di s qu ot a t i on a lc on c e ptoft r u t h,Is h owe dt h a tQui n e ,i nh i spa pe r“ Not e sont h eTh e or yofRe f e r e n c e ” , considered the theory of truth as an extra part, parasitic on the target language. 30 The main task of this dissertation is to first set up a f or mu l a t i onofQu i n e ’ sdi s qu ot a t i on a l i s ma n dabr i e f i n t r odu c t i onofHor wi c h’ smi ni ma l i s m,a n dt h e nt oe x a mi n et h e i rpe r f or ma n c ei na c c ommoda t i n gs ome general issues for deflationary theories. For further discussions of this constraint, please cf. Chapter Five. 31 After all, all deflationists are inclined to see their theories of truth as applicable to sentences of languages that are free from the threat posed by these problems. Besides, it is a bonus if a theory of truth is capable of handling these problems nicely. But this alone is not a good reason to prefer one theory over the others. 32 As indicated before, the quoted and disquoted sentences have the same interpretation. Moreover, the disquotational theory cannot just consist in some axioms instantiating the disquotational schema. This will be further clarified in the following. 29 35 Disquotational definitions of true sentences The disquotational schema cannot be turned into a definition simply because objectual quantifications do not quantify into quotation marks. However, substitution quantifications can. The problem of utilising a substitution quantification in the di s quot a t i ona ls c he ma“ ∏p( ‘ p’i st r uei fa ndonl yi fp) ” ,whe r et hes y mbol‘ ∏’r e pr e s e nt s the universal substitution quantification, is that, normally, a substitution is successful only when the resultant sentence is true. I should spend some extra space to emphasis this, for some philosophers seem to underestimate the consequence of utilising substitution quantifications in explicating truth. Alston, for instance, in A Realist Conception of Truth uses universal substitution quantification in its traditional sense to formulate a general account of propositional truth—that is, ∏p(The proposition that p is true iff p) (p. 28).33 Further, he is fully aware of the criticism of utilising it in defining truth. However, he thinks that this is absolutely fine. The three points he made are as follows: first, universal substitution quantification is only used but not mentioned in the schema, so it is not clear that truth must be mentioned to give someone an understanding of this schema; second, and more fundamental according to Alston, this schema is not meant to be a definition of truth, and hence there is no danger of the kind whereby a definition is invalidated by circularity; third, and equally important according to him, this schema does not aim to be r e duc t i ve ,t or e duc e‘ t r ue ’t ot e r msa tamor ef unda me nt a ll e ve l( p.29) . Wha t Al s t on misses, it seems, is that whether or not this schema is meant to be a definition of truth is not so relevant to the circularity criticism. Even if this schema is not a definition, the truth theory in which it is embedded is nevertheless circular. So one should be more prudent 33 I change the symbol of universal substitution quantifier he used to suit ours. 36 when using substitution quantifications to explain the concept of truth. But if the main purpose of the truth predicate is to abbreviate infinite conjunctions or disjunctions, then substitution quantifications can be taken as explicating this particular feature. In other words, substitution quantifications can be used not to quantify over the l e f tha nds i deoft hedi s quot a t i ona ls c he mabutt her i g htha nds i de .I n“ TheDe f l a t i ona r y Conc e ptofTr ut h”( Fi e l d1986) ,f ori ns t a nc e ,Fi e l ds ugg e s t st ha tt hedi s quot a t i ona l i s t should utilise substitution quantifications in their formulation of truth to capture the (possibly infinite) list of true sentences. The explication of the substitution quantifiers, following this line of reasoning, is as follows: ( SQ)Thenot a t i on‘ ( ∑p) ( … p…) ’e nc ode st hedi s j unc t i onoft hes e nt e nc e s( ope nor closed)t ha tr e s ul tf r om r e pl a c i ngt hef r e eoc c ur r e nc e sof‘ p’i n‘ ( … p…) ’bye; Thenot a t i on‘ ( ∏p) ( … p…) ’e nc ode st hec onj unc t i onoft hes e nt e nc e s( ope nor c l os e d)t ha tr e s ul tf r om r e pl a c i ngt hef r e eoc c ur r e nc e sof‘ p’i n‘ ( … p…) ’bye. (p. 99) If this is acceptable, then it seems possible to attain a disquotational definition of truth. I ns pi r e dbyFi e l d’ sr e a s oni ngons ubs t i t ut i onqua nt i f i c a t i ons ,Da vi di nCorrespondence and Disquotation (David 1994) provides a thorough analysis of the possible candidates of a disquotational definition. Da vi dbe g i nswi t ht hemodi f i e dve r s i onofTa r s ki ’ ss c he ma( T)or ,i not he rwor ds , Convention T—t ha ti s ,‘ p’i st r uei fa ndonl yi fp—and proposes a simplified definition of di s quot a t i ona ls e nt e nc et r ut hbyr e pl a c i ngt he‘ i fa ndonl yi f ’i n( T)byde f i ni t i ons i g n ‘ =DF’ . Thede f i ni t i oni sa sf ol l ows : 37 ( D*)‘ p’i sat r ues e nt e nc e=DF p. (David 1994, p. 64) This simplified version has quite a few problems, such as the difficulty of forming a reading of it in terms of ordinary terms. Among the problems, the most important two are raised by Tarski. (D*) is not a complete definition of truth, according to Tarski, because it c ha r a c t e r i s e st hephr a s e‘ t r ues e nt e nc e ’onl ywi t hi nc e r t a i nt y pe sofc ont e xt . Thei de ai s that the truth predicate is not just designed to attach to quotations of sentences, but to attach in some other contexts to a specific sentence picked out by a definite description, or to sentences with universal or existential generalisations over them. In other words, this definition cannot account for blind ascription or what Quine calls oblique generalisations34. The s ei nc l udes e nt e nc e ss uc ha s :“ Thes e nt e nc ewr i t t e nbyTa r s kii sa t r ues e n t e nc e ” ,“ Eve r ys e nt e nc ei nTa r s ki ’ sbooki sat r ues e nt e nc e , ”or“ Thes e nt e nc e s with thef e a t ur e‘ p or not p’i st r u e ” .( D*)i snoteliminative either, according to Tarski, s i nc eonec a nnote l i mi na t et hephr a s e‘ t r ues e nt e nc e ’f r oma l lc ont e xt si nwhi c hi toc c ur s on the basis of the definition alone.35 David argues correctly that the deficiency of (D*) is that it only provides us conditional information about truth. One knows how to employ the notion of disquotational truth on every given occasion only if one knows (D*) and all other relevant facts that help us sort out which sentences are picked out by which descriptions and which sentences are in the range of which quantificational phrases (p. 72). This is a fatal deficiency for a disquotational definition of truth because it would render the concept of truth substantive. The definition of disquotational sentence truth, then, has to employ substitution quantifications and the identity function to resolve these 34 That is, generalisations on sentences with the same feature. Etchemendy (1988, pp. 53 f.) points out a formal reason for preferring eliminative definitions to noneliminative ones—ar e a s onwhi c hDa v i dt h i nk smus tpl a yar ol ei nTa r s k i ’ sc on s i de r a t i on( Da v i d1994, p. 71). The reason is that an eliminative definition of a term would automatically ensure consistency relative to the theory on the basis of which the term is defined. That is, an eliminative definition cannot lead to paradoxes, unless they are already derivable without the defined term, because this definition allows us to do away with the definiendum in favour of its definiens. 35 38 problems. The major definition according to David is the following: (D) x is a true sentence =DF for some p, x i si de nt i c a lwi t h‘ p’ ,and p (p. 74); or x is a true sentence =DF ( ∑p) ( x =‘ p’&p)( p.89). The corresponding generalisation implied by (D) according to David is (GD): (GD) (x) (x i sat r ues e nt e nc e↔ ( ∑p) ( x =‘ p’&p) ) .( i bi d. ) David then proves that schema (T) is entailed by (GD). Do sentence types as the truth bearers create a modal problem? Aque s t i onf ordi s quot a t i ona lt he or i e soft r ut hi nge ne r a li st owha te xa c t l yi s‘ t r ue ’ applicable? Is it sentence types or sentence tokens? Many, if not most, advocates of disquotationalism incline to choose types over tokens.36 And David thinks that this causes a modal problem for the instances of the disquotational schema. But I have to detour a bit because, unlike others, Quine takes sentence tokens as the primary bearers of truth. As discussed at the beginning of this section, it is untenable to develop accounts of the notion of sentential meaning, or of a fact of the matter, without the involvement of any given linguistic framework. Equally, it is only by reference to visible or audible sentences that we may comprehensibly say which sentence meaning or proposition we have in mind when uttering or writing a sentence (Quine 1990b, p. 77). For Quine, the truth vehicles are eternal sentences—that is, eternalised declarative sentences—in which indexical terms are supplanted by names, addresses or other identifying particulars as needed, and the tenses of sentences are dropped by instead using exact time or related predicates such 36 Except Field. Because of his adherence to physicalism, for him it is sentence tokens to be applied by the truth predicate. 39 a s‘ e a r l i e rt ha n’t of i l li n( p.78) .Sot he s ee t e r na ls e nt ences do not include most utterances in everyday discourse that are deemed to be either true or false, because in general such utterances are not thus refined. The only linguistic expressions which physically exist are those uttered or inscribed. These utterances or inscriptions are tokens of sentences or other linguistic expressions. The issue here is whether eternal sentences, as the vehicles of truth, are taken as tokens or types of sentences. Sentence types are inevitably abstract; if this is not the case there is no way to explain the possibility of linguistic behaviour in the first place. The abstract character of sentence types does not suggest that sentence types amount to the sets of sentence tokens. Otherwise the types of never uttered or inscribed sentence tokens are the empty sets, and thus these types are identical to each other. Still, the abstract character of sentence types should not worry us too much, for this is a general feature of every semantic theory of linguistic behaviour. However, one of the important features of a sentence type is the sameness in the physical appearances of its tokens, regardless of their meaning or meanings. For the one and the same physical string of sounds or signs can be two distinct tokens of different languages, or different evolving stages of one language. Consequently, according to Quine (1970, p. 14), truth or falsity should be ascribed to eternalised sentence tokens rather than types of a particular language at a particular time. But this is not very promising. Sentence tokens have physical existence. It follows that a s e nt e nc ei sne ve rt r uei fi tha sne ve rbe e nut t e r e dori ns c r i be d.Fur t he r ,‘ A’a nd‘ B’mi g ht bebot hut t e r e dori ns c r i be dt r ues e nt e nc et oke ns ,butt he i rc onj unc t i on‘ Aa ndB’mi g ht fail to be at r ues e nt e nc et oke ni fnoonee ve rut t e r e dori ns c r i be di t . Wor s e ,“ Snowi s whi t e ”a ppe a r sont het wos i de sofi t sdi s quot a t i ons e nt e nc e“ ‘ Snowi swhi t e ’i st r uei fa nd 40 onl yi fs nowi swhi t e ”a st wodi s t i nc tt oke ns .He nc ei tdoe snots e e mr i g htt oc l a i mthat we substitute talk of a token for talk of another token used to report or describe reality. So it is sentence types rather than tokens that are the truth bearers. The eternalisation of sentence types can also be dropped as long as we preclude vague, ambiguous, indexical, self-referential, or non-truth-apt sentences as part of the target language. Fixing an interpretation of the target language can help us to set aside the above-mentioned features of sentence types. Recall that according to David, the disquotational schema (T) implied by the disquotational definition of truth (D) has a serious problem with regards to its modal status. All the instances of (T) should be necessary.37 But according to the correspondence theory they are contingent. Consider, for instance: ( 1)‘ Snowi swhi t e ’i st r uei f fs nowi swhi t e . ( 1)i snotne c e s s a r i l yt r ue ,be c a us e‘ s nowi swhi t e ’c oul dha veac ont e ntot he rt ha nt heone t ha ts nowi swhi t e .I f‘ s nowi swhi t e ’ha di t sc ont e nta ss nowi sg r e e na nds nowwe r e indeed green, t he nt hes e nt e nc e‘ s nowi swhi t e ’woul dha vebe e nt r ue ,but( 1)woul dnot ha veb e e nt r ue .Ont heot he rha n d ,i fs nowwe r ewhi t e ,j us tl i kei ta c t ua l l yi s ,but‘ s no wi s whi t e ’ha di t sc ont e nta ss nowi sg r e e n,t he n‘ s nowi swhi t e ’a nd( 1)woul dha vebot hbe en false. The problem is that the disquotational schema cannot get the counterfactuals of its instances right. The gravity of this objection consists in the fact that the contingency of (1) indicates the contingency of the schema (T) and thus the contingency of the definition 37 The plausibility of intensional concepts is sensitive to authentic Quinean philosophers. However, even Qu i n ec on s i de r st h er e l a t i onbe t we e n‘ pi st r u e ’ ,wh e r ep i st h en a meof‘ p’ ,a n d‘ p’s h ou l dbes ome t h i ng very similar to necessary equivalence. He calls it cognitive equivalence and Field follows him. There is no n e e dt oi n v ol v eQui n e ’ sr e j e ction of the concept of necessity here. It will suffice to consider the possibility t h a tt h er e l a t i onbe t we e n‘ pi st r u e ’a n d‘ p’c ou l dbede r i v e df r ombot ht h eus u a lmoda la c c ou n ta n da separate one that might be favoured by Quine. 41 that implies it. The only possibility of deriving the contingency of the schema (T) from a necessary truth, according to the correspondence theorist, is to rest on a contingent premise concerning the contingent relations between sentences and their contents or the like. The correspondence or representation theory of sentence-truth can provide a good explanation of the contingency of the schema (T), whereas the disquotational theory has certain difficulties meeting this challenge. One way to meet the challenge, according to David, is to contest the assumption that every genuine definition has to be necessary. Disquotationalists might argue that the above-mentioned requirement is only plausible when applied to those definitions that attempt to reveal the nature of their definiendum. Thus this requirement is not applicable to the disquotational definition of sentence truth. This, however, cannot be the final answer from the disquotationalist, since she still does not offer any account of the modal status of her definition. If the corresponding generalisation entailed by (D) is just a material biconditional, then the disquotationalist owes us an explanation of why (D) is better or more preferable than the following definition of sentence truth: x is a true sentence =DF ( ∑p) ( x =‘ p’&p&( t he r ea r epi g sors umme r si nAus t r i aa r emuc hwa r me r than in Tucson)) (David 1994, pp. 131-2). The answer is not too difficult to get, of course, for the truth predicate is a device only for abbreviating relevant lists. But (D) still should have an interpretation stronger than just coextensionality. Ac c or di ngt oDa vi d,t hes ol ut i onc a nbef oundi nMc Ge e ’ sVagueness, Truth, and Paradox. Let M be my idiolect that by definition is the set of sentences that are intelligible to me. The reformulation of (1), it can be claimed, is necessarily true: 42 ( 2)‘ Snowi swhi t e ’i st r uei nM i f fs nowi swhi t e . Da vi dt he nc l a i ms :“ [I]n the language that I actually speak,i ti sne c e s s a r yt ha t‘ Snowi s whi t e ’i st r uei f fs nowi swhi t e .Butt hi sdoe snote nt a i lt ha tt he r ei s n’ ta l s oar e a di ngof( 1) unde rwhi c hi ti sc ont i nge nt ”( Da vi d1994,p. 133) .I not he rwor ds ,i fIa c t ua l l ys pe a ka di f f e r e ntl a ng ua g e ,s a y ,N,i tc ou l dbene c e s s a r yt ha t‘ Snowi swhi t e ’i st r uei nNi f fs now is green. The contingency of (T) then rests on the contingent fact that I speak the language that I actually speak. Da vi d’ smos ti mpor t a ntc r i t i c i s mc a nbef oundi nhi sa r g ume nt ,foreign intruder. Here is the quote from David: My idiolect, M, contains a large part of English. If any language different from M c on t a i nsas e nt e n c et h a ti si de n t i c a lt o‘ Sn owi swhi t e ’buti st r u ei f fa l lwh a l e sa r ef i s h , then this sentence will be a foreign intruder. It will come out true, according to (D), even though it is untrue. (David 1994, p. 138) Thef o l l owi ngquot ema ke smor ea ppa r e ntDa vi d’ si nt e r pr e t a t i onofdi s quot a t i ona l i s m.He claims: [ I ] fa nyl a n gu a g edi f f e r e n tf r omM c on t a i n sas e n t e n c et h a ti si de n t i c a lt o‘ Sn owi sg r e e n ’ but is true iff all whales are mammals, then this sentence will also be a foreign intruder. It will come out untrue, according to (D), even though it is true. (ibid., emphasis mine) Many philosophers agree with David on this point. Brendel, for instance, in Brendel 2000 sides with David and views this as the consequence of failing to respect the distinction between object- and meta-languages. Following this, a disquotationalist does not have a c ount e r pa r toft het r a ns l a t i onr e l a t i oni nTa r s ki ’ sConve nt i onT. There are three important features ofDa vi d’ si nt e r pr e t a t i onofdi s quot a t i ona l i s m.Fi r s tof all, what David means by saying that something in quotation is identical to something 43 disquoted, is that they are identical sentence types onl y .Se c ond,i nt e r msofDa vi d’ s understanding of the disquotational schema, the right hand side and the left hand side are disjointed. Sentences on the right side are those in use to report or describe what happens in the world, and those quoted on the left hand side are just sentence types relative to a given language. Third, as a consequence of the first and the second, when the meaning or content of the left hand side sentence varies from one interpretation to another or from one language to another, the meaning or content of the right hand side sentence is fixed because the whole schema or definition is formulated in a particular language. Given t he s et hr e ef e a t ur e s ,t hequot e ds e nt e nc e“ Snowi sg r e e n”i nDa vi d’ sa r g ume nta c t ua l l y has as its meaning or content, that all whales are mammals. So it is actually true, but turns out to be untrue according to the disquotational definition of truth. To make his interpretation of disquotationalism more thorough, David also considers Fi e l d’ sf or mul a t i onoft hedi s quot a t i ona lt he or yoft r ut hi n“ TheDe f l a t i ona r yConc e pt i on ofTr u t h”( Fi e l d1986) .I nt ha ta r t i c l e ,Fi e l ds ugge s t st ha ti ti snotf e a s i bl et oa ppl yt he disquotational truth predicate to sentences that one does not understand. After all, our understanding of the disquotational truth of a given sentence is nothing more than the understanding of that sentence itself. It makes no sense to talk disquotationally about the truthfulness of a sentence that one does not understand. It follows that the truth predicate is not applicable to every language, but only my own idiolect. This leads to the revision of (D): (DM) x is a true M-sentence =DF ( ∑p) M(x =‘ p’& p) .( Da vi d1994,p.141) However, according to David, this newly revised (DM), providing restrictions both on the range of the variable x a ndt hes u b s t i t ut i onc l a s sof‘ p’ ,c a nnotove r c omet hef or e i g n 44 intruder problem. The reason is that (DM) characterises a conjunctive predicate—namely, t hepr e di c a t e‘ x is true and x is a M-s e nt e nc e ’( p.142) . Thi sc onj unc t i vepr e di c a t e indicates that sentences can be true but not M-sentences, and thus generates the gap allowing the foreign intruders to sneak in. The only remaining alternative for construing the disquotationalist position, David suggests, is to interpret the disquotational definition as defining a simple, non-conjunctive truth predicate for the sentences of my idiolect—that is, it is a truth predicate that is undefined for sentences outside my idiolect. The newly revised definition is: (DM) x is trueM =DF ( ∑p) M(x =‘ p ’&p) .( i bi d. ) This should do the trick, because sentences from the intersection of two idiolects, say, M and N in the case of the foreign intruder problem, turn out to be both trueM and not trueN, or vice versa, and this is not a problem at all. David provides several reasons to believe that this newly refined version does no better than its predecessors. Most seriously, the foreign intruder problem does not go away, be c a us ei ti ss t i l lpos s i bl ef ort het r ut hpr e di c a t e‘ t r ue M’ t oa c c ountf ors omeN-sentences, which are not supposed to be accounted for by it. The only way to circumvent the foreign intruder problem, David argues, is to exclude potential foreign intruders entering the overlap area, and do so without invoking any substantive resources such as a representational theory of content. Again, the main issue in the foreign intruder argument is that, without the help of any theory of content, there is no decent way to employ the truth predicate disquotationally. This of course suggests that the disquotationalist cannot account for the strong modal status of schema (T) and its instances. 45 The interpretation and the resultant criticism of disquotationalism made by David are incorrect and misleading. Most astonishingly, if the main function of the truth predicate is to allow semantic ascent—that is, to substitute talk of sentences for talk of reality—then byDav i d’ sde f i ni t i on, whe nones ubs t i t ut e st a l kof ,s a y ,“ Snowi swhi t e ”f ort a l kofs now’ s be i ngwhi t e ,onec a na l wa y ss ubs t i t ut et a l koft hes e nt e nc e“ Snowi swhi t e ”t ha tha s arbitrary c ont e nt ,f ort a l kofs now’ sbe i ngwhi t e .I fDa vi di sr i g ht ,t he ni ti sr e a l l ya mystery why the disquotationalist would have proposed something similar to this in the very first place. It is not just that on this interpretation the disquotational schema of truth does not form an explicit definition of truth, but it is a schema for sentence types that disregard what meaning or content they might have. It is plainly wrong to disjoint the two sides of the biconditionals. The schema has to be formulated relative to a given language with a fixed interpretation manual. That is, (DQ)‘ p’i st r uei f fp, whe r et hef i r s t‘ p’i sr e pl a c e dbya nys e nt e nc eoft heba s el a ng ua g ea ndt hes e c ond ‘ p’byt hes a mes e nt e nc ewi t ht hes a mei nt e r pr e t a t i on. Any sentence that is incidentally identical to a given sentence type does not cause any problem. Speakers of a given language would take it as a sentence of the language they speak. If this creates any trouble, then it is an empirical matter and will eventually be discovered through practical reasoning. Further characterisation of this formulation is best seen from the context of explaining the major features or functions of this theory, and this will not be discussed until Chapter Five. Qui ne ,i n“ Tr ut ha ndDi s quot a t i on”( Qui ne1976) ,g i ve sust hee xa c tr e a s ont obe l i e vet ha t (DQ) is the proper formulation of the disquotational schema. In a different kind of logic, 46 Quine explains, in which the work of quantifiers and variables are managed by constants, the complication resulting from the semantic concepts of reference and satisfaction is dispensable (p. 309). Since the p o i ntofdi s c us s i ngt hi spa r tofQui ne ’ sr e a s oni ngi st o emphasise that the mentioned and used sentences in the instances of (DQ) are exactly the s a mes e nt e nc ewi t ht hes a mei nt e r pr e t a t i on,Iwi l ll e a veoutt het e c hni c a l i t yofQui ne ’ s account. Employing a language with full set-t he or e t i cs t r e ng t ha ndQui ne ’ s predicate-functor logic, which is not stronger than the standard logic of quantification and identity, we can get a good explication of the disquotational schema.38 According to Quine, the definition of truth for a given object language is settled in a metalanguage c ont a i ni ngt hi sobj e c tl a ng ua ge ,t h ede f i ni e ns( ,t ha ti s ,t hede s i g na t i ons i g n‘ Δ’ )a nda notation for naming the expressions of the object language.39 Pa r a phr a s i ngTa r s ki ’ st r ut h schema with a dummys y mbol‘ p’a sf ol l ows : ( Ds )‘ p’i st r uei fa ndonl yi fp, and forming a corresponding designation schema also with a dummy symbol as follows: ( De s )Δ( ‘ p’ )=p, i ti sc l e a rt ha t Δ( ‘ p’ )i se qua lt ot hec l a i mt ha t‘ p’i st r uea c c or di ngt o( Ds )a nd( De s ), whe r et hedummys y mbol‘ p’r e pr e s e nt sas e nt e nc e . Thei nt e r me di a t ec onc e pti nduc t i ve l y defined here is not satisfaction, but designation. This inductive definition is destined to make every equation of (Des) with any one name replacing the dummy symbol, come out true. It turns out that truth is a special case of the designation schema, where again the dummy symbol represents a sentence (pp. 309-310).40 38 According to Quine, th el a n gu a g ee mpl oy e di si nc onn e c t i onwi t hSc h önf i nk e l ’ sc ombi n a t or yl ogi c . I argued previously that the distinction between an object language and a metalanguage is not essential to my formulation of disquotationalism. I maintain this distinction here for the sake of convenience only. 40 In Quine 1995, the disquotational character of the truth predicate can be seen as a special case for the denotation of no-pl a c epr e di c a t e .“ I nt h ez e r oc a s e ,t h e r ebe i n gn ot h i ngf ort h en o-place predicate to be true of, de n ot a t i onr e du c e ss i mpl yt ot r u t hou t r i gh t . ”( p.65) 39 47 This is quite similar to what Quine claimed in Quine 1953a. In that paper, he claimed that the theory ofr e f e r e nc es houl dbec ha r a c t e r i s e da sf ol l ows :‘ p’i st r ue -in-L of every p and not hi nge l s e ;‘ p’na me s -in-L p a n dnot hi nge l s e( i nwhi c h‘ L’i snotaf r e eva r i a bl er a ng i ng ove ra l ll a ng ua ge s ) . Thedi f f e r e n c ebe t we e nQui ne ’ snot i onofdi s quot a t i ona ndTa r s ki ’ si s t ha t ,i nQui ne ’ sa c c ount ,a f t e rt hel a s tme t a l i ng ui s t i cs i g nha sbe e ndi s pos e dof ,j us tt he formula of the object language quoted in the disquotational truth schema (or the de s i g n a t i onone )r e ma i ns .I nTa r s ki ’ sa c c ount ,wha tr e ma i nsi se qui va l e ntto the quoted formula.41 Thi ss ug ge s t st ha tt hes t r onge rdi s quot a t i ona lpr ope r t yunde rQui ne ’ sa c c ount is not preserved under the recursive definition of satisfaction. We should also consider whether or not any version of designation deserves that name, insofar as it fulfils (Des). For instance, unlike the designation accounta tha ndi nwhi c hΔ( ‘ S’ )i sunwoundi nt oS, a not he ra c c ountunwi ndsΔ( ‘ S’ )i n t oS1 because S and S1 happen to be the same object, the same function. The virtue of the disquotational account, Quine argues, is that it e xa c t l yunwi ndsΔ( ‘ S’ )i nt oSa ndΔ( ‘ S1 ’ )i nt oS1 respectively.42 Since Quine is only i nt e r e s t e di na c c ount i ngf ort het r u t hofi nt e r pr e t e dl a ng ua g e s ,t hequot e ds e nt e nc e s‘ S’ a nd‘ S1’a ndt hos edi s quot e da r ee xa c t l yt hes a mes e nt e nc e swi t ht hes a mei nt e r pr e t a t i on. Again, what are unwound are not sentence tokens but sentence types. Otherwise, the whole point of disquotation is missed. 41 Th ewor d“ di s qu ot a t i on ”h e r ei son l yme a n tt obet h epr oc e du r eofr e mov i ngt h equ ot a t i onma r k si nt h e T-s c h e maa n dt h edi s qu ot a t i ona ls c h e ma .Howe v e r ,i ti sa r g ua bl ewh e t h e rorn otTa r s k i ’ st he or yi swith the feature of disquotationalism. For instance, Devitt, Romanos and many others think it is. 42 The usual way to define logical particles depends on truth and falsity. For instance, negation in the logical sense can be defined in terms of two conditionals: <p>i st r u e→ <not p> is false; <p>i sf a l s e→ <not p>i st r u t h( Fol l owi ngHor wi c h ,wedi s t i ngu i s h‘ not’f r om‘ n ot ’ ,whi c hi si ni t sor di na r ys e ns e . )I t seems that, unless there is another method to avoid this, logical particles might cause some trouble of circularity. Nevertheless, Paul Horwich considers a way out in Truth. For instance, negation can be defined not in terms of a truth table, but in terms of a set of theses: p → not not p; not p → not p; p or not p. 48 (D), (DM) and (DM) are of course not in harmony with my formulation of the disquotational theory. Besides mine, there are two objections made by Field and Halbach. Fi e l di nFi e l d1994wa r nsust ha t ,i ft hes or tofi nt e r pr e t a t i onsl i keDa vi d’ sha sge ne r a l applicability, then the truth predicate no longer serves the purpose of abbreviating infinite conjunctions or disjunctions, or expressing blind ascription. Halbach also makes a correct observation in Halbach 1999 and Halbach 2000 on this sort of interpretation, which he c a l l sLe vy ’ sa r g ume nt . Tha ti s ,t he r ea r enone c e s s a r yt r ut hsi nt e r msofs uc ha n i nt e r pr e t a t i on.“ Anys e nt e nc e , ”hec ont i nue st oa r g uei nHa l ba c h2001,“ e ve nat a ut ol ogy , could have meant something else and thus be false; therefore no sentence is necessarily t r ue . ” The r ei sa not he rpr obl e mf orDa vi d’ s( D) ,( DM)a nd( DM). That is, do we have substitution quantifications in ordinary languages? Field, for instance, suggests in Field 1986 as follows: [ I ] f‘ t r u e ’i ss i mpl yade v i c eofi n f i n i t ec on j u n c t i on ,t h e nweh a v eas e r i ou sn e e df ora predicate of truth only because (or, only if)wedon ’ th a v eas u bs t i t u t i onqu a n t i f i e ri n English.43 (p. 58) Grover, Camp, and Belnap might disagree with this assumption, claiming that the substitution quantifier in English is actually expressed by the truth predicate (serving as a ‘ pr os e n t e nc e ’a c c or di ngt ot he i rt e r mi nol ogy ) .Da vi d( 1994)c l a i mst ha twec a nha vea weaker or a stronger position toward this question. The weaker position is that, although the formulation of the disquotational definition of truth is not entirely based on ordinary language, the definition plus the metalinguistic explication of substitution quantifiers 43 Field mentions this in his footnote number 4. However, there are some different opinions regarding the assumption that the sole purpose of the truth predicate is a device of infinite conjunctions or disjunctions—f ori n s t a n c e ,Gupt ai n“ Mi n i ma l i s m” . 49 makes it clear enough to understand the disquotational definition of truth. Besides, if there is a device in an ordinary language that can abbreviate the infinite disjunctions or conjunctions, the truth predicate is dispensable anyway. The reason why the truth predicate still has a role is the very reason why we cannot give an account of truth purely in terms of ordinary terms. The stronger position, on the other hand, is that the metalinguistic explication of substitution quantifications can be managed to become part of our ordinary language. That is, the truth predicate is dispensable as far as we can in principle learn substitution quantifications to abbreviate the infinite list (p. 109). The weaker position, it seems, is more comprehensible than the stronger one since the intuitive understanding of the truth predicate is clearer than the understanding of substitution quantifications. This is because there seems to be no intuitive way that we can formulate substitution quantifications without thedummys y mbol‘ p’ . Toa s s umet ha t the metalinguistic explication of substitution quantifications is learnable in principle and thus that the truth predicate can be replaced by substitution quantifications in an ordinary language seem less promising.44 Moreover, Quine does not accept substitution quantification. Since there is no restriction on the grammatical category of the admissible substituends, the substituted expressions are not just for sentences but for any symbol or sequence of symbols in a sentence as well. In Philosophy of Logic, Quine makes the distinction between objectual quantification and substitution quantification. In standard logic, names are redundant—t ha ti s ,‘ Fa’c a n 44 The issue on which David concentrates is this: How can the disquotational definition of truth be comprehensible in ordinary terms? What concerns me here is the role played by substitution quantifications in the explication of the disquotational theory of truth. On this particular concern, it see mst h a tDa v i d’ s viewpoint also helps make some clarification. 50 45 a l wa y sbepa r a phr a s e di nt o‘ ( x)(Ax & Fx) ’ . So the singular terms contained in objectual quantification are variables only. It is another story for substitution quantification. Names are essential to this non-standard quantification. Consequently, it is deviant when the domain of discourse is rich. For, there are nameless things in the discourse even if we grant it an infinitude of names (p. 92). Furthermore, according to Quine, substitution quantification for sentences is defined in terms of the interchangeability of true substituted expressions, as well as the interchangeability of false expressions, in all contexts salva veritate. Therefore, it is circular to employ substitution quantification in formulating an explicit disquotational definition of truth. For these two reasons, it is not possible for Quine to accept Davi d’ sva r i ousde f i ni t i ons . Ac c or di ngt oGupt a( 1993)Da vi d ’ sde f i ni t i onsa r ea l s onotr i g htbe c a us e“ Thes e nt e nc e wi t ht hef or m‘ φor not φ’i st r ue ”a ndt hei nf i ni t ec onj unc t i oni te xpr e s s e sa r enots e n s e equivalence. We do not need to know the meaning of every disjunct to know that the s e nt e nc ewi t ht hef or m‘ φor not φ’i st r ue .Ne ve r t he l e s st hi ss e e mst obeapr obl e mf o rt he disquotational theories of truth in general. It raises two questions: 1) What does it mean to say that the truth predicate expresses the infinite conjunctions and disjunctions in certain cases? 2) How much deductive power is desirable according to the disquotational theory? 3. Conservativeness and Substantiveness If the disquotational theory consists only of the disquotational schema, (DQ), as axiom schema and all of its instances as axioms, it will be too weak to account for all the facts with regard to the concept of truth of the target language. For instance, it implies all the Orbe t t e r ,wec a nu s e‘ ( x) (Ax) ’a n d‘ ( x)( y)(Ax & Ay & (x y) ) ’t og e t h e rt oi n di c a t et ha tpr e di c a t eAi s only true of one thing (Quine 1970, p.25). 45 51 T-s e nt e nc e soft hei ns t a nc e sof“ Thes e nt e nc ewi t ht hef or m‘ p or not p’i st r ue ”butnot this very sentence itself. Moreover, all the instances of the disquotational schema are held, but not the schema itself. So, it seems, the theory of truth needs to include something else. The reason I choose to focus on the issue of the conservativeness of the disquotational theories, is to show that loosening my previous formulation of disquotationalism to include additional axioms for the purpose of enforcing deductive power, is a risky matter. But first let us turn to question one given at the end of the last section. Namely, what does it mean for a disquotationalist to say that the truth predicate helps express infinite conjunctions or disjunctions in the cases of blind ascriptions or sentences with the same feature, s uc ha s“ Eve r ys e nt e nc ewi t ht hef or m‘ p or not p’i st r ue ” ? Expressibility First of all, not every infinite conjunction or disjunction is expressible with the help of the truth predicate. For instance, the infinite conjunction of sentences that have less than ten words is not expressible in terms of the truth predicate. Although it is most likely false, we can hardly find any reason to see how the truth predicate helps express this generalisation. The reason is clear. The instances of it are not derivable from it. So we may say that the truth predicate only helps express an oblique generalisation over sentences with the same syntactic feature. Therefore, the truth predicate does not express all infinite conjunctions or disjunctions. However, an infinite conjunction of sentences with the same syntactic feature is not necessarily expressible in terms of the truth predicate. All sentences, for instance, with the 52 f e a t ur e‘ i fp or not p,t he np’t og e t he rf or ma ni nf i ni t ec onj unc t i onnote xpr e s s i bl ei nt e r ms of the truth predicate, for the truth-value of every instance of it depends on the truth-value of‘ p’ .I tf ol l owst ha ts omec ont r a di c t or yi ns t a nc e ss uc ha s“ I fs nowi swhi t eors nowi s notwhi t e ,t he ns nowi swhi t e ”a nd“ I fs nowi snotwhi t eors nowi swhi t e , then snow is notwhi t e ”a r ede r i va bl e . The r e f or e ,t hee xpr e s s i bi l i t yofi nf i ni t ec onj unc t i onsor disjunctions in terms of the truth predicate is not subject to the condition of their having the same syntactic feature only, but is also subject to the condition of their instances having consistent truth-value.46 But this second condition is not obtainable because if there is not sufficient deductive power, then an axiomatic disquotational theory in which only instances of the disquotational schema are axioms does not even fix the extension of the truth predicate. Halbach (1999) correctly observes that, if a model-theoretic definition oft r ut hi spr e s uppos e d,t hef or mu l a“ ( x) ( φ( x) →Tx) ”i nde e de xpr e s s e st hei nf i ni t e conjunction over a set of sentences specifiedbyt hef unc t i on‘ φ’ . Tha ti s ,“ ( x)( φ( x) →Tx) ” hol dsi namode li fa ndonl yi ft hei nf i ni t ec onj unc t i on‘ φ( x) ’hol dsi nt hes a memode l (Halbach 1999, p. 7). Without doubt, this solution is not available for the disquotationalist. Another option suggested by Halbach is provability. The second condition to be satisfied can be thus characterised as the following: an infinite conjunction or disjunction is expressible in terms of the truth predicate if and only if the valid or provable translation of it in the truth-theoretic language is provable (Halbach, 1999, p. 9). To be precise, suppose the target language is a first-order language L0 plus a recursively enumerable theory T0 that is formulated in classical logic and is with some deductive power not 46 In Chapter Five, what sorts of oblique generalisations are expressible in terms of the truth predicate is i mpor t a n tt omya r g ume n tbywhi c hGu pt a ’ ss e n s ee qu i v a l e n c ec r i t i c i s mi sr e j e c t e d. 53 exceeding first-order logic. And further suppose a language L1 as L0 together with the disquotational truth predicate, and a disquotational theory of truth T1 as T0 along with the additional T-sentences of L0 as axioms. Hence, an infinite conjunction can be expressed in terms of the truth predicate in T1 if and only if it can be proved in T0. If an infinite c onj unc t i on“ ( x) ( φ( x) ) ”c a nbepr ove di nT0, then (x) ( φ( x) →Tx) can be proved in T1.47 But does saying that an infinite conjunction is provable by all its conjuncts and some inferential rules in T0 not amount to saying that the corresponding generalising sentence is derivable from all its instances and some inferential rules in T1? This is untenable because the disquotational theory T1 consisting of T-sentences as axioms and T0 does not have the deductive power exceeding first-order logic. Tarski (1956b) is probably the first one who notices this. Without the help of some infinitary rule such as ω-rule, the provability of infinite conjunctions in T0 or the derivability of the generalising sentences in T1 is out of our reach. Moreover, it is not sure at this stage that T1 has the capacity to prove that any sentence is not true if its negation is true, or that all propositional tautologies are true. And any infinite conjunction or disjunction can only be expressed by attaching the truth pr e di c a t et ot hec or r e s pondi ngg e ne r a l i s i ngs e nt e nc e .I not he rwor ds ,“ a l ls e nt e nc e swi t h t hes y nt a c t i cf e a t ur e‘ pa ndnotp’a r et r ue , ”f ori ns t a nc e ,i sa l s oal e g i t i ma t ege ne r a l i s i ng s e nt e nc ef ort hei nf i ni t ec onj unc t i on“ ( x) ( φ( x) ) ” . The s ec ons e que nc e si mme di a t e l yf ol l ow when it is shown that T1 does not fix the extension of the truth predicate. But this seems to be rather unbearable. With what sort of additional apparatus can disquotational theories similar to T1 make this disadvantage disappear? This leads us to one of the main topics of this section, namely, conservativeness. 47 Halbach (1999) offers a detailed argument showing t h a t( φ( x) →Tx) is not provable in T1. 54 Conservativeness Shapiro (1998) and several others make a forcible attack on disquotationalism, claiming that it will be inevitable for disquotationalism to imply some substantive consequences. It will be an easy job to prove this if disquotationalists concede to the constraint that all disquotational theories have to be conservative over logic. For a trivial argument offered by Halbach (2001) simply showst ha ti npur ei de nt i t yl og i c ,whe r e“ ( x) (x = x) ”i s pr ova bl ea nd“ ( x) (x ≠x) ”i sr e f ut a bl e ,“ ( x) (x = x)≠( x) (x ≠x) ”a nd“ xy(x ≠y) ”a r e consequences of T1.48 Claiming that the above argument employs the identity axioms to which the truth predicate is already applied is not a shelter for disquotationalists because this makes every newly introduced symbol conservative trivially. No rules or axioms of logic can in principle be extended to the expanded language with any new symbol (p. 179). So, standard disquotational theories such as T1 imply that there are at least two objects. But this is not surprising, for standard disquotational theories already commit to the abstract objects—namely, sentence types—according to my analysis in the last section. At any rate, conservativeness over logic is a formidable Lorelei for disquotationalists. In the following, I turn to another kind of conservativeness, conservativeness over the base theory formulated by the target language. If the base theory is Peano arithmetic, the conservativeness of standard disquotational theories is quite straightforward according to Halbach (2001). The base theory together with T-sentences does not license the derivation of any arithmetical sentence that is not provable in Peano arithmetic. Further, if rules of inference and axioms of first-order 48 T1 is of course presupposed to have the deductive power of pure identity logic here. 55 propositional logic are allowed to apply to sentences with the truth predicate, the standard disquotational theory T1 is capable of fixing the extension of the truth predicate. It can also be proved that a sentence is not true if its negation is true, and that a sentence with the syntactic form of any propositional tautology is true. Here is one demonstration. For a given T-sentence, (1) Tx ↔ x, where x is the name of x, giving that the rule of modus tollens in propositional logic be applicable to sentences with the truth predicate, we can get (2) Tx ↔ x f r om‘ x’ . Re pl a c e‘ x’a nd‘ x’i n( 1)by‘ x’a nd‘ x’ ,wege t (3) Tx ↔ x. From (2) and (3), (4) follows: (4) Tx ↔ Tx. So any given sentence is not true as long as its negation is true. Giving (5): Tx ↔ Tx, (6) Tx Tx follows. And (7) can be derived from (1), (2), and (6): (7) x x. Giving the T-sentence of (7), we get (8) T x x ↔ x x. From (7) and (8): (9) T x x follows. So any sentence with the syntactic form of any propositional tautology can be 56 proved to be true in a similar fashion. Any sentence, it was proved, cannot be both true and not true.49 However, what can be proved are sentences with some tautological syntactic feature but not the corresponding generalisation because of the lack of an infinitary rule of inference. So the applicability of all inferential rules of first order logic to disquotation sentences is the default position of the disquotational theory regarding (DQ).50 Furthermore, the disquotational theory of truth is not just an axiomatisation of truth but a theory about the axiomatisation of truth within which all uses of the truth predicate can be explained by its axioms and other truth-free explanatory resources.51 As stated previously, not all generalisations should be proved by the theory of truth. For s ome ,“ a l lt hec ons e que nc e soft hege ne r a lt he or yofr e l a t i vi t ya r et r ue ”i sa mongt he m. Most importantly, Halbach (2001) correctly makes the claim that some mathematical generalisations should be left not proved by the theory of truth. For instance, if all mathematical generalisations are provable in terms of the disquotational theory and the base theory of Peano arithmetic, substantive consequences such as the consistency of ZF plus the continuum hypothesis follow. For most disquotationalist accounts by which the disquotational theory of truth is characterised as consisting of T-sentences of the target language as axioms and the disquotational schema as an axiom schema only, the generalisations of different sets of sentences with the same syntactic feature is to play a role of reasoning similar to the role 49 Fixing the extension of the disquotational truth predicate seems to be extremely important. Otherwise the disquotational theory fulfils the material adequacy condition trivially by implying all T-sentences of the target language sentences but not the NT-s e n t e n c e s“ Tx ↔ x” . 50 Th i si sa l s oHor wi c h’ sde f a u l tpos i t i on . 51 In Horwich 1998b, Horwich emphasises the methodological point that not all the facts regarding truth are about truth only. Therefore, it is perfectly proper to make use of theories about those matters to explain these sorts of facts (p. 7). I am with Horwich on this point. 57 played by infinite conjunctions or disjunctions. Whether or not these conjunctions or disjunctions are derivable from the truth theory and other suitable explanatory resources is not that important.52 Equa l l y ,whe t he rornot“ Al ls e nt e nc e swi t ht hef or m‘ p or not p’ a r et r ue , ”f ori ns t a nc e ,i sde r i va bl ef rom all its instances is less important. What matters is wha ts or tofs e nt e nc e st he yi mpl y ;i nt hec a s eof“ Al ls e nt e nc e swi t ht hef or m‘ p or not p’ a r et r ue , ”wha ti si mpor t a nti st ha ti ti mp l i e sa l lt hes e nt e nc e s“ or not , ”whe r e‘ ’ stands for a sentence of the target language. If we are inclined to strengthen our disquotational theory, it seems quite natural to add some further inductive axioms similar t ot hos er e s ul t i ngf r omt hea xi oma t i s a t i onofTa r s ki ’ sr e c ur s i vede f i ni t i onofs a t i s f a c t i on. For instance, Halbach claims the following to be one of the inductive axioms: For all closed terms t: Pt is true if and only if Pval(t) (for all one-predicate symbols P and similarly for predicate symbols with more places; val(t) of a term is defined in terms of a reference scheme).53 The exact content of these new axioms depends on the base theory formulated in terms of the target language. Halbach (2001) and Field (1999) are two advocates of this expansion of the disquotational theory of truth. They both try to defend disquotationalism against the attack issued by Shapiro (1998) that the disquotational theories of truth embracing the above-mentioned inductive axioms are not conservative over the base theory. Halbach (2001) concludes, If the deflationist understands his claim that truth is not a substantial notion as implying that his truth theory has no substantial consequences, he commits a mistake. (p. 188) [The deflationist] claimed only that truth serves the purpose of generalisation, and he does well to parse this purpose in a strong way such that his truth theory also proves 52 Please cf. the second section of Chapter Five for a more elaborate discussion. The last clause in the brackets is a paraphrase. Halbach chooses Peano arithmetic as the base theory so that val(t) of a term is defined within it. 53 58 c e r t a i ng e n e r a l i s a t i on s . Th a ti ti n c i de n t a l l ya l s oy i e l dsc e r t a i n‘ s u bs t a n t i a l ’c ons e qu e n c e s is merely a secondary effect. (p. 189) Butbec a r e f ul ,c e r t a i n‘ s ubs t a nt i a l ’c ons e que nc e smight be very unwelcome. In the following I would like to show two reasons for not embracing the additional inductive axioms of reference and satisfaction. One reason relates to the need for an accurate f or mul a t i onofQui ne ’ sdi s quot a t i ona l i s m. Theot he rrelates to disquotationalism in general. It was repeatedly emphasised in the first two sections that Quine could hardly agree with t hee xpl a na t or yor de rofTa r s ki ’ ss e ma nt i ct he or yoft r ut h.Mor epr e c i s e l y ,a nyg i ve n reference scheme is acceptable only if it can preserve the distribution of truth. Hence, the notion of truth is explanatorily more basic to the notion of reference. Additionally, Horwich dissents from the idea of bringing inductive axioms of reference and satisfaction into a deflationary theory of truth, for the following reasons.54 First, he thinks that only the axiom schema, but not inductive axioms, of reference and satisfaction can make sense—that is, they have to be infinite too; second, many kinds of propositions or sentences cannot have their truth explained on the basis of the referents of their parts—for instance, statements of probability or counterfactual conditionals; and most importantly to his minimalism, propositions constructed from the primitive concepts that are not expressed in our language are not explainable (Horwich 1998b, pp. 136-137). The second reason is that my objective regarding the issue of conservativeness and s ubs t a nt i ve ne s si sdi f f e r e ntf r om Fi e l d’ sa ndHa l ba c h’ s .Fi e l dt r i e st of i nda na c c e pt a b l e 54 Th e s er e a s on sa c t u a l l yc omef r omHor wi c h’ sdi s a ppr ov a loft r u t hc on ditional semantics, but it also seems suitable to see them as objections to bringing inductive axioms of reference and satisfaction into the disquotational account of truth. 59 way to embrace the inductive axioms without being accused of inflating the disquotational theory of truth. Halbach, on the other hand, handsomely admits that the revised version does have some substantial consequences. But this does not amount to an inflation of disquotationalism for him. Field and Halbach use Peano arithmetic as the base theory in both of their arguments, for the disquotational theory of truth is destined to fail if it cannot survive in giving an explanation of the truths of mathematical theories. This is fine, but my interests are different. I would like to show that embracing the inductive axioms might cause some problems in accounting for the truths of natural languages.55 I ne vi t a bl yi nHa l ba c ha ndFi e l d’ sr e a s oni ng ,ar e f e r e nc es c he mei si nvol ve di ne xplaining the truths of any given natural language. Even if the inductive axioms do not cause any inflation, introducing a reference scheme into the account might. Take English as the target language, and consider any given belief system formulated in English that is comprehensive enough to include as many as empirical truths as the base theory. Then the reference scheme introduced is not supposed to have any empirical, semantical, and ontological features to lend any substantive flavour to the disquotational theory of truth. Fori ns t a nc e ,t hi sba s et he or ys houl di nc l udes omee mpt yna me ss uc ha s‘ t heKi ngof France Louis XVIII’or‘ phl og i s t o n’ . Tos pa r edi s quot a t i ona l i s munne c e s s a r yi nf l a t i on, substantive theories of reference such as a causal theory have to give way to some de f l a t i ona r yr e f e r e nc et he or i e s . Amongt he m, Ta r s ki ’ si snowor s et ha na nyot he r s . Th e r e are three ways to explain empty names. First, there are no empty names in the reference scheme. That is, the reference scheme does not enlighten us with regard to empty names. 55 Halbach admits in Halbach 2001 that he is taking the advantage of employing Peano arithmetic as the base theory by not being in need of providing any consideration with regard to a selection of the theory of reference. 60 They then disappear from the list of T-sentences. This is untenable because empty names appear essentially in many sentences. Second, empty names behave just like others, as if they refer to something—t ha ti s ,‘ t he King of France Louis XVIII’ ,f ori ns t a nc e , vacuously refers to the King of France Louis XVIII.56 This is also not practical because any name behaving in this way cannot be used in the inductive axioms. Again, the corresponding T-sentences are not implied by this reference scheme and the inductive axioms. Finally, empty names refer to nothing.57 He nc e“ TheKi ngofFr a nc eLoui s XVIII i sba l d”i spr e -determined to be satisfied by no sequences of objects. But this does notg i veust hene e de dpi e c eofi nf or ma t i ont oa l l owust ode t e r mi newhe t he rornot“ The King of France Louis XVIII is bald or the King of France Louis XVIII i snotba l d”i sa n instance of the infinite conjunc t i one xpr e s s e dby“ Al ls e nt e nc e swi t ht hef or m‘ p or not p’ a r et r ue . ”I fnos e que nc e sofobj e c t ss a t i s f y“ TheKi ngofFr a nc eLoui sXVIII i sba l d” pur e l ybe c a us e‘ t heKi ngofFr a nc eLoui sXVIII’i sa ne mpt yna me ,t he n“ TheKi ngof France Louis XVIII i snotba l d”i sa l s onots a t i s f i e dbya nys e que nc eofobj e c t s .