The New Theory of Reference: Kripke, Marcus, and Its Origins (Synthese Library)

THE NEW THEORY OF REFERENCE SYNTHESE LIBRARY STUDIES IN EPISTEMOLOGY, LOGIC, METHODOLOGY, AND PHILOSOPHY OF SCIENCE M...

Author: P. Humphreys | J.H. Fetzer (editors)

40 downloads 701 Views 33MB Size Report

This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!

Report copyright / DMCA form

DOWNLOAD PDF

THE NEW THEORY OF REFERENCE

SYNTHESE LIBRARY STUDIES IN EPISTEMOLOGY, LOGIC, METHODOLOGY, AND PHILOSOPHY OF SCIENCE

Managing Editor: JAAKKO HINTIKKA, Boston University

Editors: DIRK V AN DALEN, University of Utrecht, The Netherlands DONALD DAVIDSON, University of California, Berkeley THEO A.F. KUIPERS, University ofGroningen, The Netherlands PATRICK SUPPES, Stanford University, California JAN WOLEN-SKI, Jagielionian University, KrakOw, Poland

THE NEW THEORY OF REFERENCE: KRIPKE, MARCUS, AND ITS ORIGINS Edited by

PAUL W. HUMPHREYS University of Virginia, Charlottesville, VA, U S.A.

and JAMES H. FETZER University of Minnesota, Duluth, MN, US.A.

KLUWER ACADEMIC PUBLISHERS DORDRECHT / BOSTON / LONDON

Library of Congress Cataloging-in-Publication Data is available.

ISBN 0-7923-4898-2

Published by Kluwer Academic Publishers, P.O. Box 17, 3300 AA Dordrecht, The Netherlands. Sold and distributed in North, Central and South America by Kluwer Academic Publishers, PO Box 358, Accord Station, Hingham, MA 02018-0358, U.S.A. In all other countries, sold and distributed by Kluwer Academic Publishers, Distribution Center, P.O. Box 322, 3300 AH Dordrecht, The Netherlands

Printed on acid-free paper

All Rights Reserved

© 1998 Kluwer Academic Publishers

No part of this publication may be reproduced or utilized in any form or by any means, electronic, mechanical, including photocopying, recording or by any information storage and retrieval system, without written permission from the copyright owner.

TABLE OF CONTENTS

PAUL W. HUMPHREYS and JAMES H. FETZER / Introduction

vii

PART I: THE APA EXCHANGE 1.

QUENTIN SMITH / Marcus, Kripke, and the Origin of the New Theory of Reference

3

2.

SCOTT SOAMES / Revisionism about Reference: A Reply to Smith

13

3.

QUENTIN SMITH / Marcus and the New Theory of Reference: A Reply to Scott Soames

37

PART II: REPLIES 4.

SCOTT SOAMES / More Revisionism about Reference

65

5.

JOHN P. BURGESS / Marcus, Kripke, and Names

89

6.

JOHN P. BURGESS / How Not to Write History of Philosophy: A Case Study

125

QUENTIN SMITH / Direct, Rigid Designation and A Posteriori Necessity: A History and Critique

137

7.

PART III: HISTORICAL ORIGINS 8. 9.

DAGFINN F0LLESDAL / Referential Opacity and Modal Logic, §§ 16-19

181

STEN LINDSTROM / An Exposition and Development of Kanger's Early Semantics for Modal Logic 203

10. QUENTIN SMITH / A More Comprehensive History ofthe New Theory of Reference

235

Name Index

285

Subject Index

287 v

PAUL W. HUMPHREYS AND JAMES H. FETZER

INTRODUCTION

On January 20th, 22nd, and 29th, 1970 Saul Kripke delivered three lectures at Princeton University. They produced something of a sensation. In the lectures he argued, amongst other things, that many names in ordinary language referred to objects directly rather than by means of associated descriptions; that causal chains from language user to language user were an important mechanism for preserving reference; that there were necessary a posteriori and contingent a priori truths; that identity relations between rigid designators were necessary; and argued, more tentatively, that materialist identity theories in the philosophy of mind were suspect. Interspersed with this was a considerable amount of material on natural kind terms and essentialism. As a result of these lectures and a related 1971 paper, 'Identity and Necessity' (Kripke [1971]), talk of rigid designators, Hesperus and Phosphorus, meter bars, gold and H 20, and suchlike quickly became commonplace in philosophical circles and when the lectures were published under the title Naming and Necessity in the collection The Semantics of Natural Language (Davidson and Harman [1972]), that volume became the biggest seller in the Reidel (later Kluwer) list. l The cluster of theses surrounding the idea that a relation of direct reference exists between names and their referents 2 is now frequently referred to as 'The New Theory of Reference'. 3 On December 28, 1994, Quentin Smith read a paper at the Eastern Division meetings of the American Philosophical Association that produced a different kind of sensation. In his paper, Smith suggested that most of the major ideas in the New Theory of Reference had been developed by Ruth Barcan Marcus in the period between 1946 and 1961. 4 Smith argued that Kripke had erroneously been given credit for these ideas and, more contentiously, that Kripke had heard some of these ideas at a lecture Marcus gave in February 1962, had unconsciously assimilated them (while not properly understanding them at the time) and had later incorporated them into his Princeton lectures. Word of Smith's paper had spread before the meetings and, perhaps attracted by a provocative sentence in the published abstract, 5 an overflow crowd assembled to hear the session. The proceedings were in sharp contrast to the sleepy atmosphere prevailing at most APA sessions. Smith was heckled by some members of the audience, others shouted down the chairman when he tried to limit the length of Smith's responses, and a few listeners walked out in apparent protest. The task of responding to Smith fell to Scott Soames6 who, in Vll

viii

PAUL W. HUMPHREYS AND JAMES H. FETZER

a paper twice as long as Smith's, provided a detailed set of arguments designed to show that all of Smith's claims on Marcus' behalf were groundless. Smith then read, in great haste due to time constraints, a response to Soames that was itself almost as long as Soames' reply. The session lasted for three hours and the polarized differences of opinion about the merits of the case - and the propriety of holding the session at all - continued to be discussed in the hallways at the convention, in a lead story in a mass circulation magazine, 7 in letters to the editor of the APA Proceedings,8 and elsewhere. 9 It is important to note that although Smith apparently was the first to publicly argue the misattribution claims, his paper did not emerge from a vacuum. In particular, at least as early as the mid-1980's, Ruth Marcus had sent out many copies of an eight page letter in which she made a case for her own priority in many of these matters. 10 What tended to be lost in the acrimonious debate were several important historical questions: how did the various ideas that are now known as the New Theory of Reference come to be developed? Was the picture conveyed by the casual attributions in the literature, that Kripke had created the theory from whole cloth, accurate? Was, as Smith originally claimed, the theory essentially due to Marcus or was the true history one of a cumulative set of contributions by various philosophers, with a particularly successful version of the theory eventually augmented by and effectively dramatized by Kripke? These are serious questions and ones familiar to historians dealing with older ideas. In the interests of starting an objective discussion of this question, we decided to publish the original APA papers in Synthese. We had originally planned to devote a second issue of Synthese to this topic, but the first issue generated so much interest that we arranged for the present collection of papers to be published in place of that second issue. This is an especially appropriate venue, for many of the original papers that bear on the issue originally appeared either in Synthese or in The Synthese Library. ADVICE TO THE READER

