UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Quine on opacity in modal and doxastic contexts Dickson, Mark William 1995

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

Item Metadata

Download

Media
831-ubc_1995-059464.pdf [ 14.04MB ]
Metadata
JSON: 831-1.0088185.json
JSON-LD: 831-1.0088185-ld.json
RDF/XML (Pretty): 831-1.0088185-rdf.xml
RDF/JSON: 831-1.0088185-rdf.json
Turtle: 831-1.0088185-turtle.txt
N-Triples: 831-1.0088185-rdf-ntriples.txt
Original Record: 831-1.0088185-source.json
Full Text
831-1.0088185-fulltext.txt
Citation
831-1.0088185.ris

Full Text

QUINE ON OPACITY IN MODAL AND DOXASTIC CONTEXTS by MARK WILLIAM DICKSON BA (Hons.)j The University of British Columbia, 1986 MA, The University of British Columbia, 1989 A THESIS SUBMITTED IN PARTIAL FULFILMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY in THE FACULTY OF GRADUATE STUDIES (Department of Philosophy) We accept this tkesis as conforming ftp the recmired standard THE UNIVERSITY OF BRITISH COLUMBIA September 1995 © Mark William Dickson, 1995 In presenting this thesis in partial fulfilment of the requirements for an advanced degree at the University of British Columbia, I agree that the Library shall make it freely available for reference and study. I further agree that permission for extensive copying of this thesis for scholarly purposes may be granted by the head of my department or by his or her representatives. It is understood that copying or publication of this thesis for financial gain shall not be allowed without my written permission. (Signature) Department of £ K l lc?S 0 f a <-f The University of British Columbia Vancouver, Canada Date l^ lVW %^£ \Ofi{ DE-6 (2/88) Abstract Quine has been mainly opposed to sentences that feature cross-quantification. That is, he is critical of sentences that involve quantifying into a context that Quine labels 'opaque1. Quine's opposition to cross-quantification grew out of an earlier attack on the notion of combining quantification theory and modal logic. Quine initially dismissed, in 1943, cases of quantifying into modal contexts as meaningless. Later in the same year, Alonzo Church argued that there was a meaningful way to quantify into modal contexts, thus vindicating the notion that quantification theory could be merged with modal logic. In 1956, in "Quantifiers and Propositional Attitudes," Quine pointed out that quantification into belief contexts, though indispensable, also features the improper quantification into opaque contexts. In the same paper, Quine introduced the distinction between a relational and a notional sense of propositional attitude ascriptions. The former sense concerns the problematic sentences that feature cross-quantification. In the thesis that follows, I appropriate Quine's terminology and critically evaluate his reasons for rejecting the relational idiom in both modal and doxastic contexts. Such an evaluation reveals some startling results in the philosophy of language. One of the major problems that Quine sought to address was that of reconciling the evident significance of instances of the relational idiom with their many alleged difficulties. Quine restricted himself to acknowledging the idiom's meaningfulness ii in doxastic contexts. Most of Quine's criticisms of the relational idiom are argued by me to be unsound. It is contended that some of Quine's criticisms involve the improper exploitation of ambiguities inherent in such sentences. This fallacy is exposed and subjected to a critical evaluation. The exposure of this fallacy, which I term 'the relational fallacy' is a novel contribution to the philosophy of language. Another novel contribution to the philosophy of language is my critique of Quine's use of semantic ascent to account for intuitively meaningful relational modal sentences. A third, slightly less novel, contribution to the philosophy of language involves extending Quine's temporary view that there are meaningful relational sentences in doxastic contents to the analogous observation that there are meaningful relational sentences in modal contexts. iii TABLE OF CONTENTS Abstract ii Table of Contents iv Acknowledgement ix Chapter I. An Introduction to Quine's Criticisms of Cross-Quantification 1. Opening Remarks 1 2. Comments on Terminology and Stylistic Conventions 11 3. Quine's Objections to QML and Criticisms of the Relational Idiom 15 4. General Observations on Quinean Criticisms of the Relational Idiom 28 Chapter II. Quine's Initial Criticisms of QML and his Positive Account of Modality 5. General Remarks 36 6. Quine on Substitutivity of Identicals, Existential Generalization, and Unique Specification: Initial Stages of Quine's Critique of the Relational Idiom 37 6.1 Substitutivity of Identicals 37 6.2 Existential Generalization 45 6.3 Unique Specification 53 iv 7. Semantic Ascent 57 8. Quine•s Position in "TGMI" 66 Chapter III. Quine on Employment of the Relational Idiom as a Response to his Three Initial Objections to QML 9. General Remarks: The Two Versions of the Relational Idiom 72 10. Quine on the Solution of Church and Carnap: Three Arguments 73 10.1 Quine's First Argument [1943] 76 10.2 Quine's Second Argument [1953] 78 10.3 Quine's Third Argument [1961] 79 11. Quine on Smullyan's Use of Russell's Theory of Definite Descriptions In Conjunction with Modal Logic 81 12. The Relational Fallacy 92 13. Kaplan on Quine's Alleged Theorem 97 14. Quine's Inconsistent Skepticism: Structural Parallels Between Modal and Doxastic Phenomena 102 Chapter IV. Quine on Essentialism 15. General Remarks 115 16. Quine's Two Arguments Against Essentialism [1953-1960] 118 v 16.1 Quine's Argument Against Essentialism in FLPV and "TGMI" 118 16.2 Quine's Second Argument Against Essentialism: WO 120 17. Quine on the Reversion to Invidious Aristotelian Essentialism 127 18. F^llesdal on Quine's 'Conflation' of Two Notions of Essentialism 131 Chapter V. Stages 3 and 4 of Quine's Criticisms of Cross-Quantification 19. General Remarks 20. Quine on the Role of Attributes in "Q&PA" 21. Second Quinean Volte-Face 21.1 Quine!s Charge Concerning the "intolerable oddity" of the Relational Idiom in WO 21.2 Quine's Corresponding Charge Regarding the Collapse of Modal Distinctions 22. Quine's Reversal of Position in Pursuit of Truth Chapter VI. Attributes as Applied to Modal Phenomena and Quine on their Identity and Individuation 23. General Remarks 166 24. Attributes [Properties] Applied to vi 136 137 141 144 156 161 Modal Phenomena 171 25. No Entity Without Identity 175 26. The Individuation of Attributes 178 Chapter VII. Quine's Contribution to the Problem of Transworld Identity 27. General Remarks 190 28. Quine's Comments in "WA" 197 29. A Corresponding Position Concerning Belief Worlds 204 Chapter VIII. Quine on "IR" 30. General Remarks 209 31. A Critique of "Intensions Revisited" 211 32. Vulgar Occurrence of Terms in the Relational Idiom 222 Chapter IX. A Positive Account of the Relational Idiom 33. General Remarks 231 34. On the Semantic Significance of the de re/de dicto Distinction 232 35. Further Considerations on Modal Attributes 243 Notes 249 vii Glossary 278 Bibliography 287 Appendix A-F011esdal's Defence of Quine on Modality 290 viii Acknowledgement To my mother, Katherine Dickson whose contribution to this thesis was not purely genetic. Special thanks to Garola Ellis, for her substantial devotion of time and energy towards preparation of several draft copies. ix Quine on Opacity in Modal and Doxastic Contexts Chapter I. An Introduction to Quine's Criticisms of Cross-Quantification § 1. Opening Remarks W. V. Quine has made at least seventeen charges against combining quantification theory with modal and propositional attitude operators. Such operators include modal ones ('necessarily' and 'possibly') as well as ones associated with propositional attitude ascriptions (for example, doxastic operators). These objections arguably began in 1937. These problems overlap, to a certain extent, Quine's objections to the notions of meaning, synonymity (synonymy), and analyticity. He in fact treated all these notions as being on a philosophical par. Thus he writes that the "notion of synomymity in turn is as important and as needing of analysis as that not only of analytic, but of meaning; for the three are interdefinable." Of these seventeen charges, sixteen stand out as criticisms of what has been termed the relational idiom. What this idiom amounts to will shortly be explained. Some of Quine's charges have been aimed at accounts of propositional attitude ascription, while others have been aimed at Quantified Modal Logic (hereafter, 1 2 QML), that branch of logic that employs the familiar quantifiers 'all' and 'some' in conjunction with the modal operators 'possibly' and 'necessarily'. The history of QML has a curious span: while quantifier logic was begun by Frege in the nineteenth century and later popularized by Russell, modal logic was initially developed by Aristotle in the fourth century B.C. Modal logic was treated as a branch of propositional logic, a system of logic, one of whose differences from quantifier logic resides in the inclusion of variables: while quantifier logic features variables which may be bound by quantifiers, propositional logic does not. It was not until the twentieth century that attempts were made to combine quantifier logic with modal logic. The primary goal of this thesis is to provide a systematic explication both of Quine's various criticisms of the relational idiom as well as the most tenable responses to these criticisms. Such a systematic account of Quine's criticisms and the responses to them has, to my knowledge, never been given. It is true that some of Quine's criticisms and some of the solutions to these objections have been gathered since the early sixties. However, these collections have either been incomplete or concentrated on problems to some extent independent of problems surrounding the relational idiom. I have in mind here especially certain papers by Dagfinn F011esdal and David Kaplan. Both F0llesdal and Kaplan give partial treatments of Quine's criticisms of the relational idiom. 2 3 Throughout this thesis, I argue that there is a close connection between modal and doxastic phenomena. This point has been recognized by different philosophers, including, at various times, Quine himself. There are, it is true, limits to this connection between modal and doxastic phenomena. The exact nature of this connection is investigated at length. I feel that, as a result of my investigation of this connection, it will become clear that Quine is inconsistently sceptical (to paraphrase David Kaplan) in his opposition towards the relational idiom in modal, but not doxastic, contexts. For a number of years, it has seemed to me that such a systematic treatment of Quine's criticisms of the relational idiom is presupposed by some of the writings of Saul Kripke. If one reads some of Quine's papers that are critical of quantification into modal or doxastic contexts and then reads some of Kripke's early papers that score several points with the very use of such quantification, one gets the impression that Kripke must have some sort of refutation of Quine's criticisms, which are mainly directed at the relational idiom in both modal and doxastic contexts, in mind. (For the definition of opaque context, see § 4.1 and the definitions of instances of the relational idiom in modal and doxastic contexts are discussed later in this section.) Yet Kripke has given relatively few explicit arguments against Quine's criticisms of the relational idiom. In a word, Kripke's critique of Quine's criticisms is an enthymeme, and an elaborate one at that. Hopefully, in the 4 process of reaching the above-mentioned primary goal of this thesis, the suppressed premises in Kripke's critique will be made explicit. These suppressed premises constitute a largely negative reaction that is provoked by Quine's critique. Treatments of Quine on modality have usually focused on his account of analyticity, rather than on his account of the relational idiom. In this thesis, Quine's account of analyticity will by and large be avoided. Questions such as "Are there synthetic a priori truths?" are exceedingly difficult to leave unresolved. Nevertheless, this question is strictly separate from Quine's particular rejection of the relational idiom. In general the topic of analyticity as Quine addressed it in "Two Dogmas of Empiricism" will be avoided. A secondary goal of this thesis is to spell out the solutions for solving an expanded version of Quine's problem of accounting for meaningful instances of the relational idiom. Quine's initial problem was to account for the meaningfulness of instances of the relational idiom in doxastic contexts. I expand Quine's problem to include meaningful instances of the relational idiom in modal contexts. These solutions are developed in various parts of the thesis, but are explicitly put forth in Chapter IX, § 35. I urge that these solutions must reflect a number of contentions. Foremost among these contentions is that modal and doxastic phenomena, in light of their close similarities, ought to be accorded a similar solution to an expanded version of Quine's problem. This is why 5 Quine's initial problem, which covers only doxastic phenomena, should, I think, be expanded to cover modal ones as well. Another such contention is that the distinction between the relational and the notional idiom is not merely one that occurs in belief utterances but instead has semantic significance. Quine himself gave, over the course of his career, various glimpses of a positive account which stand in marked contrast to his usually negative outlook on the relational idiom in both contexts. I attempt to gather together these glimpses of a positive treatment of the relational idiom. The solution to Quine's problem of having to account for meaningful instances of the relational idiom that I intend to argue tentatively for was not, however, suggested by Quine. Instead, the account that I favor comes from Kripke. Some version of the doctrine of essentialism must, I argue, be accepted if quantification into modal contexts is held to be meaningful. I entirely agree with Quine's view that "the way to do quantified modal logic, if at all, is to accept Aristotelian essentialism." After arguing that instances of the relational idiom in modal contexts are meaningful, I intend to consider two ways to account for this significance. The first solution that I consider is a Quinean solution to an extended version of Quine's problem of accounting for cases of indispensable sentences. This solution largely resembles Quine's initial solution to his problem of accounting for indispensable sentences that feature 6 quantification into doxastic contexts; where Quine invoked attributes, I tentatively invoke modal attributes. It should be stressed that this particular solution is not at all supported by the author. The second solution centres around Saul Kripke's use of possible worlds. In general, the idea is to view the relational senses of necessity and possibility as fixing an object in the actual world and then ascribing certain modal properties to him. This second solution is endorsed by the author. As I urge, there are a few versions from which to choose. It is not my intention to enter into a review of all the objections to essentialism that have been given by a variety of philosophers. Rather, I concentrate exclusively upon Quine's critique of essentialism. Quine's critique is discussed in § 16. "One of the chief attractions of the Leibnizian metaphysic of possible worlds is that -as Kripke, Hintikka, and others demonstrated in the early 1960s- it enables us to give a semantical underpinning to much of the machinery of formal 0 logic." I take it that the comment regarding much of the machinery of formal logic refers to the structure of QML, since this was the principal concern of Kripke and Hintikka in the early 1960s. Possible worlds can assist philosophers in accounting for the truth and consistency of modal sentences and the validity and soundness of arguments that feature modal sentences. As I show in Chapter VII, Quine rejects possible worlds because of a variant of the so-called problem of 7 transworld identity. I disagree with Quine in this matter and the reasons are given in Chapter VII. In "Quantifiers and Propositional Attitudes" (hereafter, "Q&PA") [1956], Quine felt forced to admit that certain sentences that contain quantification over propositional attitude operators are meaningful. He first coined the term 'relational' in connection with ambiguous sentences, such as in Quine's own example in "Q&PA." The term, as used by Quine, describes a sense of an idiom. The relational idiom is contained in 1 of 2 possible interpretations of Quine's famous ambiguous example: (1) Ralph believes someone is a spy. On the one hand, Ralph might simply share the ordinary belief in spies. That is, Ralph might merely be in the position of an ordinary person who believes that there are worldwide organizations of spies. This thoroughly general belief is captured in the following sentence: (1*) Ralph believes that there are spies. (1*) can be restated as (1**) Ralph believes that (3x) (x is a spy). (1*) is an instance of the notional idiom. However, (1) might be interpreted in the following relational fashion: (1*) There is someone (in particular) whom Ralph 8 believes to be a spy. (1*) can be restated as (1'') (3x) (Ralph believes x to be a spy). (1**) would be true if the following sentence were true: (1**') Ralph believes of the man in the brown hat that he is a spy. The notional sense of belief, seen in (1*) and (1**), is a binary relation between a believer and a sentence or proposition. The contents of the believers' notions are paramount. Hence, the term 'notional'. The relational sense of belief, seen in (1*) and (1'*), involves a ternary relation between a believer, an object of belief, and an attribute. Unlike the notional sense of belief, where the contents of the believers' notions are of paramount importance, the paramount importance for the relational sense of belief is played by the terms of the relation in question. Hence, the term 'relational'. The following modal sentence exhibits a similar ambiguity: (2) The author of Hamlet might not have written Macbeth. On the one hand, (2) might be interpreted (taking the 'might' as the modal operator 'it is possible that' for the notional idiom and the modal operator 'possibly' for the relational idiom) as (2*) It is possible that: the author of Hamlet did not 9 write Macbeth. (2*) is an instance of the notional idiom and can be restated as (2**) 0 (3x) (x is the author of Hamlet and x did not write Macbeth). However, (2) might be interpreted in the relational sense as (2') There is exactly one author of Hamlet, and that same individual possibly did not write Macbeth. (2*) can be restated, using Russell's Theory of Definite Descriptions (hereafter, "TDD"), as (2'*) (3x) (x is an author of Hamlet and no one else is and 0 (x did not write Macbeth)). Though Quine accepted (1) as meaningful and attempted to reconcile this meaningfulness with problems that centred on the alleged logical difficulties of sentences such as (1), he was not ultimately satisfied with the results of his solution. These logical difficulties are discussed at length in Chapters II. and III. For this reason, Quine remains sceptical of the relational idiom. Quine's attack upon this idiom has been extremely lengthy and began in his published writings with his "Whitehead and the Q Rise of Modern Logic." As is urged later in this thesis, it is the relational/notional distinction in the context of propositional attitude ascriptions about which Quine has been thoroughly equivocal, at times championing it, attacking it at 10 others. He has been far less ambivalent in his opposition to the relational idiom in modal contexts. The criticisms that Quine has made of QML are often viewed as being criticisms of the relational idiom in general. In large measure, such an interpretation is given implicit support by Quine's papers, which only specifically mention the relational idiom at certain points. Yet, it is arguable that many of Quine's criticisms directed at QML have little direct bearing on the status of the relational idiom. Perhaps Quine's most famous objection to the modalities is contained in his "Two Dogmas of Empiricism" (hereafter, "TDE"), in which he argues that the analytic/synthetic distinction is, in general, without foundation. Many philosophers, Quine included, have seen the adverb 'necessarily' as announcing that the prefixed sentence that follows the adverb will be an analytic one. Similarly, many philosophers conceive of the adverb 'possibly1 as announcing that the prefixed sentence that follows the adverb will be a synthetic one. So Quine's famous attack on the intelligibility of the analytic/synthetic distinction has considerable bearing, or so one would think, on his objections to QML. A language containing the adverb 'necessarily1 "is intelligible only in so far as the notion of analyticity is already understood in advance." ° One of the objectives of this thesis is to show that Quine' s position on the relational idiom can be shown to be untenable even though one may remain largely agnostic with respect to the views 11 outlined in "TDE." It would be a mistake to hold that Quine's criticisms of the relational idiom are a mere subset of his more famous criticisms levelled at the analytic/synthetic distinction. There is a great degree, however, of similarity between the two groups of criticisms. Many, though not all, of Quine's criticisms of the relational idiom have at least three similarities with his rejection of the analytic/synthetic distinction: (a) Quine has sought to show that the sentences of QML can be treated as instances of either analytic or synthetic sentences (and thus are open to the sort of arguments he made in "TDE"); (b) both sets of criticisms show Quine's explicit emphasis upon substitutional criteria as a paramount sentential feature; and (c) both sets of criticisms show Quine's scepticism towards intensional attributes with regards to the failure of the specification of identity conditions. But despite these similarities, in one sense, at least, the merits of Quine's argument against the analytic/synthetic distinction are irrelevant when it comes to evaluating his criticisms of the relational idiom. The majority of his criticisms of the relational idiom are in fact independent of his argument against the analytic/synthetic distinction. § 2. Comments on Terminology and Stylistic Conventions In "Q&PA," Quine wrote of "the relational sense" of belief 12 (and other propositional attitudes) as opposed to "the notional sense." In Word and Object (hereafter, WO) , Quine used the term 'transparent sense of belief, a term that appears to be interchangeable with 'relational sense of belief. In this thesis, Quine's terminology will be appropriated. Accordingly, I will speak of 'instances of the relational/notional idiom1. Senses and idioms are semantic notions, and they are not to be confused with syntactic notions. Among this latter category, I will include ' de re sentences', 'transparent sentences1, ' de dicto sentences', and 'opaque sentences1. The de re/de dicto distinction plays an integral role in this thesis. The terms formally originated with the Scholastic philosophers. Literally, de re means 'of things1 and de dicto means 'of words1. Here is a close paraphrase of Graeme Forbes' initial definition: A sentence S with a modal operator 0 is a modal de re sentence iff either (i) 0 prefixes variables that occur free within S; or (ii) 0 prefixes an individual constant within S; or (iii) 0 is both followed by variables that occur bound and is preceded by a quantifier which binds the variables in question. A sentence S with a modal operator 0 is a modal de dicto sentence iff S is not a modal de re sentence. A variable that occurs free within S is merely one that is not bound by a quantifier. An individual constant is the device employed by the logician to represent a name. Both variables that occur free and individual constants are indicative, on this syntactic definition, of de re modal sentences. One might think that it would be unproblematic to apply this syntactic definition of modal sentences to doxastic cases. 13 I attempt to modify Forbes' technical definition of a de re modal sentence to cover doxastic de re sentences in Chapter III, § 6.2. The principal difficulty is that an individual constant within a doxastic sentence does not allow one to conclude that the sentence in question is a de re one. This issue is more completely explained in § 11. It should be remarked that the topic of opacity for Quine covers concerns that are both syntactic and semantic in nature. Quine sometimes appears to shift from syntactic matters to semantic ones without alerting the reader. One example of this involves Quine's analyses of what I call the Church/Carnap solution and the solution of Smullyan (see § 10-11 in Chapter III). The former analyzed solution involves semantic concerns, while the latter analyzed solution involves syntactic concerns. I follow Quine's lead in merging both semantic and syntactic matters. When I am speaking of individual sentences, I will endeavour to use the term ' de re' or the term 'de dicto' . When I am speaking of the meaning of a given sentence, the sentence will be said to express a relational or a notional sense of either belief or necessity. In display-numbers, most de re sentences are followed by a prime symbol ('). Similarly, most de dicto sentences are followed by an asterisk (*). For example, whereas (1') indicates the presence of a de re sentence, (1*) indicates the presence of a de dicto sentence. However, certain sentences elude exact classification using the de re/de dicto distinction. Some 14 sentences instead combine features of both de re and de dicto sentences. A good example of such inexact classification of sentences will arise when we consider alternative analyses of de re modal sentences using Russell's TDD. When this method is applied to de re sentences, the result will be an analysis that combines elements of both types of sentences. (I hasten to add that such de dicto analyses also indicate, because of their failure to provide proper analyses, that the sentence being analyzed is irreducibly de re. See pp. 59-62.) Accordingly, such sentences are preceded by display-numbers that contain both prime and asterisk symbols. Because Quine often speaks of entities to which he is only committed in order to explain the apparent meaningfulness of certain sentences, and because he has set forth strict conditions to be met by those entities to which he is committed, it is important to distinguish between heuristic entities and ones which are legitimate. The former include attributes and propositions and other entities that Quine considers to be intensionalistic. The latter include entities such as physical objects and classes and other entities that Quine considers to be extensionalistic. In the main, Quine*s various objections to modal logic as well as doxastic phenomena will be handled as follows: expositions of the various arguments will be followed by their respective critiques. (There will be no formal announcement, however, indicating that the exposition will be followed by the 15 critique.) Exceptions, however, are made in the cases where I am in substantial agreement with Quine. In those cases, I simply present an evaluation of the relevant points. For example, there is the case of Quine's rejection of the solution devised by Church and Carnap as a response to Quine's initial concerns about quantified modal logic. With some, not very crucial, reservations, I am substantially in accord with Quine's view of this matter. There are several points in this thesis where reference will be made to earlier sentences without the repetition of them. Rather, only display-numbers typically are referred to. To prevent any confusion that may result, a complete list of display-numbers, along with their respective sentences, is repeated in Part B of the glossary (p. 279). § 3. Quine's Objections to QML and Criticisms of the Relational Idiom This section will briefly list, and distinguish, Quine's various criticisms of the relational idiom. This section will also indicate where I intend to consider these criticisms more fully. Two of the most well-known objections that Quine has advanced against QML are contained within Quine's "Reference and I 0 Modality" (hereafter, "R&M"). However, this paper makes a total of five objections. (Of these objections, h stem from papers that Quine had written prior to 1953, the year in which 16 "R&M" first appeared.) The first is the familiar substitutivity of identicals. The following pair of sentences is true: (3) 9 is necessarily greater than 7, (4) The number of planets = 9. Yet the sentence (5) The number of planets is necessarily greater than 7, is, one might think, false. Yet (5) appears to have been legitimately obtained from (3) and (4) by invoking Leibniz's Law. Thus, the argument from (3) and (4) to (5) implies, for Quine, either a restriction on Leibniz's Law or on modal sentences. Since Quine cannot countenance a violation of Leibniz's Law, he feels that he has produced an effective 1 ? criticism of QML. By the time that concerns were raised over the phenomena of the substitutivity of identicals in From a Logical Point of View (hereafter, FLPV) [1953], there had already been suggestions that the problem does not arise if sentences such as (5) are granted a relational, as opposed to a notional, interpretation. These suggestions are covered extensively in Chapter II. Whereas a de dicto reading of (5) is false, a de re reading of (5) is true; Quine's criticism, it has been argued, evaporates if a relational interpretation of modal sentences such as (5) were granted. The following is a de dicto symbolic interpretation of (5): 17 ( 5*) • (3x) [ NPx A (Vy) ( WPy - x = y ) A ( x > 7 ) ] . The following is a de re symbolic interpretation of (5): (5 ' ) (3x) [ ( NPx A (Vy) ( NPy - x = y)) A D ( x > 7 ) ] . A similar treatment of (6) The number of planets is possibly less than 7, can be performed. All of the remaining criticisms of QML which Quine has submitted are, without exception, also criticisms of the relational idiom. It could be held that rather than overlooking (5*) and other instances of the relational idiom in modal contexts, Quine is more concerned with discrediting it. I will show that the arguments Quine uses to discredit de re sentences such as (5') are either invalid or unsound. The second of Quine's objections to QML contained in "R&M" is the one based on the phenomena of existential generalization. If it is true that (3) 9 is necessarily greater than 7, then it is also true that (3') {3x) x is necessarily greater than 7. Quine now asks what is this thing that is necessarily greater than 7. If this thing that is necessarily greater than 7 is the number of planets, then the result, according, to Quine, 18 conflicts with the supposed fact that (5) is false. This second objection to QML constitutes Quine's first criticism of the relational idiom. It is discussed in Chapters II and III. The third objection to QML contained in "R&M" concerns the unique specification of modal conditions. Quine holds that being necessarily greater than 7 is not a trait of a number and, hence, "9 is necessarily greater than 7" is meaningless. For "x > 7" is a necessary consequence of some, but not all, of the conditions that uniquely specify x. Whether a consequence is necessary or not depends upon its formulation. This objection to QML, much like his second, is thus an attack on the supposed meaningfulness of de re modal formulae. The argument here is that necessity (and by implication, possibility) is applicable to the connection between certain consequences, like 'x > 7' and particular ways of specifying x. Quine argues that 'x > 7' is a necessary consequence of only the first formula, even though both formulae uniquely determine the same number x: (7) x = /x + /x + /x * V"x (8) there are exactly x planets. My argument against this objection is much the same as my earlier responses: that Quine is exploiting the ambiguity present in many modal sentences. This is not to say that I consider the second and third objections to be identical; they simply rely on the same mistake, namely, one that I term 'the relational fallacy'. This fallacy involves the improper 19 exploitation of an ambiguous sentence in order to obtain a reduction to absurdity. (This fallacy plays a significant role in disputing some of Quine's charges, and it is developed at length in Chapter III.) That is, Quine's construal of the necessary consequence is intelligible only if it is interpreted as an instance of the notional idiom. This is, I urge, a mistake; as before, an instance of the relational idiom should be used in order to allow correct substitution of the variables. This third objection to QML constitutes Quine's second criticism of the relational idiom. Quine did come to terms with the relational idiom in modal contexts from 1943 to 1947, when he tentatively allowed the solution given by Alonzo Church and Rudolf Carnap to his concerns raised in "Notes on Existence and Necessity" (hereafter, "NE&N") [1943]. Their solution depended crucially upon construals of modal formulae which were instances of the relational idiom. Quine came to reject this position, which I term, in § 8, the ' Church/Carnap solution1 , because of its reliance on the notion of analyticity. That was the reason given in the first version of "R&M," which made its appearance in 1953 as part of FLPV. This objection not only constitutes the fourth problem that Quine has with QML contained in"R&M," but also the third explicit criticism of the relational idiom. The argument was subsequently improved upon in the second edition of FLPV, which made its appearance in 1961. These arguments are critically evaluated in Chapter III. To be more precise, whereas 20 Quine's initial rejection of the Church/Carnap solution marks his third explicit criticism of the relational idiom, his subsequent argument against the Church/Carnap solution marks his ninth explicit criticism of the relational idiom. The fifth objection to QML contained in "R&M" concerns the charge that invidious Aristotelian essentialism must be resorted to if one is to employ the relational idiom. This objection to QML marks Quine's fourth explicit criticism of the relational idiom. Quine became aware that sentences such as (5) can be given a different kind of relational interpretation by 1953, when he wrote "R&M." By 'a different kind of relational interpretation1 , I have in mind the proposal of Arthur F. Smullyan, who applied TDD to modally ambiguous sentences. This criticism of the relational idiom constitutes the fourth, and final, explicit criticism of it in "R&M." Quine's argument that Smullyan's use of Russell's TDD invokes Aristotelian essentialism can be distinguished from Quine's argument that essentialism is incoherent. In "Three Grades of Modal Involvement" (hereafter, "TGMI") [1953] and elsewhere, he argues that such interpretations are untenable because of the attendant ontological strain. The ontological strain is simply the alleged incoherence of essentialism. These arguments against essentialism constitute Quine's fifth criticism of the relational idiom. Both of Quine's arguments against essentialism (there are two) are argued in Chapter IV to 21 be unsound. Quine's sixth criticism of the relational idiom is that it violates his principle that legitimate entities must have adequate identity criteria. For Quine, the universe is populated only with extensional objects. Any object that does not satisfy this extensional definition of identity is ipso facto not to be admitted into Quine's ontology. Significantly, attributes are in contravention of the definition. Quine would hold that there is no extensional condition c such that, for any attributes P and Q, P = Q iff c. Interestingly enough, it was Quine who first broached the possibility of handling the relational idiom by invoking attributes in "Q&PA." But Quine's endorsement of attributes was half-hearted at best; he is much more content with banning the relational idiom altogether. I recognize that this interpretation of "Q&PA" is somewhat contentious; my position is argued for in Chapters V. and VIII. Quine's half-hearted endorsement of attributes in 1956 was hinted at in several of his papers; his treatment of this matter was nowhere more thorough than in "On the Individuation of Attributes" (hereafter, "IA") [1975]. It should be noted that Quine has never specifically criticized an account of de re modality that invokes attributes. One would imagine that he does not feel this to be necessary since (a) he did not fully support the use of attributes in "Q&PA" in order to account for the idiom in doxastic contexts and (b) he has made known his precise reasons for rejecting 22 attributes. Effectively, then, Quine has argued against the solution that I favour. In "Intensions Revisited" (hereafter, "IR") [1977] Quine appears to be on the verge of extending his 1956 treatment of propositional attitudes to cover cases of de re modality. But Quine argues instead that there is simply no reason for philosophers to be perturbed by the relational idiom in either doxastic or modal contexts. He makes two main contentions in "IR": (i) that his former view that there is a practical difference between notional statements and relational statements was mistaken; (ii) both the notion of essence and the notion of knowing or believing are dependent on context to be meaningful, (i) appears to be an entirely new line of thought. Quine follows this argument with an attack on Kaplan's notion of a vivid designator, which he regards as being on a par with that of a rigid designator. (ii) appears to be little more than a restatement of his third objection to QML in "R&M"; Quine demonstrates that this criticism also applies, in his view, to accounts of propositional attitude ascription. In effect, (ii) is nothing more than an extension of an unsound argument, (i) is a more complicated argument, but I will argue in Chapter VIII that it is nonetheless untenable. Quine's seventh criticism of the relational idiom involves the transparent sense of belief. Quine argues to the effect that acceptance of the relational idiom leads to clearly unacceptably high levels of belief. He contends that acceptance of the 23 relational idiom in doxastic contexts results in ascribing total belief on the part of an agent. WO [1960], features this seventh criticism of the relational idiom, which is that the transparent sense of belief, the sort of belief that one would have if instances of the relational idiom were meaningful descriptions of belief, is ' intolerably odd' . His view that the relational idiom in doxastic contexts is intolerably odd is his only criticism of the idiom which is not clearly applicable to instances of the relational idiom in modal contexts. It could be argued, however, that Quine attempted to extend this charge of intolerable oddity to encompass the relational idiom in modal contexts as well. This apparent application of Quine's argument occurs in § 21.2. Quine's eighth criticism of the relational idiom also occurs in WO. It is concerned with Quine's supposed proof that an attempt to "legitimize quantification into modal position" ( WO, p. 197, p. 156 of this thesis) will result in the collapse in distinction between sentences which are true and those which are necessarily true. I argue that this particular criticism appears to be a very close cousin to the seventh one. In particular, the reason that they are both unsound seems to me to be identical, namely, that they both turn on exploiting the ambiguity between de re and de dicto construals of pertinent sentences. Quine's eight criticism is discussed in § 21.2. Quine*s ninth criticism of the relational idiom occurs, as we have seen, in the second edition of FLPV, when Quine rewrote 24 certain paragraphs in "R&M." Quine more convincingly argues against the Church/Carnap solution by showing that it is incoherent. I should say that the third and the ninth criticisms of the relational idiom are unique in that I find no fault with them. (Recall that it is Quine's initial rejection of the Church/Carnap solution that I am referring to as the third criticism of the relational idiom.) Quine's ninth criticism is discussed in § 10.3. Quine's tenth criticism of the relational idiom concerns his view that instances of it betray a confusion between use and mention. Quine indicates that this view of his stems from the thirties. Quine has indicated that much of the initial interest in modal logic (and this of course includes QML) had questionable motivations which were a simple use/mention confusion. Such a charge is very appropriate when it is levelled against C.I. Lewis. Lewis failed to distinguish between metalanguage and object language in his treatment of possibility. Nevertheless, most of what C.I. Lewis proposed concerning modality is valid as well as useful (for de re modal theorists). Certainly the accounts of QML given by Kripke managed to avoid Quine's criticism that modal logic was born in sin. This criticism is discussed in § 8. His eleventh criticism of the relational idiom very likely also has deep roots for Quine: the view that instances of the relational idiom violate Occam's Razor. Occam's Razor provides, it is true, intuitive support for the view that there are no 25 instances of the relational idiom. Running counter to this intuition is the position that such instances are meaningful, and that a semantic theory is required to account for this meaningfulness. I hold that sentences expressing the relational idiom in at least modal and doxastic contexts are both meaningful and are meaningful not merely because they are disguised instances of notional sentences. Nor do I hold that de re sentences are meaningful in virtue of being analyzable in terms of de dicto ones. Quine's tenth criticism is discussed in § 24. The twelfth and thirteenth criticisms that Quine made of the relational idiom occurred in his 1976 essay "Worlds Away" (hereafter, "WA"). The twelfth criticism of the relational idiom is based on transworld identification. Unlike the eleven previous criticisms, the problems of transworld identification did not have their beginnings with Quine. Quine argues that the predicates which apply to objects that exist in more than one possible world will be fatally stricken. Quine's ontological principles are also violated, he feels, by the postulation of possible worlds. "WA" made its appearance in 1976. I would contend that the most compelling answer to the problem of transworld identification was made six years prior in Kripke's Naming and Necessity. Kripke basically argued that the problem of transworld identification, that is, the problem of individuating an object which exists in more than one possible world, does not 26 arise because the object can be stipulated to exist in different worlds prior to any concern about individuating the object. Quine's twelfth criticism is discussed in § 28. The thirteenth criticism that Quine makes is precisely the same as the one he makes with respect to transworld identification, except that now his target is transbelief world identification. Again he thinks such identification seriously impugns our notion of a physical object. Quine's thirteenth criticism is discussed in § 29. The fourteenth and fifteenth criticisms that Quine makes of the relational idiom are similarly paired in one paper, namely, "IR" which made its appearance in 1977. In that paper, Quine attacks the notion of a vivid designator, which is a term that an agent uses "when there is a specific thing that he believes it designates." His attack on vivid designators constitutes his fourteenth criticism of the relational idiom. Quine's fourteenth criticism is discussed in § 31. In "IR," Quine unveiled an argument which is directed at the supposedly significant distinction between de re (or relational) sentences and their de dicto (or notional) counterparts. Quine holds that Kripke's notion of rigid designator is dependent upon context. This attack marks the fifteenth criticism of the relational idiom. Like Quine's fourteenth criticism, his fifteenth criticism of the relational idiom is discussed in § 31. Quine's sixteenth criticism of the relational idiom 27 directly contradicts his claim, made in WO, that transparent senses of belief have both tolerably and intolerably odd results. In Pursuit of Truth [1990], Quine is critical of what he earlier saw as being a tolerably odd result: with a transparent sense of belief, an agent can unknowingly have contradictory beliefs. This is Quine's last novel criticism that has been directed towards the relational idiom. (I argue that this final criticism marks a substantial change of views, on Quine's part, with his 1960 criticism that the so-called "transparent sense of belief" is intolerably odd.) This criticism is discussed in § 22. Quine has given reasons for barring instances of the relational idiom, both technical and nontechnical. I am almost exclusively concerned with the former category in the body of the thesis; the latter one is, however, of some interest. The following citation is taken from WO: In ordinary non-philosophical usage 'possibly' usually serves merely as a modestly impersonal rewording of what is really a personal idiom of propositional attitude after all: 'I am not sure but what'. Ordinarily the construction 'necessarily' does not, curiously enough, carry the corresponding sense 'I am sure that'. Often it connotes rather a propositional attitude of purpose or resolve. Sometimes, also, 'necessarily' and 'possibly' provide a condensed way of saying that a sentence follows from or is compatible with some fixed premiss understood as background. And sometimes they provide little more than a variant style for 'all' and 'some1 .^ One of the results that I hope to make clear in the course of this thesis is that, at root, Quine is sceptical of philosophical usage of modal terms if that usage differs from 28 the non-philosophical kind. § 4. General Observations on Quinean Criticisms of the Relational Idiom Quine's arguments against the relational idiom in modal and doxastic contexts are mainly of two types: firstly, there are arguments which do not allow for the existence of any ambiguity between the notional and relational idiom. In general, I think this is a direct result of Quine's predominant tendency to view the meaningfulness of instances of the relational idiom as more apparent than real; their supposed meaningfulness is best accounted for by providing an analysis in terms of instances of the notional or de dicto idiom. Secondly, there are arguments which are especially tailored for the relational idiom in both contexts. Quine's criticisms of the relational idiom in modal contexts are largely applicable to doxastic contexts as well. They are suggestive of a philosophical theory that treats modal and doxastic phenomena as being integrally related. Quine himself is very close, at various times during his career, to suggesting such a link between these two phenomena. I believe that Quine's criticism of the relational idiom encompasses principally five stages. The first stage covers the publications prior to "R&M," which appeared in 1953. In papers such as "Designation and Existence" [1939], "Whitehead and the 29 Rise of Modern Logic" [1941], "NE&N," and "The Problem of Interpreting Modal Logic" [1947], Quine fails to acknowledge that statements are occasionally given a relational interpretation. The second stage is characterized by Quine's use of objections to QML that are distinct from those associated with the phenomena of substitutivity and existential generalization. In "R&M," Quine unveils a new objection to QML, the one that was previously referred to as the problem of unique specification. This second stage is also characterized by the recognition by Quine that there have been two attempts to counter his objections by employing the relational idiom in modal contexts. In "Q&PA" Quine, for the first time, points out that certain statements are ambiguous in that they may be interpreted by users of language as conveying one of two very different meanings. This marks the third stage of his critique of the relational idiom. As has been pointed out, Quine's famous example is (1) Ralph believes someone is a spy. Prior to "Q&PA," Quine would not have allowed (1) to be considered a counter example to his objections to quantifying into doxastic contexts. The 1956 paper constitutes the third stage of Quine's approach to the relational idiom. In 1960, Quine published WO, which ushered in the fourth stage of Quine's criticisms of the relational idiom. The finding 30 against the relational idiom in doxastic contexts on the grounds that the distinction between propositions believed to be true and those which are merely true evaporates arguably has an analogue in modal contexts. That is, Quine also argues in ¥0 that quantification into modal contexts results in an evaporation of the distinction between true modal propositions and propositions that are merely true. The fifth and final stage of Quine's criticisms of the relational idiom dates from 1977, when Quine wrote "IR." This paper contains the argument in that there is no sharp distinction between instances of the relational and notional idiom. Quine is correct that essentialism of at least some kind will have to be countenanced by philosophers seeking to overturn the first four objections in "R&M." Some space is therefore devoted to the different kinds of essentialism in § 17. My acceptance of certain essentialist theses is steeled by Quine's own view that essentialism must be entertained if it is first assumed that the relational idiom in modal contexts is meaningful. I am willing to make this assumption in light of sentences such as (9) The author of Hamlet might not have written Hamlet. Kripke holds, that sentences such as (9) are intuitively meaningful. Kripke's view is supported by a convincing argument. I intend to endorse and expand Kripke's argument to cover 31 doxastic cases as well as modal ones. It is true that certain philosophers, including Kaplan and, at various times, Quine himself, have assumed that the relational idiom in doxastic contexts is meaningful. However, they have not argued in favour of such a position in modal cases. In no way is Quine opposed to de dicto modality (that is, he has no objections to modal operators serving as prefixes of sentences). The relational, as opposed to notional, idiom in ascriptions of propositional attitudes has proved to be a source of unease for Quine. As was earlier mentioned, in "Q&PA" Quine was willing to countenance the meaningfulness of sentences such as the previous (1*) There is someone (in particular) whom Ralph believes to be a spy. However, de re doxastic sentences like (1*) pose something of a problem for Quine, in that they are not only prima facie meaningful, but also that they are bona fide examples of quantification into what he thinks is an oblique context. But it can be shown that Quine's position on the relational idiom in doxastic contexts is somewhat inconsistent with his corresponding position that such instances in modal contexts are meaningless. Such an argument should prove useful and interesting. The first three of Quine's objections to QML clearly demonstrate his unwillingness to use modal operators as anything but modifiers of sentences. His fifth objection to QML, and his 32 second criticism of relational modality, inform us as to the reason that Quine is unwilling to allow modal operators to be embedded within the scope of a quantifier, to allow modally de re formulae. Quine's concerns about avoiding a bloated ontology with respect to modalities make two appearances in his writings. The first occurrence of this criticism is to be found in "On What There Is." The second occurrence is in "TGMI." Evidently, Quine's opposition to possible entities is simply Occam's Razor-he feels that they are an unneeded addition to his universe. This criticism is, in my opinion, the most powerful one in Quine's arsenal. I think it can be overcome, principally by showing that Quine's preferred way of dealing with the relational idiom conflicts with our intuitions. Quine is correct, I think, in arguing that the logic of belief is on a par with the logic of modality. Like his occasional assumption of the meaningfulness of the relational idiom, Quine's views regarding the interconnectedness of the logic of belief and the logic of modality are not subjected to the rigors of argument. The final part of this thesis is concerned with the parallel significance of the relational idiom in doxastic contexts (propositional attitude ascriptions) and Quine's criticisms of relational belief. Allow me to indicate some of the broader distinctions that exist between Quine's approach towards modal and doxastic phenomena as compared with my own. For reasons that are made clear in Chapter III, Quine rejects Smullyan's solution to the 33 three initial objections to QML in "R&M." I support Smullyan's application of Russell's TDD to account for ambiguous modal sentences, such as our previous (5). However, I would extend Smullyan's solution to cover unambiguous modal sentences, such as our previous (9). Quine would actually agree with this assessment of (9) as being unambiguous. Only Quine would view (9) as unambiguously de dicto. I, on the other hand, view (9) as being unambiguously de re. I previously applied Russell's TDD to (5). If a similar application of Russell's TDD is made to (9), it becomes immediately clear that this sentence is indeed an unambiguously de re sentence. Applying Russell's TDD to (9) would yield the two following de dicto sentences, (9*) 0 [(3x) ({Hx A (Vy) (Hy - x = y) ) A ->Hx) ] and (9**) 0 i[(3x) {(Hx A (Vy) {Hy - x = y) ) A Hx) ] . Now, I do not think that either of these symbolizations would constitute an adequate analysis of (9); the internal sentence within (9*) would be a contradiction if its quantifier were instantiated (that is, the sentence prefixed by the '<>' within (9*) would be a contradiction if its quantifier were instantiated) and (9**) does not convey the same meaning as does (9). But a de re interpretation of (9), perfectly captures the meaning of the original: 34 (9*) (3x) [(Hx A (Vy) (Hy - x = y) ) A O-tfx] . Having argued both that Quine's criticisms of de modality are unsound and that the relational idiom in modal contexts is coherent, the route is open to providing a semantic theory which accounts for this coherence. The version of essentialism that I defend is contrasted with other types of essentialism, both trivial and nontrivial. The manner in which the meaningfulness of the relational idiom in modal contexts is accounted for also applies to the meaningfulness of the relational idiom in doxastic contexts. Curiously enough, this solution was first outlined by Quine himself in "Q&PA." Yet Quine did not adopt a comprehensive approach to modality and belief in this paper. When he did recognize the merits of a comprehensive approach in "IR" it was part of his general rejection of all instances of the relational idiom, or at least, all the ones in modal and doxastic contexts. As I have indicated, the goals of this thesis have the following motivation: I argue that the relational idiom is meaningful in both doxastic and modal contexts. In order to account for the meaningfulness of the relational idiom in such contexts, semantic theories need to be provided. These semantic theories in turn result in an ontology which constitutes an eyesore for those who have a love of desert landscapes. Quine's objections have produced numerous responses. I intend to give a critical explication of Quine's objections to the relational idiom in modal 35 and doxastic contexts and to combine this with an elaborate discussion of precisely where I think the difficulties arise with his treatment. Before proceeding further, it is useful to summarize briefly the main points of this chapter: (1) Quine has criticized the relational idiom in both modal and doxastic contexts for a variety of reasons in his published writings, beginning in 1941 and ending in 1990; (2) The sixteen criticisms of the relational idiom in either modal or doxastic contexts that Quine put forward may be roughly divided into five stages; (3) Though occasionally willing to entertain the notion that the relational idiom in doxastic contexts is significant, Quine is utterly unwilling to entertain the analogous notion concerning the relational idiom in modal contexts; (4) Quine devised two treatments of the relational idiom in doxastic contexts that takes into account their significance, but did not do likewise for the relational idiom in modal contexts; and (5) a treatment of the relational idiom in modal contexts can be fashioned pace Quine's negative treatment in doxastic contexts. This account will follow a defense of essentialism against the attacks given by Quine. Throughout the thesis, a possible worlds semantics will be assumed. The intuitive significance of instances of the relational idiom in modal contexts will be argued for. Finally, all of Quine's criticisms of the relational idiom in both doxastic and modal contexts will be urged to be, with certain noteworthy exceptions, unsound. 