‘ The King of France Louis XVIII’i sj us tadr a ma t i cc a s e .Ma nyt he or e t i c a lt e r msf a c eas i mi l a r problem. And these problems of reference should not be the business of any disquotational theory of truth in the very first place. So committing to the inductive axioms causes serious problems to which the standard disquotational theories are not committed. This will be further clarified in the following chapter. Final Remarks Asar e s ul t ,Qui ne ’ sdi s quot a t i ona l i s mi sbe t t e rof fnott oe mbr a c et hei nduc t i vea xi omsof 56 Cf .t h ef ol l owi n gc h a pt e r :“ I n s c r u t a bi l i t yofRe f e r e n c e ,Ont ol og i c a lRe l a t i v i t ya n daDi s qu ot a t i on a l Scheme of Refere n c e ” . 57 If singular terms are paraphrased as one-place predicates, it is important not to consider them to be true of the class of all things that it is true of. Otherwise empty names fallaciously turn out to be coextensive. 61 r e f e r e nc ea nds a t i s f a c t i on.Howe ve r ,Qui ne ’ sdi s quot a t i ona lt he or yi snotj us ta bouta theory of truth but also about a t he or yofr e f e r e nc ei nge ne r a l .‘ Se ma nt i ca s c e nt ’a l s o a ppl i e st os i ng ul a rt e r ms ,g e ne r a lt e r ms ,a ndpr e di c a t e s .He nc et hee xpl i c a t i onofQui n e ’ s di s quot a t i ona l i s mi snotc ompl e t eunt i lQui ne ’ st he or yofr e f e r e nc ei ss ubj e c t e dt o thorough investigation. 62 Chapter III Inscrutability of Reference, Ontological Relativity and a Disquotational Schema of Reference The central issue of this chapter is the role played by a disquotational schema of reference i nQui n e ’ sont ol og i c a lt he or y .I nt e r msof the analysis of this issue, it is also important to see whether or not we should incorporate inductive axioms of reference and satisfaction i nt ot hedi s quot a t i ona lt h e or yoft r ut h. Al t houg hImus ti ne vi t a bl ydi s c us sQui ne ’ s ontological theory, it is not my intention to evaluate its significance and plausibility.1 On the contrary, by concentrating on one important argument against the thesis of ontological relativity, my analysis will illustrate the essential nature of a disquotational schema of reference wi t hi nQui ne ’ sont ol og i c a lt he or y .Ba s e dont hi s ,t hewhol epi c t ur eofQui ne ’ s disquotational theory of truth should be further clarified. To begin with, two related theses of Quine will be briefly introduced, namely, the thesis of inscrutability of reference and the thesis of ontological relativity. The thesis of 1 Fori n s t a n c e ,Fog e l i ni n“ Qu i n e ’ sLi mi t e dNa t u r a l i s m”c r i t i c i s e sQu i n e ’ si n de t e r mi n a c ys e ma nt i c sa s incompatible with his naturalism. Although it seems implausible, I am not tempted to include the related discussion in this chapter. Moreover, Fodor in The Elm and the Expert offers an argument against the thesis ofi ns c r u t a bi l i t yofr e f e r e n c ea n dGe or g a l i si n“ Re f e r e n c eRe ma i n sI ns c r u t a bl e ”t r i e st ot a c k l ei t .Bot hof them seem to show some insufficiency in a proper understanding of the target thesis. This can be exemplified in the foll owi n g .Fodorc l a i ms :“ [ G] i v e nt h eda t aI n f[ or ma nt ]pr ov i de s ,t h el i n gu i s t sk n owt h a t I n fh ol d‘ t h e r e ’ sar a bbi t ’t r u ei nas i t u a t i oni f fh eh ol ds‘ t h e r e ’ sa nu r p[ ,t h a ti s ,un de t a c h e dpr o pe rr a bbi t pa r t , ] ’t r u ei nt h a ts a mes i t u a t i on ’ ”( Fodor199 4,p .63 ) .Bu th owc a nt h ei nf or ma n th ol dt r u ebot h‘ t h e r e ’ sa r a bbi t ’a n d‘ t h e r e ’ sa nu r p’ ?Ge or g a l i st h i n kst h a tFodor ’ sa r g ume n tbe gst h equ e s t i on .Forh i m,Fodor simply presupposes that we have to interpret certain general terms in the standard interpretations of them while the non-standard interpretation of them assumed for the reductio is gone. And this causes a disagreement in the truth-values of certain sentences between linguists. However, this is exactly what the reductio is about. According to the target thesis, linguists will provide incompatible reference schemes even if the totality of empirical truth is preserved. Fodor intends to show, in terms of his argument, that incompatible reference schemes will inevitable cause disagreements in the truth-values of certain sentences between linguists, and Georgalis seems to miss the point. Again, this is not my primary task of this chapter. 63 i ns c r ut a bi l i t yofr e f e r e nc ei sa sf ol l ows :i ft her e f e r e nc eoft het e r msofa ni nf or ma nt ’ s language is not uniquely decidable by a field linguist purely on empirical grounds, then what these schemes say about the reference of the terms of the translated language is therefore not an empirical feature of those terms. That is, the linguist can provide more than one translation manual within which the reference schemes of the terms can be radically different from each other, but still be compatible with all (possible) pieces of empirical evidence. It follows that the choice of one particular reference scheme is arbitrary or, in other words, these terms do not uniquely determinate a reference scheme. What of the thesis of ontological relativity? One simplified version of ontological r e l a t i vi t yc a nbeputi nt hi swa y :“ Wha tma ke ss e ns ei st os a ynotwha tt heobj e c t sofa theory are, absolutely speaking, but how one theory of objects is interpretable or re i nt e r pr e t a bl ei na not he r ”( Qui ne1968,p.50) .I twa she l di nCha pt e rTwot ha tt hema i n reason not to incorporate the inductive axioms with regard to reference and satisfaction i nt oQu i ne ’ sdi s quot a t i ona lt he or yoft r ut hi st hee xpl a na t or ypr i or i t yoft heconcept of truth. This claim, however, is plunged into turbulence when both of the aforementioned theses are under fire. Hartry Field considers the thesis of ontological relativity to be incompatible with the correspondence theory of truth. If a correspondence theory of truth is about the explanation of the notion of truth in terms of the relation between words on the one hand and the extralinguistic objects that they are about on the other, then the only noteworthy correspondence, following the thesis of ontological relativity, is the one between the wor dsofonet he or ya ndt hewor d sofa not he r( Fi e l d1974,p.199) .Fi e l d’ sa mbi t i oni st o show that, granted the thesis of inscrutability of reference, the correspondence theory of 64 truth still prevails (perhaps in a more general sense, however). Donald Davidson accepts the thesis of inscrutability of reference and a relativised concept of ontology, but he a gr e e swi t hFi e l d’ sa r g ume nta ga i ns tQui ne ’ st he s i sofont ol og i c a lr e l a t i vi t y ,c r i t i c i s i ngi t as falling short of providing an intelligible way to justify itself. Roger Gibson, Jr. on the ot he rha nd,c e ns ur e sbot hFi e l da ndDa vi ds on’ smi s i nt e r pr e t a t i onofQui ne ’ st he or i e s .One i mpor t a ntf e a t ur eofhi si nt e r pr e t a t i onofQui ne ’ sont ol og i c a lr e l a t i vi t yt he s i si sthat speakers of every language are normally uncritical of their own languages and thus make available a way to end the ultimate relativisation of ontology by maintaining a word-world correspondence relation between words and extralinguistic objects. This chapter has five sections, in which the following five tasks will be undertaken. First, Iwi l le l uc i da t eQui ne ’ sr e a s oni n gwi t hr e s pe c tt or e l a t i vi s i ngont ol ogy .Se c ond,Iwi l l s howt ha tFi e l d’ sa r g ume nti nf a vourofac or r e s ponde nc et he or yoft r ut hi sinvalid. Third, Iwi l ls howt ha tGi bs on’ si nt e r pr e t a t i onofQui ne ’ sont ol og i c a lr e l a t i vi t yt he s i si si nc or r e c t . Fourth, I will show that a disquotational schema of reference of our mother tongue is essential to enable us to continue common sensically doing science and living daily. Finally, I will show that a disquotational schema of reference does not and cannot provide any support to the reasoning that, in order to redeem the lack of deductive power of the disquotational theory of truth, one should explicate the disquotational conception of truth in terms of the inductive axioms of reference and satisfaction. And more notably all instances of the disquotational schema of reference do not constitute a genuine reference scheme.2 2 Th eph r a s e‘ t h edi s qu ot a t i ona ls c h e maofr e f e r e n c e ’i sus e dt h r ou gh outt h i sc h a pt e ri ni t sg e n e r a ls e ns e . That is, it includes satisfaction as well. Any point made in relation to the disquotational schema of reference is equally applicable to the disquotational schema of satisfaction. This is the default position and will not be emphasised hereafter. 65 1. Qui ne ’ sManoe uv r e In the quoted passage about ontological relativity, the point seems to be that, although it is improper to ask what terms of a theory (or language) refer to in an absolute sense, it does make sense to ask this question in a relativised way. More plainly, the only sensible way to talk about reference is to relativise it to a background language or a background theory. The simplest and probably least questionable manner of explicating ontological relativity mi g htbei na c c or da nc ewi t hQui ne ’ sa c c ountofpr oxyf unc t i on—that is, for any given theory we can provide a function which maps its universe on to the universe of a background theory. This mapping can be a one-to-one or many-to-one relation, depending on our purpose in utilising it. And the universe of a background theory is, if not identical to, then normally broader than the object one. In the case of a many-to-one relation, we use this relation as a means of ontological reduction. According to Quine, it is not necessarily worrying that we must grant the ontological profligacy of the excessive universe of the background theory in order to operate the reduction. This is because the worry is no more pressing than in the case of reductio ad absurdum, where we assume a falsehood so as to disprove it (Quine 1968, p. 58). So the excessive universe is eventually dispensable. In the case of a one-to-one relation, a given theory is always reducible to a background theory, according to Quine, and it can also form an argument in favour of the thesis of inscrutability of reference. Put simply, for any given theory in which a name refers to an object x according to its reference scheme, one can produce a proxy function, say, ø to map x on to ø(x) in the reference scheme of the background theory; similarly, if a predicate refers to (is true of) each thing x such that F(x) in the object theory, it refers to each thing x in the background theory such that Fø(x); along with an appropriate 66 characterisation of a relation like satisfaction in, presumably, each case, the truth conditions assigned by the background theory are equivalent to those assigned by the first theory, thus the inscrutability of reference. However, it seems to be one step further to hold the thesis of ontological relativity in terms of the thesis of inscrutability of reference. If there are neither semantical nor physical facts to determine the choice of reference scheme uniquely, then the only significant way to talk about the reference of any given theory (or language) is to relativise it to another theory (or language) no matter how arbitrary the choice is. The consequence of doing so is the risk of infinite regress from one theory or language to another. For Quine, we end the infinite relativisation of ontology in practice by acquiescing in our own language and taking its words at face va l ue( i bi d. ,p.49) .Le tusc a l lt hi sQui ne ’ sma noe uvr e . Wi t hr e s pe c tt ot hi s ,bot hFi e l d and Davidson hold an opposing position. The proxy function argument is the one that really indicates the mystery of reference, since proxy function can produce more dramatic cases than those in radical translation. Byt hi sIme a nt ha tnor ma l l y ,i nt r a ns l a t i nga ni nf or ma nt ’ sl a ng ua g ewewoul dnotus ea reference scheme that maps itself on to, for instance, the cosmic complement of its objects. In translating a remote language it is more likely to be a limited case in which the mi s s i o noft hef i e l dl i ng ui s ti st of i ndawa yt ounde r s t a ndori nt e r pr e the ri nf or ma nt ’ s language in a more ordinary vein. However, it is different in the case of ontological relativity since it is in a more theoretical vein. That is, in the ordinary vein the field linguist works hard to reduce the amount of competing manuals, while in the theoretical vein it is supposed to show that the interpretation of the ontology of any given language or theory could be varied drastically but that the sentence-to-sentence structure is still 67 preserved. Probably because of this difference, some philosophers such as Field are prompted to assume that providing a solution to the inscrutability of reference in the case of radical translation is sufficient to answer the mystery of reference issued from the thesis of ontological relativity. In a similar sense, perhaps, Gibson does not make a clear distinction between these two cases, either. Fie l d’ sr e a s oni ngi sonl yj us t i f i e dwhe nt he thesis of ontological relativity can be successfully renounced. It will be shown, however, i nt hef ol l owi ngdi s c us s i on,t ha tadi s quot a t i ona li nt e r pr e t a t i onofQui ne ’ sma noe uvr ec a n save him from the criticism, made by Field, that he commits the fallacy of the myth of the museum, whilst also maintaining the thesis of ontological relativity. But let us first shed s omel i g htonFi e l d’ sa r gume nt . 2. Fi e l d’ sAr gume ntag a i ns tOnt ol o gi c a lRe l at i v i t y If the idea of signification3 is to be explained relative to a translation manual, Field argues, then this relativised way of talking about the reference scheme of a given theory or language is useless. There is no difference between saying that a term T used in one language signifies the set of rabbits, relative to a translation manual M, and saying that M translates T a s‘ r a bbi t ’ ,i ft heonl ys e ns i bl ewa yt ot a l ka boutt her e f e r e nc es c he meofa theory or language is to interpret or reinterpret it in another theory or language. More precisely, (1) for every predicate T, set {x | Fx}, and manual M, T signifies {x | Fx} relative to M if and only if M maps T ont o‘ F’( Fi e l d1974,p.203) . This is not satisfactory because we cannot quantify into quotation marks in an objectual fashion. According to Field, we can make some adjustment to avoid this problem by 3 Qu i n eu s e s‘ de n ot a t i on’ . 68 paraphrasing (1) into the following: (2) for every predicate T, set {x | Fx}, and manual M, T signifies {x | Fx} relative to M if and only if M maps T onto some term which signifies {x | Fx} (ibid.). Fi e l d’ sobj e c t i ont o( 2)i st ha tt her e l a t i vi s e de xpl i c a t i onofa nyg i ve nr e f e r e nc es c he me has to rely on unrelativised semantic notions applied to our own language. The assumption that the ontology of our own language, unlike others, is fixed, is what Field cannot accept. He thinks that this makes Quine a victim of the myth of the museum (p. 204). That is, if the ontology of our own language is fixed and the relativisation ends in our mother tongue, all the reference schemes of different theories or languages seem to be fixed eventually by relativising them to our language. It is, however, hard to see the difference between the myth of the museum and this account in which the exhibits are the referents, and referential words of different languages are just like expository labels for different language speaking visitors. Davidson agrees with Field on this point and renders a similar argument (Davidson 1979, pp. 230-1), which will not be discussed. Moreover, Field cannot accept the arbitrariness of the choice between the competing translation manuals due to the fact that there are no semantical or physical facts to show the adequacy of choosing between any of them. Instead, he proposes a reformed account of certain semantical notions. Although there is no empirical ground for choosing one translation manual over another to relativise the reference scheme of a given theory or language to another, it can be held that a general term x of a given language partially signifies the sets of objects of all competing manuals. A simplified example can be put as follows: the term gavagai partially signifies the set of rabbits (relative to a translation manual M1) and the set of undetached rabbit parts (relative to M2) but fully signifies either. 69 The problem for Field is that the indeterminacy of reference scheme occurs hand in hand with the translation of the identity predicate. Generally this is not a problem for Quinean s e ma nt i c sbe c a us et het r a ns l a t i onof‘ ga va g a i ’ ,s a y ,i sf i xe da sf a ra st het r a ns l a t ion of the i de nt i t ypr e di c a t ei s .I not he rwo r ds ,‘ ga va g a i ’c a nbet r a ns l a t e da ss i g ni f y i ngt hes e tof rabbits, if the identity predicate is translated in certain relevant ways. However, this is a di f f i c ul t yf orFi e l d’ sa c c ount ,be c a us eoft her a ndom c ombi nations of the translations of ‘ ga va g a i ’a ndt hei de nt i t ypr e di c a t e . Tos ol vet hi spr obl e m,Fi e l dc l a i mst ha ti ti s inevitable to that some sort of correspondence relations be introduced to define relevance in terms of them. It follows that these causal relations between terms and extralinguistic objects can provide a way for Quine to avoid committing the myth of the museum fallacy. Da vi ds onont heot he rha ndc or r e c t l yc r i t i c i s e sFi e l d’ sr e f or me da c c ountont heg r ounds that although in a standard theory of truths, semantic roles are assigned to the parts of a sentence to explain what makes it true, Field confuses the issue by overemphasising the importance of the independent account of the semantic properties of reference, naming, signification, and so on (Davidson 1979, p. 236). The confusion is the failure to notice the difference between explaining truth, given the theory, and providing evidence that the theory is true of a speaker or community (ibid.). In addition to the assumption that there are causal connections between terms of a language and extralinguistic objects, it can be imagined that these connections do not play any role in the explanation of the use of these terms in sentences. Consequently, these connections are irrelevant to an account of that particular language. Establishing a reference scheme of a given theory or language relies on evidence concerning the adequacy of a theory of truth at the sentential level. 70 Fur t he r ,i ts e e mst ha tFi e l d’ sa c c ountdoe snotpr ovi dear e a s ona bl es ol ut i ont o the alleged failure of the thesis of ontological relativity. This account is disputable in the case of radical translation because the main purpose of it is, after all, to reduce the number of the competing translation manuals.4 It is different in the case of the thesis of inscrutability of reference. According to this thesis, the reference scheme of any given theory or language is in principle i nde t e r mi na t e . Thepr oxyf unc t i ona r g ume nts howst ha tt het e r m‘ r a bbi t ’ , for instance, can be reinterpreted by permuting the universe to map it on to, say, the cosmic complement of rabbits. We can even permute the universe of our language or theory and map it on to itself. It follows that there could be many, if not infinitely many, r e f e r e nc es c he me sofEng l i s h.‘ Ra bbi t ’woul dpe r ha pss i g ni f yma nyt hi ng st ha ty ouc a n imagine as far as the truth conditions assigned by the new reference scheme are equivalent to the old one. It can be claimed that, in some, if not in most cases, some general terms of a given theory or language might be boiled down to partially signifying e xa c t l yt hes a mes e t sofobj e c t s .Fori ns t a nc e ,t heg e ne r a lt e r ms‘ r a bbi t ’a nd‘ unde t a c he d r a bbi tpa r t s ’a nd‘ r a bbi ts t a ge s ’a l lpa r t i a l l ys i g ni f yt hes e tofr a bbi t s ,t hes e tofunde t a c he d rabbit parts, a ndt hes e tofr a bbi ts t a g e s . Thi si snoti nc ompa t i bl ewi t hFi e l d’ sa c c ount . Unless Field can prove that this is in principle impossible, I can hardly see how his account can work and how to establish any causal, correspondence relations between words and objects. 3. OnGi b s on’ sCr i t i c i s m ofFi e l d’ sandDav i d s on’ sAr g ume nt s Whe t he rFi e l d’ sa ndDa vi ds on’ sa r g ume nt sa r ec onc l us i veornot ,t he r ea r ea tl e a s tt wo 4 I nt h e‘ Pos t s c r i pt ’ofFi e l d1974,Fi e l dc l a i ms : Even though Quine was not sympathetic to a causal theory of reference, he did often grant that the stimulus meanings of terms could be used to rule out alternative reference schemes. (p. 219) Still, I think, this is in the case of radical translation rather than ontological relativity. 71 wa y st oa c c ountf orQui ne ’ sma noe uvr e .Onec a nma keadi c hot omybe t we e none ’ s mother tongue and translated languages, and thus hold that the relativisation of ontology only happens in translating (/interpreting) another language. Or, one can find a way to a c c ountf orQui ne ’ sma noe uvr ewi t houtvi ol a t i ngt het he s i sofont ol og i c a lr e l a t i vi t y .Iwi l l discuss these respectively in this and the following sections. Gi bs onha sas t r ongc ount e r vi e wt oFi e l d’ sa c c us a t i ont ha tQui nec ommi t st hemy t hoft he mus e um f a l l a c y .Si nc eDa vi ds ononl yoppos e st hepa r tofFi e l d’ sa r g ume nt a t i oni nf a v our of the correspondenc et he or yoft r ut h,Gi bs on’ sde pr e c a t i onofFi e l d’ spos i t i ona ppl i e s e qua l l yt oDa vi ds on’ svi e w.I nt hef ol l owi ngdi s c us s i onIwi l lf i r s tma kec l e a rGi bs on’ s argument, and then indicate my disagreement with it afterwards. Gibson thinks that Field and Davidson failed to notice in Quine the significant distinction in the thesis of inscrutability of reference and the thesis of ontological relativity in the epistemological vein—in which there is no fact of the matter on the denotation of the terms of a particular language—and in the ontological vein—in which the denotation of t het e r msofone ’ sunc r i t i c a l l ya c c e pt e dl a ng ua g ei sama t t e roff a c t( Gi bs on1988,p.143) . The idea is that we must conduct our epistemological investigation under an ontological framework. And this ontological framework is nothing but our scientific enterprise. There is, however, another problem regarding the choice between competing scientific theories that are empirically equivalent to, but logically incompatible with each other, insofar as t hi smi g htunde r mi neGi bs on’ sa c c ount . Tha ti s ,t het he s i sofunde r -determination is now the problem.5 It is very likely that incompatible theories formulated in the same language 5 The thesis of under-determination in the broadest and probably most appropriate sense is: providing all possible evidence or observation data, there are many, if not infinitely many, theories that are empirically 72 might have different reference schemes. Thus the aftermath of ontological relativity hits Gi bs on’ sont ol og i c a lf r a me wor ka f t e ra l l .I ti snots ur pr i s i ngt ha tGi bs ondoe snott hi nki t this way, since he and Dagfinn Føllesdal would like to persuade Quine to adopt his sectarian position regarding competing comprehensive scienti f i ct he or i e s . Tha ti s ,“ on eof two systems of the world must be deemed false if we know them to be empirically e qui va l e nt ”( Qui ne1986b, p. 156). So we have to pin down our (scientific) theory no matter what the reason is, and then we have the reference scheme in terms of the theory wepi c k .Buti s n’ tt hei de aofont o l og i c a lr e l a t i vi t yt os howt ha t ,f ora nyg i ve nt he or yor language, its universe can be permuted to map it on to something else or even itself? To this, Gibson argues that we uncritical English speakers can of course go critical, refusing t ot a ke‘ r a bbi t ’ ,s a y ,a ti t sf a c eva l ue ,butwene e dnotdot hi s .Ont heot he rha nd,t hef i e l d linguist has to go critical in the case of radical translation because she does not have any other choice (Gibson 1988, p. 144). Whether or not our willingness to go critical to our home language has any essential power of overturning the presumably ultimate regress of relativisation seems to be a minor point to him. Gi bs on’ si nt e r pr e t a t i onofQui ne ’ sma noe uvr e —that is, ending the regression of background languages is, in practice, by acquiescing in our mother tongue and taking its words at face value—is that the reference scheme is fixed by acquiescing to homophonic translation in the home language (Gibson 1988, p. 139). This is of course in an ontological vein a ndf orGi bs oni tf ol l owst ha twha tt het e r msofone ’ sunc r i t i c a l l y accepted language denote is a matter of fact (p. 143). Indeed, Gibson even claims that some, if not all, are behavioural facts. The uncritical speakers of English would probably equivalent but logically incompatible. The explication of the notion of empirical equivalence is that theories are empirically equivalent if and only if they imply exactly the same set of observational consequences. 73 a s s e ntt ot heque r y“ Doe s‘ r a bbi t ’r e f e rt or a bbi t s ? ”whi l eouri ndi g e nousf r i e nddoe snot s t a ndac ha nc et oa s s e ntt ot hequ e r y“ Doe s‘ ga va ga i ’r e f e rt or a bbi t s ? ” This is a rather bizarre way to argue, because what Gibson shows is merely the fact that the indigenous informant does not acquire the ability to speak English. It seems to be mor es e ns i bl et oa s kourf i e l dl i ng ui s twhe t he rornots hea g r e e st ha t‘ g a va g a i ’r e f e r st o rabbit. She is in any rate the expert who provides translation manuals. And following the same reasoning of sectarianism, she can choose one of the competing manuals and thus de c i det her e f e r e nc es c he meofhe ri nf or ma nt ’ sl a ng ua ge . Thi si spr oba bl yve r y controversial, because there seems to be a difference between the choice of a scientific theory, which has facts of the matters to back it up, and the choice of a translation manual, whi c hdoe snot .Butt hi sdoe snots e e mt oc a us ea nyt r oubl ei fwea doptGi bs on’ s interpretation anyway.6 If, presumably ,s hea s s e nt st oourque r y“ Doe s‘ g a va ga i ’r e f e rt o r a bbi t s ? ”wha tne xt ?I st hi sabehavioural fact that can show the incorrectness of both of the theses of inscrutability of reference and ontological relativity?7 Surely not, and not because this is not a behavioural fact, but because it is illegitimate. The major dispute so far on either the thesis of inscrutability of reference or the thesis of ontological relativity i st ha tt he r ei snowa yt ode t e r mi nehowt ous et hes e ma nt i ct e r ms‘ r e f e r ’ort hel i ke uni que l y . The r e f or e ,“ Doe s‘ r a bbi t ’( or‘ g a va ga i ’ )r e f e rt or a bbi t s ? ”i snotal e g i t i ma t e question, because the only sensible way to ask this question is if there are some related s e ma nt i cf a c t st ohe l puss i ng l eoutt heus a geof‘ r e f e r ’ .Buta c c or di ngt oGibson, there are no semantic facts. 6 7 For the simplicity sake of argument, I will ignore this problem. Gibson is a full-blooded Quinean, so he surely does not accept the existence of semantic facts. 74 Mor e ove r ,e ve ni func r i t i c a lEng l i s hs pe a ke r sdoa s s e ntt ot heque r y“ Doe s‘ r a bbi t ’r e f e r t or a bbi t s ? ”i ft hec ount e r f a c t ua lc ons i de r a t i oni sdr oppe d,i ti sdubi oust ha tt he ywi l l assent to or dissent from this sort of queries uniformly. For instance, some, if not most, unc r i t i c a lEng l i s hs pe a ke r smi g hta s s e ntt ot heque r y“ Doe s‘ At l a nt i s ’r e f e rt oAt l a nt i s ? ” or“ Do e s‘ phl og i s t on’r e f e rt ophl og i s t on? ”Gi bs o ndoe snots howust ha tunc r i t i c a l speakers of English would uni f or ml ya s s e ntt oe ve r yque r yoft hef or m“ Doe s‘ n’r e f e r st o n( s ) ” . Wor s e ,i fs omeunc r i t i c a lEng l i s hs pe a ke r sdoa s s e ntt ot heque r y“ Doe s ‘ phl og i s t on’r e f e rt ophl og i s t on? ”wha ts houl dGi b s one xpl a i nt hi sword-world correspondence relation? Therefore, there is no such demarcation as Gibson expects to be be t we e ns t a y i ngunc r i t i c a lorg oi ngc r i t i c a lt oone ’ shomel a ng ua g e .I nSe c t i onFour ,I will show that what Gibson shows us is not a word-world correspondence reference scheme of a given language constituted by taking words at their face value, but a disquotational schema of reference that is essential to our daily communication as well as scientific discourse. Tobee xa c t ,t heg r a vi t yofQui ne ’ si nde t e r mi na c yt he s i si st hede ni a lofs e ma nt i cf a c t s that can absolutely determine certain semantic features such as in the case of radical t r a ns l a t i on,a nda l lt her e l a t e dphy s i c a lf a c t sc a ndonobe t t e ront hi spoi nt .Fi e l d’ s position is that, even if we grant the assumption that there are no semantic facts, the thesis of ontological relativity does not follow. Consequently, the reference scheme of a given theory or language is neither explicable in the absolute sense of a word-world relation nor in the relativised sense of a word-word relation. This predicament can be solved, Field claims, by introducing the causal relations between words and extralinguistic objects and thus maintaining a more general sense of word-world correspondence relations between 75 them. Surprisingly, Gibson is eager to retain this word-world correspondence relation8 a ndnot i c e st ha tt hepoi ntQui nema dei n“ Re pl yt oEl e ve nEs s a y s ”( Qui ne1981a) is so deserving of putting out the welcome mat—na me l y ,“ Ius et hewor d[ ‘ r e f e r ’ ]t or e l a t e linguistic expressions to objects. Real obj e c t s . ”( p.243) .Gi bs on’ sma i npos i t i oni sa s follows: However, so long as linguists A and B continue to conform unquestionably in their usages of‘ r a bbi t ’a n d‘ un de t a c h e dr a bbi tpa r t s ’ ,j u s ts ol on gdo‘ r a bbi t ’a n d‘ u n de t a c h e dr a bbi t pa r t s ’r e f e r ,r e s pe c t i v e l y ,t or a bbi ta n dun de t a c h e dr a bbi tpa r t s .An dt h i s …pa i rofr e l a t i on s , u n l i k et h epa i rofr e l a t i on s‘ g a v a g a i ’ / ‘ r a bbi t ’a n d‘ g a v a g a i ’ / ‘ u n de t a c h e dr a bbi tpa r t s ’ ,i s not of the word-word type but of the word-wor l dt y pe :‘ r a bbi t ’ / r a bbi t ,‘ u n de t a c h e dr a bbi t pa r t s ’ / u n de t a c h e dr a bbi tpa r t s . Qu i n e ’ spoi n ti st h a tt h e s e …wor d-world relations, too, can be disturbed, but they need not be. And so long as they remain undisturbed, it makes pe r f e c t l yg oods e n s et os a yt ha t‘ r a bbi t ’r e f e r st o( i . e .de n ot e s )r a bbi t sa n d‘ un de t a c h e d rabbi tpa r t s ’r e f e r st o( i . e .de n ot e s )u n de t a c h e dr a bbi tpa r t s . ( Gi bs on1988, p.1 40) Iwi l lr e nde radi f f e r e nti nt e r pr e t a t i onofQui ne ’ sma noe uvr e .ButIwoul dl i ket of i r s t poi ntouthowmi s l e a di ngGi bs on’ se xpl i c a t i oni s . If, presumably, there are word-world correspondence relations between words and extralinguistic objects, real objects, in our own language, and others can only have a relativised word-word relations in their interpreting reference schemes of my language, then it seems to make perfectly good sense to claim that this establishes a solid case of private language. After all, nobody is in principle able to grasp my understanding of my own language, except me. Of course, there is an important difference that might cause some trouble to my 8 Gi bs onc e r t a i n l ydoe sn ota c c e ptFi e l d’ sc or r e s pon de n c ea c c ou n toft r u t h ,bu th ol ds that the thesis of ontological relativity does not preclude a correspondence theory of truth, at least in some general sense of ‘ c or r e s pon de n c e ’s u c ha sDa v i ds on i a n( Gi bs on1988, p.1 40) . Th i si sn ote n t i r e l ywr on g .Bu tIc a nh a r dl y a c c e ptGi bs on’ sr e a s oning. 76 criticism—na me l y ,Ius e‘ myownl a ng ua g e ’i ns t e a dof‘ ourhomel a ng ua ge ’ . Thi si s , however, exactly the point. If it is the case that the reference scheme of our language is f i xe ds ot ha t‘ r a bbi t ’onl yr e f e r st or e a lr a bbi t s ,t he ns oi st her e f e r e nc es c he meof my language. But how do I know that other people who speak English uncritically do use ‘ r a bbi t ’t or e f e rt or a bbi t s ?I ft he r ea r enos e ma nt i cf a c t sa boutt hi st he nt he r emus tbe physical or, probably, behavioural facts to help me figure it out. It nevertheless follows that there should also be physical or behavioural facts to pin down the reference scheme ofa nyf or e i g nl a ng ua g e . AsQui nes a y si n“ Ont ol og i c a lRe l a t i vi t y ” : If it is to make sense to say even of oneself that one is referring to rabbits and formulas and not to rabbit stages and Gödel numbers, then it should make sense equally to say it of s ome on ee l s e . Af t e ra l l …t h e r ei sn opr i v a t el a ngu a g e .( p.47) My argument is clear. If the reference scheme of my language can be fixed by a set of word-world correspondence relations as Gibson expects it to be, there must be either semantic or physical/behavioural facts to support that. And similarly my fellow English speakers should equally be able to fix their reference scheme. In fact, we should have exa c t l yt hes a mer e f e r e nc es c he mei nwhi c h‘ r a bbi t ’r e f e r st or e a lr a bbi t si nawor d-world correspondence sense. Since Gibson holds that there are no semantic facts, this fix of ontology has to depend on physical or behavioural facts. Just what kind of facts, then, fixes the ontology of a home language but not the one of a remote language? Certainly notGi b s on’ sbehavioural fact that our jungle informant fails to reply affirmatively to our que r y“ Doe s‘ g a va g a i ’r e f e rt or a bbi t s ? ”AsIs a i dbe f or e ,ouri nf or ma nt fails to respond otherwise rather because he takes our question as gibberish. The difference between home and remote languages in this sense should only be a matter of degrees but not of kinds. If Gi bs on’ spos i t i oni mpl i e st ha tone ’ shomel a ng ua gei spr ivileged over other translated 77 languages, it follows, however, that all related indeterminacy theses fail. But this means t ha tt he r ea r es e ma nt i cf a c t s ,a ndt hi si sc ont r a r yt oGi bs on’ sor i g i na ls uppos i t i on. Therefore, there is no way to fix the reference scheme of our language as well as our i nf or ma nt ’ s .I fGi bs oni ns i s t sont hepoi ntt ha tt hedi f f e r e nc ei sama t t e rofki nds ,t he nhe has to respond to the question that what kind or kinds of physical/behavioural facts can fix our ontology but not our informa nt ’ s .Mor e ove r ,hea l s oha st og i veas a t i s f a c t or y a ns we rt oamor eg e ne r a lque s t i on:howi si tt ha tIc a nde t e r mi nemyf e l l ows pe a ke r ’ s reference scheme? and what is the difference between this determination and the one in the case of radical translation? If the answers are not satisfactory, especially to the second question, Gibson has to face the charge, even from Quine, of conceding to the existence of private languages. Quine also makes an analogy between staying aboard/rocking the boat and his manoeuvre (ending the regression of background languages is, in practice, by acquiescing in our mother tongue and taking its words at face value)/contemplating a permutational mapping ofourl a ng ua geoni t s e l forunde r t a ki ngt r a ns l a t i oni n“ Thi ng sa ndThe i rPl a ces in The or i e s ”( Qui ne1981b,p.20) . Gi bs on’ sr e a di ngoft hi sa na l ogyi st ha tr e f e r e nc edoe s not go inscrutable unless we rock the boat. Otherwise there is a fact of the matter to que s t i onsl i ke“ Doe s‘ r a bbi t ’r e f e rt or a bbi t s ? ”i no urhomel a ng ua g e( Gi bson 1988, p. 145) . Tha t‘ f a c toft hema t t e r ’i ss uppos e dl ys e ma nt i cr a t he rt ha nphy s i c a lorbe ha vi our a l . What I am not sure is whether or not Quine himself is absolutely consistent on the issues of inscrutability of reference and ontological relativity. But it is almost certain that Gi bs on’ sa c c ount ,whi c ha s s i g nse ve r yr e f e r e nt i a lt e r m ofourownl a ng ua geawor d-world correspondence relation to a specific object or set of objects, causes a discrepancy 78 between the relativisation of reference scheme in the case of mother tongue and the relativisation of the reference scheme in the case of remote languages. This is exactly what Field argues against. And this is just what we cannot accept. The staying aboard/rocking the boat analogy should have a more sensible interpretation. 4. Disquotational Schema of Reference and Semantic Ascent of Terms Da vi ds ona ppr ove sofFi e l d’ sc r i t i c i s mt ha tt he r ei snoi nt e l l i g i bl ewa yt oj us t i f yt he concept of ontological relativity in either an absolute or a relativised sense. For him, there is no intelligible way to stride the first step in a so-called relativisation of ontology. The original reason to relativise the reference scheme of a given language or theory to another is that no empirical evidence can fix its reference scheme. That very problem, however, materialises again in the interpreting language or theory. This occurs accordingly in every attempt to relativise any given language or theory. It is then hard to see why the ontological regress should have gone on in the first place. Except for this, Davidson does agree that it makes sense to say about a relativised concept of ontology. His reasons are twofold. On the one hand, all acceptable reference schemes of a given language for a speaker or a community have a common feature that they ascribe equivalent truth conditions to all sentences. On the other hand, these schemes are distinguishable from each other not on empirical grounds, but because they employ predicates that cannot have the same extension (Davidson 1979, p. 238). This relativised concept of ontology, however, only helps us answer questions about reference but not reference itself (p. 239). Thi si sDa vi ds on’ ss t a ndpoi ntont het he s i sofont ol og i c a lr e l a t i vi t y . Myi nt e r pr e t a t i onofQui ne ’ sma noe uvr ei snoti nc ompa t i bl ewi t hDa vi ds on’ spos i t i onon 79 the thesis of ontological relativity. I think that Davidson makes a more feasible and more prudent interpretation than Gibson does. Quine does not offer us an account of word-world correspondence relations between words of our home language and e xt r a l i ng ui s t i cobj e c t s . Whe nhes a y st ha tt heus eof‘ r e f e r ’i st or e l a t el i ng ui s t i c expressions to real objects, he does not pin down the reference scheme of English in whi c h‘ r a bbi t ’r e f e r st oreal rabbits rather than real undetached rabbit parts or anything e l s e . Thi si sa ppa r e nti nQui ne ’ sr e pl yt oPa ul A.Rot h( Qui ne1986a) in The Philosophy of W. V. Quine: Within the home language, reference is best seen (I now hold) as unproblematic but t r i v i a l ,onapa rwi t hTa r s k i ’ st r u t hpa r a di gm. Th u s‘ Lon don’de s i g n a t e sLon don (whatever that i s )a n d‘ r a bbi t ’de n ot e st h er a bbi t s( wh a t e v e rthey are). (p. 460) Those italics are the key to my interpretation.9 In Quine 1953a, Quine holds that, except the truth paradigm, Tarskian paradigms on ‘ i st r ueof ’( t ha ti s ,‘ de not e ’or‘ s i g ni f y ’ )a nd ‘ na me ’r e l a t i vet oag i ve nl a ng ua g ea r ea l s os us t a i na bl e( pp.134-6). They can be characterised as follows: ( 1)‘ ’i st r ue -in-L of every ( 2)‘ ’na me s -in-L thing and nothing else; and nothing else. (1) holds when any one general term is written in the two blanks and (2) holds when any name is written in the two blanks. According to these three paradigms, I think we should i nt e r pr e tQui ne ’ sma noe uvr ea ssemantic ascent of terms—as offering a disquotational s c he maofr e f e r e nc et oone ’ sownl a ng ua g e .I n“ Qui nea ndt heCor r e s ponde nc eThe or y ” (1974), Field thinks that our disquotational schema of reference is incompatible with i nde t e r mi na c yoft r a ns l a t i on.I nt he“ Pos t s c r i pt ”( 2001)oft hi spa pe r ,hedisagrees with hi ms e l font hi spoi nt .Ho we ve r ,hedoe snotc ha ngehi smi ndonhi sc r i t i c i s mt oQui ne ’ s 9 Th es a mepoi n ti sa l s os h owni n“ Tr u t ha n dDi s qu ot a t i on ” .Pl e a s ec f .t h epr e v i ousc h a pt e r . 80 ontological relativity thesis. It follows that he at least does not go in for this disquotationalist approach. Some commentators overlook the significance of this point. For instance, Hylton in “ Qui neonRe f e r e nc ea ndOnt ol ogy ”( Hy l t on2004)ma i nt a i nsas i mi l a rpoi ntt oGi bs on’ s . After his citation of Quine 1986a, (cited above in the previous paragraph) he says: The point here, I think, is that while we are speaking our familiar language (or a language whose translation into it is well established), the inscrutability of reference simply gets no grip. (p. 142) Butr e me mbe rQui nes a y st ha t‘ London’de s i g na t e sLondon( wha t e ve rthat i s )a nd‘ r a bbi t ’ denotes the rabbits (whatever they are) right after he says that reference within the home language should be best seen onaparwi t hT ar s k i ’ st r ut hpar adi gm. He sees T-sentences i nadi s quot a t i ona ls e ns e ,s oi st ha t‘ London’de s i gna t e sLondon( wha t e ve rthat is) and ‘ r a bbi t ’de not e st her a bbi t s( wha t e ve rthey are). Similar to (DQ), of course, the term quoted and not quoted in an instance of the disquotational schema of reference have to be regarded as the same word type with the same interpretation.10 This is probabl yabi tr e mi ni s c e ntofSt e phe nLe e ds ’a c c ountofont ol og i c a lr e l a t i vi t yi n “ Howt oThi nka boutRe f e r e nc e ”( Le e ds1973) .Le e dst a ke st het he s i sofont ol og i c a l relativity to amount to a philosophically uninteresting triviality that (semantical) questions, whatever their subject matters are, must be asked relative to a background language (p. 488). He continues his argument in the following way: To speak of reference, we must pretend that certain questions about our language have been answered—these questions are questions about the reference of the terms in our language; the answer is given by the Tarski schema for reference in the background 10 Horwich holds the same view. Please cf. the following section. 81 theory. (p. 490, my italics) Thi spr e t e ndi ngha si t spr a c t i c a lv a l uea sar e s ul tofi t s‘ ha r ml e s sc onve ni e nc e ’ . The ontological talk, in which the subject matter is not the relations between language and the world but some important relations such as intertranslatability held between languages, is in principle possible without ever speaking of reference (pp. 491-2). This account of ont ol og i c a lr e l a t i vi t yi sne ve r t he l e s si nc ompa t i bl ewi t hQui ne ’ sna t ur a l i s mi nt ha tt he existence of objects depends on our ontological commitments and they must in turn reside within a (scientific) theory. Surely the ontological talk is not about relations between languages but between language and the world. Or better, it is about relations between theories and the world.11 Theory as a whole does owe its ontology to the world in the sense that its observational consequences have to be compatible with the world observed. It is only when theories are empirically equivalent, the ontological talk seems to be bl ur r e d . Tode ve l opa ni nt e r pr e t a t i onofQui ne ’ sphi l os ophya c c or di ng l y ,bya s s umi ngt ha t there is an objective position to be occupied to claim the truthfulness of any theory, or that the importance of ontology is not about the relations between language and the world but be t we e nl a ng ua g e s ,i sa nut t e rmi s unde r s t a ndi ng .Exc e ptf ort hi s ,Le e ds ’a c c ountdoe s offer a valuable observation that Tarskian paradigms have to play a role of great ma g n i t udei nt hei nt e r pr e t a t i onofQui ne ’ st he s i sofont ol og i c a lr e l a t i vi t y .Buti ti snota matter of pretending, but of offering a disquotational schema of reference to make possible ontological talk. And all the instances of this schema do not constitute any genuine reference scheme that can possibly be given. 11 Qu i n edoe sh a v et h ei n c l i n a t i ont ou s e‘ l a ngu a g e ’a n d‘ t h e or y ’i n t e r c h a ng e a bl y . An ot h e rf a miliar t e r mi n ol ogyi s‘ c on c e pt u a ls c h e me ’t h a tc a us e sade ba t ebe t we e nDa v i ds ona n dQu i n e .( Th a ti s ,“ Ont h e Ve r yI de aofaCon c e pt u a lSc he me ”a n d“ Ont h eVe r yI de aofaTh i r dDog ma ” . )Th i si sh owe v e rn ota mong the main issues here. 82 Constructing a disquotational schema of reference is not doable in translating or i nt e r pr e t i ngal a ng ua g eot he rt ha nt het r a ns l a t orori nt e r pr e t e r ’ shome language. For i ns t a nc e ,“ ‘ Sc hne e ’r e f e r st os now”i snota ni ns t a nc eofadi s quot a t i ona ls c he maof reference in English.12 This is already apparent in the discussion of the designation schema in the previous chapter. Therefore, to take the disquotational schema of reference a st hepr ope ri nt e r pr e t a t i onofQui ne ’ sma noe uvr ei st heonl ywa yt oma kes e ns eofi t . A disquotational schema of reference is crucial to the thesis of ontological relativity and even the thesis of ontological commitments.13 There are at least two aspects of the thesis of ontological relativity, and the disquotational schema of reference is essential to both. One way to see this is in terms of so called notational relativity—that is, the universe of our language or theory can be permuted to map it on to itself. In this aspect, the universe is kept intact. So the relativity comes from the different notational permutation between referential terms and the members of the given universe.14 However, it is inevitable for the notational permutation to depend for its significance on a pre-established di s quot a t i ona ls c he maofr e f e r e nc e .Fori ns t a nc e ,i f‘ s now’unde rt hepe r mut a t i onr e f e r st o mi l ki n s t e a dofs now,t hi sc a nonl yma kes e ns ei f‘ s now’r e f e r st os nowa nd‘ mi l k’r e f e r s to milk disquotationally. Otherwise we have difficulties in understanding what the word ‘ mi l k’s t a ndsf ori nt hec a s eofpe r mut a t i on. Anot he rwa yt os e eont ol og i c a lr e l a t i vi t yi si n terms of the inscrutability of reference in which the universe is varied, as pointed out in 12 However, in terms of a na de qu a t et r a n s l a t i onma nu a l“ ‘ Sc h n e e ’r e f e r st os n ow”mi gh tbec on c e i v e dof indirectly. 13 Ag a i n ,Idon oti n t e n dt ode f e n dQu i n e ’ sv i e wpoi ntonon t ol og i c a lc ommi t me nt s .Fori ns t a n c e , Az z oun i i n“ On“ OnWh a tTh e r eI s ” ”u r g e su st h a tt h e r emi gh tbes omeother criteria for recognizing what a discourse commits us to (CRD in his abbreviation). He might be right, but this does not concern me anyway. 14 This notational permutation is not necessary on a global scale. And the permutation on a small scale reminds usoft h er e c ons t r u a lofpr e di c a t e si n t r odu c e dbyQu i n ei n“ OnEmpi r i c a l l yEqu i va l e n tSy s t e msof t h eWor l d” . 83 t hef i r s ts e c t i on.Fora nyg i ve nt h e or yi nwhi c hage ne r a lt e r m,s a y ,‘ r a bbi t ’r e f e r st o rabbits according to its reference scheme, one can produce a proxy function, say, ø to map rabbits on to ø(rabbits) or , say, undetached rabbit parts in the reference scheme of the ba c kg r oundt he or y .Si mi l a r l y ,t oma kes e ns eoft hi sweha vet or e l yonr e a di ng“ ‘ Ra bb i t ’ r e f e r st or a bbi t s ”a nd“ ‘ Unde t a c h e dr a bbi tpa r t s ’r e f e r st ounde t a c he dr a bbi tpa r t s ” disquotationally. In the case of ontological commitments, the crucial nature of the disquotational schema of reference is even more apparent. When Quine says that the answer to the question of whether or not there is an F, say, depends on our decision to assign or not assign a value to the variable of an objectual quantifi c a t i on“ x(Fx) ”i nar e g i me nt a t i onofours c i e nt i f i c theory into the language of first-or de rl og i c ,hec e r t a i nl ydoe snots pe c i f ywha t‘ F’r e f e r s t o.I nt h ea bovec a s e ,wha t‘ F’r e f e r st oi sofc our s eF( s ) . Tha ti s ,i f‘ F’a mount st o‘ r a b bi t ’ t he n‘ F’r e f e r st or a bbi t s ,ori f‘ F’a mount st o‘ unde t a c he dr a bbi tpa r t s ’t he n‘ F’r e f e r st o undetached rabbit parts, and so on. Indeed, the thesis of ontological commitments tells us that the comprehensive scientific theory held by us commits us to certain sorts of objects, but what it does not tell us is what these objects really are. This is why in Quine 1995 Quine says: Ic on c l u de …t h a twh a tma t t e r sf ora nyobj e c t s ,c on c r e t eora bs t r a c t ,i sn otwh a tt h e ya r e but what they contribute to our overall theory of the world as neutral nodes in its logical structure. (pp. 74-75) Without the disquotational schema of reference, none of this is possible. In Word and Object, Quine holds that semantic ascent of terms usually happen when we are talking about some abstract objects such as miles or points. In the case of physical 84 objects such as wombats or kangaroos, this normally would not happen. The usefulness of semantic ascent of terms is as follows: The strategy of semantic ascent is that it carries the discussion [of what there is] into a domain where both parties are better agreed on the objects (viz., words) and on the main terms concerning them. (p. 272) My borrowing of this strategy is not entirely unreasonable. In the same page, Quine states: “ Se ma nt i ca s c e nt ,a sIs pe a kofi t , a ppl i e sa ny whe r e . ”Hea l s oc l a i mst hef ol l owi ng :“ we have talked more of words than of objects even when most concerned to decide what t he r er e a l l yi s :wha tobj e c t st oa dmi tonourowna c c ount ”( p.270) .I ns pe a ki ngof Qui ne ’ sma noe uvr ea ndt het hesis of ontological relativity, this strategy is essential. Without it the manoeuvre fails miserably. And this can be well grasped in my illustration of the disquotational schema of reference. We simply substitute talk of terms for talk of extralinguistic objects. What should not be confused with the disquotational schema of reference is that we do not render any fixed reference scheme of our mother tongue. Inscrutability of reference and ontological relativity are inevitable in doing empirical semantics. They represent the mystery of the semantic concept of reference. Conversely, the disquotational schema of reference and satisfaction are the reason why scientific research can still be carried on without being disturbed by the semantic mystery. If the ontology of one language or theory has to be reinterpreted in a background language or theory so that it can be understood by others, and this process is infinite, then, presumably, one has to understand all these theories or languages in order to understand the original ontology. This makes the thesis of ontological relativity the most inflationary theory ever, and the most inflationist theory one can imagine.15 It seems that the 15 Th i si si n s pi r e dbyA.Gu pt a ’ sc r i t i c i s mt oP .Hor wi c h’ smi n i ma l i s t i ct h e or yoft r u t hi n“ Mi n i ma l i s m” (collected in Philosophical Perspectives, Volume 7, Issue Language and Logic, p. 365). 85 disquotational schema of reference can provide a shelter for the thesis of ontological relativity from this criticism. It is also crucial to scientific theories so that they can have a significant way to talk about their reference schemes without the infinite regression of reinterpretation.16 Thi ss c he mai sa l s oc ompa t i bl ewi t hQui ne ’ sgeneral view on ontology. After all, he states: St r u c t u r ei swha tma t t e r st oat h e or y …n ott h ec h oi c eofi t sobj e c t s … Th eob j e c t s ,or values of variables, serve merely as indices along the way, and we may permute or supplant them as we please as long as the sentence-to-sentence structure is preserved. (Quine 1981b, p. 20) Therefore, the disquotational schema of reference can be seen as a part of the more general account of disquotationalism. I n“ On t ol ogya ndI de ol ogy ”( Qu i ne1951) ,Qui nee xpl a i nst hedi s t inction between ontological and ideological commitments as follows: Given a theory, one interesting aspect of it into which we can inquire is its ontology: what entities are the variables of quantification to range over if the theory is to hold true? Another no less important aspect into which we can inquire is its ideology … wh a ti de a s can be expressed in it? (p. 14) One of the values of the disquotational schema of reference is to help us axiomatise some familiar mathematical theories without question begging (Quine 1960, pp. 272-273). In other words, by going up a level we can separate the target theory from its ideology. For instance, in axiomatisation of a geometrical theory, one is in trouble with not being sure that one deduces some familiar formulas of the axiomatised theory purely from its axioms without utilising any part of further geometrical knowledge. To avoid this unwelcome embarrassment one can turn to the device of disinterpretation, acting as if one only 16 Marian David in Correspondence and Disquotation offers a disquotational account of reference to contribute to the recursive definition of disquotational truth (pp. 110-119). My interpretation, on the other hand, shows that there is a distinct need to provide a disquotational schema of reference. 