Because much of this dispute inevitably revolves around published work, one must be careful to note which version of a text is the relevant one. Both Kripke's and Marcus' work have been reprinted, sometimes many times, and changes and additions have been made in some of them, often by the authors themselves. So a few words of caution might be helpful to those readers - of whom we hope there will be many - who wish to read the original sources and to make up their own minds on the merits of the arguments presented in the collection. Marcus' paper 'Modalities and Intensional Languages' was presented at a meeting of the Boston Colloquium for the Philosophy of Science in February 1962. II It was published in an issue of Synthese XIII, noA (December 1961), pp.303-322, followed by Quine's response, op. cit., pp. 323-330. As often

INTRODUCTION

IX

happens with academic journals, that issue did not actually appear in print until some time later. 12 However, some reprintings of Marcus' paper, e.g., in her collection Modalities (Marcus, 1993) and especially in Zabeeh et al. (1974), pp. 838-853, contain changes and additions and are thus not reliable sources for historical research. In addition, a considerable portion of the disagreement between Smith and Soames revolves around remarks made by Marcus and Kripke during the discussion that occurred after Quine's response to Marcus' paper 'Modalities and Intensional Languages', this discussion being published in Synthese XIV, nos. 2/3 (September 1962), pp. 132-143. Naturally, the arguments of Smith, Soames, and others were based on the published transcript of that discussion (Marcus et al. (1962)). Having discovered that the early sessions of the Boston Colloquium were recorded on audio tape, we became curious as to whether the original recording of the discussion still existed. After an extended search, Robert S. Cohen, a former Director of the Boston Colloquium, located a large collection of unlabelled tapes and sent them to us. The relevant discussion turned out to be on the last tape sampled and a direct transcription was made from the tape. \3 It turned out that there were major differences between what was actually said at the discussion and what was eventually published in Synthese, most notably in the contributions of Ruth Marcus. Indeed, even a contemporary transcript that was circulated to the participants for approval contains many inaccuracies, omissions and changes. 14 We had originally hoped to include our new transcript as part of this collection. To this end we circulated a letter to all the participants in the original discussion informing them of our intention. Shortly thereafter, both editors were informed by Professor Marcus that she was adamantly opposed to our publishing the new transcript. As of this date, Professor Marcus is the only person who has voiced any objection to us about our publishing the new transcript. We hope, nevertheless, that at some point in the future this important part of the historical record will be made available to the philosophical community through an appropriate forum. THE PAPERS

Some comments are necessary on a few of the papers in order to provide relevant background information. The first section here, Part I, reproduces, as far as possible, the papers read at the 1994 APA session. ls Smith's paper is reprinted in its original form, with an addition to one footnote. The response by Scott Soames now includes four notes that were not part of the original. Consequently, with full concurrence on both sides, Quentin Smith has included replies to this new material in his response to Soames. These are clearly identified in the text. Part II begins with a new paper by Scott Soames that expands upon the criticisms he made in 1994 of Smith's arguments. This is followed by two

x

PAUL W. HUMPHREYS AND JAMES H. FETZER

papers by John Burgess. The first (originally published in Philosophical Studies, reprinted here with an additional footnote) directly addresses the extent to which, if at all, Marcus' work anticipates Kripke's. This is followed by a second paper by Burgess in which he attacks the historiographical methods used by Smith in his two APA papers. The section concludes with a third paper by Smith extending his earlier arguments and responding to Soames' and Burgess' previously published papers. Part III begins with an excerpt from Dagfinn F011esdal's dissertation Referential Opacity and Modal Logic. F011esdal's dissertation was submitted to the Harvard University philosophy department on April 3 1961. 16 It has previously been available only in a mimeographed version, containing slight revisions and additions, from Oslo University Press (F011esdal (1966)). F011esdal's work is especially interesting because one finds in it explicit assertions of theses that later become widely accepted. One important example is this: "This solution leads us to regard a word as a proper name of an object only if it refers to this one and the same object in all possible worlds. This does not seen unnatural. Neither does it seem preposterous to assume as we just did, that if a name-like word does not stick to one and the same object in all possible worlds, the word contains some descriptive element.! eThis attitude towards proper names is not unlike that of [Neil] Wilson e.g., in The Concept of Language.}" (Ff1JllesdaI1961 §17)

Here, the proper names are terms such as 'Hesperus' and 'Phosphorus' and the quoted sentences occur in the course of a lengthy collection of arguments designed to show that a definite description ~[X(r, can be treated as a genuine singular term if and only if the condition, ~3yDllx(¢>(x) == x = yr, holds. In addition, the characteristic formula, IIxlly(x = y J D(x = y)), can also be preserved and substitutivity of identity can be maintained in modal contexts with an ontology of non-intensional objects if this condition is imposed. There is much of interest here, including a discussion of 'Aristotelian essentialism', but the reader can make his or her own way through the material. Next is Sten Lindstrom's paper on the development of modal semantics in the 1950's, focussing on the contributions of the Swedish logician Stig Kanger and of Saul Kripke, with attention given to work of Richard Montague and of Jaakko Hintikka. These contributions are highly relevant to the development of the New Theory. This is in part because the development of modal semantics involves delicate issues of how to handle names and descriptions in modal contexts and also because Quine's arguments against the very possibility of modal logic rest, amongst other things, on the difficulties of quantifying in to such modal contexts and on the role of referentially opaque contexts. Moreover, individual variables can be viewed as the archetypes of rigid designation (they have no descriptive content) and at least some of these priority disputes in reference revolve around the differences between asserting, say, the necessity of identity relations between variables and asserting the necessity of identity relations between names. Lindstrom's careful analysis concludes that Kanger's

INTRODUCTION

Xl

contributions to modal semantics (e.g., Kanger, 1957), which were made earlier than Kripke's, are distinctive and different in kind. Lindstrom contends that Kanger's semantics are adequate for logical necessities, whereas Kripke's are adequate for metaphysical necessities. This part then concludes with a long, final paper by Smith that attempts to put the central dispute into a broader historical context. We believe that there is more than enough material here to allow the reader to decide which of the various claims to priority has merit. It would hardly be surprising, however, if any attribution to a single figure turns out to be a gross oversimplification. The historical development of ideas is often a subtle process with contributions coming from many sources, even when individuals undeniably produce singularly creative advances. There is such a thing as philosophical novelty, but modern philosophers rarely work in such isolation as to insulate then from contemporary research. It is of great importance to properly attribute ideas to their originators, and we hope that this collection has served that end. THREE SUGGESTIONS