36 Chapter II. Quine's Initial Criticisms of QML and his Positive Account of Modality § 5. General Remarks This chapter deals primarily with both Quine's negative account of QML and his positive account of modality. With respect to the former, I propose to introduce the three initial objections to QML that Quine collectively put forward in "R&M": (i) the problem based on the principle of substitutivity of identicals (hereafter, (SI)); (ii) the problem based on the phenomena of existential generalization (hereafter, (EG)); and (iii) the problem of unique specification (hereafter, (US)). Taken together, these three objections constitute, for purposes of this thesis, the initial stage of Quine's criticisms of the relational idiom. These three objections are transitional with respect to the relational idiom in that this idiom is only partly the principal object of Quine's charge. Instead, Quine's main target is modal logic, the logic of possibility and necessity. The transition that I have in mind is from objections to QML to more direct criticisms of the relational idiom. With respect to the latter objective of this chapter (that is, the positive account of modality given by Quine), I intend to explain briefly Quine's method of semantic ascent. Quine's positive account of modality is introduced through innocuous 37 examples. I then develop an argument to the effect that this account cannot cope with certain modal sentences, namely, those modal sentences that are instances of the relational idiom. The analysis of Quine's method of semantic ascent is followed by a discussion of yet another of Quine's papers, namely, "TGMI." There, Quine gives an account of modality that may be considered to be both neutral and positive. He describes three degrees of modal usage, or "involvement" as he calls it in the title of the paper. Finally, this chapter concludes with certain brief comments regarding Quine' s 1962 admission that, in his opinion, modal logic was conceived in the sin of confusing use with mention. Though modal logic is Quine's main target in "R&M," he also succeeds in indicting and (temporarily, at least) condemning belief. This verdict of 'philosophical guilt by association' is facilitated by what Quine saw (again, temporarily) as the connecting link between modal and doxastic contexts: both contexts are referentially opaque. This technical term is defined in § 6.1. § 6. Quine on Substitutivity of Identicals, Existential Generalization, and Unique Specification: Initial Stages of Quine's Critique of the Relational Idiom § 6.1 Substitutivity of Identicals In "NE&N," Quine introduces the notion of an occurrence of an expression as being not purely designative. Occurrences of 38 expressions that are purely designative refer, as the name implies, to the objects that are designated. Quine has consistently held that failure of substitutivity shows that the occurrence of a name exhibiting such failure is not 'purely designative' (later, not ' purely referential' ) . The following passage is taken from Quine 's 1941 paper "Whitehead and the Rise of Modern Logic": ... the statements: 0(number of planets < 7), 0(9 < 7) would be judged as true and false respectively, despite the fact that they are interconvertible by interchanging the terms ' 9' and 'number of planets' , both of which designate the same object. Using Quine' s own terminology, the occurrence of '9' in the expression "<>(9 < 7)" is not purely designative. This citation appears to be the earliest occasion at which Quine voices a concern about the failure of substitutivity in modal contexts. It appears to establish that Quine regards failure of substitution as a means by which one can demonstrate the inviability of modal statements, even if these statements are instances of the notional idiom. Later, he would transform this point more explicitly into a charge of an invalid argument in "The Problem of Interpreting Modal Logic" In "NE&N," as well as in "R&M," Quine early on introduces the following principle, which is one "of the fundamental principles of identity": "given a true statement of identity, any one of the two terms may be substituted For the other in any true statement and the result will be true." Quine gives the following 39 justification for this principle, which is called either the principle of substitutivity or Leibniz's Law: When a singular term is used in a sentence purely to specify its object, and the sentence is true of the object, then certainly the sentence will stay true when any other singular term is substituted that designates the same object ... He shows that this principle fails in modal contexts. His argument features premises which are collectively true, but the conclusion is false: (3) 9 is necessarily greater than 7. (4) The number of planets = 9 (Therefore) (5) The number of planets is necessarily greater than 7. However, (5) seems to be false. Yet (5) appears to have been legitimately obtained from (3) and (4) by invoking Leibniz's Law. Thus, the argument from (3) and (4) to (5) implies, for the reader, either a restriction on Leibniz's Law or on modal sentences. Quine contends that application of Leibniz's Law is unsound when performed on notional sentences. I believe that Leibniz's Law is, with qualification, a sound principle. Quine's conclusion from the failure of substitutivity is that the occurrence of ' 9' in (3) does not make singular reference to the object 9. The failure of substitutivity of a singular term a in a referentially opaque context is indicative of that occurrence of a as not being purely referential. Though Quine does not say so, I feel that the failure 40 of substitutivity of a singular term in a referentially opaque context also indicates that a restriction upon Leibniz's Law will be required, at least in de dicto sentences. (If Leibniz's Law is not thus restricted, then one would be faced with invalid sequence like the ones Quine points to.) Quine introduces the notion of referentially opaque context and holds that referential opacity is present both in doxastic and modal contexts . The basic idea underlying referentially opaque contexts is that they are contexts that do not preserve truth-value of sentences after substitution of one coreferential singular term for another has taken place. However, Quine introduces, in "NE&N," the notion of referentially opaque contexts syntactically and failure of substitution is not a defining characteristic of the contexts. Examples include 'it is possible that' and 'it is believed that1 . Substitution is a standard which, for Quine, is indicative of the status of given contexts, be they doxastic or modal. A fuller definition of referentially opaque contexts will be given in § 6.2 after a discussion of existential generalization. More precisely, application of Leibniz's Law was a standard by means of which Quine determined the viability of modal and doxastic contexts. "R&M" made its appearance in 1953, and, as noted, it was an amalgamation of papers dating back to the early forties. Later in this thesis, it is noted that Quine appears to have significantly altered his stance towards various points of contention. But it should be stressed that the three objections in "R&M" make a regular appearance in Quine's writings; "R&M," hi apart from its initial appearance, reared its head twice more, in 1961 and 1980. Substitutivity, in particular, has remained an abiding concern for Quine, as can be seen in Pursuit of Truth. It is not entirely clear from "R&M" what role Quine is ascribing to (SI) as criticism of QML, although some broader results are loosely implied. In such a broad vein, we may interpret Quine as concluding from (SI) that either Leibniz's Law must be restricted in its application or that QML must be rejected. Yet such an interpretation of Quine inaccurately blurs the distinction between Frege, who first attached significance to failures of substitutivity, and Quine. Quine contrasts his own approach to (SI) to that of Frege: Failures of substitutivity of identity, moreover, were in Frege's view unallowable; so he nominally rectified them by decreeing that when a sentence or term occurs within a construction of propositional attitude or the like it ceases to name a truth value, class, or individual and comes to name a proposition, attribute, or "individual concept." (In some ways this account better fits Church, who has sharpened and elaborated the doctrine.) I make none of these moves. I do not disallow failure of substitutivity, but only take it as evidence of non-referential position; nor dp I envisage shifts of reference under opaque constructions. This passage was taken from WO. We have already seen Quine's stance in WO with respect to failures of substitutivity. Quine had substantially the same view of failures of substitutivity some seventeen years earlier. As evidence for this latter claim, consider the following citation from "NE&N": Failure of substitutivity reveals merely that the occurrence to be supplanted is not purely designative, and that the statement depends not only upon the object but on the form 42 of the name. For it is clear that whatever can be affirmed of the object remains true when we refer to the object by any other name. So Quine's attitude about the failure of substitutivity appears not to have substantially altered from 1943 to 1960, despite the evolution from designation to position. Quine's views on the phenomena of (SI) in NE&N may be seen as constituting one of several solutions to Frege's antinomy of the name-relation. Frege's antimony arises from the substitution of coreferential names or definite descriptions. The problem is that an agent might comprehend a sentence featuring one particular name and yet not understand that this name or definite description is coreferential with other names or definite descriptions. For example, an agent might comprehend the following sentence: (10) Jean Chretien cancelled the helicopter contract, and yet not realize that (11) Jean Chretien = the Prime Minister of Canada The point is that our agent might readily agree to (10), but would not readily be in accord with (12) The Prime Minister of Canada cancelled the helicopter contract. Frege recognized that there was a different cognitive valuation accorded to sentences like (10) and (12), despite the truth of sentences such as (11). 43 Carnap discusses Frege's puzzle and six possible responses in Meaning and Necessity. "NE&N," with its emphasis upon extensionality, is one of the solutions to Frege's antinomy given consideration by Carnap. Like Frege, Quine draws a sharp line between extensional and nonextensional contexts. But whereas Frege assigns a separate nominatum to a name occurring in a nonextensional context, Quine takes the simpler route of holding that there is no assignment to be made in the first place. The following passage by Carnap contains his assessment of Quine's solution: The advantage of Quine's method in comparison with Frege's consists in avoiding the immense multiplication of entities in corresponding names to which the latter method leads. But Quine's method pays a high price for this simplification by restricting the name-relation ( 'designation' ) to extensional contexts and grouping all nonextensional contexts together with contexts in quotation marks and, further, by imposing narrow restrictions upon the use of variables in modal sentences. Frege was the first to stress the oddity of coreferential singular terms conveying different cognitive value in pairs of sentences. Quine's solution to Frege's puzzle is by and large in accord with the spirit of his objections to the relational idiom. Quine comes very close to making a full blown objection to QML in "R&M" on pp. 152-153. Quine nearly condemns QML on the grounds that the following sufficient condition is required "for the intelligibility of quantifying into modal contexts": "the universe is to consist of things which are never contingently identical, which are necessarily identical if identical at all."5 This near condemnation of QML occurs when Quine is discussing the Church/Carnap solution. (This matter is returned to in § 10.2.) 44 The principle of substitutivity is occasionally conjoined with the following de re modal formula in order to obtain an apparently unsatisfactory metaphysical result: (Vx)D (x = x) . By universal specification, we obtain: • (a = a) The allegedly unsatisfactory result comes after an ordinary statement of identity has been made, say, that Samuel Clemens is identical to Mark Twain, and the previously given modal statement of identity, jointly yield (13) •(Samuel Clemens = Mark Twain). However, sentences like (13) are not unsatisfactory. Notice that by universal specification one obtains (14) •(Samuel Clemens = Samuel Clemens) and then by substitution of "Mark Twain" for the 2nd occurrence of "Samuel Clemens" one obtains (13), which is different from (15) R( Samuel Clemens = the man named 'Mark Twain>) -whereas the first (that is, (13)) is true, the second (that is, (15)) is false. As will be indicated in Chapter IV, this argument undergoes a slight transformation from its first appearance in "R&M" in 1953 to its later appearances in the same paper in its 45 1961 and 1980 versions. In brief, Quine uses the argument as a way of demonstrating that any quantified modal logic that countenances quantification into modal contexts must also be committed to some version of essentialism. § 6.2 Existential Generalization (ii) Quine believes that one can eliminate singular terms by paraphrase. Quine writes that the phenomenon of referential opacity has . . . been explained by appeal to the behaviour of singular terms ... Ultimately the objects referred to in a theory are to be accounted not as the things named by singular terms, but as values of the variables of quantification. So, if referential opacity is an infirmity worth worrying about, it must show symptoms with quantification as well as in connection with singular terms . This is why Quine moves on to consider problems based on the phenomena of existential generalization; he realizes that the problem based on the phenomena of (SI) might be met by someone using Russell's TDD. In 1939, the appeal to existential generalization lay, according to Quine, in its ability to separate two of the three main metaphysical theories which purport to give an account of abstract objects. The three theories are (i) conceptual ism, which holds that abstract objects are mere concepts in the minds of men; (ii) realism, also known as platonism, which holds that abstract objects are, in some way, legitimate entities that occupy the universe independently of the contents of minds; and (iii) nominalism, which holds that there are no abstract objects, despite 46 our having words or inscriptions that seem to refer to such entities. Quine first discussed existential generalization, and its motivation, in "Designation and Existence." In this paper Quine argued that it was difficult to articulate the differences between realism and nominalism: If the nominalist who renounces such abstract entities as horse, unicorn, and appendicitis does not thereby foreswear any of the ordinary uses of these words, nor take issue on any factual questions of zoology and medicine, then what does his renunciation amount to? Any appeal to nature, such as was involved in the case of Pegasus and hyperendemic fever [Quine's mythical disease that supposedly killed or maimed four fifths of the population of Winnipeg in 1903-M.W.D.], seems now to have become irrelevant. What is left but a bandying of empty honorifics and pejoratives-"existent" and "non-existent," "real" and "unreal"? In "Designation and Existence," the problem for Quine is one of formulation: how does one formulate the difference between the realist and the nominalist? Quine*s answer is to appeal to the principle of existential generalization. Quine notes that if a word, such as "appendicitis," designates an entity, then a meaningful statement, such as (16) Appendicitis is dreaded, implies the consequence (16*) (3x) (x is dreaded). That is, if (16*) were not a valid consequence of (16), then "appendicitis" fails to designate an entity. This suggests a way to formulate the differences between the nominalists and the realists. Whereas a nominalist would cite the inability to obtain 47 a sentence such as (16*) from a sentence such as (16), a realist would, at least generally, accept this entailment. Quine does not hold that the nominalist would regard terms such as "appendicitis" as meaningless. Indeed, the nominalist would look on such terms as being both meaningful and useful in context. (This point is important to keep in mind, especially in regard to both the second and third objection to QML in "R&M.")13 The problem of existential generalization with regards to modality was raised for the first time by Quine in "NE&N." In this paper, Quine does not spell out the connection between existential generalization and modality. Instead, he makes certain points concerning the relation between designation and quantification with respect to statements such as (17) Giorgione was called 'Giorgione' because of his size, and (18) Barbarelli was called 'Giorgione' because of his size. Quine's use of quotes is extremely significant with respect to his treatment of modality and belief. Quine views quotes around proper names as causing the analogous types of phenomena, namely, the generation of opacity, as are caused by modal and doxastic operators. It is not significant that Quine never repeated the rationale for (EG) mentioned above in "R&M." In the citation just provided, Quine is considering a hypothetical criticism of nominalism, one 48 that he shortly rejects in "Designation and Existence." By the time that Quine wrote "R&M," Quine had ceased to defend the nominalistic theories that he and Nelson Goodman had advocated. Quine enlists the assistance of the motivation (that is, the rationale) behind (EG) in the defence of nominalism. Hence, it is understandable that Quine would cease to give this particular rationale for (EG) after he had given up on nominalism. Quine touches on the phenomena of (EG) as it applies to modality by contrasting the statement (3*) Necessarily (3x) (x > 7), with the following expression: (3') (3x) x is necessarily greater than 7. Quine interprets (3*) as being the same as (3**) Necessarily something > 7. He holds that (3**) "still makes sense, being in fact a true statement ..." 5 However, (3') is logically equivalent to (3'') There is something that is necessarily > 7. (3* ' ) is meaningless, Quine holds, for "would 9, that is, the number of planets, be one of the numbers necessarily greater than 7?" If this were the case, he holds, it would contradict his earlier view regarding the falsity of 49 (5) The number of planets is necessarily greater than seven. This latter move on Quine's part has long puzzled me. My current view is that Quine has simply exploited, in a thoroughly obfuscatory manner, the ambiguity of (5) so as to interpret it strictly as a de dicto sentence. That is, it would make sense to interpret every ambiguous sentence like (5) as being a de dicto sentence if one is firmly committed to the view that there are in fact only de dicto sentences even where, apparently, there are genuine de re sentences. We are now in position to define what Quine has termed ' referentially opaque' contexts: they are contexts that do not allow the principle of substitutivity of identicals and that do not allow existential generalization. In "R&M," Quine points out that the phenomenon of (SI) and the problem based on the phenomenon of (EG) apply to both modal and doxastic contexts. The first stage of Quine's criticisms of the relational idiom, then, exhibits a more or less uniform treatment of these two contexts. As is shown in subsequent chapters, this facet of Quine's thought changes considerably over time. Quine recognized that an objection based on the logical operation of (EG) also applied to doxastic phenomena. He notes that existential generalization is, as he puts it, "unwarranted" when applied to sentences such as (19) Penelope is unaware that Tully denounced Catiline, 50 since this leads to (19') (3x) (Penelope is unaware that x denounced Catiline). Quine then asks what is this object, that denounced Catiline without Philip's [Penelope's] having become aware of the fact? Tully, that is, Cicero? But to suppose this would conflict with the fact that (11) [that is, (11) Philip [Penelope] is unaware that Cicero denounced Catiline-M. W. D.] is false. However, there is a fallacy here that is precisely analogous to the one Quine has committed with respect to application of existential generalization on modal sentences such as (5). That is, Quine's (11) is a de dicto sentence. But if we use the more appropriate de re interpretation, no oddity arises. Here is a de re variant of Quine's (11), only this time in ordinary, and only slightly tortured, English: (19'') Tully is such that Penelope is unaware that he denounced Catiline. (19 * * ) is true, though Penelope may not be aware that she is unaware. Quine is aware that existential generalization on a modal or doxastic sentence will never lead to a de dicto statement. His example from the 1943 paper is that the de dicto statement (19*) Philip [Penelope-M. W. D.] is unaware that (3x) (x denounced Catiline), (is quite straightforward and in no danger of being inferred by 51 existential generalization from ..." our previous (19). The positive account of instances of the relational idiom in doxastic contexts that I defend allows for Penelope to be aware of Tully's having denounced Catiline even though Penelope would not agree to this fact because of a lack of information. Quine demonstrates that there is a distinction between attributing a relational pair of beliefs that are contradictory on the one hand, and attributing a belief in a contradiction on the other in "Q&PA." ' For this reason, in WO, Quine referred to this as 'tolerable oddity'. I do not differ with Quine as to the truth value of (19); I share with him the ability to interpret (19) as a de dicto statement. It might be thought that Forbes' definition of the distinction between modal de re and modal de dicto sentences can be easily extended to cover the doxastic case. But this would be a mistake. Here is an attempted doxastic version of Forbes' definition: A sentence S with a doxastic operator D is a doxastic de re sentence iff either (i) D prefixes variables that occur free within S; D prefixes an individual constant within S; or (iii) D is both followed by variables that do not occur free and is preceded by a quantifier which binds the variables in question. A sentence S with a doxastic operator D is a doxastic de dicto sentence iff it is not a doxastic de re sentence. But such an extended version of Forbes' definition is ultimately unsatisfactory. For it forces one to conclude that the presence of an individual constant within a doxastic sentence is an indication of its being de re, as opposed to its being de dicto. As we will see in Chapter V, Quine has found a way in English to provide a semantic distinction between the relational and the 52 notional sense of belief. His method does not involve this particular difficulty with the syntactic definition of doxastic sentences alluded to above. But there is a way that we could switch to a formal language that will alleviate our difficulties. This method was first proposed by Michael Dummett. The general idea is to subsume the predicates of a given formula into one single predicate. For de re doxastic sentences such as (19), the result would be IXx Bt Dxic] c, where ' Bt' represents Tom's believing, ' k" represents Catiline, and ' c represents Cicero. Dummett's notation is meant to disallow substitution and existential generalization within the square brackets. However one can substitute coreferential singular terms that are outside the brackets. In this case, one cannot substitute for the ' k* , which is within the square brackets, but one can substitute ' 1" (my representation of Tully) for the coreferential c . This is a way to characterize de re doxastic sentences that avoids the difficulty mentioned above when Forbes' syntactic definition of the de re/de dicto distinction is extended to cover doxastic contexts. For de dicto doxastic sentences, one can simply invoke Quine's suggestion, in "Q&PA," of a dyadic relation between a believer and a proposition: Bt Dck. 53 In this case, we add the stipulation that neither of the individual constants that follow the 'D' are open to either substitution or in existential generalization. The supposed failure of existential generalization in modal contexts allowed Quine to conclude that such instances of quantifying in were meaningless. This finding of Quine*s would later cause a problem, because instances of the relational idiom in doxastic contexts were later acknowledged to be meaningful. For example, the earlier (1*) There is someone (in particular) whom Ralph believes to be a spy. is viewed by Quine as being a meaningful instance of quantifying into a doxastic context. Quine's ambivalence on the question of meaningful instances of the relational idiom in doxastic contexts has never been satisfactorily resolved. Quine is, however, much more adamant when it comes to rejecting instances of the relational idiom in modal contexts. This is a difficult point to appreciate given that Quine supplied examples of the relational idiom in propositional attitude ascriptions before he wrote "R&M." § 6.3 Unique Specification (iii) Quine's next objection to QML in "R&M," like the previous one based on (EG) concerns the meaningfulness of the relational idiom in modal contexts. Quine's argument is that necessity is 54 applicable to the connection between certain consequences, like 'x > 7' and particular ways of specifying x. Quine specifically holds that 'x > 7' is a necessary consequence of only the first formula, even though both formulae uniquely determine the same number x: (7) x = fx + TTX + ffx # Vx (8) there are exactly x planets. I call this objection the phenomena of unique specification (US). As we will see in Chapter IV, my argument against (US) challenges the assumption that Quine is correctly describing the nature of traits. There is no argument in favour of this assumption, either in "R&M" or elsewhere. There are, roughly speaking, three ways that Quine's objection based on (US) may be interpreted. The first two involve taking Quine as making objections rooted in metalinguistic notions. The first of these interpretations would have Quine holding that the first of the following formulae would be true and the second false: (7 *) x = Sx + Ix + -fx * Jx \ x > 7 (8*) There are exactly x planets ui- x > 7. The second interpretation of the problem of (US) would have us draw the following contrast between (7**) "x = Jx + Jx + Jx * Jx" N "x > 7" and 55 (8**) "there are exactly x planets" i= "x > 7" using the semantic (model-theoretic) notion of consequence. The left side of (8**) apparently contains a quantifier. Use of the predicate "NP" (with the intended interpretation "is a number of the planets") allows us to obtain (5*) "NPx A (Vy) (NPy - y = x)" H- "X > 7. " It should be pointed out that none of the expressions contained in quotation marks is a sentence. However, it could be held that given any interpretation anything that satisfies the left side of (7**) also satisfies the right side of (7**) in that interpretation and so (7**) is true. This is in sharp contrast with (8**). There is an interpretation in which something satisfies the left side of (8**) but which does not satisfy the right side of (8**). (8**) is therefore false.22 However, any metalinguistic interpretation lacks sufficient similarity with Quine's second objection in "R&M," the one based on the operation of (EG) . This point is crucial because Quine states that the third objection is intended as another way (that is, in addition to the second objection) to argue for the supposed meaninglessness of quantification into modal contexts. The second objection does not involve any metalinguistic notion. Accordingly, the third objection should not involve any metalinguistic notion, either. It may appear that whereas (SI) and (EG) make linguistic points, (US) makes a metalinguistic one. Initially, this view seems 56 quite plausible for whereas Quine's conclusion to his first two objections to QML is that modal contexts are not ones "accessible to pronouns referring to anterior quantifiers," his conclusion to the third objection to QML is that being necessarily greater than 7 is not a trait of a number. But this claim is belied by Quine's own statement that (US) argues "the meaninglessness of (30) in another way." Points about meaninglessness are always linguistic ones. I think the problem of (US), as Quine presents it, makes points that are both linguistic and nonlinguistic. Why does Quine embark on another argument for the meaninglessness of (30)? Immediately subsequent to the problems based on the principle of (SI) and on the operation of (EG), Quine writes that in the present section these contexts [modal and doxastic ones-M. W. D.] have been found referentially opaque by a criterion having nothing to do with singular terms, but with the miscarriage of quantification. The reader may feel, indeed, that in this ... criterion we have not really got away from singular terms after all; for the discrediting of the quantifications (29)-(31) turned on an expository interplay between the singular terms 'Tully' and ' Cicero' , '9' and ' the number of planets' , 'Evening Star' and 'Morning Star' . Actually, though, this expository reversion is avoidable ... In other words, Quine wishes to show the meaninglessness of quantification into modal contexts without relying upon the use of singular terms. As I have indicated, this is exactly what Quine does. The third alternative interpretation of the problem of (US) would be to view Quine as holding that at least the following sentence makes some sense, whereas its counterpart does not: 57 (7***) Q[x = Jx + Jx + fx * fx - (x > 7)], (8***) •[there are exactly x planets - (x > 7)]. This interpretation does not suffer from the defect that I spoke about. It also explains Quine's notion of necessary consequence without recourse to metalinguistic notions. The ' -*' and the ' D' are part of the object language. Nevertheless, Quine's objection appears to me to require at least some argumentation. I would hold that the universal closure of (7***) is true. However, the universal closure of (8***) is false. I will return to the problem of (US) in Chapter IV. § 7. Semantic Ascent One reason that Quine objects to merging quantification theory with modal logic in "R&M" involves the account of semantic ascent. According to Quine in ¥0, semantic ascent is the "shift from talking in certain terms to talking about them." ° This is how Carnap divested "philosophical questions of a deceptive guise" and set them forth "in their true colors." Quine is quick to point out that his application of semantic ascent is wider than that of Carnap. "Semantic ascent, as I speak of it, applies anywhere. "There are wombats in Tasmania' might be paraphrased as ' 'Wombat1 is true of some creatures in Tasmania', . . . ° Such routine examples, though their philosophical significance may be of thoroughly dubious quality, allow Quine to explain the overall importance of semantic 58 ascent. This overall importance is that it allows one to debate the existence of various objects. If one does not ascend to talk of ' mile1 , for example, one cannot engage in a discussion over whether or not miles exist: "Of course there are miles. Wherever you have 1760 yards you have a mile." "But there are no yards either. Only bodies of various lengths." "Are the earth and moon separated by bodies of various lengths?" The continuation is lost in a jumble of invectives and question-begging. When on the other hand we ascend to ' mile' and ask which of its contexts are useful and for what purposes, we can get on: we are no longer caught in the toils of our opposed uses. Quine goes on to point out that this avoidance of question-begging is useful not just in philosophy, but also in the natural sciences and mathematics. It would be a simple matter to produce a very long list indeed of objects whose existence could not be effectively debated without an appeal to semantic ascent; Quine himself gives three examples. As for mathematics, Quine points to the axiomatization which allowed mathematicians to distinguish the logical vocabulary from the axiom system itself. There is a special reason, however, that semantic ascent is important to philosophy. "Most truths of elementary logic contain extralogical terms; thus 'If all Greeks are men and all men are on mortal . . . ,,JU This is sharply in contrast with the truths of physics that contain only physical terms. Quine then concludes that while "we can expound physics in its full generality without semantic ascent, we can expound logic in a general way only by 11 talking of forms of sentences."J  He then points out that whereas 59 physics can be expounded in a thoroughly general fashion by "quantifying over non-linguistic objects, while the dimension of generality wanted for logic runs crosswise to what can be got by such quantification."J Thus, semantic ascent is needed in logic to counter this crosswise running inherent in the generality of logic. Quine introduces both the notion of, and the rationale for, his method of semantic ascent by employing the sort of innocuous examples given above. However, I concentrate here on Quine's use of semantic ascent with respect to modal sentences. This position of Quine's dates back to when he was a follower of Carnap. Quine specifically applies the notion of semantic ascent to modalities in the following: A statement of the form "Necessarily. . . ' is true if and only if the component statement which 'necessarily1 governs is analytic, and a statement of the form 'Possibly. . . ' is false if and only if the negation of the component statement which 'possibly' governs is analytic. * The primary difficulty with Quine 's account of semantic ascent lies with the strong intuition, supported by arguments of varying strength, that there are in fact irreducibly de re sentences in both modal and doxastic contexts. There is strong intuitive appeal to holding that some de re sentences are both meaningful and true. With respect to semantic ascent, this fact is particularly bad inasmuch as semantic ascent is applicable only to de dicto sentences. (There appear to be irreducibly de re sentences in temporal and causal contexts as well.) Quine never extended his account of semantic ascent to cover doxastic cases. I will therefore 60 initially limit my comments upon the intuitive appeal of de re sentences to modal contexts only. Consider again the following sentence: (9) The author of Hamlet might not have written Hamlet. This sentence most plausibly means that the individual who actually wrote Hamlet possibly did not write Hamlet. The intuitive idea is that on the one hand Shakespeare might not have written Hamlet, and on the other hand, Shakespeare is identical to the author of Hamlet. I now attempt to show that Quine' s method of semantic ascent fails to provide an adequate analysis of (9). The first interpretation of (9) using semantic ascent is given in the following: (9***) 'The author of Hamlet wrote Hamlet' is not analytic. The immediate difficulty one encounters is the apparent fact that the interior sentence of (9***) is indeed analytic. I shall return to this point shortly. For the present, it is important to point out that both (9) and the interior sentence of (9***) contain a definite description, which allows us to avail ourselves of Russellian analyses of the initial sentence (9). These various analyses must be considered in order to argue that (9***) is an inadequate treatment of (9). This done, it follows that the semantic ascent account is untenable. Because there are two sentential operators in (9) (that is, the negation and possibility operators), there are no less than 61 three Russellian analyses of (9). Firstly, the definite description in (9) could have scope over the entire sentence: (9'') Exactly one thing wrote Hamlet, and whatever wrote Hamlet might not have written Hamlet. (9' * ) corresponds to the de re interpretation of (9) that was given with the above-mentioned symbolization (9') (see p. 34). For this reason, Quine could not accept (9''). Secondly, the definite description in (9) could be embedded within the scope of the modal operator of possibility but the negation operator would not have scope over the definite description: (9"*) It is possible that: exactly one thing wrote Hamlet, and whatever wrote Hamlet did not write Hamlet. (9''*) is a de dicto interpretation of (9) . It is plainly absurd, since it says that a contradiction is possible. The third analysis of (9) fully embeds the definite description within the scope of both the possibility and negation operators: (9"**) It is possible that it is not the case that: exactly one thing wrote Hamlet and whatever wrote Hamlet, wrote Hamlet. (9"**) is also a de dicto interpretation of (9). Unlike (9*'*) however, (9,f**) is not plainly absurd. Since Quine can accept neither (9'') (because this reading is de re) nor (9''*) (because this reading asserts that a 62 contradiction is possible) , the only viable alternative is (9''**). Though meaningful, it does not mean the same thing as (9). Since the second conjunct of the embedded sentence of (9*'**) (that is, "whatever wrote Hamlet wrote Hamlet") is analytic, (9''**) is equivalent to (9''***) It is possible that it is not the case that: exactly one thing wrote Hamlet. Intuitively, neither (9''***) nor its parent (9»»**) convey the same meaning as (9) does. It should be stated that this intuition is the direct result of an assumption as to the correctness of Russellian analyses of sentences such as (9'). That is, Quine's account of semantic ascent appears to fail if it is assumed that Russell's Theory of Definite Descriptions is the correct method of analyzing sentences which feature singular terms. Now, it is not necessary to assume that Russell's theory is the correct account of reference. Nor would I assume that Russell's theory is the only correct account of definite descriptions. Yet a limited part of Russell's theory seems to me to be correct, namely, that part which breaks definite description statements into more manageable chunks. This raises the question, "What sort of analysis of definite descriptions would Quine favour if not Russell's?" It should be pointed out that Quine endorsed Russell's theory in "On What There Is." I think that Russell's analysis of descriptions is acceptable so long as one does not confuse this account with the one given 63 by Russell of reference. There are powerful arguments against accepting the so-called Theory of Indirect Reference. Russell's view that proper names are in fact definite descriptions is, I think, wrong. Yet his contention that definite descriptions have certain logical features seems tome to be unassailable. The result obtained above implies that either Russellian analyses such as the one just given fail or Quine's account of semantic ascent is faulty. Somewhat analogous results can be obtained if Russell's theory of definite descriptions is applied directly to (9***). We obtain the following: (9****) 'Exactly one thing wrote Hamlet, and whatever wrote Hamlet wrote Hamlet' is not analytic. Since this sentence is an application of Quine's method of semantic ascent, there is no room for a de re interpretation of (9***). Quine is committed, I think, to the view that any proposed instance of the relational idiom is either (i) in fact an instance of the notional idiom or (ii) that the proposed sentence is false; or (iii) that the proposed sentence is meaningless. Since the second conjunct of the interior sentence of (9****) is analytic, Quine would have to hold that it is the first conjunct of the interior sentence which makes (9****) not analytic. In addition to the strong intuition that (9****) does not express the same meaning as is expressed by (9), it is here argued that holding, as Quine must, that the embedded sentence of (9****) is not analytic, will involve 64 his being undiscriminating in a highly unacceptable way. To show that Quine would be guilty of being undiscriminating if he were to champion this particular analysis of (9), a similar modal candidate may be studied. Consider once more the following sentence: (2) The author of Hamlet might not have written Macbeth. Quine's account of semantic ascent allows us to obtain the following: (2***) 'The author of Hamlet wrote Macbeth' is not analytic. Applying a combination of Quine's account of semantic ascent and Russell's Theory of Definite Descriptions to (2***) would yield the following: (2****) 'Exactly one thing wrote Hamlet and whatever wrote Hamlet wrote Macbeth" is not analytic. In this case, Quine would hold that the interior sentence of (2****) is not analytic. His reason would not be because the first conjunct of the interior sentence could be false, but because the second conjunct could be false. This position is a perfectly general one for an adherent of Quine's semantic ascent. This is a difficult point to follow, and therefore it would be helpful to consider yet another example: (20) The person delivering pizza possibly 65 smashed the delivery truck. Applying Quine's account of semantic ascent would yield (20*) 'Exactly one person delivered pizza and whoever delivered pizza did not smash the delivery truck1 is not analytic. Symbolically, (20*) becomes (20**) ' (3x) [(FxA (Vy) (Py -* x = y) ) A -Sx] ' is not analytic. In this case, Quine's account of semantic ascent would have us conclude that the interior sentence is not analytic because the negation sign (-i) attaches to the subformula "Sx" and not to the entire sentence. The non-analyticity of (20) has nothing to do with the nonexistence of the person delivering the pizzas; the germane predicate is 'not smashed'. His nonexistence is a point that does not even arise; it seems absurd to suggest otherwise. Whenever a modal sentence is interpreted by Quine's method of semantic ascent that contains a definite description, a further predicate, and a negation sign, the negation sign would not prefix the definite description. Rather, the negation sign would prefix the other predicate, as in our examples. If one abandons this general position and gives no independent explanation for it (that is, no reason other than that one thereby avoids contradiction), then one is being unacceptably undiscriminating. And indeed one 66 finds no such independent reason offered by Quine. The nonanalyticity of Quine's analysis of (9) as compared with his analysis of other modal sentences that feature definite descriptions arises in strikingly different ways . This result should be expected, given that Quine never was inclined to consider even the possibility that there were meaningful instances of the relational idiom in modal contexts. Simply labelling, as Quine would be compelled to do, (9***) as "not analytic" just is not sufficiently specific, since, on a Russellian treatment of the definite descriptions this label covers too many bases. I have finished arguing for the failure of semantic ascent to analyze correctly all de re modal sentences. It is worth pointing out that whereas in 1956 Quine countenanced the meaningfulness of de re belief ascriptions, he did not use the method of semantic ascent in an attempt to analyze them. The reason for this is quite simple: certain de re sentences, in either modal or doxastic contexts, cannot be subjected to an analysis such as the one given by an application of semantic ascent. § 8. Quine's Position in "TGMI" "There are," Quine writes in "TGMI," "three different degrees to which we may allow our logic, or semantics, to embrace the idea of necessity."J0 These degrees begin with a very parsimonious account of modality, where "necessity is expressed by a semantical predicate attributable to statements as notational forms hence 67 on attachable to names of statements." ' Quine notes that this account of modality is reconcilable with the policy of "admitting statements within statements [that are-M. W. D. ] truth-functionally only (apart of course from such contexts as quotation, which are referentially opaque)." This policy is referred to by Quine as a one of extensionality. There is a strong connection between Quine's method of semantic ascent and his first degree of modal involvement in "TGMI." They are identical. The second degree concerns an account of necessity that features a statement operator 'nee'. Quine holds that this degree is "a premeditated departure from extensionality." Quine also writes that Truth-functional occurrence is by no means the rule in ordinary language, as witness occurrences of statements governed by 'because', 'thinks that', 'wishes that', etc., as well as 'necessarily'. Modal logicians, adopting 'nee', have seen no reason that an adequate logic might adhere to a policy of extensionality. But Quine argues that in fact this statement operator is referentially opaque and hence that this approach is not significantly different from the approach involving appeal to contexts of quotation. The third degree of modal involvement is the quantification into modal contexts and is characterized by Quine as "necessity as a sentence operator."" This final degree, the most drastic that Quine considers, forces a reversion to the metaphysical doctrine of essentialism. Quine's views regarding essentialism 68 are critically evaluated in Chapter IV. This third degree of modal involvement is resurrected by Quine in "IR." This is further discussed in Chapter VIII. The third degree of modal involvement has thus been a persistent feature of Quine's critique of the relational idiom in modal contexts. In 1962, Quine revealed another long standing objection to modal logic, one which was not specifically directed at instances of the relational idiom. In this sense, the charge bears more similarity to Quine's three primary objections to QML in "R&M" than his later criticisms of the relational idiom in modal contexts. Some discussion concerning quantified modal logic is required in order to appreciate some of the subtleties connected with this objection. Earlier in this century, C.I. Lewis categorized different systems of propositional modal logic. The most important of these systems, and the one which has dominated discussion, was the system Lewis called S5. S5 differs from its logical brethren chiefly in the level of complexity, with the other systems being progressively more and more simple than S5. (They are "more simple" in the sense that the axioms of these other systems are also contained in S5 but not vice versa. ) Lewis made heavy use of the notion of strict implication. Crucially, Lewis' systems did not make mention of variables and quantifiers, which are some of the subjects of quantification theory. By the time that Lewis had devised these systems, quantification theory had been fully developed. It was thus inevitable that logicians would seek to combine the two types of 69 logic. Even though these sentences combine modal operators with quantifiers, the particular difficulties of merging quantifier theory with modal propositional logic are not yet clear. The particular difficulties that I have in mind concern the difference using a sentence (or formula) and mentioning one. There is a well-understood distinction between object language and metalanguage. To put the matter briefly, C. I. Lewis did not comprehend this distinction when he discussed modal propositional logic; his notion of strict implication is arguably both a syntactic and a semantic notion. On the one hand, strict implication is similar to the syntactic notion of material conditional, while, on the other, it is similar to the metalinguistic notion of semantic implicature. Quine put this matter in its proper historical context in his "Reply to Professor Marcus": [C. I., not David] Lewis founded modern [propositional only] modal logic, but Russell provoked him to it. For whereas there is much to be said for the material conditional as a version of 'if-then' , there is nothing to be said for it as a version of 'implies'; and Russell called it implication, thus apparently leaving no place open for genuine deductions between sentences. Lewis moved to save the connections. But his way was not, as one could have wished, to sort out Russell's confusion of ' implies' with ' if-then' ' Instead, preserving that confusion, he propounded a strict conditional and called it implication. Quine was thus correct when he held that modal logic was conceived in the sin of confusing use and mention, even if the original sin, in this case, was committed by Russell and Whitehead." This difficulty was, to some extent, extended by Barcan when she began the task of merging quantification theory with Lewis' modal propositional logic. In "TGMI," Quine does point out that 70 the use-mention confusion will occur when one uses 'nee' as a statement operator when, more properly, one should use 'Nee' as a semantical predicate. However, the engendering of this particular illusion can easily be avoided if use and mention are scrupulously kept apart. Quine also recognized this point, at least by 1962. Thus, he writes that it is logically possible to like modal logic without confusing use and mention. You could like it because, apparently at least, you can quantify into a modal context by a quantifier outside the modal context, whereas you obviously cannot coherently quantify into a mentioned sentence from outside the mention of it. Still, man is a sense-making animal, and as such he draws little comfort from quantifying into modal contexts that he does not think he understands. Applying these remarks to my two interpretations of my previous examples of ambiguous sentences such as (5), I think Quine would say, one can ascertain that any de dicto construal is absolutely free of use-mention errors, but any de re construal can be ridden with such errors. The main points of this chapter are primarily concerned with either (a) Quine's negative account of QML or ( b) Quine's positive account of modality, and can be summarized as follows: (1) his objections to QML are threefold and do not appear to have altered with time; (2) Quine*s initial objection to QML, the problem of (SI), can be seen as providing a solution to Frege's antimony of names; (3) Quine attacks the notion that quantification into modal contexts is meaningful with both the problems of (EG) and (US); (4) these latter two objections to QML serve double duty as the first two criticisms of the relational idiom in modal contexts 71 that appeared in Quine' s writings; (5) all three of the objections to QML are, if sound, also objections to a combination of doxastic sentences and quantification theory; (6) Quine's positive account of modality, his method of semantic ascent is viable only if philosophers confine themselves to de die to modality; (7) the method of semantic ascent is faulted for being insufficiently discriminating as to the reasons it gives for saying that this or that modal sentence is not analytic; (8) Quine provided an analysis that incorporated three different levels of modal involvement, thereby combining both his negative accounts of QML and his positive account of modality; and (9) in 1962, Quine made clear his view that modal logic was closely linked to confusions of use and mention of given modal sentences. 72 Chapter III. Quine on Employment of the Relational Idiom as a Response to his Three Initial Objections to QML § 9. General Remarks: The Two Versions of the Relational Idiom What I believe to be the basic flaw with Quine's three initial objections to QML will be the principal subject of this chapter. In the previous chapter, I urged that if Quine uniformly interpreted the premises and the conclusions of these objections using the relational idiom, then these objections did not appear to be sound. This point has been, at best, only cursorily established thus far. It will be expanded in the current chapter through a careful analysis of the two versions of the relational idiom in modal contexts that were in turn provoked by Quine' s initial objections. This discussion is contained in § 10 and § 11. Quine rejects both of the versions of the relational idiom in modal contexts that were given as replies to his three initial objections to QML. I agree with him regarding the first of these rejections but disagree with him regarding the second. This discussion is followed by an analysis of a fallacy that I believe is inherent in all three of Quine's objections. I shall term this error the 'relational fallacy1. It involves exploiting the ambiguities in modal and, occasionally doxastic, sentences in order to obtain certain desired results. The results that Quine wishes to obtain involve the three objections reviewed in the previous chapter. I shall attempt to 73 show that Quine' s initial objections to QML in "R&M" do not follow if one remains strictly faithful to a relational reading of the ambiguous sentences. My account of the relational fallacy is followed by a discussion of David Kaplan's appraisal of the first two of Quine's three objections. Kaplan rejects Quine's view that quantification into modal contexts is meaningless; Kaplan accuses Quine of invoking an invalid argument. This in turn is followed by a discussion of Quine's 'inconsistent skepticism' towards modal versus doxastic contexts. There, I urge that there is a structural parallel between some of Quine's modal objections and a doxastic puzzle put forward by Bertrand Russell in "On Denoting." § 10. Quine on the Solution of Church and Carnap: Three Arguments The first three objections to QML mentioned in the previous chapter are provocative with respect to instances of the relational idiom in modal contexts in that responses to them force one to adopt instances of the relational idiom as an answer to the objections. This, at any rate, fairly characterizes the first two responses to Quine's initial objections to QML. Intensional languages that can accommodate the relational idiom have been worked out in some detail in the recent past, beginning during World War II. Church and Carnap were the first to concentrate on the problems of merging modal logic with quantification theory. Even though 74 they recognized the need for, and some of the advantages of, the relational idiom, their efforts to create a satisfactory approach to QML ultimately became mired in certain problems raised by Quine. Carnap and Church both presented the necessity operator as being somewhat interchangeable with analyticity. (Strictly speaking, the relation between analyticity and the necessity operator is not one of equivalence. While necessity, when viewed as an operator, is just an operator taking a sentence into a sentence, analyticity is a predicate of sentences.) It is not surprising, in light of Quine's critique of analyticity, that Quine would be sceptical of their arguments. As we shall see, however, Quine has two distinct arguments directed at Carnap and Church. I shall argue that it is Quine's subsequent argument that is fundamentally damaging to the modified solution given by Church and Carnap. Church was the first to respond to Quine's "NE&N" with the view that quantifying into modal contexts was legitimate, provided that the variables being quantified take only intensional values. I shall call this "the Church/Carnap solution." This was the first solution to be specifically given as a response to Quine's earlier criticisms. It relied on the relational idiom in modal contexts. Quine first criticized the work of Church and Carnap on intensional languages in "The Problem of Interpreting Modal Logic" [1947]. Quine notes that Carnap "adopted it in an extreme form, limiting the range of his variables to intensional objects throughout his system." This was the system that Carnap developed in Meaning and Necessity. It was Quine's objection to Carnap's account, both 75 in 1943 and in 1945, as I shall show, which constitutes his first, and indecisive, argument against the Church/Carnap solution. This solution, as Quine sees it in "R&M," is to "repudiate all such objects [objects such as a number x that may be uniquely determined by two conditions that are not analytically equivalent to each other-M. W. D.] and retain only objects x such that any two conditions uniquely determining x are analytically equivalent." This solution is designed to get around the results of Quine's account of the problem based on the phenomenon of (EG). Church's response to Quine's criticisms is twofold: the first centres on the ambiguity in modal statements that feature definite descriptions; the second concentrates on the conclusion that Quine draws to the effect no variable within a non-extensional context can refer to a quantifier previous to that context. Regarding the former criticism, Church points out that substitutivity does indeed hold if the statement is given a de re interpretation. Church made his initial point in reaction to the paper on Whitehead, some time before "NE&N" made its appearance. Church's own solution does allow for the relational sense, but only by explaining modalities purely in terms of language. The ambiguity alluded to by Church becomes much more telling when it is conjoined with the observation that a fallacy of exploiting this ambiguity has taken place. This is what I shall term 'the relational fallacy1: that Quine switches from notional to relational readings in "R&M" as well as in other papers. The relational fallacy is explained below. (See § 12.) 