86 understands the logical vocabulary but not those peculiar terms with regard to the axiomatised geometrical system. The problem with this method, according to Quine, is t ha ti tl e a dsust ot hef ol l owi ngc ons e que nc e :“ wene ve rknowwha twea r et a l ki nga bout , 17 nor whether what we are saying is t r ue . ” Qui net he nc l a i mst ha t ,wi t hFr e ge ’ s achievement of a full formalisation of logic, semantic ascent as a device against question-begging turns out to be available (ibid., p. 273). He then says: Given the deductive apparatus of logic in the form specified operations on notational forms, the question whether a given formula follows logically from given axioms reduces to the question whether the specified operations on notational forms are capable of leading to that formula from the axioms. (ibid.) What concerns us in the axiomatisation of a mathematical theory is the danger that its ideology is always inseparable from the axiomatisation. Semantic ascent as described above shows a way to avoid this danger. Another value is in the evaluation of a neoteric scientific theory (Quine 1990b, p. 81). Qui ne ’ se xa mpl ei sEi ns t e i n’ ss p e c i a la ndg e ne r a lt he or i e sofr e l a t i vi t ybywhi c hourb a s i c conceptions of space and time are challenged tremendously.18 The evaluation of this new theory is cumbersome with or without the ideology of Newtonian physics. Without this particular ideology, the evaluation would not proceed in the first place. With the help of this ideology, however, the evaluation begs the question. It is just like taking the premise that kanji (which is bot ht hepr on u nc i a t i onof‘ Chi ne s ec ha r a c t e r ’i nChi ne s ea ndi n Japanese) is purely a cultural product of Japan, to evaluate whether or not Japanese kanji is under the influence of Chinese characters. In terms of the semantic ascent of terms, Quine then says: 17 Th i si sQu i n e ’ sc i t a t i onofRu s s e l l ’ sMysticism and Logic and Other Essays (p. 75) in Word and Object (p. 273). 18 A related point is also made in Quine 1960, p. 272. 87 But by semantic ascent one could compare the new and old theories as symbolic structures, and so appreciate that the new theory organized the pertinent data more simply than the old. Simplicity of symbolic structures can be appreciated independently of those basic conceptions. (ibid.) In other words, by taking the special and general theories of relativity and Newtonian physics as symbolic structures, one can compare the two with regard to which sentence-to-sentence structure is capable of organising empirical data more simply. 5. Disquotational Schema of Reference and Disquotational Theory of Truth In the last chapter, I considered two reasons to reject the idea of incorporating inductive axioms of reference and satisfaction into the formulation of a disquotational theory of truth. The central issue there was what sort of theory of reference for a given natural language we can get to obviate the potential difficulties of our axiomatised disquotational theory of truth suggested by the incorporation.19 The first reason is that the theory of reference involved has to deal with the explanatory order of semantic concepts, as indicated there that this is crucial to Quinean semantics. The second reason is that the theory of reference has to be without any substantive feature, so that the resultant inductive axioms are be able to properly imply nothing but all disquotation sentences of the truth language within which the disquotational theory of truth of the base language is constructed. It seems that, now, we have a candidate here, namely, the disquotational schema of reference. Nelson is probably one of the philosophers who might object to this reasoning, but for the 19 As stated there, the incorporation of inductive axioms only arrests our interest when this is about to give a formulation of disquotational theory of truth for any given natural language. 88 wrong reason. The main rationale for incorporating inductive axioms into the disquotational theory of truth is to render a way to explain how the semantic ascent of sentences helps express some infinite conjunctions or disjunctions with the same syntactic feature. The disquotational schema of reference is not supposed to amount to explicit definitions of the semantic concepts of reference and satisfaction. If it were, then it would have violated the explanatory order between semantic concepts. Since it is not, it follows that all disquotation sentences are material biconditionals. This, however, according to Nelson (1997), causes a problem of compatibility between the disquotational theory of truth and the thesis of ontological relativity. In the case of notational relativity in which the universe is not varied, suppose one produces a proxy function, say, ø to map a on to ø(a) in the reference scheme of the background theory and similarly to map a predication F(a), that F refers to (is true of) each thing a such that F(a) in the object theory, to Fø(a) in the background theory. The T-s e nt e nc e“ ‘ F( a) ’is true iff F(a) ”i sama t e r i a l biconditional according to the above analysis. Nelson indicates that, in terms of ontological relativity, according to which F(a) iff Fø(a) ,wec a ns ubs t i t ut e‘ Fø( a) ’f or ‘ F( a) ’i nt heT-sentence. The resultant biconditional i s“ ‘ F( a) ’i st r uei f fFø( a) ”a ndi ti s certainly not a T-sentence. This is apparently not a problem to my formulation of the disquotational theory, (DQ) ,i nt hel a s tc ha pt e r ,f or ,a f t e rs ubs t i t ut i ng‘ Fø( a) ’f or‘ F( a) ’i n “ ‘ F( a) ’i st r uei f fF( a) ” ,“ ‘ F( a)’i st r uei f fFø( a) ”i snotadi s quot a t i ons e nt e nc ea ny wa y .I t is the conclusion of two premises—na me l y ,t hebi c ondi t i ona l“ F( a) iff Fø(a) ” ,r e s ul t i ng f r omus i ngt hepr oxyf unc t i on‘ ø’t oma pF( a) on to Fø(a), and the disquotation sentence “ ‘ F( a) ’i st r ueiff F(a) ” . The best way to see the disquotational schema of reference in general is to see it as based 89 on the pre-theoretic intuitive concepts of reference and satisfaction of any given natural language. It does seem right, intuitively, that speakers of any language incline to understand the pre-theoretic referential and satisfaction relations of their languages in the disquotational sense. For instance, every English speaker normally accepts the disquotational schema of reference with no regard to whether or not the referent, say, of t hel e ge nda r yc i t yof‘ At l a nt i s ’o rt hemy t hi c a lc r e a t ur e‘ Pe ga s us ’e xi s t .I ft he r ea r ewor ds s uc ha s‘ r a bbi t ’a nd‘ London’i nEng l i s ha swe l la swor dsl i ke‘ phl og i s t on’a nd‘ Vul c a n’ , pe opl ea r er e a dyt oa c c e ptt ha t‘ p h l og i s t on’de not e sphl og i s t ona nd‘ Vul c a n’r e f e r st o Vul c a na sf a ra st he ya r er e a dyt oa c c e ptt ha t‘ r a bbi t ’de not e sr a bbi t sa nd‘ London’r e f e r s to London. There are no ontological reflections involved here. The problem of incorporating inductive axioms of reference and satisfaction in formulating a disquotational account of the semantic concept of truth, is that it is difficult to find a suitable theory of reference to account for the explanatory order and to enable the inductive axioms to imply all disquotation sentences. Providing that the thesis of ont ol og i c a lr e l a t i vi t yhol ds ,Ha l ba c h’ si de aofi nc or por a t i ngi nduc t i vea xi omsofr e f e r e nc e and satisfaction to enhance the deductive power of the disquotational theory of truth is therefore not satisfactory. For no matter whether or not the disquotational schema of reference goes finite or infinite, it is not a genuine reference scheme. Otherwise, it contradicts the thesis of ontological relativity.20 Even if we assume that the thesis of ontological relativity is untenable, to consider a disquotational schema of reference as providing a genuine reference scheme is problematic. This will be shown in the following 20 McGee is another philosopher who is in favour of accounting for disquotational truth in terms of the n ot i on sofr e f e r e n c ea n ds a t i s f a c t i on . Al t h oughh edoe sn ota c c e ptQu i n e ’ si n de t e r mi n a c yt h e s e s ,t h e implausibility of his account will not be discussed, even briefly, until Chapter Five. 90 di s c us s i onofHor wi c h’ sr e a s oni ng . Andt hec r i t i c i s mt oHor wi c h’ sa c c ounti se qua l l y a ppl i c a bl et oHa l ba c h’ s. Horwich is another important advocate of a disquotational theory of reference and s a t i s f a c t i on. Ac c or di ngt ohi m,Qui ne ’ st he s i sofont ol og i c a lr e l a t i vi t ydoe snothol d. Hence he can consistently see the instances of the disquotational schema of reference (and satisfaction) as providing a genuine reference scheme. That is, Th er ou g hi de ai st h a tou rme a n i ngwha twedobyt h ewor d“ r e f e r s ”c on s i s t si nou r disposition to accept sentences such as Tokens of *London* refer to London, Tokens of *the highest mountain* refer to the highest mountain, and so on. Since an ambiguous s i n gu l a rt e r m,“ n” ,wh e nus e da st h e s ee x a mpl e st o designate a referent, must be understood in just one of its senses, it will not (if any s e ns eof“ n” )bet r u et os a yt ha tevery token of“ n”( i n di v i dua t e dph on ol og i c a l l y )r e f e r s to n. This is why the expression-types deployed in disquotational reference specifications are articulated using star quotation marks—indicating that they are i n di v i du a t e d…i nt h eba s i sofme a ni ng . Thus our meaning wha twedoby“ r e f e r s ”c on s i s t si nou ri n c l i n a t i ont oa c c e pt instances of the disquotational schema Tokens of *n* refer to n, whe r ewh a ta r es u bs t i t u t e df ort h et wooc c u r r e n c e sof“ n”a r eu n de r s t oodi nt h es a me way.21 (Horwich 1998a, p. 118) In the followingc ha pt e rIwi l ls howt ha tHor wi c h’ sr e j e c t i onoft het he s i sofont ol og i c a l relativity fails. So we could tentatively say that one can always find a proxy function ø to map n of the disquotational reference scheme on to ø(n) in the reference scheme of the background theory, and similarly to a predication F(n). 21 Hor wi c hdoe sn otu s e‘ di s pos i t i on’a n d‘ i n c l i n a t i on’i n di f f e r e n t l y .I naf oot n ot e( 23)of“ ADe f e ns eof Mi n i ma l i s m”( Hor wi c h2001a), Horwich makes it clear that he distinguishes inclinations from dispositions to accept instances of the equivalence schema because of consideration regarding the liar paradoxes. However, it seems that he does use them indifferently here. 91 Butwedonotha vet o.Hor wi c h’ sa c c ountha st hedi f f i c ul t yofpr ope r l ye xpl a i ni nga l l relevant facts regarding reference. Let me first make some clarification of his schema of reference. For Halbac h’ sa c c ount ,t hepr obl e mi st ha ti tf a l l ss hor tofe xpl a i ni ngs ome di s quot a t i ons e nt e nc e ss uc ha s“ ‘ Pe ga s usi saf l y i nghor s e ’i st r uei fa ndonl yi fPe g a s u si s af l y i nghor s e . ”Hor wi c h’ sa c c ount ,ont heot he rha nd,i si mmunet ot hi sc r i t i c i s m,s i n c e he does not attempt to explicate disquotational or minimalist truth in terms of the inductive axioms of reference and satisfaction. He is fully aware of this sort of problem, and hence he modifies his schema as follows: (R) Tokens of *n* refer, if at all, to n.22 (Horwich 1998a, p. 119) He further provides some principles to help form a corresponding schema accounting for our attribution of reference to terms of foreign languages. To simplify my discussion, (R) can be restricted as only applying to referential words of a home language with a fixed interpretation, when indexical, vague, and ambiguous words along with other related matters are precluded.23 An account of reference surely need not tell us which words refer and which do not. But it has to tell us why some words refer and why some do not. For instance, a causal theory of reference would explain that a word refers if and only if a causal relation between it and its referent can be established. Otherwise, the word does not refer. Since Horwich regards it asf a l s et ha t‘ At l a nt i s ’a nd‘ t hel a r g e s tpr i menumbe r ’ refer to Atlantis and the largest prime number, because of there being no such things to be r e f e r r e dt o,t he‘ i fa ta l l ’i n( R)i st oma ket hede ma r c a t i onbe t we e nwor dst ha tr e f e ra n d those that do not. Thepr obl e mo fHor wi c h’ sa c c ounta r i s e sf r om t hi sa ddi t i ona lc ondi t i on 22 Th es e n t e n c e“ Pe g a s u si saf l y i ngh or s e ”i sade c l a r a t i v es e n t e n c e ,a n da c c or di n gt oHor wi c h’ s equivalence schema, there is a corresponding proposition expressed by it. As a result, that the proposition t h a t“ Pe g a s u si saf l y i n gh or s e ”i st r u ei fa n don l yi fPe g a s usi saf l y i ngh or s ei sa na x i omoft h ee qu i v a l e n c e s c h e maofpr opos i t i on a lt r u t h .I ti sc l e a rt h e nwhyHor wi c hdoe sn ota n dwou l dn ota c c e ptHa l ba c h’ s approach. 23 Hor wi c h’ sdi s qu ot a t i on a ls c he maofr e f e r e n c ea n da l li t sa xi omsa r es i mpl i f i e di nn ot a t i ona s“ ‘ n’r e f e r s to n”h e r e a f t e r . 92 s t oodf orby‘ i fa ta l l ’ .Fori ns t a nc e ,wha tdoe st he‘ i fa ta l l ’s t a ndsf ori nt hec a s eof “ ‘ Phl og i s t on’r e f e r s ,i fa ta l l ,t ophl og i s t on” ? Hor wi c h’ sf or ma lc ha r a c t e r i s a t i onof( R)might be helpful; that is (R1) (x) (Tokens of *n* refer to x iff n = x). (ibid.) This is similar to his formulation of the disquotational schema of reference in Truth; that is: (R2) (x) (<d> refer to x iff d = x), whe r e‘ <d>’s t a ndsf orwor dt y pe s .( Hor wich 1998b, p. 116) Apparently, if (R1)i st hef or ma lc ha r a c t e r i s a t i onof( R) ,t he n“ Toke nsof*n* refer to x”a t the left hand side of (R1) is definitely meaning that tokens of *n* refer disquotationally to x. Otherwise, that tokens of *morning star* refers to evening star if and only if morning star = evening star is an instance of (R1). But this is not an instance of (R).24 The deflationary account of reference, according to Horwich, admits no substantive relation of reference. Although there is really a relation of reference, this relation is without any underlying nature (Horwich 1998a, p. 123). And all the facts regarding reference can be explained by all the axioms/instances of the disquotational schemata of reference and satisfaction (ibid.). Thus the ‘ i fa ta l l ’c a nnotbee xpl i c a t e da se nc a ps ul a t i ngwi t hi ni tt ha t tokens of *n* (or <d>) bear some non-semantic relation r to something x. This is not just inflationary and circular, but absurd. (R) is then saying that, if there is really an object to which tokens of *n* bear a substantive referential relation, tokens of *n* refer to n. The absurdity rests on the assumption that a substantive referential relation always boils down to a disquotational referential relation, so the disquotational reference scheme is the only 24 Mor e ov e r ,‘ i fa ta l l ’i n( R)c a n n otber e a da st h a ti ft ok e n sof* n* refer they refer to n. For one should rather regard the account of whether or not tokens of *n* refer as the account of reference, and it is far from clear that this account would be not substantive. 93 reference scheme. Therefore, what Horwich has in mind must be that some reference-free auxiliary explanatory resources can explain, not which word refers, but whether or not something that is referred exists. For instance, if there is a suitable explanation of the e xi s t e nc eoft a c hy on,t he n‘ t a c hy on’r e f e r st ot a c hy on. The r ei snos ui t a bl ee xpl a na t i onof t hee xi s t e nc eofAt l a nt i s ,s ot he r ei snot hi ngt ober e f e r r e dby‘ At l a nt i s ’ . Thi s ,howe ve r , leads to three problems. First, according to what criterion does one acquire the suitable explanatory resources by which what there really is, is explained? It cannot be our commonsense, for it is very unlikely that commonsense provides all the answers of what there is. It seems that we do not have other options but to regard a comprehensive scientific theory as the auxiliary explanatory resources. But it is more than apparent that different theories are very likely to commit us ontologically to different entities. For instance, for a disquotational schema of reference based on a comprehensive theory within which the theory of phlogiston is a br a nc h,‘ phl og i s t on’r e f e r s ,a ndr e f e r st ophl og i s t on.Sowepr oba bl yha vet or e s or tt ot h e notion of the division of linguistic labour to depend for the suitable selection of our 25 c ompr e he ns i ves c i e nt i f i ct he or yo ne xpe r t s ’pr of e s s i ona lknowl e dge . This leads us to the second problem. I ts e e mst ha tHor wi c h’ sde f l a t i on a r ya c c ountofr e f e r e nc e ,ba s e de s s e nt i a l l yont he disquotational schema of reference and satisfaction, cannot explain that in some cases different theories are actually logically equivalent, if not identical, but formulated in 25 In Horwich 1998a, Hor wi c hbor r owsPu t n a m’ si de aof“ t h edi v i s i onofl i ngu i s t i cl a bou r ”t oe x pl a i nwhy the regularities of experts, with different specialised knowledge, are the paradigm samples of the regularities of the use of all kinds of words. So it seems quite reasonable to use this notion to explain why s omewor dsdon otr e f e ri nt e r msofs omee x pe r t s ’s pe c i a l i s ed knowledge of whether or not there is something to be referred to. 94 different languages, or even the same language. For the auxiliary explanatory resources are supposed to be born from the selected comprehensive theory. Suppose there are two comprehensive theories T1 and T2, and according to both of them there are electrons and mol e c ul e s . Ac c or di ng l y ,t ha t‘ e l e c t r on’r e f e r st oe l e c t r onsa ndt ha t‘ mol e c ul e ’r e f e r st o molecules are axioms of the disquotational theory of reference. Suppose it is further found out that T1 and T2 are logically incompatible in the sense that what T1 says about electrons is about molecules in T2 and what T1 says about molecules is about electrons in T2, and vice versa. In ot he rwor d s ,Qui ne ’ sr e c ons t r ua lofpr e di c a t e sc a nr e nde rT1 and T2 to be logically equivalent.26 Regarding a disquotational reference scheme as a genuine reference scheme seems to undermine our strategy of reconstrual of predicates. This strategy, according to Quine, does not only work for one pair of predicates but many ( Qui ne1975,p.320) .I ts e e msr e a l l ydi f f i c ul tf orHor wi c h’ sdi s quot a t i ona la c c ountof reference to utilise the strategy of reconstrual of predicates to discern the logical equivalence of the two theories. Further, some cases are much less trivial than the r e c on s t r ua lof‘ e l e c t r on’a nd‘ mol e c ul e ’ .ForQui ne ,t her e c ons t r ua lofpr e di c a t e si s a ppl i c a bl et oe ve nPoi nc a r é ’ si ns t a nc et ha tonef or mul a t i onofc os mol ogyt a ke ss pa c ea s finite, but depicts all objects as shrinking in proportion as they move away from centre, while another formulation takes space as infinite. This is not doable if one considers Hor wi c h’ sdi s quot a t i ona ls c he maofr e f e r e nc ea spr ovi di ngage nui ner e f e r e nc es c he me . 26 For reconstrual of predicates, Quine says: [T]he general way of reconstruing an n-place predicate is by supplying an open sentence in n variables, not caring whether there happens to be a word in our language with the same extension as the open sentence. By a reconstrual of the predicates of our language, accordingly, let me mean any mapping of our lexicon of predicates into our open sentences (n-place predicates to n-variable sentences). (Quine 1975, p. 320) I nt h ec a s eof‘ e l e c t r on ’a n d‘ mol e c u l e ’ ,t h epr e di c a t e s‘ mol e c u l e ’a n d‘ e l e c t r on’c a nbema ppe dt o t h er e s pe c t i v eope ns e n t e n c e s‘ xi sa ne l e c t r on ’a n d‘ xi sa nmol e c u l e ’ . 95 Fina l l y ,t het hi r dpr obl e mi smor es e ve r et ha nt hef i r s tt wo.Hor wi c h’ sdi s quot a t i ona l account of reference cannot account for some very important facts of reference. In the case of evaluating a neoteric scientific theory discussed toward the end of last section, one cannot properly evaluate any neoteric theory if one always sticks to the original theory. Thei ns t a nc er a i s e dbyFi e l di n“ The or yCha ngea ndt heI nde t e r mi na c yofRe f e r e nc e ” (Field 1973) is a good example showing that some important fact of reference is not 27 e xpl i c a bl ei nt e r msofHor wi c h’ sa c c ountofr e f e r e nc e . That is, for Newtonian physics ‘ ma s s ’di s quot a t i ona l l yr e f e r st oma s s ,butf orEi ns t e i ni a nphy s i c s‘ ma s s ’i sa na mbi g uous term referring either to rest mass or to relative mass. Or, consider Aki ba ’ si ns t a nc ei n “ Ca nDe f l a t i oni s mAl l owf orHi dde nI nde t e r mi na c y ? ”( Aki ba2002) : The pre-Ne wt oni a ns ’t e r m‘ h e a v i e rt h a n ’i st obec on s i de r e di n de t e r mi n a t ebe t we e nmore weighty than and more massive than from post-Newtonian viewpoint; the pre-Newtonians themselves did not have this distinction.28 (p. 225) In other words, there are examples of hidden indeterminacy of reference in pre-Newtonian formulation of physics.29 It is hard to see how a deflationary theory of r e f e r e nc es uc ha sHor wi c h’ sc a na c c ountf ort hi spr obl e m.ForHor wi c h,‘ ma s s ’r e f e r s ,i f a ta l l ,t oma s s ;‘ r e s tma s s ’r e f e r s ,i fa ta l l ,t or e s tma s s ;a nd‘ r e l a t i vema s s ’r e f e r s ,i fa ta l l , t or e l a t i vema s s .Buthowt oe xpl a i nt her e f e r e nt i a li nde t e r mi na c yofNe wt on’ st e r m ‘ ma s s ’ ?I ti sve r yl i ke ly that some referential indeterminacy is hidden in our formulation of the selected comprehensive theory too, and it will always be possible for a future t he or yt oc ont a i nhi dde ni nde t e r mi na c y .Hor wi c h’ sdi s quot a t i ona la c c ountofr e f e r e nc ei s too deflationary to account for referential indeterminacy. A disquotational schema of 27 Wh a ti n t e r e s t smei son l yFi e l d’ si ns t a n c eoft h ec ompa r i s onbe t we e nNe wt on i a n‘ ma s s ’ / ma s sa n d Ei n s t e i n i a n‘ ma s s ’ / ma s sbu tnoth i sopi n i onsont h ei s s u e sbe h i n dt h i si n s t a n c e . 28 Th i si sa l s oapr obl e mi nHor wi c h’ sus e -theoretic account of meaning. I will discuss this in the following chapter. 29 A hidden indeterminacy by definition is not a manifest indeterminacy of a given language, such as vagueness, relative to a particular time period. The hidden indeterminacy is also objective—that is, it is not decidable in terms of our awareness of it. 96 reference is useful in scientific reasoning, but serious problems emerge when it is seen as providing a genuine reference scheme. So I can conclude that a disquotational schema of refe r e nc ec a nnotpr ovi deHa l ba c h’ sa c c ountt hene e de dr e f e r e nc es c he mef ore xpl i c a t i ng the disquotational concept of truth in terms of the inductive axioms of reference and satisfaction. And it is a mistake for Horwich as well as for Halbach to see a disquotational account of reference as providing a genuine reference scheme. Final Remarks Thi st e mpor a r i l yput sa ne ndt omya na l y s i sofQui ne ’ sdi s quot a t i ona lt he or yoft r ut h. Many issues are left without discussions until Chapter Five in order for a more integrated a na l y s i s .Ne xtont hea ge ndai sPa ulHor wi c h’ smi ni ma l i s m.I fonec ommi t sone s e l ft ot he intelligibility of the concept of a proposition, and indeed the existence of propositions, it seems that minimalism is attractive. In the next chapter, I will summarise what mi ni ma l i s mc ons i s t si na nd,mor ei mpor t a nt l y ,Iwi l ldi s c us sHor wi c h’ sus e -theoretic theory of meaning according to which the legitimacy of a minimalist theory of propositional truth can be well established. 97 Chapter IV On Horwi c h’ sMi ni mal i s tThe or i e sof Truth and Meaning Re a di ngt hef i r s tc ha pt e rofPa ulHor wi c h’ sTruth (Horwich 1998b), one can almost feel the enthusiasm in his attempt to show how concise the elaboration of the minimal theory of truth is. The chapter has only eight pages to expound to the minimal theory. Of course, however, the theory is not that concise after all. After the first chapter, there are six more to go. Although very many articles are devoted to the debate over deflationary theories of truth, it is not the task of this chapter to take a stand on these matters. Instead, this chapter aims to offer a brief introduction of minimalism. For minimalism has little to distinguish it from disquotationalism, especially in terms of my analysis in the following chapter, except that for a minimalist like Horwich, the primary truth bearers are propositions. This difference is, however, certainly significant with respect to the problem of the expressibility of the truth predicate discussed in Chapter Two. For his theory of truth to be philosophical interesting, Horwich has to show us that it is possible to come up with an account of propositions without any involvement of the concept of truth. In considering this issue, this chapter consists of four sections. The first section is a brief introduction of t hemi n i ma lt he or yoft r ut h. Thef ol l owi ngs e c t i onr e l a t e st oHor wi c h’ sc r i t i c i s mof Qui ne ’ ss c e pt i c i s mc onc e r ni ngme a ni ng . Thet hi r ds e c t i ona na l y s e sa ndc r i t i c i s e st he c e nt r a li de aofHor wi c h’ sus e -theoretic account of meaning. And the final section considers his account of compositionality. 98 1. The Framework of the Minimal Theory of Truth Thef o r mul a t i onofHor wi c h’ smi ni ma l i s ta c c ounti si nde e ds i mpl e .Hor wi c h( 1998b) claims that for any declarative sentence (1) p there is an equivalent sentence ( 1’ )Thepr opos i t i onthat p is true, and the equivalence schema i sg i ve nwhe nc ombi ni ng( 1)a nd( 1’ )wi t hal og i c a l connective of biconditional—that is, The proposition that p is true if and only if p.1 Or following Horwic h’ swa ybyp ut t i ngt hedummys y mbol‘ p’i nt oa ng l ebr a c ke t st o represent a proposition, we can convert the equivalent schema into (E) <p> is true if and only if p. I ns t a nc e sof( E) ,s uc ha s“ <Snowi swhi t e >i st r uei fa ndonl yi fs nowi swhi t e ” , constitute not the axioms but the expressions of them of the minimal theory of truth, for according to Horwich, the axioms are also propositions.2 He further explains that axioms such like “ <<Sn o wi swhi t e >i st r uei fa ndonl yi fs nowi swhi t e >”c a nbes e e na sa pplying the pr opos i t i ona ls t r uc t ur eor ,i not he rwor ds ,t hepr opos i t i ona lf or m of“ <<p> is true if and 3 only if p >”t ot hepr opos i t i on“ <Snowi swhi t e >” . (cf. Horwich 1998b, pp. 17-20) There are three important features of the minimal theory, or, rather, of deflationary 1 Just what sort of equi v a l e n c er e l a t i onh ol dsbe t we e n( 1)a nd( 1’ )i swor t hpu r s u i ng . Th i swi l lber e s e r v e d for the next chapter. 2 However, in the following the distinction between the expression of an axiom and the axiom itself is dropped for the sake of simplification in some discussions. 3 This is based on what Horwich calls the Fregean Principle. Please see section four of this chapter. In a ddi t i on ,Hor wi c ht a k e st h epr opos i t i on a lf or mss u c hl i k e“ <<p> is true if and only if p >”a sun i v e r s a l propositions. 99 theories in general. First, every instance of the equivalence schema is an axiom of the minimal theory. Second, the entire conceptual and theoretical role of truth is completely explicable in terms of these axioms and the equivalence schema (E) (Horwich 1998b, p. 5).4 Finally, all the facts involving truth and certain other matters can be also explained on the basis of the minimal theory, along with appropriate theories of those other matters (p. 7). Disquotationalism, on the other hand, will be totally unattractive to philosophers, and even undeserving the name of deflationism, if it does not conform to these three features. The major task of this chapter, then, is to specify the major difference between minimalism and disquotationalism. Except for this difference between them, and the pe c ul i a rc ons e que nc e soft he i rdi ve r g e nc e ,t hepr obl e msr e g a r di ngHor wi c h’ smi ni ma l i s m a ndQui ne ’ sdi s quot a t i ona l i s ma r es i mi l a r . Thi si swhyt he s epr obl e mswi l lber e s e r ve df or the next chapter. But first let us examine whether or not it is possible to formulate the minimal theory di f f e r e nt l y .Hor wi c ha ppa r e nt l y‘ g oe si nf i ni t e ’whe ne xpl i c a t i nghi smi ni ma lt he or y ,a nd he explains why this is inevitable in accounting for semantic concepts (cf. Horwich 1998b, the fifth objection). Matthew McGrath thinks otherwise.5 Taking propositions as objects, we can use objectual quantifications to form finite theories of propositional truth. In “ We a kDe f l a t i oni s m”( Mc Gr a t h1 997b) ,f ol l owi ngHor wi c h’ si de a ,Mc Gr a t ht hi nkst ha t propositional forms or structures are functions from entities to propositions, and are expressed by schematic sentences (p. 73). For instance, the propositional form <p & q> is 4 For some general fact regarding truth it is inevitable to employ schemata to do the explication. Mc Gr a t h’ st h e or yi samodi f i c a t i onofwha tEr n e s tSos apr opos e si n“ Th eTr u t hofMode s tRe a l i s m” (Philosophical Issues 3, Science and Knowledge, pp. 77-95). The finite theory proposed by Sosa is as follows: For all propositions P, P is necessarily equivalent to the proposition that it is true. The corresponding principle for deriving instances is (PE) If <p> entails <q>, then if p, then q. 5 100 a function, CONJ, which, given propositions P and Q, returns a proposition CONJ (P, Q) (ibid.). Similar functions include DISJ for disjunction, IFF for bi-conditional, and so on. Most importantly, the truth predicate is seen as a function, namely, Tr. Tr is to be the propositional function that takes an entity and returns the proposition with respect to it, that it is true. (E) then can be seen as a function E-prop, which is decomposable into the functions IFF and Tr in such a way that for all propositions P, E-prop(P) = IFF(Tr(P), P) ( p.74) .Mc Gr a t h’ sf i ni t et he or y ,weak deflationism is as follows: (FTP) For all propositions P, P is materially equivalent to Tr(P). In order to derive instances of (FTP), McGrath introduces a principle of material implication: (PMI) If <p> materially implies <q>, then, if p, then q. If this is doable, then the minimal theory should be reformulated in this fashion, providing s omeunpr obl e ma t i cpr e mi s e s ,s uc ha s“ <Snowi swhi t e >=EXP( ‘ Snowi swhi t e ’ ) ”a nd 6 “ Tr ( ‘ Snowi swhi t e ’ )=<‘ Snowi swhi t e ’i st r ue >”( i bi d. ) . But is it? Whether or not the material e qui v a l e nc er e l a t i onbe t we e n‘ P’a nd‘ Tr ( P) ’i sf e a s i bl ei snot what concerns me here.7 What interests me is whether or not (FTP) along with (PMI) successfully formulate the minimal theory. According to Adam Kovach, however, this is nott hec a s e .I n“ De f l a t i oni s ma ndt heDe r i va t i onGa me ”( Kova c h1997) ,Kova c ha r g ue s 6 McGrath also provides finite theories of sentence, belief, and utterance. The basic idea is to use a function ‘ EXP’t oe x pl a i nt h er e l a t i onbe t we e napr opos i t i ona n dt h es e n t e n c ee x pr e s s i ngi t . Th ef unc t i on ,EXP , “ t a k e sa ne n t i t ya sa r g ume n ta n dr e t u r nst h epr opos i t i oni te x pr e s s e s( Mc Grath 1997b, p. 74). According to McGrath, theories of utterance and belief truth can be accounted for in terms of the theory of sentence truth. Details of this will not be discussed in the following. However, there are some complications of this account. According to McGrath, weak deflationism inflates sentence, utterance, and belief truths while keeping propositional truths deflated. That is, the first three need an inflation of the theory of meaning to ensure the EXP function is doing its work properly. Putting aside these complications, however, there is a serious flaw in this account, which will be discussed shortly. 7 This will be discussed in Chapter Five. 101 that (PMI) has to either be a generalisation governed by substitution quantifications or go i nf i ni t ec a s ebyc a s e ,f ori tc ont a i nsbot hna mepos i t i ons ,t ha ti s ,t hepa i rof‘ <p>’a nd ‘ <q>’ ,a nds e nt e nc epos i t i ons ,t ha ti s ,t hepa i rof‘ p’a nd‘ q’ .ButMc Gr a t hr e j e c t st hei de a of formulating the finite theory in terms of substitution quantifications. So it is, after all, inescapable that the finite theory has to go infinite. For instance, in explaining the capacity of his finite theory in accounting for the general fact regarding truth that, for all propositions P, Q, if P entails Q and P is true, then Q is true, McGrath gives us the following argument. Three further principles are needed. That is, (P1) For all propositions P, Q, if P entails Q, then P materially implies Q; and (P2) For all propositions P, Q, CONJ(<P materially implies Q >, P) entails Q; and (P3) If CONJ(<p>, <q>) entails <r>, then if p and q, then r.8 The argument runs as follows: 1. For all P, Q, CONJ(<P entails Q >, <P is true>) entails CONJ(<P materially implies Q >, <P is true>) 2. For all P, Tr(P) is necessarily equivalent to P Assm., (P1) Assm., Weak Defl. 3. For all P, Q, CONJ(<P entails Q >, <P is true>) entails CONJ(<P materially implies Q >, P) 1, 2 4. For all P, Q, CONJ(<P entails Q >, <P is true>) entails Q 5. For all P, Q, CONJ(<P entails Q >, <P is true>) entails <Q is true> 8 ( P3)i sac on s e qu e n c eofSos a ’ s( PE) ,a c c or di n gt oMc Gr a t h .Pl e a s es e ef oot n ot e5. 102 3, (P2) 2, 4 6. For all P, Q, if P entails Q and P is true, then Q is true 5, (P3) The problem comes from the dubious (P3), which has the exact problem of (PMI). It is obvious that in (P1) and (P2) we have objectual universal quantifiers and not in (P3). So wi t hi n( P3)t he‘ p’ ,‘ q’ ,a nd‘ r’n o ti na ng l ebr a c ke t sa r es e nt e nc e sr a t he rt ha npr opos i t i ons . Therefore, the quantifier occurring in 6 has to be a substitution one. Similar mistakes appear in every argument provided by McGrath while utilising (PMI) or principles in a similar fashion. In a footnote, McGrath indicates that Horwich goes for a principle similar to (PMI) in proving that minimalism can account for some general facts regarding truth.9 What McGrath fails to appreciate is that Horwich consistently goes infinite in explicating minimalism. Most importantly, Horwich actually does not use a principle similar to (PMI). It is nothing but a typographical error. Howe v e r ,Kova c h’ sobs e r va t i oni snotne wst oMc Gr a t h.I nhi sLonger Notes of“ We a k De f l a t i oni s m” ,numbe r2t obee xa c t ,Mc Gr a t hr e f e r st oDa vi ds onwhowr i t e si n“ The Fol l yofTr y i ngt oDe f i neTr ut h”a sf ol l ows : Why, though, does Horwich not try generalizing his schema by quantifying over propositions? The answer should be: because then we would have to view ordinary sentences as singular terms referring to propositions, not as expressing propositions. (p. 273) Da vi ds on’ sc l a i m,a c c or di ngt oMc Gr a t h,i spe r ha psj us tt ha tHor wi c h’ ss c he ma( E)i snot possible to generalise without utilizing sentential variables—that is, the eligible g e ne r a l i s a t i onof( E)i s“ Fora l lp, <p> is true if and only if p” .Mc Gr a t hbe l i e ve st ha tt hi s claim is falsified by his finite theory of propositional truth. Mc Gr a t h’ sbe l i e fi s ,a c c or di ngt oKova c h,i l l -founde d.Fort hi sMc Gr a t ha r g ue si n“ Re pl y 9 Please cf.t h es e c on ds e c t i onoft h ef ol l owi ngc h a pt e rf orHor wi c h’ sa r g ume n t . 103 t oKova c h”( Mc Gr a t h1997a) that (PMI) can either be taken as inferential principles or as providing support for an unstated generalisation. (PMI) can be read as inferential principles permitting, for instance, the inference that is allowed from formula of the form “ <p> materially implies <q>”t ot hec or r e s pondi ngf or mul a“ I fp, then q”( p.582) .Or , alternatively, (PMI) can be read, for instance, in the case that true conjunctions have true conjuncts, as providing support for the following generalisation. (G) For all P, Q, if <CONJ(P, Q) is true> materially implies <P is true>, then if CONJ(P, Q) is true, then P is true. (ibid.) The supposed clarification being accomplished by reading (PMI) as inferential principles or providing support to an unstated generalisation does not throw light on any prospect of clarifying (PMI). Reading (PMI) as inferential principles has exactly the same problem mentioned above, namely, appealing implicitly to substitution quantifications. In Between Deflationism and Correspondence Theory (McGrath 2000), McGrath explains what he means by reading (PMI) as a principle of inference.10 In terms of (PMI), as an inferential pr i nc i pl e ,onec a ni nf e r“ I fHe s pe r usi sapl a ne t ,t he nt he r ei sapl a ne t ”f r om “ The pr opos i t i ont ha tHe s pe r usi sapl a ne te nt a i l st hepr opos i t i ont ha tt he r ei sapl a ne t ” (McGrath 2000, p. 31). But this falls short of its explanatory purpose. To avoid the utilisation of substitution quantifications in (PMI), one has to go case by case to include a corresponding inferential principle for each inference.11 On the other hand, the only sensible way to read (G) without any involvement of substitution quantifications, it seems, is that, for all P, Q, if <CONJ(P, Q) is true> materially implies <P is true>, then IF(Tr(CONJ(P, Q)), Tr(P)).12 But this reading is still about schematic sentences and is 10 Mc Gr a t ha c t u a l l yu s e sSos a ’ s( PEI )i nh i se x pl a n a t i on .Bu tt h epoi n ti st h es a me . I ts e e mst h a t( PMI )u n de rt h i sc on s i de r a t i onboi l sdownt oHor wi c h’ sc r i t i c i s mofs u bs t i t u t i on quantifications. Please cf. Horwich 1998b, pp. 25-26. 12 ‘ I F’i sa l s oapr opos i t i on a lf un c t i onf or‘ i f -t h e n’ . 11 104 t ot a l l ydi f f e r e ntf r om ( PMI ) . Ac c or di ngt oMc Gr a t h’ snot a t i on,‘ <p>’me a nsapr opos i t i on i n( PMI )a nd‘ p’doe snot .Ot he r wi s e‘ <p>’s houl dbet a ke na s‘ <<p>>’ ,i f‘ p’me a nsa pr opos i t i onr a t he rt ha nas e nt e nc e ,but‘ <<p>>’doe snote ve nma kes e ns e .I ft hi si sa c or r e c ti nt e r pr e t a t i onof( PMI ) ,t he n“ i fs nowi swhi t et he ng r a s si sgr e e n”s houl df ol l ow f r om“ <Snowi swhi t e >ma t e r i a l l yi mpl i e s<Gr a s si sg r e e n>” .Si mi l a r l y ,“ i f<s nowi s whi t e >i st r ue ,t he ns nowi swhi t e ”f ol l owsf r om “ <<Snowi swhi t e >i st r ue >ma t e r i a l l y i mpl i e s<s nowi swhi t e >” .But‘ s nowi swhi t e ’a nd‘ gr a s si sg r e e n’noti na ng l ebr a c ke t s are sentences. In order to derive instances from (FTP), (PMI) has to utilise substitution quantifications to do the generalising work. After showing that the minimal theory is in no position to go finite, it is time to move on. It is clear that minimalism and disquotationalism diverge in virtue of appealing to different truth bearers. Horwich rejects the idea of taking sentence types of a given l a ng ua g ea st r ut hbe a r e r s ,ont heba s i st ha tt hes e nt e nc et y peofEng l i s h,s a y ,“ Ia m 13 hung r y ”i st r ueononeoc c a s i onbutf a l s eona not he r . It is the content of a sentence type that is true or false. Curiously enough he does not rest his criticism on the usual grounds that the notion of a language is elusive, or that the relation between a sentence type and its content is contingent. As indicated in the second chapter, the disquotational schema within the additional truth language is the necessary device to disquote a given sentence type of an interpreted language to perform semantic ascent. Thus the two sentences are the same sentence with exactly the same empirical content. An explanation for how we can do this is required, but it is not reasonable to simply assume that the sentence in 13 Horwich does not use an indexical instance to make this point. But the instance he uses to indicate that it is the content that is true or false, is ex pr e s s e dbyaQu i n e a ne t e r n a l i s e ds e n t e n c et y pe“ Os c a rwa sh ung r ya t mi dda yon1J a n u a r y1988”( Hor wi c h1998b, p. 16). The problem of indexicality, it seems, is at least one of the reasons for him to have reservations about taking sentence types as truth bearers. 105 quotation marks and the sentence disquoted have different empirical contents. The reason that Horwich does not fall for this line of criticism probably has something to do with a similar consideration.14 For it is also indispensable for him to identify what a particular proposition is, in terms of an interpretation of the sentence type of a given language expressing it. It just does not do to say that a linguistic expression expresses a particular pr opos i t i on,bya ddi ng“ t hepr op o s i t i ont ha t ”i nf r ontofi t .Howe ve r ,Hor wi c h’ sc r i t i c i s m doe snots e e mf a i rbe c a us ei nde xi c a l i t yi se ve r y body ’ spr obl e m. Perhaps it does not have to be fair anyway. Intuitively speaking, if there is an appropriate theory of proposition, minimalism does seem to be preferable to disquotationalism. It has two advantages over disquotationalism. First, minimalism can provide a more general concept of truth; one applicable to linguistic items (regardless even of the language they are in) that express one and the same proposition. Second, the expressibility problem explicated in Chapter Two seems to be solved if the primary truth bearers are propositions. Apparently, all instances of the generalising proposition can be derived from it, and vice versa. At least, this was held at one time by Horwich, in Horwich 1998b.15 Of course in the present case, this theory of proposition has to be completely independent from any involvement of the concept of truth. Otherwise it is circular. But, on the other hand, we should not forget that the virtue of disquotationalism is its absolute separation from any theory of meaning. In other words, given an interpretation of all sentences of a given 14 When explaining how to take utterances as truth bearers (Horwich 1998b, p. 100-101), Horwich uses the device of star marks to indicate that *I am hungry*, say, is a sentence type. He further derives a schema (Du), (Du) (u *p* )→ ( u is true iff p) ,wh e r e‘ u’i sa nu t t e r a n c e ,t h ef i r s t‘ p’qu ot e dbys t a rma r k si sa s e n t e n c et y pe ,a n dt h es e c on d‘ p’i sas e n t e n c e , f r omt h edi s qu ot a t i on a ls c h e maofs e n t e n c et y pe s .Het h e nc or r e c t l yc l a i mst h a twha ti sputi npl a c eof‘ p’i s given a uniform interpretation. 15 Please cf. the second section of Chapter Five. 106 language, the concept of truth is fully characterised by a disquotational theory. Even Hor wi c h’ st he or yoft r ut hha st oa ppe a lt oa ni nt e r pr e t a t i onofas e tofs e nt e nc et y pe sofa language to interlink which sentence type with which proposition. So it seems that minimalism and disquotationalism have their distinctive and different virtues with respect to how they treat the relation between sentence types and propositions expressed by them. Thusonec a nonl ys a yt ha t ,i fHo r wi c h’ spoi ntofvi e w in Horwich 1998b holds, the advantage of taking propositions as primary truth bearers is its reconciliation with the expressibility of infinite conjunctions or disjunctions in terms of the truth predicate. However, it will be shown in Chapter Five that this is incorrect. A peculiar feature of minimalism follows from taking propositions as the truth bearers. Thoug hmi ni ma l i s mi si n s pi r e dbyTa r s ki ’ st r ut hpa r a di g m,t he r ei snos uc hahi e r a r c hyof languages similar to the case of disquotationalism, that is constituted at each stage by the target languages and the truth languages. Since every instance of the equivalence schema i sa l s oapr opos i t i on,e a c hi ns t a nc ec a na l s ot a ket hepl a c eof‘ p’i n( E)a ndf or maf ur t he r axiom of the minimal theory of truth. This process goes on and on and on. This actually makes sense, for there are all kinds of natural languages but no language of propositions. Ta r s ki ’ st r e a t me ntoft hel i a rpa r a d ox,howe ve r ,i snota va i l a bl et oHor wi c h.Hor wi c hd oe s not provide any solution to the Liar in Truth (1998b). But he certainly does not accept Ta r s ki ’ ss ol ut i on,f ort hi s“ s ma c k sofove r ki l l ”( p.41) .Ma nyi ns t a nc e sof‘ p’i n( E)a r e wi t ht het r ut hpr e di c a t e .Fori ns t a nc e ,e ve r yi ns t a nc eof( E)c a nr e pl a c e‘ p’a ndf or m another axiom of the minimal theory. This actually creates a pretty serious problem for him.16 Discussion of this will be reserved for the next chapter. 16 I tc a u s e sat e ns i onbe t we e nHor wi c h’ ss ol u t i ont ot h eg e n e r a l i s a t i onpr obl e mc r i t i c i s e dbyma nyot h e r s , f ori n s t a n c e , An i lGu pt ai n“ Mi n i ma l i s m”a n d“ ACr i t i qu eofDe f l a t i on i s m” ,and his solution to the Liar. 107 Although minimalism does not depend for its characterisation and analysis of the concept of truth on a tenable theory of proposition; it is not compatible with just any theory of proposition. For minimalism to be philosophically interesting, Horwich has to successfully provide a theory of proposition that is without any involvement of the concept of truth whatsoever. Or else one might incline to prefer disquotationalism to minimalism. Horwich does have a response to this. He endorses the idea of meaning as use. Among others, Horwich counts Quine among the advocates of the use theory of meaning, though reluctantly so. The predominant difference between their views is that they reach entirely oppos e dc onc l us i ons .Ba s e donhi sa na l y s i soft hei nf or ma nt s ’di s pos i t i onst ol i ng ui s t i c behaviours under certain occasions, Quine establishes the thesis of indeterminacy of translation/interpretation according to which to seek a theory of proposition is simply out of the question. On the contrary, Horwich holds a use-theoretic account of word meaning and then argues that a proposition expressed by a sentence can be analysed purely in terms of the meanings of its sentential constituents and the way of combining them. It is not just that the use-theoretic account is independent of the concept of truth, but the analysis of compositionality—that is, the ways of combining the meanings of the sentential constituents—is also free from any complication caused by the concept of truth. More, Horwich claims that his account of compositionality is deflationary. It is compatible with any theory of meaning. His use-theoretic account of word meaning and his account of compositionality will be analysed and criticised respectively in the third and fourth sections. But first let us come to see, in the following section, how Horwich 108 c a nqu e l lQui ne ’ squa l ms . 2. Que l l i ngQui ne ’ sQual mso rQue l l e db yQui ne ’ s Qualms? I nor de rt oc l a r i f yHor wi c h’ si nt e r pr e t a t i onoft het he s i sofi nde t e r mi na c yoft r a ns l a t i o n,i t is necessary to make plain the central idea of his use-theoretic account of word meaning. The account, according to Horwich, is inspired by Wittgenstein’ st houg htt ha tt he meaning of a word is constituted by its use, namely, from the regularity governing the deployment, or, in other words, its overall use, in terms of the sentences in which it appears (Horwich 1998a, p. 3 and p. 6).17 Put it in another way, the meaning constituting property of a word stems from the acceptance property A(x) that gives the circumstances in which certain specified sentences containing it are accepted (ibid. p.45).18 Further, it is not just that any regularity governing our acceptance of all the sentences containing a given word will do: it also has to be explanatorily basic. Tos ur mountQui ne ’ sba r r i e r ,Ho r wi c hf i r s tdi s t i ngui s he sf i vedi f f e r e ntf or msofs c e pt i c a l vi e wsc onc e r ni ngme a ni ng ,na me l y( Hor wi c h’ sc a ps ) ,MEANI NGHOLI SM, MEANING INDETERMINACY, MEANING INTERDEPENDENCE, MEANING DISTRIBUTION, and INDETERMINACY OF EVERYDAY PROPERTY.19 MEANING HOLISM does not c ons t i t ut eama j orpa r tofQui ne ’ st he s i s ,a ndt hel a s tvi e wme nt i one di snott he 17 I ti sn otc l e a rwh a t‘ ov e r a l lus e ’me a ns .I nc r i t i c i s i ngwha th ec a l l sMEANI NGHOLI SM,Hor wi c h c h a r a c t e r i s e st h i st h e s i sa ss a y i n gt h a t ,“ t h eme a n i ngofat e r mde pe n dsonevery single aspect of its overall use”( Hor wi c h1998a, p. 61; my italics) .La t e ri nhi ss umma r yoft h i spos i t i on ,h es a y s ,“ t h eme a n i n gofa word depends on its overall us e ”( i bi d. , p.6 5) . Th e r es e e mst obeadi f f e r e n c ebe t we e n‘ e v e r ys i n g l ea s pe c t ofi t sov e r a l lu s e ’a n d‘ i t sov e r a l lu s e ’ . Th ef or me rs e e mst obes u gg e s t i ngt h a te v e r ys pe a ke r ’ smood, i n t e n t i on ,orpe r s on a lh i s t or yi si n c l u de di nt h ee l a bor a t i onofag i v e nwor d’ sme a n i ng . Whe r e a st h el a t t e r doe sn ots e e mv e r ydi f f e r e n tf r omHor wi c h’ sownf or mu l a t i onofh i sus e -theoretic account, due to the e qu i v oc a l i t yof‘ de pe n don ’ .Howe v e r ,wes h ou l dov e r l ookt h i sn owa n dr e t u r nt oi tl a t e r . 18 There are many distinct features of this account, such as non-revisionism. But that is not in need of clarification in this chapter. 19 The name of the last view does not come from Horwich. He does not give this view a name. 109 indeterminacy Quine has in mind. So what concerns me here is the second, third, and fourth.20 MEANI NGI NDETERMI NACY,a c c or di ngt oHor wi c h’ sf or mul a t i on,i st hef ol l owi ng thesis: [T]here may be no objective fact of the matter as to whether a given property of a word is, or is not, part of what constitutes the meaning of that word. For example, there is no objective fact as to whether a given acceptance property is explanatorily basic. (Horwich 1998a, p. 61) The thesis of MEANING INTERDEPENDENCE can best be explicated by giving an example of Horwich—for a given pair of words, αand β, what constitutes the meanings of those words is a non-conjunctive relational fact, Rαβ, according to which the meaning-constituting properties of αand βare (y)Rxy and (x)Rxy respectively (ibid.). Apparently, αor βhaving its meaning, depends on the other to mean what it means as c ha r a c t e r i s e da bove . Thi si se s pe c i a l l yi mpor t a ntt oHor wi c h’ sa c c ountofwor dme a ni ng , for it is vital for a word to have a certain meaning-constituting property, that other words in the sentences containing it have their own meanings. MEANING DISTRIBUTION on the other hand is most essential to my discussion of Hor wi c h’ sQQQa c c ountbe c a us ei nt e r msofhi si nt e r pr e t a t i on,Qui ne ’ squa l msma i nl y result from this thesis. It is useful to explain MEANING DISTRIBUTION by means of 20 Although the last view will not be discussed hereafter, I would like to point out one minor inadvertency ofHor wi c h’ sa c c ou n t .I NDETERMI NACYOFEVERYDAYPROPERTY,a c c o r di n gt ohi m,i sabr e e di n g ground for philosophical paradoxes. This is because certain predicates are normally used within some r e s t r i c t e ddoma i n .“ [ I ] fi twe r ee x t e n de dbe y on dt h a tdoma i n,i twou l dc onf l i c twi t hot h e rus er e g u l a r i t i e s . Philosophical paradoxes often originate in thi swa y . ”( Hor wi c h1998a, p. 64) In a footnote of this point, Horwich raises the Liar as an instance—that is, our inclination to accept every instance of the equivalence schema occurs in normal contexts that preclude propositions involving the concept of truth. But according to the minimal theory of truth every instance of the equivalence schema can be substituted into the schema to form a further axiom. So the problem of the Liar could not possibly be about this point. This is probably just a slip anyway. 