We conclude with three suggestions. First, that others who are interested in this important period of twentieth-century philosophy would do well to continue the research that is begun here in order to provide the philosophical community with a full account of the development of ideas that have been so influential on contemporary philosophy. Secondly, that the philosophical community develop more refined criteria for what counts as a genuinely novel contribution to an area of research and for attributing ideas to one individual rather than to another. This will not be an easy task, but it is one that is long overdue and, if accomplished, might help to resolve heated debates that disagreements over priority tend to generate. Thirdly, rather than suggesting that the APA repress public discussion of these issues, it would be preferable for that organization to formulate guidelines governing the appropriate attribution of sources. It is our impression as editors of some experience that many authors are very conscientious about citing sources, yet others are quite negligent about this, where the number of those who need to be reminded to cite prior work is growing. Such guidelines, which would have to rely on a successful outcome to the second project above and would necessarily have to be somewhat flexible, could then serve as a working standard for authors, publishers, editors and referees.

August 29, 1997 Corcoran Department of Philosophy University of Virginia Department of Philosophy University of Minnesota, Duluth

xii

PAUL W. HUMPHREYS AND JAMES H. FETZER ACKNOWLEDGEMENTS

The editors could not have carried out the difficult task of compiling this volume without the help and cooperation of the contributors. We gratefully acknowledge their assistance, despite occasional deep differences of opinion along the way. Saul Kripke and Ruth Marcus, the two individuals most directly affected by this dispute, were each invited on a number of occasions to contribute their own perspectives on the issue. It is unfortunate that they declined to participate, at least in the form of written papers. Finally, we take the opportunity here to thank those individuals who acted as referees and who in many cases offered unusually detailed comments to the authors. In addition we are grateful to Jaakko Hintikka, the editor-in-chief of Synthese Library, for his support of this project, to Robert S. Cohen, and to Dagfinn Fellesdal. NOTES Naming and Necessity was reissued in 1980 by Harvard University Press with a new preface by

Kripke and as the author put it: 'Obvious printing errors have been corrected, and slight changes have been made to make various sentences or formulations clearer'. (Kripke (1980), p. I). 2 The directly referential nature of terms is not restricted to names - some descriptions and ~enera1 terms also have this feature according to the theory. This terminology appears, for example, on the front cover of the 1977 anthology Naming, Necessity, and Natural Kinds (Schwartz (1977)), which describes itself as 'A challenging collection setting forth the new theory of reference expounded by Keith S. Donnellan, Saul Kripke, and Hilary Putnam .. .'. 4 In his response to Soames, Smith extends this period to 1943 when Barcan began to write her dissertation. 5 Proceedings and Addresses of the American Philosophical Association, 68:1, p. 94. The abstract reads: "It is arguable that the most serious and widespread error in the history of recent analytic philosophy is the misattribution of the New Theory of Reference to Kripke. This theory includes the epistemic argument that proper names are directly referential and are not equivalent to definite descriptions, the modal argument that proper names designate rigidly and are not equivalent to contingent definite descriptions, the argument that there are a posteriori necessities, the necessity of identity, and related ideas. Marcus had developed the New Theory by 1961; Kripke attended Marcus' presentation of her theory in 1962 and subsequently repeated with elaboration her ideas in Naming and Necessity (1972), without the expected attributions. This omission caused readers of Naming and Necessity to misattribute the New Theory of Reference to Kripke." This abstract had apparently not been read by the APA program committee when it accepted Smith's paper, but it should be noted that the APA did not at that time require an abstract to accompany a colloquium raper until after the paper had been accepted. Soames agreed to act as respondent after at least two other philosophers had declined invitations to do so. 7 Lingua Franca Volume 5, No.5 (January/February 1996), pp. 28-39. 8 Proceedings and Addresses of the American Philosophical Association 69:2 (November 1995), p. 121; ibid 69:5 (May 1996), pp. 141-2. 9 See e.g., The Australian, May 15, 1996, Higher Education Supplement, pp. 27-28. (This is an edited version of the Lingua Franca article.) 10 We note the existence of this letter because it affected the interpretation that some philosophers have placed upon the origins of the dispute, upon the nature of the dispute itself and, more bizarrely, even upon the authorship of Smith's original paper. The editors have a copy of the letter and have checked its contents against another copy. Aside from handwritten additions, and some short prefatory remarks, the contents of the two copies are the same. Earlier, and tangentially related, claims of misattributions can be found in Fogelin (1987), pp. 241-246, and F0llesdal (1994).

INTRODUCTION

XllI

II There is some unclarity in the literature about the exact date of the paper. It is identified in a footnote to Marcus (1961) as having been presented on February 7th 1962, but a corresponding note to the discussion suggests that it was given on February 8th 1962, the same day as Quine's response and the ensuing discussion. The reprint (Wartofsky (1963» also identifies the date as February 8th, which seems to be the most likely. The difference is important because the lapse of a day between the paper and discussion would mean that the comments during the discussion were more likely to be the product of considered reflection and, possibly, informal discussion. Marcus notes that her paper was announced under the title 'Foundation [sic] of Modal Logic'. 12 Marcus' paper, Quine's reply, and the discussion were also published in Wartofsky (1963), pp. 77-116. 13 The tape also contains Quine's comments on Marcus' paper, later published as (Quine (1961». It does not, however, contain a recording of Marcus's talk itself, nor did any of the other tapes that Professor Cohen located for us. 14 For example, it transcribes the spoken 'successor function' as 'Sheffer function'. Our understanding of the process that lead to the published version is that a preliminary transcript was circulated to the participants, changes were made at that stage, and the revised transcript was then circulated with additional changes being possible at that point. Correcting the page proofs also rresumably provided an opportunity for further emendation. S We make no claim that the papers here published are exactly those read at the APA. For example, the tumultuous atmosphere at the meetings led to parts of (at least) Smith's response to Soames being skipped on that occasion. 16 Follesdal says in the preface to Follesdal (1966) "I also owe a special thanks to Mr Saul Kripke, who in the spring of 1961 imparted to me virtually all the basic ideas on iterated modalities that he presented in his 1963 and 1964 papers [Follesdal must mean Kripke (l963a)(l963b) -ed!" but there is no indication that the indebtedness goes beyond that.