76 § 10.1 Quine's First Argument [1943] Quine's rejection of the Church/Carnap solution constituted a reversal of views, albeit not a very sharp one, on Quine's part. The following is taken from one of Quine's letters to Carnap, from 1943: I see no fallacy in your solution, and would merely set the advantages over against the price: adoption of an ontology of intensions to the exclusion of individuals, classes, etc. (except as "manners of speaking" introduced by contextual definition). The following presents itself as a plausible principle: modal contexts (of the contemplated kind) can be permitted (in the usual unrestricted way) only in a language which countenances known individuals or other extensions as values of its variables. Apparently, then, Quine believed as early as 1943 that some of his objections to combining modal logic with quantification theory could be overcome if the relational idiom in modal contexts, or rather some version of it, was brought into play. This passage makes clear that Quine found no fallacy in the view later termed by Carnap to be an "erroneous conception," namely, that one will be restricted, if this solution is adopted, to a universe exclusively composed of intensional entities. However, the passage also makes clear that Quine desired a language with a semantics that would admit only extensions, not intensions. Carnap, in Meaning and Necessity, has an extensive review of Quine's position on modality that takes exception to Quine's point concerning the adoption of an ontology that is exclusively composed of intensional entities. He argues that "whatever is said in these languages about individual concepts can be paraphrased 77 in terms of individuals." Carnap's main point (he calls it "decisive") is that there is no objection against regarding designators in a modal language as names of intensions and regarding variables as having intensions as values, provided we are not misled by this formulation into the erroneous conception that the extensions have disappeared from the universe of discourse of the language. Is Carnap correct to view Quine's charge as embodying an erroneous conception? His argument in Meaning and Necessi ty strikes me as more of a conclusion that Carnap has not argued for, something of a Fait accompli, rather than substantial reasoning. Why is it an erroneous conception on Quine's part that, if Carnap*s answer were adopted, then the universe will be depopulated of extensional entities? Carnap gives only these, rather unconvincing premises to support his conclusion: ...it is not possible for a predicator in an interpreted language to possess only an extension and not an intension or, in customary terms, to refer only to a class and not to a property. Similarly it is impossible for a variable to be merely a class variable and not also a property variable. This passage appears to assume that the universe contains both extensional as well as intensional entities. But such an assumption constitutes mere question-begging on Carnap's part; he has to argue, and not assume, that the universe contains both intensional and extensional entities. At any rate, the issue between Carnap and Quine has a distinctly unresolved ambience to it. Yet, as I shall urge, the decisiveness of Quine's arguments against the Church/Carnap increases with time. 78 § 10.2 Quine's Second Argument [1953] It is interesting to point out that the first two (of a total of three) versions of "R&M" differ with respect to the criticisms of the Church/Carnap solution that are endorsed by Quine. These first two versions of "R&M" constitute Quine's second and third arguments against the Church/Carnap solution. In the initial version, published in 1953, Quine had this to say about the Church/Carnap solution's reliance on instances of the relational idiom in modal contexts: Observe now the extravagant price of thus purifying the universe. Concrete objects are banished in favor of what Frege called senses of names, and Carnap and Church have called individual concepts. Numbers are banished in favor of some sort of concepts which are related to the numbers in a many-one way. Classes drop out in favor of class-concepts or attributes, it being understood that two open sentences which determine the same class still determine distinct attributes unless they are analytically equivalent. Unrestricted quantification into modal sentences has been bought at the price of adopting an ontology of exclusively intensional or idealistic type. It is by no means the most glaring evil of such an ontology that the principle of individuation of its entities rests invariably on the putative notion of synonymy or analyticity. In short, Quine in 1953 had two complaints regarding the Church/Carnap solution: firstly, that the solution is metaphysically unsatisfactory, in that it effectively bans extensional entities; secondly, it is indirectly defective, in that it invokes the notion of analyticity, a concept that Quine subjected to the charge of unintelligibility in "Two Dogmas of Empiricism." Note that the first charge subtly presupposes a principle that is a reversal of the spirit of Occam's Razor, a principle that can be explained 79 as "Never postulate too few entities where evidence requires that there be more." For Quine's first complaint in the 1953 version of "R&M," which initially occurred some ten years earlier, is that this reversal of the spirit of Occam's Razor is being violated. § 10.3 Quine's Third Argument [1961] In 1961, however, the second edition of FLPV made its appearance. This allowed Quine to make certain changes to his earlier papers. One of these changes concerns the Church/Carnap solution. Here, then, is Quine's 1961 refutation of the Church/Carnap proposal: to state the matter without recourse to singular terms, it is that the requirement...'any two conditions uniquely determining x are analytically equivalent'-is not assured merely by taking x as an intensional object. For think of Fx as any condition uniquely determining x, and think of ' ph as any nonanalytic truth. Then ' p . Fx uniquely determines x but is not analytically equivalent to 'Fx" , even though x be an intensional object. So the claim that analytically equivalent conditions may be used to repudiate intensional concepts in the Church/Carnap account simply does not do the job. Contingently true sentences may always be conjoined that will render the solution worse than useless. Any attempt to ban the conjunction in question would be arbitrary and I cannot think of a good independent reason that such conjunctions would not be allowed. Quine is thus correct to hold that "nothing is gained by limiting the universe to intensions." Why did Quine offer this third argument against the 80 Ghurch/Carnap solution in 1961 (in the second edition of FLPV) whereas before he was content with only two? Quine correctly saw that his third argument was more decisive than the original two, which still make an appearance in the revised form of "R&M." Thus he writes that "we can quickly see that the expedient of limiting the values of variables to them [that is, to individual concepts-M. W. D.] is after all a mistaken one." It is at this point that Quine introduces the argument given above. To gauge the decisiveness of Quine' s argument, consider some condition F such that F uniquely determines some individual concept x. Further assume that there is another condition G such that G also uniquely determines x, and that ' Fx' is analytically equivalent to ' Gx" . Quine's argument shows that by simply conjoining an arbitrarily chosen contingent truth ' p' , it will still be possible to uniquely determine x. Clearly, 'Fx A p" is not analytically equivalent to ' Gx* , because ' p" is not an analytic truth. So the Church/Carnap solution is untenable. In addition, Quine's third argument against the Church/Carnap solution is absolutely independent of any theses that may require argumentation (for example, the above-mentioned reversal of the spirit of Occam's Razor and his concerns over the invocation of analyticity). I imagine that Quine's first and second argument against the Church/Carnap solution would not convince philosophers who do not share Quine's concerns about the analytic/synthetic distinction. Quine's third argument avoids this problem. The main result of Quine' s argument against the Church/Carnap 81 solution that Quine draws is that the only course open to the champion of quantified modal logic is to meet my strictures head on: to argue in the case of 9 and the number of planets that this number is, of itself and independently of mode of specification, something that necessarily, not contingently, exceeds 7.... This is how essentialism comes in: the invidious distinction between some traits of an object as essential to it (by whatever name) and other traits of it as accidental. I do not say that such essentialism, however uncongenial to me, should be uncongenial to the champion of quantified modal logic. On the contrary, it should be every bit as congenial as quantified modal logic itself. § 11. Quine on Smullyan's Use of Russell's Theory of Definite Descriptions In Conjunction with Modal Logic As noted, Quine held in "R&M," that there were two possible responses to the phenomena (that is, the phenomenon of (EG) and that of (US)) that he drew attention to. After dismissing the Church/Carnap solution, Quine concentrates on the account given by Smullyan in "Modality and Description." Smullyan developed his solution based on a suggestion of Church. Smullyan was the first to give a logically satisfactory rationale for the de re/de dicto distinction. Thus, Smullyan writes that the difficulty [alluded to by Quine in "NE&N"-M. W. D.] is obviated if we draw a distinction. We must distinguish between statements of the following forms: (d) The so-and-so satisfies the condition that it is necessary that Fx, and (e) It is necessary that the so-and-so satisfies the condition that Fx. J Smullyan's (d) corresponds to Forbes' definition of a de re modal 82 sentence and (e) likewise corresponds to his definition of a de dicto modal sentence. Smullyan realizes that the distinction that he is drawing attention to seems capricious and predicts that the "reader at this stage is bound to feel as though he were being i f\ asked to distinguish between Tweedledum and Tweedledee. " In fact, however, the de re/de dicto distinction is a very real one. This distinction is significant for calling into question Quine's argument in "The Problem of Interpreting Modal Logic." This argument is a direct ancestor of the one in "R&M" which, again, proceeds as follows; (3) 9 is necessarily greater than 7. (4) The number of planets = 9. Yet the sentence (5) The number of planets is necessarily greater than 7, seems to be false. Yet (5) appears to have been legitimately obtained from (3) and (A) by application of the law of (SI). If Smullyan is correct, then it is entirely likely that (5) is ambiguous and can be symbolized, using Russell's TDD, in either one of the two following ways: (5*) • (3x) KNPx A (Vy) (NPy -+ x = y) ) A (x > 7) ] , and (5 * ) (3x) [ ( NPx A (Vy) ( NPy - x = y ) ) A D ( x > 7 ) ] . 83 The reason the distinction between these two symbolizations is so important is that the invalid sequence that Quine is alluding to does not arise if (5) is given a de re construal, as in (5*). For this particular instance of the relational idiom in modal contexts is true (as opposed to the falsity of the de dicto version (5*). So the oddity never arises. (Notice that (5*) does not follow by (SI) from (3) and (4) and that (5') does, again using (SI). However, since (5') is true, there is no fallacy.) Smullyan does not actually mention the example that Quine uses (that is, (3), (4)-(5)). Nor does Smullyan actually give the two relevant construals of (5) (that is, (5*) and (5')). Instead, as in the previous citation from "Modality and Description," he uses ' Fx" as any predicate and thereby makes valuable points concerning the form of Quine's argument. Smullyan shows that there is no oddity if the ambiguous modal statement is given a de re interpretation in the following: ... the valid argument-form is rather N(Fy) y = (u) (4>x) :. [ (i x) (4>x)] . N(F(u) ( 4>X) ) . For the second premise of this argument is, by definition, equivalent to (3x) (<pz =z z = x:y= x) , which, in conjunction with the first premise, yields (3x)((f>z =z z = x:y = x.H(Fy)). This, by Leibniz's Law gives (3x)(<f>z =z z = x.N(Fx)) 84 which is the same proposition as [(u)(^)].N(F(lX)(^)).18 Smullyan does not make specific points regarding the objects that are being referred to by the definite descriptions present in the ambiguous sentences. Rather, he goes on to combine Russell's TDD with the class abstract approach adopted by Quine in "New Foundations for Mathematical Logic." Smullyan never makes the simple point that instances of the relational idiom in modal contexts ascribe metaphysical characteristics to objects. How does Quine receive this line of objection? The following passage substantially contains Quine's position on Smullyan's solution to Quine's objections to QML: ...Smullyan took the alternative course of challenging my criticism itself. His argument depends on positing a fundamental division of names into proper names and (overt or covert) descriptions, such that proper names which name the same object are always synonymous...He observes quite rightly on these assumptions, that any examples which, like (15)-(20) and (24)-(25), show failure of substitutivity of identity in modal contexts, must exploit some descriptions rather than just proper names. Then he undertakes to adjust matters by propounding, in connection with modal contexts, an alteration of Russell's familiar logic of descriptions. As stressed..., however, referential opacity remains to be reckoned with even when descriptions and other singular terms are eliminated altogether. [Quine's (15) is my (9) 9 is necessarily greater than 7, his (20) is my (21) 9 is possibly less than 7, his (24) is my (4) the number of planets = 9, 85 and his (25) is my (21) the Morning Star = the Evening Star. In short, Quine still believes that the phenomenon of (EG) shows that there will be problems with opacity even when singular terms have been eliminated. As was briefly (and, inconclusively) argued in the previous chapter, Quine has committed, in the course of this argument, a fallacy that I will term in § 12 'the relational fallacy". Without seriously anticipating this argument, let me merely contend that Quine's use of the argument based on the phenomenon of (EG) is as unsound the second time that it is employed as it was in its first use. In brief, I shall urge that Quine's argument against Smullyan's solution (to the three initial charges levelled at QML) relies heavily upon an assumption of some of the characteristics of de dicto modality when more properly he ought to subject the characteristics of de re modality to scrutiny. Quine views Smullyan's position on the relational idiom to be the lesser of two evils, where the Ghurch/Carnap solution is the other evil. Thus he writes, following his argument against the latter solution, that . . . the only hope of sustaining quantified modal logic lies in adopting a course that resembles Smullyan's, rather than Church and Carnap, in this way: it must overrule my objection. It must consist in arguing or deciding that quantification into modal contexts makes sense even though any value of the variable of such a quantification be determinable by conditions that are not analytically equivalent to each other. The only hope lies in accepting the situation illustrated by (32) and (33) and insisting, despite it, that the object x in question is necessarily greater than 7. This means adopting an invidious attitude toward certain ways of uniquely specifying x, for 86 example (33), and favoring other ways, for example (32), as somehow better revealing the 'essence' of the object. Consequences of (32) can, from such point of view, be looked upon as necessarily true of the object which is 9 (and is the number of the planets), while some consequences of (33) are rated still as only contingently true of that object. Quine's conviction that the separation of names from definite descriptions, when applied to modal discourse, leads to essentialism of the invidious kind is taken up in the next chapter, as are Quine' s three arguments against essentialism. For now, let me simply show the benefits of this separation in conjunction with instances of the relational idiom in modal contexts. These benefits convince me that Smullyan's approach is the correct one. There are intuitive reasons, I first contended last chapter, for holding that the relational idiom is intuitively meaningful. Let me now argue further for this contention. I am critical of the lack of argumentation on Quine's part for making precisely the opposite assumption, namely, the assumption that instances of the relational idiom in modal contexts are not meaningful. A good place to begin arguing that such instances are meaningful is to appeal to some examples. Russell's TDD will allow us to highlight a variety of scope ambiguities as well as to assist us in piquing our intuitions concerning what I allege to be the meaningfulness of such instances of the relational idiom. Consider again (2) The author of Hamlet might not have written Macbeth. This sentence is an ambiguous one in that it might mean, on the 87 one hand, that, possibly, the author of Hamlet did not write Macbeth. On the other hand, it might mean that there is (in fact) one, and only one author of Hamlet, and this individual possibly did not write Macbeth. The difference between these two interpretations may well appear to be a flight of fancy, and Smullyan's comment about distinguishing between Tweedledum and Tweedledee seems appropriate here as well. To see better the logical distinctions between the two, let us briefly consider some logical translations: (2*****) 0 [(3x) (Hx A (Vy) (Hy - x = y) ) A -*Mx] and ( 2 ' " ) (3x) [ ( Hx A (Vy) ( Hy - x = y) ) A 0 ^Mx] . The first is de dicto, and the second is de re. The two are very distinct logically as well as ontologically. On the logical front, one cannot even conclude from (2*****) that there is an author of Hamlet, whereas this conclusion does follow from (2 , , ,). On the ontological front, one must interpret, (assuming a possible worlds semantics) (2*''), but not (2*****), as selecting an individual from the actual world (in which an individual wrote Hamlet) , and then concluding that there is one world in which this individual did not write Macbeth. That is, there is a possible world in which the individual who wrote Hamlet (in the real world) did not write Macbeth in that world. Kripke's argument from Naming and Necessity regarding the 88 Quinean view that a property can not be held to be essential or accidental to an object independently of its mode of description is extremely important and relies on intuitions of ordinary persons : Suppose that someone said, pointing to Nixon, 'That's the guy who might have lost1. Someone else says 'Oh no, if you describe him as "Nixon", then he might have lost; but, of course describing him as the winner, then it is not true he might have lost'. Now which one is being the philosopher, here, the unintuitive man? It seems to me obviously to be the second. The second man has a philosophical theory. The first man would say, and with great conviction, 'Well, of course, the winner of the election might have been someone else. The actual winner, had the course of the campaign been different might have been the loser, and someone else the winner; or there might have been no election at all. So, such terms as "the winner" and "the loser" don't designate the same objects in all possible worlds. On the other hand, the term "Nixon" is just a name of this man . Kripke uses his argument to defend the notion of a name being a rigid designator (roughly, a name that designates the same thing in all possible worlds). Note, however, that precisely the same considerations (that is, the intuitive approach of the ordinary man) also count as reasons for holding that the following statement makes intuitive sense: (9) The author of Hamlet might have not written Hamlet. After all, (9) has the same form as Kripke's (22) The winner of the presidential election in 1968 might have lost the election. Thus, I would conclude that (9) is intuitively meaningful. Since (9) contains a definite description, we can apply Russell's TDD to provide an analysis. The same considerations that 89 were earlier applied to (2) do not apply to (9) as I shall now show. Suppose that a de dicto reading of (9) were given. It would look like this: (9*) 0 [(3x) ((Hx A (Vy) (Hy - x = y) ) A --Hx) ] But such a rendition is unsatisfactory, because it says that it is possible that something satisfies contradictory conditions; a patent falsity. Nor does the following de dicto construal of (9) appear to be satisfactory: (9**) 0 ->[(3x) ((Hx A (Vy) (Hy - x = y) ) A Hx) ] , since it does not convey the same meaning as (9). The only interpretation that captures the intuitive meaningfulness of (9) is, I submit, the following: (9') (3x) [(Hx A (Vy) (Hy •* x = y)) A O-^Hx] . This sentence is an instance of the relational idiom in modal contexts. Barcan Marcus has argued that, in 1980, Quine admitted that Smullyan's use of Russell's TDD was entirely legitimate. More precisely, she contends that "of course it was a mistake for Quine to claim that Smullyan had 'altered' Russell's logic of descriptions." If true, this position would constitute a major reversal on Quine's part, since in the first two editions of FLPV, he was slightly critical of Smullyan's solution. In both the 1953 90 and 1961 versions of "R&M," Quine rejects Smullyan's use of Russell's TDD because "Smullyan allows difference of scope to affect truth value even in cases where the description concerned succeeds in naming." It was for this reason that Quine charged that Smullyan had engaged in "an alteration of Russell's familiar logic of descriptions." Barcan Marcus has claimed that Quine was mistaken here in that Smullyan's treatment was in fact an exact employment as laid out in 'On Denoting' and exemplified there by Russell in his analysis of apparent substitution failures in contexts of epistemological attitudes. It is a central point of Russell's theory that in such contexts even singular descriptions which succeed in denoting one thing must be unpacked. She claims that Quine thereby "missed ... an innovative feature" of Russell's theory. Was Quine incorrect in his view that "change in the scope of a description was indifferent [for Russell-M. W. D.] to the truth value of any statement unless the description failed to name"?28 It certainly is true that the 1980 version of "R&M" gave Barcan Marcus the impression that Quine thought he was incorrect in 1953 and 1951. In the 1980 version of "R&M," Quine has deleted any reference to Smullyan's alteration of Russell's TDD. Now Quine claims that Smullyan has taken "a leaf from Russell.... scope is indifferent to extensional contexts. But it can still matter in mtensional ones." Barcan Marcus views these remarks as representing some sort of significant change on Quine's part: In summary, by 1980 it is finally seen that by fully employing the theory of descriptions and allowing for fixed reference of ordinary proper names the substitution failure is dispelled. But now Quine points out that the successful analysis places 91 a modal operator in the scope of a quantifier and in front of an open sentence which means 'adopting an invidious attitude towards certain ways of uniquely specifying [an object] x" and 'favoring other ways ... as somehow better revealing "the > in essence" of the object . Did Quine change his mind regarding Smullyan's "alteration" of Russell's TDD? I would argue that, contrary to Barcan Marcus' belief, Quine's third version of "R&M" shows no serious deviation from the earlier versions. This becomes clear once we consider a few modal sentences with definite descriptions and compare them to a few doxastic sentences with definite descriptions. Consider again (6) The number of planets is possibly less than 7. Since there is a number of planets, it is clear that the definite description in (6) has succeeded in naming. According to Russell's TDD, the truth-value of (6) ought not to change if the scope of the description is altered. Equally clear is that the truth-value of (6) will indeed be different depending on whether it is given a de dicto or de re interpretation. (6) is false when it is given a de re interpretation but true when given a de dicto interpretation. So Russell's point about truth-values not changing when the description fails to name is false when Russell's TDD is conjoined with modal logic, which is what Smullyan did. But Russell's point does hold true when TDD is applied to doxastic sentences. Consider again (l*'*) Ralph believes of the man in the brown 92 hat that he is a spy. Assuming that the definite description succeeds in naming, it would not matter if this sentence were given a de dicto or a de re interpretation; the truth-value would not change. So Quine was correct to say that Smullyan had altered Russell's TDD in both 1953 and 1961. Furthermore, I'm not convinced that Quine reversed himself on this issue, as Barcan Marcus appears to hold. For, as she herself notes, Quine still has reservations about the unequal truth-value outcomes of intensional (modal), as opposed to extensional (doxastic) contexts-which I take it, was the sole basis of his initial charge that Smullyan had changed Russell's TDD. Finally, note that Quine's charge that invidious ways of uniquely specifying an object x will have to be adopted if Smullyan's solution is adopted is hardly new to 1980. As we have seen, Quine has made this point as early as 1953. This leads me to suppose that Quine' s changes to the text of "R&M" in 1980 that Barcan Marcus mentioned are properly regarded as being stylistic changes, rather than fundamental revisions. § 12. The Relational Fallacy The objection based on the phenomenon of (EG), which is the second of Quine's criticisms of quantified modal logic in "R&M" involves a subtle mistake. I believe that there is a fallacy committed routinely by philosophers who share with Quine this 93 refusal to admit that there are genuine instances of the relational idiom in any or at least some contexts. This is a general fallacy at work which Quine, among other philosophers of language, has made that concerns a scope confusion with respect to the de re/de dicto distinction. The fallacy typically works in the following fashion: first, the significance of an instance of the relational idiom is called into question; second, the instance in question has its truth value evaluated by exploiting the ambiguity between de re and de dicto construals. This exploitation of the modal or doxastic sentence involves improperly ascertaining the truth or falsity of the de re version by appealing to the truth or falsity of the de dicto one. I shall call this fallacy the Relational Fallacy. Individual constants within a de dicto sentence are substituted with other terms, be these others individual constants or definite descriptions. This substitution plays a major role in the relational fallacy; substitution should only take place with de re (or relational) sentences. Substitution is valid for de re sentences but invalid for de dicto ones. "R&M" contains what I take to be several clear cases of the relational fallacy. They are, for the most part, associated with the problem based on the principle of (SI). It is illegitimate to substitute, Quine would hold, the singular term '9' with the coreferential singular term ' the number of planets' in our previous (6) The number of planets is possibly less than 7, because such substitution would yield the evident falsehood 94 (6') 9 is possibly less than 7. (I have already commented on Quine*s view that (5) is false. I am now trying to show that Quine has also exploited the ambiguity of other modal sentences in "R&M.") But (6) is true only if it is given a notional or de dicto reading. It is false if it is given a relational or de re meaning. Yet another example of Quine's committing the Relational Fallacy involves his view that (23) Necessarily if there is life on the Evening Star then there is life on the Evening Star, is true while (24) Necessarily if there is life on the Evening Star then there is life on the Morning Star, is false. An analysis similar to (5) is possible whereby a relational construal of (24) turns out to be true. In effect what Quine does when he argues against quantification into a modal context is to rely only on the de dicto reading. But to argue against quantification into a modal context, he ought to be relying on the de re reading. For the de dicto reading does not involve quantification past a modal operator, whereas the de re reading does involve such quantification. And then one quickly sees there is no problem with substitution at all, for the premises (that is, in this case, (23) and the identity statement (21)) are collectively true and so is the de re conclusion 95 (24') (3x) (necessarily if there is life on the Evening Star then there is life on x). There is another case in "R&M" that involves the relational fallacy. Recall that Quine considers (3') to sustain damage from the question, "What is this thing that is necessarily greater than 7?" If one answers that it is 9 that is necessarily greater than 7, Quine would simply substitute '9' with 'the number of planets', and thereby obtain what he considers to be a falsehood. There is at least one problem with this. Quine has in effect pointed out that the de dicto interpretation of (5) ((5*)) is false. Yet it appears that the objection that concerns the phenomenon of (EG) was supposed to apply specifically to relational constructions. The following citation from "R&M," which is taken directly after Quine introduces his objection to QML based on the phenomenon of (EG), supports this contention: Note that (30) and (31) [Quine's (30) is my (3') (3x) (x is necessarily greater than 7) while his (31) is my (24')-M. W. D.] are not to be confused with: Necessarily (3x) (x > 7), Necessarily (3x) (if there is life on the Evening Star then there is life on x), which present no problem of interpretation comparable to that presented by (30) and (31). The difference may be accentuated by a change of example: in a game of a type admitting of no tie it is necessary that someone of the players will win, but there is no one player of whom it may be said to be necessary that he will win. Thus it appears that Quine did not engage in some sort of conflation of de dicto construals with their de re counterparts. Rather than 96 conflating the de dicto and de re construals, he appears to be talking specifically about de re construals. But if this is indeed the case, that Quine specifically was objecting to a relational interpretation, why did Quine appeal to the truth-value of a notional interpretation? That is, if Quine specifically was objecting to a relational interpretation, then why did Quine claim that the de re sentence had the truth-value of the de dicto cognate? At first glance, this appears to be blatantly inconsistent. This problem for Quine's criticism of existential generalization appears to stem from Quine's inexplicable refusal to admit that there are genuine instances of the relational idiom in modal contexts. This inexplicability is compounded by his third objection to QML, a problem that also appears to betray a certain confusion of the de dicto/de re distinction. For when Quine holds that "Necessary greatemess than 7 makes no sense as applied to a number x . . . ," he appears not to be discussing the relational interpretation (that is, his (30) and (31) and my (3') and (24' ), respectively). The application of the modalities to numbers, as opposed to sentences, seems to me to be the defining feature of the relational interpretations. But the objection based on the phenomenon of (US) was evidently intended to be another way (in addition to the objection based on the phenomenon of (EG) ) to argue for the meaninglessness of relational constructions. So it does not seem fitting that Quine should rely on considerations specific to de dicto construals. Actually, there is less difficulty accepting this third 97 objection than there is the second. For there is no occurrence of the relational fallacy in Quine's objection based on the phenomenon of (US) , whereas there is a genuine problem with respect to the relational fallacy and the objection based on the phenomenon of (EG). § 13. Kaplan on Quine's Alleged Theorem Kaplan contends in "Opacity" that Quine gave an invalid argument in "NE&N" concerning the interpretation given to any language that merges quantification and intensional contexts. Kaplan terms this interpretation the "alleged theorem." Kaplan presents it thus: the theorem "says that in a sentence, if a given position, occupied by a singular term, is not open to substitution, then that position cannot be occupied by a variable bound to an initially placed quantifier." Quine's alleged theorem "imposes serious restrictions, commonly unheeded, upon the significant use of modal operators..." The following is Kaplan's reconstruction of the argument: Step 1: A purely designative occurrence of a singular term in a formula is one in which the singular term is used solely to designate the subject. [This is a definition.] Step 2: If an occurrence of a singular term in a formula is purely designative, then the truth value of the formula depends only on what the occurrence designates not on how it designates. [From 1.] Step 3: Variables are devices of pure reference; a bindable occurrence of a variable must be purely designative. [By standard semantics.] 98 Notation: Let <j> be a formula with a single occurrence of ' x' , and let <j>a, (j>(3, and <J>Y be the results of proper substitution of the singular terms a, (3, y for ' x1 . Step 4: If a and (3 designate the same thing, but <|>a and <|>|3 differ in truth value, then the indicated occurrences of a in <J>a and of p in <{>P are not purely designative. [From 2.] Now assume 5.1: a and (3 are co-designative singular terms, but (j)a and <|)(3 differ in truth value, and 5.2: y is a variable whose value is the object co-designated by a and (3. Step 6: Either <J>a and (j>y differ in truth value or <j>(3 and <J>Y differ in truth value. [From 5.1, since (j>a and <|>|3 differ.] Step 7: The indicated occurrence of y i-n <I>Y ^ s n o t purely designative. [From 5.2, 6, and 4.] Step 8: It is semantically incoherent to claim that the indicated occurrence of y i-n <|>Y ^ s bindable. [From 7 and 3.] Kaplan holds that a simple error in Quine's argument for the theorem has been at the root of several other errors by Quine. The error has occurred, according to Kaplan, at Step 4, which does not follow from Step 2. The most that can be concluded on the basis of Step 2 is that at least one of a in <J>a or p in <J>p is not purely designative. Unfortunately for Quine, he needs the result that both variables are not purely designative. Thus, Step 7, which is the conclusion of the alleged theorem, does not follow. However, Kaplan's point is altered by Quine; Quine changes the alleged theorem and its argument, as we will now see. In his reply to Kaplan in the Schilpp volume, Quine appears to hold that Kaplan's reconstruction of his argument is faulty with respect to Steps 2 and 4. To meet Kaplan's criticism, Quine changes these two steps. Here is Quine's own reconstruction: Step 2': If a position in a formula is purely referential, then the truth value of the formula depends only on what the 99 term in that position designates and not on how it designates. Step A' : If a and (3 designate the same thing, but $a and <|)P differ in truth value, then the position of a in <j>a (and of p in <|>p) is not purely referential. " Despite having the merit of updating the terminology from 'occurrence' to 'position' , it is not clear, at first glance, how Quine can infer his (somewhat modified) conclusion. It might appear, again, that the most Quine can conclude, this time on the basis of his updated premises, is that at least one of the positions (either a in (j>a or the position of p in (j>p) is not purely referential. Thus, the modified conclusion Step 8': It is semantically incoherent to claim that the position of y in <J>y is bindable, does not appear to follow. It might be thought that there is a difficulty with Kaplan's objection concerning the claim that one of a or p is not purely designative; it is difficult to see how one of them can not be while the other one is, given that a = p (that is, "a" and "p" are coreferential). However, they are different terms and the notion of being purely designative originally was applied to occurrences of terms and subsequently to positions in a formula-frame. Quine does dispute Kaplan's reconstruction, although he does not dismiss it for this reason. There is another difficulty concerning Quine's argument as reconstructed by Kaplan. It seems to me that <|>a and <|)P differ in truth value only if (ji is a de dicto construal of the modal sentence in question. If we grant the plausible assumption that <|> is an 100 instance of the relational idiom, then it simply does not follow that the two sentences differ in truth value. The assumption made at Step 5.1 just doesn't hold for de re sentences. So such cases are outside the applicability of the alleged theorem. Quine's response to Kaplan does indeed mark an advance over his previous position on pure designation. Quine's point would be even more convincing had he shown that the notion of a term's position can be defined as a syntactic entity. Patton has given his permission to quote his following comment in full: The position of a term in a sentence, like the scope of a quantifier, doesn't sound like a syntactic entity, but could be plausibly identified as one. For a quantifier: the formula it is prefixed to. For a term, its position in Y [where Y = _ t_ ( t being the term) ] could be defined as _x_ or as _o_, or the like: Y itself but with t removed and some dummy "position-filler" like ' x' or 'o' put in for it. Even without this "syntactizing" of the notion of a term's position, we could have this principle or axiom governing the notion: where <j>Y contains neither a nor (3, and <J>a is the result of putting tokens of a for all tokens of y in (j>y, and similarly for {j>p, a and (3 have the same position in <j>a and <|)p, respectively. Roughly, the idea is that you know the position of X in Y if you know what X is, what Y is, and where X goes in F-you can separate Y into (i) tokens of X, and (ii) a "frame" in which they go. So I think Quine can meet this charge of a fallacy. (But will his "modified conclusion" still be enough for his argument? Evidently he thinks so.) Two observations come to mind when we consider this comment: firstly, Quine did not, as Patton implies, explicitly state the "modified conclusion"-I did; secondly, it seems difficult to imagine that position could be anything but a syntactic notion. It may be the case that Quine has succeeded in defeating Kaplan's critique of his argument. It certainly is the case provided that position is a syntactic notion. (To make sense of Quine's 101 response to Kaplan, we must grant the assumption that, for Quine, position is a syntactical notion.) But do the results apply to a in <j>a if (J) is an instance of the relational idiom? Consider again (5) The number of planets is necessarily greater than 7. The various construals of (5) are, once more, as follows: (5**) It is necessary that: there is exactly one thing that is the number of planets and that thing is greater than 7. (This is the de dicto version of (5).) (5'') There is exactly one thing that is the number of planets and it is necessary that that thing is greater than 7. (This, again, is the de re version of (5).) It is not relevant to Quine' s theorem that (12^) [my (5*)-M. W. D.] is false and (12r) [my (5')] is true since neither contains a name corresponding to "the number of planets." Thus, (11) and (12) do not constitute a counter-example to the claim that the occurrence of " x" in "x is necessarily greater than 7" is bindable. Robinson's point is that the position of '9' in (3) 9 is necessarily greater than 7 is clearly different from the position of any expression in either (5**) or (5'*) -- which are the two possible analyses of (5) --and, hence, Quine's alleged theorem does not apply. 102 There is a further point to be made in connection with the alleged theorem. It involves Church's earlier point concerning the ambiguity inherent in modal expressions that feature definite descriptions. Church's point, along with that regarding the relational fallacy, constitutes an intuitive argument against Quine's objection based on the principle of (SI) and the one based on the operation of (EG). Before leaving the topic of Quine on position, let me reiterate that I think Quine's move to talk of position from his earlier talk of designation only partly saves his alleged theorem. The move to talk of position does address some of Kaplan's concerns. Calvin Normore has indicated that Kripke's view that one can always quantify into a position occupied by a proper name contradicts Quine contention that distinctions should be made between referentially opaque and referentially transparent positions of proper names in sentences. Normore is correct to point to this problem. I tend to favor Quine's stance over Kripke's, principally because I think that a de dicto doxastic sentence can contain a proper name. (This is why I doubt that Forbes' syntactic definition of the de re/de dicto distinction in modal contexts can be extended to doxastic contexts.) § 14. Quine's Inconsistent Skepticism: Structural Parallels Between Modal and Doxastic Phenomena Quine defends his alleged theorem in the following: 103 Purported proofs aside, is the alleged theorem true? It says, in a word, that a position that resists substitutivity of identity cannot meaningfully be quantified. Does this mean that no meaning can be assigned to such quantification? Surely no; meaningless expressions are precisely the ones that stand unencumbered and receptive to the sum of any meanings we like. What I was pointing out [in "NE&N"-M. W. D.] was the muddle that besets our original conception of quantification in such cases. Such quantifications do commonly serve good purposes and seem to make good sense until scrutinized in the light of what the quantifier is literally taken to mean. So in "Quantifiers and Propositional Attitudes", I explored paraphrases that might serve those same good purposes while not straining our conception of quantification. Yet this strikes me as being somewhat contradictory. After all, it would seem from this passage that Quine views his objections to QML in "NE&N" as being only secondary concerns with respect to the meaningfulness of a given sentence, even when that features a nonextensional context. This certainly is not the impression that one receives from an initial reading of the alleged theorem in "NE&N." It would appear from Quine's passage just cited that the question of whether a given sentence is meaningful or not is, as Quine would put it in "WA," free for the thinking up. Should not the meaningfulness of a given sentence in a nonextensional context depend upon whether one can apply the alleged theorem to it or not? As we have seen in the previous section, Quine applies his alleged theorem to certain modal sentences to show that they are meaningless. There are grounds, I urge, to hold that modal sentences ought to be treated in precisely the same manner as propositional attitude sentences. I believe that there are intuitive reasons for treating propositional attitude ascriptions as being roughly 104 on a semantic par with modal phenomena. By 'semantic par1 , I have in mind a very close connection between semantic phenomena which appear, at least initially, to be vastly different. Propositional attitude ascriptions are associated with construals of belief; on the other hand, modal phenomena concern necessity and possibility, two quite different things, one might think. Furthermore, the manner in which philosophers have traditionally accounted for these two sorts of phenomena leads one to the conclusion that the gap between them is one that cannot be bridged. Linsky is one philosopher who argues for this sort of account. He reports that "Possible-worlds semantics worked fine for necessity and possibility, I found, but not at all well for belief." Theorists of both the major schools of reference are united in their hostility towards granting equal status to modal and doxastic phenomena. However, despite the apparent differences between these two kinds of phenomena, they bear a striking number of resemblances. One resemblance between them can be seen by transforming Russell's doxastic puzzle into a modal one with the same structure and by transforming Quine's modal puzzle into a doxastic one with the same structure. Perhaps more than any other philosopher, Quine has given (strictly tacit) approval to this notion. As is argued in the discussion on "Intensions Revisited" (hereafter, "IR"), Quine comes very close to endorsing this view in his argument against Kaplan's vivid designators. That is, Quine appears to be on the point of arguing that if vivid designators, the kind of designator used in doxastic contexts, are allowed, so too must 105 rigid designators, the kind used in modal contexts, be allowed. Since rigid designators are, for Quine, unacceptable, so too are vivid designators. The only way this position could be considered to be viable is, I think, if it is conjoined with the assumption that modal and doxastic phenomena are on a semantic par, and have to be treated as one set of closely related phenomena. It would be consistent to reject rigid designation and accept vivid designation only if one were willing to deny that there is any close connection between the two sorts of puzzles. The significance of a semantic par between modal and doxastic phenomena would not be lost on contemporary philosophers of language. Many philosophers who would defend the Theory of Direct Reference (hereafter, TDR) want to draw a sharp line between modal and doxastic phenomena for the following (once again, implicit) reason: if such phenomena were on a sematicpar, then the relational idiom would be meaningful in both contexts. Yet if this were true, some version of the doctrine of essentialism would thereby have to be defended. Yet these same philosophers of language feel that there are sufficiently compelling negative arguments to dismiss essentialism. Before addressing some of these arguments against essentialism, allow me to spell out the intuitive arguments in favor of treating propositional attitude sentences and modal sentences as having great similarities. Consider firstly Russell's doxastic puzzle from "On Denoting": 106 (25) George IV wondered whether Sco t t was the au thor of Waverley. (26) Sco t t = the au thor of Waverley (Therefore) (27) George IV wondered whether Scott was Scott. Now consider once more Quine's modal puzzle: (3) 9 is necessarily greater than 7, (4) The number of planets = 9 (Therefore) (5) The number of planets is necessarily greater than 7. It is important to point out that I concur with Smullyan's suggestion that a relational construal of (5) will be the correct, and not absurd at all, conclusion to draw. I argue that because it is obvious that these modal and doxastic puzzles are structurally interchangeable, it is highly intuitive that there is a semantic par between modal and doxastic phenomena. Consider a modal version of Russell's puzzle about George IV: (28) It is possible that the number of planets is greater than 9. (4) The number of planets = 9 107 (Therefore) (29) It is possible that 9 is greater than 9. Consider now a doxastic version of Quine's puzzle about necessity: (30) Ronald Reagan wondered whether Mt. Everest is higher than Mt. Olympus. (31) Mt. Everest = the highest mountain in the world (Therefore) (32) Ronald Reagan wondered whether the highest mountain in the world is higher than Mt. Olympus. What, if anything, is new about the puzzles that I called 'the modal version of Russell's puzzle about George IV and 'the doxastic version of Quine's puzzle about necessity1? Strictly speaking, they are not new puzzles. Instead, they are new combinations of old puzzles . Russell' s original puzzle begins with a premise that features a definite description and a name. Its conclusion features two occurrences of the same name and no definite description. Quine's original puzzle begins with a premise that features two names. Its conclusion features one name and one definite description. In both cases, the second premise asserts the identity of a name and a definite description. Though obviously similar, structurally they are distinct. What the second generation of puzzles shows, I think, is that the original puzzles have further 108 applications than those envisioned by their authors. Russell's original puzzle poses a problem for modality as well as belief; so too does Quine's puzzle. It should be noted that Russell' s doxastic puzzle and Quine' s modal puzzle played fundamentally different roles in their respective philosophies of language. Russell viewed his doxastic puzzle as a problem to be solved. It was one of three main motivations for the TDD. That is, the problem posed by Russell's puzzle was one to be overcome. This was not the case with Quine's puzzle, where, as we have seen, the problem of (SI) posed by Quine's puzzle functions as a criticism of quantified modal logic. Inasmuch as I view Quine's puzzle as posing a problem that I think can be solved by introducing the relational idiom as a way of clarifying hidden scope distinctions, it is clear that I favor the approach adopted by Russell. Would Quine have viewed the Russellian version of the modal puzzle as also representing an insuperable criticism of quantified modal logic? Possibly he would. Yet that would also, arguably, mean invalidating a close kin of one of Russell's three motivations for the TDD, a theory that Quine has supported. Russell no doubt would see the puzzle about Mt. Everest as constituting a fourth reason to accept the Theory of Definite Descriptions. A logical theory may be tested by its capacity for dealing with puzzles, and it is a wholesome plan, in thinking about logic, to stock the mind with as many puzzles as possible, since these serve much the same purpose as is served by experiments in the physical sciences. The other three puzzles are also presented by Russell as puzzles 109 to be solved. I mentioned already the puzzle about the author of Waverley. The remaining two are concerned with (a) the law of excluded middle and (b) the apparent inconsistency of denying existence. Philosophers who hold that modal and doxastic phenomena are to be accounted for entirely through the use of independent treatments are committed to a counterintuitive position. For the construction of the modal and doxastic puzzles given above is by no means an enterprise of no consequence; it strongly indicates that these phenomena are at least two sides of the same coin. If this is true, then there is a presumption in favor of a unified treatment that would account for both sorts of phenomena. Russell's puzzle was first shown in a doxastic context; Quine's puzzle was first displayed in a modal context. It could have been otherwise, namely, both Russell's and Quine's puzzle could have been framed in reverse contexts. Then philosophers would have different, yet closely intertwined, phenomena to account for. So the two phenomena are indeed intrinsically linked. I would submit that all four puzzles-Quine's original modal puzzle and its doxastic counterpart and Russell's original doxastic puzzle and its modal counterpart-can best be addressed by invoking the distinction between the relational idiom and the notional idiom. Russell came upon this distinction in "On Denoting" with respect to his doxastic puzzle: To return to George IV. and Waverley, when we say, "George IV. wished to know whether Scott was the author of Waverley," we normally mean "George IV. wished to know whether one and 110 only one man wrote Waverley and Scott was that man" ; but we may also mean: "One and only one man wrote Waverley, and George IV. wished to know whether Scott was that man" . In the latter, "the author of Waverley" has a primary occurrence; in the former a secondary. 