110 the above instance—namely, the meaning-constituting relational fact, Rxy, is symmetric so that Rxy = Ryx. In the present case, αand βonly have a joint meaning-constituting property. It follows that some pairs or clusters of words might not have a meaning-constituting property of their own (ibid. p. 62). Horwich then claims that, it seems to him, this is what Quine argues in Chapter Two of Word and Object and “ Ont ol og i c a lRe l a t i vi t y ” .Hef ur t he rc onc l ude st ha tt he r ei sa not he ra l t e r na t i vedr a wnby MEANING DISTRIBUTION according to which the idea of word meaning is not unde r mi ne dbyQui ne ’ sa r g ume nt s . Thi si s ,howe ve r ,r e a l l yas ur prising reading of Word and Object chapter two. In Chapter Nine of Meaning,“ Qu e l l i ngQui ne ’ sQua l ms ” ,Hor wi c hpr oc e e dswi t hhi s c r i t i c i s monQui ne ’ ss c e pt i c a lvi e wr e ga r di ngme a ni ng . Ac c or di ngt ohi si nt e r pr e t a t i on , Quine argues that translating a remote language is not to rely our evaluation of competing translation manuals on their equating terms of a translated language with those of translating languages on the basis of their having the same meaning. Rather, it is the capacity of various manuals to preserve assertability that distinguishes adequate manuals from inadequate ones. In other words, it is the pragmatic consideration that makes the demarcation. And there are no grounds to tell which is correct and which is not, as far as they all conform to pr a gma t i cc ons i de r a t i ons .Hor wi c he xt e ndshi sf or mul a t i onofQu i ne ’ s analysis a bit further. First, according to the pragmatic point of translation, any translation ma nua li sa de qua t ei fi tpr e s e r ve sa l lt hei nf or ma nt s ’di s pos i t i onst oa s s e nta nddi s s e nt . Second, based on the first point, it can be argued that there are many non-equivalent but nevertheless adequate translation manuals by means of which the same term of the translated language is mapped onto different non-synonymous terms of the translating 111 language. Third, Quine thereby rests his arguments on the conclusion that there are no such things as meaning entities (pp. 197-198). I nor de rt oa r g uea g a i ns tQui ne ’ squa l ms ,Hor wi c hs t a r t swi t ht hei nf e r e nc ef r omt he second step to the third. In what sense, that is, can we say that there are no meaning entities by showing that there are multiple non-equivalent but adequate translation manuals? For the sake of carrying out his criticism, Horwich believes that he has s uc c e s s f ul l yr e f or mul a t e dQui ne ’ smos timportant argument to include the following three premises (cf. Horwich 1998a, pp. 198-199). (I) Two manuals with regard to their respective translations between a remote language L and English are both adequate but are incompatible with each other in the sense that according to one of them the translation of a foreign word v i s‘ e’a nda c c or di ngt ot he ot h e ri ti s‘ e*’ .‘ e’a nd‘ e*’a r enotc o-referential for English speakers. (II) If a word has a meaning then an adequate translation of it must have the same meaning. (III) If two English (context-insensitive) referring expressions have the same meaning then speakers of English regard them as co-referential. According to these premises, it is easy to derive the conclusion that no words have a meaning. For, in terms of (I) and (II) ,‘ e’a nd‘ e*’ha vet hes a meme a ni nga nd,c ombi ni ng this with (III) ,‘ e’a nd‘ e*’s houl dbec o-referential for speakers of English. But according to (I) ,‘ e’a nd‘ e*’a r enotc o-referential. Therefore v does not have a meaning. Hor wi c h’ sc r i t i c i s mha st woma j orpa r t s ,a ga i ns t( I) and (II) respectively. He first argues that, even if (I) is granted, (II) is untenable. For the pragmatic consideration of adequate 112 translation manuals only ensures that each one of them is useful or justified. Although the amount of adequate manuals could be many, only one of them is true and thus preserves meaning (Horwich 1998a, p. 199). Modifying (II) into (II’ ) ,t ha ti fawor dha sar e f e r e n t , then an adequate translation of it must have the same referent, nevertheless does not help. Horwich has a much stronger standpoint than I do on the disquotational schema of r e f e r e nc e .Forhi m,t ha tt het r a ns l a t i ngt e r ms ,‘ e’a nd‘ e*’ ,r e f e rdi s quot a t i ona l l yt oe and e* respectively is perfectly determinate. It does not follow that v can be regarded as referring to both e and e* because they are not co-referential. So (II’ )ha st ober e j e c t e d too. Horwich then turns to criticise the first premise. In terms of proxy function, which I discussed in the third chapter, he thinks that Quine provides us unlimited instances of translations of words. Let f be a 1-1 function that takes each physical object into its ‘ c os mi cc ompl e me nt ’ .Fore a c hna me‘ a’ ,t he r ei sac or r e s pondi ngs i ng ul a rt e r m‘ a*’s uc h that a* = f(a), and for each primitive predic a t e‘ G’ ,a na ddi t i ona lpr e di c a t e‘ G*’s uc ht ha t {x}{G*[f(x) ]↔ G(x)}, a nds i mi l a r l yf orr e l a t i ona lpr e di c a t e s .I tf ol l owst ha t ,f ore a c ha t omi cs e nt e nc e‘ G(a) ’ , G(a)↔ G*(a*) . Anda nye xt e ns i ona ls e nt e nc e‘ p’i st he r e bye qui va l e ntt ot hes e nt e nc e 21 ‘ p*’ . Fors pe a ke r soft hes a mel a ng ua g e ,‘ p’a nd‘ p*’a r ec o-assertable. Suppose that D(x1, x2…)i st het ot a l i t yofa s s e r t a bi l i t yf a c t sf oras pe a ke r ’ sdi s pos i t i onst oa s s e nta nd dissent, D(p1, p2…)i se q ui va l e ntt oD(p1*, p2*…) . Wi t ha na ddi t i ona la s s umpt i on( A) : 21 If ol l owHor wi c h’ sf or mul a t i on . 113 (A) Any manual of translation that preserves assertability is adequate, one can suppose that, for a language H, which has sentences h1, h2…a nds oon,a nda language J, which has sentences j1, j2…a nds oon, ama nua lt r a ns l a t i ngf r om H to J is adequate according to (A) as far as it maps h1 onto j1, h2 onto j2,…,a nds oon.I f( A)i s correct, then (I) can be derived from (A) and the consideration regarding proxy function. In other words, given any manual M and a suitable proxy function, one can always produce another manual M’that is as adequate as M but gives referentially different words as the translations of words in the translated language. (A) is untenable, according to Horwich. Assertability preservation is too narrow a requirement for adequate translation manuals. If the pragmatic value of a translation manual M is to help us communicate smoothly with our foreign informant in the way that we substitute what we would have said with the fellow speakers of our language with the corresponding foreign utterances in terms of M, then this manual does not assure us that we will always produce correct predictions of what the informant will say under certain c i r c ums t a nc e s .Mor ei mpor t a nt l y ,wec a nnotpr e di c tt hei nf or ma nt ’ sinferential behaviours.22 Therefore, the collapse of premise (I) is the consequence of the rejection of (A). Horwich then continues to construct what he thinks is a more reasonable account of translation, based on his use-theoretic account of word meaning. Thef i r s ti mpr e s s i onofHor wi c h’ sa c c ountofQui ne ’ squa l ms ,i st hequa l mt ha tt he s ea r e r e a l l ynotQui ne ’ squa l ms .Hor wi c h’ sr e f or mul a t i oni sa c t ua l l ynot a reformulation of Qui ne ’ sa r g ume nt . Thi sr e f or mul a t i on,ormor ea c c ur a t e l yt hi sf or mul a t i on,c ome sf rom 22 Please cf. Horwich 1998a, pp. 202-206. 114 two claims of Quine—namely, reference is inscrutable, and adequate translation manuals a r ec ompa t i bl ewi t ht het ot a l i t yo ft hei nf or ma nt s ’l i ng ui s t i cdi s pos i t i onsofa s s e nta nd dissent. But these two claims do not jointly together form one argument. Further, to reformulate the alleged argument of Quine residing in Word and Object in terms of the consideration regarding proxy function is not chronologically correct. For the idea of proxy function does not even appear in Word and Object. In addition, Horwich does not s e e mt ounde r s t a ndt hei de aofpr o xyf unc t i onpr ope r l y . The r e byhi sc r i t i c i s m ofQui n e ’ s f a i l ur et opr e di c tt hef or e i g ni nf or ma nt ’ si nf e r e nt i a lbe ha vi our si smi s l e a di ng .Mor e ove r , it is dubious that Quine only takes assertability preservation as the one and only qualification of adequate translation manuals. Let us consider these points one by one. Hor wi c h’ sf or mul a t i oni sc onf us e d.Qui ner a i s e st hereductio argument with regard to the inscrutability of reference at the beginning of section 12, Synonymy of Terms, of the second chapter of Word and Object. To make this argument plain I need to first introduce the definition of stimulus meaning. The definition of stimulus meaning comes from Qui ne ’ ss pe c ul a t i onont hel e a r ni ngpr oc e s sofahomel a nguage and the radical translation of a remote language—the stimulus meaning of a given utterance is the ordered pair of affirmative and negative stimulus meanings.23 The affirmative stimulus meaning of a given sentence is the class of all stimulations, for a given speaker, which would prompt his assent to the sentence, and the negative stimulus meaning is the class of all the stimulations that would prompt his dissent from it.24 So a translated sentence and a 23 Hor wi c h’ sus e -theoretic account of word meaning heavily relies on the idea that to identify the meaning-constituting property of a word, is to find the acceptance property A(x) according to which the circumstances in which certain specified sentences containing the word are accepted are given. I assume that he does not have a strong opinion against the concept of stimulus meaning. At least it is not a serious problem in his formulati onofQu i n e ’ sa r gume n t . 24 Th ewor d‘ woul d’i su s e dt oe n a bl et h ede f i n i t i ont oi n c l u dea l lpos s i bl ev e r ba ldi s pos i t i on soft h e 115 translating one are stimulus synonymous if and only if they have the same stimulus meaning.25 The reductio argument is intended to show that, even if the stimulus s y nony mousr e l a t i onbe t we e nt hei nf or ma nt ’ sone -wor ds e nt e nc e“ Ga va ga i ”a ndour translating one-wor ds e nt e nc e“ Ra bbi t ”i sgr a nt e d,‘ ga va g a i ’a nd‘ r a bbi t ’a st e r msa r enot thereby intuitively synonymous.26 The argument runs as follows: 1. Stimulus synonymy between one-word sentences can guarantee the intuitive synonymous relation of the corresponding two terms. (the reductio) 2. I f“ Ga va g a i ”a nd“ Ra bbi t ”a r es t i mul uss y nony mous ,t he n‘ ga va ga i ’a nd‘ r a bbi t ’a r e intuitively synonymous. 3. I f‘ g a va g a i ’a nd‘ r a bbi t ’a r ei nt ui t i ve l ys y nony mous ,t he nt he ya r ec o-referential. 4. I ti spos s i bl et oma i nt a i nt hes t i mul uss y nony mousr e l a t i onof“ Ga va ga i ”a nd“ Ra bbi t ” while t hee xt e ns i onof‘ ga va g a i ’i sva r i e df r om t her e f e r e ntof‘ r a bbi t ’ . 5. Therefore, stimulus synonymy between one-word sentences cannot guarantee the intuitive synonymous relation of the corresponding terms. Let me elaborate this point a bit further. Quine eventually arrives at a definition of stimulus synonymy of terms, in terms of the definition of stimulus analytic sentences. The definition of stimulus analytic sentences is as follows: sentences are intrasubjective or intersubjective stimulus analytic respectively if and only if a speaker or all speakers of a speech community would assent to it respectively, when given any stimulation.27 Thus s pe a k e r ’ sa s s e n t i n gt o,ordi s s e n t i n gf r om,t h eg i v e ns e n t e n c e . Th e r ei s ,ofc ou r s e ,af l a v ourof counterfactuals in this explication and Quine is famous on denying that counterfactual conditionals have any philosophical significance. However, this does not deprive him of legitimately employing such a grammatical usage of the subjunctive. Quine makes it clear in the following passag e :“ i t sus e[ Th a ti s ,t h e u s eof‘ woul d’ ]h e r ei sn owor s et h a ni t su s ewhe nwee x pl a i n s‘ xi ss ol u bl ei nwa t e r ’a sme a n i ngt h a tx wou l ddi s s ol v ei fi twe r ei nwa t e r ”( Qu i n e1960,p. 33) . 25 This is not quite right. Some socially widespread collateral information might cause some sort of problem. Since this does not affect my elaboration, it is dropped for the sake of argumentation. 26 I ti sn otobv i ou sh owt oc ount“ Ga v a g a i ”a saon e -word sentence. But, again, for the argument sake, this issue is temporarily dropped. 27 For Quine, stimulus analyticity comes in degrees. For a speech community within which all members 116 t wog e ne r a lt e r ms‘ F’a nd‘ G’a r es t i mul uss y nony mousf oras pe a ke rorf ors pe a ke r sof the same speech community respect i ve l yi fa ndonl yi f“ Al lFs are Gs ,a ndvi c eve r s a ”i s stimulus analytic to her or to them respectively. This also extends to cover singular terms—t wos i ng ul a rt e r ms‘ a’a nd‘ b’a r es t i mul uss y nony mousf oras pe a ke rorf or speakers of the same speech communi t yr e s pe c t i ve l yi fa ndonl yi f“ a = b”i ss t i mul us analytic to her or to them respectively. However, the concepts of stimulus synonymy of sentences, stimulus synonymy of terms, and stimulus analyticity, are not the candidates for reconstructing their intuitive counterparts (Quine 1960, p. 66). It is clear, then, that premise (II)ofHor wi c h’ sf or mul a t i oni sa ni nc or r e c ti nt e r pr e t a t i onofQui ne ’ st he or y . Quine never rests the adequacy of a translation manual on conferring the same meaning of the translating word to the translated one. They are only held to be stimulus synonymous. It is noteworthy that what is crucial in the above argument is not the inscrutability of the reference regarding the translating words in different manuals. Instead, it is the referent of the translated wor d‘ ga va g a i ’t ha ti si ns c r ut a bl ei nt e r msoft hes t i mul usme a ni ngof “ Ga va ga i ” . Thes t i mul uss y nony mousr e l a t i onbe t we e n‘ ga va g a i ’a nd‘ r a bbi t ’a st e r ms also does not pin down the reference of the former term. Further, the reference of ‘ ga va g a i ’de pe ndsont hei ma g i ne df i e l dl i ng ui s t ’ sa na l y t i chy pot he s i s ,bywhi c ht he translation business of her is guided in the very first place. Part of the reason for introducing the argument of the inscrutability of reference is to be a preface of the discussion of analytic hypotheses. In discussing the phenomenon of the inscrutability of r e f e r e nc e ,Qui neur g e sust ha tt h es t i mul usme a ni n gof“ Ga va g a i ”doe snoti de nt i f yt he have certain specialised knowledge, there is a set of sentences that are stimulus analytic to those members but not to those that are not a member of that community. 117 r e f e r e nc eof‘ ga va ga i ’a sr a bbi t s ,unde t a c he dr a bbi tparts, or stages of rabbits. The reason is that what a gavagai is, involves quantitative but not qualitative identity. And quantitative identity is not just about the specification of the identity predicate, but it is also about the specification of the whole syntactic device, such as the singular and the plural, the copula, and the articles and pronouns, and so on. It is only possible to identify ‘ ga va g a i ’a sr e f e r r i ngt oar a bbi tr a t he rt ha nunde t a c he dr a bbi tpa r t sors t a ge sofar a bbi t , when the translation of certain syntactic device is determined. For instance, if the i nf or ma nt ’ ss e nt e nc er e ga r di ngi de nt i t yi st r a ns l a t e da s“ Is this the same rabbit as that? ” 28 t he n‘ ga va ga i ’i smor el i ke l yt obet r a ns l a t e da sr e f e r r i ngt or a bbi t s . On the other hand, ‘ ga va g a i ’c a na l s obet r a ns l a t e da sr e f e r r i ngt ounde t a c he dr a bbi tpa r t s .Fori ns t a nc e ,we t r a ns l a t et hes a mei nf or ma nt ’ ss e nt e nc ea s“ Are these the same undetached rabbit parts as those? ”Butt he s ec ons i de r a t i onsh a venot hi ngt odowi t ht hei nt ui t i veme a ni ng ( s) of ‘ ga va g a i ’a nd‘ r a bbi t ’ .Ofc our s e , t r a ns l a t i ng‘ ga va g a i ’a sr e f e r r i ngt or a bbi t s ,unde t a c he d rabbit parts, or rabbit stages is not entirely indeterminate in terms of empirical inquiry. For instance, if our linguist queries the indigenous informant on whether or not the gavagai is the same one of yesterday, in a sentence of the translated language of course, a ndi fouri nf or ma nt ’ sdi s pos i t i oni st oa s s e ntt oi t , t he na tl e a s tourf i e l dl i ng ui s tc a n knowf ors ur et ha t‘ g a va g a i ’doe snotr e f e rt or a bbi tstages. For rabbit stages have their unique position in a space-time continuum. The point is rather that asking questions regarding reference can only make sense after the related syntactic device is determined. Again, premise (II) or the modi f i e dve r s i onoft hea r g ume nti snotQui ne ’ s . The r ei sonemor et hi ngIwoul dl i ket os a ya boutt hes e c ondpr e mi s eofHor wi c h’ s 28 The underlinings indicate the related syntactic device. The same applies to the second sentence with underlined words. 118 formulation. To say that, based on the thesis of the indeterminacy of translation, Quine attempts to refuse the abstract entities of meaning is a bit overstated. What Quine offered usf r om “ TwoDog ma sofEmpi r i c i s m”t hr oug ht oWord and Object, is the sceptical view that there is no methodologically and epistemologically consistent way to account for the concept of meaning and the existence of meaning entities. If there is no consistent way to provide any tenable account regarding meaning, then this concept is the least thing we should care about while doing empirical semantics, much less the existence of meaning entities. In other words, to argue against meaning entities metaphysically is not what concerns Quine. As indicated three paragraphs back, the idea of a proxy function does not appear in Word and Object. The argument of the inscrutability of reference is later promoted by many commentators as one of the most important arguments supporting the thesis of translation indeterminacy. But the point made originally in the second chapter of Word and Object is to be the curtain raiser of the considerations of analytic hypotheses. In order to tear down the alleged argument of Quine, however, it is essential for Horwich to formulate it in t e r msoft hei de aofapr oxyf unc t i on. Ac c or di ngt oHor wi c h’ si nt e r pr e t a t i onoft hi si de a ,a suitable proxy function that preserves truths of the original theory/language maps each name, each singular term, each general term, and all kinds of predicates onto their corresponding counterparts of the background theory/language. He criticises the idea of a proxy function, as the basis of the indeterminacy of translation, by saying, Th u s ,i na ppr opr i a t ec i r c ums t a n c e sIt h i nka n ds a y“ Th a ti sar a bbi t ” ;bu tIwi l ln ott h i n k “ Th ec os mi cc ompl e me n toft h a ti st h ec os mi cc ompl e me ntofar a bbi t ” —even though I might be disposed to work it out and assent it if queried. Consequently, if we use such facts about actual usage, together with a translation scheme, to generate expectations about what foreigners will say under similar circumstances, then only one of the various 119 assertibility-preserving schemes will be right. (Horwich 1998a, p. 205) But this is a fairly strange interpretation of the idea of a proxy function. In Word and Object,ag i ve nt r a ns l a t i onma nu a lma ppi ng‘ g a va g a i ’ont o‘ r a bbi t ’doe snotba s et hi s mapping on their being stimulus synonymous as terms. It is actually because we premeditate to translate the syntactic device of the remote language in a corresponding wa y .I fwedi dt hi sdi f f e r e nt l y ,t he n‘ g a va g a i ’mi ghtbema ppe dont o,s a y ,‘ unde t a c he d r a bbi tpa r t s ’ .Pr oxyf unc t i onswi t hr e ga r dt or e l a t i vi s ation of ontology are another matter, ahi g h l yr e l a t e donepe r ha ps .I ti snons e ns et os a yt ha t‘ r a bbi t ’r e f e r st or a bbi t si na n absolute sense. It can only be asked, relative to a background theory T, to what does the t e r m‘ r a bbi t ’oft heobj e c tt he or yr e f e r ?I nde e d,wewi l lnott hi nka nds a y“ The cosmic complement of that i st hec os mi cc ompl e me ntofar a bbi t ”i fwet hi nka nds a y ,i n a ppr opr i a t ec i r c ums t a nc e s ,“ That i sar a bbi t ” . Wewi l lr a t he rt hi nka nds a y“ That is the cosmic complement of a rabbit (relative to a background theory T) . ”As ui t a bl epr oxy function maps the universe of the object theory onto the universe of the background theory. Although the background theory contains new predicates of each of the corresponding old ones, the proxy function at hand is not designed for giving a mapping of them. According to the above characterisation, even the stimulus synonymous relation of ‘ ga va g a i ’a nd‘ r a bbi t ’a st e r ms ,f orabi l i ng ua lorabi l i ng ua lc ommuni t y ,doe snotf i xt he i r reference, for the stimulus meaning does not fix the reference of the given term or terms. Re me mbe rt ha tt hede f i ni t i onofs t i mul uss y nony myoft e r msi s :“ Al lFs are Gs, and vice ve r s a ”i ss t i mul usa na l y t i c ,“ ‘ Al lg a va g a i sa r es t a ge sofr a bbi t s ,a ndvi c eve r s a ’i ss t i mul us analytic”i sa sg ooda s“ ‘ Al lg a va g a i sa r er a bbi t s ,a ndvi c eve r s a ’i ss t i mul usa na l y t i c ”f or 120 the sake of defining their stimulus synonymy. Then, why does Horwich choose to criticise the idea of proxy functions as his major launching pad? Why does he interpret this idea in this way? My best guess is that it is to dowi t hhi st he or yofr e f e r e nc e . Ana l og ous l yt omyi nt e r pr e t a t i onofQui ne ’ st he or yof reference in Chapter Three according to which the disquotational schema of reference is i ndi s pe ns a bl et oQui ne ’ st he ory, the disquotational schema of reference is essential to Hor wi c h’ st he or yt oo. Thedi f f e r e nc ei st ha tHor wi c hc ons i de r st ha tt hedi s quot a t i ona l schema of reference and all its instances provide a genuine reference scheme. More precisely, his theory of reference is that nothing more lies behind the trivial instances of the disquotational schema of reference, (R). The disquotational schema of reference (R) is: (R) Tokens of *n* refer, if at all, to n, where the star quotation marks indicate that the expression-type in quote is de-ambiguised to avoid the multiple referential relation of ambiguous expression-types (Horwich 1998a, p. 119). He further derives a conditional schema (RT) for foreign terms: (RT) *n* is the correct translation of w → ( x)(w refers to x iff x = n).29 The reason Horwich believes that there is one and only one correct translation manual is revealed in the last sentence of our quotation two pages back, namely: Consequently, if we use such facts about actual usage, together with a translation scheme, to generate expectations about what foreigners will say under similar circumstances, then only one of the various assertibility-preserving schemes will be right. (Horwich 1998a, p. 205 my italics) 29 (RT)i sa c c e pt a bl es ol on ga swec h a ng et h ewor d‘ c or r e c t ’t o‘ a de qu a t e ’a n da dd“ r e l a t i v et oama nu a lM” after bot ht h ef i r s ta ppe a r a n c eof‘ w’a n dt h el a s ta ppe a r a n c eof‘ n’ . 121 In other words, if presumably a list of all possible translation manuals in terms of suitable proxy functions is given, only one of them is correct because it genuinely matches what foreigners have in mind. But the idea of a proxy function is also applicable to home languages—that is, foreigners can also permute the universe of their language onto, say, itself. Considering his interpretation of the idea of a proxy function, Horwich seems to pick the wrong target. He should criticise the thesis of ontological relativity instead of the thesis of translation indeterminacy. For holding true both the disquotational schema of reference, in an absolute sense, and the thesis of ontological relativity is inconsistent. Hor wi c hha snotf i ni s he dhi sque l l i ngofQui ne ’ squa l msy e t .Hef ur t he rc l a i ms that the preservation of assertability is not a suitable criterion of the adequacy of translation manuals. Some, but of course not the real c or r e c tone ,wi l lf a i lt opr e di c tt hef or e i g ne r ’ s inferential behaviour. Regarding the case addressed in our quoted words, Horwich continues to say, For we infer h*[ Th a ti s ,“ Th ec os mi cc ompl e me n toft h a ti st h ec os mi cc ompl e me n tofa r a bbi t . ” ]f r omh [ Th a ti s ,“ Th a ti sar a bbi t . ” ] ,n ott h eot h e rwa yr ou n d. Wef i r s tobs e r v e( or i ns omeot h e rwa ydi s c ov e r )t h a t“ Th a ti sar a bbi t ”i st r u e ,a n dt h e n ,g i v e nt h ede f i n i t i onof h* in terms of h,wer e a s ont ha t“ Th ec os mi cc ompl e me n tofthat is the cosmic complement ofar a bbi t ”mu s ta l s obet r u e .Mor e ov e r ,t h ef a c tt h a twer e a s oni nt h i swa yi sabe h a vi ou r a l fact about us—a fact about our behaviour with respect to h and h*. Consequently a fully adequate translation manual should preserve such behavioural relations, and not merely dispositions to assent and dissent. (Horwich 1998a, p. 206) This is the strongest reason why Horwich conceives of being assertability preserving as not sufficient to identify the one and only adequate translation manual. I suspect this is t her e a s onwhy ,a c c or di ngt ohi sc l a s s i f i c a t i on,hea t t r i but e sQui ne ’ squa l mst o MEANING DISTRIBUTION. If, in termsofas ui t a bl epr oxyf unc t i on,‘ r a bbi t ’unde ra particular interpretation I1 of a language L c a nbema ppe dont o‘ t hec os mi cc ompl e me n t 122 ofar a bbi t ’unde ra not he ri nt e r pr e t a t i onI2 oft hes a mel a ng ua g e ,a nd‘ t hec os mi c c ompl e me ntofar a bbi t ’unde rI1 can bema ppe dont o‘ r a bbi t ’unde rI2, then there is a phe nome nonofMEANI NGDI STRI BUTI ONbe t we e n‘ r a bbi t ’a nd‘ t hec os mi c c ompl e me ntofar a bbi t ’wi t hr e s pe c tt ot ha tpr oxyf unc t i on. Ac c or di ngt ohi s use-theoretic account of word meaning, however, the predicat e‘ t hec os mi cc ompl e me nt ofar a bbi t ’i sac ompl e xonewhos eme a ni ngi nt ur nha st obec ons t r uc t e di nt e r msoft he me a ni ngoft hes i mpl epr e di c a t e‘ r a bbi t ’ .Het he r e f or ec onc l ude st ha tt hephe nome nonof MEANING DISTRIBUTION is nothing but an illusion. Inte r pr e t i ngQui ne ’ squa l msa s MEANING DISTRIBUTION, however, is really an illusion. Considering the idea of a proxy function, Horwich seems to go backwards in constructing the view of MEANING DISTRIBUTION. In the light of his disquotational schema of referenc e ,‘ r a bbi t ’r e f e r st o r a bbi ta nd‘ t hec os mi cc ompl e me ntofar a bbi t ’r e f e r st ot hec os mi cc ompl e me ntofa r a bbi t . Theme a ni ng sof‘ r a bbi t ’a nd‘ t hec os mi cc ompl e me ntofar a bbi t ’a r ene e de dt o make sure that these two instances of the disquotational schema come out right. Hence, it seems that their meanings also have a mapping relation similar to the one between the r e f e r e ntof‘ r a bbi t ’a ndt her e f e r e ntof‘ t hec os mi cc ompl e me ntofar a bbi t ’i nt e r msofa suitable proxy function. This, however, is not what Quine has in mind. St i l l ,Iha vet oc l a r i f yHor wi c h’ squa l msove rt hea de qua c yc r i t e r i onoft r a ns l a t i on manuals. Right after the cited words in the last paragraph, Horwich continues to argue, [A fully adequate translation manual] should translate h into a sentence of J from which J-speakers infer the sentence into which it translates h*. That is, since we infer h* from h, and adequate translation, T, should be such that J-speakers infer T(h*) from T(h). But suppose that T(h*) = j* and T(h) = j, whereas T*(h*) = j and T*(h) = j*; and suppose that J-speakers infer j* from j. Then T* will not correctly predict the inferential r e l a t i on s ,e v e nt h oug hi tma ys a t i s f yQu i n e ’ sa de qu a c yc on di t i on( A) ,pr e s e r v i nga l lt h e 123 assertibility facts. (Horwich 1998a, p. 206) The r ea r et wopr obl e msofHor wi c h’ sa r g ume nt .Fi r s t ,t hei nt a ng i bl epa r tofHor wi c h’ s argument is why the inference within our speculation regarding translation business also t a ke spa r ti nouri nf or ma nt ’ smi nd ?Whe nQui nec l a i mst ha tt her e f e r e nc eofat e r m‘ n’of a remote language is inscrutable in translation, it is the translator who finds there to be no e vi de nc et os uppor the rt r a ns l a t i onoft hewor d‘ n’a sr e f e r r i ngt om or m*. It is we who are doing the translating. We want to know which translation manual, T or T*, is correct or adequate in the light of this sort of speculation. It is we who are puzzled, not our informant. There is no such a parallel of inferential behaviours between our informant and us while doing translation. This confusion, it seemst ome ,c ome sf r om Hor wi c h’ st a ki ng the disquotational schema of reference in an absolute sense. Without recognising this, his criticism is at best misleading, if not mistaken. Thes e c ondpr obl e mi st ha tHor wi c h’ sa r g ume nts howst ha theg e t st hewhol ei dea of the indeterminacy thesis wrong. In the case of radical translation, Quine shows that in translating a remote language, the field linguist tries to establish semantic facts of the t r a ns l a t e dl a ng ua ge .Howe ve r ,Qui ne ’ st he s i si st ha t ,e ve ni fwege tt he totality of assertability conditions of our informants, the translation manual is not fixed. What we are not doing in translating a remote language is to translate its semantic theory. In terms of the linguistic practice of our informants, if there are semantic facts to be established, then we can form a general semantic theory. If the field linguist already has difficulties to fix the reference scheme of the translated language, the inference given by Horwich is to make the determination of translation more impossible. True, if the proxy function argument is valid, then reference is inscrutable and the ontology of a given language or a 124 given theory has to be relativised to be comprehended. Hence translation of another language is indeterminate. However, the indeterminacy thesis does not hinge its significance on the validity of the proxy function argument. The notion of proxy function is used to make a point regarding semantical reasoning. Therefore, to actually infer h from h* or to infer h* from h is irrelevant, is incidental. In addition, it is hard to see, under a proper understanding of the idea of a proxy function, why we cannot do it the other way round and infer h from h*?I nde e d,i nEng l i s h‘ r a b bi t ’ i ss e ma nt i c a l l ys i mpl e rt ha n‘ t hec os mi cc ompl e me ntofar a bbi t ’ ,butwhyone a r t hdowe think that we are authorised to confer this semantic feature to other languages? What sort of evidence do we have to support our assumption that it is an adequate characterisation of this semantic feature of a remot el a ng ua g e ?Thi si swha tQui ne ’ squa l msa r ea bout .I na normal vein, the field linguist might arrive at such a manual. But this is because of the projection by her, from something similar to our semantical considerations. In the vein of philosophical theorisation, this projection does not prove anything. It seems to me that what Quine has reservations about—that is, do certain semantic and synthetic features of our language coincide with other languages?—turns out to be a presupposition of Hor wi c h’ sa c c ount.30 Moreover, does Quine really accept the claim that the only criterion of the adequacy of translation manuals is assertability preservation? There are two clarifications to make. First, translation of a remote language is an empirical enterprise. Our imagined field linguist has nothing more than normal induction and scientific methods to help her form a 30 Please see the following two sections. 125 translation manual or manuals. She does not have any evidence other than the observed occasions, under which indigenous utterances are produced, and the assent/dissent di s pos i t i onsofhe ri nf or ma nt st oc e r t a i nque r y i ngi ndi g e nousut t e r a nc e s .Hor wi c h’ s qualms over the licensing of correct expectations and inferential behaviours from adequate translation manuals are only a matter of empirical inquiry. If our field linguist is competent, she will make right queries. For our field linguist, the totality of her i nf or ma nt s ’l i ng ui s t i cdi s pos i t i onst oa s s e nta nddi s s e nt ,a ndt hec or r e s pondi ngobs e r ve d occasions, are the sole depository of empirical evidence. And the translation manual she f or msi snotonl yg e ne r a t e df r omhe rs t udyofhe ri nf or ma nt s ’l i ng ui s t i cdi s pos i t i ons ,but a l s of r om he rc onj e c t ur eofhowt os e g me ntt hei ndi g e nousi nf or ma nt s ’ut t e r a nc e si n terms of her analytic hypotheses, by which the translation business is guided in the very first place. Second, assertability preservation is not the only criterion for adequate translation manuals. An adequate manual has to preserve truth. Although Quine does not explore much of this point in Word and Object, he later agrees with Davidson that this is necessary.31 In addition, it is possible for some indigenous stimulus analytic sentences to be translated into ours that are not stimulus analytic. The instance given by Quine is our translation of a community-wide be l i e fof“ Al lr a bbi t sa r eme nr e i nc a r na t e ” ,whi c hi snot stimulus analytic in English (ibid. p. 69). This consideration is certainly about truth preserving. The process of translating consists, no doubt, in confusions and misunderstanding, but this is the signature of empirical inquiry and is no more than what we are caught by with respect to normal inductive method. There will always be conflicts, if the inferences of the indigenous informants are misplayed. It is part of the field l i ng ui s t ’ st a s kt os or tt he s et hi ng sout . Wi t hr e g a r dt ot het ot a l i t yofhe ri nf or ma nt s ’ 31 Th i sc e r t a i n l yc a u s e sapr obl e mt oQu i n e ’ st h e or yoft r u t h .Iwi l ldi s c u s st h i si nt h ef i n a ls e c t i onof Chapter Five. 126 linguistic dispositions, our field linguist does not have the luxury to hope for any evidence beyond this. The common disturbance in translation, however, does not reside in the problem of pr e s e r vi ngi nf or ma nt s ’i nf e r e nc e s ,s i nc et hemi s unde r s t a ndi ngofot he r s ’l ogi c a lr e a s oni ng is easily located. The difficulty of translation involves ordinary beliefs, such as in the case of rabbits, men, and reincarnation. Consider folk Chinese religious beliefs as an example. The following are some popular Chinese superstitions: good gods always defend their disciples from attack by evil spirits, disease and disease-like symptoms are indications of demonic possessions, and good gods effect their power through their mortal mediums to repulse demonic attacks. In particular, superstitious Chinese believe that drinking a cup of water mixed with the ashes of Chinese incense sticks dedicated to good gods can cure diseases. It is clear that these beliefs put together explain why superstitious Chinese hold that it is true that ash-water has magical potency. To organise and formulate an inference from these superstitious beliefs in terms of her translations of them is not difficult for our field linguist. What is difficult is to first translate these beliefs themselves, because they are devoid of reasonable scientific justifications and because they are so different from wha tt hef i e l dl i ng ui s tbe l i e ve s . Thi se xa mpl ede mons t r a t e showi na de qua t eHor wi c h’ s recognition of the translation business is. For he thinks that the one and only adequate translation manual is always able to generate expectations regarding how the foreigner wi l lr e s pond.Hea l s ot hi nkst ha ts uc hama nua lwoul dl i c e ns eus“ t oha vee xa c t l yt he expectations we would have had if we were dealing with members of our own linguistic community—modul ot hes ubs t i t ut i onoft hec or r e s pondi ngf or e i g ns e nt e nc ef orourown” (Horwich 1998a, p. 205). What will members of the English speaking community 127 respond to with respect to the translating sentences of the beliefs above characterised, if they do not hold them true? The expectations between how our informants will respond to a query and how our fellow speakers will respond to the translating sentence of the queried one are not always the same. An adequate translation manual, according to Hor wi c h’ se l a bor a t i on,pr oba bl yj us ta mount st oag ui de bookofda i l yc onve r s a t i ons ,a nd not even a very good one. How to keep the balance between holding our analytic hypotheses, conforming to the totality of linguistic dispositions of indigenous informants, and preserving all truths held by them is the major task of the field linguist. And it is for this reason that the translation business is pragmatic. Hor wi c ht hi nkst ha tQui ne ’ sargument at best shows the indeterminacy of word meaning, but not sentence meaning (Horwich 1998a, p. 200). But it is apparent that this is not the case. The concept of stimulus synonymy is only an ersatz of its intuitive counterpart. Qui ne ’ st he s i sofi ndeterminacy of translation shows us that searching an empirical a c c ountofi nt ui t i veme a ni ngofwor dsa nds e nt e nc e si si nva i n.Hor wi c h’ sc r i t i c i s m of Qui ne ’ squa l msdoe snotf a i lbe c a us eofi t si nc ons i s t e nc ybuti t si na c c ur a c y .Hor wi c hd i d not quell Quine’ squa l ms ,buthi sa c c ountofme a ni ngi sve r yl i ke l yt obeque l l e dby Qui ne ’ squa l ms . 3. OnHo r wi c h’ sUs e -Theoretic Account of Word Meaning Hor wi c h’ st he or yofme a ni ngc ont a i nsma nya s pe c t s ,s uc ha st her e l a t i onsbe t we e nhi s account of meaning and the concepts of truth, reference, and normativity, which will not be the main focus in the following two sections. In order for a full account of proposition to be given, Horwich provides his use-theoretic account of word meaning and then 128 develops his account of compositionality, which explains how the constituents of a sentence combine together. My major task in the two consecutive sections is to analyse them respectively. Hor wi c h’ smos ti mpor t a ntf or mul a t i onofag i ve nwor d’ sme a ni ngi sg i ve nbyt he following: (M) The meaning of a word, w, is constituted by a use property, u(x), according to which the overall use of w is governed/best explained by an explanatorily basic regularity, where an acceptance property A(x) possessed by w gives the circumstances in which certain specified sentences containing it are accepted (cf. Horwich 1998a, p. 3, p. 6, p. 7, and p. 45).32 Many things within (M) are in need of clarification, but some will be set aside for the following section, such as the idea of a Fregean proposition. At issue here are, first, the relations between the use properties, regularities, and the acceptance properties. Second, what does it mean to say that one property is constituted by the other? And finally, how can we make sense of (M)? I will start with the second one. From Truth to Meaning, Horwich rejects the idea of a reductive analysis of semantic not i onss uc ha s ,i nt hepr e s e ntc a s e ,“ x means F”i sr e duc t i ve l ya na l y s a bl ei nt e r msofa relation T such that T(x, f). The reason for adopting a form of reductive analysis normally, is that it preserves logical form shared by both the reduced property and the one to which the former reduces (Horwich 1998a,p. 22) .I nt hea bovec a s eof‘ x means F’ ,t hec ommon logical form is the relational schema Rxy. Horwich c a l l st hi st he“ Cons t i t ut i onFa l l a c y ” 32 For Horwich, the meaning of a predicate is a concept—t h a ti s ,t h eme a n i ngof‘ r e d’i st h ec on c e ptRED. The illustration of this reasoning is left out because it is not relevant to my discussion below. 129 (ibid. p. 25). According to Horwich, it is always possible to reduce a superficial relational property to a monadic underlying property.33 It is clear then, that what Horwich rejects is not a reduction from one property to another, but that this reduction has to preserve the logical form of the reduced property. Further, in the philosophical vein, to say that one property is reduced to the other implies that they are identical. But this is not what Horwich has in mind. Rather, that a property P constitutes anther property Q is what makes Q reduce to P. The instance given by Horwich is that being water reduces to being made of H2O. That is, first, they apply to the same thing and, second, the facts about being water are best explained by that being water and being made of H2O apply to the same thing. AsSc hi f f e ri ndi c a t e si n“ Hor wi c honMe a ni ng”( Sc hi f f e r2000) ,t hi si snotqui t er i g htf or the constitution relation between meaning properties and use properties.34 For what has t obee xpl a i ne di ns uc hac ons t i t ut i onr e l a t i ona r enotf a c t sa boutawor d’ sme a ni ng property, but facts about the use of words that have that meaning property (p. 531). In other words, what has to be explained are facts about the acceptance of those sentences containing the word. Here I come to explicate the relations between the use properties, regularities, and the acceptance properties. Normally one will go for the property that is indispensable for the explanation of these facts to be the meaning-constituting property. That is, normally, the acceptance property A(x) possessed by a word w should be what the meaning property of w reduces to. This is, however, not the case according to Horwich. 33 The three italic wor dsof‘ r e du c e ’wi t h i nt h i spa r a g r a phi n di c a t ewh a tHor wi c hh as in mind when he talks of a reduction. However, as Schiffer argues, Horwich does not provide us an acceptable explanation of why the reduction should not happen the other way round. 34 In discussing the instance raised in the last paragraph, Schiffer says that, for properties and , constitutes if (a) and apply to the same things, and (b) facts about is explained by (a) (p. 530). Pr o ba bl ybe c a us eHor wi c hus e s‘ r e du c e ’a n dh eu s e s‘ c on s t i t u t e ’ ,h emi s pl a c e s‘ ’f or‘ ’i n( b) . 130 He says: [F]or each word, w, there is a regularity of the form [:] All uses of w stem from its possession of acceptance property A(x), where A(x) gives the circumstances in which certain specified sentences containing w are accepted. (Horwich 1998a, p.45) It is clear that what constitutes meaning properties is not the acceptance properties but the use properties, according to which certain regularities are given in the form depicted in the quoted words. There is good reason for such a detour. It is possible for a word, v, to have the conjunctive (acceptance) proper t y‘ A(x) and B(x) ’s ot ha twea r ec ompe l l e dt o wrongly admit the synonymous relation between w and v (ibid. p. 58, footnote 5). According to Schiffer, this is not exactly correct. It is not that w and v would be synonymous, but v would mean both A(x) and B(x) (Schiffer 2000, p.531). And this is not what we want to have. Moreover, Horwich claims that it follows from his theory that it is not possible to explain why a given meaning property reduces to the acceptance property to which it happens to reduce. Schiffer agrees with him on this point, because one cannot deduce that a word has the meaning property it does, from premises including that the word has the particular acceptance property that it does (ibid. p. 534). Otherwise, a word, v, with a conjunctive acceptance property is confused with another word, w, which has one of the conjuncts as its acceptance property. After we have sorted these things out, we can move to the third issue, namely, how to make sense of (M). The first entangled part is the ideat ha tawor d’ soverall use can be g ove r ne dbyar e g ul a r i t y ,e xpl a na t or i l yba s i cornot .Sc hi f f e r ’ se xa mpl ei st ha tas opr a no mi g htwa r m uphe rvoc a lc hor dsbys i ng i ng‘ os s e ous ’t e nt i me si ns uc c e s s i on( Sc hi f f e r 2000,p.531) . Thi sus eof‘ os s e ous ’i s ,a ndshould, without doubt not be governed by a r e g ul a r i t ya c c or di ngt owhi c h,s a y ,t her e s toft hewor d’ sus ei sg ove r ne d. AsSc hi f f e r 131 indicates, this is not a genuine difficulty for Horwich. When he talks about the overall use of a word, Horwich is talking about the overall use of a word that is explained by the acceptance ofs e nt e nc e sc ont a i ni ngt hewor d. Wha t ,t he n,wi l lc ounta sape r s on’ s a c c e pt a nc eofs e nt e nc e st ha tc on t a i nt hewor d?Ac c or di ngt oHor wi c h,i ti sape r s on’ s uttering the sentences containing a given word assertively to herself that counts as her private acceptances of those sentences (Horwich 1998a, p. 94). But this account does not really solve the problem. In a metaphysical sense, what Horwich elaborates is that it is impossible for two words, w and v, to be exactly alike except that they do not have the same meaning property. That the word w or v has a meaning property, m(x), is a consequence of its having a use property, u(x). On the other hand, it is also metaphysically impossible for w and v to be exactly alike except they have different use properties. A word having its use property is the consequence of its having a certain meaning property. In the light of this consideration, the meaning property of a word is metaphysically indistinguishable to its use property. That is to say, a word w has m(x) if and only if it has u(x).35 However, in an epistemological sense, the regularity governing the overall use of a given word is determined by its use in empirical inquiries, not the other way round. For, if it is the other way round, then we should rather reduce use properties to meaning properties. This amounts to saying that the overall use of a word is guided by its having certain meaning property. In other words, the meaning property of a given word w, acts like a norm that constrains the use of it. And this is what 35 Based on this feature, Horwich reduces the meaning property of a given word to its use property. This is with a shred of doubt though. As Schiffer indicates, metaphysical equivalence is a two-way street. But to reduce one to the other presupposes that there is an asymmetry between them. Horwich does not offer us a reasonable account with regard to this (Schiffer 2000, p. 535). 132 Horwich does not accept.36 Therefore, the problem is how to account for alien uses? It will not do if they are dismissed as incorrect use by the regularity. For an alien use might as well be explained by some other regularity. If the use of a word is incorrect according to a regularity of its use, then alien uses of a word are correct in the sense that they are also governed by a regularity according to which most uses of that word are also explained. It is trivial to say that, given a regularity of the overall use a word, some uses are correct if they are governed by the regularity and some uses are not if they are not governed by it. The question we asked is rather to request a justification of the distinction between what is granted to go into the use fact of a word and what is not. This leads us to the elaboration of two further questions of (M) regarding the notions of explanatory basicness and best explanation. And my suspicion is that these two do not contribute to the full determination of use properties and thus meaning properties of words, as Horwich wishes them to do. Horwich sometimes varies the content of (M) by saying that what is governed by the regularity is the basic use of a given word. In distinguishing his account from MEANING HOLI SM,whi c he mbr a c e se ve r ya s pe c tofawor d’ sove r a l lus e ,hec l a i mst ha tnot every use of a given word is included in the construction of explanatorily basic use regularity. Some use facts of a word constitute the use of a given word and some other facts do not, and what makes the distinction is the difference between explanatorily basic use properties and the rest (Horwich 1998a, p. 60). He so explains, In other words, the way to pick out the particular use property of a word that comprises wha twec a l l“ t h eus e ”i st of i n dt h eus epr ope r t yt h a tpr ov i de sthe best explanation of all the others. (ibid., my italics) 36 Cf. the first, the second, the third, the eighth, and the tenth chapters of Meaning. 133 37 Fort hes a keofc l a r i t y ,‘ t heus e ’f r om nowoni ss pe c i f i e da s‘ t heba s i cus e ’ . What concerns his account of word meaning then, is to first construct the explanatorily basic r e g ul a r i t yofag i ve nwor d’ sbasic us e .Howe ve r ,i ti si mpl i e dbyHor wi c h’ si l l us t r a t i on above that there are other regularities capable of accounting for the basic use of a given word. The only difference between these regularities is the capacity of delivering the best explanation of the not basic use of the word. His example of this use distinction invokes t hewor d‘ pl a ne t ’ .Ourdi s pos i t i ont oa c c e pt“ The r ea r eni nepl a ne t s , ”s oHor wi c he xpl a i ns , does not constitute the use—that is, the basic use—of‘ pl a ne t ’ ,whe r e a sourdi s pos i t i o nt o a c c e pt“ Pl a ne t sor bi ts t a r s ”i spl a us i bl yt hec ons t i t ue ntofthe use (ibid. pp. 59-60). He continues to elaborate this point by saying that, within the range of use of a given word, there is indeed some sort of distinction, one that is similar to the analytic/synthetic di s t i n c t i on. Thef a c tt ha twes ome t i me sa c knowl e dg et ha tt he r ei sac ha ngeofawor d’ s meaning is due to the fac tt ha tt hewor d’ se xpl a na t or i l yba s i cus epr ope r t y ,bywhi c ht he basic use of it is explained, changes (ibid.). This, however, strikes me as very implausible. The distinction between the basic use of t hewor d‘ pl a ne t ’a ndt heot he rpa r tofi t sus ei saconsequence of the scientific theory he l dt r uea tapa r t i c ul a rt i me . Wh a tg oe si nt ot heme a ni ngof‘ pl a ne t ’i spur e l ya contingent fact of scientific inquiry. In terms of Ptolemaic astronomy, all stars and planets orbit earth in one way or the other. It certainly follows that a different explanatorily basic r e g ul a r i t yg ove r nst heba s i cus eoft hewor d‘ pl a ne t ’ . The r ei snouni f i e dwa yt o distinguish the basic use of each word from what is not basic. This is probably all right. 37 Hor wi c h’ sdi s t i n c t i onbe t we e nthe use, or in other words, the basic use, and not basic use of a word is not a n a l og oust ot h ei n t u i t i v edi s t i n c t i onbe t we e n‘ l i t e r a lus e ’a nd‘ e x t e n de du s e ’ .Mu c he x t e nde du s ei se i t h e r based on the shape or pronunciations of a word, or is a deliberate misuse of it. In either case, the extended use is not explained by the literal use. 134 The genuine problem is that this is really impractical.38 The only three examples Horwich offers to elaborate his account are not helpful—t ha ti s ,‘ a nd’ ,‘ r e d’ ,and ‘ t r ue ’ . The a c c e pt a nc epr ope r t yoft hewor d‘ a nd’i sa sf ol l ows : ( AND)Thea c c e pt a nc epr ope r t yt ha tg ove r nsas pe a ke r ’ sove r a l lus eof“ a nd”i s( r oug hl y ) hi st e nde nc yt oa c c e pt“ p and q”i fa ndonl yi fhea c c e pt sbot h“ p”a nd“ q” . (Horwich 1998a, p. 45) But can (AND) account for uses similar to the italic ‘ a nd’t wos e nt e nc e sba c k ?We l l ,i t can account for the italic ‘ a nd’t hr ee sentences back, but not always for other sentences containing this word. (AND) is about the acceptance property of the logical connective ‘ a nd’i nor di na r yEng l i s h. Thi sc onne c t i ve ,howe ve r ,i sa l s ous e da sac onne c t i vet o connect two or more grammatical subjects or objects in ordinary English. I do not think t ha te ve r ys e nt e nc ec ont a i ni ngt hewor d‘ a nd’us e di nt hi ss e ns e ,s uc ha s“ Rome oa nd J ul i e ta r el ove r s , ”ort hel a s t‘ a nd ’i n( AND) ,c a nbebest explained in terms of (AND). In most cases, it just transfers a two or more places primitive predicate into a complex compound of sentences. This brings us to ponder what it means to say that the explanatorily basic use property is about the basic use of a given word. What makes certain use the basic use of a word, w, is certainly not in terms of whether or not the use property, ub(x), governing that certain use is explanatorily basic.39 This is utterly circular. For what makes ub(x) explanatorily basic is its capacity to best explain the rest of w’ sus e .Soub(x) depends for its feature of explanatory basicness, on a previous distinction between the basic and not basic use of w. 38 For instance, who can give the distinction between what sorts of dispositions of accepting sentences c on t a i n i ngt h ewor d‘ t a bl e ’a r ea bou tt h eba s i cus eofi ta n dwhi c ha r en ot ?Ca nf u r n i t u r ede s i gn e r sor manufacturers? 39 Th et i ny‘ b’i n‘ ub(x) ’i n di c a t e st h a ti ti se x pl a n a t or i l yba s i cwi t hr e g a r dt ot h eba s i cu s eofw. The same a ppl i e st ot h et i ny‘ b’i nt h en e x tpa r a g r a ph . 