REFERENCES Davidson, D. and Harman, G.: 1972, Semantics of Natural Langage. D. Reidel Publishing Company, Dordrecht. Fogelin, R.: 1987, Wittgenstein (2nd edition). Routledge, London. Follesdal, D.: 1961, Referential Opacity and Modal Logic. Unpublished Ph.D. thesis, Harvard University. Follesdal, D.: 1966, Referential Opacity and Modal Logic. Filosofiske Problemes Nr. XXXII, A. Naess (ed), Universitetsforlaget: Oslo. (A slightly supplemented reprinting of Follesdal. (1961). Follesdal, D.: 1994, 'Stig Kanger in Memoriam', in D. Prawitz, B. Skyrms and D. Westerahl (eds), Logic, Methodology and Philosophy of Science IX. Elsevier, 885-888. Kanger, S.: 1957, Provability in Logic. Stockholm Studies in Philosophy, Vol. I, Almqvist and Wiksell, Stockholm. Kripke, S.: 1963a, 'Semantical Analysis of Modal Logic I: Normal Propositional Calculi', Zeitschriftfur Mathematische Logik und Grundlagen der Mathematik, 9, 67-96. Kripke, S.: 1963b, 'Semantical Considerations on Modal Logic', Acta Philosophica Fannica, lb, 83-94. Kripke, S.: 1971, 'Identity and Necessity', in Munitz, 135-164. Kripke, S.: 1972, 'Naming and Necessity', in Davidson and Harman, 253-355 and 763-769. Kripke, S.: 1980, Naming and Necessity. Harvard University Press, Cambridge. Marcus, R.: 1993, Modalities. Oxford University Press, Oxford. Marcus, R., W. Quine, S. Kripke, D. Follesdal, and 1. McCarthy.: 1962, 'Discussion on the Paper of Ruth B. Marcus', Synthese, 14,132-143. Marcus, R.: 1961, 'Modalities and Intensional Languages', Synthese, 13, 303-322. Munitz, Milton K.: 1971, Identity and Individuation. New York University Press, New York. Quine, w.V.0.: 1961, 'Reply to Professor Marcus', Synthese, 13, 323-330. Schwartz, S.: 1977, Naming, Necessity and Natural Kinds. Cornell University Press, Ithaca. Wartofsky, M.: 1963, Boston Studies in the Philosophy of Science: Proceedings of the Boston Colloquiumfor the Philosophy of Science 1961/1962. D. Reidel Publishing Company, Dordrecht. Zabeeh, F., E. Klemke, and A. Jacobson.: 1974, Readings in Semantics. University of Illinois Press, Urbana.

PART I THE APA EXCHANGE

QUENTIN SMITH

MARCUS, KRIPKE, AND THE ORIGIN OF THE NEW THEORY OF REFERENCE

The New Theory of Reference in the philosophy of language became widespread in the 1970s and is still flourishing today. The New Theory implies that many locutions (e.g., proper names) refer directly to items, which contrasts with the traditional or old theory of reference, which implies that names and relevantly similar locutions express descriptive senses or are disguised descriptions. The New Theory encompasses such notions as direct reference, rigid designation, identity across possible worlds, the necessity of identity, a posteriori necessities, singular propositions, essentialism about natural kinds, the argument from the failure of substitutivity in modal contexts that proper names are not equivalent to contingent definite descriptions, and related ideas and arguments. Some of the contributors to the development of this theory include Kripke, Kaplan, Donnellan, Putnam, Perry, Salmon, Soames, Almog, Wettstein and a number of other contemporary philosophers. The point of this paper is to correct a fundamental and widespread misunderstanding about the origins of the New Theory of Reference; the main misunderstanding is that it is widely believed that Kripke originated the major ideas of this theory, presented in his 1972 article on "Naming and Necessity" (Kripke, 1972) and his 1971 article on "Identity and Necessity" (Kripke, 1971). The fact of the matter is that the key ideas in the New Theory were developed by Ruth Barcan Marcus, in her writings in 1946-47 (1946; 1947) and especially in her 1961 article on "Modalities and Intensional Languages" (reprinted with small changes in (Marcus, 1993)). "Modalities and Intensional Languages" was presented in February, 1962 at the Boston Colloquium for the Philosophy of Science; Marcus' commentator was Quine and Kripke participated in the discussion which followed. II

I shall begin by adducing a number of representative instances where this historical misunderstanding is evinced. The point of the following series of quotations is to illustrate the nature and extent of the misunderstanding of the origin of the New Theory of Reference. 3 P. W. Humphreys et al. (eds.), The New Theory of Reference © Kluwer Academic Publishers 1998

4

QUENTIN SMITH

If one picks out at random any essay in the philosophy of language or essentialism in the past twenty-five years or so, then one will most likely find the New Theory of Reference attributed to Kripke (and some others, e.g., Kaplan and Donnellan) and that Marcus' name will only occasionally be mentioned. A recent example is Recanati's article in Philosophical Studies: "My starting point will be the ... notion or rigidity, introduced by Saul Kripke in the philosophical literature" (Recanati, 1988: 103). Paging through recent issues of Nous, one finds David Braun listing the proponents of the new "direct reference" theory of names: "Kripke (and) Donnellan's view strongly suggest direct reference. Almog, Kaplan, Salmon, Soames and Wettstein explicitly advocate versions of Direct Reference for proper names" (Braun, 1994: 465, n. 1). It is notable that the originator of the new "direct reference" theory, Marcus, is not even mentioned. Many misattributions of Marcus' ideas to Kripke and others can be found in the writings of contemporary philosophers, but what is even more surprising is that the leading developers of the New Theory do the same. Apart from Marcus, no one has done more to develop the New Theory of Reference than David Kaplan. David Kaplan, in some of his published works, attributes the New Theory of proper names to Kripke. He correctly notes that "the term 'rigid designator' was coined by Kripke to characterize those expressions which designate the same thing in every possible world in which that thing exists and which designate nothing elsewhere". But Kaplan proceeds to identify falsely the thesis or claim that proper names are rigid designators as Kripke's specific claim (rather than Marcus'); Kaplan writes: "He (Kripke) uses it in connection with his controversial, though, I believe, correct claim that proper names, as well as many common nouns, are rigid designators" (Kaplan, 1989: 492; my italics). Of course, Marcus did not make a claim for the rigidity of common nouns, and this idea is correctly credited to Kripke, but the same does not hold for the rigidity of proper names, as we shall shortly see. Even in cases where we find philosophers listing a series of contributors to the New Theory of Reference, Marcus is typically not mentioned. John Perry, who, along with Kaplan, is the leading exponent of the New Theory of indexicals, writes that "Lessons learned from the works of Donnellan, Kaplan, Kripke, Putnam, Wettstein and other New Theorists of Reference have convinced me to accept two theses. ... First, the references of the singular terms do not depend on Fregean senses, or identifying descriptions in the mind of the speaker. ... Second, each of these utterances expresses what David Kaplan has called a 'singular proposition'" (Perry, 1988: 108). The relevant lessons, however, were first taught by Marcus and repeated with elaboration by the named individuals. In Nathan Salmon's book on Reference and Essence, he writes that versions of the New Theory of Reference were "developed, to a considerable extent independently, by several contemporary philosophers of semantics, most notably Keith Donnellan, David Kaplan, Saul Kripke, and Hilary Putnam" (Salmon, 1981: 3). Another leading contributor