5 Russell's discovery of the relational idiom in doxastic contexts appears to have been largely forgotten until Quine made his rediscovery in "Q&PA." That Quine came upon the distinction between the relational idiom and the notional idiom in doxastic contexts is somewhat ironic. Given that, as was argued in this chapter, the relational idiom in modal contexts was used as a solution by Church, Carnap, and Smullyan, it would have been more natural for Quine to have uncovered the distinction in doxastic contexts as a solution to concerns about substitutivity. After all, Quine did point out in various papers before "Q&PA" that there were substitutional problems with doxastic contexts. His rediscovery of the distinction in doxastic contexts appears to have been more the result of a direct appraisal of ambiguity rather than (like Russell) a method of solving an unpalatable problem. In fact, as has been shown, the relational idiom in doxastic contexts poses, and does not assist in solving, problems for Quine. If instances of the relational idiom can indeed constitute a solution to the original Russellian and Quinean puzzles, then they also constitute a solution to the second generation of puzzles. The doxastic version of Quine's puzzle can be solved in a fashion that is analogous to the solution given by Smullyan to Quine's original puzzle that functioned as an objection to QML. That is, a relational construal of (32) is sufficient to deal with this Ill puzzle in precisely the same fashion that (5') was sufficient to deal withQuine's original puzzle. Russell's solution to his puzzle about George IV depends on a relational construal of his first premise ((25)). Similarly, the modal version of Russell's puzzle can be solved by a relational construal of (28): (33) There is one and only one number of planets and that number is possibly less than 9. (33) is false. Symbolically, (33' ) (3x) ( (NPx A (Vy) (NPy -* x = y) ) A 0(x < 9) ) . Use of Russell's TDD is one way (perhaps the only way) to proceed if one is going to argue that the number of planets, independently of the mode of specification, is necessarily greater than 7. That is, Russell's TDD would allow for scope distinctions that would separate relational from notional interpretations; the relational ones would allow for quantifying in. This, as we have seen, was Smullyan's proposal. It is one that is supported by the author. Quine has a case in which a singular term is treated as occurring either in non-referential position or as being in purely referential position. I quote the following passage from WO in full: If we understand the sentence: (2) The commissioner is looking for the chairman of the board 112 in such a way as to be prepared to affirm it and yet to deny: (3) The commissioner is looking for the dean even though, by recent appointment and unknown to the commissioner, (4) The dean = the chairman of the board, then we are treating the position to the right of 'looking for' as not purely referential. On the other hand if, aware of the commissioner's persistent avoidance of the dean, we are still constrained by (2) and (4) to treat (3) as true, then we are indeed treating the position as purely referential. This case is about the closest that Quine ever comes to recognizing the indispensability of the relational idiom in doxastic contexts, apart from "Q&PA." I would think that the same results could be achieved for what I interpret as the indispensability of the relational idiom in modal context. And Quine's case about the number of planets necessarily being greater than 7, can be pressed into service in order to fulfil this function. If we understand (3) in such a way as to affirm it and yet to deny (5), even though (4) is true, then we are treating the position to the left of 'necessarily is greater than 7' as not purely referential. If, on the other hand, we are compelled to treat all three of these latter sentences as true, (we might initially have regarded (5) as false) , then we are indeed treating the position as been purely referential. My own view is that we should view the de dicto sentence (5*) as false and the de re sentence (5') as true. We should also treat the position of ' x" in ' •( x > 7)]' in (5') as being purely referential. 113 The first through third objections in "R&M" have been found to be unsound. I have argued that there is a systematic ambiguity between the de dicto and the de re interpretations of the sentences contained in most of these objections. This systematic ambiguity is exploited by Quine in favor of de dicto construals. Such exploitation is consistent with Quine's theory of modality (his account of semantic ascent). If one adopts the account of semantic ascent, then one is a fortiori committed to de dicto construals of the relevant sentences. (The analytic/synthetic distinction, which forms the basis of Quine's account of semantic ascent, applies only to statements, not to objects; so the de re versions are automatically shut out.) Substitution and quantifying into a modal context are legitimate operations only in relational contexts; Quine incorrectly uses substitution and quantifying in as tests by appealing to the truth value of statements in notional, rather than relational, constructions. If Quine allows for the relational idiom, then there is no legitimate objection, since the sentence in question (that is, (5)) is true. This means that the argument from the premises (3) and (4) to the conclusion (5) is not invalid. However, and this is an instance of the relational fallacy, Quine exploits the ambiguity inherent in (5) in the way previously described. In brief, then, Quine*s initial objections to QML are unsound. Nevertheless, these criticisms have been treated by other philosophers as if they were sound. The main points of this chapter may be summarized as follows: (1) the three initial difficulties raised by Quine provoked two 114 sorts of solution, both of which allow for quantification past a modal operator; (2) one, the Church/Carnap solution was criticized for three different reasons by Quine; (3) Quine's final objection to the Church/Carnap solution is both convincing and sound; (4) the second solution to the three initial difficulties raised by Quine involves the use of Russell's TDD to distinguish de re from de dicto modal instances and is due to Smullyan; (5) I disagree with Quine's rejection of Smullyan's solution--the benefits, such as the explanation behind ambiguous modal sentences such as (2) as well as unambiguous ones, such as (9), are too large; (6) There is a fallacy which undermines his position on QML; (7) Kaplan's view that there is a simple logical error in Quine's Alleged Theorem, is very convincing; and (8) Quine is guilty of inconsistent skepticism with respect to his denial of the meaningfulness of the relational idiom in modal contexts but not in doxastic ones. 115 Chapter IV. Quine on Essentialism § 15. General Remarks This chapter is primarily concerned with Quine's account of the doctrine of essentialism. Before proceeding with this subject, however, it would be useful to undertake a short review of the main accomplishments up to this point. First, we provided, in Chapter I, a comprehensive overview of Quine's objections to QML as well as his criticisms of the relational idiom in modal and doxastic contexts. Second, Quine's three initial objections to QML were discussed in Chapter II. Third, I urged, in Chapter III, that two of these three objections to QML are fundamentally unsound. Fourth, it was also contended in Chapter III that both Russell's original doxastic puzzle and Quine's modal puzzle have applications further afield than those envisioned by their authors. Fifth, it was demonstrated, in Chapter III, that there have been two attempts to answer Quine's objections to QML that have invoked the relational idiom. I agreed with Quine that the Church/Carnap solution fails, but disagreed with him regarding the viability of Smullyan's solution which uses Russell's TDD. Quine's critique of Smullyan's solution is twofold: on the one hand, Quine argues that defenders of quantification into modal contexts are constrained to revert to what he calls Aristotelian essentialism; on the other hand, Quine has two arguments to the 116 effect that essentialism is an incoherent doctrine. Taken together, these two arguments against the coherency of essentialism constitute his fourth criticism of the relational idiom. His argument that quantifying into modal contexts requires a reversion to Aristotelian essentialism marks Quine's fifth criticism of the relational idiom. Consider the following remark by Barcan Marcus: It may be that what concerns Quine has finally little to do with interpretations of modal logic or use-mention confusions or senselessness. It may be that what lies behind his concern is that it presents one of the challenges to his attack on any analytic/synthetic distinction where even logical truths are included among analytic truths. Recall that on Quine's characterization of essential truths, predicates formed from valid sentences like being green, or not green or being self-identical are also to be included among the predicates which he takes to be invidious. Barcan Marcus' point here seems not to be stated strongly enough. That is, it is highly probable that Quine's attitude regarding the analytic/synthetic distinction was the chief cause of his reluctance to accept essentialism. To his credit, Quine recognizes that the problem of substitutivity of identicals would have no force if one adopts Aristotle's doctrine of essentialism. If Aristotle's doctrine were indeed correct, there would be no substantial objections to de re modality. There would thus be no reason to disallow relational interpretations of ambiguous modal sentences. Take, for example, my earlier sentence (2) The author of Hamlet might not have written Macbeth. As indicated before, the relational interpretation of (2) is (2') There is exactly one author of Hamlet, and that same 117 individual possibly did not write Macbeth. In symbols, (2') is ( 2 • " ) (Bx) [ (Hx A (Vy) (Hy - x = y) A 0 ->Mx] . If indeed some version of Aristotle' s doctrine were correct, there would be nothing wrong with an interpretation of (2) along the lines of (2') and (2'* » ) . Moreover, the same point holds even if the modal sentence in question is unambiguously relational. Consider again (9) The author of Hamlet might not have written Hamlet. I have urged that sentences of the form of (9) must be given a relational interpretation. (9) must be interpreted as (9'') Exactly one thing wrote Hamlet, and whatever wrote Hamlet might not have written Hamlet. In symbols, this is (9-) (3x) [(Hx A (Vy) (Hy - x = y) ) A O-Hx] . The type of essentialism that I am interested in features modal attributes that account for the relational interpretations of sentences such as (2) and (9). Let us begin by evaluating Quine's pair of arguments that are designed to show the incoherence of essentialism. I shall attempt to show that these arguments are unsound. 118 § 16. Quine's Two Arguments Against Essentialism [1953-1960] § 16.1 Quine's Argument Against Essentialism in FLPV and "TGMI" At the outset of FLPV, Quine formulates his criticism of essentialism. I think that it is worthwhile to see Quine's argument in its original form in 1953: The Aristotelian notion of essence was the forerunner, no doubt, of the modern notion of intension or meaning. For Aristotle it was essential in men to be rational, accidental to be two-legged. But there is an important difference between this attitude and the doctrine of meaning. From the latter point of view it may indeed be conceded (if only for the sake of argument) that rationality is involved in the meaning of the word 'man' while two-leggedness is not; but two-leggedness may at the same time be viewed as involved in the meaning ' biped1 while rationality is not. Thus from the point of view of the doctrine of meaning it makes no sense to say of the actual individual, who is at once a man and a biped, that his rationality is essential and his two-leggedness accidental or vice versa. Things had essences, for Aristotle but only linguistic forms have meanings. Meaning is what essence becomes when it is divorced from the object of reference and wedded to the word. Quine thus combines an argument against essentialism with one which shows the relevant difference between the doctrine of meaning and that of essentialism. Later in 1953, in "TGMI," Quine transformed this combination of arguments into a more precise criticism of essentialism. In "TGMI," Quine held that quantification into modal, and presumably doxastic, contexts was "not prima facie absurd if we accept some interference in the contextual definition of singular terms." Cross-quantification, the defining characteristic of 119 instances of the relational idiom is, however, still problematic for Quine because it forces an acceptance of Aristotle's doctrine of essentialism. Again, Quine provides no reason for the reader to think there is anything particularly Aristotelian about the type of essentialism that he is about to attack. Here is Quine's second argument against essentialism in full: There is yet a further consequence, and a particularly striking one: that of Aristotelian essentialism. This is the doctrine that some of the attributes of a thing (quite independently of the language in which the thing is referred to, if at all) may be essential to the thing, and others accidental. E.g., a man, or talking animal, or featherless biped (for they are in fact all the same things), is essentially rational and accidentally two-legged and talkative, not manly qua man but qua itself. More formally, what Aristotelian essentialism says is that you can have open sentences-which I shall represent here as ' Fx" and ' Gx" -such that (54) (3x)(nec Fx. Gx. -necGx) . An example of (54) related to the falsity of (53) [(53) nee (the number of planets = 9)-M. W. D] might be: (3x)[nec(x > 5). there are just x planets. ~nec(there are just x planets)], such an object x being the number (by whatever name) which is variously known as 9 and the number of planets. Clearly, the primary defect in several of Quine's other criticisms of the relational idiom (referred to before as a failure on Quine's part to distinguish between relational and notional constructions of given formulae) will not be sufficient here to discredit the objection. For Quine is assuming that the concrete individual in question is being featured in a referential (that is, a relational) sense. There still remains room for an ambiguity, however, and I shall explain this shortly. On one of these relational readings, the problem lies not in the failure to make 120 the relational/notional distinction, but, as we will see, rather (effectively) in a fallacy of the form (Vx) • (Fx -* Gx) T Fa T •Ga. F This form of inference is invalid in all standard systems of QML. § 16.2 Quine's Second Argument Against Essentialism: WO Quine attempts to "evoke the appropriate sense of bewilderment"5 when one is dealing with the distinction between necessary and contingent attributes. Quine borrows from the argument against essentialism that he had previously given in FLPV as well as in "TGMI. " Though similar to the argument given in the previous section, I think the arguments are sufficiently different to be considered distinct. Accordingly, I refer to this argument as 'Quine's second argument against essentialism1 . Defenders of this distinction may want to hold (34) All mathematicians are necessarily rational but are not necessarily (that is, are contingently) bipedal. They might also wish to hold that (35) All cyclists are necessarily bipedal but are not 121 necessarily (that is, are contingently) rational. But the following premise is true: (36) Someone is both a mathematician and a cyclist. Therefore, (C) Either this concrete individual is both necessarily rational and contingently bipedal or necessarily bipedal and contingently rational. [(G) is a close paraphrase of Quine's rhetorical question, "Is this concrete individual (the one who satisfies my (36)). (C) is supposed to follow from (34)-(36).-M. W. D.] necessarily rational and contingently two-legged or vice versa?" (WO, p. 199) What conclusions can be reached concerning this argument? Since (36) is true and (C) is false, it follows that either (34) or (35) is false or that there is some problem with quantification into modal contexts. Quine opts for this latter conclusion to his 'evocation of bewilderment' argument; Just insofar as we are talking referentially of the object, with no special bias toward a background grouping of mathematicians as against cyclists or vice versa, there is no semblance of sense in rating some of his attributes as necessary and others as contingent. Some of his attributes count as important and others as unimportant, yes; some as enduring and others as fleeting; but none as necessary or contingent. In other words, since there is no procedure which will decide between either of the disjuncts of (C), it is meaningless to use the distinction between necessary and contingent attributes. 122 There are alternative interpretations of (34) and (35); they are ambiguous. To show that Quine is relying upon the fallacy referred to in the previous chapter, premises (34), (35), (36), and (C) must first be symbolized: (34') (Vx) (D(Mx - Rx) A ->0{Mx - Bx) ) (35') (Vx) (Q(Cx - Bx) A -<•( Cx - Rx) ) (36' ) (3x) (Mx A Cx) and ( C ) (3x) [(DRx A iDBx) V (DSx A ^QRx) ] . ( C ) does not follow from (34*), (35'), and (36'). Barcan Marcus first pointed this out in "Modalities and Intensional Languages." So Quine's "bewilderment" does not arise. There is an even stronger conclusion that Quine could have made to be used in his "evocation of bewilderment" argument. This even stronger conclusion is not (G) or its symbolization ( C ) , but is instead the conjunction of (34) and (35). This conclusion is stronger in that it would constitute, or at least appear to constitute, a more powerful argument against essentialism. This matter will be returned to shortly. In a later paper, Barcan Marcus recognizes that the translations might be made with the modal operators fully embedded. This forms the basis for a second relational interpretation. The modal operators in both (34') and (35') are not fully embedded. Barcan Marcus does not fill in her suggestion; so I shall. Here 123 then are the modified premises of Quine's argument with the operators fully embedded: (34**) (Vx) ((Mx - ORx) A (Mx - ->OBx) ) and (35 ' ' ) (Vx) ( ( Cx - UBx) A ( Cx -* ->ORx) ) . Both (36) and (36*) are existential. Let a be this cycling mathematician: (36'' ) Ma A Ca. Using the rule of universal instantiation, the following can be obtained: (37) (Ma - QRa) A (Ma - ~UBa) UI, (34'') (38) ( Ca - UBa) A ( Ca - -Of?a) UI, (35''). We can then use the rule of ' &E' to obtain the four following conditionals: (37' ) Ma -* DRa (37' * ) Ma - -iQBa (38') Ca - D5a (38« * ) Ca -* ->DRa. When (36' • ) is conjoined with the conditionals (37*)-(38'* ) allows us to obtain the two following contradictions: 124 ( 3 9 ) URa A ->ORat a n d ( 4 0 ) DBa A -UBa. Al though (C") (3x) [(DRx A -^ORx) A (QBx A -iQBx) ] follows from (34''), (35'*), and (36'), this interpretation is harmless, since (34'') and (35**) are false. This follows if we use a possible worlds semantics. Let Fermat be a mathematician in the actual world who happens to be irrational in another world. Since Fermat fails to satisfy the formula "Mx -+ DRx, " it follows that (34'') is false. Similarly, let Bauer be a cyclist in the actual world but without one of his legs in another world. Since Bauer fails to satisfy the formula " Cx - QBx," it follows that (35*') is false. This de re interpretation of (34) and (35) might appear to be the strongest condemnation of essentialism that Quine could have used. But I wish to emphasise that essentialism emerges unscathed from this condemnation, since both of the interpreted premises are false. There is another, more complete, interpretation of Quine's argument against essentialism, also by Barcan Marcus. In the Perspectives on Quine volume, she notes that there is an ambiguity in the previously mentioned (34) between the de dicto and the de re interpretations. In the prior paragraph, I spelled out some 125 of the ambiguities that can arise among competing de re interpretations. Are we saying that implicit.. . . is ' •( x) ( Mx^Rx) or ( x) ( Mx^ORx) ... If the reading of the implicit conditional premises is like the latter [ de re] with the modal operator attached to the consequent, the premises are inconsistent. If more plausibly the former then nothing baffling follows. Barcan Marcus' point about the earlier, de dicto, premises still holds even if we use de dicto versions of (34) and (35), which are as follows: (34*) •( Vx) (Mx - Rx) A ->D(Vx) (Mx - Bx) and (35*) •( Vx) (Cx -* Bx) A ->•( Cx - Rx) I should add that a de dicto version of Quine's conclusion, along the lines of (34*) and (35*), would be a radical departure from the original. So we'll simply ask if either (C ) or (C ' ) follows from (34*), (35*) and (36). Nothing baffling follows from the de dicto version because nothing follows about which traits are possessed essentially (or necessarily) by a given individual. Though the premises are collectively true and the conclusions are false, one cannot obtain either (C* ) or (C**) by the rules of QML from those premises. Let there be two possible worlds, a real world and a world p. Let both worlds contain exactly three things: Bauer, Einstein, and Fermat. In the real world Einstein and Fermat are mathematicians and rational, Bauer and Einstein are cyclists and bipedal. Fermat is neither a cyclist nor bipedal, and Bauer 126 is neither a mathematician nor rational. In p, none of Bauer, Einstein, or Fermat are mathematicians, rational, cyclists, or bipedal. But this does not evoke any bewilderment at all, for the rules of QML do not allow either conclusion to be derived from (34*), (35*), and (36).9 Barcan Marcus goes on to note that Quine does not repeat this argument after 1962. It "does not surface again in Quine's later critiques of modal logic." There are then, at least three ways that one can interpret Quine' s second argument against essentialism. Two of these interpretations are de re, and the third is de dicto. I am in substantial agreement with Barcan Marcus when it comes to the absence of baffling results in the case of the de dicto interpretation. The one area that we may not be in total agreement relates to the fully embedded de re interpretation given above; I say this because Barcan Marcus does not fully spell out the fully embedded de re interpretation. In no way does she surrender the gate to Quine when it comes to either of the two de re interpretations. Instead, Barcan Marcus no doubt would accurately question the truth of premises such as (34) and (35), on either interpretation. The main point that Barcan Marcus is making, then, is that none of the interpretations of (34), (35), (36), and (C) succeeds in evoking our bewilderment. It should be stressed that the second argument against essentialism given by Quine in WO is not substantially different from his first argument in "TGMI" and in FLPVin that both versions attack the notion that it makes sense to separate essential from 127 accidental properties. Yet the first and second of Quine' s arguments against essentialism are distinct from his objection in "R&M," where he holds that escape from the problems of (SI) and of (EG) and of (US) necessitates a reversion to Invidious Aristotelian Essentialism (hereafter, I.A.E. ). While the first problem concerns the untenability of essentialism, the second problem shows that this untenable doctrine must be adopted by philosophers seeking to avoid the earlier objections to QML (mentioned in Chapter II). We turn now to this second problem. § 17. Quine on the Reversion to Invidious Aristotelian Essentialism Is Quine's criticism that quantification into modal contexts can be sustained only if one reverts to essentialism compelling? In a thoroughly qualified sense, the answer to this question is 'yes' . Quine, in his 1961 version of "R&M," pointed out that the friends of QML are committed to the view that if individuals x and y are identical, then they are necessarily identical. This particular argument was referred to above on p. 43. So it is true that some version of essentialism will have to be adopted if one wishes to use QML. But what is the precise nature of the essentialism that one will have to adopt? Before answering this question, let us consider Quine's affirmative argument from "TGMI": How Aristotelian essentialism as above formulated is required by quantified modal logic can be quickly shown. Actually something yet stronger can be shown: that there are open 128 sentences ' Fx and ' Gx fulfilling not merely (54) [seeabove-M. W. D.] but: (x)(nec Fx . Gx .~nec Gx) , i.e.: (x) nee Fx . (x) Gx . ( x) ~nec Gx. An appropriate choice of ' Fx' is nee : ' x = x . And an appropriate choice of' Gx' is 'x = x.p', where in place of ' p' any statement is chosen which is true but not necessarily true. Surely there is such a statement, for otherwise 'nee' would be a vacuous operator and there would be no point in modal logic. This argument has considerable kinship with the stance Quine later adopted both in WO as well as in the second version of "R&M." Kaplan admits that quantification into modal contexts can be sustained only if some form of essentialism is accepted, but he holds that only a benign form of essentialism is required. That was put in the preceding paragraph in the negative. As we shall see, Quine's comments on unique specification appear to lead more to what Kaplan terms 'Benign Quinean Essentialism' (hereafter, B.Q.E. ) than they do to I.A.E. Quine does not provide a compelling argument that I.A.E. would be the conclusion of an argument based on the premises in either "R&M" or "TGMI." Kaplan's argument allows for the assumption that Quine' s earlier criticisms of QML are sound. I have already questioned this assumption. Kaplan's argument may, however, still be a worthy contribution if it succeeds in illuminating the various strands of essentialism that appear to be invoked by a systematic rejection of Quine's criticisms. To be precise, Quine*s criticism that the only way that QML can be "sustained" will be if one assumes I.A.E. This position will now be scrutinized in detail. What then is Quine's 129 view as to the characterization of (I.A.E.)? That is, in what way is Aristotelian essentialism offensive, at least according to Quine? The answer is that Quine views the essentialist as being committed to sundry types of necessary properties, as was witnessed in the de re interpretations of (34) and (35). I disagree. Instead, I would opt for a more moderate approach to essentialism. Kaplan holds that cross-quantification is legitimate even if a less virulent strain of essentialism is invoked. This latter view relies on the notion of logical truth. One way in which logic is not invidious is in the Fungability of individuals. Thus if the valuation of <j) by p is logically true, any valuation .... by an isomorphic assignment g will also be logically true ... But discrimination in favour of logical truth hardly seems invidious. You can't be harmed by logical truth. As Barcan Marcus has speculated, it is possible that Quine' s primary motivation for attacking essentialism is that he recognizes that any version of the doctrine will be in conflict with the views expressed in "TDE." Both she and Kaplan have noted that the instances of essentialism which are attacked by Quine are not cases of invidious essentialism, but are instead benign in nature. Here is a version of essentialism that is genuinely invidious: I.A.E. : (3F) Ox) (3y) (UFx A Fy A ^UFy) Kaplan in particular holds that Quine's argument against essentialism does not yield invidious results. This leads him to a reconstructed version of the argument which he contends, rightly so, is an ad hominem one. This innocuous form of essentialism is 130 contained in the following principle: (E) (3x) {D(x = x) A [ U = x) A P] A ->D[(x = x) A P]}, where, again, "P" represents an arbitrarily chosen contingent truth. As Kaplan notes, (E) is hardly baffling. (E) is consistent with both I.A.E. as well as the following kind of essentialism: B.Q.E: (VG) (Vx) (Vy) {(DGx A Gy) -* DGy} B.Q.E. is a contradictory of I.A.E. Kaplan doubts that, by itself, (E) will result in I.A.E. "I cannot believe that benign essentialism of the kind exhibited in (E) could have been Quine's target." Here is the previously mentioned ad hominem reconstruction provided by Kaplan in order to explain why Quine thought of (E) as leading to I.A.E.: (i) Adoption of a relational sense of necessity (or acceptance of quantification in) allows one to formulate I.A.E. claims, (ii) Those who adopt such a sense must wish to assert such claims, (iii) Such claims are unjustifiable. (ii) is false. However, one could still adopt a relational sense of necessity without wishing to assert I.A.E. claims. Kaplan was cited above reconstructing Quine's rationale that if one uses the relational sense of necessity, then one will be able to formulate I.A.E. Barcan Marcus goes one step further, pointing out that the essentialism of Kripke and Putnam is not invidious. I concur with her claim. Both Kripke and Putnam, I would contend, have the sort of essentialism that is neither I.A.E. nor B.Q.E. There is yet a third alternative, which will be explained 131 shortly. Aristotle viewed essential properties, Barcan Marcus notes, as sortal. "Objects which have them, have them necessarily but there are objects which do not or might not have them. Being an entity is a non-sortal property as presumably would be being self-identical." Yet (according to Barcan Marcus and Kaplan) Quine, who is opposed to I.A.E., concentrates on non-sortal properties 1R when he attacks the doctrine. That is, (according to Barcan Marcus and Kaplan) the sort of essentialism that Quine is attacking is not the same doctrine that Aristotle defended. Here, roughly speaking, is Aristotle's doctrine: Moderate Aristotelian Essentialism (hereafter, M.A.E.): (3x) (3F) (3G) (Fx A Gx A UFx A -UGx) . It should be noted that M.A.E. is an existential generalization of (E). An advocate of M.A.E. would defend a very limited number of properties being necessary. M.A.E. is neither as metaphysically innocuous as B.Q.E. nor as metaphysically offensive as I.A.E. The necessary and possible properties of M.A.E. are modal attributes; their legitimacy is tentatively defended in Chapter VI. § 18. F0llesdal on Quine's 'Conflation' of Two Notions of Essentialism Before leaving the topic of Aristotelian essentialism altogether, there is a curious problem that can be solved at this 132 point. It concerns the charge that Quine failed to distinguish between the Church/Carnap solution and the reversion to Aristotelian essentialism. More precisely, F011esdal has held that Quine conflated two versions of essentialism, one logical while the other is metaphysical. F011esdal writes that Quine in 1961 was satisfied that he had refuted Carnap and Lewis' views on the modalities and had shown that quantified modal logic requires what he then called 'essentialism1 . He also acknowledged that the formal difficulties that he had brought to light in his writings on quantified modal logic could be overcome if one introduced the notion of "genuine" singular terms. When some modal logicians in the seventies argued that there was a problem of identifying an object from one possible world to another and appealed to a notion of essence for this purpose, Quine seems to have regarded this as just another manifestation of his old claim that quantified modal logic requires essentialism. However, in my view, Quine was wrong here. I would argue that there is scant evidence for this position, which is given as follows: "The only form of essentialism that Quine's argument shows modal logic to be committed to, is essentialism in the weak sense . . . " The view that Quine was dealing with two versions of the doctrine of essentialism is not borne out by his comments in "R&M," where he specifically mentions that it is Aristotelian essentialism that will have to be relied upon in order to evade the objection based on the principle of (SI), and the two arguments for the meaninglessness of the relational idiom in modal contexts (the objection based on the phenomenon of (EG), and the problem of (US)). The claim that Quine has conflated two versions of essentialism derives its plausibility from the fact that two philosophers (C.I. Lewis and Carnap) developed accounts of modality which sought to 133 incorporate analyticity. Both treated modality in terms of the analytic/synthetic distinction. Quine's own account of semantic ascent is consistent with the approaches developed by these two philosophers. But it would be a mistake to hold that essentialism was somehow being invoked in two separate ways by Quine. F0llesdal appears not to have grasped a simple point: the question of merits vs. demerits of essentialism does not even arise, in the second version of "R&M," where Quine is discussing the Church/Carnap solution to his three initial objection to QML. Moreover, Quine himself disagrees that he conflated two competing versions of essentialism: Mine was a single inclusive conception of essentialism embodying just one general condition: "an object, of itself and by whatever name or none, must be seen as having some of its traits necessarily and others contingently." Which traits these might be is left open to begin with. F0llesdal grants my contention of the early fifties and before, that some degree of essentialism thus defined is called for in quantifying into modal contexts. The necessity dedicto conceived by Garnap and C.I. Lewis did not suffice. It is clear that Quine was specifically attacking the merits of Aristotelian essentialism because he correctly saw that it would be used by proponents of the relational idiom. Therefore, F011esdal is incorrect in his contention that only the logical modalities are attacked by Quine. It is true that Quine switches between attacks on the logical and metaphysical modalities, but he does not ignore either notion nor does he conflate them. The charge that Quine is conflating two rival versions of essentialism also fails to note the fact that Quine clearly distinguished two accounts of quantification into modal contexts. 134 That is, F</>llesdal omits any mention of Smullyan' s use of Russell's TDD as a separate account, from that of Church/Carnap, of quantification into modal contexts. This is a serious defect, especially when one considers that Quine' s view that quantification into modal contexts requires essentialism is meant to apply only to the account of Smullyan, and not to that of Church/Carnap. F0llesdal is partially correct in that the notion of essentialism in Quine's writings from 1941-1961 had a different motivation from the notion of essentialism in the writings of the seventies. And he is correct in pointing out that Barcan Marcus and Terence Parsons have shown, contra Quine, "that most systems 10 of quantified modal logic have anti-essentialist models... " So, F0llesdal's analysis is not altogether faulty. His confusion is easy to understand in light of (1) his failure to detect the presence of Smullyan's solution alongside that of Church and Carnap; (2) his correct assessment of the distinct motivations that underlay the discussion of essentialism in the two periods mentioned above; and (3) his lack of discussion of Quine's use of the two separate versions of his argument involving the conjoining of a contingent truth p. (F011esdal may have mistakenly thought that the argument directed against essentialism in "TGMI" was identical to his argument against the Church/Carnap solution given in the 1961 version of "R&M.") Quine's arguments against essentialism have been distinguished from his argument that use of Russell's TDD will constrain one to adopt essentialism. The two categories are of course connected; 135 it is only because essentialism is held to be unacceptable that it apparently constitutes a valid criticism of the relational idiom in modal contexts. Quine's contention that essentialism, as it is presented by Quine, is properly characterized as invidious, has been shown to be false. It has been argued in this chapter that Quine's charge that defenders of quantification into modal contexts, where such quantification is intended to illustrate the relational idiom, are committed to I.A.E. is untenable. It has also been urged that the two arguments that Quine has used against the doctrine of essentialism are unsound. More precisely, Quine's argument is inadequate -- in its most natural interpretation it is invalid, given a second possible interpretation the premises need not be accepted by the essentialist. This is not to say that essentialism would withstand every critique that might be fielded against it. But it is safe to conclude that Quine does not have an explicit argument against the doctrine at his disposal. As I shall show in Chapter VI however, Quine does have some more implicit arguments against essentialism that he is able to use. Nor is Quine finished with the topic of modality and quantification. There are three more novel arguments that Quine has made in connection with these two subjects. One of these arguments will be analyzed in the following chapter, namely, that the distinction between ordinary truths and necessarily true statements collapses if one attempts to merge quantification theory with modal logic. 136 Chapter V. Stages 3 and h of Quine's Criticisms of the Relational Idiom § 19. General Remarks This chapter is primarily concerned with two periods of Quine's critique of the relational idiom. The first period is marked by the paper "Q&PA." This period corresponds to what I termed in Chapter I as the third of principally five stages of Quine's critique of the relational idiom. As was first indicated, also in Chapter I, this paper is significant for a number of reasons, but we are mainly concerned here with his invocation of attributes as a way of accounting for the evident significance of certain sentences expressing the relational sense of belief. The second period is marked by the book WO, where he unveils his seventh and eighth criticisms of the relational idiom. This second period corresponds to what I termed in Chapter I as the fourth stage of Quine's critique. This chapter will additionally focus on a third period that is marked by the book Pursuit of Truth, where Quine appears to reverse substantially the position he earlier took in WO. Each of these works is highly significant in terms of Quine's receptivity towards the relational idiom in doxastic contexts. Whereas the previous chapter was concerned exclusively with modality, this chapter focuses mainly on Quine's problem of reconciling meaningful instances of the relational idiom with its 137 alleged logical difficulties. The exception to this occurs in § 21.2, where the topic of modality is again discussed, this time in connection with an evaluation of Quine's charge that quantification into modal contexts results in the collapse of the distinction between a necessarily true sentence and one which is merely true. The strategy that is followed in this chapter is quite straightforward: selected arguments will be taken from each of "Q&PA," WO, and Pursuit of Truth. They are then critically discussed. It is my contention that the first two works are dramatically distinct from the third. § 20. Quine on the Role of Attributes in "Q&PA" In "Q&PA" Quine concedes that the relational idiom in doxastic contexts is meaningful. It is argued, however, that this endorsement on his part is half-hearted at best. Quine distinguishes between relational and notional constructions of belief and holds that we are "scarcely prepared to sacrifice the relational construction 'There is someone whom Ralph believes to be a spy' ... " Later in the same paper, Quine writes that Belief contexts are referentially opaque; therefore it is prima facie meaningless to quantify into them; how then to provide for those indispensable relational statements of belief. . . ?2 This stance marks something of a change on Quine's part; in his early papers on the topic, Quine held that, presumably all, 138 propositional attitude ascriptions were referentially opaque. Thus, he writes, in "NE&N," that (S) Philip is unaware that Tully denounced Catiline, is but "an example of another common type of statement in which names do not occur designatively. " The salient point, or at least one of them, is that Quine does not, in 1943-1953, treat ascriptions of propositional attitudes any differently than modal sentences. By 1956, this has changed. As we shall see, after 1956, Quine has adduced a plethora of reasons to reject the relational idiom in doxastic contexts. The first change of mind on Quine's part is referred to as his first volte-face and the latter is referred to as his second volte-face. The present section is concerned with the first volte-face. (Quine's first volte-face is largely identical with the third stage of his critique of the relational idiom.) It is not simply doxastic phenomena which motivate Quine to distinguish relational and notional constructions. Belief is one of the propositional attitudes, and also the most widely discussed one. But there are others, including wishing, seeking, and striving. Some of Quine's examples of relational constructions of these latter sort of propositional attitudes include: (41) (3x) (Witold wishes that x is president), (42) (3x) (x is a dog. x talks. I seek x), and 139 (43) (3x) (x is a lion. Ernest strives that Ernest finds x) . Quine is correct to imply that relational constructions of some of these other propositional attitudes are necessary to language and that they are therefore "indispensable." Quine accounts for the evident meaningfulness of instances of the relational idiom in doxastic contexts through the use of attributes in the following fashion: . . .we may recognize also a triadic relation of belief among a believer, an object, and an attribute, thus: (15) Ralph believes z{z is a spy) of Ortcutt. For reasons which will appear, this is to be viewed not as a dyadic belief between Ralph and the proposition that Ortcutt has z( z is a spy), but rather as an irreducibly triadic relation among the three things Ralph, z{ z is a spy), and Ortcutt.4 If Quine's (15) were viewed as an instance of dyadic belief, then, according to Quine, it would be illegitimate to apply existential generalization to 'Ortcutt'. But this would undo the essential point about relational constructions of belief. Prior to invoking attributes to account for the evident meaningfulness of instances of the relational idiom in doxastic contexts, Quine considers an alternate solution. He calls it an obvious move to distinguish between two senses of belief: belief\, which disallows (11) [w sincerely denies '...". wbelieves that ... (p. 185 of "Q&PA") ] , and belief2, which tolerates (11) but makes sense of (7) [Quine ' s ( 7 ) is my (1' ' ) (3x) ( Ralph believes that x is a spy) -M. W. D.]. For beliefi, accordingly, we sustain (12)-(13) [(12) Ralph believes that the man in the brown hat is a spy, (13) Ralph does not believe that the man seen at the beach is a spy (p. 185 of "Q&PA")] and ban (7) as nonsense. For belief2, on the other hand, we sustain (7); and for this sense 140 of belief we must reject (13) and acquiesce in the conclusion that Ralph believes2 that the man at the beach is a spy even though he also believes2 (an<^ believesj ) that the man at the beach is not a spy. But Quine does not pursue this line of reasoning. Instead, he goes on to offer an account of the role of attributes in explaining the evident meaningfulness of the relational idiom in belief contexts which he is scarcely prepared to sacrifice. I agree with Quine that belief is a relation. Quine explicitly, but not exclusively, endorses the view that belief is a relation between a mental agent and a sentence, both in "Q&PA" and in "IR." The sentence (44) The winner of the presidential election in 1968 is believed by Ralph not to have won the election in 1968. expresses a relation, according to Quine, between Ralph, the winner of the presidential election in 1968 and (in "Q&PA") an attribute. (One can modify this approach to include either the property or the attribute of having won the 1968 presidential election in the threefold set of terms). Certain of Quine's views on ontology are discussed below in Chapter VI. Given these views, in conjunction with all the criticisms of the relational idiom mentioned above, it is surprising that Quine invoked a trio of intensional entities (that is, attributes, and propositions) in "Q&PA." It is principally due to the early occurrence of the bulk of Quine's objections that 141 this endorsement of these three intensional entities should be seen as half-hearted at best. Thus, he writes that intensions "are creatures of darkness, and I shall rejoice with the reader when they are exorcised, but first I want to make certain points with the help of them."' § 21. Second Quinean Volte-Face The clear implication that is gathered from "R&M" is that contexts that are referentially opaque are illegitimate and therefore, one would assume, ripe candidates for being sacrificed. Moreover, it is infuriating to discover that Quine apparently thought of propositional attitude ascriptions as being worthy candidates for sacrifice (because they are referentially opaque) in both the 1953 and the 1961 versions of "R&M." Recall that "Q&PA" made its appearance in 1956. The picture in brief is this: (1) in 1953, instances of the relational idiom in doxastic contexts (and indeed all propositional attitude ascriptions) were suitable candidates for sacrifice; (2) in 1956, such instances of the relational idiom became candidates that we are "scarcely prepared to sacrifice"; and (3) in 1961, the earlier position from 1953, namely, that we are prepared to sacrifice such instances of the relational idiom, was reverted to. At the bottom of the first page of "Q&PA," Quine sheds some light on the manner in which his views have evolved on the relational idiom: "This paper appeared in the Journal of Philosophy 53 (1956), summing up some points which I 142 Q had made from 1952 onward." There are three ways in which we could account for these apparent volte-faces on Quine's part. One option, perhaps the simplest one, would be to hold that the 1961 version of "R&M" is nothing more than a rewrite of the earlier version. On this line of thought, the 1961 version would be expected to repeat all the substantial lines of thought from the earlier one. Therefore, any change of opinion on Quine * s part later than 1953 but earlier than 1961 would merely appear to constitute two changes while in reality there would be only one change of mind. But this option fails, because there are indeed changes of a substantial sort from the 1953 version to the 1961 version. We have already seen an example of this above (see Chapter III): whereas Quine's more convincing refutation of the Church/Garnap solution in the later paper centres on conjoining a contingent truth to an analytic condition, in 1953 Quine was content to argue against the Church/Carnap solution solely on the basis of what he takes to be the incomprehensible notion of analyticity. Therefore, if there were but one change of mind on Quine*s part regarding the readiness to sacrifice relational constructions of belief from 1953 ("R&M") to 1956 ("Q&PA"), one would not find cases of substantial change of opinion. But one does find such cases. It follows, by modus tollens, that there is not one change on Quine's part but rather at least two. The second explanation to account for the two apparent changes is to hold that Quine did not think it particularly germane, in 143 either version of his paper, to his dual topics in "R&M," to discuss the apparent meaningfulness of relational constructions of belief. This option appears plausible until one reflects that Quine discusses not only the referential opacity of doxastic/modal contexts, but also claims of the significance of relational constructions in QML. In other words, Quine is quite willing to discuss features that modal and doxastic sentences have in common earlier in the paper (again, both in the 1953 and 1961 versions), but this willingness to discuss resemblances ends later in "R&M" when he is only willing to raise the topic of claims of meaningfulness in modal contexts. Why this change of mind? The apparent meaningfulness of instances of the relational idiom in doxastic contexts, a subject which is dramatically similar to the one concerning claims of the apparent meaningfulness of such instances in modal contexts, is not mentioned in either 1953 or in 1961. This deafening silence on Quine's part is, I think, indicative of his willingness to return to what I have called his earlier position, namely, that propositional attitude ascriptions are referentially opaque and thus worthy of sacrifice. The third option to account for the two apparent changes on Quine's part, (the first, 1953-1956; the second, 1956-1961) is just to countenance that these transformations are genuine ones. (This is the option that the author finds preferable.) It has the merit of being consistent with the substantial changes of thought one finds in the text of "R&M." It also is consistent with the finding in "Q&PA" that instances of the relational idiom in doxastic 144 contexts are indeed meaningful. Finally, it is consistent with the finding, soon to be discussed, that Quine's endorsement of attributes was half-hearted at best. The third part of this detour in Quine's thought actually predates 1961; in 1960, Quine wrote WO, in which he again attacked the supposed meaningfulness of the relational idiom in doxastic contexts. His argument there, which is reviewed extensively below, is that an agent will end up believing every truth and every falsehood if a relational construction of belief (Quine uses, in WO, the term 'transparent sense of belief) is allowed. In this sense, the second version of "R&M" is but a continuation in a line of thought that Quine had begun in 1960. It seems to me that this is another piece of supporting evidence that there were two shifts, rather than just one, of thought on Quine's part from the period beginning in 1953 and ending in 1961. § 21.1 Quine's Charge Concerning the "intolerable oddity" of the Relational Idiom in WO Quine has produced several arguments against the relational idiom following his endorsement in "Q&PA" of that idiom in doxastic contexts. In WO, Quine's main attack on the relational idiom in doxastic contexts takes the form of the criticism directed at the belief states that are expressed by the relational idiom. His argument against the transparency of belief seeks to show that an agent will believe everything (every arbitrarily chosen truth 145 and falsehood) if this transparency is countenanced. Quine is correct to call this result an "intolerable oddity." Let us first turn to an evaluation of the charge of what Quine terms tolerable oddity. (This order follows Quine*s own argument in WO; Quine initially considers what he terms "tolerable oddity" of Tom's beliefs, thereby laying the groundwork for his conclusion of "intolerable oddity.") If there are genuine instances of the relational idiom in doxastic contexts, then there are cases where a believer has beliefs even though he or she might not give assent to sentences seeming to express those beliefs. This might, admittedly, appear to be odd at first glance. Yet its occurrence is fairly common. Quine discuses this tolerable oddity in the following citation from ¥0: .. . see what urgent information the sentence ' there is someone whom I believe to be a spy' imparts in contrast to ' I believe that someone is a spy' (in the weak sense of 'I believe there are spies') ... Surely, therefore, the transparent sense of belief is not to be lightly dismissed. Yet let its urgency not blind us to its oddity. "Tully," Tom insists, "did not denounce Catiline. Cicero did." Surely Tom must be acknowledged to believe, in every sense, that Tully did not denounce Catiline in that Cicero did. But he still must be said also to believe, in the referentially transparent sense, that Tully did denounce Catiline. The oddity of the transparent sense of belief is that it has Tom believing that Tully did and that he did not denounce Catiline. This is not yet a self-contradiction on our part or even on Tom's, for a distinction can be reserved between (a) Tom's believing that Tully did and that Tully did not denounce Catiline, and (b) Tom's believing that Tully did and did not denounce Catiline. But the oddity is there and we have to accept it as the price of saying such things as... that there is someone whom one believes to be a spy. Quine appears, in this passage, to be on the verge of pointing out that relational construction of belief will allow for lack 146 of verbal agreement from mental agents. Belief-ascriptions might not always aim at fully reporting "cognitive contents" (though beliefs might be "purely composed" of them). But in fact Quine does not assert this, and is instead satisfied with showing that this oddity does not approach the level of self-contradiction. As far as I know, Quine never again raises the subject of tolerable oddity of the relational idiom, except when he directly reverses his position in Pursuit of Truth, some thirty years later. Calvin Normore has devised an objection to Quine's use of the attribute that bears striking resemblances to Quine' s position on tolerable oddity. The following are Normore's remarks: ... consider: having forgotten his grade school chemistry Albert believes that he is drinking a glass of water but not that he is drinking a glass of H^O. Call the glass of stuff in question X. Albert believes of X that is a glass of water but does not believe of X that it is a glass of H20. (Those sentences are in the relational idiom.) So the attribute of being a glass of water must be different from that of being a glass of H20. Surely this is a case of tolerable oddity, as described by Quine. Albert would no doubt fail to give verbal assent to the declarative statement D Albert is drinking a glass of 1^0, while fully agreeing with the declarative statement D' Albert is drinking a glass of water. Contradictory beliefs are a problem but are not on a par with believing in a contradiction. Albert has the former but does not 147 have the latter. Normore concludes that the attributes in question must be different. I find this conclusion to be debatable (see the discussion on attributes and properties on p. 183). More important is the point that the property of being a glass of water is equivalent to the property of being a glass of F^ O. Kripke's problem of Puzzling Pierre seems to me to be yet another variant on Quine's phenomena of tolerable oddity. As a result of early childhood experience of London, Pierre believes that "Londres est jolie" but as a result of less pleasant adult experience he believes that "London is not pretty". Pierre does not express a belief in a contradiction, such as the belief that "Londres est jolie et Londres n'est pas jolie" or the belief that "London is not pretty and London is pretty" . This latter pair of beliefs would clearly be intolerable. The picture is somewhat complicated by Kripke's use of the presumably bilingual Pierre, whose inconsistent beliefs evidently depend to an extent upon faulty translation. (They obviously do not depend entirely upon faulty translation; two experiences drawn from different stages of the same life plays a role as well as Pierre's bilingualism. ) But this complication ought not to deter us from contending that Pierre's oddity is a tolerable one. Quine's argument for the intolerable oddity of the transparent sense of belief proceeds as follows: Now if this much oddity on the part of the transparent sense of belief is tolerable, more remains that is not. Where ' p" represents a sentence, let us write 'dp* (following Kronecker) as short for the description: 148 the number x such that ((x = 1) and p) or ((x = 0) and not P). Since Tom believes Cicero did denounce Catiline, and granting Quine's "logical acumen" principle as applied to Tom, we get "(3) Tom believes that 6(Cicero denounced Catiline) = 1. But, whenever 'p1 represents a true sentence, 11 6p = 6(Cicero denounced Catiline)...." Quine now applies the transparency of belief to (3) to obtain Tom believes that dp = 1. At this point, the hypothesis about Tom's logical acumen is once more used, this time to obtain the result that (4) Tom believes that p. But ' p' represented any true sentence. Repeating the argument using the falsehood 'Tully did not denounce Catiline1 instead of the truth 'Cicero denounced Catiline', we establish (4) also where ' p" represents any falsehood. Tom ends up believing everything. The notion of a transparent sense of belief is thus rendered by Quine to be nonsensical. Fortunately, his argument is unsound, as I will now show. Quine's conclusion (4) does not follow from the transparency of belief and the supposed logical acumen of Tom. Patton has given a thorough critique of Quine's argument concerning the intolerable oddity of transparent belief, and the key points of Patton's argument will be summarized here. Assuming, for the sake of argument, that (3) satisfies the law of (SI), as well as that 6p 149 = (Cicero denounced Catiline), Quine would be correct to conclude that Tom believes that 6p = 1. But it is improper to conclude from this result and Tom's supposed logical acumen that (4) Tom believes that p. Quine views this acumen as being central to the content of Tom's beliefs. For proof of this claim, consider the following citation from WO: We may suppose that poor Tom, whatever his limitations regarding Latin literature and local philanthropies, is enough of a logician to believe a sentence of the form '6p=l" when and only when he believes the sentence represented by ' ps . Patton concentrates on Quine's use of logical acumen. He writes that big trouble does arise over Tom's logicality.... if we ask how this "logical acumen" is supposed to work, we will see that nothing properly so called, and indeed nothing plausibly ascribable to Tom, could make this generalization true, given what the context of Quine's argument makes clear: that the generalization must refer (perhaps inter alia) to transparent beliefs.16 If, as he ought to, Quine were to stay with the transparent sense of belief, it would be impossible for him to reach his conclusion that Tom ends up believing everything. Tom's logical acumen, on any conception of it that Quine could defend as plausible, would have Tom being actively aware of the sentence represented by 'p" iff he believes a sentence of the form '6p=l'. Why would this condition be true if it were being applied to sentences representing a transparent sense of belief (that is, to instances of the relational idiom in doxastic contexts)? My tentative answer is that this condition is not true when applied to sentences representing a transparent sense of belief. 