135 Following the discussion so far, (M) is derived from three more fundamental principles, (Ub), (Mb), and (BP). They are: (Ub) The range of use of a word, w, is divided into two categories, basic and not basic, which in turn are decided by a non-intentional way, sk(x)40; (Mb) The basic use of a word, w, is governed by an explanatorily basic use property, ub(x), according to which the basic use of w is governed by an explanatorily basic regularity, rb(x), where an acceptance property Ab(x) possessed by w gives the circumstances in which some specified sentences containing it are accepted; (BP) If the not basic use of a word, w, are best explained by the explanatorily basic use property, ub(x), then (M). (Mb) is least problematic among the three. (Ub), which is about the distinction between the basic and not basic use of a given word, and (BP), which is about the idea of best explanation, are in need of clarification. Whi l ei l l us t r a t i nghi si de aofi mpl i c i tunde r s t a ndi ng ,Hor wi c hbor r owsPut na m’ si de aof ‘ t hedi vi s i onofl i ng ui s t i cl a bour ’t oe xpl a i nwhyt her e g ul a r i t i e sofe xpe r t s ,wi t hdi f f e r e nt specialised knowledge, are the paradigm samples of the regularities of the use of all kinds of words. But there seems to be another reason for appealing to this idea. The distinction between the basic and not basic use of a given word also has to appeal to an expert who is specialised in the relevant field. For, in many cases, laymen can hardly tell the difference be t we e nt heba s i ca ndnotba s i cus eofawor d,s a y ,‘ pos t -mode r n’ . The r e f or e ,i ts e e ms r e a s on a bl et os a yt ha tt hedi s t i nc t i onbe t we e nawor d’ sba s i ca ndnotba s i cus ei sde c i d e d 40 Th ewor ds‘ n on-intentional way, sk(x) ’wi l lbef u r t h e re x pl i c a t e dl a t e r . 136 by an expe r t ’ ss pe c i a l i s e dknowl e dge ,sk(x). Although Horwich does not consider this, it seems that it is a necessary component of his theory. These two roles played by experts in Hor wi c h’ sus e -theoretic account, will be analysed in the following. As Schiffer ar g ue s ,Put na m’ si de adoe snotf ul l ye xpl a i nwhyt her e g ul a r i t i e sg ove r ni ng the overall use of a word given by different experts, would eventually boil down to the s a met hi ng( Sc hi f f e r2000,p.534) .Hor wi c h’ sr e s pons et ot hi smos tg e ne r a lc r i t i c i s mi s that what suits us is not every regularity but the explanatorily basic one .Sos omee xpe r t s ’ regularities are either best explained in terms of the explanatorily basic one or dismissed in terms of the consideration within the specialised field. This explanation, however, generates two further questions. For Horwich, if there is no objective fact regarding which is more explanatorily basic than the other, then this is the case of what he calls MEANING INDETERMINACY. This indicates that Horwich does not intend to explain away this situation in terms of some canons of scientific method, such as simplicity, explanatory power, or conservativeness. This seems right, because these canons are used on a pragmatic basis to evaluate rival theories rather than the embodied regularities. The real worry comes from the idea of best explanation. As indicated above, there is no constraint for finding only one regularity being capable of explaining the basic part of a wor d’ sus e .Butonl yoneoft he mc a nbe s te xpl a i nt henotba s i cuse of the word, according to Horwich. And if there are two equally good regularities, it will be a case of MEANI NGI NDETERMI NACY. Thus ,i nwha ts e ns edoe s‘ be s te xpl a na t i on’wor ka sa criterion to somehow fix the regularity or, in some cases, regularities to be explanatorily basic in the above characterisation? It seems to me that which explanation is the best, depends on our explanatory purpose(s). One explanation probably has its value as making 137 the whole explanatory business simple, and the other may have its value as making it easy to comprehend. If what makes a regularity explanatorily basic results from a pragmatic consideration, then there will always be another regularity that is equally explanatorily basic in accordance with a different pragmatic consideration. If this view is correct, then for a word, w, there are many use properties that can explain, if not best explain, the overall use of w. Eve ni fHor wi c h’ sa c c ountc oul ds uc c e s s f ul l ypi c koutt hee xpl a na t or i l yba s i cr e g ul a r i t y or regularities by its or their capacity of best explaining the not basic use of a given word, it would have another problem. That is, the distinction between the basic and not basic use of a word is problematic. I have argued at length that this distinction itself is in need ofc l a r i f i c a t i on,a ndt hebe s tg ue s s ,i ts e e ms ,i st oi nvokee xpe r t s ’s pe c i a l i s e dknowl e dge , sk(x), to make the distinction. Horwich does not provide us a general account of what is to c ounta sa ne xpe r t . Thi si soneofSc hi f f e r ’ si mpor t a ntc r i t i c i s msofHor wi c h’ sa c c ountof word meaning (Schiffer 2000, p. 532). It is dubious for such an account anyway. It is not problem-proof, though, if such an account is given. For it seems very unlikely for this account to discriminate in favour of some experts butnott heot he r s . Af t e ra l l ,“ t he di vi s i o nofl i ng ui s tl a bour ”i snot“ t hedi vi s i onofl i ng ui s t i cl e g i s l a t or ” ,pr ohi bi t i ngs ome s pe c i a l i s t i cus eofawor di nt e r msofs omeot he re xpe r t s ’opi ni on. Ana s t r ol og i c a le xpe r t , for instance, might make a very different distinction between the basic and not basic use oft hewor d‘ pl a ne t ’f r om a na s t r onomi c a le xpe r t .“ The r ea r eni nepl a ne t s ”s e e mst obea s i g ni f i c a ntus ef a c tr e g a r di nga s t r ol ogy .I tc e r t a i nl yf ol l owsf r omHor wi c h’ sa c c ountt ha t they have very different explanatorily basic regularities governing the overall use of the wor d‘ p l a ne t ’ . Andi ti snota na dvi s a bl el i neofa r g ume nt ,a ga i ns tt hea bovec r i t i c i s m,t o 138 c l a i mt ha tt hea s t r ol og i c a lr e g ul a r i t yg ove r ni ngt heba s i cus eoft hewor d“ pl a ne t ”i sn ot able to best explain, or even explain the rest of the use of it, especially the use fact regarding astronomy. For the same applies to the astronomical regularity. This is much worse than the cases of MEANING INDETERMINACY. The claim that there is a set of regularities, which combined together govern the overall use of a given word, seems to make sense. It is very unlikely, even in terms of commonsense, that there is a single regularity to be constructed to account for the overall use of a word. People probably switch the use property in different contexts, such as the use property of the word “ pl a ne t ”i na na s t r ol og i c a lc ont e xta ndt heot he ronei na na s t r onomi c a lc ont e xt . How to make sense of (M)? The criticism boils down to the following. First, an account of providing best explanation of something in terms of something else has to be given. Without this, (M) is not justified because which regularity is capable of providing best explanation to the not basic use of a word is very likely to depend on our explanatory purpose(s). Second, a general account of what counts as an expert has to be given. Wi t houtt h i sa c c ount ,i nvoki ng“ t hedi vi s i onofl i ng ui s t i cl a bour ”i snotc ompl e t e . Thi r d, e ve ni ft hes e c ondf e a t ur ei sf ul f i l l e d,Hor wi c h’ sa c c ounti ss t i l lnots uc c e ssful. For there seems to be no way to reconcile different opinions from different specialised fields regarding what the basic use of a given word is. If these questions are not answered, the whol ee nt e r pr i s eofHor wi c h’ se xpl a na t i onofwor dme a ni ngc ol l a pses. The reason is that, following his reasoning, either there is no way to decide which meaning property and use property are the meaning property and use property of a given word, or there seems to be no way to decide to which use property the meaning prope r t ys houl dr e duc e .Hor wi c h’ s account seems to create its own version of indeterminacy of meaning. 139 The r ei sonel a s tpoi ntIwa ntt oma ker e ga r di nghowt oma kes e ns eof( M) .Hor wi c h’ s elaboration of what is an acceptance property is less than clear. According to (M), an acceptance property, A(x), gives the circumstances in which certain specified sentences containing the word, w,a r ea c c e pt e d. Wha tdoe s‘ c i r c ums t a nc e s ’me a n?I nhi si ns t a nc eof t hewor d“ pl a ne t ” ,“ Pl a ne t sor bi ts t a r s ”i saus ef a c tt ha tconstitutes the basic use of that word. But in what circumstance does one accept this sentence? It is highly unlikely that what it means is the necessary theoretical facts for accepting this sentence, for Horwich emphasises that the explanatorily basic regularity along with other necessary apparatus can best explain the overall use of a given word. So circumstances should be something other than theoretical facts. It is not clear, then, in what circumstance does one accept “ Pl a ne t sor bi ts t a r s . ”Ac c or di ngt oQui ne ’ sdi s t i nc t i onbe t we e noc c a s i ona nds t a ndi ng sentences, which depends on how close the sentences are with their stimulus meanings, most theoretical sentences are standing sentences and thus do not have particular circumstances for their acceptance.41 Therefore, one of the genuine difficulties of Hor wi c h’ sa c c ounti showt oa c c o untf ort hea c c e pt a nc epr ope r t i e soft he or e t i c a lt e r ms . The r ea r eot he rmor es e ve r edi f f i c ul t i e sf orHor wi c h’ sa c c ountonamor ege ne r a lba s i s . Ac c or di ngt oHor wi c h’ sa c c ount , the meaning of almost every word is determinate at a particular time period of the evolving stages of a given language, except the cases of MEANI NGI NDETERMI NACY. Thi si sduet ot hef a c tt ha thea ppe a l st oe xpe r t s ’ opinions with regard to the explanatorily basic use property of every word, and these experts exist only relative to a particular time period. However, there is always hidden 41 Thepa r t i a lr e a s onf orQu i n e ’ sc on s t r u c t i n gat e c h ni c a lde f i n i t i onofs t i mul u sme a ni ngi st os pe c i f yt h e circumstances in which certain sentences are queried. 140 indeterminacy at the evolving stages of a given language, according to Ken Akiba (2002), as I indicated in Chapter Three.42 A hidden indeterminacy by definition is not a manifest indeterminacy of a given language, such as vagueness, relative to a particular time period.43 The hidden indeterminacy is also objective—that is, it is not decidable in terms of our awareness of it. Thei ns t a n c eg i ve nbyAki bai st her e l a t i ona lpr e di c a t e ,‘ he a vi e r t ha n’ ,r e l a t i vet oapr e -Newtonian language. There is a hidden indeterminacy of the use of ‘ he a v i e rt ha n’i napr e -Newtonian language—t ha ti s ,i ti si nde t e r mi na t ebe t we e n‘ mor e we i g ht yt ha n’a nd‘ mor ema s s i vet ha n’f r om t hepos t -Newtonian viewpoint (p. 225). The be s tr e p l yIc a nc omeupwi t hi nt e r msofHor wi c h’ sa c c ounti st ha tt heus epr ope r t ya nd he nc et heme a ni ngpr ope r t yof‘ h e a vi e rt ha n’i napr e -Newtonian language are determinate accordingt ot hee xpe r tore xpe r t s ’opi ni ona tt ha tpa r t i c ul a rt i mepe r i od. Later in a post-Newtonian language, the use property and the meaning property of it simply changed. So the best way to answer this question is to appeal to a change of meaning as a result of empirical inquiry in general, or scientific inquiry in particular. My intention here is not to evaluate this objection, but to question the envisaged way of replying to it. I nc r i t i c i s i ngHor wi c h’ sa c c ount ,Sc hi f f e rr a i s e save r yg oodobj e c t i on. Tha tis, if the a mbi t i onofHor wi c h’ sa c c ounti st oe xpl a i nt hepublic or communicative linguistic behaviours of speakers of a particular language, then it will flagrantly fail (Schiffer 2000, p. 535). I would like to further elaborate a related point regrading Hor wi c h’ sa c c ounto f 42 Ak i badoe sn otr a i s et h i squ e s t i onf orHor wi c h’ st h e or yofme a ni ngbu tf orde f l a t i oni s tt he or i e sof truth and reference in general. But it is suitable to make the point here. 43 Ak i bae mph a s i s e st h a tt h i sha sn ot hi ngt h i ngt odowi t hQui n e ’ st h e s i sofi n de t e r mi n a c yoft r a ns l a t i on because the case he raises is only concerning a single language. But for Quine indeterminacy even happens in the home language. Horwich agrees with Quine that the matter of meaning is at the end of the day about translation. What he does not agree with is that there is any indeterminacy with regard to this subject matter. 141 what he calls implicit understanding. On the one hand, Horwich claims that, even if a sophisticated scientific inquiry is r e qui r e di nor de rt or e a c ht hec onc l us i ont ha tt hewor d‘ dog ’ha sac e r t a i ne xpl a na t or i l y basic use property, u(x) ,pe opl es t i l lunde r s t a nd‘ dog ’pe r f e c t l ywe l l( Hor wi c h1998a, p. 16). On the other hand, he also claims that, when a member of a linguistic community has only a minimal understanding of‘ dog’ ,notaf ul lori nc ompl e t eknowl e dg eofi t ,t he communal language meaning, namely, u(x), may be correctly attributed to him (ibid., pp. 17-18) .Forpopul a rwor dss uc ha s‘ dog ’ ,‘ r e d’ ,a nde ve n‘ a r t hr i t i s ’ ,t hi smi g htwor k.For s ophi s t i c a t e dwor dss uc ha s‘ s up e r ve ne ’i n“ p supervenes on q if and only if p metaphysically necessitates q, ”i ti sdubi ous .I na ddi t i on,Hor wi c h’ se xpl a na t i onoft he distinction between full, incomplete, and minimal understanding of a word is in terms of t hede g r e eofs i mi l a r i t ybe t we e nwha ti tme a nsi none ’ si di ol e c ta ndwha ti tme a nsi nt he communal language. And this degree of similarity in turn is constituted by the degree of similarity between the explanatorily basic use property (ibid., p. 18). This is really puzzling. First of all, there does not seem to be any possibility of constructing a general technical criterion by which a speaker would be qualified as capable of understanding each word of the communal language. Secondly, it is difficult to see how the degree of similarity between the explanatorily basic use property of a word for a speaker, and the one for a community, has anything to do with the attribution of the communal language meaning of that word to the speaker. Here is an analogy. Suppose that someone bases his understanding of the rules of soccer on his understanding of the rules of Celtic football. How do we make sense of the attribution of the communal language meaning of the rules of playing soccer to him, in terms of the degree of similarity between his understanding of 142 Celtic football and his understanding of soccer? Further, since Horwich claims that words of different languages have the same meaning as far as they have the same use property, the degree of similarity of the use property is also applicable between a foreign word and awor dofmot he rt ong ue . Ta keQu i ne ’ sf a mous‘ ga va g a i ’a sa ni ns t a nc e .Eng l i s hs pe a ke r s do not have a corresponding word for it, because it is accepted in certain specified sentences containing it, when a particular kind of insect appears. A new English word, say, ‘ r a bbi t -f l y ’ ,i spr oba bl yi nvented for this reason. So English speakers have two use pr ope r t i e sr e g a r di ng‘ r a bbi t ’a nd‘ r a bbi t -f l y ’ .Le tusf ur t he rs uppos et ha tt he r ei snos uc ha distinction between rabbits and rabbit-flies in the remote language. How do we make sense of the degree ofs i mi l a r i t yb e t we e na nEng l i s hs pe a ke r ’ sus epr ope r t i e sofbot h 44 wor dsa ndouri ndi g e nousi nf or ma nt ’ sus epr ope r t yof‘ g a va ga i ’ ? Finally, if we want to make sense of why people can have public communication, then the degree of dissimilarity between speake r s ’us epr ope r t i e sa ndt hes o-called communal use property of every word should also be taken into account. In many cases, what is difficult to explain is why and how people can understand each other while having different use properties of many words. An additional point about the change of meaning in public languages should be addressed. An instance regarding two four-word Chinese idioms can illustrate this well—namely, 45 “ Mor ni ng-three-evening-f our ”a nd“ Mor ni ng -Chin-evening-Chu” . These are used indifferently by most Chinese nowadays, but they actually had a very different meaning in the old days. Let us use p and q r e s pe c t i ve l ya ss hor t ha nd.Wha t‘ p’s a y sc onc e r nst r i c ke r y . 44 Having the synonymous relation between words of different languages depend on possessing the same use property is highly dubious. The discussion of this is reserved to the following section. 45 It is hard to find the exact translations of them, so they are written in the fashion of combining the four corresponding English words with dashes. 143 It comes from a story. A person has three monkeys and he feeds them four chestnuts every morning and evening; due to his poverty he decides to reduce the amount of chestnuts; he tells the three monkeys that, from now on, they will only have three chestnuts for br e a k f a s ta ndf ourf ordi nne r . Themonke y s ’pr ot e s toft hi sde c i s i onf or c e s him to try something different. He tells the three monkeys that, from now on, they will have four for breakfast and three for dinner. The monkeys happily accept this new offer, feeling that their bargain works because the amount of chestnuts for breakfast increases. So this story is saying that the three monkeys are tricked by their master because of their simple-mi nde dne s s . Wha t‘ q’s a y si sa c t ua l l ya boutpe opl ebe i ngi nc ons t a nti na l lki nd sof relationships. It comes from a story with regard to a small country that is sandwiched between two powerful and opposing countries, Chin and Chu. In the morning the messenger of Chin was received in audience by the King of the small country, and the Ki nga c c e de dt oChi n’ spr opos a lo fa l l i a nc e .La t e ri nt hee ve ni ngthe messenger of Chu wa sr e c e i ve di na udi e nc ea ndt heKi ngc ha nge dhi smi ndt oa c c e det oChu’ spr opos a lo f alliance. The original meaning of this idiom is about double-mindedness and it is extended to mean the inconstancy occurs in all kinds of relationships . Ther e a s ont ha t‘ p’ i sus e di ndi f f e r e nt l ywi t h‘ q’t oda yi sma i nl ybe c a us et hef i r s ta ndt het hi r dwor dsoft he m are the same—t ha ti s ,‘ mor ni ng ’a nd‘ e ve ni ng’ —a ndma nype opl emi s t a ket he‘ t hr e e ’a nd ‘ f our ’i n‘ p’a s“ wi t ht hr e eorf ourwome norme n” .I t could be argued, however, that the change of meaning might be caused by many reasons. This is just one of the cases. But a c c or di ngt oHor wi c h,wes ome t i me sa c knowl e dget ha tt he r ei sac ha ng eofawor d’ s meaning that is due to the fact that t hewor d’ se x planatorily basic use property changes (Horwich 1998a, pp.59-60). Contrary to the last point I made, the change of meaning in this particular instance does not result from the change of the similarity between the use 144 property of the mass majority and the use property of the experts. It rather results from the dissimilarity between them. Ie ndt hi ss e c t i onwi t hadi s c us s i onr e ga r di ngHor wi c h’ sobj e c t i ont oKr i pke ’ ss c e pt i c a l pa r a d oxbe c a us eHor wi c h’ sr e pl yt ot hepa r a doxs e e mst os ugge s tt hedownf a l lofhi sown theory of meaning. The sceptical paradox applies to dispositionalism in general, insofar as it is claimed that the reference of linguistic expressions is not determined by our dispositions to linguist behaviours. One major instance given by Kripke is the i ndi f f e r e nc eoft heus ef a c tof‘ pl us ’ofourl i ng ui s t i cc ommuni t ya nd‘ quus ’ofa n imaginary community. The divergence between the two, by definition, is about their extensions, when numbers are ungraspably huge, so huge that none of us can talk about t he m,a nda tt hi spoi ntt h et r i pl e sofnumbe r st owhi c h‘ quus ’a ppl i e sa r es l i g ht l ydi f f e r e nt t ot ha tof‘ pl us ’ .Hor wi c h’ sobj e c t i ont ot hi si ma g i na r yc a s ei si t sque s t i onbe gg i ngna t ur e . He argues in the following way: For to insist, without argument, on there being such a possibility is simply to beg the question against the use theory of meaning. What does seem intuitively right is that there are possible complex expressions—de f i n a bl ei nt e r msof‘ pl u s ’ —whose extensions di v e r g ef r omt h a tof‘ pl us ’i nt h eslight way imagined. But no complex expression—since it will inevitably bear certain use relations to its constituents—can have exactly the same use as a primitive expression. Therefore, in order to construct his counter example to the use theory, Kripke must assume in addition that there could be a primitive t e r m( onapa rwi t h‘ pl u s ’ )t h a ti sc oe x t e n s i v ewi t hon eoft h os epos s i bl e complex expressions, and that, if there were such a primitive, its use (including dispositions concerning the divergent cases)wou l dbei de n t i c a lt oou ru s eof‘ pl u s ’ . (Horwich 1998a, p. 223) Al e xa nde rMi l l e rc or r e c t l ya ddr e s s e sapr ot e s tt ot hi sa r g ume nt ,t ha t‘ quus ’i s semantically complex but not syntactically complex, and hence has no constituents to which it might have use r e l a t i ons( Mi l l e r2000) .Hec ont i nue st oa r g ue ,us i ngGoodma n’ s 145 f a mousi ns t a nc eof‘ bl e e n/ g r ue ’ ,t ha te ve ni fHor wi c h’ spoi nti sr i g ht ,t hec a s eof ‘ bl ue / gr e e n’a nd‘ bl e e n/ g r ue ’s t i l la ppl i e s( i bi d.pp.167-168). For, in terms of the 46 f or mul a t i onof‘ bl e e n’a nd‘ gr ue ’ ,t hef ourpr e di c a t e sa r emut ua l l yde f i na bl ei npa i r s . Thewor d‘ bl e e n’i snots e ma nt i c a l l ymor ec ompl e xt ha n‘ bl ue ’i s . Howe v e r ,Hor wi c h’ sque s t i onbe gg i nga r g ume ntdoe snots e e mt obee xa c t l ywha tMi l l e r thinks it is. After the words cited above, Horwich says: But it is plausible to suppose that a foreign word could acquire such an extension only e i t h e ri twe r ede f i n e di nt e r msofs omewor dme a n i ngt h es a mea s‘ pl us ’ ,ori twe r e applied in some definite way to certain triples that, given our limitations, are beyond our de f i n i t er a n g eofa ppl i c a t i onof‘ pl u s ’ ;a n di nn e i t h e roft h e s ec a s e swou l di tbec or r e c tt o say that the use of the foreign word is exactly l i k eou ru s eof‘ pl us ’ .( i bi d. ) Thel i neofHor wi c h’ sde f e nc ei ss i mi l a rt ot heonehea dopt st oa r g uea ga i ns tQui ne ’ s proxy function in the case of indeterminacy of translation. The basic idea there is that we i nf e r‘ t hec os mi cc ompl e me ntofar a bbi t ’f r om ‘ ar a bbi t ’be c a us et hef or me ri sac ompl e x term bearing some use relations to the latter. The exact reason is of course unknown, but Horwich somehow seems to confuse these two very different argumentations as being onea n dt hes a me .I nhi sa r g ume nta ga i ns tpr oxyf unc t i on,t hel i nei sc l e a rt ha t‘ t he c os mi cc ompl e me ntofar a bbi t ’i si ndeed semantically and synthetically complex. In the c a s eo f‘ pl us ’a nd‘ quus ’ ,t he r ei snor e a s onwha t s oe ve rt ot hi nkt ha t‘ quus ’i sa s y nt a c t i c a l l yc ompl e xe xpr e s s i on . Bute xc e ptt hi ss l i p,Hor wi c h’ spoi nti sdi f f e r e ntf r om that made by Miller. Horwich see mst oa r g uet ha t‘ quus ’a nd‘ bl e e n’ ,bet he ys y nt a c t i c a l l y complex or not, are constructed from the actual pr e di c a t e s‘ pl us ’a nd‘ bl ue ’ . Al t houg hi n t hec a s eof‘ bl ue / g r e e n’a nd‘ bl e e n / g r ue ’t he ya r emut ua l l yde f i na bl e ,t hepa i rof 46 That is, X is grue iff x is green at or before a certain time t or blue after t. X is bleen iff x is blue at or before a certain time t or green after t. X is blue iff x is bleen at or before a certain time t or grue after t. X is green iff x is grue at or before a certain time t or bleen after t. 146 ‘ bl e e n / g r ue ’i sc ons t r uc t e df r omt hepa i rof‘ bl ue / gr e e n’ ,nott heot he rwa yr ound.I not he r wor ds ,unl e s swef i r s tc ons t r uc t‘ bl e e n/ g r ue ’f r om t hea c t ua lpr e di c a t e s‘ bl ue / g r e e n’ ,t he r e is no way and no need to define the latter in terms of the former. Creating some eccentric predicates in the light of the actual ones does not confer the former the same use relations between the latter and the rest of our linguistic enterprise. The idea behind this defence, it seems, is that the eccentric predicates lack the interconnections of use facts of the complicated reticular structure of our linguistic enterprise within which the actual predicates, names, singular and general terms are nodes of the structure. However, by the same token, why should the same or very similar reticular structure of use relations of onel a ng ua g ea ppl yt oa not he rve r ydi f f e r e ntl a ng ua ge ?I not he rwor ds ,Hor wi c h’ se nt i r e use-theoretic account of meaning depends for its plausibility on some contingent use facts of a language, in the present case, English. Different languages share the total contingent use facts of English. This is hardly acceptable. It will be shown in the following section t ha tt hi si smor ea ppa r e nti nHor wi c h’ sa c c ountofc ompos i t i ona l i t y . 4. Inflating Compositionality According to Horwich, to unde r s t a ndone ’ sownc ompl e xe xpr e s s i ons ,oneonl yne e ds one ’ sunde r s t a ndi ngoft he i rpa r t sa ndhowt he ya r ec ombi ne d.Mor es pe c i f i c a l l y ,a sf a ra s one has figured out how a sentence is constructed from primitive constituents and providing that one knows the meanings of those constituents, one automatically qualifies as understanding the sentence (Horwich 1998a, p. 155). Nothing else is needed for this sort of implicit unde r s t a ndi ng . Thi si st heg e ne r a li de aofHor wi c h’ svi e won compositionality. So in providing an account of word meaning and the knowledge of how words composed in complex expressions, the meanings of complex expressions are 147 revealed. Two things are noteworthy. First, since this understanding is implicit, how constituents of complex expressions combine, in terms of his principle of compositionality, does not involve process of any kind. Second, recollections of the meanings of words (complex expressions being themselves an inference or within one), the pragmatic contexts of complex expressions, and what the world is like, are not related i nonewa yora not he rt oone ’ sa u t oma t i cunde r s t a ndi ngofc ompl e xe xpr e s s i ons . The ya r e entirely irrelevant. Furthermore, Horwich claims that his principle of compositionality deserves to be called deflationary. By this he is saying that the principle of compositionality is compatible with all kinds of theories of word meaning. So the subject oft hi ss e c t i onboi l sdownt ot hef ol l owi ngs e ve r a li s s ue s .Fi r s t ,wha ti sHor wi c h’ s principle of compositionality and what is it deflationary about? Second, is this principle able to explain the use fact regarding complex expressions? Finally, does it make sense to say different languages somehow share the same set of synthetic features? Hor wi c h’ spr i nc i pl eofc ompos i t i o na l i ty can be put as follows: (PC) A synthetically complex expression has its meaning as determined by the meanings of its constituents and, together with, the manner they combine. Thef i r s tt a s ki st oe xpl a i nwha tt hemodeofc ombi na t i oni s . Tous eHor wi c h’ sfavourite e xa mpl e“ Dog sba r k”a ndhi sc a pi t a l i s i ngwa yofs pe c i f y i ngt heme a ni ng -constituting pr ope r t yt oe xpl a i nt ha tt heme a n i ngof“ Dog sba r k” ,x, results from putting terms whose meanings are DOG and BARK, in that order, into a schema whose meaning is NS V (Horwich 1998a, p. 156).47 Since the meaning-constituting properties of words reduce to 47 For notational reasons, the meanings of schemata are not just capitalised but is also italici s e d,a n dt h e‘ S’ on the other hand is not italicised in order to indicate the plural ending. This somehow foreshows the difficulty of his account. 148 their use properties, all capitalised words can also be replaced by use properties. The most i mpor t a ntpa r toft hi sf or mul a t i oni sc e r t a i nl ya boutt hes c he ma“ ns v”and its meaning, or rather its having meaning.48 The general idea behind this specification of the meaning of a complex expression is what Horwich calls the Fregean Principle. That is, (FP) The meaning of the result of applying combinatorial procedure P to a sequence of primitives = The result of applying combinatorial procedure P to a sequence of primitives. (ibid. p. 180) Horwich continues to explain (FP) in a footnote: Note that a combinatorial procedure, P, acts on any sequence of functions and objects…Th e r e f or eP can perfectly well act both on a sequence of words (including schemata) and on the sequence of their meanings. (ibid.) This is really a peculiar explanatory strategy. And it is not Fregean. John Collins in “ Hor wi c h’ sSc he ma t aMe e tSyntactic Structures”( Col l i ns2003)obj e c t st ha t ,nor ma l l y , one does not assign semantic values to schemata because schemata are abstract descriptions of classes of phrases made from schematic symbols describing classes of words.49 It is the phrases or words they desc r i beha veme a ni ng s .“ [ I ] ti snott hes c he ma t a t ha tc ompos e swor dme a ni ng s , ”Col l i nsc ont i nue s ,“ [ but ]i ti st heme a ni ng st he ms e l ve s , a si twe r e ,t ha tc ompos e ,a xi om s c he ma t ame r e l yde s c r i bet hi s . ”( Col l i ns2003,p.427) This is quite right. However, let me bypass this problem and continue my analysis. Even if the above-mentioned problem is put aside, in what sense can one say that (PC) is deflationary? The most important claim regarding the deflationary nature of (PC) is of course its compatibility with all kinds of theories of word meaning. Given that a 48 Cf .t h el a s tn ot e . Th es a menot a t i ona ppl i e st o“ ns v” . Fori n s t a n c e ,on edoe sn ots a yt h a tt h ee qu i v a l e n c es c h e ma“ Th epr opos i t i onthat p is true if and only if p”i st r u ebu tt h a ti t si n s t a n c e sa r e . 49 149 Da vi ds oni a nt r ut hc ondi t i ona la c c ounti st hema j orr i va lofHor wi c h’ sa c c ount sofwo r d meaning and the principle of compositionality, and a Davidsonian account falls short of an inflation of the theory of t r ut h,a c c or di ngt oHor wi c h,a na c c ountl i keHor wi c h’ swi t h no such inflation is hence deflationary. This seems to be the implicit reasoning of Horwich. However, this is only correct when a Davidsonian account defines compositionality in terms of satisfacti on.I ft hi si st hec a s e ,t he nHor wi c h’ sa c c ounti si n this sense deflationary. But it is not. On the contrary, that such a Davidsonian account does not employ schemata, much less conferring meanings to them, is a virtue of it. Jerry Fodor and Ernie Lepore are two strident critics of the claim of the compatibility between the deflationary nature of (PC) and all kinds of theories of word meaning. Among various arguments they made together or separately, two are discussed in the following. The first one is about principle of uniformity: (PU) The semantic values of phrasal expressions of syntactic category c are constituted in the same way as the semantic values of those of its lexical parts of category c.50 (PU) and (PC) jointly imply that the stereotype theory of word meaning is false. (PC) says that the meanings of complex expressions are constructed from their constituents. (PU) says that the meaning of any phrasal expression is constituted in the same way as its constituents are. Take a stereotype theory as a ni ns t a nc e ,t heme a ni ngof‘ pe tf i s h’ ,s a y , a c c or di ngt o( PU)i sa l s oas t e r e o t y pe . Thes t e r e ot y peof‘ pe tf i s h’ ,howe ve r ,i snot 51 c ons t i t ut e dbyt hes t e r e ot y pe sof‘ pe t ’a nd‘ f i s h’ . So if (PC) and (PU) both hold, any 50 Th i sf or mu l a t i oni sPh i l i pRobbi n s ’ si n“ Wh a tCompos i t i ona l i t ySt i l lCa nDo”( p.3 30) . Th i si sbe c a u s et h es t e r e ot y pea s s oc i a t e dwi t ht h eph r a s e‘ pe tf i s h’h a ss omef e a t u r e st h a ta r en otf e atures oft h es t e r e ot y pea s s oc i a t e dwi t ht h ewor d‘ pe t ’ort h es t e r e ot y pea s s oc i a t e dwi t ht h ewor d‘ f i s h’ .For i n s t a n c e ,l i v i n gi nah omea qua r i umi sn otaf e a t u r eoft h es t e r e ot y pea s s oc i a t e dwi t h‘ f i s h ’or‘ pe t ’ . 51 150 stereotype theory is false. Therefore, (PC) is not compatible with all kinds of theories of wor dme a ni ng . Thi si sa l s oPhi l i pRobb i ns ’ ss t r ongde f l a t i oni s mpos i t i on,a c c or di ngt o which (PC) together with other widely-accepted semantic principles such as (PU) do not set constrains on the nature of word meanings. Hor wi c h’ sr e pl yi se a s yt oa nt i c i pa t e .Her e j e c t s( PU)out r i g ht .( PU)a nd( PC)a r es i mpl y i nc ompa t i bl ewi t he a c hot he ri nHor wi c h’ sa c c ount .I n“ De f l a t i ngCompos i t i ona l i t y ” (Horwich 2001), this is pretty much his response. Horwich hardly explains why (PU) is wrong, except to say that we should resist the impulse to accept (PU) (ibid., p.380). But this is not an answer at all, especially when (PU) seems to be intuitively correct. If Horwich is right, then it is difficult to explain why t he r ei sac onc e ptof‘ hor s e ’butno c onc e ptof‘ whi t ehor s e ’ . Horwich provides another reply: the meaning of the phrase ‘ pe tf i s h’s houl dbet her e s ul t ofs ubs t i t ut i ngwor dsme a ni ngwha t‘ pe t ’a nd‘ f i s h’do—in this case, the stereotypes associated with them respectively—i nt oas c he mame a ni ngwha t‘ a n’doe s ;t hephr a s e itself does not possess anything similar to what its constituents possess (Horwich 2001, p. 380) .Hor wi c h’ sa r g ume nti st ha t ,i fonea c c e pt st ha tt heme a ni ngofawor di se ng e nde r e d by its associated stereotype (or inferential role, or recognitional capacity), one does not have to accept that the meaning of a complex is composed by the properties possessed by its constituents. He then continues to say that this would be an instance of the deflationary view of compositionality (ibid.). According to this, given either the stereotype theory or indeed any theory of word meaning and (PC), (PU) does not hold. I think Horwich is right on this one, though (PC) alone, it seems, trivially dismisses the 151 stereotype theory as unqualified. For the stereotype theorists simply do not accept the i de at ha tt hes t e r e ot y pea s s oc i a t e dwi t h,s a y ,‘ pe tf i s h’i st hef us i onoft hes t e r e ot y pe s 52 a s s oc i a t e dwi t h‘ pe t ’a nd‘ f i s h’ . And this is one of the important reasons for them to adopt (PU). I seriously doubt that this can be explained the other way round, for the incompatibility of (PC) and (PU) is straightforward. To form an argument in terms of the assumption that one who holds (PC) also holds (PU) does not seem to be feasible. In ot he rwor ds ,t hea r g ume ntf r om Fodora ndLe por e ,a ndt heonef r omRobbi ns ’ ss t r ong deflationism too, seem to deviate from the right line of argumentation at the very beginning. The difficulty, it seems, is about (FP). It might be less problematic for the stereotype theorists to accept both (FP) and (PC) to be applicable to sentences, but with respect to phrases this is entirely different. Accepting (FP) is equal to accepting that schemata have meanings. And the meanings of schemata force the stereotypes of words in a sequence of a phrase to fuse. So the issue here is not that (PC) together with (PU) imply the falsity of the stereotype theories, but that the stereotype theories are not compatible with (PC) in the cases of phrases.53 In terms of this fact, the stereotype theories probably should be rejected outright, for they cannot explain why one can understand the composite meaning oft hephr a s e‘ pe tf i s h’wi t houta nyi nvol ve me ntoft he stereotype of pet fish. It follows, however, that (PC) has little to do with being deflationary or not, because not every theory of word meanings is compatible with (PC). The second argument is quite different from the first one. Fodor and Lepore continue to da nc et oHor wi c h’ st une .Butt hi st i met hea r g ume ntis based on the less controversial 52 The problem of this line of argument is implicit in this very sentence. If one agrees with (PC), the me a ni ngof‘ pe tf i s h’wi l ln otbeas t e r e ot y pebu tac ons t r u c t i onpr ope r t yr e s u l t i ngf r oma ppl y i ngt h e s t e r e ot y pe sa s s oc i a t e dwi t h‘ pe t ’a n d‘ f i s h’i n t ot h es c h e ma‘ a n’wh os eme a n i n gi s“ A N” . 53 Strangely enough, the italic ‘ ph r a s e ’e x c e r pt e df r omHor wi c h’ swor dsa tt h ebe g i nn i ngoft h i spa r a g r a ph i sa c t u a l l yr e pl a c e dbyt h ewor d‘ s e nt e n c e ’i nHor wi c h’ sa r t i c l e . 152 pa r tof( PC) ’ sa ppl y i ngt os e nt e n c e s .Iwi l lf ol l owHor wi c h’ sf or mul a t i onoft hi s argument: (1) To understand certain complex expressions indicates the fact that some other complex expressions built from the same elements are unde r s t ood.Fori ns t a nc e ,“ J ohn s nor e s ” / “ Fl ounde r ss wi m”a n d“ J ohns wi ms ” / “ Fl ounde r ss nor e ” .( ‘ s y s t e ma t i c i t y ’ ) (2) Therefore, the meaning of a word does not depend on the complex expression in whi c hi ta ppe a r s .( ‘ c ont e xt -i n d e pe nde nc e ’ ) (3) But the stereotype assoc i a t e dwi t ht hewor d‘ s wi m’i n“ Fl ounde r ss wi m”i snott he s a mea st hes t e r e ot y pea s s oc i a t e dwi t ht ha twor di n“ J ohns wi ms . ” (4) The r e f or et heme a ni ngof‘ s wi m’i snote ng e nde r e dbya na s s oc i a t e ds t e r e ot y pe . (Horwich 2001, p. 381, footnote 15) Horwich summarises his objection to this argument in three points. First, premise (3) has t og o. Thes t e r e ot y pea s s oc i a t e dwi t ht hewor d‘ s wi m’i snotva r i e di ne ve r ypl a c ei t occurs. Second, even if this argument were persuasive it could tell against only the stereotype theory but not every theory of word meaning in general. Third, this argument does not depend on compositionality but on context-independence. If the consequence is to prevent any theory of word meaning from being compatible with (PC), that is because of context-independence rather than compositionality. Let me discuss these issues in a reverse order. Tot het hi r dpoi nt ,i ti sha r dt os e ehowHor wi c h’ sr e pl yc a nwor k. Thi si sbe c a us e context-i nde pe nde nc es e e mst obet hec ons e que nc eof( PC) .I fawor d’ sme a ning depends on the contexts in which it appears, and if (PC) also holds, then the meaning of a syntactical complex expression and the meanings of its constituents are mutually 153 de t e r mi ne d. Thi si sc i r c ul a ra ndc ont r a r yt oHor wi c h’ sownvi e w. The r e f or e , context-independence is the consequence of (PC). The second point is a bit complicated. On the one hand, Fodor and Lepore argue that (PC) sets constraints on all kinds of theories of word meaning. On the other hand, this argument is concentrated on the incompatibility between (PC) and the stereotype theory in particular. It might be all right for Horwich to object to this argument in terms of its failure of proving that (PC) sets constraints on all kinds of theories of word meaning. But it is devastating for him to accept this argument even in a counterfactual sense, for this suggests that he concedes that (PC) is not deflationary. To the first point, this is the typical reply we expect. But it is only natural for Horwich to make the claim that the stereotype associated with the word ‘ s wi m’i snotva r i e di ne ve r ypl a c ei toc c ur s ,i ti snotf ort hes t e r e ot y pet he or i s t st oma ke such a claim. As a matter of fact, Fodor and Lepore specifically explain what they mean by saying that the stereotype associated with the wor d‘ s wi m’i n“ Fl ounde r ss wi m”i snot t hes a mea st hes t e r e ot y pea s s oc i a t e dwi t ht ha twor di n“ J ohns wi ms ” . The ys a y : [T]he stereotype of people-swimming is much different from the stereotype of flounder-swimming since the latter, but not the former, adverts to the exercise of fins. (Fodor and Lepore 2001, p. 365) Ic a nn ots e ehowHor wi c h’ sr e pl yi sa c t ua l l yr e s pondi ngt ot hi sc l a r i f i c a t i on. Thi si spr oba bl ynoti mpor t a nta ny wa y ,i fwec ons i de rRobbi ns ’ sc r i t i c i s m ofHor wi c h’ s claim that (PC) is deflationary as correct. According to Robbins, (PC) is not deflationary even if it is seen as claiming something weaker, something similar to what Robbins calls weak deflationism—that is, (PC) itself, not together with other semantic principles, is compatible with all kinds of theories of word meaning. According to Robbins, a superstrong holistic theory of word meaning, for instance, is not compatible with (PC). 154 By definition, superstrong holism is the view that the meaning of a word in a language is at least partly constituted by the meanings of other words and phrases in the language, including phrases of which the word is a constituent (Robbins 2001, p. 333). It is clear t ha ts upe r s t r onghol i s mi snotc ompa t i bl ewi t hFodora ndLe por e ’ s ‘ c ont e xt -i nde pe nde nc e ’ .Ofcourse superstrong holism is not plausible, especially to Hor wi c h.Butt hi si sbe s i det hepoi nt .Ca nwewe a ke nHor wi c h’ sc l a i m bys a y i ngt ha t( PC) is only compatible with plausible theories of word meaning? I am afraid not. This means that (PC) indeed sets c ons t r a i nt sont he or i e sofwor dme a ni ng .SoHor wi c h’ s( PC)i snot deflationary. I twa sa l r e a dya r g ue di nt hedi s c us s i onofFodora ndLe por e ’ sf i r s ta r g ume ntt ha tt he stereotype theories of word meaning are perhaps too implausible themselves to be compatible with (PC). Superstrong holism, on the other hand, provides us an i nc ont e s t a bl ec a s et ha ti ti si t s e l fi nc ompa t i bl ewi t h( PC)a c c or di ngt oHor wi c h’ sown view. Horwich is probably so enthusiastic to establish a deflationary semantic enterprise that he fa l l ss hor tofa ddr e s s i ngSc hi f f e r ’ sc omme ntt ha tt hewor ds‘ de f l a t i ona r y ’a nd ‘ i nf l a t i ona r y ’a r emor ee mot i ver h e t or i ct ha nphi l os ophi c a l l ys i g ni f i c a ntnot i ons . In addition, the two sample sentences in the argument two paragraphs back are with different sc he ma t aa c c or di ngt oHor wi c h’ sa c c ount .I ti s‘ n vs ’f or“ J ohns wi ms ”a ndi ti s ‘ ns v’f or“ Fl ounde r ss wi m” .I ts e e mst ha tbot hs i de st a ket hi sa sami nordi f f e r e nc e .But is it? Thes ke t c hi ne s sofHor wi c h’ sa c c ountofs y nt he t i cs c he ma t ama ke si tve r ydi f f icult to 155 understand (PC). It is so sketchy that every construction property, resulting from applying the meanings of the constituents of a sentence to a syntactic schema in a specific order, is unique. It even cannot tell the difference between well-formed s c he ma t a ,s uc ha s‘ n vs ’ f or“ J ohns wi ms , ”a ndi l l -f or me done s ,s uc ha s‘ vs n’f or“ Swi msJ ohn. ”Tobe g i nwi t h , this account fails to explain why there are structurally synonymous relations between sentences. Horwich disposes of this sort of question in terms of the fine-grained characteristic of his account. Structural synonymy, it seems, only makes sense when considering a coarser-grained account of sentential meanings, according to which the synonymous relations are explained in terms of their having the same semantic structure (Horwich 1998a, pp. 163-164). The instance considered by him only concerns the active and passive, but there are many varieties. For example, how are we able to explain the s y nony mousr e l a t i onbe t we e n“ Hor wi c h’ st he or yi snote a s yt opr ove , ”“ Topr ove Hor wi c h’ st he or yi snote a s y , ”a n d“ I ti snote a s yt opr oveHor wi c h’ st he or y ” ?Ea c hof t he mi swi t hauni quec ons t r uc t i o npr ope r t y .Exc e ptf ort hi s ,Hor wi c h’ sa c c ounti sa l s o not able to explain structural ambiguity. According to Horwi c h’ sa c c ount ,e ve r ys e nt e nc e has a unique construction property. If this is not the case, it is because either one or some words it contains is vague or is ambiguous or has more than one explanatorily basic use property. Structural ambiguity is left out. Appa r e nt l yc e r t a i ns y nt a c t i cs c he ma t as uc ha s‘ a 54 ns and ns v’ ,whe r e‘ a’me a nsa na dj e c t i vea nd‘ and’me a ns‘ a nd’ ,a r ea mbi g uous . We do not know whether or not the second noun is bounded by the adjective. Further, this account is too sketchy to explain complex syntactic features, such as the plural and singular, tense, and many others. If the point is to provide an account of compositionality 54 Idon ote v e nk n owh owt of i l li nt h ec on n e c t i v e‘ a n d’e x c e ptbya ddi ng‘ and’t ot his schema. This is a g e n e r a ldi f f i c u l t yf orun de r s t a n di ngHor wi c h’ sa c c oun t .Col l i n sc or r e c t l yobs e r v e st h a tt h es y mbol‘ be’i na s c h e ma ,s a y ,‘ n be a’i sn oti t s e l fme a n i ng l e s s . 156 to explain ordinary use facts regarding complex expressions, then that every sentence has a unique construction property does not help. Fur t he r ,Col l i nsc r i t i c i s e sHor wi c h’ sa c c ounta snotbe i nga bl et or e c onc i l ewi t ht he general view of syntactic theories developed lately. It seems that certain syntactic features do not associate with syntactic schemata but with words they contain. One instance Col l i nsg i ve si st hedi f f e r e nts y nt a c t i cf e a t ur e sbe t we e nt hewor d‘ pr oba bl e ’ ,whi c hon l y t a ke sf i ni t ec l a us e s ,a ndt hewor d‘ l i ke l y ’ ,whi c ht a ke sbot hf i ni t ea ndi nf i ni t ec l a us e s . Moreover, these features are not determined by any a priori reason, gleamed from reflection upon what language must be like or what it is for (Collins 2003, p. 418). Syntactic schemata do not explain this sort of syntactic feature. If any account of meaning or linguistic structure is inconsistent with such restrictions, it should be rejected outright ont heg r oundst ha ti tmi s de s c r i be st hef a c t s( i bi d. ) . AsCol l i nsc ha r a c t e r i s e s ,Hor wi c h’ s syntactic schemata are flat. It lacks the necessary hierarchical structure imposed on a collection of words to accommoda t et he i rva r i ouss e ma nt i cf e a t ur e s .I fHor wi c h’ sa c c ount aims to explain how people can understand complex expressions of a language, such as English, in terms of their implicit understanding of nothing more than a class of words and a class of syntactic schemata of that language, then this account fails miserably. On the one hand, this account fails because of the reasons characterised above. On the other hand, if the syntactic features of words are absorbed into the range of implicit understanding of the use properties of words, which is very unlikely to be accepted by Horwich, then all schemata are redundant. If the acceptance property of a given word also gives circumstances in which the 157 syntactic features of the word are accepted, then the use property does not just govern the use of the word but also its syntactic connections to other words. If this is the case, then it is not that easy to distance the use-theoretic account from MEANING HOLISM. Further, apparently not every syntactic schema is applicable to every word. Horwich owes us an explanation of this. In addition, some words of languages such as Chinese, which do not have a syntactic item as clauses, are not translatable. Fi na l l y ,e ve ni fHor wi c h’ sa c c ounti sc a pa bl eofc opi ngwi t ha l lt hese difficulties, the attempt to construct a general account of proposition fails. In terms of my personal painful experience, translating more than ten books from English to Chinese, the syntactic structures of these two languages are very different. The following three points themselves are only examples. First of all, Quine is right on the point that there is no place-to-place corresponding between sentences of different languages. There is not even a guarantee that sentences have similar length. For instance, the Chinese translation of the 55 s e nt e nc e“ Pl a ne t sor bi ts t a r s ”c ont a i nsa tl e a s tt e nChi ne s ec ha r a c t e r s . Second, Chinese has a very different grammar from English. The following are several examples. (1) There are no plural endings in Chinese. (2) There is no such a concept of clauses in Chinese grammar. If someone writes a sentence containing clauses exactly in an English fashion, it will be corrected. (3) In English, there is a particular grammatical device to distinguish countable nouns from uncountable ones ,s uc ha sones a y s‘ ag l a s sofwa t e r ’i ns t e a dof‘ awa t e r ’ .But 55 Ev e nQu i n e ’ sc ons i de r a t i oni sv e r yEn gl i s h .Fori n s t a n c e ,youc a n n ott e l lwh e t h e rorn ot“ Ga v a g a i ”i sa sentence. This leads to the question of how to segment utterances of a remote language. The reasoning behind this is that pronunciations of words of English, or Latin phyla in general, contain syllable or syllables. Words with many syllables consist in many alphabets accordingly. But Chinese words do not have t h e s ef e a t u r e s . Ac c or di ngt ot hepr on u n c i a t i onof“ Ga v a g a i ” ,i ti sa bou tt h r e eCh i n e s ec h a r a c t e r sf ors u r ei f it is Chinese. 158 this does not hold in Chinese. Almost every noun has a particular kind of terms like ‘ g l a s s ’i n‘ ag l a s sofwa t e r ’ .I ns t e a dofs a y i ng‘ at r e e ’ ,f ori ns t a nc e ,Chi ne s es a y‘ a__ oft r e e ’ . (4) More importantly, verbs do not have tense variations.56 And this can go on, and on, and on. Third, similar to English, Chinese relies both on words and phrases to form sentences. But there is no corresponding relation between a word and the complexes containing it that can be preserved in an English-Chinese or a Chinese-English direction. The sentence “ Pl a ne t sor bi ts t a r s ”i sa l mos tape r f e c te xa mpl ef ort hec ha r a c t e r i s a t i ona bove .For i ns t a nc e ,Chi ne s edonots a y‘ apl a ne t ’or‘ ma nypl a ne t s ’buts a ys ome t hi ngl i ke‘ a__of pl a ne t ’a nd‘ ma ny__ofpl a ne t ’ . ( Re me mbe r ,nopl ur a le ndi ng s . )Mos ti mpor t a nt l y ,t he reason for using at least ten Chinese characters to translate it is simply because the c or r e s pondi ngwor dof‘ or bi t ’i nChi ne s ei sonl yanoun.Chi ne s edoe s not have a c or r e s pondi ngve r bt ot heEng l i s hve r b‘ or bi t ’ .Soweha vet ot r a ns l a t et hi st hr e ewor d English sentence into something whose English translation is similar to the sentence “ Pl a ne t si nt he i ror bi t smovea r ounds t a r s . ”Si nc et he r ei snoc or r e s pondi ngve r bof‘ or bi t ’ in Chinese, inevitably we must use an alternative Chinese verb. This is not enough to t r a ns l a t e“ Pl a ne t sor bi ts t a r s ”be c a us et hes ubs t i t ut e dChi ne s eve r bdoe snotma kepl a i n that planets move around stars in their orbits. The transla t i onoft henoun‘ or bi t ’ha st of i l l in to ensure the translation to be correct. In order to explain this deviant case, what sort of s y nt a c t i cs c he ma t ac a nHor wi c hi nvoke ?Thewor d‘ or bi t ’be i ngbot hanouna ndave r b in English or being only a noun in Chinese is entirely contingent. I do not see how 56 Therefore, Chinese does not have a corresponding grammatical device as subjunctive conditionals. This is probably the reason why many Chinese-speaking people have difficulty with understanding counterfactuals properly. 159 Hor wi c hc a ne xpl a i na wa yt hea bovepr obl e m oft r a ns l a t i ng“ Pl a ne t sor bi ts t a r s ”i nt o Chinese. Thef u n da me nt a ldi f f i c ul t yofHor wi c h’ sus e -theoretic account of word meaning and compositionality is its fine-grained character. According to this account, words of different languages are synonymous if and only if they have the same use property. The wor d‘ c e l i ba t a i r e ’ ,f ori ns t a nc e ,i st r a ns l a t e da s‘ ba c he l or ’r a t he rt ha n‘ unma r r i e dma n ’ because in terms of the fine-g r a i ne dc ha r a c t e rofHor wi c h’ sa c c ount‘ c e l i ba t a i r e ’a nd ‘ ba c h e l or ’a r ebot hpr i mi t i ve sbut‘ unma r r i e dma n’i snot( Hor wi c h1998a, p. 163). We a l r e a dys a wt hi sl i neofr e a s oni ngi nma nypl a c e sofHor wi c h’ sa c c ount .Howe ve r ,i n Chinese the translating wordsofbot h‘ ba c he l or ’a nd‘ unma r r i e dma n’a r ec ompl e x e xpr e s s i ons . Thec or r e s pondi ngChi ne s ephr a s eof‘ ba c he l or ’i sac ombi na t i onofa two-wor dphr a s ewhos eme a ni ngi ss i mi l a rt ot hewor d‘ s i ng l e ’i n‘ as i ng l el i f e ’a nda word whose meaning is similar to ‘ ma n’ .Se nt e nc e sofdi f f e r e ntl a ng ua ge sa r e synonymous, according to Horwich, if and only if they result from applying words with the same use properties in an exact order in a schema shared by those languages. But we already saw the flaw of this way of thi nki ngi nt hei ns t a nc eof“ Pl a ne t sor bi ts t a r s ”a ndi t s Chinese translation. They do not even fit in the same syntactic schema. It is inexplicable to believe that all contingent facts of English words and sentences are preserved in foreign languages. The particular semantic and synthetic features of foreign words are f i ne l ygr oundi nt ome a ni ng l e s spi e c e sbyHor wi c h’ smi l l .Qui ne ’ squa l ms ,c ha r a c t e r i s e d properly, explain why this failure happens. To look for a word-word corresponding manual for translation, one has to be clear first what a word is. Unless this is solved, the so-called fine-grained semantics can hardly begin. But in solving what a word is for a 160 r e mot el a ng ua ge ,i tt ur nsoutt ha tt r a ns l a t i oni si nd e t e r mi na t e .Hor wi c h’ sa c c ountdoe snot quell Qui ne ’ squa l msbuti sque l l e dbyt he m. Final Remarks I nt hes e c onds e c t i onIc ons i de r e dhowQui ne ’ ss c e pt i c i s mt owa r dme a ni ngi s mi s unde r s t oodbyHor wi c h. AsIha vea r g ue di ns e c t i onst hr e ea ndf our ,ne i t he rHor wi c h’ s use-theoretic account of word meaning nor his account of compositionality can work. This raises doubt regarding the sensibleness of constructing an account of proposition without any complication caused by the concept of truth. This further weakens the rationale of adopting the minimalistic rather than the disquotational approach toward the deflationary concept of truth. 