MARCUS, KRIPKE, AND THE ORIGIN OF THE NEW THEORY OF REFERENCE

5

to the New Theory, Howard Wettstein, writes such things as "New theorists like Keith Donnellan, David Kaplan, Saul Kripke, John Perry, and Hilary Putnam - and my own work falls into this tradition - proffer an account (of direct reference that is anti-Fregean)" (Wettstein, 1986: 185-186). Quotations such as these can be picked out at will from the current literature. I was once myself among the guilty, calling Marcus' theory of proper names "the KripkeDonnellan theory of proper names" (Smith, 1987: 387). Occasionally her work is alluded to in the literature on the New Theory! but her central role is overlooked. It seems to me that, from the point of view of the history of philosophy, correcting this misunderstanding is no less important than correcting the misunderstanding in a hypothetical situation where virtually all philosophers attributed the origin of the Theory of Forms to Plotinus. This correction is the aim of the following several sections. In these sections, I will outline some of the basic ideas in Marcus' work that also occur in Kripke's "Naming and Necessity" and "Identity and Necessity". As I shall explain later, I believe the two main causes of the subsequent failure of philosophers to mention her work is that attributions to Marcus did not appear in the relevant places in "Naming and Necessity" and "Identity and Necessity" and that many philosophers may have been insufficiently familiar with Marcus' earlier contributions. III

I shall quote some passages from Marcus' 1961 article that reveal six main ideas she contributed to the New Theory. First, let us start with the idea that proper names are directly referential and are not abbreviated or disguised definitions, as Frege or Russell and most philosophers up to the 1970s believed. Marcus writes: But to give a thing a proper name is different from giving a unique description .... (An) identifying tag is a proper name of the thing .... This tag, a proper name, has no meaning. It simply tags. It is not strongly equatable with any of the singular descriptions of the thing. (1961: 309-310).

This is the basis of the contemporary "direct reference" theory of proper names, where proper names are argued not to be disguised descriptions. For example, "Scott" refers directly to Scott and does not express a sense expressible by such a definite description as "the author of Waverly". A second idea that Marcus introduces is that we can single out a thing by a definite description, but this description serves only to single it out, not to be strongly equatable with a proper name of the thing. She says: "It would also appear to be a precondition of language [especially assigning names] that the singling out of an entity as a thing is accompanied by many ... unique descriptions, for otherwise how would it be singled out? But to give a thing a proper name is different from giving a unique description" (1961: 309). This idea later became widely disseminated through Kripke's discussion of how

6

QUENTIN SMITH

reference-fixing descriptions are sometimes used to single out a thing as a bearer of a name, but that the names are not disguised descriptions. Kripke writes: "It seems plausible to suppose that, in some cases, the reference of a name is indeed fixed via a description [but that the description is not "part of the meaning of the name"] (Kripke, 1972: 276). Kripke, however, added the novel idea (and this is one of the main original ideas in "Naming and Necessity") that in other cases names' reference may be secured by a historical causal chain stemming back to the original "baptism" (Kripke, 1972: 298-303). A third component of the New Theory of Reference introduced by Marcus is the famous modal argument for the thesis that proper names are directly referential rather than disguised contingent descriptions. Contrary to Nathan Salmon's claim that "The modal arguments are chiefly due to Kripke" (Salmon, 1981: 24), they are chiefly due to Marcus and presented by Kripke ("Naming and Necessity" (1972: 269ft) without reference to Marcus' earlier statement of them. Let us begin with this passage, where Marcus is discovering (10) and (15) (in her notation). (10) and (15) are (10) (15)

The evening star eq the morning star Scott is the author of Waverly

The symbol "eq" stands for some equivalence relation. Types of equivalence relation include identity, indiscernibility, congruence, strict equivalence, material equivalence and others. Marcus wants to argue that the equivalence relations to be unpacked in (10) and (15) are not strong enough to support the relevant theses of the "disguised contingent description" theory of proper names. She writes (1961: 308-309): If we decide that "the evening star" and "the morning star" are [proper] names for the same thing, and that "Scott" and "the author of Waverly" are [proper] names for the same thing, then they must be intersubstitutable in every context. In fact it often happens, in a growing, changing language, that a descriptive phrase comes to be used as a proper name - an identifying tag - and the descriptive meaning is lost or ignored.

Marcus will find that not all of the relevant expressions are names for the same thing. They are not intersubstitutable in modal contexts; Marcus writes: Let us now return to (lO) and (15). If they express a true identity, then 'Scott' ought to be anywhere intersubstitutable for 'the author of Waverly' in modal contexts, and similarly for 'the morning star' and 'the evening star'. If they are not so universally intersubstitutable - that is, if our decision is that they are not simply proper names for the same thing; that they express an equivalence which is possibly false, e.g., someone else might have written Waverly, the star first seen in the evening might have been different from the star first seen in the morning - then they are not identities. (1961: 311)

Marcus' modal argument shows why the 'disguised contingent description" theory of proper names is false. Since (10) and (15) do not express identities, the expressions flanking "is" are not proper names for the same thing. In (10) and (15) a weaker equivalence relation should be unpacked, for example, by a

MARCUS, KRIPKE, AND THE ORIGIN OF THE NEW THEORY OF REFERENCE

7

theory of descriptions. (By contrast, the sentence "Hesperus is Phosphorus" evinces an identity sign flanked by the two expressions; thus, it passes Marcus' modal test for containing two proper names of the same thing.) This modal argument goes back to Marcus' formal proof of the necessity of identity in her extension of S4 (Barcan, 1946; 1947), which is a fourth component she introduced into the New Theory of Reference. She showed that (T)

(xly) § O(xly)

is a theorem of QS4, QS4 being her quantificational extension of Lewis' S4. The quadruple bar here means strict equivalence. Since identities are necessary, a failure of intersubstitutivity in modal contexts will show that a proper name does not express the relevant descriptive sense. If "Scott" is not intersubstitutable with "the author of Waverly", "Scott" does not express the sense expressed by this definite description. This opens the door to the theory that proper names do not express descriptive senses but instead are directly referential. Of course, this argument does not prove that proper names do not express senses, merely that they do not express senses of contingent definite descriptions. Marcus' modal argument is consistent with the idea of L. Linsky (1977) and A. Plantinga (1978) that proper names express senses expressible by necessary definite descriptions, which are definite descriptions that express modally stable senses. For example, "Scott" may express the modally stable sense of "the actual author of Waverly". In order to rule out this modally stable descriptive theory of proper names, one needs further argumentation, such as the epistemic argument that proper names are directly referential. If the descriptive theory of proper names is true, Le., proper names are defined by descriptions, then "Venus is the evening star" should express a truth knowable a priori, i.e., knowable merely by reflection upon the concepts involved. But it cannot be known a priori that Venus is the evening star; this is known a posteriori, through observation of the empirical facts. As Marcus writes: You may describe Venus as the evening star, and I may describe Venus as the morning star, and we may both be surprised that, as an empirical fact, the same thing is being described. But it is not an empirical fact that (17) Venus I Venus. (1961: 310)