150 However, this condition would be true when applied to sentences representing an opaque sense of belief (that is, to instances of the notional idiom). For this reason, I think that (3) should be construed as a de dicto or opaque sentence and not as a de re or transparent one, assuming that we are to make sense of Quine's position on Tom's logical acumen. It is certainly true that Quine believes, albeit mistakenly, that he is condemning the transparent sense of belief. A transparent sense of belief would be represented by an instance of the relational idiom in doxastic contexts. But if, as I believe to be the case, Quine is relying on the opaque sense of belief to explain Tom's logical acumen, then the previous step that Tom believes that 6p=l does not follow. For he could no longer rely on the transparency of belief. Quine's indictment of the transparent sense of belief thus relies crucially upon the wrong sense of belief. Therefore, Quine's argument collapses. Incredibly, Quine still holds in WO that a transparent sense of belief must be accepted, even though he concludes that this sense of belief is intolerably odd. Thus, he writes that the "need for cross-reference from inside a belief construction to an indefinite singular term outside is not to be doubted." Cross-reference is a feature of the transparent sense of belief, which is expressed by use of the relational idiom in doxastic contexts. So Quine is caught between a need (for transparent belief) and an intolerably odd place. Recognizing that he is precariously close to being ensnared in a web of contradiction, Quine offers this puzzling solution: 151 In general what is wanted is not a doctrine of transparency or opacity of belief, but a way of indicating, selectively and changeably, just what positions in the contained sentence are to shine through as referential on any particular occasion...A way of doing that is to agree to localize the failure of transparency regularly in the 'that1 of 'believes that' in the 'to' of believes to' , and not in the ' believes' . Thus we may continue to write ' Tom believes that Cicero denounced Catiline' when we are content to leave the occurrence of 'Cicero' and 'Catiline' nonreferential, but write rather: (5) Tom believes Cicero to have denounced Catiline if we want to bring 'Cicero' into referential position. Similarly we can get 'Catiline' into referential position thus : (6) Tom believes Catiline to have been denounced by Cicero. If we want to get both into referential position, we are driven to something like: (7) Tom believes Cicero and Catiline to be related as denouncer and denounced. On this convention 'believes that' is unequivocally opaque...What was offered before as an idiomatic equivalent of (2) remains legitimate, however: 'There is (or was) someone whom Tom believes to have denounced Catiline'. ° There are two points that are prompted by this proposed solution. Firstly, it is unclear how Quine's solution to his objection that the transparent sense of belief is intolerably odd is supposed to function. It is important to appreciate that this point rests on the implicit assumption that Quine's objection is a tenable one--an assumption that, I believe, to be a thoroughly dubious one for reasons that I have mentioned above. The point in question thus should be phrased as a conditional: if Quine's criticism of the intolerable oddity of the transparent sense of belief is sound, then it is unclear how Quine's solution is supposed to function. Quine's solution appears to come down to an exclusion of 'that' and 'of clauses in cases of the transparent sense of belief. 152 However, this exclusion would not materially alter the damage supposedly incurred by his charge of intolerable oddity; Quine's argument could be paraphrased to exclude any 'that' clauses. Quine simply does not explain in sufficient detail just how his solution is intended to work. In brief, it is unclear how Quine's supposed solution to his supposed criticism constitutes an advancement of his philosophy of language. Secondly, even if we grant that Quine's solution adequately solves his criticism of the intolerable oddity of cases of the transparent sense of belief, then once again Quine has contradicted himself. This time, the problem is concerned with the first of his two criticisms of quantification into opaque contexts in "R&M." This problem was referred to in previous chapters as the one based on the phenomenon of (EG) and as Quine's first explicit criticism of the relational idiom. One might think that Quine only applied this problem to quantification into modal contexts, but this view, as I first indicated in Chapter II, is mistaken. In both the 1953 and 1961 editions of "R&M," Quine applies the step of existential generalization to (19) and obtains (19') (3x) (Penelope is unaware that x denounced Catiline). This fact has also been mentioned in Chapter II. The years of the two editions of FLPV and the appearance of WO are significant. The first edition of FLPV occurred seven years prior to WO, while the second edition of FLPV occurred one year after WO. There were substantive changes between the two editions of FLPV. Yet the 153 proposed solution that Quine submitted in WO is arguably at sharp variance with the position adopted regarding the problem based on the phenomenon of (EG) in both editions of FLPV. I have the suspicion that the only reason that Quine has for excluding 'that' and 'of clauses in cases that feature the transparent sense of belief is to block the premises in his argument for intolerable oddity that feature such clauses. In particular, the premises that are Quine's (3), his premise Tom believes that 6p = 1, and (4) presumably all go by the wayside. But is this really true? Cannot (3) be paraphrased as (3') Tom believes 6(Cicero denounced Catiline) to be identical with 1? I can see no reason why not. Similarly, Quine's intermediate premise could be paraphrased as Tom believes 6p to be identical with 1. Finally, (4) could be paraphrased as (4') Tom believes p. Furthermore, the arguments that are given by Quine in support of the steps in the 'intolerably odd' criticism are, as I have suggested, defective, since Tom's logical acumen will not allow Quine to obtain his intermediate step. His proposed solution does 154 not remedy this defect. It is therefore an ignoratio elenchi. I frankly fail to see that this solution on Quine' s part to his objection of the intolerable oddity of the transparency of belief advances his case at all. The solution given by Quine therefore does not allow him to avoid being ensnared in a web of contradiction. Quine's charge of the intolerable oddity of the relational idiom in doxastic contexts depends to a large degree on his assumption that Tom will believe sentences having the form '&p = l1 iff he believes the sentence represented by 'p". There is no reason to grant this assumption where Tom is having a relational belief. On the other hand, Quine*s assumption is plausible if Tom's beliefs are opaque ones. Opaque beliefs are conveyed through the use of instances of the notional idiom. This notion of the conveyance of opaque beliefs is, however, a profound mistake as it is transparent belief which Quine is criticizing. Transparent beliefs are conveyed through the use of instances of the relational idiom. Thus, Quine is only apparently engaged in a critique of sentences expressing the relational sense of belief. As in the case of the problem of (EG), Quine is evaluating, or rather claiming to evaluate, a quantified doxastic sentence that features cross-quantification. (See § 4.2 and § 9.) He should not introduce a de dicto interpretation of this sentence. For, as indicated above, in Chapter III, the de re interpretation is the one that also features cross-quantification and not the de dicto one. Quine does much the same thing when he evaluates de re belief 155 as would be contained in a sentence like (45) Tom believes of Tully that he denounced Catiline. The oddity that (in WO) is tolerable does not arise unless a de dicto interpretation of (45) is given. The oddity concerns the ability of an agent to entertain beliefs that can be described (by another) as contradictory. But, as in the case of the problem posed by the phenomenon of (EG), the de re interpretation is the one that should be evaluated. The de re interpretation will be the one that correctly features cross-quantification. I would readily concede that the contents of an agent's belief could be one way or the other. Yet (45) is not about the contents of Tom's belief, but is instead a de re ascription of his belief. (The relational idiom in doxastic contexts are ascriptions of belief.) What is conveyed by such a de re ascription of belief? It is true that some of the content of Tom's belief will be conveyed by a de re ascription of his belief, but, as Patton has stated, we "fail to learn something that would correspond to a referring expression that would/could be used in expressing..." Tom's belief. Thus, a de re ascription of Tom's belief will not provide complete information about the content of his belief. On pain of inconsistency, then, Quine ought not to construe the transparency of belief as resulting in differing contents of belief. For Quine is clearly attempting to employ belief contents expressed by de dicto sentences (whether they are true or false) where he should more properly use contents expressed by de re 156 sentences in conjunction with Tom's supposed logical acumen. The result will not be that Tom believes everything, but rather that a new interpretation of belief, along the lines of Quine's own belief2 (in "Q&PA"), will have to be recognized. § 21.2 Quine's Corresponding Charge Regarding the Collapse of Modal Distinctions Quine argues in a symmetrical fashion that the distinction between truths and modal truths collapses if quantification is allowed into modal contexts. Let me now turn to a critical evaluation of this argument, which also made its appearance in WO. Immediately prior to this argument, Quine again contends that the necessity operator is opaque. Here is the relevant passage, which is given in its entirety: Such illustrations of opacity depend on the existence of appropriately stubborn objects. The stubbornness of 9 consists in its being specifiable in ways that fail of necessary equivalence (e.g. as numbering the major planets and succeeding 8), so that various traits (such as greaterness than A) are entailed with necessity under one specification of 9 and not under another. Now if we narrow the universe of objects available as values of variables of quantification so as to exclude such stubborn objects, there ceases to be any such objection to quantifying into modal position. Thus we can legitimize quantification into modal position by postulating that whenever each of two open sentences uniquely determines one and the same object x, the sentences are equivalent by necessity. Schematically we can put the postulate as follows, using ' Fx' and ' Gx (now) for arbitrary open sentences and using ' Fx and x only1 as short for ' ( w) ( Fw if and only if w - x)' : (4) If Fx and x only and Gx and x only then (necessarily ( w) ( Fw if and only if Gw)) . But this postulate annihilates modal distinctions; for we 157 can deduce from it that 'Necessarily p" holds no matter what true sentence we put for 'p". The argument is as follows. Let ' p" stand for any true sentence, let y be any object, and let x = y. Obviously then (5) (p and x = y) and x only and (6) x = y and x only. By (4), next, with its ' Fx" taken as 'p and x = y" and its Gx as ' x = y" , we can conclude from (5) and (6) that (7) Necessarily (w) ( (p and w = y) if and only if w = y) . But the quantification in (7) implies in particular ' (p and y - y) if and only if y = y" , which in turn implies 'p1; so from (7) we conclude necessarily p. Quine's argument regarding the collapse of modal distinctions, if (4) is allowed, appears to be valid. It is important to point out that, using Forbes' definition, Quine's (7) is a de re sentences, in virtue of the variable that occurs free in (7). Some instances of (4) are de dicto, whilst others are de re. The variable in (4) could occur free, in which case it would be de re. In the case of (7), the formula after the modal operator, that is, ' ( w) ( (p and w = y) iff w = y) s , contains both a bound and an unbound variable; the ' v? is bound, while the ' y is unbound. Depending on what formulae replace "F" and " G" (which are schematic letters), (4) will possibly contain free occurrences of a variable within the scope of the necessity operator. In the case under consideration, both "F" and " G" are replaced by formulae that contain free occurrences of "y" and so this instance of (4) 158 is de re. For example, suppose that there were a possible world, (3, in which x wrote Hamlet but where x did not write Macbeth. From this it follows that (4) is false in this interpretation. What Quine's argument succeeds in showing is that one can't legitimize quantification into modal position by postulating that whenever two sentences uniquely determine the same object x, the sentences are equivalent by necessity. If this particular argument has a familiar ring to it, that is because Quine used the same reasoning against the Church/Carnap solution. Quine' s (4) is nothing more than a repetition of the vindication of modal sentences that is given by the Church/Carnap solution. Oddly enough, F011esdal, who reviewed both the Church/Carnap solution and Quine's argument regarding the collapse of modal distinctions in "Quine on Modality," fails to point to this obvious similarity. The argument firstly assumes that 9 is a stubborn object since the occurrence of ' 9' in (46) Necessarily 9 > 4, is held to be not purely referential since 9 = the number of major planets. Quine also concludes on this basis that the necessity operator is opaque. He then goes on to argue that the stubbornness of 9 consists "in its being specifiable in ways that fail of necessary equivalence ... so that various traits (such as greaterness than 4) are entailed with necessity under one specification of 9 and not under another." It is his solution to this problem, a problem that is, I am convinced, not genuine, 159 which gives rise to the troublesome postulate in (A). It looks as though Quine has once more confused instances of the relational idiom with those of the notional. Again, I would concede that the following notional or de dicto sentence is false: (47) It is necessary that: the number of major planets > 4. Symbolically, using Russell's TDD, (47) becomes (47*) • (3x) \_{NPx A (Vy) {NPy - x = y ) ) A ( x > 4 ) ] . On the other hand, the relational or de re counterpart of (47) is true: (47') The number of major planets is necessarily > 4. Symbolically, (47') becomes (47 ' ' ) ( 3x) [ ( NPx A (Vy) ( NPy - x = y ) A D ( x > 4 ) ] . The problem of the alleged stubbornness of 9 results from Quine's quantification into (46), with the conclusion being (47). This is wrong. Quine should, on pain of inconsistency, reach the more innocuous de re conclusion (47**). (It is more innocuous in that, unlike both (47) and (47*), it is true.) The reason that he does not do this is simple enough: he is not prepared to accept any instances of the relational idiom in modal contexts because of its supposed metaphysical defects. Thus, Quine is content to deal solely with the notional idiom in modal contexts, even when the subject is quantification into modal contexts. 160 It can be concluded, therefore, that Quine's argument concerning the collapse of the distinction between true beliefs and truths is not on a par with his argument concerning the collapse of the distinction between truths and necessary truths. In the former case, Quine is dealing, or at least claiming to deal, with a transparent sense of belief, the kind of belief that places no restrictions on substitution of singular terms. By way of contrast, even while he is discussing quantification into modal contexts, he places severe restrictions upon substitution instances. Quine is not dealing with a transparent sense of necessity, which is what one would expect if his arguments were on a par. Quine has not successfully argued that the distinction between truths and belief in truths (or the distinction between falsehoods and belief in falsehoods) collapses. I think Quine' s argument in ¥0 shows that one particular method of legitimizing quantification into modal position just does not work. However, this does not mean that all attempts to legitimize quantification into modal contexts will not work. In particular, modal distinctions will not collapse if quantification is merged with modal logic in the manner proposed by Smullyan. Unlike the premises in Quine's argument for the collapse of modal distinction, Smullyan's use of Russell's TDD allows for cross-quantification past modal operators. So, in conclusion, Quine's argument for the collapse of modal distinctions does not pose an insurmountable problem for the relational idiom in modal contexts. The same conclusion holds for the similar argument that Quine directed at 161 the relational idiom in doxastic contexts. § 22. Quine's Reversal of Position In Pursuit of Truth In 1990, Quine did something of a volte face with respect to the criticism that acceptance of the relational idiom in doxastic contexts implied intolerable oddity. As we have just seen, his earlier position was that acceptance of relational sense of belief results in Tom's believing every truth and every falsehood. In the following passage, however, Quine introduces the notion of an empty name, presumably assuming that the name has no content for Tom: The voice of the ascriber of a propositional attitude sometimes blends with that of the man in the attitude, causing de re intrusions and consequent ambiguities. "Tom believes that Tully denounced Catiline' is perhaps false de dicto and apparently true de re, if Tom knows a bit about Cicero but not his other name. Or is it clearly true even de re? Perhaps, confusing 'Tully' with 'Lully' , Tom staunchly denies ' Tully denounced Catiline' , while asserting 'Cicero denounced Catiline'. We can still credit Tom with believing that somebody or other denounced Catiline, but of whom does he believe it? M. Tullius Cicero? Yes and no. Such is the dubiousness of quantifying into an intensional idiom, thus (1) 3x(Tom believes that x denounced Catiline), which calls for Tom's having in mind at least one specific person, as opposed to the unproblematic (2) Tom believes that 3x(x denounced Catiline). Consider the parallel as in the case of George IV: (3) 3x( George IV wonders whether x wrote Waverley) , (A) George IV wonders whether 3x(x wrote Waverley) . 162 Whereas (4) is clearly false, (3) strikes one as true, since the king knew Scott by sight and for his poems, and wondered only about the anonymously published novel. On the other hand (1) should perhaps be reckoned false, if 'Cicero' is little more than an empty name to Tom. As I will now show, this citation, when conjoined with one from the previous chapter, evokes its own appropriate sense of bewilderment. This matter constitutes the clearest example of a contradiction on Quine's part in his war on the relational idiom. In 1960, in WO, he held that contradictory beliefs on the part of Tom were a 'tolerable oddity' . Evidently, Quine has changed his mind, though he does use the same mental agent (Tom, with his inexact grasp on Roman history). Moreover, the attempt to obtain a reductio ad absurdum from the transparent sense of belief is at considerable odds with this view, which comes 30 years later than the one in WO. That is, Quine has apparently forgotten that Tom is supposed to be reduced to the unsatisfactory position of believing every truth and every falsehood, which should, I assume, include truths with 'empty names'. So, ex hypothesi, Tom must believe and not believe the same truth, even though it does feature an empty name. Quine' s position in Pursuit of Truth, that "(1) should perhaps be reckoned false, if 'Cicero' is little more than an empty name to Tom," bears striking similarity with Quine's three initial criticisms of QML in that it appears to rely on an exploitation of the ambiguity between instances of the notional and relational idioms. It really would be the case that a sentence like 163 (48) Tom believes that Cicero denounced Catiline, would be false if (48) were given a de dicto construal and ' Cicero' were an empty name to Tom. But, Quine*s example assumes that a de re construal of Tom's belief is being given. It would be irrelevant to the truth or falsity of Tom's belief that 'Cicero' were an empty name due to Tom, if the belief in question were given a de re construal. In a way, however, Quine's final criticism of the relational idiom is consistent with his earlier stance that he took in "R&M." The resemblance that I have in mind between these two writings involves the role of the relational fallacy. I pointed out earlier that Quine evaluates instances of de re modality in an incorrect fashion in that he inappropriately appeals to the truth or falsity of a de dicto construal of an ambiguous sentence. He also commits the relational fallacy in the passage just cited; Quine's (4) is the de dicto construal and his (3) is a de re construal. That Scott' is an empty name for George IV would be irrelevant to the question of the truth of (3); it would, of course, be relevant to the falsity of (A). Such is the dubiousness of criticising quantification into intensional contexts. To reach charitably the source of Quine's confusion, look no further than the change in terminology in the relevant cases of ¥0 and Pursuit of Truth. For in WO, Quine is speaking of the intolerable oddity of a transparent sense of belief, while in Pursuit of Truth, he is speaking of the truth or falsity of 164 quantifying into an intensional idiom. But how much significance ought one to place on this shift in terminology? Not much, in my view. It seems to me that conflicting beliefs were tolerable at one point and intolerable some thirty years later. To be precise, Quine held in WO that Tom could both believe that Tully did (in the transparent sense of belief) and that Tully did not (in both the transparent and opaque senses of belief) denounce Catiline. Quine's example in WO has Tom insisting that Tully did not denounce Catiline; 'Tully1 would be an empty name in this sense that Quine used in Pursuit of Truth. As we have seen, empty names became, for Quine, indicative of the falsity of instances of the relational idiom some thirty years later. The tolerable has become intolerable. It also seems to me that Quine's attempt in WO to show that a transparent sense of belief will have a mental agent believing everything was invalidated by his later attempt to show that the sentences expressing the relational sense of belief "should perhaps be reckoned false." According to the earlier picture in WO, Tom is supposed to end up believing everything; the presence of empty names ought not to count towards Tom believing falsely. Recall that my position in this chapter is that Quine' s initial view of the 'tolerable oddity' of a transparent sense of belief was a satisfactory one but that his argument regarding the 'intolerable oddity' of a transparent sense of belief was not. This is still my position, which means that I also reject Quine' s argument in Pursuit of Truth. The four main accomplishments of this chapter are, firstly, 165 a discussion of Quine' s use of attributes to account for the evident meaningfulness of the relational idiom in doxastic contexts; secondly, a critical discussion of what I have termed his ' second volte-face which is the second time that Quine reversed himself on the question of admissibility of the relational idiom in doxastic contexts; thirdly, the flaw in Quine's argument that the transparent sense of belief is intolerably odd has been pointed out; and fourthly, a discussion of Quine's reversal was made with respect to his two previous arguments in WO that are concerned with two kinds of oddity of the relational sense of belief. 166 Chapter VI. Attributes as Applied to Modal Phenomena and Quine on Their Identity and Individuation § 23. General Remarks This chapter is concerned with extending Quine' s own solution, in "Q&PA," from the problem of indispensability of the relational idiom in doxastic contexts to the analogous problem, as I see it, of the indispensability of the relational idiom in modal contexts. As we saw in the previous chapter, Quine is torn between this indispensability of the relational idiom, on the one hand, and his view that it is "conspicuously unsatisfactory in logical ways," by which he means the supposed referential opacity of the relational idiom in doxastic context, on the other. I wish to stress, at the very outset, that the author is not in favor of this analogous, and therefore Quinean, solution to this problem of accounting for the indispensability of the relational idiom in modal contexts. Instead, the author finds a solution prompted by Kripke's writings to be much more satisfactory. This second solution will be discussed briefly in the following chapter. Three reasons are prominent behind what I see as a requirement to extend Quine's use of attributes to provide a treatment of the relational idiom in modal contexts. The first reason is that the existence of parallels between modal and doxastic phenomena, argued for in Chapter II, furnishes limited justification for similar 167 analyses of doxastic and modal phenomena. Given that Quine has provided an analysis of doxastic phenomena that incorporates attributes, there is some justification for providing an analysis of modal phenomena that incorporates attributes as well. The second reason, also alluded to in Chapter II, is that Quine holds, in "R&M," that modal contexts are referentially opaque. That is, they are as conspicuously unsatisfactory in logical ways as are doxastic phenomena. Quine himself makes this point about the shared referential opacity of doxastic and modal phenomena. He perceives that the unsatisfactory nature, in logical terms, of the relational idiom in doxastic contexts is at odds with their indispensability. I previously indicated reasons for extending this stand on the indispensability of the relational idiom in doxastic contexts to the indispensability of the relational idiom in modal contexts. Since a precedent has been set with Quine invoking attributes in order to account for the evident indispensability of the relational idiom in doxastic contexts, attributes should be invoked to account for the evident indispensability of the relational idiom in modal contexts as well. The third reason is that Kripke's argument, cited above in Chapter III, to the effect that modal de re sentences like (22) The winner of the presidential election in 1968 might have lost the election. are intuitively meaningful, is just another way of arguing that such sentences are also indispensable. Quine, as we have seen, 168 is highly resistant to this conclusion. With respect, I would urge that he is wrong regarding this second horn of his dilemma. The reader may have noticed that we have come to a curious point in my attempt to show that attributes ought to play a similar role in the treatment of the relational idiom in both contexts. It is this: it was acknowledged that attributes are invoked by Quine in order to cope with a dilemma. And I indicated why I thought the second horn, that is, their character of indispensability, was a feature of modal contexts as well, contra Quine. But I have already argued against the other horn that Quine holds is present in both modal and doxastic contexts, namely, their supposed referential opacity. This curious feature of extending attributes to provide a treatment of the relational idiom in modal contexts is returned to towards the close of this chapter. Suffice it to say that the author is not blind to this apparent inconsistency and that an attempt is made to rectify the perceived problem. Quine's concerns about extensionality manifest themselves mainly in two different ways : firstly, there is his general concern about avoiding failures of extensionality; secondly, hypothetical entities that appear to defy an acceptable, for Quine, method of individuation are not admissible in his ontology. Now, both these sorts of concerns have considerable bearing on an evaluation of Quine's criticisms of the relational idiom. Regarding the first set of concerns, consider the following passage from ¥0: An opaque construction is one in which you cannot in general supplant a singular term by a codesignative term (one referring to the same object) without disturbing the truth value of 169 the containing sentence. In an opaque construction you also cannot in general supplant a general term by a coextensive term (one true of the same objects) , nor a component sentence by a sentence of the same truth value without disturbing the truth value of the containing sentence. All three failures are called failures of extensionality. It was these sorts of considerations that shaped Quine*s views of belief and the modalities. If one were to accept Quine's views concerning the alleged logical difficulties with propositional attitude ascription, then it may appear that one would be forced to adopt his solution which makes use of attributes and propositions. Quine would disagree, principally because of the lack of identity conditions for attributes. This matter will be returned to below; but it should be pointed out that Quine does consider an alternate way to account for the meaningfulness of de re doxastic sentences in "Q&PA" that embodies extensionalistic concerns. This was his use of language. One might consider this approach to have been free of concerns about identity conditions, but that is precisely the concern that Quine raises. This concern may more properly be called an intensionalistic one, but it is interesting to note that at one point at least Quine was willing to apply such concerns to extensionalistic solutions. Quine defends his use of intensional entities in a predictable fashion for an extensionalist: Let it not be supposed that the theory which we have been examining is just a matter of allowing unbridled quantification into belief contexts after all, with a legalistic change of notation. On the contrary, the crucial choice recurs at each point: quantify if you will, but pay the price of accepting near contraries like (15) and (22) [(15) Ralph believes z 170 (zis a spy) of Orcutt and (22) Ralph believes z (z is not a spy) of Orcutt-M. W. D.] at which point you choose to quantify. In other words: distinguish as you please between referential and non-referential positions, but keep track, so as to treat each kind appropriately. The notation of intensions, of degree one and higher, is in effect a device for inking in a boundary between referential and non-referential occurrences of terms. Quine does not, therefore, view the invocation of attributes in "Q&PA" to be a mere stylistic change in order to meet the perceived threat of referential opacity. The second sort of concerns that Quine has expressed regarding extensionality has less to do with opaque restrictions than it has to do with the exclusion of entities. Furthermore, such entities might play a role in an alternate account of belief and the modalities. A good example of this, and the one that I am chiefly interested in, of this would be attributes. Quine himself showed, in "Q&PA," that attributes could be used to account for the relational idiom in doxastic contexts. But Quine was never very deeply committed to his solution, principally because he thought that entities like attributes could not satisfy his ontological concerns. In this chapter, I intend to show that one can keep attributes while satisfying at least the spirit, although not the letter, of Quine*s concerns. Quine has consistently shown an admiration for desert landscapes. That is, he has consistently demonstrated his preference for an ontology with a minimum of types of individuals. This preference is directly behind much of Quine's rejection of the relational idiom. He wants to have his ontology as sparse as it 171 can be. In this chapter, another ontological principle of Quine's is investigated, namely, his principle that identity criteria must be spelled out for legitimate entities. Quine uses both principles in connection with his critique of the relational idiom in modal and doxastic contexts. (OR) and the dictum that there is no entity without identity are employed by Quine to cast doubt upon the purported existence of possible worlds, attributes, propositions, and intensionally conceived relations. Quine's use of the latter principle to argue that attributes cannot be individuated is discussed at length in this chapter. The strategy that will be followed in this chapter is threefold. Firstly, it is explained how Quine's treatment of attributes can be extended to cover modal phenomena. This use of attributes is a Quinean solution to a Quinean problem. Secondly, Quine's dictum that there is no entity with no identity is tentatively endorsed in order to evaluate this Quinean solution from the point of view of a hypothetical Quinean philosopher. Thirdly, the precise problem with using attributes, namely, that they cannot be individuated, will be fully spelled out. § 24. Attributes [Properties] Applied to Modal Phenomena To move an expression in the notional idiom to what Quine would call the referential position, he invokes the puzzling step of exportation. Exportation is critically evaluated in Chapter VIII; though I reject Quine*s logical step, it does produce some 172 dividends that are useful. Foremost among them is the use of the primitive predicate ' Bel' . David Kaplan makes active use of this notation in his "Quantifying In." He also draws certain parallels with the primitive modal predicate 'Nee1 . In this section, we will consider Nee and Pos as binary relations. These binary relations will be explained for philosophers who are willing to countenance the type of essentialism that was referred to in Chapter IV as M.A.E. 'Nee1 and 'Pos' are predicates that play an analogous role to that of modal attributes. Both primitive modal predicates and modal attributes will be analyzed as functions as part of a possible worlds semantic account. An adherent of M.A.E. (and this includes the author) would want to hold that a person is necessarily a person but that a person would not necessarily be an ambassador. The first could be expressed using an intuitively meaningful instance of the relational idiom. This state of affairs can be expressed using modal attributes as legitimate entities. Every person has the attribute z (z is necessarily a person) but nothing has the attribute z( z is necessarily an ambassador). As I will contend in § 35, the use of attributes as legitimate entities is one way to account for the indispensability of the relational idiom in modal contexts. As we shall see in the following chapter, Quine has little use for possible worlds semantics. However, many philosophers of language do and it is interesting to point out that modal attributes can be understood in such a way as to be consistent with such a semantical treatment. We will consider ordinary attributes first, 173 and then move to their modal brethren. Let K be the set of all possible worlds, let D be a function that assigns a domain of individuals to every possible world w e K, and let U be the set of all possible things. U is the union of all values of the function D. Attributes can be understood as functions that assign to each possible world we Ka subset of U. Consider the attribute of being red, R. R would be the function that assigns to every possible world w the set of all red things in w. A binary relation, such as being a parent of, would be the function that assigns to a possible world w the set of ordered pairs <x, y> such that x is a parent of y in w. Both the primitive modal predicates 'Nee1 and 'Pos' , and modal attributes can be understood in a similar fashion. Consider the attribute z{z possibly is not the author of Hamlet). Everything in w is possibly not an author of Hamlet. This attribute is the function that assigns to every possible world w the domain of w. Analogous remarks apply to the primitive modal predicate ' Pos' . Similarly, the modal attribute z{ z is necessarily a person) is the function that assigns to every possible world w the set of all things in w that are necessarily persons. One can combine a respect for a principle of individuation of attributes with an account of the meaningfulness of the relational idiom in modal and doxastic contexts. A principle of individuation of attributes has been defended. However, any lingering agnosticism over possible worlds will have to go by the wayside if one defends the method mentioned above for individuating 174 attributes. For this method of individuating attributes holds for all possible worlds, not just the actual one. Possible worlds are thus required. So be it. The individuation of attributes clears the road for an account that is analogous to the one that Quine proposed in "Q&PA." What is the desirability of introducing attributes to account for the meaningfulness of the relational idiom? Attributes fend off referential opacity for de re sentences, making straightforward (nonopaque) sense for them. Assuming that de dicto sentences in doxastic contexts are indeed referentially opaque, attributes fulfil the desired function. That is, attributes do allow one to account for the supposed indispensability of the relational idiom in doxastic contexts. Using predicates certainly does not constitute a medical breakthrough for Quine; as was discussed in Chapter II, Quine favored, at one time in his career, the metaphysical theory of nominalism. A nominalist will only use extensional entities, such as sentences and predicates. The predicates that we have just been considering have been ones of the primitive modal variety. This was done to lend substance to an expanded Quinean solution to an expanded version of the Quinean problem of having to account for the indispensability of the relational idiom in doxastic contexts. Quine's own views on modal predicates will be subjected to analysis in Chapter VIII. 175 § 25. No Entity Without Identity There is general acceptance of at least the spirit of (OR), which is that entities should not be needlessly postulated. (OR) is one ontological principle that Quine adheres to. It serves to further justify Quine in criticizing the relational idiom, as we will later see in § 31. Another ontological principle is that there is no entity without identity. Quine implies that if this identity condition cannot be met by an allegedly genuine entity, then it is meaningless to use that entity. Thus he writes that we "must separate the serious predications from the silly ones." Quine uses the following argument to support his principle that there is no entity without identity. "We have an acceptable notion of class, or physical object, or attribute, or any other sort of object, only insofar as we have an acceptable principle Q of individuation." That is, the identity criteria of an entity must be given in order that the existence of that entity be legitimately, as opposed to heuristically, postulated. One plausible rationale is that there should be criteria of identity such that it is clear when there are two entities and when there is only one. Any attempt, therefore, to account for the relational idiom by employing attributes seems doomed to failure. For the attribute of being a creature with a heart may well be co-extensive with the attribute of being a creature with a kidney. There is no doubt that Quine was aware that the postulation of attributes might be submitted as a possible means of dealing 176 with the problems one encounters when it is granted that the relational idiom makes intuitive sense. It was, after all, Quine himself who first thought of the solution when he employed attributes as a way of dealing with propositional attitude ascription statements in "Q&PA." There is the subsidiary principle that governs reductionistic attempts to order ontology. It is this secondary role which links (OR) with Quine's principle that there is no entity without identity. For (OR) will, if one already accepts Quine's principle, yield the result that intensional entities are beyond necessity and thus should not be postulated. There is an obvious linkage between the notion of being "beyond necessity" and identity criteria that are insufficient: if a potential entity has insufficient identity criteria, then it cannot play a satisfactory role in explaining linguistic or semantic phenomena. Another motivation for adhering to the principle that there is no entity without identity derives from the intuition of many philosophers that true statements are true in virtue of facts that satisfy all the rigors of discriminability. That is, certain philosophers do not want the truth of any statement to hinge on facts of the matter that involve entities that do not rigorously meet extensionality requirements. For example, the truth of statements should not, according to Quine and certain other philosophers, depend on an entity such as a proposition or an attribute that cannot be arranged in discriminable sets. This in turn supports the idea that the postulation of intensional 177 entities is, for Quine, purely for heuristic purposes and thus has no ontological significance whatsoever. This is why I think Quine's endorsement of attributes as a way to account for the significance of the relational idiom in doxastic contexts in "Q&PA" was thoroughly half-hearted. Let us scrutinize Quine's rationale for excluding entities without identity conditions from one's ontology. In "Ontology and Reduction," Richard E. Grandy writes when we are giving an interpretation of a language it is otiose to have objects in the model which are unnecessary. Thus, for example, a model with undistinguishable objects or one in which a certain subset could be dropped should not be seriously contemplated as interpretations of the language. That is not to say that they are not perfectly good mathematical objects or that they should be ruled out in studying model theory, but simply that, if we are considering the interpretation as an interpretation of a language which is used by speakers, then, the extra objects would be excess baggage since the truth value of no sentence would be changed by dropping them. Thus, the rationale behind Quine's concerns over his principle that there is no entity without identity is just one instance of his rationale for (OR) : if an entity is postulated without identity conditions, then its postulation was unnecessary. Quine's explanation of the competing factors behind the need for individuation is an appealing view. He correctly points out that there are at least two points worthy of note. Firstly, there is the point that inexact individuation "has nothing to do with 1 0 vagueness of boundaries." Though we are accustomed to physical objects being seemingly co-extensive with one another, there is a principle of individuation which constitutes a definition of 178 identity. "Specification is one thing, individuation another." Secondly, classes of physical objects can easily be individuated, "since their identity consists simply in the identity of their members." One common definition of identity is that of Russell. Russell's definition of identity works perfectly well when the entities being grouped in classes can satisfy certain criteria which are well individuated independently of class membership. For example, a sample of HjO is identical to a sample of mercury iff the two samples belong to the same classes. But the sample of f^ O can be individuated from a sample of mercury independently of the failure of the Russellian definition. The two samples could be individuated from one another on the basis of location-at-the-same-time, or of chemical composition. This sets the stage for Quine's question of the principle, if any, of the individuation of attributes. I shall argue in § 26 that there is indeed such a principle. § 26. The Individuation of Attributes Quine does not fully explain his opposition to attributes in "Q&PA," although he does come very close. He points out that there "are good reasons for being discontent with an analysis that leaves us with ... intensions.... [since they are-M. W. D.] less economical than extensions (truth values, classes, relations) and that they are more narrowly individuated. The principle of their individuation, moreover, is obscure. 179 However, Quine does not offer a complete explanation of his charge of obscure individuation concerning attributes. As we shall shortly see, a complete account of this topic was offered almost two decades later by Quine. Suppose that one were to accept the principle that there is no entity without identity. Can one give adequate identity conditions for attributes? The question is a critical one in that Quine invoked attributes, albeit for heuristic purposes alone, in order to account for the supposed meaningfulness of the relational idiom in doxastic contexts. One of his students, whom Quine referred to as 'Zedsky', applied the Russellian definition of identity to attributes. This definition states that if a and b are in exactly the same sets, then a = b. More formally, (Vz) ((a € z) « (b e z) ) -* (a = b) It can be shown that if unit sets are a feature of set theory, then the consequence in the previous conditional follows from the antecedent. Assume {\/z) (a e z) » (be z) By instantiation, a e {b} <* be {b} S i n c e b e {b} 180 it follows that a e {b}. But Vx( x e {b} « x = b) . So a = b. Zedsky's idea was that attributes A and B are identical iff the classes that they belong to are identical. Quine has referred to this solution as "exasperating" because it presupposes a prior method of individuation of the very same entity that is the subject of the exercise of individuation, namely, attributes. When applied to attributes, Russell's definition of identity is viciously circular. Russell's definition relies on the notion of set (or class) and the criterion of identity for sets is the principle of extensionality, which relies on the notion of individual. "Zedsky is evidently caught in a circle, individuating attributes in terms of identity of classes whose individuation depends on that of attributes."18 At this point, Quine anticipates a possible objection that Zedsky might make in order to save his application of Russell's **definition of identity to the individuation of attributes. Zedsky's response might centre on the fact that the Russellian definition. .. . does not really mention identity 181 of classes. When expressed in words it seems to do so: attributes are identical when the classes that they belong to are identical. But it is simpler in symbols, and says nothing of identity of classes. Quine sees this as a possible rebuttal to the charge of the vicious circularity of Zedsky's attempt to individuate attributes. This is why Quine concludes that the "real reason why the formula [the Russellian definition of identity] does not clarify the individuation of attributes is not that it mentions identity of attributes, but that it mentions classes of attributes at all." Quine makes what I take to be an indirect criticism of the relational idiom when he calls into question the legitimate ontological standing of attributes. The indirect, and partially reconstructed, criticism of the relational idiom in both modal and doxastic contexts is as follows: the use of attributes to provide an explication of the alleged meaningfulness of the relational idiom is needed; any rejection of the legitimacy of attributes is ipso facto a rejection of the relational idiom. This indirect criticism is open to three sorts of counter argument: (1) one might agree with Quine that attributes violate the dictum that no kind of entity should be employed for which there are no criteria of identity, but also hold that a modification of attributes is possible where, in principle at least, such entities could satisfy extensionalistic criteria; (2) one could hold that it is inappropriate to place identity conditions upon intensionalistic entities such as attributes; (3) one might treat attributes as being on a par with classes, which are identical 182 to each other whenever their members are identical. This alternative should be seriously considered only if no identity condition for attributes are available. Nor do I have sympathy for the third option. How then can attributes be individuated, thereby satisfying Quine's dictum that there is no entity without identity? Quine himself put forward a possibility where the question of individuating attributes from one another becomes a question of whether the same attribute is expressed by two open sentences. Quine's proposal, in "IA," was that attributes F and G are identical iff ( Vx) D (Fx « Gx) .22 This criterion of identity has at least one thing in its favor: no mention is made of set theory. Quine notices this advantage. For this reason, this definition appears to escape the vicious circularity of Zedsky's original proposal. David Lewis made a similar proposal in On the Plurality of Worlds. Quine was of course quick to dismiss his own version of the proposal: For my part, however, I find this sort of account unsatisfactory because of my doubts over the notion of analyticity-to say nothing of modal logic. (If Quine had even bothered to address Lewis's version of the proposal, one can reasonably expect him to have been equally dismissive. This identity criterion is itself an example of a de re modal formula.) It might appear that this proposal provides a definition of the identity of attributes. Kripke came very close to making a similar proposal in his 183 papers on modal logic in the early sixties. The idea would be to use possible world semantics to argue that Property F is identical to property G iff F(a) = G{a), for all possible worlds a (that is, iff "•( Vx) ( Fx « Gx)" is true in all possible worlds). ^ Curiously enough, there is a passage in Naming and Necessity in which Kripke distinguishes the notion of attribute from that of property: The type of property identity used in science seems to be associated with necessity, not with a prioricity, or analyticity: For all bodies x and y, x is hotter than y if and only if x has higher mean molecular kinetic energy than y. Here the coextensiveness of the predicates is necessary, but not a priori. The philosophical notion of attribute, on the other hand, seems to demand a priori (and analytic) coextensiveness as well as necessary coextensiveness. It is unclear to me why Kripke says what he does regarding attributes. I cannot see why the notion of attribute should require a priori and analytic coextensiveness. Such coextensiveness does not play a role in my sense of modal attributes. Nor do I believe that such coextensiveness was a feature of Quine's brief use of attributes to account for the indispensability of the relational idiom in doxastic contexts. I accept Kripke's proposal as if it were a proposal for the identity of attributes because I do not agree with his distinction between properties and attributes. Furthermore, I would hold that Kripke's properties are adequate for analyzing relational belief. A defense will now be given of my view that Kripke was in error to distinguish 184 attributes from properties in the fashion just indicated. In defending his view that properties are not the same as attributes, Kripke gives every sign of defending a particular view of the traditional notion of the attribute. This view, as we have just seen, is that a priori, and analytic, coextensiveness as well as necessary coextensiveness play an integral role in the traditional notion of the attribute. I would argue that there is little or no historical evidence that supports this view. Aristotle is probably the first philosopher to have invoked attributes. For example, he invoked attributes fairly routinely in the Metaphysics. I hasten to admit that I am not an expert on Aristotle. However, even a cursory reading of Aristotle's works indicates that neither a priori coextensiveness nor necessary coextensiveness were integral to Aristotle's use of attributes. It is true that Aristotle was in part preoccupied with questions relating to individuation, but these questions are concerned with TO the notion of matter, and not the notion of attribute. In brief, there is little reason to accept Kripke's point. Therefore, I feel justified both in not following his distinction between properties and attributes and applying his method of property individuation to the individuation of attributes. Although Quine's proposal from "IA" is an example of a de re formula, Kripke's proposal for the identity of properties is instead a de dicto formula. This is why Quine's proposal is defective, and not because of any doubts over the notion of modal logic. In terms of property identity, the de dicto is more powerful 185 than the de re. This become clear when the following model is considered: W = {a, 0} D(a) = {1} 0(P) = {1,2} F(a) = {1} = G(a) = G{ (3) F O ) = {1,2} Quine's de re " ( Vx) Q (Fx **• Gx)" will be true in a, but Kripke's de dicto "• ( Vx) (Fx « Gx)" is false in both a and (3.30 One advantage with Kripke's proposal is that de re modal theorists will have a way to identify attributes, thereby freeing them for the sort of duty that Quine envisioned for them in "Q&PA." Thus, one can agree (at least hypothetically) with Quine that there is no entity without identity and still maintain that the attribute is a legitimate entity. This result allows one to make use of attributes to account for the meaningfulness of the relational idiom in both modal and propositional attitude contexts. Later in his paper on individuation principles for attributes, Quine notes that "an old tendency still persists among philosophers to view classes and attributes as radically unlike."0 He calls this tendency to contrast these two entities a false one and a "serious misconception."J For Quine, all modal sentences are modally de dicto in nature. To repeat a previous point, the method of semantic ascent does not apply to de re modal sentences since they are not even 186 countenanced. Natural numbers, by way of contrast, could be said to be countenanced, in some sense, even by proponents of a reductionist analysis. Frege did recognize natural numbers and so too did von Neumann and Zermelo. Quine did not recognize the relational idiom, at least not until 1956, by which time he had already given his account of semantic ascent. But this last comment is wholly ambiguous, in that it conflates the question of Quine's not recognizing the relational idiom for ontological reasons with the question of Quine's not recognizing the meaningfulness of the relational idiom. To be sure, Quine does in fact oppose the relational idiom both on the grounds that it contravenes dearly held ontological principles and on the grounds that quantification into opaque contexts is meaningless. But, and this is an important 'but' , the method of semantic ascent is not strictly analogous to reductionist accounts of say, real numbers to sets of rational numbers. Semantic ascent is not a response given by Quine to address his ontological concerns regarding the relational idiom. If Quine's method of semantic ascent works at all, it works for de die to modal sentences. The principles discussed in this chapter and certain previous ones complement each other. Concerns about extensionality additionally serve to avoid a bloated ontology. Concerns about a bloated ontology prompt one to adopt other ontological principles, in this case the (OR), that will reduce the number of ontological orders. Yet I would urge that both of these principles ought to be respected. Furthermore, the principle that there is no entity 187 without identity is, intuitively at least, a sound one. This is very close to my position. I do not reject propositions, classes, or attributes simply because they are intensional entities. But I do accept a large measure of Quine's criticisms of attributes and other intensional entities. I defend the spirit of (OR). So, for example, if I could be convinced that instances of the relational idiom could be successfully replaced by instances of the notional idiom, then I would be compelled, by my support for the spirit of (OR), to abandon the relational idiom. But this is not the case. As mentioned in previous chapters, there are positive reasons for adhering to genuine instances of the relational idiom in both modal and doxastic contexts. Similar remarks apply to my evaluation of Quine's dictum that there is no entity without identity. I accept the spirit of this view but would take exception to its being applied to possible worlds. There are, I contend, good arguments to support this position. Identity principles rarely have unrestricted application. Defenders of identity principles typically hold that there are certain entities that are primitive and so are not affected by the identity principle. Frege, for example, held the view that identity was a relation between objects and thus that it was a category mistake to speak of identity between concepts. Quine, at least temporarily, recognized that his notion of language was plagued by extensionality problems. So he would not have wanted to apply his dictum that there is no entity without identity to this particular notion. In brief, Quine's dictum seems to me to 188 be satisfactory, but unrestricted application of it does not. Kripke's argument for the intuitive significance of the previous (22) The winner of the presidential election in 1968 might have lost the election, is relevant. This significance can, as we saw in the previous section, be accounted for by using binary relations such as the ones seen in § 24. (22) can be symbolized by (22* ) (3x) ( ( WPx A (Vy) ( WPy •* x = y) ) A Pos [ z ( z is not a WP) , x] ) . The main points of this chapter are concerned with extending Quine's solution in "Q&PA" that involves the deployment of attributes as a way to account for the relational idiom in doxastic contexts to an analogous treatment of the relational idiom in modal contexts and may be summarized as follows : (1) Both primitive modal predicates and modal attributes can be understood as functions that can be invoked in order to account for the meaningfulness of sentences that express the relational sense of modality; (2) this use of primitive modal predicates and modal attributes is part of an account that employs possible worlds semantics; (3) Quine does not view, in "Q&PA," the invocation of attributes to be a mere stylistic change in order to counter the effects of referentially opaque doxastic contexts; (4) given that modal 189 sentences such as (9) are as indispensable as Quine's more famous (1), an account analogous to the one contained in "Q&PA" is a definite requirement; (5) The dictum that there is no entity without identity is basically a sound one, but Quine is mistaken in using it to exclude possible worlds; (6) Quine's charge in "IA" that any attempt to individuate attributes along the lines of Russell's definition of identity will be viciously circular is a legitimate observation; (7) Quine does not fully explain his opposition to attributes in "Q&PA"; (8) Quine does offera complete articulation of this opposition to attributes in "IA"; (9) Quine considers, and immediately rejects, a method of individuating attributes; and (10) Quine is only accidentally correct in rejecting this method of individuating attributes. 