161 Chapter V Be t we e nQui ne ’ sDi s quot at i onal i s m andHor wi c h’ sMi ni mal i s m Almost every adherent of different deflationary theories of truth, holds that there is some sort of equivalence relation stronger than material equivalence between the two sides of all biconditional instances of the disquotational schema.1 This feature of the relation is essential to the explanation of three major factors regarding the success of a deflationary theory of truth. The first of these is the functional role of the truth predicate to do the semantic ascent required to express infinite conjunctions and disjunctions. The second is the adequacy of the explanatory role in terms of which all facts involving truth are explained along with other suitable explanatory resources, and the third is the proper ascription of the truth predicate. Many philosophers locate their criticisms of disquotationalism in general, including minimalism, in one factor or another. In my di s c us s i onsoft he s ec r i t i c i s msIwi l la l s opr ovi demye va l ua t i onofQui ne ’ s di s quot a t i ona l i s ma ndHor wi c h’ smi ni ma l i s m. This chapter thus consists of four sections. The first three discuss the three factors for the success of a deflationary theory of truth me nt i o ne da bove . Thef our t hque s t i onsQui ne ’ sa t t i t udet owa r dsdi s quot a t i ona l i s m. 1. The Equivalence Relation and Semantic Ascent In order correctly to set the explanatory order regarding the equivalence relation and semantic ascent, I will begin my discussion with the sense equivalence relation. 1 As said in Chapter One, the deflationary theories of truth considered in this dissertation are disquotational and minimalist theories only. 162 1.1 The sense equivalence relation Many opponents of disquotationalism claim that the sense equivalence relation holds be t we e nt r ut ha s c r i pt i ons e nt e nc e s ,s uc ha s“ ‘ Snowi swhi t e ’i st r ue ” , and the sentences to whi c ht het r ut hpr e di c a t ei sa s c r i b e d,i nt hi sc a s e“ Snowi swhi t e ” . Al t hought hi ss e e mst o be problematic for disquotationalism, it is in fact equally problematic for minimalism as well. But let us look at disquotationalism first. Ri c ha r dSc ha nt zi n“ HowDe f l a t i o ni s t sDi ve r gef r omTa r s ki ”( Sc ha nt z2000)ma ke st he following claim: The general idea on which [the disquotational account] is based is that a sentence of the f or m‘ si st r u e ’h a st h es a meme a ni ngort h es a mecognitive content as the sentence s. (p. 318) I nt hec a s eof“ ‘ Snowi swhi t e ’i st r uei fa ndonl yi fs nowi swhi t e , ”hec i t e sQui ne ’ s wor dsof“ c a l l i ngt hes e nt e nc et r ue ,wec a l ls nowwhi t e ”i nPhilosophy of Logic, to support this reading of disquotationalism. However, Schantz continues to discuss the functional role of the truth predicate that cancels the effect of semantic ascent, and then to the expressibility of infinite conjunctions and disjunctions. No explanation is provided as to why the cited words are taken to support the claim that the sense equivalence relation hol dsbe t we e ni ns t a nc e ss uc ha s“ ‘ Snowi swhi t e ’i st r ue ”a nd“ Snowi swhi t e ” .I n“ The Me t a phy s i c sofTr ut h” ,De vi t ta l s oc l a i mst ha twhe nt het r ut ht e r mi sa t t a c he dt ot he quotation name of a statement, it yields a statement that is equivalent to that statement, s uc ha st hec l a s s i ce qui va l e nc er e l a t i onbe t we e n“ ‘ Snowi swhi t e ’i st r ue ”a nd“ Snowi s white (p. 583). If this sort of criticisms concentrates only on the truth ascription of singly given sentences without further elaboration, then they are based on the redundantist 163 readings of disquotationalism. In fact, some of opponents make this plain. For instance, Wol f g a ngKünnei n“ Di s quot a t i o na l i s tConc e pt i o nsofTr ut h”( Künne2000)di s t i ng ui s h e s the disquotationalist redundancy thesis from the disquotationalist abbreviation thesis of infinite conjunctions and disjunctions.2 And he criticises them separately.3 This line of criticising disquotationalism, and minimalism as well, is not very promising, for none of the major supporters of disquotationalism go for the sense equivalence explanation explicitly.4 Fori ns t a nc e ,f ol l owi ngi nQui ne ’ sf oot s t e ps ,Ha r t r yFi e l dr e j e c t st he interpersonal and intrapersonal synonymous relations between linguistic expressions in general. However, if there is no sensible connection between the redundantist reading and the abbreviation reading of the disquotational theory of truth, then it is hard to make sense of either. Of course, the abbreviation reading is partially the consequence of reading disquotationalism in the cases of singly given sentences in a redundantist way. A more 2 Künne raises an argument inspired by Quine, Argument From Entailment, against the redundantist r e a di n g .Ba s e donDa v i ds on’ sobs e r v a t i oni n“ Wh a ti sQu i n e ’ sVi e wofTr u t h ”t h a tt h ei ns t a n c e softhe di s qu ot a t i on a ls c h e maa r egu a r a n t e e dt obet r u eon l yi nt h eve r ys pe c i a lc a s e ,n a me l y ,f orag i v e npe r s on’ s idiolect, Künne reformulates the disquotational schema as ( D* )‘ p’i st r u eas used by me now if and only if p. “ ‘ Sn owi swhi t e ’a sus e dbymen owi st r u e ”h e n c ee n t a i l s ( a )“ Sn owi swh i t e ”a sus e dbymen owi sme a ni ngf u l , and (b) Something is sometimes used by someone. Howe v e r ,“ Sn owi swhi t e ”e n t a i l sn e i t h e roft h e m. In a footnote Künne indicates that Quine does not have very much sympathy for this argument. But I do not t h i n kt h a tQu i n eobj e c t st oe i t h e r( a )or( b)a sac on s e qu e n c eof( D* ) ,a tl e a s tn ot( a ) .I n“ Me a n i n ga n d I n f e r e n c e ”t h ea r gume n ti n t r odu c e dbyQu i n ei sa bou tt h ei nf e r e n c ef r om‘ Fa’or‘ ~Fa’t o(x) (Fx ~Fx). What Quine objectst oi st h ea s s umpt i ont h a tt h eme a ni ngof‘ Fa’e x i s t si f‘ Fa’i sme a ni ngf u l ,bu th eh a sn o ob j e c t i ont os ome t hi ngl i k e( a ) . Wh a ti sr e j e c t e di st h ei nf e r e n c ef r omt h eme a ni ngf ul n e s sof‘ Fa’t ot h e e x i s t e n c eoft h eme a n i ngof‘ Fa’ .Fu r t h e r ,( b)i sa mbi g u ou s because the very something that is used could be the sentence itself or the meaning of the sentence. Quine only disagrees with the assumption that the latter s ug g e s t st h ee x i s t e n c eoft h eme a ni ngof‘ Fa’t a k e na sa ne n t i t y .I ne i t h e rc a s ei ti sn otr i gh tto think that Qu i n edi s a g r e e swi t hKünn e ’ sa r g ume n t .For( a )i se n ought odi s c e r nt h edi f f e r e n c ei nme a n i ngbe t we e n “ ‘ Sn owi swhi t e ’a sus e dbymen owi st r u e ”a n d“ Sn owi swh i t e ” . 3 Künne leaves out Quine as one of those that go for the sense equivalence explanation. 4 In the following, however, it will be shown that some deflationists are committed to this explanation implicitly. 164 suited and hence weaker manner is to regard the disquotational truth predicate as providing the necessary mechanism to express infinite conjunctions and disjunctions, a r g ua bl yi nva r i ousde g r e e s .Eve ni nt hi swe a ke rs e ns e , Ani lGupt a( “ ACr i t i queof De f l a t i oni s m” ,Gupt a1993a) urges us to acknowledge that, in order for the deflationary truth predicate to express infinite conjunctions and disjunctions, the sense equivalence relation in the instances of the disquotational schema, and minimalist equivalence schema a swe l l ,i se s s e nt i a l .Gupt a ’ sc r i t i c i s mi sa ppl i c a bl et obot ht hedi s quot a t i ona ls c he maa nd Hor wi c h’ se qui va l e nc es c he ma .Fort hes a keofs i mplicity, I will only consider the disquotational schema in the following discussion. Gupta begins his criticism with an example according to which the expressive power of the truth predicate is manifested. Suppose we wish to affirm all sentences of the form ____ & snow is white. Our wish to affirm all the instances of this form amounts to affirming the following infinite conjunction by filling the blank of the form with all sentences: (1) [Sky is blue & snow is white] & [Chicago is blue & snow is white ]& …. The disquotational function of the truth predicate makes (1) equivalent to ( 2)[ ‘ Skyi sbl ue ’i st r ue&s nowi swhi t e ]&[ ‘ Chi c a g oi sbl ue ’i st r ue&s nowi swhi t e ] &…. As far as the sentence names are what a universal objectual quantification ranges over, us i ngt hepr onomi na lva r i a bl e‘ x’ ,( 2)c a nbef or mul a t e da s (3) For all sentences x: [x is true & snow is white]. Ba s e dont hi ss e t t i ng ,Gupt a ’ sa r g ume ntg oe sa sf ol l ows .For( 3)t oe xpr e s s( 1) ,( 1)a nd( 2) 165 need to be equivalent. In order for the truth predicate to play its role in forming generalisations such as (3), (3) and (1) have to be equivalent too. This is accepted by almost everybody, as the claim that the truth predicate has the function of expressing some infinite conjunctions.5 The difference between disquotationalism and non-disquotationalism lies in how strong the equivalence relation is and whether or not this is the only function played by the truth predicate. The disquotation function of the truth predicate makes certain that (1) and (2) come out equivalent. As for the claim made by the disquotationalist that the truth predicate is only a device of disquotation, the equivalence relation between (1) and (2) has to be strong enough to justify it. For (1) to be expressible in terms of (3), the relation between (1) and (2) can be varied from material, necessary, to sense equivalence. The material and necessary equivalence relations only yield that T-sentences are true or necessarily true respectively. But substantive truth theorists have no problem agreeing with the necessary truth of T-sentences. Indeed, according to Gupta, they have to.6 Only when the relation between (1) and (2) is sense equivalence, is the claim that the truth predicate is only a device of disquotation and nothing else justified. Under this strong reading of the equivalence relation, however, the claim that (3) expresses (1) is plainly false. (3), as a universal statement, and the conjunction of its instances, namely, (2) do not have the same sense.7 For Gupta, this is the consequence of confusing what Quine calls the difference between affirming the universal and affirming all its instances (Gupta, 1993a, p. 63). Gupt a ’ sa r g ume nti st wof ol d.Ont heoneha nd,i ts howst ha tbot ht hedi s quot a t i ona la nd 5 Infinite disjunctions are also expressible in terms of the same claim. Since the main issue here is about the expressibility of infinite conjunctions only, the part regarding infinite disjunctions is dropped from now on. 6 Argument for this, please cf. Gupta 1993a, p. 69. 7 Künne raises a similar argument, Argument From Conceptual Overloading. Please cf. Künne 2000, p. 187-188. 166 the minimalist theories of truth can at most explain all the instances of generalisations regarding truth, but not those generalisations themselves. On the other hand, the meaning of‘ t r ue ’i snotf i xe dbywha tHor wi c hc a l l st hei nc l i na t i ont oa c c e pt ,wi t houte vi de nc e ,a ll the instances of the equivalence schema, for the acceptance of these instances cannot sufficiently explain the particular use of the truth predicate in truth generalisations. The main point made by Gupta is the complaint that the disquotational and the equivalence schemata at most can explain every instance of a given truth generalisation but not the generalisation itself. Or, in other words, all the instances of a given generalisation can be derived from these schemata and other auxiliary resources, as long as these resources are truth-free, but not the generalisation itself. This complaint results from the fact that, without an infinitary inference rule, how can one explain the use of the t r ut hpr e di c a t ewi t hi nat r ut hg e ne r a l i s a t i on,s uc ha s“ Eve r ys e nt e nc eoft hef or m‘ p or not p’i st r ue ” ,i nt e r msoft hee xpl a na t i onoft heus e soft het r ut hpr e di c a t ewi t hi na l lofi t s instances? If accepting the sense equivalence reading is not an admissible alternative, then disquotationalism and minimalism are not capable of explaining all the uses of the truth predicate. This criticism, on the other hand, does indicate that, as far as the disquotationalist and the minimalist can find a way to explicate the reasoning of infinitary inference, the sense equivalence reading and the related problems are dispensable. However, discussions with regard to the reasoning of infinitary inference are highly related to the disquotational and the minimal accounts of the adequacy of the explanatory role of the truth predicate in terms of which all facts involving truth are explained. The r e f or e ,Gupt a ’ sc r i t i c i s mi si n s i g ni f i c a nt ,i fi nt hef ol l owi ngs e c t i ont her e a s oni ngof infinitary inference can be explained. 167 However, Gupta does not only raise one argument only. But the discussion of his further arguments is postponed. Gupta does not contest the claim that instances of the disquotational schema are necessarily true. According to him this is an inspirational insight. However, some opponents of disquotationalism and minimalism even disagree with this claim. For them, disquotationalism and minimalism cannot even make sense of the necessary equivalence relation of the instances of the disquotational schema or the equivalence schema. So I turn to this issue first. 1.2 The necessary equivalence relation We have discussed, in Chapter Two, the criticism made by Marian David that the i ns t a nc e soft hedi s quot a t i ona ls c he ma ,s uc ha s“ ‘ Snowi swhi t e ’i st r uei fa ndonl yi f s nowi swhi t e , ”a r enotne c e s s a r yt r ut hs .Hei snota l onei nt hi sl i neof criticism. The scope of this criticism is different from one opponent to another. For instance, James Van Cl e vei n“ Minimal Truth Is Realist Truth”( Cl e ve1996,p.869)a ndPa s c a lEng e li nTruth (Engel 2002, p. 47) only ascribe this problem to disquotationalism whose primary truth bearers are sentences. McGrath, on the other hand, includes minimalism as well (McGrath 1997b, pp. 79-84; McGrath 2000, pp. 35-39). The basic idea behind this line is that truth depends on meaning. Whether or not an instance of the disquotational schema holds, depends on the contingent fact of what the quoted sentence in the instances means. Künne presents an argument according to this reasoning (Künne 2000, p. 180). Let the sentences (S) Snow is white 168 and ( T)“ Snowi swhi t e ”i s true embed in a modal context, as consequents in a counterfactual conditional: ( MS)I fwea l lc a met ous et hes e nt e nc e“ Snowi swhi t e ”f ors a y i ngt ha tgr a s si sr e d,i t would not be the case that snow is white. ( MT)I fwea l lc a met ous et hes e n t e nc e“ Snowi swhi t e ”f ors a y i ngt ha tgr a s si sr e d,i t would not be the case that “s nowi swhi t e ”i st r ue . ( MT)i st r ue ,but( MS)i sf a l s e .Kü nnet he na r g ue st ha t ,i ft hea s c r i pt i onoft r ut ht o“ Snow i swhi t e ”i n( T)we r enot hi ngmor et ha nt hea s c r i pt i onofwhi t e ne s s to snow in (S), the a ppl i c a t i onoft hemoda lope r a t o r“ i twoul dnotbet hec a s et ha t ”t o( S)a nd( T)c oul dnot impose the assignment of different truth values to (MS) and (MT) (ibid.).8 Therefore, the instances of the disquotational schema are contingent. And the contingency of the disquotational schema is very likely to force people always to confuse (MS) with (MT) or (MT) with (MS). Mc Gr a t ht hi nkst ha ts t r ongde f l a t i ona r yt he or i e soft r ut h,s uc ha sHor wi c h’ smi ni ma l i s m, cannot get away from this attack. Although Horwich employs propositional forms to construct his minimalism, according to McGrath the instances of the equivalence schema are embedded in the same muddy field.9 This is correct. When formulating the equivalence schema, Horwich claims that for every English declarative sentence,‘ p’ , there is a corresponding equivalent sentence,“ t hepr opos i t i ont ha tp i st r ue ” .I ti se a s yt o 10 s e et ha t ,wi t hal i t t l ea dj us t me nt ,Künne ’ sa r g ume nta l s oa ppl i e st omi ni ma l i s m. In 8 Künne raises this argument against the redundantist reading of disquotationalism. But the abbreviation reading is very likely to be doomed if this is established. 9 Da v i di n“ Mi n i ma l i s ma n dt heFa c t sAbou tTr u t h ”a l s on ot e st h i s . 10 Precisely because of this reason, McGrath proposes his weak deflationism according to which only the 169 terms of the observation that minimalism eventually has to gravitate to the mud, regarding propositions as the primary truth bearers does not seem to provide a significant difference from disquotationalism in dealing with truth ascriptions in the cases of vague, ambiguous, and indexical contexts. If there is any way to dispose of these problematic cases, it will not only be available to minimalism. This line of criticism affects almost all the truth theories of the major deflationary t he or i s t s ,pe r ha pse xc e ptQui ne ’ s . The yr e s pondi noneway or another in order to maintain the important feature of disquotationalism, that all instances of the disquotational schema are necessary. Horwich on the other hand does not seem to care too much about this. Perhaps he takes this as a minor point under which the core of minimalism is intact. He does consider the instances of the equivalence schema as necessary though. In a footnote he says that the alternative strategy—the one that he prefers—is to keep the theory of truth un-modal and simple, and instead to derive the necessity of the axioms of the minimal theory from a separate theory of necessity. And the theory proposed by him is that necessary truths are specified by their being explanatorily fundamental. Establishing this theory, and providing that the minimal theory is explanatorily basic, it follows that the axioms of it are necessary.11 As David observes in “ Mi ni ma l i s ma ndt heFa c t sa bou tTr ut h”( Da vi d2000) ,t hi st he or yofne c e s s i t yi spl a i nl y f a l s e . Hes a y s ,“ [ N] a t ur a ls c i e nc ec ont a i nse xpl a na t or ily fundamental laws that are not ne c e s s a r yt r ut hs ”( p.167) .Orc o n s i de rGupt a ’ se xa mpl ei n“ ACr i t i queofDe f l a t i oni s m” , the statement that no chemical reaction will produce caustic soda from saltpetre and theory of propositional truth is deflationary and other theories of sentence, belief, and utterance truth are inflationary. 11 Please cf. Horwich 1998b, p. 21 footnote. 170 sulfuric acid is considered as necessary but is still in need of a deeper explanation by c he mi s t r yofwhyi thol ds .I ft hi ss t a t e me nti sne c e s s a r y ,t he na c c or di ngt oHor wi c h’ s theory of necessity it is explanatorily fundamental. But this contradicts the fact that chemistry provides a deeper explanation for it. Now Horwich has to reject the necessity of this statement, which might cause a cleavage between the intuitive set of necessary truths and the set of necessary truths according to his theory, and insist that chemistry is explanatorily fundamental. Eve nput t i nga s i deDa vi d’ sobs e r va t i ont ha tt hi si ns i s t e nc ei s ill-founded, there is another more serious problem. What is explanatorily fundamental normally depends on the explanatory purpose. It is hard to believe that there is an absolute order of explanatory fundamentality relative to practical purposes of explanation. It follows that Horwich fails to give us any reason to accept that the instances of the equivalence schema are necessary. For solving the problem of the modal status, Vann McGee provides a modified disquotational schema (DSI): ( DSI )“ Φ”i st r uei nmyi di ol e c ti fa ndonl yi fΦ. He then claims that all instances of (DSI) are necessary. As long as we keep the hidden indexicals straight, he so explains, we always arrive at (MT) rather than (MS).12 If this is the end of his explanation, this account is less appealing. For this account leaves e ve r y bodyi nt heda r kwhe na c c o u nt i ngf orot he rpe opl e ’ sus eoft hes a megr oupof sentences. McGee extends his idiolect disquotationalism to a global one by quelling Qui ne ’ squa l moft r a ns l a t i on.Forhi m, [ T] h ea c t u a lf i e l dl i ngu i s tn e ve rf i n dsh e r s e l fi nt h epos i t i onofQu i n e ’ sr a di c a lt r a n s l a t or , because the field linguist is willing to posit [psychological] hypotheses that Quine wishes to 12 Please cf. McGee 1993, pp. 90-91. 171 disallow to the eff e c tt h a tt h en a t i v e ’ sc on c e pt u a ls y s t e mi sbr oa dl ys i mi l a rt oh e rown . (McGee 1993, p. 99) McGee claims that the problem of the inscrutability of reference under this consideration is untenable. There are some differences one cannot mistake between the conceptual roles f ort hewor d‘ r a bbi t ’a ndf ort hephr a s e‘ unde t a c he dr a bbi tpa r t s ’ .Cons e que nt l y ,f ora n indigenous informant whose conceptual frame is at all similar to my own, the word ‘ ga va g a i ’us e dbyhe rmor ec l os e l yr e s e mbl e smyus eof‘ r a bbi t ’r a t he rt ha n‘ unde t a c he d r a bbi tpa r t s ’ .Iwi l lnotc ons i de rt hi se xpl a na t i onmor ef ul l y .Howe ve r ,l e tmepoi ntou t two things. First, there is no dispute that the field linguist is very likely to translate ‘ ga va g a i ’a s‘ r a bbi t ’r a t he rt ha n‘ u nde t a c he dr a bbi tpa r t s ’ ,a sIi ndi c a t e di nt hepr e vi ous chapter. There is even no dispute over bringing some psychological hypotheses into account in the practical vein of translation. What is in dispute is whether or not this translation is determinate. So it is better not to include psychological hypotheses in c ons i de r i ngr a di c a lt r a ns l a t i oni nt het he or e t i c a lve i n.Mc Ge e ’ sa c c ounts e e mst omi s st he point. Second, it is unclear how the relation of resemblance accounts for translation. More importantly, resemblance is not a transitive relation. Hence, the chance of a tenable account of the global disquotationalism is slim. The price of accounting for the necessary status of all instances of the disquotational schema is not low. Hartry Field gives a somewhat similar analysis in “ De f l a t i oni s tVi e wsofMe a ni nga nd Cont e nt ”( Fi e l d1994) . Tha ti s : [ A]pe r s onc a nme a n i ngf ul l ya ppl y“ t r u e ”i nt h epu r edi s qu ot a t i on a ls e n s eon l yt o utterances that he has some understanding of; and for such an utterance u, the claim that u is true (true-as-she-understands-it) is cognitively equivalent (for the person) to u itself (as she understands it).13 (p. 105) 13 There is a qualification according to Field that keeps the two from being fully cognitively equivalent. This is because the claim that u is true makes an ontological commitment to the utterance u, while the 172 This is what Field calls pure disquotational concept of truth. Further, Field explicates the notion of cognitive equivalence in a footnote: [ T] oc a l lt wos e n t e n c e st h a tape r s onu n de r s t a n ds‘ c ogn i t i v e l ye qu i v a l e nt ’f ort h a tpe r s oni s t os a yt h a tt h epe r s on’ si nf e r e nt i a lpr oc e du r e sl i c e n s eaf a i r l ydi r e c ti n f e r e n c ef r oma ny sentence containing an occurrence of one to the corresponding sentence with an occurrence of the other substituted for it.14 (p. 106) According to Field this is certainly not an instance of substitutional quantification. And he claims that this cognitive equivalence implies that this kind of inferences is empirically indefeasible and even conceptually so, if semantic paradoxes can be ignored. This c og ni t i vee qui va l e nc ebe t we e n‘ ui st r ue( a sIunde r s t a ndi t ) ’a ndt heut t e r a nc eu (as I unde r s t a ndi t )g ua r a nt e e st ha t“ u is true iff u”i sc onc e pt ua l l yne c e s s a r y ,( i fnotl og i c a l l y necessary). However, this account owes us a complete explication of the sort of inference that is appealed to regarding cognitively equivalent sentences, according to which the notions of truth and truth preserving cannot play any role. The next issue is how to apply the disquotational truth predicate to sentences that the speaker does not understand.15 This application is definitely not in a pure disquotational sense, because pure disquotational truth is by definition not about sentences or utterances that a given speaker does not understand. The account that allows the truth predicate to apply to utterances or sentences that one does not understand is therefore called by Field the extended disquotational theory of truth (1986, p. 59; 1994, p.128). The first way to explain extended disquotational truth is to take a theory of content into account. A foreign sentence is true in the extended disquotational sense as far as it is (intrapersonally or even utterance itself does not. But this is a relatively minor point and will be kept tacit in the following discussion. 14 There are certainly some qualifications of this explication. But for the sake of simplicity, they will be ignored. 15 This will be further clarified in the third section. 173 interpersonally) synonymous with a sentence that I understand. The general description of it is (ED): (ED) for any foreign sentence s, s is true iff s i ss y nony mouswi t has e nt e nc e‘ p’( t ha tI understand) and p. The notions of intrapersonal synonymy and interpersonal synonymy, according to Field, have to be explained in terms of something in which the notion of truth conditions does notpl a ya nyi mpor t a ntr ol e . Ay e r ’ sve r i f i c a t i oni s mi sa mongt hos et he or i e st ha tc a n provide the desired help. In fact, two sentences are interpersonally synonymous according to verificationism if and only if they have the same set of verification conditions. However, verificationism is undesirable because of its overly simplified character. This extended version of disquotational truth is still less preferable, according to Field, even if the theory of content is more refined. This is because it is probably better not to mix some semantic elements into the logical aspects of truth (Field 1994, p. 130). Field then proposes the notion of quasi-disquotational truth to replace extended disquotational truth in counterfactual circumstances. (For simplicity, I will use subscripts ‘ pd’ a n d‘ qd’ t odi s t i ng ui s ht hepur edi s quot a t i ona lt r ut hpr e di c a t ef r omt he quasi-disquotational truth predicate.) The definition of quasi-disquotational truth is as follows: (QD) ﭐ (S is trueqd i f f∑p( m (m is the meaning of S and @ (m i st heme a ni ngof‘ p’ ) )a nd p)). (Field 1994, p. 131) The idea is to invoke not just the notion of synonymy, but the notion of meaning as well, and to find a un-truth-theoretic way to explain it. It also appeals to an actually operator ‘ @’t ha tt e mpor a r i l yundoe st hee f f e c t soft hemoda lope r a t or( i bi d. ) .Oneoft hema i n 174 reasons why this quasi version is important, is that it is more likely to be our ordinary use of the truth predicate as maintained by Field. This is because we seem to incline to accept c e r t a i nc ount e r f a c t ua l ss uc ha s“ I fweha dus e d‘ Snowi swhi t e ’di f f e r e nt l yi twoul dha ve ha dt het r ut hc ondi t i onst ha tg r a s si sr e d. ”Andt oa voi dt hea c c us a t i ont ha tt heor di na r y way ofus i ng‘ t r ue ’i snott hepur e l ydi s quot a t i ona ls e ns e ,weha vet oa ppe a lt os ome t hi ng similar to quasi-disquotational truth. There is another option for explaining the truth ascription to sentences that one does not understand though, and that is to define the truth of foreign sentences as being true relative to a correlation of it to one of our sentences. A sentence is true relative to the correlation if our sentence that is in correlation with it is true in the purely disquotational sense (Field 1994, p.128) .Ofc our s e ,Qui ne ’ st r a ns l a t i ont he or yi sa mongt hos et ha tc a n explain the correlation between foreign sentences and our own. The feature of this option is that it is lacking an objective synonymy relation. The correlation, such as a translation between sentences, is highly context-sensitive so that it does not make sense to talk about correct or incorrect translations, but good or bad ones. Consequently, the application of the truth predicate to sentences that I do not understand does not have objectivity either. This option allows us to give a quasi-disquotational explanation of the counterfactual (MT). That is, I nc on s i de r i ngc oun t e r f a c t u a lc i r c ums t a n c e su n de rwh i c hweu s e d“ Sn owi swhi t e ”i n certain very different ways, it is reasonable to translate it in such a way that its disquotational truth conditions relative to the translation are that grass is red. (Field 1994, p. 133) Si nc eFi e l df ol l owsi nQui ne ’ sf o o t s t e psi nhol di ngas c e pt i c a lvi e wt owa r dt henot i ono f 175 synonymy, intrapersonal and int e r pe r s ona l ,i nt hePos t s c r i pt( 2001)of“ De f l a t i oni s tVi e ws onMe a ni nga ndCont e nt ” ,heg o e sf ort hes e c ondopt i ona ndma ke ss omef ur t he r refinements to the definition of quasi-disquotational truth: (Qd) S is trueqd (at a possible world v) iff S is to be translated by a sentence of mine (in the actual world) that is purely disquotationally true (at v). (2001, p. 152) Appa r e nt l y ,Fi e l d’ sa c c ountus e st hec onc e pt i onsofpur edi s quot a t i ona lt r ut ha nd extended disquotational truth to explain the necessary status of all instances of the disquotational schema, and further uses the conception of quasi-disquotational truth to make sure that we will always reach something similar to (MT) in counterfactual circumstances. According to McGrath, this is not satisfactor y .Hepoi nt soutt ha tFi e l d’ ss t r a t e gyi st o conceive of the relevant modal intuitions as special cases of intuitions about truth in other languages (McGrath 1997b, p. 82). The language used by our counterfactual selves in the relevant world is treated as another language, truth in which can thus be explained in the extended disquotational way as truth in actual other languages (pp. 82-83). The question McGrath raises is this: the issue is not just whether or not the counterfactual intuitions can be accounted for by deflationist, but whether she can explain the facts these intuitions r e g i s t e r( p.83) . Wene e dt oknowwhyi nt hec ount e r f a c t ua li nt ui t i on,( MT) ,“ Snowi s whi t e ”woul dha vet hedi s quot a t i ona lt r ut hc ondi t i onsr e l a t i vet oat r a ns l a t i ont ha tg r a s sis red. Why do these objective counterfactual intuitions relate to practical translations? The r ei ss t i l la not he rf e a t ur eofFi e l d’ sa c c ountt oc ons i de r . Thee xt e nde d- and quasi-di s quot a t i ona lt r ut hsdonotove r s t r i det hegi ve ns pe a ke r ’ sownpur edi s quot a tional truth, unless we are given some practical hypotheses such as that the 176 translation/interpretation manuals of people speaking the same language tend to be convergent. Most importantly, McGee and Field, especially Field, seem to be committed to the sense equivalence relation implicitly. Theda ma g ef r om a r g ume nt ss i mi l a rt oKünne ’ si snotonl yt ode f l a t i ona r yt he or i e sof truth. If the instances of the disquotational schema and the equivalence schema are contingent, then the disquotational schemata of reference and satisfaction are also doome d.I fwea l lc a met ous et h es e nt e nc e“ Snowi swhi t e ”f ors a y i ngt ha tg r a s si sr e d, t he n“ Snowi swhi t e ”woul dnotha veha dt hedi s quot a t i ona lt r ut hc ondi t i onst ha ts nowi s whi t e .I nCha pt e rTwoIme nt i one dHa l ba c h’ sreply to this sort of argument, which, a c c or di ngt ohi st e r mi nol ogy ,her e f e r st oa s‘ Le vy ’ sa r g ume nt ’ . Thema i npoi nti st ha t , under this sort of counterfactual reasoning, there are no necessary truths. For instance, “ Thep r i menumbe rbe t we e nf ou ra nds e ve ni ss ma l l e rt ha ns i x”woul dha venotbe e n ne c e s s a r i l yt r uei ft hephr a s e‘ pr i menumbe r ’we r eus e ddi f f e r e nt l y ,a s ,f ore xa mpl ei fi t we r eus e dt ome a nwha t‘ e ve nnumbe r ’me a ns . The r e f or e ,i ti snotj us tt ha twes houl d follow Field (1994) in emphasising that there are logical uses of the truth predicate that a r ei mmunef r om Le vy ’ sa r g ume nt ,butt ha tt he r emust be uses of the truth predicate to which this sort of arguments does not apply (Halbach 1999, p. 173).16 This leads us to the final discussion of the equivalence relation. 1.3 The material equivalence Theor i g i nofdi s quot a t i ona l i s mc ome sf r om Qui ne ’ si de a sofs e ma nt i ca s c e nta ndde s c e nt . 16 Halbach also gives us a quite different answer to the modal problem in Halbach 2000. Since his main concern is to provide a solution to the generalisation problem, his solution will be discussed in the following section. 177 The disquotational schema of truth therefore is supposed to grasp these ideas. The peculiar feature of semantic ascent and descent, as I discussed in the second section of Chapter Two, is that the mentioned sentence and the used sentence are exactly the same—the same sentence with the same empirical content or the same interpretation. It simply does not make sense to substi t ut et a l kofwhe t he r“ Snowi swhi t e ”i st r uef ort a l k ofwhe t he rs nowi swhi t e ,whe nt h es e nt e nc e“ Snowi swhi t e ”i n“ ‘ Snowi swhi t e ’i st r ue ” is used for saying that grass is red. Therefore, counterfactual reasoning does not directly apply to the instances of the disquotational schema. Though disquotationalism takes sentences as the primary truth bearers, this does not amount to its being the case that it is a degenerate case by which the fundamental feature of semantic ascent and descent cannot be correctly grasped. Furthermore, this does not suggest that counterfactual reasoning does not apply to sentences with the truth predicate, but that counterfactual r e a s on i nga bouts e nt e nc e swi t ht het r ut hpr e di c a t ei sde r i va t i ve .Fori ns t a nc e ,Fi e l d’ s quasi-disquotationalism is derived from pure and extended disquotationalism.17 More precisely, we should understand the disquotational schema in terms of the same sentence being mentioned and used relative to a language with a fixed interpretation scheme. This understanding of the disquotational schema is difficult to form in a s c he ma t i cf or mul a t i on.Mc Ge e ’ si di ol e c tdi s quot a t i ona l i s m,t ha t“ Φ”i st r uei nmyi di ol e c t i fa ndonl yi fΦ,a ndFi e l d’ spur edi s quot a t i ona l i s m,t ha t“ u is true (true-as-she-understands-i t ) ”i sc og ni t i ve l ye qui va l e nt( f ort hepe r s on)t o“ u itself (as she unde r s t a ndsi t ) ” ,c a nnotf ul l yi l l umi na t et hi spoi nt . Howe ve r ,ba sed on the proper understanding of the disquotational schema, the instances of this schema only need to be 17 Whether or not Fiel d’ squ a s i -disquotationalism is tenable is not at issue here. 178 held materially equivalent. So we may follow my characterisation in Chapter Two to formulate the disquotational schema as an axiom schema in a truth theory, within which all disquotation sentences of the base language sentences are axioms, as follows: (DQ)‘ p’i st r uei f fp, whe r et hef i r s t‘ p’i sr e pl a c e dbya nys e nt e nc eoft heba s el a ng ua g ea ndt he s e c ond‘ p’byt hes a mes e n t e nc ewi t ht hes a mei nt e r pretation. Us i ngGupt a ’ se xa mpl e ,( 3)e xpr e s s e s( 1)be c a us ea ppl y i ngt hedi s quot a t i ona ls c he mat o each instance of (3) in terms of the material equivalence reading, cashes out exactly each corresponding conjunct of (1).18 In Truth, Horwich says the following: [T]he explanatory basic fact about our use of the truth predicate is our tendency to infer i n s t a n c e sof‘ Th epr opos i t i onthat p i st r u e ’f r omc or r e s pon di n gi n s t a n c e sof‘ p’ ,a n dvi c e v e r s a ,wh e n e v e r( a )e a c h‘ p’i sr e pl a c e dwi t ht ok e n sofa nEng l i s hs e n tence, (b) these tokens are given the same interpretation as one another, (c) under that interpretation they e x pr e s st h ec on t e n tofas t a t e me nt( apr op os i t i on ) ,a n d( d)t h et e r ms‘ t h a t ’a n d ‘ pr opos i t i on ’a r eg i v e nt h e i rEn g l i s hme a n i ng s .( p.126, t h eitalics in (b) are mine) Therefore, my reasoning regarding how to interpret the disquotational schema of truth is noti nc ompa t i bl ewi t hHor wi c h’ se xpl a na t i onoft hee qui va l e nc es c he maoft r ut h. The necessary status, in addition, is not essential to the explication of the expressibility of infinite conjunctions in terms of the truth predicate. Granted, the material equivalence relation of the disquotational schema, (DQ), is a bit stronger than the usual one, for the sentences mentioned and used are restricted to the same sentence with the same interpretation. This is precisely why Quine takes truth as immanent to home languages, f orTa r s ki ’ sT-sentences are not automatically instances of (DQ).19 The expressibility of (1) 18 Ia mj u s tbor r owi n gGu pt a ’ sownc a s et os h owh i smi s c on c e pt i on . Th ec r i t i c i s mt h a ti ti si n c oh e r e n tt o use (1) to (3) in constructing the argument still holds. 19 This is why I made the claim in the second section of Chapter Two, that the stronger disquotational pr ope r t yun de rQu i n e ’ sa c c oun ti sn otpr e s e r v e dun de rt h er e c u r s i v ede f i n i t i onofs a t i s f a c t i on . 179 in terms of (3) in the above instance, depends on both the material equivalence relation and the additional restriction about the sentence mentioned and used in every instance of (3). However, this restriction does not yield any modal relation to the schema and its instances. If it should be held that the disquotational schema yields instances with a necessary status, then it has to be because of other reasons.20 In terms of all the deflationists I discussed so far, only McGrath and Horwich propose the material equivalence relation for explicating the expressibility of infinite conjunctions using the truth predicate. In the last chapter we saw that McGrath proposes his finite theory of propositional truth as that, for all propositions p, p is materially equivalent to Tr(P). In the postscript of the second edition of Truth, in answering the criticism from Gupta mainly discussed in 1.1, Horwich replies: The generalizing function of truth is perfectly well fulfilled as long as instances of the truth schemata, understood as merely material biconditionals, are accepted. Bu tt h i spoi n t[ Th a ti s ,Gu pt a ’ sc r i t i c i s moft h es e n s ee qu i v a l e n c er e l a t i onbe t we e n‘ Th e statement that p i st r u e ’a n d‘ p’ ]doe sn otun de r mi n et h emi ni ma l i s ts t or ya bou tt h e function of truth; for, as just mentioned, that function requires merely that the generalizations permit us to derive the statements to be generalized—which requires merely that the truth schemata provide material equivalences. (Horwich 1998b, p. 124) By applying the propositional form or, in other words, propositional s t r uc t ur eof‘ <p> is true iff p’t ot hepr opos i t i on,s a y ,“ <Snowi swhi t e >” ,t he yc a ne ns ur et ha tt he pr opos i t i onsonbot hs i de sof‘ i fa ndonl yi f ’a r ee xa c t l yt hes a mepr opos i t i ont ha ts nowi s white.21 The functions of semantic ascent and descent in turn make sure that the material 20 However, considering for what reason one should consider disquotation sentences as necessary, will not be discussed.. 21 Hor wi c hc l e a r l yma k e st h i spoi n ti n“ ADe f e ns eofMi n i ma l i s m” ,p.153. 180 equivalence relation is enough to explicate the expressive function of the truth predicate.22 In considering the case of disquotationalism, requiring that the sentences mentioned and used in every instance of the disquotational schema are the same sentences relative to a language with a fixed interpretation scheme is vital to a proper understanding of the disquotational schema, (DQ). This, on the other hand, also shows that although the disquotational schema and the equivalence schema are not sense equivalent, they are still capable of explaining the expressing of infinite conjunctions.23 Whether or not disquotationalism and minimalism can explain general facts involving truth—that is, the affirmation of infinite conjunctions—will be discussed in the following section. But it is clear that so far I cannot find anything giving minimalism the advantage over disquotationalism. Thes e c onda s pe c tofGupt a ’ sc r i t i c i s mr e s t sonHor wi c h’ sa c c ountoft heme a ni ngof‘ i s t r ue ’ .I nMeaning, Hor wi c h’ se xpl i c a t i onoft heme a ni ngoft hewor d‘ t r ue ’i sa sf ol l ows : ( TRUE)Thea c c e pt a nc epr ope r t yg ove r ni ngourt ot a lus eoft hewor d“ t r ue ”i st he i nc l i na t i ont oa c c e pti ns t a nc e soft hes c he ma‘ t hepr opos i t i onthat p is true if and only if p’ .( p.45) I naf o ot not e( 23)of“ ADe f e ns eo fMi ni ma l i s m”( Hor wi c h2001a), Horwich makes it clear that he distinguishes inclinations from dispositions of accepting instances of the equivalence schema because of consideration regarding the liar paradox. This, in turn, is one of the important issues of the coming section. Now, Gupta claims that, for 22 Providing that the reasoning of infinitary inference is well explained by disquotationalism and minimalism. 23 This is probably one of the reasons why Horwich would like to keep his theory of truth un-modal and simple. 181 deflationism to dethrone the truth-conditional account of meaning, the instances of the 24 e qui va l e nc es c he maha vet oe xpl a i nt heme a ni ngors e ns eof‘ t r ue ’ . By this, he means t ha tourunde r s t a ndi ngoft heme a ni ngof‘ t r ue ’de pe ndse nt i r e l yonourpr i or understanding of the meaning of all (declarative) sentences of a given language. Again, this cannot be correct, according to Gupta, because the conceptual resources of understandi ng‘ t r ue ’a r ee nor mou s l yma s s i ve .I ti snotdi f f i c ul tt of i ndt heba s i c assumption of this criticism—that is, the sense equivalence relation. However, in terms of my clarification of the understanding of the disquotational schema, it is without doubt that t hi sa s s umpt i oni st o t a l l ywr ong , a ndhe nc es oi sGupt a ’ sa r g ume nt . Qui nec e r t a i nl ywoul dnota g r e ewi t hHor wi c h’ s( TRUE)a se xpl i c a t i ngt hemeaning of ‘ t r ue ’ .Ne ve r t he l e s s ,t heg e ne r a li de abe hi ndi ti spr oba bl ya c c e pt a bl ei nt ha ti nQui ne ’ s terms, all instances of the disquotational schema are stimulus analytic.25 Following my clarification of the understanding of the disquotational schema three paragraphs back, it is quite clear that a proper understanding of the disquotational schema can be nicely separated from any account of meaning, provided that these accounts of meaning are non-truth-conditional.26 Wha ti se s s e nt i a lt oourunde r s t a ndi ngof‘ t r ue ’i st hepe c ul i a r feature of the disquotational schema, that the sentences mentioned and used in every instance are exactly the same sentence with the same interpretation. What is not essential is what the sentence mentioned and used means. The truth predicate is really transparent 24 Gu pt a ’ sa r g ume n ti sba s e donDumme t t ’ sa r g ume n ti nh i sc e l e br a t e dpa pe r“ Tr u t h ” ,a c c or di n gt owh i c h the sense equivalence relation is also regarded as essential to deflationism. 25 Stimulus analytic sentences are by definition those sentences that speakers would assent to under any circumstance. 26 There are, of course, some necessary conditions for the applicability of the truth predicate; for instance, with the syntactic feature of declarative sentences and with assertoric content. But this is not unique to disquotationalism, but is true of deflationary truth theories in general, especially the minimal theory of propositional truth. Please see section 3. 182 in this sense. It also nicely accommodates the fact that what is true depends doubly on meaning and reality. Further, it is unnecessary to restrict the language, in the present case, English, to be relative to a particular space-time frame—that is, the language can be seen a se vol vi ngove rt i me . The r ei snone e dt oa doptMc Ge e ’ si diolect disquotationalism or Fi e l d’ spur edi s quot a t i ona l i s m,be c a us et heus e soft het r ut hpr e di c a t eofpe opl es pe a ki ng the same language incline to coincide with one another. For, unlike McGee and Field, the l a ng ua g ewea r et a l ki nga bouti snota ny one ’ si diolect. The impression that the extension of the disquotational truth predicate varies each time new terms are introduced into the l a ng ua g e ,i sc a us e dbyt hede f i ni t i ona lf or mul a t i on,s uc ha sDa vi d’ s( D) ,( DM) ,a nd( DM) discussed in Chapter Two, according to which the range of truth ascription depends on the substitution class. It is now clear that this is nothing but a misconception. If any new term introduced into a language causes problems to the disquotational theory, then it equally causes problems to substantive theories too. English speakers would accept instances of t hedi s quot a t i ona ls c he ma ,s uc ha s“ ‘ Snowi swhi t e ’i st r uei fa ndonl yi fs nowi swhi t e , ” noma t t e rwha t . Anot he rr e a s onnott oa doptMc Ge e ’ sorFi e l d’ sa c c ount s ,i st he prominent role played in them by the notion of content. Be f or ewemoveont ot hef ol l owi ngs e c t i on,onet hi ngs houl dbenot e d.I n“ Ana l y t i c i t y , Ca r na p ,Qui ne ,a ndTr ut h”( Da vi d1996) ,Da vi dobs e r ve st ha tQui ne ’ sa r g ume nti n“ Two Dog ma sofEmpi r i c i s m”a ga i ns tCa r na p’ sde f inition of analyticity—that is, the definition in terms of the semantical rules or, in other words, meaning postulates of a given language—has a similar impact on other semantic concepts, including the concept of truth. 183 Ca r na p’ si de ai st ha t ,onc ey ouc hoose a convenient language form, there is a set of (analytic) sentences, each of which is true according to the semantical rules of this language. In other words, when the language is fixed, a certain set of analytic sentences is also fixed. The drawback of this account, Quine argues, is that it only labels some s e nt e nc e sa s‘ a na l y t i c ’or‘ s e ma nt i c a lr ul e ’a ndl e a ve st he m une xpl a i ne d.Fur t he r ,t hi s account transfers the unclearness of the concept of analyticity to the concept of semantical rules (cf. Quine 1953b section 4). Ac l os e ri nve s t i ga t i ona ndc r i t i c i s m oc c ur si n“ Ca r na pa ndLog i c a lTr ut h”( Qui ne1954) , written three years later.27 One of the main points is about the distinction between legislative and discursive postulation. Legislative postulation institutes truth by convention and discursive postulation is merely the selection of certain true expressions (from a pre-existing body of truths) as a basis from which to derive others (p. 111). There is also a similar distinction between legislative and discursive definitions. Following these di s t i n c t i ons ,mos ts e ma nt i c a lr ul e sa ndpos t ul a t e si nCa r na p’ srule book should be classified as discursive ones, if Carnap is attempting to explicate the intuitive concept of analyticity. These semantical rules and postulates simply state pre-existing truths, whose selection is arbitrary, or relative to our demonstrating purposes or interpretation. Consequently, there are no statements that are true by these rules or postulates. Otherwise, all true sentences should be considered as semantical rules or postulates since they are true by t he ms e l ve s .I fCa r na p’ sa t t e mpti st oc ons t r uc tane w( a r t i f i c i a l )l i ng ui s t i cs y s t e m to explicate the intuitive concept of analyticity, then those rules and postulates are legislative even though this legislative procedure is still arbitrary. If the reason for 27 “ TwoDogma sofEmpi r i c i s m”wa swr itten in 1951. 184 constructing a formalised language L is to capture the intuitive concept of analyticity, then this legislative procedure can never be successful. These rules, or postulations, are constructed in language L and hence do not belong to ordinary languages. The analytic statements in language L, then, are not coincident with the analytic statements in ordinary languages. It follows that the attempt at capturing the intuitive concept of analyticity fails28. In a nutshell, if it is impossible to define the concept of analyticity as analytic-in-L for a variable L ranging over all languages in terms of a restricted, indexed concept of analyticity, then to define true-in-L for a variable L ranging over all languages in a similar fashion is also not possible. Surprisingly, David so complains, Quine later conceded that the debate between him and Carnap merely concerns the explanandum (David 1996, p. 284).29 Da vi d’ ss ur pr i s e ,I think, results from hi smi s c onc e p t i onofQui ne ’ si de aa boutt r ut h.I ti sr a t he rdubi oust ha t analyticity can sensibly be defined as S is analytic for L (for both variable S and L). For i ns t a nc e ,t hec l a i mt ha t“ Ba c he l or sa r eunma r r i e dme n”i sa na l y t i cf ora nyl a ng ua g eL is problematic. This claim implies that it makes sense to talk about the analyticity of this sentence in Mandrin, say. But for Mandrin speaking people who do not understand Eng l i s h,t hi ss e nt e nc ei snot hi ngmor et ha ng i bbe r i s h.Ont heot he rha nd,Ca r na p’ s account consists of two parts. First is about the definition of analyticity relative to a given language. Then utilising the notion of translation to see whether or not the above-mentioned instance can make sense in another language. Although Quine has a strong view against the absoluteness of meaning and translation, this is irrelevant to the 28 M.G. Wh i t ei n“ Th eAn a l y t i ca n dt h eSy n t h e t i c : AnUnt e na bl eDu a l i s m”( 1950)ma deas i mi l a rc r i t i c i s m a sf ol l ows :“ Wh a ts omeph i l os oph e r sdoi st opr e t e n dt h a tn a t u r a ll a ngu a g e sa r er e a l l yqu i t el i k et h e s e artificial languages; and that even though there is no rule-book for them, people do behave as if there were such a rule-book ”( i bi d. :82) . 29 Quine admits this in Dear Carnap, Dear Van, Quine to Carnap, 1951-3-29. 185 explanatory procedure ofCa r na p’ sa c c ount .Si mi l a r l y ,‘ t r uei nL’( va r i a bl eL) should be explicated in a similar fashion. I n“ No t e sont heThe or yofRe f e r e nc e ” ,Qui nes a y s: I nTa r s k i ’ sc on s t r u c t i ons ,mor e ov e r ,weh a v ea ne x pl i c i tg e ne r a lr ou t i n ef orde f i n i n g truth-in-L for individual languages L which conform to a certain standard pattern and are well specified in point of vocabulary. We gave indeed no similar single definition of ‘ t r u e -in-L’f orv a r i a bl e‘ L’ ;butwh a twedoh a v es uf f i c e st oe n dow‘ t r u e -in-L’ ,e v e nf or v a r i a bl e‘ L’ ,wi t hah i ghe n oug hde g r e eofi n t e l l i gi bi l i t ys ot h a twea r en otl i k e l yt obe averse to using this idiom. (p. 138) I nhi sp a pe r ,Da vi dt wi s t s‘ t rue-in-L for individual languages L’a nd“ ‘ t r ue -in-L’f or variable L”i nt o‘ i nde xe dt r ut hpr e di c a t e sr e s t r i c t e dt oi ndi vi dua lf r a g me nt sofEng l i s h ’ a nd‘ t r u e -in-Eng l i s h’ . Thi si st hec ons e que nc eofhi sf a i l i ngt oa ppr e c i a t eQui ne ’ s reasoning as characterise da bove .Cons i de r i ngDa vi d’ smi s i nt e r pr e t a t i onof disquotationalism, this is not surprising. The general routine is not available in defining a na l y t i c i t y ,pur e l ybe c a us e‘ i sa na l y t i c ’i snott r a ns pa r e nt . Thewor ds‘ ge ne r a lr out i ne ’c a n of course be regarded as about the disquotational schema. The corresponding truth predicates in different languages are similarly accounted for in terms of the corresponding disquotational schemata of them. Therefore, whether or not we can have a general de f i ni t i onof‘ t r ue ’depends on how strong our account of meaning is. This will be discussed in the third section. 