Here "I" is the identity symbol. If "Venus" expresses the modally stable sense expressible by "whatever is actually the evening star and morning star", then the persons designated by "you" and "I" in the passage quoted from Marcus' article should be able to know a priori, simply by reflection upon the semantic content of the expressions "Venus", "the morning star" and "the evening star" that Venus is both the morning star and the evening star. The fact that they cannot know this indicates that "Venus" does not express the modally stable sense expressed by "whatever is actually the evening star and morning star".

8

QUENTIN SMITH

Thus we have the irony that Plantinga's and Linsky's theory of proper names was refuted years before they formulated it, unbeknowst to themselves (and unbeknownst to later New Theorists)! Marcus' arguments for the "direct reference theory" make manifest her discovery of a fifth crucial component of the New Theory of Reference, the concept of rigid designation (although the name of this concept, "rigid designation" was first coined by Kripke). "Hesperus" is intersubstitutable salva veritate with either occurrence of "Phosphorus" in "Necessarily, Phosphorus is Phosphorus". Each of these two names actually designates Venus in respect of every possible world in which Venus exists and does not actually designate anything in respect of worlds in which Venus does not exist. If these two names were instead equivalent to contingent descriptions (e.g., "the morning star" and "the evening star"), they would not be intersubstitutable salva veritate in this modal context and thus would be non-rigid designators. Marcus notes in her 1970 APA paper, "Essential Attribution", presented at a symposium at which Kripke was one of the symposiasts, that "individual names don't alter their reference, except to the extent that in (respect of) some worlds they may not refer at all" (1971: 194). Although I have used the "rigid designation" terminology, Marcus does not use it, since Kripke's introduction of this expression in his "Identity and Necessity" (1971) assimilated proper names to some descriptions (viz., modally stable descriptions), which obscure their different semantic properties. Marcus' points can be accommodated, consistently with the continued use of "rigid designators", if we make the following classification, which is familiar to those working with the New Theory of Reference. Adopting the genus/species terminology, we may say that the genus is rigid designators, and the different species are (a) proper names, (b) referentially used definite descriptions (in Donnellan's sense), (c) attributively used definite descriptions that express a modally stable sense, (d) uses of indexicals, (e) natural kind terms, and certain other expressions. We avoid assimilating proper names to some modally stable descriptions, since proper names refer directly, whereas attributively used definite descriptions that express modally stable senses refer indirectly, via the expressed sense. A sixth idea introduced into the New Theory of Reference by Marcus is the idea of a posteriori necessity. Recall our earlier quotation of Marcus' remarks about Venus and the evening star: You may describe Venus as the evening star, and I may describe Venus as the morning star, and we may both be surprised that, as an empirical fact, the same thing is being described. But it is not an empirical fact that (17) Venus I Venus. (1961: 85)

Consider the expression "Hesperus is Phosphorus". We do not know this to be true a priori. It is not an analytic assertion whose truth value is known by analysis of the concepts involved. Nonetheless, it is necessarily true, since both

MARCUS, KRIPKE, AND THE ORIGIN OF THE NEW THEORY OF REFERENCE

9

names directly refer to the same thing, Venus. It is true that (a)

Hesperus I Phosphorus,

whereas, as before, "I" is the sign of identity. Given Marcus' theorem of the necessity of identity, it follows that (b)

Necessarily, Hesperus I Phosphorus.

Thus "Hesperus is Phosphorus" can be viewed as a synthetic a posteriori necessary truth. This belies Salmon's historical comment about the sentence "Hesperus, if it exists, is Phosphorous". He writes of Kripke's three 1970 talks at Princeton, published in 1972 as "Naming and Necessity", that "In 1970 Saul Kripke astonished the analytic philosophical community with his claim supported by the rich theoretical apparatus of possible-world semantics and his new 'picture' of reference - that [the mentioned sentences], though synthetic and a posteriori contain necessary truths, propositions true in every possible world" (Salmon, 1986: 2).2 A more accurate statement would be that Kripke eloquently elaborated upon Marcus' idea and extended it to new sorts of items, such as "water is H 2 0". What explains the wide misunderstanding of the historical origins of the New Theory of Reference, a major movement in the history of analytic philosophy? I have already suggested that two reasons may be that many philosophers were insufficiently familiar with Marcus' earlier work and that Kripke did not attribute the relevant ideas to her in "Naming and Necessity" and "Identity and Necessity" (despite the fact that he was present when she presented her seminal work in 1962 and was undoubtedly familiar with her earlier formal papers on modal logic). There seems to me a plausible explanation of why Kripke did not make these attributions. He writes in the Preface to the 1980 edition of Naming and Necessity that "The ideas in Naming and Necessity evolved in the early sixties - most of the views were formulated in about 1963-64" (1980: 3). Of course some of the ideas in "Naming and Necessity" are genuinely new, such as the causal chain picture of the reference of names, the idea that natural kinds are rigid designators and the theory of the necessity of origins. But since most of the views in "Naming and Necessity" occur in 1961 with Marcus' article "Modalities and Intensional Language" and in her formal work on modal logic in 1946-47, we need to look at Kripke's remark about the evolution of his views from this perspective. First, recall that Marcus presented her paper "Modalities and Intensional Languages" in February 1962 at the Boston Colloquium for the Philosophy of Science. Her paper was published, in conjunction with the colloquium, in the 1961 volume of Synthese. Kripke was present and participated in the discussion, which was subsequently published. Kripke, it seems, did not wholly grasp Marcus' ideas at this time. During the discussion of the paper at the 1962 colloquium, Kripke made

10

QUENTIN SMITH

the following remark about Marcus' theory: The tags are the "essential" denoting phrases for individuals, but empirical descriptions are not, and thus we look to statements containing "tags", not descriptions, to ascertain the essential properties of individuals. Thus the assumption of a distinction between "names" and "descriptions" is equivalent to essentialism. (Marcus et aI., 1962: 142)