190 Chapter VII. Quine's Contribution to the Problem of Transworld Identity § 27. General Remarks This chapter is primarily concerned with Quine's particular stance on the problem of transworld identity. In general, I am not convinced that the difficulties alluded to by Quine are genuine. Nor am I sympathetic with Quine's attempt to extend the problem beyond the confines of modality in order to further indict belief. The significance of a model theoretic account of QML, also known as possible worlds semantics, was briefly mentioned in Chapter I. It is only within the context of possible worlds semantics that the objection to de re modality to be discussed in this chapter makes sense. For this objection concerns the existence of the same individual in more than one possible world. It should be stressed that, barring a recasting of the semantics, those who are committed to combining a system of propositional modal logic such as S5 and quantification theory are also committed to the view that the same individual can exist in more than one possible world. It is fairly easy to show that certain de re modal formulae imply transworld identification between individuals in this the actual world and individuals in other worlds. Consider the following de re sentence: (49) (3x) DFx 191 This formula will be true iff there is, in the actual world, at least one individual x such that it satisfies the predicate F and, in all the worlds that contain x, it satisfies F. Or consider, yet again, the sentence (9) The author of Hamlet might not have written Hamlet. This sentence is true iff there is exactly one individual in the actual world who wrote Hamlet, and in at least one other possible world, this same individual failed to write Hamlet. In both cases, presupposition of transworld identity occurs when the truth conditions are spelled out. Some philosophers are convinced the objection shows that the notion of a sentence expressing the relational sense of modality must be significantly altered. Most philosophers who address this issue are content with phrases such as "the problem of transworld identity," thereby implying that there is only one problem. This view is almost certainly false. In this section, I shall establish that there are at least four versions of this objection which pose a prima facie threat to the relational idiom in modal contexts. One of these versions was developed by Quine in "WA," where he argues that there are structural constraints on predicates which prevent them from expressing transworld identity. It shall be maintained that the problem being discussed has little or no merit. It shall also be established that the equivalent problems for the relational idiom in doxastic contexts are also illegitimate. Quine was among the first philosophers of language to argue that the problem of 192 transworld identity suggests a more or less analogous problem for doxastic contexts that are relational. This position of Quine's has its beginnings in the sixties, although it becomes more pronounced in the seventies. Earlier, he had consistently treated the two sets of phenomena in a different fashion. (See the quote from Kaplan on the next page and note 6.) As I have indicated in Chapter V, Quine first endorsed a more or less symmetrical approach to doxastic and modal phenomena in WO, when he argued that both doxastic and modal distinctions disappear if one adopts either a transparent sense of belief or attempts to legitimize quantifying into modal position. It may seem odd that Quine did not explicitly endorse such a symmetrical approach to doxastic and modal phenomena earlier in his writings, but there are good reasons why Quine did not do this. Though it is true that Quine knew about the relational idiom in modal contexts as early as 1943, he did not treat the doxastic counterpart until 1956. Quine has never treated the former as being as indispensable as the latter, and I believe this goes a long way in explaining the fact that modal attributes were never used to explain the relational idiom in modal contexts, which would have been the analogue of his account for the relational idiom in doxastic contexts that he gave in "Q&PA." So when it comes to positive accounts of the relational idiom in both contexts, there is a definite asymmetry. Kaplan also notices this asymmetry: "Indeed the analogies between the relational senses of belief and necessity are so strong that I have often wondered why Quine's scepticism 193 with regard to Nee did not extend to Bel." With negative accounts (that is, with problems), however, one does detect a symmetrical approach. One sees this first in WO, and then later in Quine's papers from the 1970's. As we shall see, Quine views the problem of transworld identity as adversely affecting the relational idiom in both contexts. The problem, or at least, a version of the problem, made its first appearance in a letter by Arnauld to Leibniz: Besides, Sir, I do not know how by taking Adam as the example of a singular nature one can conceive of many [possible-M. W. D.] Adams. It is as though I were to conceive of many possible varieties of myself, which is certainly inconceivable. For I cannot think of myself without considering myself as a singular nature, so distinct from any other existing or possible [being-M. W. D.] that I can as little conceive of different varieties of myself as of a circle whose diameters are not all of equal length. The reason is that these different varieties of myself would all be distinct one from another, otherwise there would not be many of them. Thus one of these varieties of myself would necessarily not be me: which is manifestly a contradiction. Arnauld is making the claim that conceiving of transworld identity is on a par with conceiving of a logical contradiction. Does this psychological version of the problem have any merit? No. Arnauld's objection could be restated as a problem about conceiving of the same entity at different times, rather than the same entity in different worlds. Superman at t is different, in certain respects, from Clark Kent at t' , even though they are one and the same being. (Superman presumably has as much a unitary nature as does Adam. ) It does not follow that there is a Superman at some time who is not Superman. Beings have a unitary nature over time, and it is difficult to see how one could fail to conceive 194 of Clark Kent and Superman as identical, given the release of sufficient information. It is not inconceivable that beings have a unitary nature which spans possible worlds as well. Thus, the psychological version of the problem of transworld identity has no merit. A more frequently occurring version of the problem concerns the adequacy of proposed criteria of identity across possible worlds. Typically, the issue of whether such criteria can be adequately spelled out is the main version of the problem which preoccupies modern philosophers. I shall call this difficulty the ontological version of the problem of transworld identity. The ontological version is distinct from Arnauld1s psychological problem in that the question is not "Does the notion of the same object inhabiting different possible worlds result in a contradiction?," with Arnauld's fallacious answer, but rather "What additional entities must be postulated in order that it may be said that the same individual exists in different possible worlds?" In Quine's view, there is but one possible world, namely, the actual one. The ontological version of the problem is thus an application of Occam's Razor, which will be further discussed in the following chapter. The psychological and ontological versions of the problem of transworld identity could be combined into yet another version of the problem: a conceptual one. That is, Arnauld's point about the inability to conceive of transworld identity could be restated in such a way as not to depend upon a logical impossibility. 195 Instead, the inability to conceive of transworld identity could be said to depend upon the absence of a clear set of criteria for identity. As far as I know, no such version of the problem of transworld identity has ever been advanced. In that such a combination yields such a fallacious problem, its lack of submission does not surprise me. Plenty of human beings routinely manage to conceive of entities which fail to have a clear set of criteria for identity. It might be thought that the difficulties with this particular version do not end there. The notion underpinning this difficulty is that a belief is identified through the facts it corresponds to. If, like most analytic philosophers, one accepts the correspondence theory of truth, and if one supported the conceptual version of the problem of transworld identity, then one would, apparently, be forced to conclude that there was no such thing as a false belief. According to this view, "only beliefs that do correspond to facts--namely, true ones--can have identities (can exist) ..." For false beliefs are merely beliefs which do not correspond to facts. But this particular piece of heresy can easily be stopped. "But a Fregean, say, would identify a belief as a (purely conceptual) proposition ... Frege ... could still hold a correspondence theory of truth ..." Related to the ontological version of the problem of transworld identity is an epistemological difficulty which I shall call the verification version. Here the question becomes, "what set of criteria for identity across possible worlds must prevail in order 196 to determine that object Oj in world Wj is the same object as O2 in W2?" Such determination is decided by properties of the alleged transworld individuals, such as continuity or chemical makeup. Quine's own version of the problem combines elements of both the ontological and the verification ones. Recall that, for Quine, an object lacking criteria of identity does not exist. He writes 1 ? in "IA" that there "is no entity without identity." Thus the importance of a clear set of identity conditions for objects which are alleged to be in more than one possible world. Leonard Linsky believes that there is a way to escape the ontological version of the problem of transworld identification problem. He writes that we must distinguish between the claim that we are unable to state a satisfactory explicit criterion for identifying across possible worlds and the claim that we are unable to make sense of such identifications." He correctly points out that the second claim is neither true nor is it entailed by the first one. A satisfactory criterion for identifying objects inhabiting different possible worlds is demanded only by those who adhere to the Quinean principle that without identity, there is no entity. There thus appear to be at least four separate versions of the problem of transworld identity where it is loosely implied that there is only one. Furthermore, there appears to be, for Quine, a fifth version of the problem. Quine' s own version of the problem is most interesting and it requires detailed analysis in order to provide a proper evaluation of it. Accordingly, I wish to 197 introduce further Quine's own brand of transworld identification criticism. I think it is an amalgamation of the ontological and the verification versions. § 28. Quine's Comments in "WA" Quine' s main contribution to the problem of transworld identity takes his liberal notion of a physical object as his point of departure. Quine shares with Nelson Goodman, among others, the view that physical objects are made up of a vast array of temporal parts. As I shall show, Quine does not concede that quantification is as unproblematic in the domain of possible worlds as it is in the domain of time slices. This is an area of complete disagreement between the author and Quine; as indicated above, reasons alluded to by Arnauld do not satisfy me that the analogy between transworld, and transtemporal, identity breaks down. I shall show that Quine*s position regarding the collapse of this analogy is similarly untenable. It is the verification version that I contend has resulted in an extremely elaborate ignoratio elenchi. David Lewis, in particular, has fashioned his metaphysical theory of counterparts in order to cope with the verification version of the problem of transworld identity. Quine is critical of the notion of counterparts, and holds that Despite man's stubborn body-mindedness, there are good reasons for the more liberal ontology of physical objects. All these objects, when I quantify over individuals, are the values 198 of my variables. What, then, would be analogous values of variables if one were to quantify over individuals in all possible worlds? Simply the sums of physical objects of the various worlds, combining denizens of different worlds indiscriminately. One of these values would consist of Napoleon together with his counterparts in other worlds, if ' counterpart' made sense; another would consist of Napoleon together with sundry utterly dissimilar denizens of other worlds. Immediately afterwards, Quine adds that "quantification over objects across possible worlds does not require us to make any sense of 'counterpart' . "^ He does not sufficiently elaborate upon this point, but what I think Quine is objecting to is the exotic definitions of a counterpart to which David Lewis has resorted to in both his Counterpart Theory and in his theory of modal realism. That is, if indeed quantification is as straightforward over possible words as it is over temporal slices, then there would not have to be a separate account of the objects of quantification. They would be as unproblematic as physical objects existing over time, and no separate account is required for transtemporal objects of quantification. So there should be no separate account of a transworld object of quantification (a definition for what does and what does not constitute a genuine counterpart). Quine's liberal notion of a physical object includes temporal parts of objects. Would it be fundamentally problematic to extend this liberal notion to also include transworld parts of objects? In "WA," Quine argues against the possibility of such an extended approach: Just as any two momentary objects in different moments are shared as time slices not by just one time-extended object but by countless ones, so any two physical objects in different 199 worlds are shared as realizations not by just one intermundane object but by countless ones. Quantification is as straightforward over the one domain as over the other, unless there is some independent trouble with possible worlds. But Quine is confident that independent trouble with possible worlds does indeed arise, owing to the predicates one must use in de re modal sentences. He argues that individuation is alive and well with quantification over time. ... in our own world the identification of a physical object from moment to moment makes sense only relative to the principle of individuation of one or another particular predicate-usually, though not necessarily, the predicate 'body' or one of its subordinates. This view commits Quine to supporting a set of sentences that are analogous to instances of the relational idiom, namely, the set of sentences in temporal contexts. For Quine would not take exception to sentences such as (50) The Cayster was bathed in by Heraclitus over the course of two days, even though (50) involves transtemporal identity. Not only are these sentences unquestionably extensional, they are so without the reliance on postulated entities that are primitive. Instead, 10 Quine relies on his notion of temporal parts. Quine also claims that the all-important predicates used in discriminating physical objects over time involve continuity through on distortion, displacement, and chemical change. Here the analogy between transtemporal and transworld identification ends, according to Quine. "The devastating difference is that the series of 200 momentary cross sections of our real world is uniquely imposed on us, for better or for worse, whereas all manners of paths of continuous gradation from one possible world to another are free for the thinking up." In Naming and Necessity, Kripke briefly, and decisively, points to a common misconception regarding possible worlds. In Kripke's own words, possible worlds "are stipulated, not discovered by powerful telescopes." Kripke simply does not agree with Quine*s view that identification of individuals within possible worlds is a legitimate enterprise. Possible worlds are stipulated, Kripke would (I think) say, by certain instances of either the relational idiom in modal or doxastic contexts . Concerns about identification do not arise, and it is for this reason that the problem of transworld identification, and in particular Quine's version of the problem, is not a genuine one. Kripke's view that identification of an actual world object that is said to exist in other possible worlds is not a genuine problem is in apparent conflict with Quine's dictum that there is no entity without identity. Yet I would be reluctant to hold that this conflict is anything more than an apparent one. Consider the case of the author of Hamlet, who, it is alleged possibly did not write Macbeth. According to Kripke' s view, we fix the reference of the actual individual who wrote Hamlet. It is at this point, before other possible worlds are stipulated, that concerns about identity (and this includes Quine' s concern about identity criteria of individuals) may or may not be appropriate. But once the other 201 possible worlds are stipulated, the identity of the author of Hamlet becomes no more and no less problematic than before. This result generalizes. There are no further problems of identity across possible worlds than there are within a single (that is, the actual) world. There are at least three other points to bear in mind when discussing Kripke's view on the stipulation of possible worlds. Firstly, a "possible world is given by the descriptive condition we associate with it." This picture of possible worlds allows Kripke to supply meaning to utterances such as "In some other possible world I would not have given this lecture today. " He writes that we just imagine the situation where I didn't decide to give this lecture or decided to give it on some other day. Of course, we don't imagine everything that is true or false, but only those things relevant to my giving the lecture; but, in theory, everything needs to be decided to make a total description of the world. * Though Kripke never explicitly says so, comments such as the citation just given seem to me to strongly imply that Kripke is a conceptualist. Secondly, there is Kripke's emphasis upon the objects in the actual world. There is no reason why we cannot stipulate that, in talking about what would have happened to Nixon in a certain counterfactual situation, we are talking about what would have happened to him. The significance of this view is that it would be nonsensical to attempt to identify Nixon in a possible world in which he is stipulated. 202 Thirdly, Kripke defends his notion of the stipulation of possible worlds against the following charge: It comes down to the same thing, because whether Nixon could have had certain properties, different from the ones he actually has, is equivalent to the question whether the criteria of identity across possible worlds include that Nixon does not have these properties. Kripke points out that the two questions are not equivalent since "the usual notion of a criterion of transworld identity demands that we give purely qualitative necessary and sufficient conditions for someone being Nixon. "ll Now, Kripke is not bothered by the demand for necessary conditions for someone being Nixon. Such necessary conditions are limited only by our imagination: For example, supposing Nixon is in fact a human being, it would seem that we cannot think of a possible counterfactual situation in which he was, say, an inanimate object; perhaps it is not even possible for him not to have been a human being. Then it will be a necessary fact about Nixon that in all possible worlds where he exists at all, he is human or anyway he is not an inanimate object. This has nothing to do with any requirement that there be purely qualitative sufficient conditions for Nixonhood which we can spell out. Notice that this sort of essentialism does not imply I.A.E., as Quine supposed in Chapter IV. Nor is this sort of essentialism as innocuous as B.Q.E. Rather, it is what I called M.A.E. I believe that this sort of essentialism is plausible. The stipulation of possible worlds does not, however, extend to the actual world. The actual world is indeed discovered. (This is another difference with Lewis' theory of modal realism, where the actual world is but one of many possible worlds; Lewis accords no primacy to the actual world. ) This is the only difference between transworld and cross-temporal identification, and it is not very 203 devastating. It simply will not be appropriate to ask whether this or that object, in this or that possible world, is identical to this object in the actual world, and vice versa. In short, one will not be required to pick out an object from a world distinct from the actual one and ask of it, "Is this object a counterpart to an object in the actual world?" Linsky made one of the more compelling arguments in support of transworld identification in "Reference, Essentialism, and Modality": Suppose someone tells us 'I did not miss this morning's lecture, but I might have' . The intelligibility to us of this statement depends upon our ability to make sense of the idea that the subject of the statement is identical with an individual in another possible world: one in which he misses this morning's lecture. Indeed, the statements, 'I did not miss this morning's lecture, but I might have' and 'I did not miss this morning's lecture, but there is a possible world in which I did, ' are full paraphrases of each other. The latter statement explicitly identifies its subject with an individual in another possible world; if this makes no sense to us, neither does the former statement. But the former statement does make perfect sense. To the extent that we understand such assertions we are able to make sense of the idea of identical individuals in different possible worlds; and we may exploit this understanding to give intuitive meaning to the statement of quantified modal logic. So much the better for quantified modal logic; for this is good enough for modal logicians to be going on with.25 Linsky is, I think, restating the position originally penned by Kripke: possible worlds are stipulated and this stipulation presupposes transworld identity. Linsky's case also brings home a point contra David Lewis: when one utters 'I did not miss this morning's lecture, but I might have', it is highly counterintuitive to suppose that there is a counterpart of me that missed this morning's lecture (that is, someone nonidentical with me), as 204 opposed to its being just plain old identical me who might have missed this morning's lecture. Kripke' s point about the stipulation of possible worlds holds only in the case of the relational idiom, as opposed to the notional idiom, in modal contexts. This point may be thought to be too obvious to acquire argumentation, but I will provide it nonetheless. Let us consider again the modally ambiguous sentence (2) The author of Hamlet might not have written Macbeth. Using TDD, the notional reading of (2) is (2*****) 0[(Vx) ((Hx A (3y) (Hy - x = y) ) A ^ Mx] . In other words, there is a possible world in which an author of Hamlet (who may or may not be identical with the author of Hamlet in the actual world) did not write Macbeth. The relational reading of (2) is, once more ( 2 » • * ) (3x) ( ( Hx A (Vy) ( Hy - x = y) ) A 0^ Mx) In other words, the individual in the actual world who wrote Hamlet did not write Macbeth in some other possible world. So, whereas (2*****) does not presuppose transworld identity, (2''') most certainly does. § 29. A Corresponding Position Concerning Belief Worlds Towards the close of "WA, " Quine shows how an analogous problem 205 with predicates arises for those who hold that belief contexts can be quantified into. That is to say, philosophers of language who contend that there are de re propositional attitudes also have to confront the objection that belief states, like possible worlds, are free for the thinking up and are thus unable to be individuated: Each belief world will include countless bodies that are not separately recognizable objects of the believer's beliefs at all, for the believer does believe still that there are countless such bodies. Questions of identity of these, from [belief-M.W.D.] world to world, remain as devoid of sense as they were in the possible worlds of alethic modal logic. Quine's objection to identification from possible world to possible world can be more fully applied: now there will be an equally unpalatable choice between belief worlds whose objects cannot be individuated and Kaplan's suggested singular term for belief contexts, the vivid designator, which Quine also holds to be meaningless. I would suggest that a response similar to the one given by Kripke concerning transworld identification can also be given concerning Quine's objection based on trans-belief-world identification. That is, the identification of belief worlds (I shall later refer to them as de re belief states) is stipulated and we do not have to worry about identifying individuals which manage to inhabit more than one belief world. Quine's argument against transworld identification is that one must choose between two unpalatable options: (a) possible worlds that are free for the thinking up, and thus escape individuation and (b) singular terms, such as Kripke's rigid designator, which 206 are alleged by Quine to be utterly dependent on context, and are 11 thus meaningless. I think an argument could be devised that would show that distortion, displacement, and chemical change can survive transworld identification and that Quine's views on unique imposition of single world objects begs the question. That is, I think that a calculus of individuals could be devised in such a way as to allow for transworld identification. The basic idea is simply to expand Quine's liberal notion of a physical object so as to include a transworld dimension. A physical object is simply the sum of all its temporal and modal parts. Does this mean that this solution is no different from David Lewis' Counterpart Theory? It certainly does not. It is interesting that both Quine and Kripke regard the notion of a counterpart as being somewhat superfluous, although for different reasons. Quine points out that Kripke's criticisms of Lewis' counterpart theory turn on the latter's view that possible worlds are given to us purely in a qualitative manner. Lewis bases the notion of a counterpart in terms of important resemblances that hold between objects in the actual world and objects in other worlds. Kripke is critical of the notion of counterpart in the following: Surely these notions are incorrect. To me Aristotle's most important properties consist in his philosophical work, and Hitler's in his murderous political role; both...might have lacked these properties altogether. Surely there was no logical fate hanging over either Aristotle or Hitler which made in any sense inevitable that they should have possessed the properties we regard as important to them; they could have had careers completely different from their actual ones. Important properties of an object need not be essential, unless importance' is used as a synonym for essence; and an object could have had properties very different from its most striking 207 actual properties, or from the properties we use to identify it.32 Thus both Kripke and Quine regard the notion of a counterpart as being beside the point. Yet I am satisfied that Kripke's position is correct and thus, there is no need for such a calculus. (One need not have a pronounced affection for desert landscapes to be reluctant to postulate needlessly entities; such needless postulation appears to me to be precisely what one would get from a calculus of individuals that would expand Quine's notion of a physical object. ) Quine's charge that possible worlds are free for the thinking up just does not apply to Kripke's notion of a possible world. My solution, then, to Quine's version of the problem of transworld identity consists not in making Quine's broad conception of a physical object even broader in order to facilitate identification, but rather in denying that there is a legitimate problem in the first place. In view of my acceptance of Kripke!s response to the verification version of the problem of transworld identity, I contend that Quine's position regarding the structural damage done to predicates by the de re modal idiom is nothing more than a further epicycle of a response to a problem which does not arise in the first place. If Quine 's position is extended to cover instances of the de re doxastic idiom, Kripke's response to the pseudo-problem of transworld identity can be successfully reused. By way of conclusion, then, it would appear that two of Quine's criticisms of the relational idiom (one in each context) can be 208 successfully addressed. 34 209 Chapter VIII. Quine on "IR" § 30. General Remarks This chapter is devoted to the ramifications of Quine' s "IR," a paper that made its appearance in 1977. Quine begins "IR" by temporarily adopting the necessity predicate 'Nee' to provide an extensional treatment of modality. His use of sentences and predicates contrasts with his use of attributes in "Q&PA," where he was attempting to come to terms with the relational idiom in doxastic contexts. In "Q&PA," Quine chose not to use predicates and sentences because of concerns regarding identity criteria for language. Quine's solutions that involve intensional entities should be seen as conditional ones: if, despite Quine's plethora of concerns regarding the relational idiom, one is still determined to quantify into modal contexts, then either the solution proposed in "Q&PA" or in "IR" is open to consideration. Despite his misgivings towards semantical reformulations in "Q&PA," Quine proceeds to outline his extensional treatment of necessity in "IR." Quine's mind appears to have changed regarding the earlier problem he had with semantical reformulation, as this problem is not mentioned in "IR." The concern mentioned in the previous paragraph regarding identity criteria for language was his problem with semantical reformulation. The impression might be taken, given Quine's concerns regarding identity criteria in Chapter VI, that 210 such concerns were exclusively directed by Quine against intensional entities. Such an impression is false, for Quine directed such concerns against extensional entities in "Q&PA." "IR" does contain a new criticism of the relational idiom, one that Quine deliberately applies to the relational idiom in both modal and doxastic contexts. This argument, which concludes by "writing off" the relational idiom in both contexts, actually could be said to be a rejoinder of sorts to my charge of inconsistent skepticism in Chapter II; Quine certainly is consistent in this renunciation of the relational idiom in both contexts. This later consistent skepticism on Quine's part was the result of Quine having perceived some of the parallels between doxastic and modal phenomena that I wrote about before. Quine does not, it is true, argue for the interchangeability of his own modal puzzle and Russell's doxastic puzzle as was done in Chapter II. But he does recognize the parallels between such phenomena in terms of the intensional versus extensional solutions mentioned on the previous page. In "IR," Quine provides a translation schema between his necessity functor ' [T and the necessity predicate 'Nee' : •FxjX^ ... xn E Nec( ' F' , xl, Xp . . . , xfl) J Quine notices the similarities between the notion of belief, where "the underlying logic ... is extensional ... [and] ... the case of 'Nee'.* A similar translation schema is then given between an intensional belief sentence functor and an extensional belief predicate: 211 B^FXjX^ ... xn -' T o m believes ' F" of x^, Xp . .., xn.' More than anywhere else, "IR" shows the extent to which Quine believes that doxastic and modal contexts of the relational idiom are interconnected. It is argued in the following section that Quine's criticism of the relational idiom in this paper is an enthymeme, with the suppressed premise being that the notional idiom, in both contexts, is in some way prior to the relational idiom. Both idioms are used to express states of affairs that are concerned either with modality or belief. § 31. A Critique of "Intensions Revisited" A cursory reading of "IR" reveals Quine to be supporting a comparatively uncomplicated objection to the relational idiom. Realizing that de re statements in modal and doxastic contexts share certain characteristics, Quine feels compelled to reject, albeit with some hesitation, the relational idiom in doxastic contexts while simultaneously, this time with no hesitation, rejecting the relational idiom in modal contexts. He realizes that a commitment to modality de re would presuppose some version of the detested doctrine of essentialism; consistency forces Quine to object to de re statements of belief. The relational idiom in modal contexts forces the use of essences, which, for Quine, are meaningful only in certain cases. He further claims that there 212 is no satisfactory method to discriminate between acceptable and unacceptable uses of essences. However, such an interpretation of Quine's position in "IR" does not reveal what I take to be an obfuscatory aspect of "IR." This objection is concerned with exportation-a logical step from an opaque sense to the cognate relational sense. Exportation uses de dicto statements as inputs and de re statements as subsequent outputs. Quine's views on exportation appear confusing to me. Quine writes in "IR" that "we must find against exportation" because of an argument given by Sleigh. Quine concluded from this that the analysis he had given in "Q&PA" of de re doxastic sentences needed to be changed. The acceptance of this argument against exportation marks something of a reversal on Quine's part from 1956, when he wrote "Q&PA." There, he held that this form of inference from a de dicto doxastic sentence to an analogous de re doxastic sentence was "implicative"--by which Quine meant that this form of inference was "generally valid." For example, the sentence (51) Ralph believes Bernard to be wearing a brown coat, is obtained using exportation from the de dicto doxastic sentence (52) Ralph believes that Bernard is wearing a brown coat. The following passage is taken from "IR": In the 1956 paper, I dwelt on the practical difference between the de dicto statement: 213 (8) Ralph believes '(3x) x is a spy' and the de re statement ' There is someone whom Ralph believes to be a spy' , that is, (9) (3y) Ralph believes 'spy' of y. I noted also the more narrowly logical difference between the de dicto statement (10) Ralph believes 'Ortcutt is a spy' and the de re statement: (11) Ralph believes 'spy' of Ortcutt, .... However, if we transcribe (10) and (11) into terms of 'Br' according to the foregoing patterns, we get: (12) Br(0rtcutt is a spy) (13) (3x) (x = Ortcutt . Br(x is a spy)), and here the existential force of (13) would seem to belie the validity of the exportation. Sleigh, moreover, has challenged this step on other grounds. Surely, he observes (nearly enough), Ralph believes that there are spies. If he also believes, as he reasonably might, that (14) No two spies are exactly of the same height then he will believe that the shortest spy is a spy. If exportation were valid, it would follow that Ralph believes ' spy' of the shortest spy, and this, having the term 'the shortest spy' out in referential position, implies (9). Thus the portentous belief (9) would follow from trivial ones, (8) and belief of (14). Quine then applies this argument to the supposed difference between de dicto and de re modality. Let us consult incidently the analogues of (10) and (11) in modal logic. Looking to the transcriptions of (12) and (13), we see that the analogous modal structures are ' DFa' and ' (3x) (x = a . DFx) ' . Does the one imply the other? Again the existential force of the latter would suggest not. And again we can dispute the implication also apart from that existential consideration, as follows [abbreviating (14) and 'there are 214 spies' in conjunction as ' 141 ]: (15) D(14z>. the shortest spy is a spy), (16) (3x) (x = the shortest spy . •(14 3. x is a spy)). Surely (15) is true. On the other hand, granted (14), presumably (16) is false; for it would require someone to be a spy de re, or in essence. The argument against exportation allows Quine to conclude that there is almost no practical difference between the relational idiom and its less controversial brother, the notional idiom. Before he states this conclusion, Quine builds further upon his argument against exportation with the simultaneous attack on Kaplan's notion of a vivid designator^ and its analogue, Kripke's notion of a rigid designator. Quine is concerned about the inability to distinguish between admissible and inadmissible cases of exportation. Exportation involves moving from a notional formula to a relational formula, where both formulae have roughly the same subject/object format. He considers, and subsequently rejects, the use of vivid designators as a means of differentiating between the two sorts of cases. The connection between exportation and vivid designators resides in Kaplan's claim that the step of exportation is legitimate only in cases where vivid designators are involved. Kaplan introduced the notion of a vivid designator, which was designed to be a doxastic counterpart to Kripke's rigid designator. I submit that exportation results in unacceptable results without exception and that exportation is not useful as a criterion for the supposed lack of practical difference between the two idioms. 215 It is important to note that Quine's claim that 'DFa' is an analogue of the occurrence of (10) in "IR" is, according to certain philosophers, false. There is no question regarding the status of (10): it is, as Quine says, a de dicto statement. However, the inclusion of the individual constant 'a' in 'QFa' would render the statement, again according to certain philosophers, to be a de re one. Forbes holds such a position. However, the chief fault with Quine' s argument lies not with its claims, dubious or otherwise, regarding its applicability to both doxastic and modal contexts. On the other hand, Sleigh's/Quine's argument does succeed in showing a defective aspect of exportation. In general, exportation is not a legitimate step. It should be noted that such a view could be damaging to an adherent of the relational idiom only if it were conceded that the relational idiom is, in some unknown fashion, dependent upon the notional idiom. Quine does not bother to argue for such a position. To be fair, the initial conclusion that Quine draws is that, "we must find against exportation." What Quine does not argue for, and this I find extremely puzzling, is the relevance (in terms of a wholesale rejection of the relational idiom, in both doxastic and modal contexts) of his finding against exportation. Without an adequate explanation of the supposed relevance of his argument against exportation, one is forced to conclude that Quine's point is an incomplete one. Quine then expands a point that he first made in "R&M" to the effect that the role played by context is of vital importance. 216 In "IR" Quine held that both vivid and rigid designators are "utterly dependent on context." That is, Quine holds that cross-quantification into both modal and doxastic contexts is utterly dependent on context. The notion of knowing or believing who or what something is, is utterly dependent on context. Sometimes, when we ask who someone is, we see the face and want the name; sometimes the reverse. Sometimes we want to know his role in the community. Of itself, the notion is empty. Quine still believes that de re statements are meaningful only in context because of their existential force. As I urged in Chapter I, we can meaningfully utter sentences such as (1) There is someone whom Ralph believes to be a spy. That is in a narrow context where we want to know the name of the spy and not some universal denotation of him, as was described earlier. That is the Quinean view. His earlier point was that the notion of necessary consequence also was dependent on context. The significance of this point is, however, very dubious. Plenty of notions are viable only within a certain context. The notion, for example, of loyalty towards one' s community, is utterly dependent on which community one happens to belong to. One would not be able to conclude that loyalty towards one's community is a worthless concept. In short, I do not agree with Quine that being dependent on context is somehow reprehensible. The passing of Kaplan's vivid designator leaves us defenseless against Sleigh's deduction of the strong (9) from (8) and belief of (14). Thus it virtually annuls the seemingly vital contrast between merely believing there are spies and suspecting a specific person." 217 The question may be asked, "Why does exportation even matter; why is this form of inference even relevant?" Earlier in "IR," Quine explains the de re/de dicto distinction in terms of the intensional necessity functor '•' that is fairly illuminating as regards this question: When the term concerned is a variable, there is nothing to distinguish; de re is de rigueur. When it is not a variable, we keep it in the scope of ' •' for de dicto: D(number of planets is odd) (false) and bring it out thus for de re: (1) (3x) {x = number of planets. •( x is odd)), (true)'* Notice that Quine did not explain the distinction by saying that when the term concerned is a variable we keep it out of the scope of the necessity functor and bring it in when the term is not a variable. This might not appear to be significant, but it seems worth bearing in mind given the fact that many speakers of English use 'and' as verbal shorthand for 'and then1. I submit that Quine has engaged in such shorthand in the passage just cited. That is I think that Quine is supposing that when the term concerned is not a variable, we keep it in the scope of ' •' for de dicto modality, and then bring it out thus for de re modality. Quine appears to view examples of the notional idiom as being, both figuratively and literally, prior to examples of the relational idiom. The former are figuratively prior to the later in that there is, for Quine, a sort of ontological primacy associated with de dicto modality. Quine's criticisms of the relational idiom in modal contexts merges with a very time honoured philosophical problem, 218 namely, that of the existence of nonbeing. How can one meaningfully utter a statement containing the word 'not'? Quine draws the parallel in the following passage in "On What There Is": Possibility, along with the other modalities of necessity and impossibility and contingency, raises problems upon which I do not mean to imply that we should turn our backs. But we can at least limit modalities to whole statements. We may impose the adverb 'possibly' upon a statement as a whole, and we may well worry about the semantical analysis of such usage; but little real advance in such analysis is to be hoped for in expanding our universe to include so-called possible entities. I suspect that the main motive for this expansion is simply the old notion that Pegasus, for example, must be because otherwise it would be nonsense to say that he is not. To understand better this point, compare the two following sentences: (51) It is possible that: there is a fat man in the doorway, and (52) There is a possible man in the doorway. Quine is implying that the latter sentence constitutes a violation of the spirit of Occam's Razor. He concludes by arguing that such violation is done solely because of the problem, first pointed to by Plato, that it is nonsense to say that a thing is not the case (for this implies, however loosely, that the thing in question does not exist). Obviously, I disagree with this latter contention; my reason here for expanding the ontology would have more to do with the intuitive meaningfulness of the relational idiom in modal 219 contexts. Because Quine's argument against exportation is formulated in such a way as to apply to both modal and doxastic contexts, I think it can be concluded that Quine's views of the ontological premise of de dicto modality also apply to de dicto belief. De re belief would be as unacceptable to Quine as would de re modality. It is difficult to explain why Quine would have held that exportation was valid. A related explanation to the one just given is this: in a formal language L names standardly have existential import. That is, a model M of L must (in standard quantifier logic) assign some member of the domain to a name. The argument shows instead that the step of exportation from the de dicto to the de re is, in certain circumstances, one that is unacceptable. I fail to see how this finding against the step of exportation in any way shows that the distinction between relational and notional idiom is more apparent than real. On the other hand, I think I can provide at least two reasons as to why Quine may have held that this argument succeeded. Firstly, Quine quite possibly held that exportation from de dicto statements was on a par with the principle of existential generalization; violation of so sacred a logical principle would be untenable. Such a rationale would, I think, have to be limited to doxastic contexts. Secondly, a de dicto statement not only says what the cognate de re statement says, it says more. Sentences such as our previous (19) Penelope is unaware that Tully denounced Catiline, 220 involve reference to Tully and Catiline and thus are true if they are interpreted as de re, but (19) may be either intended or construed as de dicto. In this latter case, in addition to informing us as to who Penelope's lack of awareness is about, it also tells us that Penelope would be willing to express, when suitably prompted, this lack of belief by using ' Tully1 . Accordingly, the de dicto construal conveys more information, namely, that Penelope would use the name 'Tully' in expressing her lack of awareness; (19) still conveys to us what it would when intended de re (where the interest is only in whom Penelope believes, or rather fails to believe, 'x denounced Catiline' of, not in how Penelope would refer to that person). So on this picture, the de dicto (19*) implies the de re (19*). If this seems fanciful, recall that (a) Quine (earlier, in "Q&PA") unquestionably assumes, but does not argue for, the view that exportation is required for the relational idiom in doxastic contexts; that (b) some sort of rationale underlying this view is required; (c) Quine's assumption in favor of exportation was dropped in "IR" with arguments against exportation in both modal and doxastic contexts; and that (d) the author has absolutely no interest in defending this rationale. With all due respect, Quine should have produced an explicit argument in favor of exportation, if not in "Q&PA," then certainly in "IR." What he did produce in "IR" were two arguments against exportation—which is curious in itself since neither of the two rationales given above for exportation has a modal analogue. So Quine has engaged in 221 philosophical overkill; he has an argument against exportation in modal contexts even though no such argument need be given. Quine apparently shares with John Searle, among others (including myself), the view that to accept the relational idiom in doxastic contexts one would also have to countenance the relational idiom in modal contexts, something which he is not prepared to do. The relational idiom, in both contexts, is to be rejected "except as idioms relativized to the . . . situation at hand."1' That is, Quine holds that the relational idiom is meaningful only in cases where the context is already understood. Quine' s position in this regard indicates, I think, that he prefers a consistent treatment of the relational and notional idioms in both modal and doxastic contexts. "We remain less cavalier toward propositional attitudes than toward modal logic only in the unquantified or de dicto case, where the attitudes are taken as dyadic relations between people or other animals and closed l it sentences." One difference, though, between Quine's treatment of the relational idiom in modal as opposed to doxastic contexts is that he has no reluctance in abandoning the former while he does display reluctance in abandoning the latter. Quine thus leaves himself vulnerable to the criticism that he is at least somewhat inconsistent." 222 § 32. Vulgar Occurrence of Terms in the Relational Idiom As was mentioned above, the relevancy of Quine's argument against exportation can be challenged. There is another way to argue in favor of Quine's view that the step of exportation must be abandoned. A better way is with the use of a fictional entity, rather than the factual one (that is, a spy) that is employed by Quine's argument. This avoids a certain degree of ambiguity that is inherent in Sleigh's/Quine's argument. But a different set of conclusions follow from this new argument against exportation. More importantly, the primary purpose of this new argument is very distinct from that of Quine/Sleigh. In fact, by breathing new life into the argument against exportation one obtains a conclusion which is diametrically opposed to Quine's view on utter dependency on context. Robinson has given a doxastic variant of Quine's/Sleigh*s argument against exportation. He does this to undermine Quine's claim that the meaning of a sentence containing "believes" depends utterly on its context. Robinson contends that the meaning of a de re doxastic sentence is not utterly dependent on context. One way to reach this conclusion is to demonstrate that the occurrence of the term "Mark Twain" in (53) Tom believes Mark Twain to have written Huckleberry Finn. is vulgar, in Kaplan's sense. If an occurrence of a term is vulgar, then it is open to substitution and existential 223 generalization. The assumption will be made that vulgarity implies nondependence on a context. "Utter" dependence on context appears to preclude the sort of cross-quantification that one would find with sentences expressing the relational sense of belief. After giving Robinson's doxastic argument in favor of the occurrence of the term "Mark Twain," two modal versions of this argument are supplied by the author. Before showing that the occurrence of the term "Mark Twain" is vulgar, Robinson firstly argues in favor of the validity of existential generalization for de re doxastic sentences. For example, that the inference from the de re sentence (54) Ralph believes Bernard to be wearing a brown coat, to (55) There is someone whom Ralph believes to be wearing a brown coat, is valid. Robinson rightly observes that, The question whether existential generalization preserves truth for sentences that contain the expression "believes ... to be" might better be considered in cases where the existential generalization is false. For example, consider the sentence (56) There is something Sean believes to be a leprechaun as a generalization of the following de re doxastic sentence: (57) Sean believes the shortest leprechaun to be a leprechaun. 