2. Explaining General Facts Involving Truth For disquotationalism and minimalism to be feasible theories of truth, the most important task is to explain the over a l lus eof‘ t r ue ’i nt e r msoft hec or r e s pondi ngs c he ma t aoft r u t h. Fori ns t a nc e ,onemi g hti nc l i net oa c c e ptt hes e nt e nc e“ ‘ Wa t e ri st r a ns pa r e nt ’i st r uei fa nd 186 only if water is H2O” . Thedi s quot a t i ona ls c he maa ndt hee qui va l e nc es c he ma ,along the line of its material equivalence feature, can well explain this. I only give an explanation here in terms of the disquotational schema. This sentence is a derivation from the scientific explanation of the transparency of water, according to which water is transparent if and only if it is H2Oa ndt hedi s quot a t i ons e nt e nc et ha t“ Wa t e ri s 30 t r a ns pa r e nt ”i st r uei fa ndonl yi fwa t e ri st r a ns pa r e nt . Among these explanations, the most important part is about general facts involving truth. Or, in other words, does affirming all instances have a relation to affirming the corresponding generalisations close enough to establish the relation that, by affirming a generalisation or an infinite conjunction, one can affirm the other? In the following, we will first see some answers provided by two disquotational theorists, Field and Halbach. Then I will discuss one of Hor wi c h’ si ns t a nc e sofmodus ponens—that is, if one sentence/proposition implies another, and the first one is true, then so is the second. By doing this, I seek to discover whether or not disquotationalism and minimalism could satisfactorily account for general facts involving truth. Moreover, I would like to show that in what reasoning one can depend the explanation of all general facts involving truth on the material equivalence relation occurring within each instance of the disquotational and the equivalence schemata. And finally, I aim to show a deficiency that is unique to minimalism, that is, the problem of the liar paradox. But let me briefly discuss some alternatives advanced by some defenders of disquotationalism. Field (1994) proposes to begin from the idea of building schematic 30 Amor ec ompl i c a t e dc a s ei sge n e r a t e dbys u bs t i t u t i n gt h ewor d‘ be c a u s e ’f or‘ i fa n don l yi f ’i nt h e sentence explained and the scientific explanation. Horwich seems to think that this is problem free (Horwich 1998b, pp. 49-50 and pp. 105-106). Although I doubt that this will work, I ignore this issue. 187 variables into the language and subjecting them to a rule of substitution.31 Unlike Horwich, Field accepts the truth schema itself, and not just its instances, as part of a language. It follows that this device naturally gives the disquotational theory of truth the desired deductive power to justify generalisations. For instance, to prove that, whenever modus ponens is applied to true premises, the conclusion is true, we can do as follows: 1. If p, then q premise 2.‘ p’i st r ue pr e mi s e 3.‘ p’i st r uei f fp a s s umpt i on,di s quot a t ionalism 4. p from 2, 3 5. q from 1, 4 6. ‘ q’i st r uei f fq a s s umpt i on,di s quot a t i ona l i s m 7.‘ q’i st r ue f r om5,6 So‘ I fp,t he nq’a nd‘ p’i st r ue ,t he n‘ q’i st r ue All general facts involving truth are supposed to be derived in a similar fashion. According to Field, there are at least three advantages of this account. First, the theory of truth does not need to be revised as the language is expanded. Second, the use of the s c he ma t i cf or mul a“ ‘ p’i st r uei fa ndonl yi fp”i nt hea xi oma t i s a t i ons e e mst omor e directly capture the core of the notion of truth. Finally, and most significantly, according to Field, the theory does not depend on our having a compositional account of the functioning of the other devices in the language. That is, the accounts of reference and satisfaction are expendable in explicating compositionality (cf. Postscript of Field 1994, p. 142). 31 Field supplements his view, in the Postscript of Field 1994, with the claim that the syntactic theory also needs to be done with schematic variables. For the sake of simplicity, the discussion of this issue is left out. 188 Presumably, the second advantage is just about my proper understanding of the disquotational schema in the discussion of the material equivalence relation. The third is important. It is about how truth commutes,bor r owi ngHa l ba c h’ st e r ms ,wi t hl og i c a l connectives. In terms of a similar procedure, for instance, it is not hard to reach the conclusion: ‘ porq’i st r uei fa ndonl yi f‘ p’i st r ueor‘ q’i st r ue . Sha pi r or a i s e sonepr obl e m ofFi e l d’ spr opos a l .Heque s t i onst hes a t i s f a c t or i ne s soft h i s proposal in accounting for sentences that one does not understand, even in the extended disquotational sense (Shapiro 2003, p. 129). Extended disquotationalism is not helpful because the substitution range of schematic variables still boils down to the same class of sentences that one understands. This will be discussed in the next section. Horwich, on the other hand, questions the reasonableness of attributing schematic reasoning to the great majority of people. This seems to be too sophisticated an activity to be attributed to ordinary people (Horwich 2001a, note 20). Halbach also gives us a quite different answer in Halbach 2000. According to his modalised disquotationalism (MD), the necessary status of the disquotation sentences, that is, the instances of the disquotational schema, can be accounted for by the notion of truth-analyticity.32 He uses the following as an axiom of truth: “ ‘ A’i st r ue ”a nd“ A”a r et r u t h-a na l y t i c a l l ye qui va l e ntf ore ve r ys e nt e nc e“ A” of the base language. (p. 163) ‘ Tr ut h-analytically equivalent ’i nt ur nc a nbede f i ne di nt e r msoft r ut h-analyticity and 32 For the preliminary formulation, please cf. the second sect i onofCh a pt e rTwo,un de rt h et i t l e‘ A preliminary set-u pofapr ope rf or mu l a t i on’ ,a n dt h et h i r ds e c t i onoft h es a mec h a pt e r ,u n de rt h et i t l e ‘ Ex pr e s s i bi l i t y ’ . 189 material equivalence: “ A”a nd“ B”a r et r ut h-analytically equivalent if and only if (the proposition e xpr e s s e dby )t hes e nt e nc e“ A if and only if B”i st r ut h-analytic (ibid.).33 In addition to all the disquotation sentences in the truth theory, two further principles are added in order to define truth-analyticity. That is, (L) Truth-analyticity is closed under predicate logic; and a reflection principle ( R)Fo re ve r yf or mul a“ A(x) ”wi t honef r e eva riable x, the following is derivable: Fora l l“ B” :i f“ A(x) ”i st r ut h-a na l y t i cf or“ B” ,t he nA(“ B” )( p.164) . Then it can be claimed in the truth theory, that all disquotation sentences (of the base language) are truth-analytic. The virtue of this account, according to Halbach, is its capacity to allow truth to commute with the connectives (p. 165). Unlike Field, the reasoning is not schematic, but is about universally quantified sentences. For instance, it can be proved that truth commutes with the connective of negation: 1. All disquotation sentences are truth-analytic. Assumption, MD 2.Fora l l“ A” :“ ‘ A’i snott r uei f fnotA”i st r ut h-analytic. from 1, (L) 3.Fora l l“ A” :“ ‘ notA’i st r uei f fnotA”i st r ut h-analytic. from 1 4.Fora l l“ A” :“ ‘ notA’i st r uei f f‘ A’i snott r ue ”i st r ut h-analytic. from 2, 3 5.Fora l l“ A” :“ notA”i st r uei f f“ A”i snott r ue . f r om4,( R) All the general facts involving truth are supposed to be proved under similar reasoning. However, there is one thing not very clear in this account. Halbach does not appeal to a notion of what Boghossian calls impure analyticity to define truth-analyticity. If the 33 Halbach is well aware of the de re character of his formulation and offers a solution to it. Please cf. Halbach 2000, pp. 163-164. 190 not i onof‘ t r ut h-a na l y t i c i t y ’i sde f i ne di nvi r t ueoft heme a ni ngoft het ruth predicate, then truth-analyticity cannot be used as an auxiliary notion to help axiomatise truth. Similar to Horwich, (except for the differences between the primary truth bearers), Halbach holds that the meaning of the truth predicate is fixed and exhausted by nothing beyond the disquotation feature (Halbach 2000, p. 168). Thus, a sentence is truth-analytic if and only if it is either a disquotation sentence or if it follows logically (not by virtue of the meanings of non-logical vocabulary) from the disquotation sentences (ibid.). Halbach also claims the following: According to the view of modalized disquotationalism, being logically provable from the disquotation sentences and being [truth-analytic] are the same. (p. 168) So truth-analyticity is conceived in terms of provability. The problem for this account is Ha l ba c h’ sc l a i mt ha tt hepr e di c a t e‘ i sl og i c a l l ypr ova bl ef r om t hedi s quot a t i ons e nt e nc e s ’ , in the sense of classical predicate logic, can be defined within the base theory (ibid.). And according to Halbach, a base theory is formulated by the base language consisting of only s e nt e nc e swi t houtt het r ut hpr e di c a t e( p.155) . The n,howc a n‘ l og i c a l l ypr ova bl ef r omt he di s quot a t i ons e nt e nc e s ’bede f i ne di nat he or ywi t hi nwhi c hdi s quot a t i ons e nt e nc e s cannot e ve nb ef or me d?It hi nkwha tHa l ba c hha si nmi ndi st ha t‘ l og i c a l l ypr ova bl e ’butnot ‘ l og i c a l l ypr ova bl efrom disquotation sentences’c a nbede f i ne dwi t hi nt heba s et he or y . Andt hepr e di c a t eof‘ l ogi c a l l yp r ova bl ef r om di s quot a t i ons e nt e nc e s ’c a nhence be defined in the truth theory. As I indicated in the last subsection, the original idea of disquotationalism comes from semantic ascent and descent. The truth predicate has its use in helping us express, as well as affirm, infinite lots of sentences, and infinite conjunctions.34 The issue regarding the 34 Again, the discussion of infinite disjunctions is left out hereafter. 191 deductive power of the disquotational theory is raised in considering the relation between affirming an infinite conjunction and affirming a corresponding (universal) generalisation. It is easy to reach the conclusion that affirming a universal generalisation implies affirming all of its instances, and through the disquotational schema, all conjuncts of the original infinite conjunction are gained. The other way round is problematic, due to lack of an infinitary rule of inference. When some infinite conjunctions have their conjuncts with syntactic or proof-theoretic features, it is easiest to see how adding some axioms and/or rules of inference can help explain the affirmation of a generalisation from the affirmation of all its instances. The truth theory accompanied with additional axioms and/or rules of inference provides derivations of the affirmation of a given generalisation from the affirmation of all its instances. Both Field and Halbach hold that the disquotational theory of truth is not simply an axiomatisation of truth, in terms of the disquotational schema as the axiom schema, but a theory about an axiomatic system of truth. Since all the axioms resulting from the instances of the disquotational schema are deductively too weak to account for certain general facts involving truth, more axioms and/or rules of inferences are needed to enhance the deductive power of the truth theory. If the additional axioms and/or rules of inference are truth-free ,t he ya r ea dde dt oot he re xpl a na t or yr e s our c e s( Fi e l d’ s ) .I fnot , t he ya r ea dde dt ot het r ut ht he or y( Ha l ba c h’ s ) . Al t houg hbot hFi e l da ndHa l ba c h emphasise that these additional axioms and/or rules of inference are not substantial, it is doubtful that they do not inflate the truth theory in one way or another. Fi e l d’ sa c c ounti nt r oduc e ss c he ma t i cl e t t e r sa ndt hec or r e s pondi ngs ubs t i t ut i onr ul ei n t o 192 our natural language, and first order propositional logic is thus absorbed into the language we speak as a part of it.35 In addition to attributing schematic reasoning to ordinary people, one striking consequence is that infinite conjunctions with certain synthetic features are all expressible. The function of the disquotational truth predicate of expressing some infinite conjunctions vanishes. The use of the truth predicate is to properly form disquotational truth conditions, and a side interest is to distinguish true generalisations from false ones. Another striking consequence is that, it seems, some model-theoretic conceptions are built-in the language spoken by ordinary people. One can de r i vet ha t ,f ora l ls e nt e nc e s ,‘ I fS,t he nS’i st r ue ,whe r et hes c he ma t i cl e t t e r‘ S’r a ng e s substitutionally over all sentences of the language.36 In other words, that the sche ma‘ I fp, t he np’i sl og i c a l l yva l i dc a nbepr ove di npr i nc i pl ebyor di na r ype opl ewhos pe a kt he natural language they speak. For Halbach, on the other hand, the disquotational truth predicate cannot play the most important role of expressing and affirming infinite conjunctions, so the notion of truth-analyticity is indispensable for this purpose. It seems that the notion of disquotational truth more or less fades out at this stage. More importantly, in order for the truth predicate to commute with objectual quantifiers, some intensional predicates of other semantic concepts are needed. The more we need, the less important the truth predicate is. Both accounts complicate the truth theory in order to establish the soundness of first order logic. But the original task is only to establish the claim that, by justifying that one can affirm an infinite lot of sentences by affirming the corresponding generalisation, infinite conjunctions can be expressed in terms of the truth predicate. Whether or not the price is so high, that the name, if not the spirit, of 35 Field proposes to develop a new kind of logic to solve the liar paradox, so he does not go for a hierarchical system. 36 I nv i e wofFi e l d’ squ a s i-disquotationalism, it seems that consideration regarding possible world semantics is comprised in his account. 193 deflationism is put in danger, is worth considering.37 Since neither Quine nor Horwich go f ort hes t r a t e g i e ss ugge s t e dbyFi e l d’ sa c c ounta ndHa l ba c h’ s ,t hedi s c us s i onwi l lbel e f t here. Now, let us return tot hedi s c us s i onofHor wi c h’ smi ni ma l i s te xpl a na t i on,b a s e dont he minimal theory of truth (MT), of the general fact mentioned at the beginning of this section. It runs as follows: 1. Log i cpr ovi de suswi t hf a c t sl i ke[ dog sba r k&( dog sba r k→ pi g sf l y ) ]→ pi gs fly, that is, with every fact of the form [p & (p → q) ]→ q. 2. Therefore, given MT, we can go on to explain every fact of the form [<p> is true & (p → q) ]→ <q> is true. 3. But from the nature of implication, we have all instances of (<p> implies <q>)→ ( p → q). 4. Therefore we can explain each fact of the form [<p> is true & <p> implies <q>]→ <q> is true. 5. And therefore, given MT, we get each fact of the form <[<p> is true & <p> implies <q>]→ <q> is true> is true. 6. But it is a peculiar property of propositions that any general claim about them—any characterization of all propositions—is made true by the infinite set of particular facts associating that characteristic with each individual proposition. 7. Therefore, in light of 5 and 6, we can explain the general fact: Every proposition of the form, <[<p> is true & <p> implies <q>]→ <q> is true>, is 37 Whether or not they make some extra metaphysical and/or epistemological commitments is not able to be discussed here. 194 true. (Horwich 1998b, pp. 21-22) This explanation is clearly unavailable to the disquotationalist, because of premise 6. This turns out to be insignificant, for Horwich later abandons this line of reasoning. Earlier, the success of this explanation was crucial for Horwich, to demonstrate the deductive power of the minimal theory, its capacity for deriving a universal generalisation from all its instances, and also the fulfilment of the necessary condition for the minimal theory to be an adequate theory of truth.38 It appears that because of premise 3, Horwich intends to explain the notion of logical implication. But this makes the whole explanation unsound, for the strong modal relation between <p> and <q> i n‘ <p> implies <q>’i smi s s i ng from premises 2, 3 and premise 4. Soi ti spr oba bl ymor ec ha r i t a bl et ot a keHor wi c h’ sa r g ume nta sbe i nga boutt henot i o nof material implication—about modus ponens. This slip does not affect my criticism of his a r g ume nta ny wa y .Is ha l lf ol l owDa vi d’ sobservation that this must be a typographical error and hereafter change 3 into the following: 3’( <p> implies <q>) ↔ ( p → q).39 The idea behind this explanation is apparently to take the general fact as an infinite c onj unc t i on,“ ‘ <Snowi swhi t e >i mpl i e s<Gr a s si sg r e e n>a nd<Snowi swhi t e >i st r ue , t he n<Gr a s si sg r e e n>i st r ue , ’‘ <Skyi sbl ue >i mpl i e s<Dog sba r k>a nd<Sky is blue> is t r ue ,t he n<Dog sba r k>i st r ue ,…” ,a ndt he nt oe xpl a i nhowwec a na r r i vea tt he conclusion that this conjunction can be expressed in terms of the minimalist truth predicate. And hence modus ponens—if one sentence/proposition implies another, and the 38 Premise 6 is new to the second edition of Truth, so is 7. It is not difficult to see why Gupta complains in bot h“ ACr i t i qu eofDe f l a t i oni s m”( 1993)a n d“ Mi n i ma l i s m”( 199 3)t h a tHor wi c h( i nt h ef i r s te di t i onof Truth, 1990) only explains all the instances of a generalisation but not the generalisation itself. I will skip the discussion of the old version. 39 Please cf. David 2000, note 19. 195 first one is true, then so is the second—can be explained in terms of the minimal theory along with some truth-f r e ee xpl a n a t or yr e s our c e s .Da vi d’ si nt e r pr e t a t i on(2000) is that Hor wi c h’ si de aofe xpl a i ni ngg e ne r a lf a c t si nvol vi ngt r ut hi st odeduce these facts from the minimal theory, along with some truth-free resources. This does not seem to be misleading, for it is Horwich himself who advances such an idea in some places.40 What David wants to show is that the deducibility of these general facts eventually boils down to an extended notion of loose deducibility. The r ea r es e ve r a lpr obl e msf orHo r wi c h’ sa r g ume nt . Themos ti mpor t a ntonei st ha t ,a sI discussed in the last chapter, premise 3 will inevitably go infinite at the level of schematic argument. Since we have the problem of being short of an infinitary rule of inference, this is probably the reason for Horwich to provide a meta-argument. Apparently it is the sixth premise that makes Horwich think he successfully proves the deducibility. In the Postscript of Horwich 1998b, Horwich says: However, it seems to me that in the present case, where the topic is propositions, we can find a solution to this problem [That is, there is no logically valid infinitary rule]. For it is plausible to suppose that there is a truth-preserving rule of inference that will take us from a set of premises attributing to each proposition some property, F, to the conclusion that all propositions have F. No doubt this rule is not logically valid, for its reliability hinges not merely on the meanings of the logical constants, but also on the nature of propositions. But it is a principle we do find plausible. We commit ourselves to it, i mpl i c i t l y ,i nmov i ngf r omt h edi s pos i t i ont oa c c e pta nypr opos i t i onoft h ef or m‘ x is F’ (where x i sapr opos i t i on )t ot h ec on c l u s i on‘ Al lpr opos i t i on sa r eF’ .( p. 137,t h el a s titalic word is mine) Supposedly, the disposition t oa c c e pta nypr opos i t i onoft hef or m‘ x is F’i ss ome t hi ngt o do with premise 5. The inference rule, which is characterised in 6, is thus to licence us 40 I na tl e a s tt wopl a c e sh es pe c i f i c a l l yu s e s‘ de r i v e ’( Hor wi c h1998 b, pp. 11-12)or‘ de du c e ’( pp. 24-25) to account for what he means to say about explaining general facts involving truth. 196 from premise 5, to derive 7. What Horwich does not tell us is what sort of nature possessed by propositions makes us implicitly commit to this rule of inference.41 David, on the other hand, offers an interpretation of this alleged nature of propositions. He says: I tmus tbebe c a u s e[ Hor wi c h ]t h i n kst h a tpr opos i t i on s …obe yt h ef ol l owi n gpr i n c i pl e :If a proposition exists in any possible world, then it exists in every possible world. But this principle holds only on some (e.g., Fregean) conceptions of the nature of propositions. (David 2000, p. 172) Russellian propositions are precluded from the minimal theory then, if this is the case. Thi sc ont r a ve ne sHor wi c h’ sa s s umpt i ont ha tmi ni ma l i s mi sc ompa t i bl ewi t hbot hFr e g e a n and Russellian propositions, and even compatible with a hybrid kind of these two (Horwich 1998b, p. 17). This also damages his account regarding the relation between beliefs and propositions (ibid., Question 31). For David, the more serious deficiency of Horwi c h’ sa s s umpt i ont ha tpr e mi s e6i sar ul eofi nf e r e nc e ,t owhi c hwea r ei mpl i c i t l y commit, is its failure of explaining how this envisaged rule might work in the absence of the assembling of infinitely many premises, or an infinite premise. Furthermore, this envisaged rule is only supposed to work for some specific objects, namely, propositions. This, according to David, shows that Horwich implicitly slackens the deducibility in need of accounting for the deductive power of the minimal theory, to a much looser one. Consequently, Horwich has to admit that substantive theories of truth can always find some equally loose deducibility to fit their explanation of their adequacy in accounting for our acceptance of all instances of the equivalence schema and all general facts involving truth. I do not think that Horwich is guilty of implicitly loosening the required deducibility. Because in Truth, second edition, Horwich states the following: 41 Armour-Garb expresses the same view in Armour-Garb 2004, note 13. 197 My inclination is to think that the principles of deductive logic are neither susceptible of, nor in need of, demonstrative justification. Instead, I would follow Quine (1953) in supposing that the logical principles relied upon in scientific theorizing are justified in so far as they are members of the simplest way of accommodating experience. I would say, moreover, that the logical principles deployed outside science in ordinary life—which may or may not differ from those used in science—are not subject to revision in light of experience. Hence these commitments are a priori; their justification is pragmatic; purely practical goals are furthered by adopting them.42 (p. 76, the first string of italics is mine) If there is no need to demonstratively justify the principles of deductive logic, within or without science, then what does it mean to say that the minimal theory, along with some truth-free explanatory resources, can derive or deduce general facts involving truth, including the principles of deductive logic such as the one I discussed?43 There seems to be an apparent incoherence. The more sympathetic interpretation is that the required deducibility is loosened in the very first place. What Horwich is guilty of is not making this explicit.44 Why Horwich eventually intends to justify those general facts by appealing to the unique nature of propositions is a separate question. In the cited words two paragraphs back, we can see that Horwich recognises the logical invalidity of the rule of inference characterised as such. He probably thinks, separately from the thought regarding explanation cum derivation/deduction of general facts involving truth, that the principles of deductive logic can be explained in an even looser sense. This shows yet another 42 This is partially the reason why Horwich does not go for the strategies advanced by Field. Moreover, the c l a i mt h a t“ t h el og i c a lpr i n ciples deployed outside science in ordinary life—which may or may not differ from those used in science—a r en ots u bj e c tt or e v i s i oni nl i gh tofe x pe r i e n c e ”r e s ul t sf r omt h en e e df or squaring this with his use theory of meaning. That is, the demarcation between common-sense logic and various kinds of logical theories. 43 Apparently it is not just about the principles of deductive logic but also about other general facts, such as that true beliefs about how to attain our goals tend to facilitate success in achieving them. To account for this general fact and not only its instances, it is inevitable that Horwich will appeal to the unique nature of propositions. 44 A sympathetic guess is that in revising the first edition of Truth these things did not coherently come to Hor wi c h’ smi n d. 198 incoherence in his account. Horwich holds that the minimal theories of utterance truth and belief truth can be constructed, with suitable auxiliary assumptions, from the minimal theory of propositional truth.45 He says: Ordinary language suggests that truth is a property of propositions, and that utterances, beliefs, assertions, etc., inherit their truth-like character from their relationship to propositions. However, the above derivations show that this way of seeing things has no particular explanatory merit. The truth-like conception for each type of entity is equally minimalistic. And by assuming any one of them we can easily derive the others. (Horwich 1998b, p. 102) But the minimal theories of utterance truth and belief truth certainly cannot provide the explanation of the general fact of modus ponens in the way characterised above, for that e xpl a na t i onde pe ndsont heuni quena t ur eofpr opos i t i ons . ThusHor wi c h’ smi ni ma l i s t account is incoherent. There is another notable fact, indicated by Armour-Grab (Armour-Grab 2004, note 13), t ha tma ke sHor wi c h’ spos i t i one nt i r e l yi nc ohe r e nt . Atthe very beginning of Truth, Question 1, that is, Horwich specifically makes the following claim: However this commitment [to the existence of propositions], though controversial and in n e e dofs omede f e n c e( …) ,i smu c hl e s ss u bs t a n t i a lt h a ni tmi g h ts e e mat first. For it presupposes very little about the nature of propositions. (p. 16) The reason given by Horwich, to relieve the worry of introducing substantial elements into the minimal theory by committing ourselves to the existence of propositions, is that this commitment presupposes very little about the nature of propositions. But premise 6 is important for Horwich to account for the loose deducibility of general facts involving truth. The adequacy of the minimal theory turns out to be absolutely hinging on not just the commitment to the existence of propositions but also on their very nature. The 45 Please cf. note 15 of the previous chapter or Horwich 1998b, pp. 98-102. 199 minimal theory is thus not very minimal, much less deflationary. In latter times, Horwich seems to modify his argument, abandoning the idea of appealing to t heuni quena t ur eofpr opos i t i ons .I n“ ADe f e ns eofMi ni ma l i s m”( 2001) ,Hor wi c h r e pl i e st oGupt aa ndSoa me s ’c r i t i c i s mt ha tmi ni ma l i s m onl ye xpl a i nshowt or e a c ha l l instances of a generalisation but not the generalisation itself.46 Gupta in Gupta 1993a and 1993b made the criticism that Horwich refers to and Soames in Soames 1997. In Horwich 1998b, the argument about modus ponens characterised above is supposed to be one of Hor wi c h’ smos ti mpor t a ntr e pl i e st ot hi sc r i t i c i s m.Ne ve r t he l e s s ,i nHor wi c h2001,i n accounting for general facts involving truth he substitutes another premise for premise 6 mentioned above. It seems that he withdraws the original strategy in Horwich 1998b, though he does not make this explicit, at least not in this article. The new premise added to truth-free explanatory resources is as follows: Whenever someone can establish, for every F, that is G, and recognizes that he can do this, then he will conclude that every F is G. (Horwich 2001, p. 157) In explicating this premise, Horwich continues: We can establish, for every proposition of the form, (p → p), that it is true. Moreover, we can become aware of this ability. For we can see that any of the propositions in question may be designated by some expression of a certain form. And we can also come to see that there is a general strategy which, given any such expression, will establish the s e n t e n c et h a tr e s ul t sf r omwr i t i n gt h ee x pr e s s i onbe f or e“ i st r u e ” .( i bi d. ) Thi sma t he ma t i c a li nduc t i ver e a s oni ngs qua r e swi t hHor wi c h’ sa t t i t udet owa rd the principles of deductive logic, and thus makes his account more coherent. If adding this 46 This is actually part of the criticism. In terms of this part, Gupta and Soames go on to conclude that the equivalence schema does not really suffice to account for all uses of the truth predicate. It follows that the me a ni ngof‘ t r u e ’i sn otf i x e dbyou ri n c l i n a t i ont oa c c e pta l li n s t a n c e soft h ee qui v a l e n c es c h e ma .Howe v e r , it is clear that the key issue here is about the explanation of how we move from affirming all instances to affirming the corresponding generalisation. 200 new premise validates the argument, and accordingly all generalisations involving truth are explained in terms of this line of reasoning, it does not seem that this line of reasoning is not available to disquotationalism.47 For the significance of this premise does not rely on the nature of propositions. And this reasoning is especially well-s ui t e dt oQui ne ’ s thought that truth, even logical truth, hinges on reality; that in each instance of the g e ne r a l i s a t i on,s e nt e nc e swi t ht hef or m‘ p → p’a r et r ue ,oure y e sa r eonr e a l i t y —that is, for instance, if snow is white then snow is white. Hence I again fail to find any reason to take minimalism as more desirable than disquotationalism. There is a possible objection to my claim that this strategy is also available to the disquotationalist. One might cast some doubt on the claim that disquotationalism is entitled to do so, by arguing that what confers the ascription of the truth predicate to an expression is the proposition designated by it. This line of reasoning is not applicable to utterances or sentences, for, without the propositions designated by them, their interpretations can vary from one case to another. Contrast this criticism to my interpretation of disquotationalism, however, and it is easy to see that it is not justified. For in every instance of the disquotational schema, the sentence mentioned and used is with a fixed interpretation. According to Armour-Gr a b( 2004) ,howe ve r ,Hor wi c h’ sne ws t r a t e gyi si nt e ns i onwi t h his treatment of the liar paradox. Remember the basic claim of the minimal theory, that bya ppl y i ngt hepr opos i t i ona lf or m or ,i not he rwor ds ,pr opos i t i ona ls t r uc t ur eof‘ <p> is true iff p’t oevery proposition, we yield an axiom. In dealing with the liar paradox, using 47 The soundness of this new premise will not be discussed hereafter, for it is not the main issue. 201 ‘ #’t oa bbr e vi a t eal i a r -pa r a doxi c a li ns t a nc e“ THEPROPOSI TI ONFORMULATEDI N CAPTI TALLETTERSI SNOTTRUE, ”Hor wi c hs a y s : But [to deny classical logic] cuts too deep; [to deny (à la Tarski) that the concept of truth can be coherently applied to propositions, such as <#>,] also smacks of overkill; and [to deny that the sentence in capital letters succeeds in formulating a proposition] goes against the fact that, for any condition C, one might happen to believe that the proposition meeting that condition is not true—whi c h…wou l di mpl yt h a t‘ Th epr opos i t i onme e t i ng condition C i sn ott r u e ’e x pr e s s e sapr opos i t i on …Th e r e f or et h eon l ya c c e pt a bl es ol u t i on is [to reject certain instances of the equivalence schema]: only certain instances of the equivalence schema are correct.48 (Horwich 1998b, p. 41) Horwich does not provide any solution in order to bar liar-paradoxical propositions in Horwich 1998b. In a footnote (21), he acknowledges what Gupta pointed out to him that the need to restrict the instantiation of the equivalence schema is in tension with the ability of the minimal theory to capture schematic generalisations (p. 42). For not every instance of a given generalisation is entailed by the generalisation. However, Horwich says, [S]uch problem cases are few and far between; so the utility of truth as a device of generalization is not substantially impaired by their existence. (Horwich 1998b, p. 42) This only justifies why we should maintain the minimal theory, because of its utility, but does not justify the theory itself. Similarly, the strategy of appealing to the nature of propositions certainly is not helpful either, and the same is true for the new mathematical inductive reasoning. Normally there are two options to deal with liar-paradoxical expressions. We can follow Ta r s ki ’ si ns i g hta ndg of orahi e r a r c hi c a ls t r uc t ur eofl a ng ua g e s .I ne a c hs t a g ewer e s t r i c t the target language to consist only of sentences without the truth predicate or without the higher order truth predicate. Or alternatively, we can adopt deviant logic to allow 48 Wor dsi ns qu a r ebr a c k e t sa r eHor wi c h’ s ,i nt h epr e v i ou spa r a g r a phofmyqu ot e . 202 truth-value gaps into which the liar-paradoxical expressions fall. Both options have their deficiencies. Going hierarchical is probably too artificial. Adopting deviant logic is, as Horwich argues, to cut too deep. Nevertheless, the first option is not available to Horwich, for he has no way to go hierarchical.49 And there is good reason why he should not, be c a us et heme a ni ngof‘ t r ue ’i st h e nnotf i xe dt hr oug houri nc l i na t i ont oa c c e ptt he instances of the equivalence schema. Horwich apparently turns down the second option for a similar reason that this jeopardises his use theory of meaning. Consider his instance of (AND) discussed in the previous chapter. Presumably, the counterparts of all logical connectives can be defined use-theoretically in a similar way. Adopting deviant logic creates problems for this. Notice though, that these two options are open to the di s quot a t i ona l i s t .I nmyf or mul a t i onofQui ne ’ sdi s quot a t i ona l i s m,Ig oa l lt hewa ydown in terms of the first option. The second option is also available to Quine, though I think that it is not very Quinean. Weha venotdi s c us s e dHor wi c h’ sr e s pons ey e t . Thet e ns i oni st ha t ,i nde a l i ngwi t ht he generalisation problem, liar-paradoxical propositions cause trouble, and in dealing with the liar paradoxes, the explanation of truth generalisations is incomplete. Presumably, any principle or additional axiom added to the minimal theory has to satisfy both sides that are in tension. That is, by restricting the instantiation of the equivalence schema, the explanation of truth generalisations does not exclude any instance. We can agree with Horwich that the utility of truth as a device of generalisation is not rejected by the few and far between cases of the liar paradox, but the minimal theory is. One way to do it is to 49 In a footnote (McGrath 1997b, p. 74, 10), McGrath suggests that we can distinguishes between all propositions, those with the truth predicate in various ways and those without the truth predicate, and that we can allocate them to different levels. Relative to each level, then, we can develop a theory of propositional truth for that level. It is doubtful that this is doable. But I will ignore this approach. 203 convert all truth generalisations into axioms of the minimal theory. But this is clearly not acceptable for Horwich for the same reason that he does not go for the two options mentioned above. Another way is to add the t e r m‘ non-pa r a doxi c a l ’t oa l lt r ut h generalisations to make them restricted. For instance, for modus ponens, if one non-paradoxical sentence/proposition implies another non-paradoxical one, and the first one is true, then so is the second. But according to Armour-Garb (2004, pp. 499-500), not all truth generalisations can be treated in the same way. For instance, true non-paradoxical disjunctions have at least one true non-paradoxical disjunct. Liar-paradoxical propositions can still be one of the disjuncts.50 As McGrath indicates, that a proposition might generate such a paradox is in some cases a contingent matter (McGrath 1997b, p. 74, note 10). It follows that there is no a priori method for identifying all and only those problematic propositions. We can conclude for this section that the minimal theory has serious problems in dealing with the liar paradox. The consequence is that its explanatory capacity for affirming truth generalisations is compromised. On the other hand, disquotationalism is in a better shape. Unl e s sHor wi c h’ smi ni ma l i s m ha sadr a s t i ca ndc ons i s t e ntc ha ngei nbot hoft hemi ni ma l theory of truth and the use theory of meaning, it seems that disquotationalism is a better choice. 3. The Proper Ascription of the Truth Predicate Many opponents incline to regard deflationary theories of truth as having genuine difficulty in accounting for truth ascription. The issue is about truth ascription to 50 Armour-Garb also gives a good counterexample in the case of blind ascription. Cf. Armour-Garb 2004, pp. 500-501. 204 non-understood sentences. The difficulty of truth ascription under this condition is divided into four types. They are, non-understood sentences of a given language of a given speaker (NUSL), non-understood sentences of different languages (NUSDL), sentences of the possible extension of a given language that are non-understood now (NUSPEL), and sentences that are in principle non-understandable to human beings (NUSH). Disquotationalism seems to be in more trouble with this difficulty. So I shall begin with a discussion of disquotationalist accounts. However, we will soon discover that this criticism focuses on disquotationalism because of some particular interpretation of this sort of theory. The criticism that disquotationalism has difficulties in accounting for the truth ascriptions to non-understood sentences, stems from various readings of disquotationalism, mainly ba s e donr e a di ng sofFi e l d’ sa c c o u nt ,pe r ha psMc Ge e ’ st oo,a nds ome t hi ngs i mi l a r . Amongot he r s ,t he r ea r eGupt a ’ ss e ns ee qui va l e nc er e a di ng ,Künne ’ sr e dunda nt i s tr e a di ng , a ndDa vi d’ sde f i ni t i ona lr e a di ng .Ia l r e a dys howe di nt hef i r s ts e c t i ont ha tGupt a ’ sr e a di ng results from an incoherent assumption and a misconception of disquotationalism, so in the f ol l owi ngonl yKünne ’ sr e a di nga ndDa vi d’ sr e a di ngwi l lbebr i e f l ydi s c us s e d. Si nc et he s er e a di ng sa r ema i nl ya boutFi e l d’ sdi s quot a t i ona lt heory, I should first elaborate his account a bit. Field claims that the pure disquotational truth predicate should be explained as true-as-one-understands-it in every instance of the disquotational schema, and each sentence on the right hand side should also be explained as what one understands. Hence the truth predicate is only applicable to sentences that a given speaker understands in the pure disquotational sense. Truth ascription to other sentences is 205 understood in the extended disquotational sense according to which the proper translation of non-understood sentences of a given speaker is needed. Two problems emerge. First, different speakers are inclined to have very different understandings of the truth predicate, and hence their ascriptions of it are different in principle. Second, there seems to be a conspicuous difference between cases of NUSL and those of NUSDL, for it is dubious whether or not the notion of translation can play any role in the cases of NUSL.51 Künne raises the following argument aga i ns tFi e l d’ spur edi s quot a t i ona l i s m.Suppos et wo people, Brad and Rachel, are talking about a newspaper report, and Brad comments on it bys a y i ng ,“ Tha ti st r ue ”whi l ei ndi s a g r e e me ntwi t hhi m,Ra c he ls a y s ,“ No,i ti snott r u e ” (Künne 2000, p. 189). The startling consequence of pure disquotationalism is that there is no disagreement between them at all. For the truth predicates used by them are different in principle.52 OrBr a dma ys a y“ I ft her e por ti st r ue ,t he nt heg ove r nme nti si nt r oubl e , ” and Rachel a gr e e swi t hhi mbys a y i ng“ Ther e por ti st r ue ” .Br a dt he nc onc l ude s ,“ The g ove r nme nti si nt r oubl e ” .I ts e e mst ha tt he ybui l dupa na r g ume ntba s e donmodus ponens ( i bi d. ) .Butt hi si sha r dl yt hec a s e ,f ora g a i nt he yus e‘ t r ue ’di f f e r e nt l y . Thi s criticism iss i mi l a rt oGu pt a ’ s : [ O] n ef i r s ta c qu i r e st h ec on c e pt‘ t r u e -in-my-present-i di ol e c t ’a n dt h e nu s i n gi ta c qu i r e s the full-f l e dg e d‘ t r u e . ’ Th epr o bl e mofe x pl a i n i ngh owon ege t sf r om ‘ t r u e -in-my-present-i di ol e c t ’t o‘ t r u e ’s e e ms …muc hh a r de rt h a nt h a tofe x pl a i n i ng‘ t r u e ’ u s i ngal i mi t e di de ol ogy ” .( Gupt a199 3,p. 81, n ot e48) Field owes us an explanation of how different speakers of the same language eventually 53 c omet ous e‘ t r ue ’i ndi f f e r e nt l y . 51 Even if the speaker uses a hybrid language, the distinction between the languages she speaks and those that she does not speak still makes sense. To simplify my discussion, only unilingual cases are in considering. 52 Imodi f yKü nn e ’ sa r g ume n tabi t to suit my argumentation. 53 Ia l s oma deas i mi l a robj e c t i ont oFi e l d’ sa c c oun ti n1. 2. Th a ti s ,t h ee x t e n de d- and quasi-disquotational 206 Now turn to the second problem. If the truth predicate is primarily applied to sentences that a given speaker understands, then there is a difference between non-understood sentences that belong to the language spoken by her and those that do not. For the latter, t het r ut ha s c r i pt i ont ot he mi se xpl a i ne dbyFi e l d’ sextended disquotationalism, according to which a non-understood sentence is disquotationally true only if there is a proper translation between it and a corresponding sentence that the speaker understands. For the former, the notion of translation has no place. It seems that Field has to use the idea of NUSPEL to explain the cases of NUSL. But there is supposed to be a difference between NUSPEL and NUSL. The consequence of pure disquotationalism, then, is the confusion between truth ascription to sentences of a given language that one does not understand now, and to sentences of the possible extension of it that one could not understand now. Da vi ds t r ong l ydi s a gr e e swi t hFi e l d’ spur edi s quot a t i ona l i s m. Thea us t e r e disquotationalism, David so argues, is its lack of reliance on any notion of meaning. However, the notion of understanding is within the same family of notions as meaning, and hence cannot legitimately to be used by a physicalist account of disquotationalism (David 1989, p. 608). For Field, and McGee as well, to reserve the necessary status of instances of the disquotational schema, the price to pay is to introduce the notion of unde r s t a ndi ngi nt ot hea c c ount .I tmi g htbec l a i me dt ha t‘ t r ue -as-one-understands-i t ’a nd ‘ a soneunde r s t a ndsi t ’a tbot hsides of pure disquotation sentences somehow cancel each other out—that is, it does not specify the understanding of the target sentence. In other t r u t hsdon otou t pa c et h egi v e ns pe a k e r ’ sownpu r edi s qu ot a t i on a lt r u t h ,un l e s swea r eg i v e ns omepr a c t i c a l hypotheses such as that the translation/interpretation manuals of people speaking the same language tend to be convergent. 207 wor ds ,t het wo‘ unde r s t a nd’a ppe a r i ngs t r i ng sofwor dsa r eonl yus e dt or e s t r i c tt her a ng e of substitutional va r i a bl e‘ u’i n“ ‘ u’i st r ue( a soneunde r s t a ndsi t )i fa ndonl yi fu (as one unde r s t a ndsi t ) ” . Ac c or di ngt oDa vi d,t hi si sonl yac he a t . Al t houg hDa vi d’ se xa mpl ei s his disquotational definition of truth, (D), which I discussed in Chapter Two, his point is also applicable to the above claim.54 The point David made is that: Of course we can keep a definition [or a full explanation of a schema] clean of u n c omf or t a bl en ot i on si fweon l yu s et h os en ot i on si na n“ i nf or ma lc omme n t a r y ”ont h e definition [or the schema]; but in this way we only keep the definition [or the schema] clean and not the theory in which the definition [or the schema] is embedded. (David 1989, p. 608) AsIi ndi c a t e de a r l i e ri n1. 2,Fi e l d ’ spur edi s quot a t i ona l i s mi si na de qua t ei nbe i ng implicitly committed to the sense equivalence relation. Conversely, this criticism is not applicable to (DQ), for the notion of interpretation involved is not about the sentences appearing at the two sides of the biconditionals. If one has any problems with it, one s houl de qua l l yobj e c tt oTa r s ki ’ sf or mul a t i onofConve nt i onT. Fur t he r mor e ,Sha pi r or a i s e sa na r g ume nta g a i ns tFi e l d’ sdi s quot a t i ona la c c ounti nSha pi r o 2003. Here is the background case on which the argument depends. A guru makes a pronouncement about set theory; a disciple thinks that everything the guru says is true; t hedi s c i pl edoe snotunde r s t a nds e tt he or y ;al og i c i a nwhodi s t r us t st heg ur u’ s set-theoretic pronouncements would like to draw their number-theoretic consequences, which the disciple does understand.55 Shapiro argues that if in addition the disciple trusts t hel og i c i a n’ sa c ume na boutl og i c( t houg hnota bouts e tt he or y ) ,t he nt hedi s c i pl eoug htt o be able to reason in terms of the notion of logical consequence, that the number-theoretic 54 The definition is as follows: (D) x is a true sentence =DF ( ∑p) ( xi s‘ p’ ,&p)( Da v i d1989, p.60 4;Da v i d1994 ,p. 89). 55 Th i si sa c t u a l l yFi e l d’ sc h a r a c t e r i s a t i onofSh a pi r o’ sc a s ei nPos t s c r i pt( 20 01)ofFi e l d199 4. Wh e nFi e l d ma det h i sc h a r a c t e r i s a t i on ,Sha pi r o’ spa pe rwa sn oty e ti npubl i c a t i on . 208 c ons e que nc eofwha tt heg ur us a y si st r ue . Theups hotofSha pi r o’ sa r g ume nti st ha t , though the disciple understands the conclusion, the reasoning is frustrated by the pure di s quot a t i ona lf a c tt ha tt hedi s c i p l ec a nnotme a ni ng f ul l ya ppl y‘ t r ue ’t ot heg ur u’ s utterances. In considering how to extend the truth schema as we expand the language, Fi e l dc onc e de st ha t“ onc ewec omet or e g a r da ne xpr e s s i onSa sa na c c e pt a bl ede c l a r a t i ve sentence, then even if we have no understanding of it (or virtually none) still we ought to a c c e ptt hec or r e s pondi ngi ns t a nc eoft het r ut hs c he ma ”( Pos t s c r i pt( 2001)ofFi e l d1994,p. 147).56 Field also acknowledges that, according to the cognitive equivalence thesis of pure disquotationalism, truth attribution to an expression not yet understood is out of the question. In order to settle this conflict, he proposes to think of this matter in terms of indeterminacy—that is, [ A]s e n t e n c eIdon ’ tu n de r s t a ndi sma xi ma l l yi n de t e r mi n a t ei nc on t e n tf orme ,a n ds oi s an attribution of truth toi t ;bu tt h ei n de t e r mi n a c i e sa r e‘ t i e dt og e t h e r ’ .( i bi d. ) Ac c or di ngt oFi e l d,t hi smodi f i e dvi e wha ndl e sSha pi r o’ sc a s ene a t l y . Thedi s c i pl er e g a r ds t heg u r u’ ss e nt e nc e sa s“ i napot e n t i a le xpa ns i onoft hedi s c i pl e ’ sl a ng ua g e ,s ot ha tt he r ei s 57 no difficul t yi nhi sc a r r y i ngoutt her e a s oni ng ”( i bi d. ,p.148) . In addition, Field argues that the guru problem is also a prima-facie problem for minimalism (ibid.). For the s y nony mousr e l a t i onbe t we e nt hedi s c i pl e ’ se xpr e s s i onsa ndt heg ur u’ si suna va i l a bl et oa mi ni ma l i s t . The r e f or e ,Hor wi c ha l s oha st opr ovi dea ne xpl a na t i onofhowt hedi s c i pl e ’ s reasoning is possible. I tdoe ss e e mt omet ha tFi e l dg e t st hewr ongi de aofSha pi r o’ sc r i t i c i s m.Fr ommy 56 It seems to me that, in the case of language expansion, the schematic reasoning introduced by Field to natural language is capable of accommodating the acceptance of the corresponding instances of the truth schema. However, as a full-blooded physicalist, Field still relies on computational psychology to explicate his pure disquotationalism. 57 Fi e l di n di c a t e st h a th i sr e pl ys t i l lh ol dsi ft h egu r u’ sl a n gu a g edi v e r g e sf r omt h edi s c i pl e ’ s .I nt h i sc a s e , t h ee x pa n s i onoft h edi s c i pl e ’ sl a n gu a g ei st ot r a ns l a t et h egur u’ sutterances. 209 formulation of disquotationalism, mainly with respect to the schema (DQ), it is not hard to find an explanation based on translation or possible expansion of the language to accommodate this problem. My theory does not hinge its interpretation of the corresponding schema on the notion of understanding. It is, then, a practical question whe t he rornott hedi s c i pl ei nSha pi r o’ sc a s ec a nr e a s oni nt e r msofhi sa s s umpt i ont ha t e ve r y t h i ngt heg ur us a y si st r ue , i nt e r msoft hel og i c i a n’ sa c ume ni nl og i c ,a ndt heno t i on of logical consequence. It depends on the justificationoft hedi s c i pl e ’ sfaith i nt heg ur u’ s ut t e r a nc e s ,a swe l la si nt hel og i c i a n’ sa c ume ni nl og i c .Le tuss uppos et ha tt hel og i c i a ni s trustworthy. If the part of everything the guru says that the disciple understands comes out true, then the disciple at least is at a departure point for some inductive reasoning. If the remaining part also comes out true after the disciple expands his language or translates t hos eut t e r a nc e si nt ohi sl a ng ua g e ,t he nt hedi s c i pl e ’ sf a i t hi sf ul l yj us t i f i e d.Howe ve r , Field is ne ve ri napos i t i ont or e ga r dSha pi r o’ sc a s ea sapr a c t i c a lque s t i on,f ort he di s c i pl e ’ swhol er e a s oni ngi sbl o c ke di nt hef i r s tpl a c e .Be c a us eoft hi s ,Fi e l dha st o a ppe a lt oi nde t e r mi na c yt or e s po n d,c umbe r s ome l ype r ha ps ,t oSha pi r o’ sc r i t i c i s m, whereas my disquotational theory does not have to. Re g a r d i ngFi e l d’ sc r i t i c i s mt oHor wi c h,It hi nkt ha thema ke sag oodpoi nt .Si mi l a rt o Fi e l d,t hedi s c i pl e ’ sr e a s oni ngi sbl oc ke da tt heve r ybe g i nni nga c c or di ngt ot hemi ni ma l theory of propositional truth. Shapir oc r i t i c i s e sHor wi c h’ smi ni ma l i s ms i mi l a r l y . The disciple is unable to understand the corresponding instances, about what the guru says, of the equivalence schema. He can only express it vacuously in terms of his faith in the guru and the logician who tells him that the set-t he or e t i cut t e r a nc eoft heg ur u’ si sal e g i t i ma t e declarative sentence. Furthermore, it is dubious why the disciple should believe in the 210 notion of logical consequence, for he only has instances of what he understands. In other words, Hor wi c h’ sma t he ma t i c a li n duc t i ver e a s oni ngha si t se f f e c t i ve ne s sonl yf or sentences that one understands. For sentences that one does not understand, there are no corresponding propositions to be grasped.58 It is the very nature of propositions, it seems tome ,t ha tma ke st hedi s c i pl e ’ sr e a s oni ng fail. That is, one does not know what proposition is expressed if one does not understand the expression (does not know the meaning of it). The fact that the disciple does not unde r s t a ndt heg ur u’ sut t e r a nc e si mpl i es that the disciple does not know what propositions are expressed by those utterances. Consequently, the disciple cannot apply t hemi n i ma l i s tt r ut hpr e di c a t et ot heg ur u’ swor ds .I not he rwor ds ,t he r ei sapr a c t i c a l problem of truth ascription for the minimal theory of propositional truth in the cases of NUSL. This also affects the explanation of general facts involving truth. The infinite conjunctions or disjunctions are not fully captured by attaching the truth predicate to generalisations, for the propositions not known by a speaker should not appear in the instances of the generalisations. The possible expansion of language does not much help here. Based on the assumption, or perhaps, as Horwich would say, common sense view, that for any declarative sentence there is a corresponding proposition, Horwich formulates his equivalence schema. For sentences that a given speaker does not understand, the function of the minimalist truth predicate that plays the role of forming generalisations, is only fulfilled by accepting the aforementioned assumption that those non-understood sentences express propositions because of their declarative feature.59 In other words, the 58 Please see the following criticism to Horwich. This assumption is not that unimportant. It is one of the main reasons why Horwich rejects moral expressivism. With respect to this, I am with Michael Smith (1994a and 1994b)in holding that deflationary theories of truth should not take any stand on issues like this. 59 211 minimalist truth ascription to many instances of a given generalisation is founded on our implicit, if not explicit, acceptance of (DQ). The function of the minimalist truth predicate only makes sense when (DQ)i spr e s uppos e d. The r e f or e ,t heve r yr e a s oni ngofHor wi c h’ s , which gives the primary position to the minimal theory of propositional truth, is in doubt. In dealing with propositions not yet expressible, Horwich refers to a possible extension of English to explain. He says: [ W] ec a nc h a r a c t e r i z et h e‘ e qui v a l e n c ea x i oms ’f orunf or mu l a t a bl epr op os i t i on sby considering what would result if we could formulate them and could instantiate those formulations in our equivalence schema. Thus we may specify the axioms of the theory of truth as what are expressed when the schema ( E)‘ <p> is true iff p’ is instantiated by sentences in any possible extension of English. Alternatively, instead of identifying the axioms indirectly in terms of how they would be expressed, we can solve the problem by directly specifying the propositional structure which all and only the axioms have in common. (Horwich 1998b, p. 19, note 3) What we do not know, according to this passage, is whether or not the minimalist truth predicate is applicable to propositions such as in the cases of NUSH. The option for the disquotationalist is clear. There is no intelligent way to apply the disquotational truth predicate to expressions (of some languages) that are in principle not understandable to us. Horwich, on the other hand, has no choice but to accept the very controversial assumption that we human beings are capable in principle of understanding all possible propositions. Again, there is no reason, as far as I can see, to give the minimal theory of propositional truth a privileged position. In terms of the first three sections, there is little difference between my formulation of disquot a t i ona l i s ma ndHor wi c h’ smi ni ma l i s m,e xc e pt ,ofc our s e ,mi ni ma l i s mi sat he or y 212 of propositional truth. If we take the equivalence schema not to apply to propositions but sentences, and providing that those conditions set up by Horwich for a proper understanding of the equivalence schema are satisfied, then we can hardly find any difference between (DQ) and (E). However, according to the analysis in this and the second section, there are some very serious deficiencies in the minimal theory of propositional truth. So we may say that disquotationalism seems to be preferable. 