This is mistaken, since Marcus clearly did not claim in her article that things have their names essentially. As Marcus later explained, "that was not my claim. Socrates might have been named Euthyphro; he would not thereby be Euthyphro" (1993: 226-227). To the contrary, her claim was that names are what later came to be described as directly referential. They are not denoting phrases, essential or otherwise. This suggests how we may understand the statement in the 1980 Preface to Naming and Necessity that "most of the views were formulated in about 196364". We should interpret this as suggesting that Kripkefirst correctly understood Marcus' theory in 1963-64 and that before this time, he did not grasp what she conveyed in the presentation he attended. In particular, it was Marcus' theory of the necessity of identities, where names flank the identity sign, and the associated ideas of direct and rigid reference that became clear to Kripke in subsequent years. We should perhaps interpret this 1980 passage from Kripke in this light: He says "Eventually I came to realize - this realization inaugurated the aforementioned work of 1963-64 - that the received presuppositions against the necessity of identities between ordinary names were incorrect, that the natural intuition that the names of ordinary language are rigid designators can in fact be upheld" (1980: 5). But why does Kripke not say instead that at this time he first came to understand Marcus' arguments for the necessity of identity and the directly referential and rigid character of proper names? In the 1972 essay, he attributes one idea to Marcus: "Marcus says that identities between names are necessary" (1972: 305). But instead of explaining how this idea and Marcus' other ideas formed the theoretical basis of "Naming and Necessity", Kripke goes on to criticize a minor aside made by Marcus, viz., that a good dictionary should be able to tell one if "Hesperus" and "Phosphorus" have the same reference (1972: 305). (But as Marcus later explained, she should have made it clear that what she had in mind was a "dictionary" that functions as an encyclopedia, where coreferring names are listed as in a biographical dictionary (Marcus, 1993: 34, n. 1).) I believe a reasonable explanation of why Kripke did not attribute the central features of the "New Theory" to Marcus is that he originally misunderstood Marcus' New Theory of Reference. When he eventually understood it, after a year or two, the insight that came made it seem that the ideas were new. I suspect that such instances occur fairly frequently in the history of thought and art. 3 Philosophy Department Western Michigan University

MARCUS, KRIPKE, AND THE ORIGIN OF THE NEW THEORY OF REFERENCE 11 NOTES Some philosophers have noted in passing some of Marcus' contributions, but have not recognized their full extent. 1. Almog (1986: 220, n. 8) goes further than most in recognizing some of Marcus' contributions (e.g., the idea of direct reference), but confines this remark to a footnote in an article devoted to the theory that proper names are rigid designators, which he believes was originated by Kripke rather than Marcus. Another example is Alan Sidelle's recognition that Marcus, not Kripke, first formulated the idea of the necessity of identity (1989: 25), but Sidelle notes this in a book devoted to the idea of a posteriori necessity, which he believes was originated by Kripke rather than Marcus. John McDowell (1994: 105, n. 28) mentions the trend towards construing some expressions as directly referential and, after referring to Kripke and Donnellan as "early proponents of this trend", writes: "See also, predating the trend, Ruth Barcan Marcus, 'Modalities and Intensional Languages', Synthese, 27, (1962), 303-22". {Material added in May 1995: It should be added that in publications subsequent to the ones I quoted from Kaplan, N. Salmon, and H. Wettstein, each of them credits Marcus for developing at least one of the ideas for the New Theory of Reference. These appear in Salmon's Frege's Puzzle, Wettstein's Has Semantics Rested on a Mistake? and in Kaplan's contribution to Modality, Morality and Belief Essays in Honor of Ruth Barcan Marcus, Walter Simmot-Armstrong et al. (eds.).} 2 Of course Salmon is expressing here the nearly universal misunderstanding of the origin of this idea. For another example, consider Sidelle's comment in his book devoted to the idea of a posteriori necessity: "Enter Kripke and his Naming and Necessity. Kripke made it very plausible that there are necessary truths that are synthetic and knowable only a posteriori. Some of the more familiar examples are 'Hesperus and Phosphorus'" (Sidelle, 1989: 2). 3 Even official histories, such as Munitz's Contemporary Analytic Philosophy, get it wrong. In the chapter where the ideas developed by Marcus are explained, we find the chapter title "Referential Opacity, Modality, and Essentialism: Saul Kripke".

REFERENCES Almog, Joseph: 1984, 'Semantical Anthropology'. in P. French et aI., eds., Midwest Studies in Philosophy, 9, 479-90. Almog, Joseph: 1986, 'Naming Without Necessity', The Journal of Philosophy, 83, 210-42. Braun, David: 1993, 'Empty Names', Nous, 27(4), 443-69. Donnellan, Keith: 1966, 'Speaking of Nothing', The Philosophical Review, 83, 3-32. Kaplan, David: 1985, 'Dthat', in A. Martinich, ed., The Philosophy of Language. Oxford University Press, New York. Kripke, Saul: 1980, Naming and Necessity. Harvard University Press, Cambridge, Mass. Kripke, Saul: 1972, 'Naming and Necessity', in D.Davidson and G. Harman, eds., Semantics of Natural Language. Reidel, Dordrecht, pp. 253-355. Kripke, Saul: 1971, 'Identity and Necessity', in M. Munitz, ed., Identity and Individuation. New York University Press, New York. Linsky, Leonard: 1977, Names and Descriptions. University of Chicago Press, Chicago. Marcus, Ruth Barcan: 1993, Modalities. Oxford University Press, New York. Marcus, Ruth Barcan: 1971, 'Essential Attribution', Journal of Philosophy, 68, 187-202. Marcus, Ruth Barcan: 1961, 'Modalities and Intensional Languages', Synthese, 13, 303-22. [Marcus] Barcan, R.: 1946, 'A Functional Calculus of First Order Based on Strict Implication', Journal of Symbolic Logic, 11, 1-16. [Marcus] Barcan, R.: 1946, 'The Identity of Individuals in a Strict Functional Calculus of First Order', Journal of Symbolic Logic, 12, 12-15. Marcus, Ruth Barcan et al.: 1962, 'Discussion of the Paper of Ruth B. Marcus', Synthese, 14, 13243. McDowell, John: 1994, Mind and World. Cambridge, Mass. Munitz, Milton: 1981, Contemporary Analytic Philosophy. Macmillan Pub. Co., New York. Perry, John: 1977, 'Frege on Demonstratives', The Philosophical Review, 86, 474-97. Perry, John: 1988, 'Cognitive Significance and New Theories of Reference', Nous, 22, 1-18.