224 Because there is a definite description in (57), we can use Russell's TDD even though the de dicto doxastic sentences (58) Sean believes that there are leprechauns and (59) Sean believes that no two leprechauns have the same height, might be true. Therefore, it seems that the de dicto doxastic sentence (60) Sean believes that the shortest leprechaun is a leprechaun is true. (57) follows from (60) by exportation. This shows that exportation is an invalid rule of inference. (60) is a true sentence, even though it concerns a false belief of the deluded Sean. (57) is false, since there is no shortest leprechaun for Sean to be believe it to be a leprechaun. At this point, Robinson uses the falsity of (57) in a radically different fashion than Quine, even though they are in accord concerning the invalidity of exportation. Robinson observes that the inference from (57) to (56) does not constitute a counter-example to existential generalization and this particular inference "seems to be the strongest candidate for such a counter-example." It follows that existential generalization is a valid form of inference for sentences containing the expression "believes .. . 225 to be." The argument for the validity of substitution has already been given in Chapter II, where it was suggested that (19 *' ) Tully is such that Penelope is unaware that he denounced Catiline, allows us to obtain (19''*) Cicero is such that Penelope is unaware that he denounced Catiline, by substituting an occurrence of the term "Tully" with an occurrence of the coreferential term "Cicero. " Recall that the story in Chapter II was that Penelope was prepared to accept "Cicero denounced Catiline" as true but was not prepared to accept "Tully denounced Catiline." Nevertheless, it was held, both (19 * * ) and (19 * * *) are true ascriptions of relational belief and stand in marked contrast with both (19) Penelope is unaware that Tully denounced Catiline, and (19**) Penelope is unaware that Cicero denounced Catiline. Both (19) and (19**) are notional ascriptions of belief. Note that the earlier substitution of the occurrence of the term "Tully" with its coreferential counterpart "Cicero" would here be invalid, since (19) is true whilst (19**) is false. Note too that (19) and 226 (19**) would be complete sentences if one were to delete the initial four words but that the same would not be true if the first four words were deleted from (19'') and from (19 * • * ) . That is, the initial 'that' in the first two sentences ((19'*), (19*'')) is not followed by an independent clause while the initial 'that' in the second two sentences ((19), (19**)) is followed by an independent clause. This latter difference is more important than the former because it shows that there is a syntactic difference separating the de die to pair (19)-(19**) from the de repair (19*')-(^••'J. So there is no syntactic reason to support Quine's view in "IR" that there is virtually no reason to distinguish the truth value, let alone the meaning, of these two pairs of sentences. Now consider a modal version of Robinson's argument against the validity of exportation: (61) There is something such that it is possible for it to be a unicorn, as a generalization of the de re modal sentence (62) It is possible for the tallest unicorn to be a unicorn. (61) is false since it is impossible for anything existing in the real world to be a unicorn. Nevertheless, the following sentences could be true: (63) It is possible that there are unicorns, and 227 (64) It is possible that the tallest unicorn is a unicorn. (63) and (64) would both be true if they were interpreted as de dictomodal sentences. (62) follows from (64) through an application of exportation. Again, this shows us that exportation is an invalid form of inference. As before with Robinson's argument for the validity of existential generalization, the modal argument against exportation can be pressed into further service. Both (62) and (61) are falsehoods. So the inference from (62) to (61) is not invalid, even though one would plausibly think that this would be the strongest candidate for invalidity of (EG). Therefore, existential generalization is a valid form of inference for modal de re sentences that contain the verb ' to be1 . Consider now a modal version of the argument that substitution is also a valid form of inference. Suppose that an essentialist were to maintain that an essential characteristic of Mark Twain was his writing of Huckleberry Finn. This state of affairs can be expressed by the following de re modal sentence: (65) It is necessary for Mark Twain to have written Huckleberry Finn. It can be argued that this is also an essential characteristic of Samuel Clemens. Nevertheless, it could also be held that the de dicto modal analogue to (65) is false: (66) It is necessary that Mark Twain wrote Huckleberry Finn. 228 It might be false, for example, if the embedded sentence within (66) were not analytic. As with the earlier pair of doxastic sentences considered in Robinson's positive argument for the validity of substitution, there are significant differences between the relevant de re and de dicto sentences. (65) is a modal de re sentence and (66) is not. (65) would not be a complete sentence if the first four words were deleted. But because the words following 'it is necessary that1 in (66) form an independent clause, it is evident that there are syntactic differences that can distinguish both the truth and the meaning of this pair of sentences. Again, Quine's claim in "IR" that the difference between de re and de dicto sentences has been virtually annulled has been shown to be false. Robinson's doxastic arguments, and my modal ones, for the validity of existential generalization and substitution show that the occurrence of terms in de re sentences are vulgar. It thus is false that the meaning of such sentences is utterly dependent on context. The main points of this chapter all concern Quine's argument against exportation in "IR" and may be summarized as follows: (1) Quine temporarily adopts the necessity predicate 'Nee' and provides a translation schema between it and the necessity functor '•' ; (2) Quine's argument, the one that apparently was due to Sleigh, succeeds in demonstrating the invalidity of exportation; (3) Robinson's argument, as well as that of the author, are even more compelling than that of Quine/Sleigh; (4) none of these arguments, 229 however, succeed in indicating the relevance of this finding against exportation that Quine wishes to attribute to it; (5) Quine rejects the relational idiom because of his exportation argument and an unclear notion of dependency on context; (6) because Quine has neglected to argue for the relevance of his exportation argument, there is no reason to accept his rejection of the relational idiom in both contexts; (7) Robinson's argument, as well as the one by the author, allow for a different conclusion to be obtained, namely, that existential generalization is a valid form of inference when performed on de re sentences in both doxastic and modal contexts; and (8) this, combined with the finding that substitution in de re doxastic and modal sentences is also a valid form of inference, shows that terms in de re modal and doxastic sentences are vulgar, and thus their meaning is not utterly dependent on context. Let us briefly return to the passage quoted from "On What There Is" that was referred to as an explanation for Quine's views on exportation. I would urge that the primary motivation for the disputed entities is not, as Quine holds of Pegasus, that they must be because otherwise it would be nonsense to say that they are not. Instead, as I have urged, the primary motivation is simply the intuitive meaningfulness of the relational idiom. But I think that there are a few arguments in favor of this position other than an appeal to intuitions. One is that all the arguments against the relational idiom appear either unjustified or unsound. It is in this category that I would place Quine's criticisms of the relational idiom in "IR." A second, to be developed in Chapter 230 IX, is that there is a semantic significance that attaches to the de re/de dicto distinction. 231 Chapter IX. A Positive Account of the Relational Idiom § 33. General Remarks There are three main objectives of this chapter. The first is to give a critique of Kripke's view that the de re/de dicto distinction is not the same as the distinction between referential and attributive uses of definite descriptions. I urge that Kripke' s account in "Speaker's Reference and Semantic Reference" is substantially, though not totally, satisfactory. How is this view relevant, one may ask, to a thesis devoted to Quine's criticisms of the relational idiom? The answer to this becomes clear when one considers the second main objective of this chapter. I intend to show that Howard Wettstein's view that there is a semantic significance of the distinction between referential and attributive uses of definite descriptions has an analogue with respect to the distinction between the relational and the notional idiom. The point here is that one might doubt that such an analogous view to Wettstein's could be viable in light of the arguments given by Kripke. In arguing against Quine's rejection of the relational idiom in modal and doxastic contexts, a variety of responses that meet Quine's concerns have been advanced. The third main objective of this chapter is the gathering together of some of these solutions, many of which have been suggested by Quine himself. His account 232 of the meaningfulness of the relational idiom in doxastic contexts is further applied to account for the meaningfulness of the relational idiom in modal contexts. This account relies on the use of attributes, entities that, admittedly, Quine rejected for reasons given in Chapter VI. But Quine also suggested (in Chapter VI) a method for legitimizing (that is, individuating) such entities. This allows attributes to assist in an account of the meaningfulness of the relational idiom in modal contexts. But, as we saw in the previous chapter, an alternate account involving the use of modal predicates could also be given to account for the meaningfulness of the relational idiom in modal contexts. I emphasize again that this solution, which involves modal attributes, was suggested by Quine himself. Quine of course would neither countenance the extension of his problem regarding the need to account for the indispensability of the relational idiom in doxastic contexts nor the sort of 'either or' theory that is presented in this chapter. (Either one uses modal attributes to account for such indispensability in modal contexts or one uses modal predicates to play this role.) This theory constitutes a positive account of the relational idiom. § 34. On the Semantic Significance of the Distinction between the Relational and the Notional Idiom Wettstein has argued that there is a semantic significance which underlies the referential/attributive distinction with respect 233 to propositional attitude ascriptions. (See note 1 for an explanation of Keith Donnellan's referential/attributive distinction.) Wettstein points out that a "referentially used description functions semantically as a demonstrative, an attributively used one functions as a property-attributor," and thus the two will "express different propositions." I think that a similar argument may be made with respect to the distinction between the relational and notional idioms. That said, it is important to acknowledge that the referential/attributive distinction is not the same as the de re/de dicto one. Kripke emphasized this point in his "Speaker's Reference and Semantic Reference" and defended it with three powerful arguments. He states that, "The view in question suggests that the de dicto case is to be identified with Donnellan's attributive use, the de re with the referential" (This is the view that he argues against.) His first is that it would be a mistake to identify de dicto uses of definite descriptions with only the attributive use. (For an account of de re and de dicto uses of definite descriptions, see § 5 or § 6.) He writes If a description is embedded in a {de dicto) intensional context, we cannot be said to be talking about the thing described, either qua its satisfaction of the description or qua anything else. Taken de dicto, "Jones believes that the richest debutante in Dubuque will marry him," can be asserted by someone who thinks (let us suppose, wrongly) that there are no debutantes in Dubuque; certainly then, he is in no way talking about the richest debutante, even "attributively." So Kripke' s view that the attributive use of a definite description cannot be identified with its use in a de dicto sentence is true. 234 The attributive use of definite descriptions ought not to be regarded as identical with the de dicto use. Kripke' s second point is that the referential use of definite descriptions cannot be identified with its use in a de re sentence. Suppose I have no idea how many planets there are, but (for some reason) astronomical theory dictates that the number must be odd. If I say, "The number of planets (whatever it may be) is odd," my description is used attributively. If I am an essentialist, I will also say, "The number of planets (whatever it may be) is necessarily odd," on the ground that all odd numbers are necessarily odd; and my usage is just as attributive as in the first case. An essentialist who utters "The number of planets (whatever it may be) is necessarily odd," is using the de re, or relational, use of the description. So, instances of the relational idiom may also feature attributive use. Kripke's third reason for rejecting the assimilation of the referential/attributive distinction into the de re/de dicto one is not quite as decisive as his previous two. It seems clear, from the following passage, that Kripke regards both the de re/de dicto distinction and the referential/attributive one, as being twofold distinctions: Russell wished to handle the de dicto-de re distinction by his notion of the scope of a description. Some have suggested that Donnellan's referential-attributive distinction can replace Russell's distinction of scope. But no twofold distinction can do this job. Consider: (2) The number of planets might have been necessarily even. In a natural use, (2) can be interpreted as true; for example, there might have been exactly eight planets, in which case the number of planets would have been even and hence necessarily even. (2), interpreted as true, is neither de re nor de dicto; that is, the definite description has neither the largest nor the smallest possible scope. 235 Kripke goes on to explain that any one of the three following symbolizations of (2) would count as Russellian analyses: (2a) 0D(3x) (There are exactly x planets and x is even), (2b) (3.x) (There are exactly x planets and 0D( x is even)), and (2c) 0(3x) (There are exactly x planets and •( x is even). (2a) is both de die to and false. Both (2b) and (2c) are, according to Forbes' definition of modal de re and de dicto sentences, de re and though (2b) is false in S5, (2c) is true. It is interesting to note that, for Kripke, (2c) is neither de re nor de dicto; he assumes that de re sentences have the largest possible scope for their descriptions and that the de dicto sentences have the smallest possible scope for their descriptions. It isn't clear if Kripke intends this assumption to be a temporary one or not. Kripke also gives two other examples of ambiguous sentences which have three possible readings: (65) Jones doubts that Holmes believes that Smith's murderer is insane, and (66) Hoover charged that the Berrigans plotted to kidnap a high American official. Kripke concludes his argument by repeating that no twofold distinction can replace Russell's notion of scope. There are two reasons why I think Kripke's third argument 236 should be discounted as one against identifying the use of a definite description in a de re sentence with the referential use of that definite description and identifying the de dicto use with the attributive use. (On the other hand, it is a very good argument indeed against identifying either of these twofold distinctions with Russell's scope distinction.) Firstly, on Forbes' definition of a de re modal sentence, and on my definition of a de re doxastic sentence, the de re/de dicto distinction is not merely a twofold one that cannot account for cases of intermediate scope such as (2c). Remember that Forbes1 definition of the distinction calls a formula (that is, any formula) that exhibits cross-quantification de re, even if, as in the case of (2c), the formula exhibiting cross-quantification is itself embedded within a modal operator. In short, Kripke is wrong to hold that the de re/de dicto distinction cannot cope with cases of intermediate scope. (On the other hand, Kripke is sort of right: no matter how you construe the definition of the de re/de dicto distinction, it is clear that it will lack the level of discriminability of Russell's scope distinctions.)' Secondly, and more directly, even if Kripke's claim about the de re/de dicto distinction were true, it would be irrelevant to his overall argument that the de re/de dicto distinction is not the same as the distinction between referential and attributive uses. Kripke evidently believes that a doxastically de re sentence is to be identified with a doxastic sentence with a description in which the description has the largest possible scope. Far from 237 showing that the two sets of distinctions (that is, the de re/de dicto one and the referential/attributive one) are not to be assimilated, Kripke has simply managed to point to an alleged similarity. Though this would not, of course, prove that the two distinctions are interchangeable, one could well imagine a defender of the view pointing to this similarity as a reason for supporting his/her position. Fortunately, as I will now show, this evident defect in Kripke's third argument may be easily repaired by combining the two problems that I have uncovered in this argument. A way to repair Kripke's third argument is simply to hold, as I would, that the de re/de dicto distinction is merely a twofold distinction, but not in the way described by Kripke. This would constitute a powerful difference between the de re/de dicto distinction and the one between the referential and attributive uses of definite descriptions. The paradigm cases of the referential/attributive distinction exclude the possibility of its being anything but a twofold distinction that could accommodate an intermediate case. The de re/de dicto distinction is a twofold distinction also, but there is a crucial difference. The difference I have in mind concerns the inability of the referential/attributive distinction to handle cases of intermediate scope of definite descriptions and the corresponding ability of the de re/de dicto distinction to handle such cases. This difference is significant, and it would constitute a third reason for denying that the two distinctions could be used interchangeably. In brief, I support two of Kripke's three reasons for not 238 identifying the referential/attributive distinction (with respect to the usage of definite descriptions) with the de re/de dicto one. Despite this support, I think that one can make an analogous argument to the one made by Wettstein regarding the semantic significance of the referential/attributive distinction. This will constitute my second positive argument for the view that the relational idiom is indispensable. The first one is that the relational idiom is intuitively indispensable. It should be noted that both arguments apply to instances of the relational idiom in both modal and doxastic contexts. This is highly important as many philosophers acknowledge not only that the referential/attributive distinction (which is held by some philosophers to be just the same one as the one between instances of the relational and notional idioms) is merely between kinds of utterances or reports, but also allege that the distinction does not travel further than the level of utterances or reports. Searle, in particular, defends this view. He writes "that there is a genuine de re/de dicto distinction, but it is only a distinction of reports." If indeed there is a semantic significance underlying the referential/attributive distinction, then it will be necessary to treat the distinction at a more philosophically sophisticated level than one which merely concentrates on utterances or reports. If true, Searle's notion would force the conclusion that there is no semantic significance underlying the de re/de dicto distinction. The author is fully aware of the fact that whereas Forbes 239 and Smullyan are content with the view that the de re/de dicto distinction is a syntactic one, Kripke and Searle appear to be content with the view that this distinction is a semantic one. One way to sort out this problem, at least roughly, is to recast Kripke's and Searle's comment regarding the de re/de dicto distinction as being about the distinction between the relational and the notional idiom. No such recasting is required when we are discussing the views of Howard Wettstein concerning the semantic significance of the distinction between the attributive and the referential uses of definite descriptions; Wettstein does not, as far as I know, extend his views regarding semantic significance to the distinction between the relational and the notional idiom. Wettstein*s argument encompasses three principal stages. Firstly, Wettstein adduces "a datum to be accounted for by any semantic account of definite descriptions." This datum is that almost all definite descriptions "fail to uniquely denote [sic] but rather apply to any number of items." His chief example is borrowed from Donnellan and involves the utterance "the murderer of Smith is insane" when the definite description is used referentially (that is, it is an instance of the relational idiom in a declarative context). When one makes this comment, whilst the utterer gestures "in the direction of Jones, the guilty party, one refers to Jones despite the fact that the uttered description, 11 given its lexical meaning, fails to uniquely specify him." This datum would occur in either doxastic or modal contexts as well as the simpler declarative ones which Wettstein relies 240 on. That is, the great majority of definite descriptions in both of these contexts fail to uniquely denote as well. Furthermore Wettstein's case could be adapted for instances of the relational idiom in either context. For example, one could utter either, (67) the murderer of Smith is believed by Ralph to be insane, or (68) the murderer of Smith is possibly insane, and precisely these same results about the failure to uniquely denote would still be obtained. So, Wettstein's datum is a general fact about utterances. Secondly, Wettstein argues that the theories of Frege and Russell (Frege's sense-reference model and Russell's theory of descriptions), according to which all occurrences of definite descriptions are, in Donnellan's terms, attributive, could not accommodate this 1 / datum. He holds that the Fregean Theory of Reference (hereafter, FTR) cannot cope with the datum because one of its tenets is that "Reference is achieved only by means of a uniquely specifying sense." Descriptions behave precisely in the same fashion as do demonstratives when they are used referentially, Wettstein urges. The fact that an utterer succeeds in referring (when uttering sentences like, "the murderer of Smith is insane," even when Smith has not been murdered) shows that reference "is secured by the fact that the context makes clear which murderer is in question." The argument that the FTR cannot cope with these sorts of 241 referential utterances proceeds by asking what sort of proposition is being expressed by such utterances: It is surely not the Fregean proposition, the subject constituent of which is a sense that uniquely specifies the murderer in question, i.e. Jones ... there is no uniquely specifying sense implicit in the utterance.... Again, it seems to me that similar considerations would apply to at least some of the instances of the relational idiom in both modal and doxastic contexts. An object is being referred to when one utters either (9) The author of Hamlet might not have written Hamlet, or (9*" ) The author of Hamlet is believed by Ralph not to have written Hamlet. Both of these sentences express what Kaplan has called a singular proposition. ° Semantic arguments against the FTR are the most convincing kind of argument. If the FTR is correct, then names refer through descriptive content of their senses. But there is an alternative theory of reference that does not feature the denoting relation as given above. This other account of reference instead contends that proper names are no more than "empty tags" which merely label objects, and for which reference is determined not by way of any conceptual content in the name but by tracing it back along some sort of causal chain leading originally from the object to a speaker's use of its name. Consider the name ' Plato' and the coreferential definite description 242 'the Athenian teacher of Aristotle1. Suppose that we discovered that Plato did not, in fact, teach Aristotle but that some other Athenian did. To whom would we be referring in the sentence: (69) Plato popularized the dialogue? The intuitive answer is that we would be referring to Plato. But the FTR would have us conclude that we would be referring to some other individual, the person who satisfied the definite description. I hold that modal sentences that can be used to argue for the theory of direct reference are also examples of the relational sense of necessity. This may seem to be a fairly tenuous connection between the theory of direct reference and the indispensability of the relational sense of necessity. But I think this connection is fairly strong. The name 'Nixon' in sentences such as (70) Nixon might have lost the election, refers directly to Nixon. It was argued before (by Kripke) that it was counterintuitive to hold that sentences like (70) were to be judged false just because Quine holds that (22) The winner of the presidential election in 1968 might have lost the election, is false. Quine's view is, Kripke has argued, counterintuitive. Both (70) and (22) are true and both are plainly relational. (22) can only be construed as an instance of the relational idiom, since an interpretation of it as an instance of the notional idiom is 243 contradictory. This fact about (22) can be conjoined with another: (22) can also be interpreted as using the definite description ' the winner of the election in 1968' either attributively or referentially. This is a significant point. The speaker could intend the truth-value of his statement to depend on what might be the case whoever the actual murderer of Smith turns out to be. The speaker is not referring to this actual-world murderer (the speaker is unable to, not knowing who it is) , but neither is he talking about whoever the description 'Smith's murderer1 fits in counter factual circumstances. If it were true that the referential/attributive distinction were identical to the distinction between the relational and the notional idiom, then one could not accept both of these facts. So, in addition to the three reasons given above for not treating the two distinctions as one, here is yet a fourth reason. § 35. Further Considerations on Modal Attributes The principal task of this thesis has been to show that the relational idiom, in both modal and doxastic contexts, is resistant to all of Quine's criticisms. This is, admittedly, largely a negative result. It was accomplished through a critical evaluation of all of Quine*s criticisms. In the course of this evaluation, small glimpses have been revealed of a more positive account of the relational idiom. This section is concerned with laying out the various solutions in order to provide a more complete positive 244 account of the relational idiom. Attributes are key entities in a positive account of the relational idiom in both doxastic and modal contexts. When Quine invoked attributes, in "Q&PA," he did so to provide a positive account of the relational idiom only in doxastic contexts. This solution was given in § 20. As was later explained, in § 25, it was the fact that Quine believed attributes could not be individuated that led him to doubt his positive account. I argued that attributes could indeed be individuated in a way that could satisfy a philosopher who had serious doubts about the cogency of Quine's charges against QML. Therefore, it is legitimate, for such a philosopher, to invoke attributes in order to account for instances of the relational idiom in doxastic contexts. We need not stop with the relational idiom in doxastic contexts. There is no reason that we cannot invoke attributes to account for the relational idiom in modal contexts as well. In various places throughout the thesis, I have either explicitly or implicitly argued for the indispensability of the relational idiom in modal contexts. I have been critical of Quine' s willingness to treat the relational idiom in doxastic contexts as indispensable but not the relational idiom in modal contexts. I also noted, in § 31, that Quine moved away from this inconsistent scepticism by rejecting the idiom in both contexts. My position, by way of contrast, is also to avoid this inconsistent scepticism, but by giving accounts of the relational idiom in doxastic as well as modal contexts. My example of a genuine (that is, an irreducibly 245 de re) sentence that expresses the relational sense of modality is, to repeat, (9) The author of Hamlet might not have written Hamlet. My account of the relational idiom in modal contexts uses modal attributes. Here is a positive account of (9) using modal attributes: (9* " •) (3x) ({Hx A (Vy) (Hy - x = y) ) A z ( z in possibly -iH) holds for x) . The basic idea is to use modal attributes in conjunction with a possible worlds semantics as outlined in Chapter VI. I wish to stress, however, that this use of modal attributes is superfluous in that possible worlds allow us to account for the meaningfulness of sentences expressing the relational sense of modality without the use of modal attributes. This account centres on the stipulation of possible worlds and was endorsed by me on pp. 200-201. Modal attributes are not used in this account. Quine accounted for the meaningfulness of sentences expressing the relational sense of belief by invoking attributes. I expanded Quine's problem to cover modal contexts. Likewise, I expanded Quine's solution by invoking modal attributes. Modal attributes are thus a Quinean motivated solution, if you will, to a Quinean motivated problem. I agreed with Quine's principle that there is no entity without identity. Hence the need for identity conditions for modal attributes. Such identity conditions can be given by 246 relying on possible worlds. Quine would disagree for metaphysical reasons. But the crucial point is that we have an account using possible worlds already and that (Kripkean) account does not invoke modal attributes. So, one who opted for a solution that invoked modal attributes would, to paraphrase Quine, be showing a marked distaste for desert landscapes. I agree with Quine that some reversion to the doctrine of essentialism will be required if one is inclined to regard the relational idiom in modal contexts as indispensable. That is, essentialism of one sort or other will be required if one is to invoke attributes in order to account for the indispensability of the relational idiom in modal contexts. But the type of essentialism that I favor is not of the invidious variety. It could be argued that every solution given in this thesis has direct roots in the work of Quine, and is hence unoriginal. To this charge, I plead not guilty. It is true that the thesis has largely been motivated by problems put forward by Quine. On the other hand, there is an obvious degree of impact of Kripke's writings upon this thesis. Without Kripke's point concerning the stipulation of possible worlds, I would have accepted Quine's criticism of transworld and transbeliefworld identity. So, if I have been unoriginal, at least I have not been slavishly devoted to but one master. Three basic points comprise this chapter. The first two are connected. First, Kripke's argument that the de re/de dicto distinction (in reality, the distinction between the relational 247 and the notional idiom) is not to be assimilated into the referential/ attributive one, was found to be two/thirds correct. The final third of Kripke's argument can be repaired. Despite the fact that there are very real differences between the two sets of distinctions, there are certain similarities. Second, some of these similarities become illuminated when one considers Wettstein's argument that there is a semantic significance to the referential/attributive distinction. An analogous argument was given to show that there is a semantic significance to the distinction between the relational and the notional idiom. This argument is important if one wishes to argue that this latter distinction does not belong somehow to speech-act theory or pragmatics, being merely one of reports or utterances. Third, the solution that invokes modal attributes was shown to be superfluous after this solution was further applied to a sentence expressing the relational sense of necessity. This solution explained how modal attributes can be used to account for the relational idiom in modal contexts. But the use of modal attributes is superfluous because of their reliance on possible worlds. The question has been raised, "At the end of the day has any good come of Quine' s long crusade against quantification into modal and doxastic contexts?" I believe that the answer is "yes," primarily because philosophers will thereby be compelled to resolve the negative reaction to Quine's crusade in a much more positive fashion. Quine's crusade has left the important legacy of certain thought provoking questions. Among these are "How do we account 248 for the evident indispensability of the relational idiom?" Even though I would substantially reject most of Quine's criticisms of the relational idiom, I think this later question must still be answered. The responses of philosophers will encompass both work in the philosophy of language and in metaphysics. So the negative reaction to Quine's negative crusade against quantification into modal and doxastic contexts will provoke a positive response. And this positive response will be of lasting benefit to philosophy. 249 Notes to Chapter I. 1. Some of Quine*s charges against merging the variables of quantification theory with modal and doxastic operators were abiding concerns long before they made their appearance among his writings. An example of this would be Quine's objection to modal logic that involves use-mention confusions. Apparently, this particular charge was voiced by Quine long before it appeared in print. Thus Quine wrote in 1962 that There was a period twenty-five years ago when I kept being drawn into arguments with C. I. Lewis and E. V. Huntington over interpretations of modal logic; and in those arguments I found it necessary to harp continually on the theme of use versus mention. And now points that Professor Marcus has urged this evening, in favor of modal logic, force me back to that same theme again. (Quine, "Reply to Professor Marcus", in The Ways of the Paradox, 2nd ed. (New York: Random House, 1966), p. 176. Thus, a purely chronological account of Quine's seventeen objections to quantifying into modal or doxastic contexts would not be reflected in an account of Quine's charges as they appeared in published form. To solve this dilemma, I shall impose the convention of introducing the objections in a chronological sequence as they were published. 2. Quine, "Quine-Carnap Correspondence," in Dear Carnap, dear Van, Richard Creath, ed., (Univ. of California Press, 1990), p. 298. Quine goes on to explain this interdefinability as follows: Synonymity is definable as sameness of meaning; and the meaning of an expression is definable, conversely as the class of its synonyms. The definition of analytic in terms of synonymity (and truth) has been seen...[on the previous page, Quine had written the following: "a statement is analytic if it can be turned into a logical truth by putting synonyms for synonyms"-M. W. D . ] ; and conversely, expressions are synonymous if, whenever one is put for the other within a statement of the form ' p => p1 , the result is analytic, (p. 298). 3. Such a combination of these two systems of logic was desirable from the outset of modal logic, at least as it was discussed by Aristotle. For Aristotle held that modal operators might be used in different ways in conjunction with symbols for individuals to yield sentences with radically different senses. 250 Aristotle was the first to recognize the distinction, or some version of it, between the de re and de dicto idiom. The following passage is taken from his Prior Analytics: The expression 'it is possible for this to belong to that' may be taken in two ways: either 'to which that belongs' or 'to which it may belong' ; for ' A may be said of that of which B" means either one of these-either 'of which B is said' or 'of which it may be said' ; and there is no difference between ' A may be said of that of which B' and ' A may belong to every 5' . It is clear then that the expression 'A may possibly belong to every B" may be used in two ways. Aristotle, Prior Analytics, in The Complete Works of Aristotle, trans. Jonathan Barnes, (Princeton, New Jersey: Princeton Univ. Press, 1984), pp. 52-53. (32b25-32) This is one of the first occurrences of the all-important notion of scope distinction in philosophy. The first sense is de dicto in that the modal operator, here denoted by 'may', has scope over a sentence that is itself devoid of modal operators ('of which B is said to belong to A" ). The second sense is de re in that the modal operator is embedded within the overall sentence ('of which B may be said to belong to A" ). One conclusion that can be drawn from Aristotle's recognition of modally ambiguous sentences is that the need for merging quantifier logic and modal logic was recognized early on. A merger of quantifier and modal logic readily allows for clarification of sentences that are modally ambiguous. Aristotle's doctrine of essentialism is to be understood as a theory which is expressible in terms of de re sentences. De re sentences, in this context, are those in which the modal operators are embedded within sentences containing either standard individual constants or definite descriptions. It should come as no surprise that the de re/de dicto ambiguity present in modal sentences is also present in other types of sentences which are modified by operators. This is certainly the case with doxastic sentences, (that is, those sentences involving belief ascriptions) as well as for those sentences involved in propositional attitude ascriptions. Aristotle was concerned with syllogism, both modal and actual. Aristotle was also conscious of the ambiguities which can arise if one were to merge what we would now call ' quantification theory' with modal operators. It is just short of astonishing that Aristotle would be able to make these points in light of the fact that he did not have quantifier logic at his disposal. (That is, Aristotle did not have quantification theory; syllogistic cannot be represented within propositional logic.) But I digress. 4. Dagfinn F011esdal, "Quine on Modality," in Words and Objections, ed. Donald Davidson and Jaakko Hintikka (Dordrecht: D. Reidel, 1975), pp. 175-185. Also see F011esdal's "Essentialism and Reference," in The Philosophy of W. V. Quine, ed. Lewis Edwin HahnGRx 251 and Paul Arthur Schilpp (La Salle, Illinois: Open Court, 1986), pp. 97-113. David Kaplan, "Quantifying In," in Words and Objections, pp. 206-242. Also see Kaplan's "Opacity," in The Philosophy of W. V. Quine, pp. 229-288. 5. Saul Kripke, "A Completeness Theorem in Modal Logic," Journal of Symbolic Logic, 24 (1959), pp. 1-15. 6. John R. Searle, in Intentionality, (Cambridge, England: Cambridge Univ. Press, 1983), holds that the de re/de dicto distinction in doxastic contexts occurs merely at the level of utterances and does not have semantic significance. T. E. Patton has pointed out, quite accurately, that the distinction is not a syntactic one. He writes, that if it is a syntactic distinction, "how can there be ambiguous sentences, like 'Philip believes that the capital of Honduras is in Nicaragua' ? (Ambiguous even though the de re reading can sound absurd.)" (Patton, Letter received by the author. November 7, 1994.) The de dicto reading attributes a false belief to Philip. 7. Quine, "Reference and Modality," in Reference and Modality (London: Oxford Univ. Press, 1971), p. 31. 8. Raymond Bradley and Norman Schwartz, Possible Worlds (Indianapolis, Indiana: Hackett Publishing Company, 1979), p. xvii. 9. W. V. Quine, "Whitehead and the Rise of Modern Logic," in The Philosophy of Alfred North Whitehead, Paul Arthur Schilpp, ed. , 2nd ed. (La Salle, Illinois: Open Court, 1951), pp. 125-164. The original Schilpp volume on Whitehead was published in 1941. 10. Quine, "Two Dogmas of Empiricism," in FLPV (Cambridge, Massachusetts: Harvard Univ. Press, 1953), p. 31. 11. Quine, "Q&PA," pp. 183-184. 12. There are three versions of "R&M" in the three editions of FLPV. Unless otherwise stated, all references to "R&M" indicate the initial edition of Quine's book. For clarity's sake, the second edition of FLPV is not relied upon. The version of "R&M" that occurs in the 1971 collection of essays entitled Reference and Modality is from the second edition of FLPV. There is only one relevant difference between the first two versions of Quine's papers and the third version. See Chapter III., § 11. 13. Though worthy of note, the fact that Quine refrains from stating his conclusion about the phenomenon of substitutivity in "R&M" using precisely the terms just given does not at all deter me. Though there are arguments for and against my interpretation of Quine's stance on the conclusions drawn from this objection, I 252 contend that my interpretation would be the strongest one available, were the objection a sound one. I hold that it is not. My reason for holding this is that Quine insists on interpreting modally ambiguous sentences as if they were not. He compounds this defect by recognizing, in the objections based on the phenomena of existential generalization and of unique specification, that the sentences in question are modally ambiguous, and then proceeds to give them only one interpretation. Quine's views about the phenomenon of substitutivity are discussed further in Chapters II. and III. 14. Quine, "IR," in Theories and Things (Cambridge, Massachusetts: Belknap Press, 1981), p. 120. 15. Quine, WO (Cambridge, Massachusetts: M.I.T. Press, 1960), p. 195. 16. However, Quine's willingness to allow for the meaningfulness of the relational idiom was a dubious venture at best. Quine's willingness to countenance intensional entities as a means of accounting for the relational idiom was never sufficiently powerful to overcome his extensional predilections. 17. Such an argument will rely on Quine' s comments in his "Reply to David Kaplan," [1986]. Quine has at times allowed for the meaningfulness of the relational idiom in doxastic contexts (albeit without much enthusiasm) even though he is usually opposed to the idiom. In "R&M," he does not mention that the idiom in doxastic contexts might be held to be meaningful even though he entertains the notion that the relational idiom in modal contexts is held by some philosophers to have meaning. Three years later, in "Q&PA," he appears to reverse himself by implying that the relational idiom in doxastic contexts presents a problem which should be solved. This second position is in turn rejected in WO. Quine's latest statement on the relational idiom appears in "IR" in which he adopts the view that there is no fundamental distinction between the relational and notional idioms. These changes of heart on Quine's part are the direct result of not having a persuasive argument (either in favour or not) of the idiom's meaningfulness. Notes to Chapter II. 1. Quine, "Whitehead and the Rise of Modern Logic," p. 142. This case is not to be confused with the more famous example of 9 being necessarily greater than 7 in "R&M"; it might be thought that the '<' is in error and that ' >' would be more appropriate, especially given Quine's later examples. In fact, however, the correct symbol in Quine's paper on Whitehead is '<' . If the symbol were '>', then the conclusion would be true, contrary to Quine's intentions. 253 2. In FLPV, Quine describes "R&M" as having "grown out of a fusion of" "NE&N," and "The problem of interpreting modal logic," Journal of Symbolic Logic 1 (1947). Quine, FLPV, p. 170. 3. Quine, "NE&N," p. 113. 4. Quine, WO, p. 142. 5. Quine, Pursuit of Truth (Cambridge: Harvard University Press, 1990). See, for instance, pp. 68-70. 6. Quine, WO, p. 151. 7. Quine, "NE&N," p. 114. 8. Rudolph Garnap, Meaning and Necessity (Chicago: The Univ. of Chicago Press, 1947), p. 137. 9. Quine, "R&M," p. 153. 10. It is true that sentences such as •(Samuel Clemens = the author of Huckleberry Finn), are false, but this should not be confused with the more innocuous •(Samuel Clemens = Mark Twain). 11. Quine, "R&M," pp. 144-145. 12. Quine, "Designation and Existence," in The Journal of Philosophy, 36, No. 26 (1939), p. 704. 13. The fact that abstract terms such as 'appendicitis1 are conceded to be meaningful by Quine is extremely important. For there is a clear presumption that if a term has meaning, then there must be some structure of ontology in virtue of which it is meaningful. I go much further than Quine in that I apply this clear presumption to the intuitive meaningfulness of instances of the relational idiom. 14. See for example, "Steps Towards a Constructive Nominalism," in Journal of Symbolic Logic 1949. 15. Quine, "NE&N," p. 791. 16. I have not replaced 'Philip' with 'Penelope' out of some sort of exaggerated sense of gender sensitivity, but rather to avoid an annoying logical ambiguity that arises with the personal pronoun 'he' when it is used to refer to either Tully or Cicero. Without the switch to 'Penelope' , the reader may be left with the 254 mistaken impression that the male personal pronoun is meant to refer to Philip. 17. Quine, "R&M," p. 147. The application of (EG) to (19), which yields (19'), is invalid but only because (19) is being construed as a de dicto sentence. If instead it were interpreted as a de re sentence, than application of (EG) to it would be, to paraphrase Quine, warranted. 18. Quine, "NE&N," p. 794. 19. Quine, "Q&PA," pp. 187-189. The two relevant sentences are (15) Ralph believes z{z is a spy) of Ortcutt, and (22) Ralph believes z(z is not a spy) of Ortcutt. Quine notes that both (15) and (22) are true. This is not, however, to charge Ralph with contradictory beliefs. Such a charge might be reasonably read into: (23) Ralph believes z( z is a spy . z is not a spy) of Ortcutt, but this merely goes to show that it is undesirable to look upon (15) and (22) as implying (23). It might be thought that this passage contradicts an earlier one in which Quine holds that the 'believes that5 construction must be eliminated in cases where there is a quantification into, as opposed to within, a belief context. His reason is that 'believes that1 is referentially opaque. Quine's point is a subtle one and it can easily appear that he is objecting to the very same point that he makes in a few short pages. As Vlastos has said, "there are times when the drudgery of saying the obvious is rewarded, and this is one of them." ("The Third Man Argument in the Parmenides", p. 325). Quine's problem is to give an account of the relational sense of belief that reflects the intuition that instances of such belief are meaningful. This is Quine's point of departure, and it is one that changes over time. Contents of belief are indicated by 'believes that4 and it cannot be the case that Ralph has a belief content which is contradictory to another. 20. This alternate way of describing the de re/de dicto distinction for doxastic sentences was prompted by a complaint that I had conflated syntactic and semantic phenomena. This complaint was made by Calvin Normore, the External Examiner of the thesis. The specific suggestion for adapting Dummett's ' X" notation came from Robinson. 255 21. In a seminar conducted by Howard Jackson, alternative interpretations were being offered to account for Quine's use of the term necessary consequence. The author suggested 'D(-' -a suggestion that Peter Apostoli correctly pointed out contains a use-mention confusion. The conventional symbol for modal necessity-' Q" -belongs in the object language but the conventional symbol for semantic implication-'h'-belongs to the metalanguage. Peter suggested that this problem could be avoided if we introduce a new metalinguistic symbol-' "i-1 -to indicate necessary consequence. Neither Apostoli nor Quine spelled out exactly how the notion of necessary consequence would differ from the metalinguistic notion of semantic implication. I confess to being unable to add sufficient clarity to this problem. However, I would note that this problem would lend strength to an interpretation wholly rooted not in the metalanguage, rather in the object language. There is no use-mention present if we interpret necessary consequence as ' -CT • Nor is there a need to introduce a new, more coherent, symbol. 22. Robinson, E-mail received by the author, April 13, 1995. There is such an interpretation if we take it to include possible worlds. That is, say "7" were assigned 7 and " NP" were assigned {7}. 23. Quine, "R&M," p. 149. 24. Quine, "R&M," p. 149. 25. That is, the universal closure of (8***) is false in an interpretation W = {a,p} D{a) = 2?(P) JVP(a) = {9} NP(P) = {7} 26. 27. 28. 29. 30. 31. 32. Quine, Quine, Quine, Quine, Quine, Quine, Quine, = WO, WO, wo, wo, wo, wo, wo, p. 271. p. 271. pp. 271-272. p. 272. p. 273. p. 273. pp. 273-274. 33. Quine, "NE&N," p. 121. See also "R&M," p. 143. 256 34. Quine, "R&M," p. 143. 35. In "On Denoting," Russell divides definite description statements into three major components. In Joseph Margolis, ed. An Introduction to Philosophical Inquiry, 2nd ed. (New York: Alfred A. Knopf, 1978), pp. 486-496. 36. Nathan U. Salmon, Reference and Essence (Princeton, New Jersey: Princeton Univ. Press, 1980), pp. 23-31. Salmon sets out three arguments against what he calls Orthodox Theories (by which he means variants of Fregean or Indirect Theories of Reference) the most compelling of which are semantical ones in articles given originally by Donnellan, Kaplan, Kripke and Putnam. 37. This is a denial of a category-mistake. On Russell's analysis, a definite description brings in two clauses, and it is them to which ' V is not prefixed. 38. Quine, "TGMI," in The Ways of Paradox (New York: Random House, 1966), p. 156. 39. Quine, "TGMI," p. 156. 40. Quine, "TGMI," p. 160. 41. Quine, "TGMI," p 42. Quine, "TGMI," p 43. Quine, "TGMI," p 160. 160. 170. 44. Quine, "Reply to Professor Marcus," in WP, p. 175. 45. Quine, "TGMI," p. 175. In reply to a message by the author received by members of the Russell List on e-mail, Francisco Rodriguez stated that Russell was made aware of the use-mention confusions that arise out of his joint work with Whitehead, Principia Mathematica. Here is Prof. Rodriguez1 comment as received on March 1, 1995: Yes, he certainly did it, although my piece of information does not come from the period after 1962, but from well before. In the academic year 1938-9 he was explicitly asked the question about the use-mention distinction in the _Principia_, by Copi, and he answered something like "well, we did not have that style of talk." Professor Kenneth Blackwell, who presides over the Russell Archives at MCMaster University subsequently forwarded a copy of Copi's letter, which is dated February 19, 1990. Here is a more complete account of the matter (that is, use-mention confusion) from the 257 letter to Professor Rodrigues: Very soon after Carnap returned to Chicago in the academic year 1938-39 I learned from him that he no longer identified logic with syntax but allowed it to include both semantics and pragmatics as well. It seem to me that this perspective borrowed from Peirce at the urging of Charles Morris, made it possible to understand certain parts of Volume I of Principa Mathemathica more clearly. Here I have in mind the difference between Primitive Propositions *1.1 Anything implied by a true elementary proposition is true. Pp. and *1.3 q. ^ .p o q. Pp. At my first conference with Russell I inquired if the difference wasn't that one was a Rule for the language (and therefore in the metalanguage) whereas the other was a statement of the language and therefore in the object language. Russell answered "Yes", and explained that in 1910 he and Whitehead hadn't yet "got on to that way of talking." He was not all that comfortable at having to reply thus to my questions, (p. 2). Blackwell, Letter received by the author, March 28, 1995. 46. Ruth C. Barcan, "A Functional Calculus of First Order Based on Strict Implication," in Journal of Symbolic Logic, 11, 1947. Carnap and Church began, before Barcan and after C.I. Lewis, to give semantic analyses of modal logic. 47. Quine, "TGMI," p. 164. It is relevant that Quine does not see this particular defect as common to all the levels of modal involvement that he discusses; it is not a feature of his third degree. Confusions between use and mention are not a feature of Quine's third level of modal involvement, the level associated with the metaphysical doctrine of essentialism. 48. Quine, "Reply to Professor Marcus," pp. 175-176. Notes to Chapter III. 1. Church, rev. of "NE&N," Journal of Symbolic Logic 8 no. 45, (1943), p. 45. Carnap also responded to Quine in 1943, holding that use of the relational idiom would affectively nullify Quine' s criticisms. See below. 2. Quine, "R&M," in Reference and Modality, p. 29. 3. Quine, "R&M," in Reference and Modality, pp. 27-28. 4. Quine, "The Quine-Carnap Correspondence," in Dear Carnap, Dear Van (Berkeley and Los Angeles: Univ. of California Press, 1990), p. 371. Quine's letter is dated Dec. 8, 1943. Quine goes on to give two positive elaborations of the remedy that he has 258 just proposed. With respect to (8) The number of planets is necessarily > 7, Quine writes that (i) Numbers may be conceived as entities proper (rather than quasi-entities introduced by contextual definition), hence as intensions. But if 9 and the number of planets are taken as numbers in this sense, then despite their arithmetical equality they must be regarded as distinct numbers; for ... no one entity can be designated by two non-synonyms.... Or (ii) Numbers may be conceived as quasi-entities, say classes, introduced by contextual definition....(The anomaly, that is, of having to say that 9 and the number of planets are ti/o numbers. ) But then the contextual definitions should be framed in such way that the sign '>' , if construed as having to do with numbers ... is prevented from turning up after an 'N' ... Naturally the same precaution would have to be taken in introducing any connective for use between "names" solely of extensions and not intensions, (pp. 371-372) 5. Garnap, Meaning and Necessity (Chicago: The Univ. of Chicago Press, 1947), pp. 198-199. 