4. Is Quine a Disquotationalist? Whether or not Quine is an advocate of disquotationalism can be discussed in two parts. One is about some inconsistent passages about which theory of truth he is in favour of. For instance, in Quine 1995 he says: Science is seen as pursuing and discovering truth rather than as decreeing it. Such is the 60 idiom of realism,a n di ti si n t e g r a lt ot h es e ma n t i c soft h epr e di c a t e‘ t r u e ’ . (p. 67, emphasis mine) The other part is about his theory on meaning. For instance, in Quine 1981b he states the following: “ I twa sl e f tt oDa v i ds ont or e c og n i z eTa r s k i ’ st h e or yoft r u t ha st h ev e r ys t r u c t u r eofa t h e or yofme a ni ng ”( p.38) . We will first spend some space to settle the former. According to Davidson, Quine is among those who are in favour of deflationism. This seems to be correct, for Quine does urge that disquotationalism is all there is to say about truth. For instance, in Quine 1990b he says the following: [T]he r ei ss u r e l yoni mpug ni ngt h edi s qu ot a t i ona c c oun t ;n odi s pu t i ngt h a t‘ Sn owi s 60 Davidson suggests that this passage seems to suggest a non-deflationary position. This is probably not t h ec a s e .Con s i de r i ngGr ov e r ’ sv i e wpoi nt( Gr ov e r2000 )t h a tt h ei mpor t a n c eof scientific pursuit is to search for what-is-true, according to his terms, the italic word two sentences back can be perfectly explained as the collection of all what-is-trues in terms of disquotationalism. 213 whi t e ’i st r u ei fa n don l yi fs nowi swhi t e .Mor e ov e r ,i ti saf u l la c c oun t :i te x pl i c a t e s clearly the truth or falsity of every clear sentence. It is even a more than full account: it imposes a requirement on the truth predicate that is too strong for any predicate within the language concerned—on pain of contradiction. (p. 93) Ofc o u r s e ,t he r ei sa l s oa na l mos ti ndi s put a bl epa s s a ge :“ Tr ut hi sdi s quot a t i on”( p. 80) . However, there are also unmistakeable passages suggesting that we should not interpret Qui ne ’ sdi s quot a t i ona l i s ma sde f l a t i ona r y . The r ei sas t r ongf l a vouroft hec ohe r e nc e theory of truth in the following passage: It is rather when we turn back into the midst of an actually present theory, at least hypothetically accepted, that we can and do speak sensibly of this and that sentence as true. (Quine 1960, p. 24) Wh e r ei tma k e ss e n s et oa ppl y‘ t r u e ’i st oas e n t e n c ec ou c h e di nt h et e r msofag i v e n theory and seen from within t het h e or y ,c ompl e t ewi t hi t spos i t e dr e a l i t y . ”( i bi d. ,e mph a s i s mine). Then again, Quine seems to be in support of the correspondence theory as well. In Quine1990b he says: Yet there is some underlying validity to the correspondence theory of truth, as Tarski has t a u gh tu s .I ns t e a dofs a y i n gt ha t‘ Sn owi swhi t e ’i st r u ei fa ndon l yi fi ti saf a c tt h a ts n ow i swhi t e ,wec a ns i mpl yde l e t e‘ i ti saf a c tt h a t ’a sv a c u ous ,a n dt h e r e wi t hfacts themselves: ‘ Sn owi swhi t e ’i st r u ei fa n don l yi fs n owi swhite. To ascribe truth to the sentence is to ascribe whiteness to snow; such is the correspondence, in this example. (p. 80, emphasis mine) Sof a ra si ti shol ophr a s t i c ,f orQu i ne ,t os a yt ha t‘ Snowi swhi t e ’i st r uei si nf a c tt o ascribe whiteness to snow. This ascription conversely does not rely on the analysis of sentential structure. The connection between our understanding of the world and the world itself is through our sense experience. The empirical slack between the meagre sensory input and the torrential output allows us the latitude of ontological relativity to vary the reference schemes of a given theory while preserving the distribution of truths 214 (or, in other words, the sentence-to-sentence structure). Quine seems to vacillate between the coherence theory and the correspondence theory. The inclination to a coherence theory of truth, according to Davidson, is a mistake. This i nc l i na t i oni ss howni nQui ne ’ si nde c i s i onbe t we e nhi ssectarian and ecumenical positions towards choosing a true theory among empirically equivalent ones (Quine 1986). Some citations might help. For sectarian position, Quine states: One of the two systems of the world must be deemed false even if we know them to be empirically equivalent. My reason for it was naturalism. (Quine 1986, p. 156) For ecumenical position, he states: I opted for truth of both systems of the world, finding it offensive to my empiricist sensibilities to declare otherwise. (ibid.) The difficulty with the ecumenical position is as follows: Two empirically equivalent systems of the world may be logically incompatible, and hence incapable of being simultaneously viewed as true. (ibid.) Af t e rpe r f or mi ngDa vi ds on’ se xp e di e nt ,t ha t“ t hei nc ompa t i bi l i t yi sr e s ol ubl es i mpl yby re-construing some theoretical ter mi nt ha ts e nt e nc ea sapa i rofdi s t i nc thomony ms ” (ibid.), the ecumenical position does not have this difficulty since empirically equivalent systems of the world might be logically non-equivalent but never in conflict. Da vi ds ondoe snota c c e ptQui ne ’ sindecision. For him, the sectarian position is completely flawed. It is hard to see how a sentence, with interpretation unchanged, can be true for one person but not another or can be true for a person at one time but not at a not he r .“ Thi sdi f f i c ul t y , ”Da vi ds ona r g ue s ,“ s e e msduet ot hea t t e mptt oi mpor t e pi s t e mol og i c a lc ons i de r a t i oni nt ot hec onc e ptoft r ut h”( Da vi ds on1990,p.306) . Andf or Davidson the import of epistemological consideration means to follow the approach of a 215 coherence theory to account for the concept of truth.61 Wea l s oha veg oodr e a s ont or e j e c tQui ne ’ si nc l i na t i ont oac or r e s ponde nc et he or y .Eve n if Quine considers his theory of truth as some sort of correspondence theory, his theory is too different from the mainstream theories considered in Chapter Two to be included as one of them. The concept of truth does have a close relation to reality. As I said before, truth is about reality, about what there is. This is what semantic ascent is about. It is by semantic ascent that we can substitute the talk of sentences for the talk of reality. Thus to s a yt h a t“ Snowi swhi t e ”i st r uei st os a yt ha ts nowi si nr e a l i t ywhi t e . Thi ss e e mst obet he reason Quine claims that his theory of truth is a kind of correspondence theory (Quine 1990b, p. 80). As long as we do not possess some sort of telepathic communicative ability that does not involve any linguistic forms whatsoever, we need linguistic expressions to help us articulate facts. Whether or not we can specify a fact without any linguistic form involved is less important here. What is important is that we have to rely on linguistic f or mst os pe c i f yf a c t si na l lki ndsofc ommuni c a t i ons .He nc eQui ne ’ spoi ntc a nbe illustrated as the claim that any given sentence of a target language is true if and only if this sentence articulates a fact. According to most realists, truths as well as facts should be independent from scientific t he or i e s .Howe ve r ,Qui ne ’ si de a soft r ut ha ndf a c tdonotr e s pe c tt hi svi e w.Oura c c ountof what reality should be concei ve dof ,a c c or di ngt oQui ne ’ shol i s ma ndt he s i sof under-determination, makes no absolute sense.62 It depends on how we construct 61 According to Colin McGinn, disquotationalism can be used to reject the classic coherence theory of truth. Please cf. McGinn 2000, pp. 194-195. 62 The thesis of under-determination is that theories or theoretical formulations can be empirically equivalent—that is, they imply the same set of empirical consequences—while being logically incompatible 216 hypotheses to interpret pieces of empirical evidence. According to holism, we have the latitude to modify our theory, to redistribute truth-values over some sentences of our theory, even logical truths, when a conflict with experience happens.63 And the modification or redistribution of truth-values can occur in more than one way. It seems that we have different ways to capture reality. The thesis of under-determination pushes this to an extreme degree, for it implies that all possible observations fail to determine the unique nature of reality.64 Therefore, what sorts of facts we have are relative to the theory we hold. Otherwise, logically incompatible theories that account for all possible observations are unintelligibly in support of the same set of facts. So, different comprehensive scientific theories claim different sets of facts. It follows that talking about the correspondencer e l a t i oni na na bs ol ut es e ns ei spoi nt l e s s .He nc eQui ne ’ s correspondence theory is not the normal kind in which most realists are interested. Even with each other in a non-trivial sense. The logical incompatibility of two theoretical formulations might s i mpl ybet h a tt h e yh a v eadi f f e r e n tu s a g ef ort wot h e or e t i c a lt e r ms ,s a y ,‘ e l e c t r on’a n d‘ mol e c u l e ’ ,which do not appear in the linguistic vehicles of their empirical consequences. What are said in theoretical f or mu l a t i onTa b ou t‘ e l e c t r on’i sa bou t‘ mol e c u l e ’i na n ot h e rf or mu l a t i onU,a n ds i mi l a r l ya bou tt h ewor d ‘ mol e c u l e ’ . Th es e e mi n g l yl ogi c a li n c ompa t ibility coming from the divergent usage of terms, for Quine, is superficial and can be dispensed with by reconstrual of predicates. Wh a t‘ r e c on s t r u a lofpr e di c a t e s ’me a ns i s‘ a nyma ppi ngofou rl e x i c onofpr e di c a t e si n t oope ns e n t e n c e s ’( Qu i n e1975,p. 32 0). 63 One might argue that logical truths are different from other sentences in the sense that they are devoid of content, or that their truth has nothing to do with their content. The disputable part is whether logic is a compendium of the broadest traits of reality or it is just an effect of linguistic convention (Quine 1970, p. 96) .“ Mu s ta l lr i g h t -mi n de dme na g r e eonl ogi c ,ori si te v e r yl a ngu a g ef ori t s e l f ? ”( i bi d. )I nt r a ns l a t i n ga remote language, however, the logic of the translated language is at worst incommensurable with ours but never in conflict, for the conflict only discredits our translation. Following this line, it seems that logic is naturally an effect of linguistic convention. But the truth predicate plus some other Quinean theses—that is, holism, indeterminacy of translation, and inscrutability of reference—s e pa r a t el og i cf r oml a n gu a g e .“ Th e truth predicate serves the crucial purpose, in oblique generalisation, of disquotation. Logical theory, despite its heavy dependence on talk of language, is already world-oriented rather than language-oriented; and the t r u t hpr e di c a t ema k e si ts o. ”( i bi d. ,p. 97 ) 64 Gi l be r tHa r ma ni n“ Me a n i nga n dTh e or y ”s ug g e s t st h a tt h e r ea r et won ot i on sofe mpi r i c a le qu i v a l e n c e . The first one is that two theories are empirically equivalent if and only if they are equally supported by all observational evidence. The second is that two theories are empirically equivalent if and only if they are compatible with a l lobs e r v a t i on a le v i de n c e .I n“ OnPoppe r ’ sNe g a t i v eMe t h odol ogy ” ( Qu i ne1974 a), Quine argues that there is no confirmational support of natural laws or the like but only non-disconfirmational support of them. (cf. ibid., pp. 218-20) There is no observational evidence that can confirm a hypothesis or a law or the like pieces of evidence only support them in the sense that they do not refute them. In other words, what the pieces of observational evidence support is the compatibility between a theory and observations. Therefore, there is no genuine distinction betweent h ec on c e pt sof‘ e qu a l l ys u ppor t e dby ’a n d ‘ c ompa t i bl ewi t h’f orQu i n e . 217 though Quine seems to express his preference for realism in Quine 1995, his account of the correspondence theory has no attraction for other realists. The resultant account of realism that Quine supports is hardly the realism that we are familiar with. Qui ne ’ si nc l i na t i ons houl dnotbea boutt hec or r e s ponde nc et he or yoft r ut h,butr a t he r about a correspondence intuition regarding truth. In considering the passage cited from Quine 1990b ( p.93) ,t hi si sa ppa r e nti nt hef ol l owi ngwor dsquot e df r om “ Comme nton La ue n e r ”( Qui ne1990a): I keep to the correspondence theory of truth, but only holophrastically: it resolves out into Ta r s k i ’ sdi s qu ot a t i on a lv e r s i onoft r u t hr a t h e rt h a nac or r e s pon de n c eofwor dst oobj e c t s . (p. 229) The correspondence intuition can be phrased neatly as follows: the sentence mentioned and used in all instances of (DQ) is exactly the same sentence with the same interpretation. I not h e rwor ds ,whe noneus e sas e nt e nc e ,f ori ns t a nc e ,“ s nowi swhi t e ”t oma kea n 65 a s s e r t i ona boutt her e a l i t y ,onec a ne qua l l ya s s e r tt ha t“ Snowi swhi t e ”i st r ue . Now let us move onto the second part of this section. My formulation of disquotationalism, based on (DQ), as I indicated above, separates the theory of meaning ni c e l yf r omt het he or yoft r ut h.I fweha veat he or yofme a ni nga ss t r onga sHor wi c h’ s , 65 McGinn in McGinn 2000 adheres to what he calls thick disquotationalism, according to which the concept of truth is simple, without empirical essence or nature, unanalysable, but substantial (p. 202). Truth is a robust property and is disquotationally definable because it is the one and only self-effacing property whose application conditions can be stated without making reference whatsoever to itself (p, 198). Moreover, truth is essentially a device of inference; by learning that a certain proposition is true one can infer a fact of the world in terms of making a transition from the truth of the proposition to that very fact. Disquotation sentences accordingly are read as follows: from left to right, they are read as reality-implying—that is, by knowing that truth applies to a proposition one comes to know facts about the world (p. 199); from right to left, the truths are supervenient on the facts but do not collapse into facts and propositions, since it is an irreducible property (pp. 205-206). Except for seeing truth as a robust property, almost all the features above can be explicated in terms of my interpretation of (DQ). What thick disquotationalism does not tell us, is how to explain the feature of semantic ascent. How can such a substantial truth predicate form generalisations by which those corresponding infinite conjunctions and disjunctions are expressed? 218 then in the cases of NUSDL the disquotational truth predicate is equally applicable to sentences of other languages. Even under this consideration, disquotationalism is still more preferable than minimalism according to the analyses in the last two sections. If such a strong theory of meaning is not available, then truth ascription varies from one translation/interpretation manual to another. Disquotationalism under this condition is still in a better shape, for minimalism inevitably collapses. The genuine problem of my f or mul a t i onofQui ne ’ sdi s quot a t i ona l i s me me r ge s from his theory of meaning, from his inclination to accept truth conditional semantics. For instance, in Quine 1974b, he says: [I]n learning language, we are learning how to distribute truth values. I am with Davidson here; we are learning truth conditions. (p. 65) I ti sc e r t a i nl yt e mpt i ngt og oa f t e rMi c ha e lWi l l i a ms ’vi e wi n“ Me a ni nga ndDe f l a t i ona r y Tr ut h”( Wi l l i a ms1999)t ha t ,pr ovi di ngFi e l d’ sdi s quot a t i ona lt he or ya st hes a mpl e ,t he concept of truth need not play an explanatory but only an expressive role in the major parts of a Davidsonian theory of meaning. For example, the imputation of logical form to alien utterances, ascribing semantic relations of reference and satisfaction to nonlogical, subsentential components of utterances, and most importantly the principle of charity.66 Establishing this view, a Davidsonian theory of meaning can join the deflationary club. Howe v e r ,a c c or di ngt oJ e f f r e yHe r s hf i e l da ndDe bor a hHa ns e nSol e si n“ Re i nf l a t i ng Tr ut ha sa nExpl a na t or yConc e pt ”( 2003) ,t hi svi e wi s totally a misconception of a Davidsonian theory of meaning. 66 Al s oc f .“ OnSomeCr i t i c sofDe f l a t i on i s m”( 2002) ,e s pe c i a l l yp.156in which Williams argues that the r ol eof‘ t r u e ’pl a y e di nt h et r u t hc on di t i onsofobs e r v a t i ons e n t e n c e si sj u s tg e n e r a l i s i nga nde x pr e s s i ng . 219 I merely indicate two points made by Hershfield and Soles, which mean that any attempt to deflationise Davidsonian theory of meaning is doomed. The first point regards Wi l l i a ms ’ sc l a i mt ha tr e ading first order logic into alien utterances is neutral with respect to the issue of whether truth is deflationary, and the point runs as follows: Williams seemingly overlooks the ontological ramifications that come along with Da v i ds on’ sa ppl i c a t i onofaTarski-style theory. In particular, [Williams] seems to ignore the fact that the assignment of first order logical form, with a standard objectual interpretation of the quantifiers, carries certain ontological import: using such a theory to generate truth conditions for utterances of a speaker is, among other things, to impute to that speaker beliefs about the existence (or non-existence) of various (sorts of) things. (Hershfield & Soles 2003, p. 36) So the imputation of logic into alien utterances is to commit an alien informant to something ontologically to which a deflationary concept of truth cannot do. The second point is about the principle of charity, which Williams characterises as the maximisation of truth and observational adequacy: Da v i ds on’ spr i n ciple of charity is a precondition of any interpretation at all. As such it is part of a rationalizing enterprise which involves attributions of belief (and meaning and desire). The attribution of contentful belief can and will involve attributions of what the interpretee counts as evidence; it will also involve occasional attributions of erroneous belief, that is, the attribution of some belief to the interpretee which the interpreter does not hold true but for which the interpreter can find an explanation for the fact that the interpretee does hold it true. But error presupposes an explanatory concept of truth, for it represents the case where what is held true comes apart from what is actually true.67 (ibid., p. 37) It does not seem to make any sense, in both of the maximisation of truth and obs e r va t i ona la de qua c y ,t oba s eFi e l d’ se xt e nde ddi s quot a t i ona l i s mt ous et hema ppi ng r e l a t i onbe t we e none ’ sownut t e r a nc e sa nda l i e none s ,t oe xpl a i nDa vi ds oni a npr i nc i pl eof 67 In Horwich 1998a, Hor wi c hc l a i mst h a t ,e x c e ptt h edi v e r g e n c ei nt h ee x pl a n a t or yor de r ,“ Da v i ds on’ s principle of c h a r i t ya r g u a bl yboi l sdownt o[ hi s ]us et h e or yofme a n i ng ”( p.72) .Con s i de r i ngt h epos s i bi l i t y oft h ea s c r i pt i onofe r r orbe l i e f ,e i t h e rHor wi c h’ si n t e r pr e t a t i onoft h i si mpor t a n tpr i n c i pl eofDa v i ds on i a n theory fails, or he implicitly employs truth substantively as an explanatory concept. In either way, his ob j e c t i on st oDa v i ds on’ st h e or ya r es i gn i f i c a n t l ywe a k e n e d. 220 charity. Final remarks At the end of the day, I do not look at any possible reconciliation with sanguinity. There are only two options left. One is to go all along the line of disquotationalism and to abandon any idea of accepting the truth conditional theory of meaning. Or we can follow Davidson in seeing truth as a substantive primitive, and hence depart from disquotational reasoning. So we may come to the conclusion that either one should adopt truth conditional semantics or, if one tends toward deflationism, one should choose the Quinean disquotational theory based on (DQ)ove rHor wi c h’ smi ni ma l i s m. 221 Conclusion At the beginning of the Introduction, I listed some issues regarding deflationism that had yet to be clarified. These were: 1. What is a proper formulation of a disquotational schema of truth? 2. What is the modal status of disquotation sentences or the axioms of the minimal theory of proposition truth? 3. How sensible are the presuppositions involved in asking these questions about the modal status of sentences or axioms? 4. How should we give the deflationary theory of truth the deductive power it needs so as so as to adequately explain all general facts involving truth? 5. What is the deflationary conception of reference? Is it feasible to explicate the deflationary conception of truth in terms of the disquotational concepts of reference and satisfaction? 6. What is an appropriate account of propositions for use in a minimalist account of truth? 7. What sort of equivalence relation occurs within disquotational sentences or minimalist axioms of truth? 8. Between disquotationalism, taking sentences as the primary truth bearers, and minimalism, taking propositions as the primary truth bears, which is the more promising approach to a deflationary theory of truth? Hopefully this dissertation has made some contribution to a more appropriate 222 understanding of disquotationalism and minimalism. The first question is of course a very general one, which is answered throughout the work. The second question has been answered by the way in which I address many criticisms of disquotationalism, and some of minimalism as well, which stem from a misinterpretation of the disquotational schema of truth. For instance, David, Künne, and many others, criticise disquotationalism for falling short of accounting for the modal status of disquotation sentences. According to Mc Gr a t h,t hi si sa l s oapr obl e mf orHor wi c h’ smi ni ma l i s m.I nt hes e c onds e c t i onof Chapter Two I showed that this is the consequence of misinterpreting the disquotational schema of truth as ascribing the truth predicate to sentence types disregarding what contents they have. This sort of misinterpretation turns the whole reasoning of disquotationalism and minimalism into a grotesquery. Further, in Chapter Five I argued that the necessary status of disquotation sentences is not indispensable when explaining all general facts regarding truth—this of course also addresses the fourth question above. If there is any reason for disquotation sentences as well as the axioms of the minimal theory to be necessary, it must be because of something else. Thes e v e nt hque s t i oni sde a l twi t hi npa r tbymydi s c us s i onofGupt a ’ sc r i t i c i s m of disquotationalism and minimalism that the equivalence relation within any instance of the disquotational schema, as well as any instance of the equivalence schema, has to be sense equivalence in order to maintain the expressing feature of the truth predicate. In Chapter Five, I showed that the sense equivalence reading does not concern anyone as long as a suitable explanation of the reasoning of infinitary inference is given. And this reasoning in turn was discussed in the second section of Chapter Five. 223 We can see that these criticisms can be answered by a proper understanding and a proper formulation of the disquotational schema. Following the clarific a t i onofQui ne ’ s reasoning on the disquotational feature of the truth predicate, namely, semantic ascent and descent, my formulation of the disquotational schema within the truth language relative to a target language is given in Chapter Two. This formulation addresses the first highly general question in the list. The schema is: (DQ)‘ p’i st r uei f fp, whe r et hef i r s t‘ p’i sr e pl a c e dbya nys e nt e nc eoft het a r g e tl a ng ua g ea ndt hes e c ond ‘ p’byt hes a mes e nt e nc ewi t ht hes a mei nt e r pr e t a t i on. Further in Chapter Five, I argued that, in terms of this formulation, the sense equivalence as well as the necessary equivalence relations are not necessary in order for the truth predicate to express infinite conjunctions. The material equivalence relation is enough for the truth predicate to do this, providing that, of course, a suitable account of the reasoning of infinitary inference is available. Thus the second, third, and seventh questions from the list are addressed by my reply to objections raised by David, Künne, and many others, which do not arise according to the proper understanding of (DQ). Granted, the material equivalence relation of the disquotational schema, (DQ), is a bit stronger than the usual one, for the sentences mentioned and used in all disquotation sentences are restricted to the same sentence with the same interpretation. This is precisely why Quine takes truth as immanent t ohomel a ng ua ge s ,f o rTa r s ki ’ sT-sentences are not automatically instances of (DQ). The expressibility of infinite conjunctions in terms of the corresponding generalisations depends on both the material equivalence relation and the additional restriction about the sentence mentioned and used in every instance of the generalisations. However, this restriction does not yield any modal relation to the schema and its 224 instances. I go some way towards dealing with question five in my discussion of the disquotational conception of reference. Along with the disquotational feature of the truth predicate, the disquotationalist and the minimalist normally adopt a disquotational theory of reference, considering the disquotational schema of reference and all its instances to provide a genuine reference scheme for the target language. Some disquotationalists, such as Halbach and McGee, even claim that the disquotational concept of truth can be explicated in terms of the disquotational concepts of reference and satisfaction. I showed in Chapter Fi vet ha tMc Ge e ’ sa c c ounti sba s e donami s unde r s t a ndi ngofQui ne ’ si nde t e r mi na c y t he s i s . Ha l ba c h’ s account, on the other hand, is different. In order to enhance the deductive power of the disquotational theory of truth, Halbach proposes to explicate the concept of truth in terms of the inductive axioms of reference and satisfaction. In Chapter Three, I argued that both claims cannot be accepted by Quine. Question five is also dealt with in much of the second and third chapters—particularly wi t hr e f e r e nc et oHa l ba c h’ spr opos a l .I nt hos ec ha pt e r sIs howe dhi spr opos a lf a c e sa problem of the explanatory order, as well as a problem of falling short of implying all disquotation sentences of the target language. If the disquotational schema of reference is c ons i de r e da spr ovi di ngage nui ner e f e r e nc es c he me ,t he nQui ne ’ st he s e sofi ns c r ut a b i l i t y of reference a ndont ol og i c a lr e l a t i vi t ya r ej e opa r di s e d.ForQui ne ’ sma ne uvr e ,whi c he n ds the infinite relativisation of ontology in practice by acquiescing in our own language and taking its words at face value, amounts to the claim that the relativised explication of any given reference scheme of other languages or theories has to rely on unrelativised 225 semantic notions applied to our own language. In additional to this, I showed that the disquotational schema of reference and all its instances, if taken as providing a genuine reference scheme, cannot account for all the facts involving reference. For instance, consider the hidden indeterminacy of reference in pre-Newtonian formulation of physics. Ic onc l ude dt ha tadi s quot a t i ona la c c ountofr e f e r e nc e ,s uc ha sHor wi c h’ s , is too deflationary to account for referential indeterminacy. A disquotational schema of reference is useful in scientific reasoning, but serious problems emerge when it is seen as providing a genuine reference scheme. I return now to how I have dealt with the fourth question. One of the most important tasks of disquotationalism and minimalism is to explain all general facts involving truth. Therefore, it is important for disquotationalism and minimalism to show that they have enough deductive power to explain why infinite conjunctions or disjunctions can be e xpr e s s e di nt e r msoft het r ut hpr e di c a t e .If ound,i nCha pt e rFi ve ,t ha tHor wi c h’ sa c c ount of explaining general facts involving truth (in Horwich 1998b) is incoherent. On the one hand, he claims that the minimal theory of truth does not become substantial because of its commitment to the existence of propositions. For it presupposes very little about the nature of propositions. On the other hand, the crucial arguments for the capacity of the minimal theory of truth to explain all general facts involving truth depend heavily for their validity on the very nature of propositions. Moreover, Horwich claims that he follows Quine in supposing that the logical principles relied upon in scientific theorising are justified in so far as they are members of the simplest way of accommodating experience. Therefore, the principles of deductive logic, according to Horwich, are neither susceptible to, nor in need of, demonstrative justification. This, however, indicates 226 another incoherence. If there is no need to justify the principles of deductive logic, such as modus ponens, why does Horwich appeal to the nature of propositions to prove that the generalisation regarding the principle of modus ponens can be derived from all its instances? Later, Horwich abandons appealing to the nature of propositions, and turns to mathematical inductive reasoning. That is, whenever someone can establish, for every F, that is G, and recognises that he can do this, then he will conclude that every F is G. This reasoning, however, generates problems regarding the liar paradox for minimalism. In sum, either the truth predicate does not apply to every proposition, or the applications pr oduc epa r a doxe s ,butt he r ei snowa yt ha tHor wi c h’ sa ccount can get away from this. Chapter four contains much of my discussion of the sixth question. In this chapter I c ons i de rHor wi c h’ sc l a i m,f ol l owi nghi si l l us t r a t i onofwha twor dme a ni ng sa nds e nt e nt i a l me a ni ng sa r e ,t ha tQui ne ’ si nde t e r mi na c yt he s i sdoes not successfully repudiate meaning entities. He argues that this is because reference is not inscrutable. Hence, he argues, Qui ne ’ spr oxyf unc t i ona r g ume ntdoe snots howt ha tt he r ea r eonl ya de qua t et r a ns l a t i o n manuals, but not a correct one. In discuss i ngHor wi c h’ sQQQa c c ount ,If i r s ti ndi c a t e d t ha thi sf or mul a t i onoft hea l l e g e da r g ume ntofQui nei sa c t ua l l ynotQui ne ’ s . The e mpha s i sofQui ne ’ si nde t e r mi na c yt he s i si si na c c ur a t e l yboi l e ddown,byHor wi c h,t ot he inscrutability of reference. Probably because of this, the importance of the proxy function a r g ume nti sove r e s t i ma t e dbyHor wi c hi ne va l ua t i ngQui ne ’ si nde t e r mi na c yt he s i s .I c r i t i c i s e dHor wi c h’ sdi s quot a t i ona lt he or yofr e f e r e nc ei nCha pt e rThr e e . Wha tHor wi c h f a i l st ounde r s t a nda boutQui ne ’ s indeterminacy thesis, is that in the case of radical 227 t r a ns l a t i on,wi t hr e g a r dt ot het ot a l i t yofhe ri nf or ma nt s ’l i ng ui s t i cdi s pos i t i ons ,ourf i e l d linguist does not have the luxury to hope for any evidence beyond this. Even if we get the totality of assertability conditions of our informants, we cannot establish semantic facts of the translated language. True, if the proxy function argument is valid, then reference is inscrutable and the ontology of a given language or a given theory has to be relativised to be comprehended. Hence translation of another language is indeterminate. In contrast to Hor wi c h’ sove r e s t i ma t i onoft hei mpor t a nc eoft hepr oxyf unc t i ona r g ume nt ,howe ve r ,t he indeterminacy thesis does not hinge its significance on the validity of this argument. The notion of proxy function is used to make a point regarding semantical reasoning. The notion itself is not a semantic notion. To actually infer a reference relation from its proxy or the other way round is irrelevant and incidental. I also di s c us s e dHor wi c h’ sus e -theoretic account of meaning in Chapter Four. Based on his view according to which the meaning property of a word is metaphysically indistinguishable to its use property, Horwich arrives at the conclusion that each word has its (explanatorily basic) use property, according to which the overall use of a word is best explained. A word of a language is synonymous with a word of another language so long as they have exactly the same use property. In terms of his account of compositionality, applying a schema with a schematic property to the constituents of the sentence in an exact order, one gets a construction property for each sentence. Similarly, two sentences of different languages are synonymous if and only if they have the same construction property. Consequently, a proposition can be explicated as the construction property of the sentence expressing it. Therefore, a successful formulation of a use-theoretic account of propositions makes a minimal theory of propositional truth more materialised. 228 Thef i r s tpr obl e mofHor wi c h’ su s e -theoretic account of meaning is that the explanatorily ba s i cus epr ope r t yofawor dde p e ndsont hedi s t i nc t i onbe t we e nt hewor d’ sba s i cus ea nd the rest. Horwich thinks that a use property of the basic use of a word is explanatorily basic if and only if it can best explain the not basic use of the word. However, he does not give us a satisfactory account of how to distinguish the basic use of a word from the rest of its use. It is very likely that this distinction eventually boils down to a contingent fact that can be known only through an empirical or scientific inquiry. This implies that there might be equally explanatorily basic use properties of a word, due to the contingency of the distinction. Secondly a ndmor ei mpor t a nt l y ,Hor wi c h’ se nt i r eus e -theoretic account of meaning depends for its plausibility on some contingent use facts of a language, in his case, English. He seems to assume that different languages all share the total contingent use facts of English. This is hardly acceptable. It is more apparent in his account of compositionality that, according to him, different languages share the same set of s y nt he t i cs c he ma t a . Thef unda me nt a ldi f f i c ul t yf orHor wi c h’ sus e -theoretic account of word meaning and compositionality is its fine-grained character. The particular semantic and synthetic features of foreign words are finely ground into meaningless pieces by Hor wi c h’ smi l l . The fourth and eighth questions are answered by how I show that the deficiencies in Hor wi c h’ sus e -theoretic account of meaning and compositionality suggest a missing link in his enterprise of deflationary semantics. Regardless of whether or not Horwich provides an acceptable account of propositions, the minimal theory of propositional truth is unable to explain all ascriptions of the truth predicate. In Chapter Five, in discussing 229 the proper applications of the truth predicate in the disquotational or the minimalist sense to sentences or propositions that one does not understand, I found that the minimalist truth predicate cannot be applied to sentences that one does not understand. For one does not know what propositions are expressed by those sentences. As a result, the minimalist truth predicate is not applicable to sentences that one does not understand. This further da ma g e sHor wi c h’ sma t he ma t i c a li nduc t i ver e a s oni ngr e ga r di ngt hemi ni ma l i s t explanation of general facts involving truth. This reasoning has its effectiveness only for sentences that one understands, but not for all propositions. The possible expansion of language does not much help here. Based on the assumption or common sense that for any declarative sentence there is a corresponding proposition, Horwich formulates his equivalence schema. For sentences that a given speaker does not understand, the function of the minimalist truth predicate that plays the role of forming generalisations, is only fulfilled by accepting the aforementioned assumption that those non-understood sentences do express propositions because of their declarative feature. In other words, the minimalist truth ascription to many instances of a given generalisation is founded on our implicit, if not explicit, acceptance of (DQ). I nt hef i na ls e c t i onofCha pt e rFi ve ,Ibr i e f l ydi s c us s e dQui ne ’ spos s i bl yincoherent attitudes toward the matter of truth. I did not see any reconciliation between his disquotational reasoning with respect to truth and his advocacy of a Davidsonian truth c ondi t i ona la ppr oa c h.I nt e r msoft hec ompa r i s onbe t we e nQui ne ’ sdi s quot a t ionalism and Hor wi c h’ smi ni ma l i s m,h owe ve r ,Ic onc l ude dt ha ti ft he r ei sas t r ongt he or yofme a ni ng , then we can have a general account of disquotational truth or minimalist truth. Even given t hi ss i t ua t i on,Qui ne ’ sdi s quot a t i ona l i s mi spr e f e r a bl et oHor wi c h’ smi ni ma l i s m.I ft he r e 230 is no strong theory of meaning, then disquotationalism has to cope with the following problem: how do speakers of different languages, or even the same language, use the truth pr e di c a t e ?ButHor wi c h’ smi ni ma l i s mi s ,ont heot he rhand, destined to fail. 231 Bibliography: Aki ba ,Ke n( 2002)“ Ca nDe f l a t i oni s mAl l owf orHi dde nI nde t e r mi na c y ? ” ,i nPacific Philosophical Quarterly, Volume 83, pp. 223-234. Alston, W. (1996) A Realist Conception of Truth, Ithaca, NY: Cornell University Press. Armour-Ga r b,Br a dl e y( 2004)“ Mi ni ma l i s m,t heGe ne r a l i z a t i onPr obl e ma ndt heLi a r ” ,i n Synthese, Volume 139, pp. 491-512. Az z ou n i , J ody( 1998)“ On“ OnWha tThe r eI s ” ” ,i nPacific Philosophical Quarterly, volume 79, issue 1, pp. 1-18. Boghossian, Paul (1990a) “ TheSt a t usofCont e nt ” ,i nThe Philosophical Review, April, pp. 247-263. ——(1990b)“ TheSt a t usofCont e ntRe vi s i t e d” ,i nPacific Philosophical Quarterly 71, pp. 264-278. Brandom, R. (1994), Making it Explicit, Cambridge, Mass.: Harvard University Press. Br e nde l ,El ke( 2000) ,“ Ci r c ul a r i t ya ndt heDe ba t ebe t we e nDe f l a t i oni s ta ndSubs t a nt i ve The or i e sofTr ut h” ,i nCircularity, Definition and Truth, edited by André Chapuis and Anil Gupta, New Delhi: Indian Council of Philosophical Research, pp.29-47. Buchanan,R.( 2003)“ Ar eTr ut ha ndRe f e r e nc eQua s i -Di s quot a t i ona l ? ” ,i nPhilosophical Studies, volume 113, no. 1, pp. 43-75. Chr i s t e ns e n,Da vi d( 1993)“ Ske p t i c a lPr obl e ms ,Se ma nt i c a lSol ut i ons ” ,i nPhilosophy and Phenomenological Research, Volume 53, Issue 2, pp. 301-321. Col l i ns , J ohn( 1993)“ Hor wi c h’ sSc he ma t aMe e tSy nt a c t i cSt r uc t ur e s ” ,i nMind, Volume 112, Issue 447, pp. 399-431. David, Marian A. (1994), Correspondence and Disquotation, New York, Oxford University press. ——( 1996) ,“ Ana l y t i c i t y ,Ca r na p,Qui ne ,a ndTr ut h” ,i nPhilosophical Perspectives, 10, Metaphysics, pp. 281-296. ——( 2000) ,“ Mi ni ma l i s ma ndt heFa c t sa boutTr ut h” ,i nWhat is Truth, edited by Richard Schantz, Berlin, New York: de Gruyter, pp.161-175. ——( 2003)“ Tr ut h,El i mi na t i vi s m,a ndDi s quot a t i ona l i s m” ,i nNoûs, Volume 23, Issue 447, pp. 599-614. Da vi ds on,Dona l d( 1979) ,“ TheI n s c r ut a bi l i t yofRe f e r e nc e ”i nInquiries into Truth and Interpretation (1984), New York, Oxford University press, pp. 227-241. ——( 1990)“ TheSt r uc t ur ea ndCont e ntofTr ut h” , in Journal of Philosophy, Volume 87, Issue 6, pp. 279-328. 232 ——( 1996)“ TheFol l yofTr y i ngt oDe f i neTr ut h” ,i nJournal of Philosophy, Volume 93, Issue 6, pp. 263-278. De vi t t ,Mi c ha e l( 2001) ,“ TheMe t a phy s i c sofTr ut h” ,i nThe Nature of Truth, edited by Michael Lynch, Cambridge, MA.: MIT Press, pp. 579-611. Di l l a r d,Pe t e rS.( 1996)“ Ra di c a l Ant i -De f l a t i oni s m” ,i nPhilosophy and Phenomenological Research, Volume 56, Issue 1, pp. 173-181. Engel, Pascal (2002), Truth, Montreal: McGill-Que e n’ sUni ve r s i t yPr e s s . Et c he me ndy ,J ohn( 1988) ,“ Ta r s kionTr ut ha ndLog i c a lCons e que nc e ” ,The Journal of Symbolic Logic 53, pp. 51-79. Fi e l d,Ha r t r y( 1972) ,“ Ta r s ki ’ sTh e or yofTr ut h”( i nc l udi ng“ Pos t s c r i pt ”( 2001) ) ,i nTruth and the Absent of Fact (2001). Oxford: Clarendon Press, pp. 3-29. ——( 1973) ,“ The or yCha ngea n dt heI nde t e r mi na c yofRe f e r e nc e ”( i nc l udi ng “ Pos t s c r i pt ”( 2001) ) ,i nTruth and the Absent of Fact (2001). Oxford: Clarendon Press, pp. 177-198. ——( 1974) ,“ Qui nea ndt heCor r e s ponde nc eThe or y ”( i nc l udi ng“ Pos t s c r i pt ”( 2001) ) ,i n Truth and the Absent of Fact (2001). Oxford: Clarendon Press, pp. 199-221. ——( 1986) ,“ TheDe f l a t i ona r yConc e pt i onofTr ut h” ,i nFact, Science, and Morality: Es s ay sonA.J . Ay e r ’ s‘ Lang uage , Tr ut handLogi c ’ , edited by G. Macdonald and C. Wright. Oxford: Basil Blackwell, pp. 55-117. ——( 1994) ,“ De f l a t i oni s tVi e wso fMe a ni nga ndCont e nt ”( i nc l udi ng“ Pos t s c r i pt ” (2001)), in Truth and the Absent of Fact (2001). Oxford: Clarendon Press, pp. 104-156. ——( 1999) ,“ De f l a t i ngt heCons e r va t i ve ne s sAr gume nt ” ,The Journal of Philosophy, 96, (pp. 533-40). Fodor, J. (1994) The Elm and the Expert, Cambridge, Mass: MIT Press. Fodor ,J . ;Le por e ,E.( 2001)“ WhyCompos i t i ona l i t yWon’ tGoAwa y :Re f l e c t i onson Hor wi c h’ s‘ De f l a t i ona r y ’The or y ” ,i nRatio, volume 14, no. 4, pp. 350-368. Fog e l i n,Robe r tJ .( 1997)“ Qui ne ’ sLi mi t e dNa t ur a l i s m” ,i nThe Journal of Philosophy, volume 94, no. 11, pp. 543-563. For be s ,G.( 1986)“ Tr ut h,Cor r e s ponde nc e ,a ndRe dunda nc y ” ,i nG.Ma c Dona l da ndC. Wright eds., Fact, Science, and Morality, Oxford: Blackwell, pp. 671-687. Ge or g a l i s ,Ni c hol a s( 2000)“ Re f e r e nc eRe ma i nsI ns c r ut a bl e ” ,i nPacific Philosophical Quarterly, volume 81, issue 2, pp. 123-129. Gibson, Roger, Jr. (1988), Enlightened Empiricism, Florida: University Presses of Florida. Gr ove r ,Dor ot hy( 1990)“ Tr ut ha ndLa ng ua ge -Wor l dConne c t i ons ” ,i nJournal of Philosophy, Volume 87, Issue 12, pp. 671-687. 233 —— ( 2000) ,“ OnLoc a t i ngOurI n t e r e s ti nTr ut h” ,i nWhat is Truth, edited by Richard Schantz, Berlin, New York: de Gruyter, pp.120-132. Gupta, A. (1993a) ,“ ACr i t i queofDe f l a t i oni s m” ,i nPhilosophical Topics Vol. 21 no. 2, pp. 57-81. ——(1993b) ,“ Mi ni ma l i s m” ,i nPhilosophical Perspectives 7: Philosophy of Language, edited by J. E. Tomberlin. Atascadero, Calif.: Ridgeview Publishing Company. ——( 2002) ,“ AnAr g ume nt Ag a i ns tTa r s ki ’ sConve nt i onT” ,i nWhat is Truth, edited by Richard Schantz, Berlin, New York: de Gruyter, 225-237. Ha l ba c h, Vol ke r( 1999) ,“ Di s quot a t i ona l i s ma ndI nf i ni t eConj unc t i ons ” ,Mind Vol. 108 (429), pp. 1-22. ——(2000), “ Di s quot a t i ona l i s mFor t i f i e d” ,i nCircularity, Definition and Truth, edited by André Chapuis and Anil Gupta, New Delhi: Indian Council of Philosophical Research, pp.155-176. ——( 2001) ,“ HowI nnoc e ntI sDe f l a t i oni s m? ” ,Synthese 126, pp. 167-194. Heck, Richar dG. ,J r .( 1997) ,“ Ta r s ki ,Tr ut h,a ndSe ma nt i c s ” ,i nThe Philosophical Review, Volume 106, pp. 533-554. He r s h f i e l d,J . ;Sol e s ,D.H.( 2003) ,“ Re i nf l a t i ngTr ut ha sa nExpl a na t or yConc e pt ” ,i n Pacific Philosophical Quaraterly, Volume 84, pp. 32-42. Horwich, Pa ul( 1996)“ Re a l i s ma ndTr ut h” ,i nPhilosophical Perspectives, 10, Metaphysics, pp. 187-197. ——(1997a)“ Da vi ds ononDe f l a t i oni s m” ,i nU.Ze g l e n,e d. ,Discussions with Donald Davidson: On Truth, Meaning, and Knowledge, London: Routledge, pp. 20-24. ——(1997b)“ I mpl i c i tDe f i ni t i on, Ana l y t i cTr ut h,a ndApr i or iKnowl e dg e ” ,i nNoûs, Volume 31, Issue 4, pp. 423-440. ——(1998a) Meaning, Clarendon Press: Oxford. ——(1998b) Truth, second edition, Clarendon Press: Oxford. ——(2001a)“ ADe f e ns eofMi n i ma l i s m” ,i nSynthese, 126, pp. 149-165. ——(2001b)“ Deflating Compositionality” ,i nRatio, volume 14, no. 4, pp. 369-385. Hy l t on,Pe t e r( 2004) ,“ Qui neonRe f e r e nc ea ndOnt ol ogy ” ,i nThe Cambridge Companion of Quine, ed. By Roger Gibson, Jr., pp. 115-150. Kovach, Adam (1997) “ De f l a t i oni s ma ndt heDe r i va t i onGa me ” ,i nMind, Volume 106, Issue 423, pp. 575-579. Kr a ut ,Pa ul( 1993)“ Robus tDe f l a t i oni s m” ,i nThe Philosophical Reviews, Volume 102, Issue 2, pp. 247-263. Künne , Wol f ga ng( 2000) ,“ Di s quot a t i ona l i s tConc e pt i onsofTr ut h” ,in What is Truth, edited by Richard Schantz, Berlin, New York: de Gruyter, pp. 176-193. 234 Le e ds , St e phe n( 1973) ,“ Howt oThi nka boutRe f e r e nc e ”Journal of Philosophy 15: pp. 485-503. Mc Ge e , Va nn( 1993)“ ASe ma nt i cConc e pt i onofTr ut h” ,i nPhilosophical Topics, vol. 21, No. 2, spring, pp. 83-111. McGinn, Colin (2000), “ TheTr ut ha boutTr ut h” ,i nWhat is Truth, edited by Richard Schantz, Berlin, New York: de Gruyter, pp. 194-204. McGrath, Matthew (1997a) ,“ Re pl yt oKova c h” , in Mind, Vol. 106, issue 423, pp. 581-586. ——(1997b) ,“ We a kDe f l a t i oni s m” , in Mind, Vol. 106, issue 421, pp. 69-98. ——(2000), Between Deflationism & Correspondence Theory, New York: Garland Publishing. Mi l l e r , Al e xa nde r( 2000) ,“ Hor wi c h,Me a ni nga ndKr i pke ’ sWi t t g e ns t e i n” , in Philosophical Quarterly, Vol. 50, issue 199, pp. 161-174. ——( 2001) ,“ OnWr i g ht ’ sAr g ume nta ga i ns tDe f l a t i oni s m” , in Philosophical Quarterly, Vol. 51, issue 205, pp. 527-531. Ne l s on,R.J .( 1997)“ Pr oxyFunc t i ons , Tr ut ha ndRe f e r e nc e ” ,i nSynthese, volume 111, issue 1, pp. 73-95. O’ Le a r y -Ha wt hor ne ,J ohn;Oppy ,Gr a ha m( 1997)“ Mi ni ma l i s ma ndTr ut h” ,i nNoûs, Volume 31, Issue 2, pp. 170-196. Qui ne , W. V.( 1951)“ Ont ol ogya ndI de ol ogy ” ,i nPhilosophy Studies 2, pp. 11-15. ——(1953a) ,“ Not e sont heThe or yofRe f e r e nc e ” ,i nFrom a Logical Point of View, Cambridge, Mass.: Harvard university Press, pp. 130-138. ——(1953b) ,“ TwoDogma sofEmpi r i c i s m” ,i nFrom a Logical Point of View, Cambridge, Mass.: Harvard university Press, pp. 20-46. ——( 1954) ,“ Ca r na pa ndLog i c a lTr ut h” ,i nThe Ways of Paradox and Other Essays, revised ed. Cambridge, Mass.: Harvard University Press, pp. 107-132. ——(1960), Word and Object, Cambridge, Mass.: M.I.T. Press. ——(1969a) ,“ Epi s t e mol ogyNa t ur a l i z e d” ,i nOntological Relativity and Other Essays, New York: Columbia University Press, pp. 69-90. ——(1969b) ,“ Ont ol og i c a lRe l a t i vi t y ” ,i nOntological Relativity and Other Essays, New York: Columbia University Press, pp. 26-68. ——(1970), Philosophy of Logic, Eaglewood Cliffs, N.J.: Prentice-Hall. ——(1974a) ,“ OnPoppe r ’ sNe ga t i veMe t hodol ogy ” ,i nThe Philosophy of Karl Popper, ed. by P. A. Schlipp, La Salle, Ill: Open Court. ——(1974b), Roots of Reference, La Salle, Ill: Open Court. ——( 1975) ,“ OnEmpi r i c a l l yEqui va l e ntSy s t e msoft heWor l d” ,Erkenntnis 9, pp. 235 313-328. ——(1976) ,“ Tr ut ha ndDi s quot a t i on”i nThe Ways of Paradox and Other Essays, revised ed. Cambridge, Mass.: Harvard University Press, pp. 308-321. ——( 1977) ,“ Fa c t soft heMa t t e r ” ,i nAmerican Philosophy from Edwards to Quine, ed. by Shahan and Merrill, Norman, Okla.: University of Oklahoma Press. —— (1981a) ,“ Re pl yt oEl e ve nEs s a y s ”i nPhilosophical Topics 12, pp. 227-243. ——(1981b) ,“ Empi r i c a lCont e nt ”i nTheories and Things, Cambridge, Mass.: Harvard university Press, pp. 24-30. ——(1981c) ,“ Re s pons e s ”i nTheories and Things, Cambridge, Mass.: Harvard university Press, pp. 173-186. ——(1986a) ,“ Re pl yt oPa ul A.Rot h”i nThe Philosophy of W. V. Quine, edited by L. Hahn and P. Schilpp, La Salle, Ill: Open Court, pp. 459-461. ——(1986b) ,“ Re pl yt oRog e rF .Gi bs on,J r . ”i nThe Philosophy of W. V. Quine, edited by L. Hahn and P. Schilpp, La Salle, Ill: Open Court, pp. 155-157. ——(1990a)“ Comme ntonLa ue n e r ”i nPerspectives on Quine, ed. by Robert B. Barrett and Roger F. Gibson, Cambridge, Mass.: Basil Blackwell, p.229. ——(1990b) Pursuit of Truth, Cambridge, Mass.: Harvard University Press. ——(1995) From Stimulus to Science, Cambridge, Mass.: Harvard University Press. Ra ms e y ,F .P .( 1927) ,“ Fa c t sa ndPr opos i t i ons ” ,Proceedings of the Aristotelian Society, suppl. Vol. 7, pp. 153-171. Robbi ns ,Phi l i p( 2001)“ Wha tCompos i t i ona l i t ySt i l lCa nDo” ,i nPhilosophical Quarterly, volume 51, issue 204, pp. 328-336. Romanos, George D. (1983), Quine and Analytic Philosophy, Cambridge, Mass.: M.I.T. Press. Rot h,Pa ul A.( 1986) ,“ Se ma nt i c swi t houtFounda t i ons ”i nThe Philosophy of W. V. Quine, edited by L. Hahn and P. Schilpp, La Salle, Ill: Open Court, pp. 433-458. Sc ha nt z ,Ri c ha r d( 2000) ,“ HowDe f l a t i oni s t sDi ve r g ef r omTa r s ki ” ,i nCircularity, Definition and Truth, edited by André Chapuis and Anil Gupta, New Delhi: Indian Council of Philosophical Research, pp.317-331. Sc hi f f e r ,St e phe n( 2000)“ Hor wi c honMe a ni ng ” ,i nPhilosophical Quarterly, volume 50, issue 201, pp. 527-536. Sha pi r o,S.( 1998) ,“ Pr oofa ndTr ut h:Thr oug hThi c ka ndThi n” , Journal of Philosophy 95, pp. 493-521. ——( 2003)“ TheGur u,t heLog i c i a na ndt heDe f l a t i oni s t :Tr ut ha ndLog i c a l Cons e que nc e ” ,i nNoûs, volume 37, no. 1, pp. 113-132. She r ,Gi l a( 1998)“ Ont hePos s i bi l i t yofaSubs t a nt i veThe or yofTr ut h” ,i nSynthese, 236 volume 117, issue1, pp. 133-172. Smith, Michael (1994a) ,“ Mi ni ma l i s m, Tr ut h-a pt i t udea ndBe l i e f ” ,i nAnalysis, 54.1, pp. 21-26. ——(1994b) ,“ WhyExpr e s s i vi s t sa boutVa l ues houl dLoveMi ni ma l i s ma boutTr ut h” ,i n Analysis, 54.1, pp. 1-12. Soames, Scott (1997), “ TheTr ut ha boutDe f l a t i oni s m” ,i nPhilosophical Issues 8, edited by E. Villanueva, Atascadero, Cal.: Ridgeview Publishing Company. ——(1999), Understanding truth, Oxford, New York: Oxford University Press. Sosa, Ernest (1993a)“ Epi s t e mol ogy ,Re a l i s m,a ndTr ut h” ,i nPhilosophical Perspectives, Volume 7, Issue Language and Logic, pp. 1-16. ——(1993b) ,“ TheTr ut hofMod e s tRe a l i s m” ,i nPhilosophical Issues 3, Science and Knowledge, pp. 77-95. Tarski, Alfred (1956a) ,“ TheEs t a bl i s hme ntofSc i e nt i f i cSe ma nt i c s ”i n Tarski, Logic, Semantics, and Metamathmatics, tr. J. H. Woodger, London: Oxford university Press. ——(1956b) ,“ TheConc e ptofTr ut hi nFor ma l i s e dLa ng ua g e s ”i nTa r s ki ,Logic, Semantics, and Metamathmatics, tr. J. H. Woodger, London: Oxford university Press. Va nCl e ve ,J a me s( 1996)“ Mi ni ma lTr ut hI sRe a l i s tTr ut h” ,i nPhilosophy and Phenomenological Research, Volume 56, Issue 4, pp. 869-875. Van Inwagen, Peter (2000), “ Ge ne r a l i z a t i onsofHomophoni cTr ut h-Se nt e nc e s ” ,i nWhat is Truth, edited by Richard Schantz, Berlin, New York: de Gruyter, pp. 205-222. Wi l l i a ms ,Mi c ha e l( 1999) ,“ Me a ni nga ndDe f l a t i ona r yTr ut h” ,i nThe Journal of Philosophy, 96, pp.545-564. ——( 2002) ,“ OnSomeCr i t i c sofDe f l a t i oni s m” ,i nWhat is Truth, edited by Richard Schantz, Berlin, New York: de Gruyter, pp.146-158. 237