6. Carnap, Meaning and Necessity, p. 199. "The intension of an individual expression is the individual concept expressed by it." (p. 42.) He gives as an example the individual concept Walter Scott, expressed by the intension of 'sb . 7. Carnap, Meaning and Necessity, p. 199. 8. Carnap, Meaning and Necessity, p. 199. 9. Quine, "R&M," in FLPV, p. 153. 10. Quine, "R&M," in Reference and Modality p. 28. 11. Quine, "Reply to Professor Marcus," in Ways of Paradox, p. 182. 12. Quine, "R&M," Reference and Modality, p. 28. 13. Quine, "Reply to Professor Marcus," p. 182. 14. Smullyan, "Modality and Description," in Reference and Modality (London: Oxford Univ. Press, 1971), pp. 17-34. Originally, this paper appeared in The Journal of Symbolic Logic, 13 no. 1, (1948), pp. 31-37. 15. Smullyan, "Modality," p. 35. 16. Smullyan, "Modality," p. 35. 259 17. Smullyan, "Modality," p. 39. 18. Smullyan, "Modality," p. 38. 19. Quine, "R&M," in Reference and Modality, pp. 29-30. 20. Quine, "R&M," in Reference and Modality, p. 30. 21. Kripke, Naming and Necessity, 2nd edition (Cambridge, Massachusetts: Harvard Univ. Press, 1980), p. 41. 22. This sort of rendition seems as ad hoc to me as its counterpart in Chapter II. , where this point is more fully developed with respect to Quine's account of semantic ascent. I can see no reason to prefix the entire statement by ' 0~>' other than the fact that it (a) keeps the overall sentence de dicto, and (b) avoids the uncomfortable consequence that arises if the ' ->' is embedded while the '0' has scope over the entire sentence. This point occurred to me initially when I was co-authoring, with Richard E. Robinson, "Modal Predicates". This paper was presented on Oct. 25, 1990 as part of the University of British Columbia Philosophical Colloquium. 23. Barcan Marcus, "Backward Look," p. 236. 24. Quine, "R&M" in the first edition of FLPV, p. 155, n. 9. The 1961 version of "R&M", reprinted in Reference and Modality, features the quotation on p. 30, n. 9. 25. Quine, "R&M," in the first edition of FLPV, p. 155. It might be thought that Quine held that Smullyan had engaged in "an alteration of Russell's familiar theory" because Russell's "On Denoting" does not mention modality. Now, it is certainly true that there is no mention of modality in "On Denoting", but this is not the reason that Quine alludes to when he further explains his point in n. 9 on p. 155. 26. Barcan Marcus , "A Backward Look at Quine's Animadversions on Modalities , " in Perspectives on Quine, Robert Barrett and Roger Gibson ed. (Cambridge, Massachusetts: Basil Blackwell, 1990), p. 236. 27. Barcan Marcus, "Backward Look," p. 236. 28. Quine, "R&M," p. 155. 29. Quine, "R&M," in the 1980 edition of FLPV, p. 156. 30. Barcan Marcus, "Backward Look," p. 236. 260 31. The fallacy as it applies to the relational idiom in modal contexts was recognized by Church and Smullyan. T.E. Patton pointed out that Quine commits this fallacy in WO where Quine argued that the relational (or transparent) idiom of belief is "intolerably odd". 32. Quine, "R&M," pp. 24-25. 33. David Kaplan, "Opacity," in The Philosophy of W.V. Quine, Lewis Edwin Hahn and Paul Arthur Schilpp eds. , (La Salle, Illinois: Open Court, 1986), p. 232. 34. Quine, "NE&N," p. 127. 35. Quine, "Reply to David Kaplan," in The Philosophy of W.V. Quine, p. 291. 36. Robinson, E-mail received by author, 20 Oct. 1994. 37. Letter received from T. E. Patton, 18 July 1994. 38. Robinson, "Quine's 'Alleged Theorem1 and Relational Belief," (Manuscript). 39. Quine, "Reply to David Kaplan," in The Philosophy of W. V. Quine, p. 291. It is difficult to know where to begin in pointing out the inconsistencies that have been constituted by this particular citation. For one thing, distinctions between what is and what is not meaningful are supposed to be, according to Quine, unintelligible. Secondly, Quine is not being entirely accurate with respect to the role played by attributes with respect to the meaningfulness of instances of the relational idiom in doxastic contexts. This passage again shows that Quine has never really made up his mind with respect to the meaningfulness of these instances. 40. Quine, "WA," in Theories and Things, (Cambridge, Massachusetts: Belknap Press, 1981), p. 127. 41. Leonard Linsky, Oblique Contexts (Chicago: University of Chicago Press, 1983) p. viii. 42. Quine actually came quite close to giving the Russellian version of Quine's modal puzzle in "NE&N": Consider the attributes determined by the respective matrices: (26) x > number of planets, (27) x > 9; 261 that is, the attribute of exceeding the number of planets and the attribute of exceeding 9. Since (26) and (27) are not logically equivalent, it follows that the attributes will not be identical. The statement: (28) The attribute of exceeding the number of planets = the attribute of exceeding 9 is false, (p. 126). Despite this apparent similarity of this argument with the Russellian modal one, it is clear that they feature fundamentally different conclusions. 43. This puzzle began life in R. E. Robinson's and M. ¥. Dickson's "Steps Towards a More Benign Modality" (unpublished). The author was responsible for the former puzzle while R.E. Robinson was responsible for the latter puzzle. 44. Bertrand Russell, "On Denoting" in An Introduction to Philosophical Inquiry, Joseph Margolis, ed. (New York: Alfred A. Knopf, 1978), p. 400. 45. Russell, "On Denoting," p. 494. 46. Quine, WO, p. 142. Notes to Chapter IV. 1. There have been three such attempts, if one includes Quine's own presentation, in "R&M," of the problem based on the operation of (EG) as well as the problem of (US). 2. Ruth Barcan Marcus, "Backward Look," p. 239. 3. Quine, "Two Dogmas of Empiricism," in FLPV, 1st ed. (Cambridge, Massachusetts: Harvard Univ. Press, 1953), p. 22. 4. Quine, "TGMI," pp. 173-174. 5. Quine, WO, pp. 198-200. 6. Quine, WO, p. 200. 7. Ruth Barcan Marcus , "Modalities and Intensional Languages , " in Boston Studies in the Philosophy of Science 1961/1962, M. Wartofsky, ed. (Dordrecht: Reidel), p. 64. Also in Synthese, Evert W. Beth, ed. (Dordrecht, The Netherlands: D. Reidel), (1961), pp. 303-322. 262 8. Ruth Barcan Marcus, "Backward Look," p. 238. I am not certain that Barcan Marcus is correct in her claim that the de dicto reading is more plausible, simply because nothing baffling, as she points out, results from it. It appears unlikely that Quine would opt for a reading which carried no force against his intended target, the detested doctrine of essentialism. On the other hand, Quine may well have been in the grips of the Relational Fallacy when he made his criticism of essentialism. This in itself would imply that a de dicto reading on Quine' s part is believable. This issue in short, is an unresolved one. 9. Given the Completeness Theorem for QML, since there are models in which (34*), (35*), and (36') are true and both (C*) and (C**) are false, it follows that neither (C*) nor (C') are derivable from (34*), (35*), and (36'). 10. Barcan Marcus, "Backward Look," p. 238. 11. Quine, "TGMI," in The Ways of Paradox, p. 174. 12. Kaplan, "Opacity," in The Philosophy of W.V. Quine, Lewis Edwin Hahn and Paul Arthur Schilpp, eds., (La Salle: Open Court, 1986), p. 252. 251. 13. Kaplan, "Opacity," p. 257. 14. Kaplan, "Opacity," p. 257. 15. Kaplan, "Opacity," p. 257. 16. Barcan Marcus, "Backward Look," p. 239. 17. Barcan Marcus, "Backward Look," p. 238. 18. I think it is clear that Barcan Marcus and Kaplan are at least partially correct in that Quine is objecting to non-sortal properties in his first objection to essentialism. But the second objection to essentialism does not fit so easily within this view. Arguably, it is I.A.E. that Quine is criticizing with his second objection to essentialism. It depends on which of the three interpretations of Quine' s argument in WO that one sees as being the correct version. In particular, if one opts for the de re interpretation where the modal operators are fully embedded, then Quine is objecting to an invidious form of essentialism. 19. Dagfinn F0llesdal, "Essentialism and Reference," in Perspectives on Quine, pp. 98-99. 20. Dagfinn F0llesdal, "Essentialism," p. 103. 263 21. The question of the merits vs. the demerits of essentialism would only arguably arise, again in the second version of "R&M", after Quine has discussed Smullyan's solution, which will, Quine contends, yield Aristotelian essentialism. 22. Quine, "Reply to Dagfinn F011esdal," in Perspectives on the Philosophy of W. V. Quine, p. 114. 23. Dagfinn F0llesdal, "Essentialism," p. 103. Notes to Chapter V. 1. Quine, "Q&PA, 2. Quine, "Q&PA, 3. Quine, "NE&N, 4. Quine, "Q&PA, 5. Quine, "Q&PA, 186 189. p. 115 187 186 6. In "Q&PA" Quine supports "irreducibly triadic" (p. 187) relations between a mental agent, an attribute and an object (or a pair or a triple or ... of objects) ... In "IR," he supplants the role of attribute with that of the predicate. In "Q&PA," Quine selects attributes rather than predicates in order to account for the meaningfulness of instances of the relational idiom in doxastic (but not modal) contexts. He chooses attributes because he is worried about the notion of predicate of L because of the difficulty of supplying sufficient identity conditions for language. In "IR," on the other hand, Quine appears to select predicates. However, he still selects predicates without giving identity conditions for language. 7. Quine, "Q&PA," p. 186. 8. Quine, "R&M," pp. 141-143. Quine writes To sum up the situation in a word, we may speak of the contexts is unaware that...' and 'believes that...' as referentially opaque. The same is true of the contexts 'knows that...' , ' says that...', 'doubts that...' , and 'is surprised that...', etc. This passage is contained in both editions of FLPV. This fact is a maddening one, for reasons given shortly. The quote mentioned in this endnote can be found on p. 20 of Linsky's Reference and Modality, which contains the second version of Quine's "R&M". Quine's solution in "Q&PA" is to devise, through the use of 264 attributes, contexts that are not referentially opaque. 9. Quine, "Q&PA," p. 183. In his autobiographical remarks in the Schilpp volume, Quine sheds no light upon his evolving stance towards the relational idiom, with the sole exception of one remark on the tension between the ability to sacrifice propositional attitude ascriptions, (that is, instances of the relational idiom in doxastic contexts) and our infrequent ability to do so: I worried [in the middle of 1964-M. W. D.] about the idioms of propositional attitudes: they are conspicuously unsatisfactory in logical ways, yet not readily dispensable. How might their purposes be served in more scientific terms? (pp. 35-36). 10. Quine, WO, p. 148. 11. Calvin Normore, External Examiner's Report received by the author, August 29, 1995, p. 3. 12. Quine, WO, p. 149. 13. Quine, WO, p. 148-149. 14. Patton, "Is Transparent Belief 'intolerably odd'?," in Dialogue, 13 no. 4, (1974), pp. 647-655. 15. Quine, WO, p. 149. 16. Patton, "Transparent Belief," pp. 650-651. 17. Quine, WO, p. 148. 18. Quine, WO, pp. 149-150. 19. John P. Stewart, for one, upholds Quine's solution. Unlike me, he believes that Quine's criticism that the transparent sense of belief is intolerably odd is a reasonable one. 20. Patton, especially pp. 650-652. 21. Patton, Letter received, 16 July 1994. 22. Quine, WO, pp. 197-198. 23. Dagfinn F011esdal, "Quine on Modality," in Words and Objections, pp. 175-185. Remarkably enough, F0llesdal, who is the only author who has dealt with Quine's argument regarding the collapse of modal distinctions utterly fails to note any of the subtleties that I call attention to. He does not bother to subject 265 Quine's colossal assumption concerning the stubbornness of objects to any sort of critical evaluation. F0llesdal merely relates that ' p = Dp' is a disastrous conclusion drawn from the following premise: (PI) ( y) ( Fy = . y = x) . (y) ( Gy = . y = x) . o N( y) ( Fy = Gy) ((PI) is taken directly from F0llesdal's article "Quine on Modality.") F0llesdal's (PI) is equivalent to Quine's (4) in WO. 24. Quine, WO, p. 197. 25. Quine, WO, p. 197. 26. Quine, Pursuit of Truth, (Cambridge, Massachusetts: Harvard Univ. Press, 1990), pp. 68-69. Notes to Chapter VI. 1. Quine, "Autobiography of W. V. Quine," in The Philosophy of W. V. Quine, p. 35. 2. Quine, WO, p. 151. 3. Quine, "Q&PA," p. 189. 4. A solution at mid-point between Kripke's and Quine's accounting of the meaningfulness of certain relational sentences has been given in an unpublished paper (Robinson and Dickson, "Modal Predicates," paper presented at Philosophical Colloquium, Oct. 25, 1990. This result initially occurred in "Modal Attributes," another unpublished work which began life in the previous year. The point was later strengthened in "Modal Predicates." This method takes away Kripke's possible worlds, thereby siding with Quine. On the other hand, this method extends Quine's account of the meaningfulness of relational sentences in doxastic contexts to one of the meaningfulness of relational sentences in modal contexts, thereby siding with Kripke. This solution allows one to combine a sceptical approach to possible worlds and the formulation of M.A.E. claims in order to close the gap between Kripkean and Quinean accounts of relational sentences. 5. Kripke typifies possible worlds as "possible states," in his forward to the 1980 version of Naming and Necessity (pp. 15-17) . Not only do I agree with this proposal, I also see it as being an effective answer to the complaint that the meaning of possible worlds has not been adequately spelled out. 266 6. As we shall see in Chapter VIII., there was a brief time in which Quine's concerns about intensional entities manifested themselves even about extensional entities. I am referring to his concerns about predicates and sentences of a language L towards the end of "Q&PA." Quine apparently set aside these concerns in his later (1977) paper "IR," where he used necessity and doxastic predicates with no mention of his previous misgivings. 7. Quine, "IA," in The Logical Enterprise (New Haven: Yale Univ. Press, 1975), p. 12. 8. Quine, "IA," p. 5. 9. The distinction between legitimate and heuristic postulation is explained above on p. 12 of this thesis. 10. Richard E. Grandy, "Ontology and Reduction," in Essays on the Philosophy of W. V. Quine, Robert W. Shahan and Chris Swoyer eds., (Norman, Oklahoma: University of Oklahoma Press, 1979), p. 71. 11. One might gain the mistaken impression that there is no difference between Quine' s principle that there is no entity without identity and the principle regarding the coextensiveness of sets as an arbiter of the legitimacy of entities. It is mistaken because whereas the first principle is an ontological dictum, the second is simply a principle concerning the individuation of entities. As we have just seen, Quine has serious doubts about the application of the second principle to attributes. The latter principle makes no mention of any particular entity--it merely describes how two entities may be individuated. The two principles do, on the other hand, both serve as indicators, for Quine, of the illegitimacy of attributes. Quine holds that identity conditions cannot be set forth for attributes; so his ontological dictum is violated. In addition, Quine holds that attributes cannot be individuated; so his principle of individuation is violated. 12. Quine, "IA," p. 3. 13. Quine, "IA," p. 4. IA. Quine, "IA," p. 4. 15. Quine, "Q&PA," p. 191. 16. Quine, "IA," pp. 4-7. 17. However, for each kind of individual i there must be a separate principle for individuating i (there is, in addition, a problem of circularity present, because some kind of individual must be taken as basic, for which no principle of individuation 267 can be given without becoming viciously circular). We may have adequate criteria for individuating numbers. Therefore, numbers could be members of sets, but without adequate criteria for individuating attributes there would be no sets of attributes (with the exception of the empty set). 18. Quine, "IA", p. 4. 19. Quine, "IA," pp. 4-5. 20. Quine, "IA," p. 5. 21. Quine, "IA," p. 7. 22. I will argue shortly that this proposal by Quine is an incorrect one, but not for the reasons that Quine holds. The proposal in question would be more compelling if instead of the de re {Vx) 0 (Fx ** Gx) , Quine had used the de dicto U ( Vx) (Fx «• Gx) . 23. David Lewis, On the Plurality of Worlds (New York: Basil Blackwell, 1986), pp. 50-51. 24. Quine, "IA," p. 7. 25. I use "F" and " G" both for predicates in the language and their intended interpretations, properties. 26. Kripke, Naming and Necessity, p. 138. 27. The analytic coextensiveness of attributes that Kripke envisions presents an unneeded complication that is not a feature of Kripke's properties. 28. To select but one instance of Aristotle's use of attributes, consider the following citation from Book VII: ... the differentiae present in man are many, e.g. endowed with feet, two-footed featherless. Why are these one and not many? Not because they are present in one thing; for on this principle a unity can be made out of any set of attributes. But surely all the attributes in the definition must be one; for the definition is a single formula and a formula of substance, so that it must be a formula of some one thing; for substance means a 'one' and a 'this1, as we maintain. (1037b22-28. ) I am not convinced that analytic coextensiveness plays an even minimal role in Aristotle's use of attributes. This matter admittedly warrants further attention, but space does not allow me to devote more attention to it. 268 29. "Symposium: The Principle of Individuation," in Proceedings of the Aristotelian Society, Suppl. Vol. 27 (1953), pp. 69-120. See also S. Marc Cohen, "Aristotle and individuation ," in New Essays on Aristotle, Francis Jef f ry Pelletier and John King-Farlow eds. (Guelph, Ontario: Univ. of Calgary Press, 1984), pp. 41-66. 30. You do not have necessary coextensiveness with Quine's version. This model was suggested to me by Robinson. Robinson, Letter received by the author, June 13, 1995. It can be shown that "VxD (Fx ~ Gx)" is derivable in QML from "DVx (Fx •* Gx)" . For example, SHOW VxD (Fx • Gx) (DD) UVx {Fx • Gx) StteW- VxD (Fx «• Gx) SHOW UVx ( Fx « Gx) ^ Vx (Fx « Gx) Fx ** Gx ' Pranise (UD) (ND) 2, KR 5, UI is a modal version of Kalish and Montague's system of derivation using rules introduced in Richard E. Robinson's Introduction to Modal Logic. (This derivation includes Kalish and Montague's rule of universal derivations (UD).) 31. Quine, "IA. 32. Quine, "IA," p. 9. 269 Notes to Chapter VII. 1. As is shown in this chapter, concerns over transworld identification can be extended to ones over belief world identification. Quine accomplished this, that is, extended such concerns, in "WA." 2. Kripke was the first to prove a completeness theorem in QML, and the system that he used was a combination of S5, as well as a variety of other propositional modal systems, and quantification theory. S5 is one of several systems of propositional modal logic first devised by C. I. Lewis. This system is the dominant one among modal logicians. 3. I realize that such an interpretation of (49) and (9) is contrary to the views of Quine, who prefers to account for such sentences with his theory of semantic ascent. This approach has already been rejected (see Chapter II. above). 4. David Lewis comes readily to mind in this regard. His conception of an individual is that it cannot be strictly identical with an individual in another possible world. His view of the relational idiom in modal contexts relies on a different set of truth conditions from the one that was given by Kripke in his modal theoretic account of QML. A good discussion of Lewis' view regarding the truth conditions of the relational idiom in modal contexts is contained in his "Counterpart Theory and Quantified Modal Logic." 5. Hintikka's work suggests such equivalent treatment. In "WA" and in "IR," Quine was responding to Hintikka's earlier suggestion. 6. Kaplan, "Quantifying in," in Words and Objections, Donald Davidson and Jaakko Hintikka, eds. (Dordrecht, Holland: D. Reidel, 1975), p. 221. 7. The Leibniz-Arnauld Correspondence, H. T. Mason trans. (New York: Manchester Univ. Press and Barnes and Noble, 1967), p. 29. The letter is dated May 13, 1686. 8. However, the problem could be restated in such a way as to appear more palatable. It might be held that the ability to conceive of the same object at different times requires the notion of temporal parts. It might also be held that there is no notion of a modal part that plays the same unifying role as do temporal parts. 9. Patton, Correspondence received by the author, 23 Oct. 1994. 270 10. Patton, Correspondence received by the author 23 Oct. 1994. 11. There are four main responses to the verification version of the problem of trans-world identity. These are (i) David Lewis' systematic recasting of modal semantics, which replaces discussion of an individual existing in more than one possible world with that of non-identical but similar counterparts; (ii) Kit Fine's search for de dicto equivalents to de re modalities; (The goal of finding de dicto equivalents of de re modal sentences can be achieved in one of two ways: firstly, restrictions can be introduced on S5 models; secondly, modal language may be modified in such a way that it can be formally proved that there are de dicto equivalent formulae to de re formulae) (iii) creation of a calculus of individuals which would make sense of trans-world identity; and (iv) the view that no need arises to provide criteria for trans-world identity. The fourth option is advocated by Kripke and it is the option preferred by the author. 12. Quine, "IA," p. 4. 13. Leonard Linsky, "Reference, Essentialism, and Modality," in Reference and Modality, Leonard Linsky, ed. (London: Oxford Univ. Press, 1971), p. 99. Originally in Journal of Philosophy, 66 no. 20, (1969), pp. 687-700. 14. Quine, "WA," in Theories and Things (Cambridge Massachusetts: Belknap Press, 1981), pp. 125-126. 15. Quine, "WA," p. 126. 16. David Lewis has devised a comprehensive theory of modality which attempts to solve the purported problem of transworld identification. A very strong impression that one receives from Lewis' writings on the Theory of Modal Realism (hereafter, TMR) is that a rejection of de re modality would be untenable. Another impression is that Lewis accepts enough of the standard (i.e., Quinean) criticisms of QML. Confronted with the dilemma of either rejecting de re modality or the objections levelled at QML, Lewis ingeniously selects a third option that apparently affords him a means of escape. He recasts the semantics of modality by doing away with individuals who occupy more than one possible world in favour of non-identical counterparts. A comprehensive account of Lewis' theory of modal realism is contained in On the Plurality of Worlds. A similar account of Counterpart Theory is to be found in "Counterpart Theory and Quantified Modal Logic", in The Possible and the Actual, Michael J. Loux, ed. (Ithaca, New York: Cornell Univ. Press, 1979), pp. 110-128. Lewis spells out his opposition to trans-world identity in the following passage, from On the Plurality of Worlds: I do not deny the existence of trans-world individuals,yet 271 there is the sense in which I say that they cannot possibly exist. As should be expected, the sense in question involves restricted quantification... It is possible for something to exist iff it is possible for the whole of it to exist. 17. Quine, "WA," p. 124. 18. Quine, "WA," p. 125. 19. Quine, "Identity, Ostension, and Hypostasis," in FLPV, pp. 65-79. 20. Quine, "WA," p. 127. 21. Quine, "WA," p. 127. 22. Kripke, Naming and Necessity, p. 44. There is some confusion in the secondary literature regarding Kripke's point about stipulation of possible worlds. In Identity and Essence, Baruch Brody ascribes to Kripke the view that essentialist statements do not presuppose trans-world identity. He then proceeds to argue at length that this view is faulty. Brody describes his understanding of Kripke's position as follows: ... we claimed, at the beginning of this section, that f a has P essentiallyl means that a has P and that in all worlds containing an object identical with a, that object has P. On this account, the meaningfulness of essentialist claims presupposes the meaningfulness of trans-world identity. But suppose we simply said that f a has P essentiallyl means that a has P and that in all worlds containing a, a has P. This account makes no reference to cross-world identity and does provide an account of the meaning of essentialist claims. But can we offer this account? Can we use a name which refers to some object in the actual world to refer to that object in some possible world? If Kripke is right, and the name in question is a rigid designator, it seems to follow that we can, and that by doing so we avoid any need to employ the notion of trans-world identity. Brody, Identity and Essence (Princeton, New Jersey: Princeton Univ. Press, 1980), p. 108. Brody's enterprise in this matter strikes me as being both curious and ill-conceived. For the view that Brody attributes to Kripke is based on a falsehood. Kripke does not deny that transworld identification exists for Nixon in one world and the Nixon who might have lost the presidential election in 1968. Rather, Kripke holds that transworld identification, though present, does not pose any difficulty. It would be more precise to say that Kripke does not contend that criteria for trans-world identity are not presupposed by essentialist statements. This interpretation of Kripke's position, i.e., the doctrine 272 of essentialism does presuppose trans-world identity but does not presuppose criteria for trans-world identity, is supported by Naming and Necessity. By way of contrast, Brody's interpretation has very little in the way of textual support. Brody does attempt, albeit unsuccessfully, to argue against Kripke's use of names as primitive objects which range over possible worlds. His argument rests on the different role played by perception in endurance over time compared to endurance over possible worlds. "Because we cannot perceive their [objects-M.W. D . ] persistence through possible worlds, names for persisting objects must be constructed from names for nonpersisting objects by use of a definition of cross-world identity." There are, then, two primary difficulties with Brody's argument against Kripke. Firstly, Brody errs when he neglects to attribute the presupposition of trans-world identity to Kripke. To repeat, Kripke does think that trans-world identity is presupposed by essentialism but he also thinks this presupposition is innocuous in that criteria for trans-world identity need not be given. Secondly, even if Kripke did argue that trans-world identity is not presupposed by essentialism, Brody's argument would not be satisfactory. 23. Kripke, Naming and Necessity, p. 44. 24. Kripke, Naming and Necessity, p. 44. 25. Kripke, Naming and Necessity, p. 44. 26. Kripke, Naming and Necessity, p. 46. 27. Kripke, Naming and Necessity, p. 46. 28. Kripke, Naming and Necessity, p. 46. In general, it is a fallacy to identify necessary conditions with necessary properties. But Kripke does not hold that the two notions are interchangeable in all circumstances. Rather, this interchangeability is limited to this conceivability criterion. 29. Leonard Linsky, "Reference, Essentialism, and Modality," p. 100. 30. Quine, "WA," p. 128. 31. This appears to be the gist of Quine's comment on p. 128 of "WA." See also n. 4 on that page. 32. Kripke, Naming and Necessity, p. 77. 33. Kripke, Naming and Necessity, p. 45. Kripke briefly points out two difficulties with Lewis' GT, one formal and one informal. 273 34. Two of Quine's charges against QML can be combined: his concerns about (EG) and his criticism of possible worlds. Quine does not want to quantify over modal operators because that would grant legitimacy to de remodality. He would have a similar problem quantifying over possible worlds. Quine does not make this criticism explicitly, but it is implied by his writings. His views on ontological commitment are by no means isolated ones. One philosopher who is entirely in accord with Quine on these matters is R.M. Martin. In the following, Martin remarks on the relation between quantification, reduction, and ontological commitment: . . . surely if persons are in some sense "reducible" to things so that quantifiers over the latter are indeed quantifiers over them, the implications for philosophy are enormous. Of course we can quantify over anything we want. The point is that what we quantify over we regard as constituting the ontology of our theory, and in so far forth as constituting the entities our theory takes as "real" in the most fundamental sense. If Davidson wants to quantify over persons in action theory, he is then taking an explicit stand as to what he regards as real. The import of doing this for his philosophy is thus very different from one who admits only quantifiers over just things that are non-persons. This is taken from Martin's Mind, Modality, Meaning and Method (Albany, New York: SUNY Press, 1983), pp. 4-5. I am substantially in accord with Martin's point, one which has far-reaching significance beyond a critique of Davidson's action theory. That is, Martin's point would be applicable if one were debating to quantify over, say, possible worlds. It is my contention both that instances of the relational idiom presuppose that de re sentences are true, and that the truth of these sentences involves an ontological commitment to certain entities. Notes to Chapter VIII. 1. Quine, "IR," in Theories and Things, (Cambridge, Massachusetts: The Belknap Press of Harvard Univ. Press, 1981), pp. 113-123. 2. Quine, "Q&PA," p. 193. 3. Quine, "IR," p. 115. Nee was originally Quine's second grade of modal involvement mentioned in "TGMI." In "TGMI," Nee is a sentence operator, but in the translation schema just given, it does much the same sort of work as the third grade of modal involvement. 4. Quine, "IR," p. 118. 5. Quine, "IR," p. 119. 274 6. Quine, "IR," p. 119. The view that exportation is generally valid ("Q&PA," p. 188.) is a conclusion for which Quine has given absolutely no argument. I think it should viewed as a thoroughly dubious assumption when it resurfaces in "IR." 7. Quine, "IR," p. 120. 8. Quine, "IR," p. 120. 9. Kaplan "Quantifying In," p. 233. 10. Quine is, to be precise, oversimplifying Kaplan's condition for exportation; Kaplan insisted on a representing name, and vividness was only one of three aspects of representation. 11. Quine, "IR," p. 120. 12. Quine, "IR," p. 121. 13. Quine, "IR," p. 121. 14. Quine, "IR," p. 115. 15. Quine, "On What There Is," in FLPV, 1st ed., p. 4. 16. This second rationale was suggested to me by Patton. Patton, Letter received by the author, 16 May, 1995. 17. Quine, "IR," p. 122. 18. Quine, "IR," p. 122. 19. Kaplan accuses Quine of an "inconsistent scepticism" in "Quantifying In". Quine's reply to this objection is contained in Words and Objections: The distinction between opaque and transparent on the modal side is the distinction between what Chisholm, reviving scholastic terminology, calls necessitas de dicto and necessitas de re. But I had a reason, . . . . , for treating belief more fully than necessity. It was that the notion of belief, for all its obscurity, is more useful than the notion of necessity. For this reason my treatment of modal logic was brief and negative; I was content to outline the opacity troubles. Kaplan's charge of "inconsistent scepticism" is off the point; the point is that some obscure notions are, on grounds of utility, more worth trying to salvage than others. (p. 344 ) . 20. Robinson, "Quine's 'Alleged Theorem1 and Relational Belief," January 30, 1995. 275 21. Kaplan, "Quantifying In," pp. 178-179. 22. Robinson, "Belief," p. 9. 23. Robinson, "Belief," p. 10. Notes to Chapter IX. 1. Wettstein, "The Semantic Significance of the Referential-Attributive Distinction," in Philosophical Studies, 44 (1987), pp. 187-196. The famous distinction between referential and attributive uses of definite descriptions was initially made in Keith Donnellan's "Reference and Definite Descriptions," in Philosophical Review, 75, (1966), pp. 281-304. It was reprinted in The Philosophy of Language, edited by A. P. Martinich, (London: Oxford Univ. Press, 1985), pp. 236-248. Donnellan explains this distinction as follows: A speaker who uses a definite description attributively in an assertion states something about whoever or whatever is the so-and-so. A speaker who uses a definite description referentially in an assertion, on the other hand, uses the description to enable his audience to pick out whom or what he is talking about and states something about that person or thing, (p. 238 of The Philosophy of Language). 2. Wettstein, "Semantic Significance," pp. 187-188. 3. Kripke, "Speaker's Reference and Semantic Reference," in The Philosophy of Language, p. 252. 4. Kripke, "Speaker's Reference," p. 252. 5. Kripke, "Speaker's Reference," p. 252. 6. Kripke, "Speaker's Reference," p. 252. 7. (2b) is certainly false in S5. 8. Kripke, "Speaker's Reference," p. 253. 9. It might be thought that, in order to make sense of Kripke' s point regarding (2) , it must be interpreted as an independent point, that is, independent from Kripke's overall argument that the two sets of distinctions must not be assimilated. Accordingly, it could be held that although both (2b) and (2c) are de re modal sentences there is at least one significant difference separating them: their truth-values. It seems to follow that the de re/de dicto distinction is inadequate to handle scope differences. But Kripke does not say that (2c) is de re. In fact, it seems quite purposeful on his part not to say this, given that he 276 expressly identifies (2a) as de dicto and (2b) as de re. All he says about (2c) is that it is "the interpretation which makes (2) true." (Kripke, "Speaker's Reference," pp. 252-253). If Kripke is making an independent point, its certainly isn't that both (2b) and (2c) are de re. So, if Kripke is making an independent point, then he is making a thoroughly ambiguous one. The elusive point of Kripke's third argument seems that there are intermediate cases of scope that are neither de re nor de dicto. But this point does not advance his overall argument, since definite descriptions such as the one in (2c) would presumably be used neither attributively nor referentially. So, Kripke' s third argument needs to be repaired. 10. Searle, Intentional!ty, p. 217. 11. Wettstein, "Semantic Significance," p. 188. 12. Wettstein, "Semantic Significance, 13. Wettstein, "Semantic Significance, 14. Wettstein, "Semantic Significance, 15. Wettstein, "Semantic Significance, 16. Wettstein, "Semantic Significance, p. 189. p. 189. p. 188. p. 189. p. 190. 17. Wettstein, "Semantic Significance," p. 190. My acceptance of Wettstein's "which murderer" should be altered an acceptance of "which murderer or at least which person". This would be consistent with my assertion that there may be no murderer. Here is a rough definition of the denoting relation present in the FTR: "the relative product of the relation of expressing, . . . [which holds between a singular term and a concept-M. W. D.] and the relation of determining, which holds between the concept and an object." From Reference and Essence, by Nathan U. Salmon, (Princeton: Princeton Univ. Press, 1981), pp. 10-11. 18. Kaplan, "Demonstratives," in Themes from Kaplan (London: Oxford Univ. Press, 1989), p. 483. 19. Salmon, Reference and Essence, p. 11. 20. This account of the relational idiom in modal contexts was initially proposed in a seminar given by Robinson at the University of British Columbia during the Fall of 1989. It was noted that Quine' s account in "Q&PA" was not duplicated with respect to the relational idiom in modal contexts. Such a solution was given in our joint paper, "Modal Attributes." This became, by the Fall of 1990, a slightly different paper, "Modal Predicates." The use of attributes to account for the relational idiom in modal 277 contexts has had great appeal to me because it is dramatically similar to a solution proposed by Quine himself. Glossary-Part A: Abbreviations 278 From a Logical Point of View Word and Object "On the Individuation of Attributes" "Intensions Revisited" "Notes on Existence and Necessity" "Quantifiers and Propositional Attitudes" "Reference and Modality" "Two Dogmas of Empiricism" "Three Grades of Modal Involvement" "Worlds Away" Occam's Razor Benign Quinean Essentialism Invidious Aristotelian Essentialism Moderate Aristotelian Essentialism Russell's Theory of Definite Descriptions Quantified Modal Logic FLPV WO "IA" "IR" "NE&N" "Q&PA" "R&M" "TDD" "TGMI" "WA" (OR) B.Q.E. I.A.E. M.A.E. TDD QML 279 Part B: Numbered Sentences (1) Ralph believes someone is a spy. (1*) Ralph believes that there are spies. (1**) Ralph believes that (3x) (x is a spy). (1') There is someone (in particular) whom Ralph believes to be a spy. (I'1) (3x) (Ralph believes that x is a spy). (I''*) Ralph believes of the man in the brown hat that he is a spy. (2) The author of Hamlet might not have written Macbeth. (2*) It is possible that: the author of Hamlet did not write Macbeth. (2**) 0 (3x) (x is the author of Hamlet and x did not write Macbeth). (2***) 'The author of Hamlet wrote Macbeth' is not analytic. (2****) 'Exactly one thing wrote Hamlet and whatever wrote Hamlet wrote Macbeth" is not analytic. (2*****) 0 [Ox) {Hx A (Vy) (Hy - x = y)) A ->Mx1 (2') There is exactly one author of Hamlet, and that same individual possibly did not write Macbeth. (2'*) (3x) {x is the author of Hamlet and no one else is and 0 (x did not write Macbeth)). (2' • •) (3x) [ (Hx A (Vy) (ify - x = y)) A 0 ->«x] . (3) 9 is necessarily greater than 7, 280 (3*) Necessarily (3x) (x > 7) (3**) Necessarily something > 7. (3') (3x) (x is necessarily greater than 7). (3'') There is something that is necessarily > 7. (4) The number of planets = 9 (5) The number of planets is necessarily greater than 7. (5*) • (3x) [ (NPx A (Vy) (NPy - x = y) ) A (x > 7)]. (5**) It is necessary that: there is exactly one thing that is the number of planets and that thing is greater than 7. (5- ) (3x) [(NPx A (Vy) (NPy - x = y) ) A •( x > 7)]. (5'') There is exactly one thing that is the number of planets and it is necessary that thing is greater than 7. (5*) "Npx A (Vy) (Npy - y = x)" » "x > 7." (6) The number of planets is possibly less than 7. (6*) 9 is possibly less than 7. (7) x = /x + /x + -fx # /x (7*) x = Vx + /x + ifx # Vx h^ x > 7. (7**) "x = /x + fx + T/"X * /x" h "(x > 7)." (7***) D[x = /x + /x + Vx # /x - (x > 7)]. (8) there are exactly x planets. (8*) There are exactly x planets "i- x > 7. (8**) "there are exactly x planets" H "(x > 7)." (8***) •[there are exactly x planets - (x > 7)]. (9) The author of Hamlet might not have written Hamlet. 281 (9*) 0 [ (3x) ( ( Hx A (Vy) (Hy - x = y) ) A ^Hx) ] . (9**) 0 --[(3x) ((Hx A (Vy) (Hy - x = y) ) A fix)]. (9***) 'The author of Hamlet wrote Hamlet" is not analytic. (9****) 'Exactly one thing wrote Hamlet, and whatever wrote Hamlet wrote Hamlet' is not analytic. (9 • ) (3x) [ ( Hx A (Vy) ( Hy - x = y)) A 0~-Hx] (9'') Exactly one thing wrote Hamlet, and whatever wrote Hamlet might not have written Hamlet. (9''') The author of Hamlet is believed by Ralph not to have written Hamlet. (9*'*') (3x) ( (Hx A (Vy) (Hy - x = y) ) A z (z possibly -^ H) holds for x). (9''*) It is possible that: exactly one thing wrote Hamlet, and whatever wrote Hamlet did not write Hamlet. (9"**) It is possible that it is not the case that: exactly one thing wrote Hamlet and whatever wrote Hamlet, wrote Hamlet. (gtt***) I t is possible that it is not the case that: exactly one thing wrote Hamlet. (10) Jean Chretien cancelled the helicopter contract. (11) Jean Chretien = the Prime Minister of Canada. (12) The Prime Minister of Canada cancelled the helicopter contract. (13) •(Samuel Clemens = Mark Twain). (14) D(Samuel Clemens = Samuel Clemens). (15) •(Samuel Clemens = the man named 'Mark Twain1). 282 (16) Appendicitis is dreaded. (16*) (3x) (x is dreaded). (17) Giorgione was called 'Giorgione1 because of his size. (18) Barbarelli was called 'Giorgione1 because of its size. (19) Penelope is unaware that Tully denounced Catiline. (19*) Penelope is unaware that (3x) (x denounced Catiline). (19**) Penelope is unaware that Cicero denounced Catiline. (19') (3x) (Penelope is unaware that x denounced Catiline). (19'') Tully is such that Penelope is unaware that he denounced Catiline. (19''') Cicero is such that Penelope is unaware that he denounced Catiline. (20) The person delivering pizza possibly smashed the delivery truck. (20*) 'Exactly one person delivered pizza and whoever delivered pizza did not smash the delivery truck1 is not analytic. (20**) ' (3x) [(PxA (Vy) (Py - x = y) ) A ->SxY is not analytic. (21) the Morning Star = the Evening Star. (22) The winner of the presidential election in 1968 might have lost the election. (22') Pos {z (z is possibly not a WP), x] . (23) Necessarily if there is life on the Evening Star then there is life on the Evening Star. (24) Necessarily if there is life on the Evening Star then 283 there is life on the Morning Star. (24') (3-x) (necessarily if there is life on the Evening Star then there is life on x) (25) George IV wondered whether Scott was the author of Waver ley. (26) Scott = the author of Waverley (27) George IV wondered whether Scott was Scott. (28) It is possible that the number of planets is greater than 9. (29) It is possible that 9 is greater than 9. (30) Ronald Reagan wondered whether Mt. Everest is higher than Mt. Olympus. (31) Mt. Everest = the highest mountain in the world (32) Ronald Reagan wondered whether the highest mountain in the world is higher than Mt. Olympus. (33) There is one and only one number of planets and that number is possibly less than 9. (33 • ) (3x) ( ( NPx A (Vy) ( NPy -+ x = y)) A 0(x > 9)). (34) All mathematicians are necessarily rational but are not necessarily (that is, are contingently) bipedal. (34*) •( Vx) (Mx -* Rx) A -<D( Vx) (Mx -* Bx) . (34*) (Vx) (0(Mx - Rx) A ^(Mx - Bx) ) . (34'*) (\/x) ((Mx - DRx) A (Mx - -^DBx) ) . (35) All cyclists are necessarily bipedal but are not necessarily (that is, are contingently) rational. (35*) D( Vx) (Cx - Bx) A -"•( Cx - Rx) 284 (35') (Vx) (D(<7x - Bx) A ^D( Cx - i?x) ) . (35 • * ) (Vx) ( ( Cx -* UBx) ' ( Cx - -QRx) ) . (36) Someone is both a mathematician and a cyclist. (36' ) (3x) (Mx A Cx) . (36'•) Ma A Ca. (G) Either this concrete individual is both necessarily rational and contingently bipedal or necessarily bipedal and contingently rational. ( C ) 3x) {{URx A -UBx) V (D-Bx A -IORX) ] . (C") (3x) KDRx A ^Di?x) A (OBx A -iQBx) ] . (37) (Ma - QRa) A {Ma - -iQBa) . (37') Ma - DRa. (37' • ) Ma - -CLBa. (38) ( Ca - QBa) A ( Ca » -.Qffa) . (38') Ca - QBa. (38* ' ) Ca - ->Qffa. (39) DRa A -^URa. (40) QBa A ->0Ba. (41) (3x) (Witold wishes that x is president). (42) (3x) (x is a dog. x talks. I seek x). (43) (3x) (x is a lion. Ernest strives that Ernest finds x). (44) The winner of the presidential election in 1968 is believed by Ralph not to have won the election in 1968. (45) Tom believes of Tully that he denounced Catiline. 285 (46) Necessarily 9 > 4. (47) It is necessary that: the number of major planets > 4. (47*) • (3x) KNPx A (Vy) (NPy - x = y ) ) A ( x > 4 ) ] . (47') The number of major planets is necessarily > 4. (47 ' * ) (3x) [ ( NPx A (Vy) ( NPy - x = y) ) A D (x > 4 ) ] . (48 (49 (50 (51 (52 (53 (54 (55 (56 (57 (58 (59 (60 (61 Tom believes that Cicero denounced Catiline. (3x) DFx. The Cayster was bathed in by Heraclitus over the course of two days. It is possible that: there is a fat man in the doorway. There is a possible man in the doorway. Tom believes Mark Twain to have written Huckleberry Finn. Ralph believes Bernard to be wearing a brown coat. There is someone whom Ralph believes to be wearing a brown coat. There is something Sean believes to be a leprechaun. Sean believes the shortest leprechaun to be a leprechaun. Sean believes that there are leprechauns. Sean believes that no two leprechauns have the same height. Sean believes that the shortest leprechaun is a leprechaun. There is something such that it is possible for it to 286 be a unicorn. (62) It is possible for the tallest unicorn to be a unicorn. (63) It is possible that there are unicorns. (64) It is possible that the tallest unicorn is a unicorn. (65) Jones doubts that Holmes believes that Smith's murderer is insane. (66) Hoover charged that the Berrigans plotted to kidnap a high American official. (67) the murderer of Smith is believed by Ralph to be insane. (68) the murderer of Smith is possibly insane. (69) Plato popularized the dialogue. (70) Nixon might have lost the election, 287 Bibliography Almog, Joseph and Perry, John and Wettstein, Howard (editors). Themes from Kaplan. New York: Oxford University Press, 1989. Anderson, Alan Ross and Marcus, Ruth Barcan and Martin, R. M. (editors). The Logical Enterprise. New Haven: Yale University Press, 1975. Barcan Marcus, Ruth. "Modalities and Intensional Languages." In Synthese 13 no. 4, (1961). . "A Backward Look at Quine's Animadversions on Modal Logic." In Barrett, Robert and Gibson, Roger (editors). Perspectives on Quine. London: Basil Blackwell, 1990. Barnes, Jonathan (editor). The Complete Works of Aristotle. Princeton, New Jersey: Princeton University Press, 1984. Brody, Baruch A. Identity and Essence. Princeton: Princeton University Press, 1980. Carnap, Rudolf. Meaning and Necessity: A Study in Semantics and Modal Logic. Chicago: University of Chicago Press, 1947. Chisholm, Roderick M. "The Logic of Knowing." The Journal of Philosophy 60 no. 25, (1963), pp. 773-795. Creath, Richard. Dear Carnap, dear Van. Berkeley and Los Angeles, California: University of California Press, 1990. Davidson, Donald and Hintikka, Jaakko (editors). Words and Objections. Dordrecht: D. Reidel, 1975. Forbes, Graeme. The Metaphysics of Modality. Oxford: Clarendon Press, 1985. Kaplan, David. "Quantifying in." In Words and Objections. pp. 206-242. . "Opacity." In The Philosophy of W. V. Quine. pp. 229-288. Kripke, Saul. Naming and Necessity. Princeton: Princeton University Press, 1980. Lewis, David. On the Plurality of Worlds. Princeton: Princeton University Press, 1986. 288 Linsky, Leonard. Oblique Contexts. Chicago: University of Chicago Press, 1983. (editor). Reference and Modality. London: Oxford University Press, 1971. Loux, Michael J. (editor). The Possible and the Actual: Readings in the Metaphysics of Modality. Ithaca: Cornell University Press, 1979. Martin, R. M. Mind, Modality, Meaning, and Method. Albany, New York: State University of New York Press, 1983. Martinich, A. P. (editor). The Philosophy of Language, 1st edition. Oxford: Oxford University Press, 1985. Mason, H. T. (editor and translator). The Leibniz-Arnauld Correspondence. New York: Manchester University Press and Barnes and Noble, 1967. Patton, T.E. "Is Transparent Belief 'intolerably odd1?." In Dialogue 13 no. 4, (1974), pp. 647-655. "On a Persistent Fallacy Regarding the De Re." In Analyse (1987). . Letter received by author. 18 July 1994. Plantigna, Alvin. The Nature of Necessity. Princeton: Princeton University Press, 1974. Quine. Willard Van Orman. "Designation and Existence." In Journal of Philosophy 36 (1939). . "Whitehead and the Rise of Modern Logic." In The Philosophy of Alfred North Whitehead. Paul Arthur Schilpp (editor). 2nd edition. La Salle, Illinois: Open Court, 1951. pp. 125-164. . "Notes on Necessity and Existence." In Journal of Philosophy 40 (1943). "The problem of interpreting modal logic." In Journal of Symbolic Logic 12 (1947). pp. 43-48. . From a Logical Point of View: 9 Logico-Philosophical Essays. Cambridge, Massachusetts: Harvard University Press, 1953. Word and Object. Cambridge, Massachusetts: Massachusetts Institute of Technology Press, 1960. 289 The Ways of Paradox of and Other Essays. New York: Random House, 1966. Philosophy of Logic. Cambridge, Massachusetts: Harvard University Press, 1970. . Theories and Things. Cambridge, Massachusetts: The Belknap Press of Harvard University Press, 1981. . "Reply to Dagfinn F0llesdal." In The Philosophy of W. V. Quine. pp. 114-115. . "Reply to David Kaplan." In The Philosophy of W. V. Quine. pp. 290-294. . The Time of My Life: An Autobiography. Cambridge, Massachusetts: Massachusetts Institute of Technology Press, 1985. . Pursuit of Truth. Cambridge, Massachusetts: Harvard University Press, 1990. Robinson, R. E. Introduction to Modal Logic. Manuscript, 1983. . "Discussion Paper." 1994. Salmon, Nathan U. Reference and Essence. Princeton: Princeton University Press, 1981. . Frege's Puzzle. Cambridge, Massachusetts: The MIT Press, 1986. Shahan, Robert W. and Swoyer, Chris (editors). Essays on the Philosophy of W. V. Quine. Norman: University of Oklahoma Press, 1978. Wedin, Michael V. "Singular Statements and Essentialism in Aristotle." In The Canadian Journal of Philosophy. Supplementary Volume (1984), pp. 67-88. Wettstein, Howard K. "Demonstrative Reference and Definite Description." In Philosophical Studies 40 (1981), pp. 241-253. "The Semantic Significance of the Referential-Attributive Distinction." In Philosophical Studies 44 (1983), pp. 187-196. . "Cognitive Significance Without Cognitive Content." In Themes from Kaplan. 290 Appendix A-F0llesdal on a Defense of Quine on Modality F0llesdal holds that the criticisms made against Quine's objections to QML are without foundation. F0llesdal considers three such criticisms. I shall attempt to show that this survey of opinions is incomplete and that there are flaws with some of the counter objections made by F0llesdal in defence of Quine. The first criticism F0llesdal argues against is that of Church. He notes that Church makes sense of quantification into modal contexts by allowing that the variables in question have an intensional range composed of attributes as opposed to classes. Here, in full, is F0llesdal's rebuttal of Church's position: However, merely restricting the range of the variables does not help one out of the difficulties; nothing is presupposed about the range of the variables in Quine's arguments, which are unaffected by any restriction on variables . . . The crucial part of Church's review is his reference to Frege's ideas concerning sense, denotation and oblique occurrences. In his later article "A Formulation of the Logic of Sense and Denotation" (1951), Church makes ingenious use of these ideas to develop a system where one quantifies into necessity contexts without this leading to any collapse of modal distinctions. In this system the variables take intensions as values in necessity contexts, but what saves the system is not that the variables are thus restricted, but that the positions of the variables are referential. There are no opaque constructions in the system. Its necessity operator notwithstanding, it is therefore not a modal, i.e., nonextensional system, and it is not evidence against Quine's thesis that one cannot quantify into referentially opaque contexts. Church has rather shown how to handle the logical modalities without making use of opaque contexts. This defense of Quine is defective in two areas: Firstly, Quine's arguments make considerable presuppositions about the setting of the variables, namely, that the variables and their 291 quantifiers are embedded within the scope of a modal operator. In other words, there is a presupposition that the formulae are de dicto. This consideration is significant if one is supposing, as I am supposing, that quantification into modal contexts makes perfect sense in the case where the relevant formulae are de re. Secondly, just because Church's operators are nonextensional in nature does not mean that they are nonmodal. Plenty of essentialist theses have been stated using nonextensional modal operators, for example, Barcan-Marcus' account of QML. Regarding the latter criticism, Church urges that rather than the option that Quine takes, Quine should hold that if a variable within a non-extensional context is to refer to a quantifier that occurs before that context, "a variable must have an intensional range--a range, for instance, composed of attributes rather than classes."3 To hold as F0llesdal does, that Church has merely succeeded in making quantification into modal contexts legitimate by making modal variables referential, and thus non-modal is putting the horse before the cart. Of course, one is tempted to reflect, quantification into modal contexts would succeed only if the occurrences are made to be referential. Interpreting modal variables as referential in order to make sense of quantification into modal contexts is precisely identical to providing an account of the relational idiom as being meaningful in modal contexts. One gets the impression from F0llesdal that no expansion (that is, no 292 reinterpretation) of the occurrences of modal variables in order to transform them into purely referential ones can be permitted on the grounds that it violates some definition of modality. I submit that F0llesdal's definition of modality, which is that modal occurrences are automatically referentially opaque, is entirely arbitrary. Hence it presents no obstacle. Carnap's criticism of Quine in "NE&N," is the second criticism of Quine's thesis to be considered by F011esdal. It is virtually identical to that of Church; the primary difference between the two approaches lies in their respective solutions to Frege's antimony of the name-relation. The third criticism that F011esdal considers is that of Hintikka. As in the case of Carnap's objection to Quine's views regarding quantification into modal contexts, F011esdal holds that Hintikka would have to accept Quine's finding of 'p = Dp ' . Again however, F011esdal provides absolutely no argument to support this view. Here is the relevant passage in question: As a matter of fact, Hintikka's semantics does make sense if the values of the variables are taken to be expressions, i.e., singular terms that are substituted for the variables. However, Hintikka does not intend to give this [Church and Carnap's account of modal semantics-M.W.D.] kind of 'substitutional' interpretation of quantification. He wants, and I think rightly, the values of his variables to be "real, fullfledged individuals". However, if this is to be the case, it is unclear what the objects are over which Hintikka quantifies, unless he supplements his semantics in such a way that Quine's thesis becomes valid in it. It is interesting to note that, in his reply to this paper, Quine does not indicate the colossal leap of faith that F011esdal is asking the reader to take. Thus F0llesdal compounds 293 his earlier lapse of required argumentation: not only does he not provide a shred of evidence to support Quine's thesis of ' p = Dp', he also fails to produce a substantial argument to show that both Carnap and Hintikka are committed to this view. I have also criticized F011esdal for falsely accusing Quine of conflating two notions of essentialism. (See § 18.) However, I do not wish to leave the impression that F0llesdal has not shed any valuable light upon Quine. He has, notably in his comments regarding causation. It is his comments regarding the viability of Quine's various charges against QML that I find to be wrong, and it is to these alone that I object. Notes to Appendix A 1. F0llesdal, "Quine on Modality," p. 177. 2. Patton has urged that "the range of the variables is independent of where (or how) the variables and their quantifiers occur. Or at least someone might say this." Letter received from Patton, 16 July 1994. Patton's view is consistent with my position that our picture of the range of the pertinent variables should not be restricted by holding that the variables must be embedded within the scope of a modal quantifier. 3. F011esdal, "Quine on Modality," p. 177. Churches review of Quine's "NE&N," is in Journal of Symbolic Logic 8 (1943), pp. 45-47. 4. F011esdal, "Quine on Modality," p. 178. 

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
https://iiif.library.ubc.ca/presentation/dsp.831.1-0088185/manifest

Comment

Related Items