12

QUENTIN SMITH

Plantinga, Alvin: 1978, 'The Boethian Compromise', American Philosophical Quarterly, 15, 129-38. Recanati, F.: 1988, 'Rigidity and Direct Reference', Philosophical Studies, 103-17. Salmon, Nathan: 1981, Reference and Essence, Princeton University Press, Princeton. Salmon, Nathan: 1986, Frege's Puzzle. MIT Press, Cambridge, Mass. Sidelle, Alan: 1989, Necessity, Essence, and Individuation. Cornell University Press, Ithaca. Smith, Quentin: 1987, 'Problems with the New Tenseless Theory of Time', Philosophical Studies, 52, 77-98. Smith, Quentin: 1993, Language and Time. Oxford University Press, New York. Soames, Scott: 1987, 'Direct Reference, Propositional Attitudes, and Semantic Content', Philosophical Topics, 15, 47-87. Wettstein, Howard: 1986, 'Has Semantics Rested on a Mistake'?, The Journal of Philosophy, 83, 185-209.

SCOTI SOAMES

REVISIONISM ABOUT REFERENCE: A REPLY TO SMITH

My task today is an unusual and not very pleasant one. I am not here to debate the adequacy of any philosophical thesis. Rather, my job is to assess claims involving credit and blame. According to Quentin Smith, the central doctrines of Naming and Necessity were developed by Ruth Marcus in her pioneering papers on quantified modal logic in the late 40's, and in her paper, 'Modalities and Intensional Languages' in 1961. 1 Smith maintains that Saul Kripke learned these doctrines from her, initially misunderstood them, and, when he later straightened things out, mistakenly took the doctrines to be his own. Finally, Kripke is supposed to have published them without properly citing her. The entire profession was allegedly fooled, despite the fact that Kripke and Marcus were among its most well known members, and their work was familiar to leading researchers in the field. For years nobody said anything. Now, more than 20 years later, Smith claims to be bringing the truth to light. In what follows I show that the charges Smith makes against Kripke are false, and that the historical picture he paints is inaccurate. However, before I begin, I want to make clear that although Smith takes himself to be championing Marcus, my criticisms are of him, not her. I take a back seat to no one in my respect, admiration, and affection for both Ruth and Saul. As you will see from my comments on particular matters of substance, Marcus, along with certain other philosophers, do deserve credit for anticipating important aspects of contemporary theories of reference. However this credit in no way diminishes the seminal role of Saul Kripke. With this in mind let me review some ofthe accomplishments of Ruth Marcus. She is, deservedly, one of the most distinguished and well-known philosophers in America. She is widely recognized and admired for her pioneering work in quantified modal logic, and for her important contributions to a variety of related topics. In 1946 she published the first axiomatic systems of quantified modal logic, together with a collection of theorems provable in the systems, plus some proof theoretic metatheorems? In 1947 she extended her systems to include second order quantification, and defined identity in terms of secondorder indiscernibility. 3 She then established a variety of proof-theoretic facts concerning this conception of identity. One of these involved a quantified version of the thesis of the necessity of identity. Specifically, her axiomatic system of second order quantified modal logic (S4) had as one of its theorems the claim that for all individuals x and y, ifx = y, then necessarily x = y.4 13 P. W. Humphreys et al. (eds.), The New Theory of Reference © Kluwer Academic Publishers 1998

14

scon SOAMES

In addition to developing these formal systems, Ruth Marcus successfully defended them in the early 1960's against Quine's influential, but ultimately wrong-headed, attacks on necessity in general, and on quantifying into formulas prefixed with modal operators in particular. In the course of developing this defense, she suggested that ordinary proper names might be Russellian logically proper names - terms whose meanings, or propositional contents, are nothing more than their referents. In the 60's and 70's she published important articles on essentialism as a plausible metaphysical doctrine, and helped clarify the minimal extent - namely its intelligibility - to which modal logic, as a system of logic, is committed to it. During the same period, she explored and defended substitutional interpretations of some quantifiers as both formally legitimate and potentially fruitful philosophically. The 1980's and 1990's have seen a series of papers on a wide range of topics, including belief, rationality, direct reference, moral dilemmas and moral consistency. This remarkable record of achievement is widely acknowledged and establishes Ruth Marcus as one of the leading philosophical logicians of the past 50 years. The question before us is whether a further accomplishment should be added to this already impressive list. I mentioned that in her early work Marcus suggested that ordinary proper names might be Russellian logically proper names. This raises a question regarding the relationship between her early work on this subject and the later development of contemporary non-Fregean theories of reference by philosophers such as Saul Kripke, David Kaplan, Keith Donnellan and others. Despite the similarity between some of the doctrines advocated by Marcus and those later incorporated into the new theories attributed to these philosophers, Marcus is generally not accorded a prominent role in the development of these theories. Quentin Smith claims there is an historical injustice here. On his view, the so-called "new theory of reference" is Marcus' theory. He claims that its main ideas were developed in her articles on modal logic in 1946 and 1947, and in her 1961 paper, 'Modalities and Intensional Languages.' Later theorists, especially Kripke, are viewed mainly as elaborating these ideas, and adding a few subsidiary points, while denying Marcus proper credit. A careful look at the historical record will show that this is not so. For example, Marcus' early papers on modal logic, in 1946 and 1947, developed certain formal systems in a language in which names do not occur at all, and in which the only singular terms are individual variables. The restricted version of the necessity of identity derived from the axioms of the system therefore does not involve names. These papers are not concerned with natural language, and do not contain formulations or defenses of any theses about proper names. Moreover, the distinction between proper names and descriptions, and the thesis that coreferential proper names are substitutable in modal contexts without changes in truth value, were not original with Marcus. These doctrines appeared in the literature on modal logic in response to Quine by Arthur

REVISIONISM ABOUT REFERENCE

15

Smullyan in 1947 5 and 1948,6 and Frederick Fitch in 1949. 7 In her later work, especially the 1961 paper cited by Smith, Marcus adopted these theses of Smullyan and Fitch, essentially suggesting that ordinary proper names might be Russellian logically proper names. However, even here names were not Marcus' main concern, and theses about their meaning and reference were only briefly indicated, rather than systematically explored and argued for. Her entire discussion of the meaning and reference of names, as well as their relations to descriptions, covers only five or six pages; and all of Smith's citations come from four of those pages. This is significant because it constitutes essentially all the evidence he provides for his claim that the central doctrines developed in the late sixties and seventies by Kripke are really due to Marcus. One clue that this claim cannot be correct comes from the observation that proper statements and defenses of the various doctrines of the so called "new theory" would not fit into such a small space. There is, of course, much more at issue than limitations of space. In what follows I will explain in detail why the historical picture Smith sketches is inaccurate. First, I will survey a series of relevant papers between 1943 and 1961 dealing with quantified modal logic and the behavior of names. These include papers and reviews by Ruth Marcus, Willard van Orman Quine, Arthur Smullyan, Frederick Fitch, and Alonzo Church. I will cite the origins in this literature of various theses about names, descriptions, identity, and modality; and I will explain the philosophical context in which those theses arose. Second, I will go through, one by one, Quentin Smith's claims about the six central ideas allegedly introduced into the new theory of reference by Ruth Marcus, and I will indicate the ways he goes wrong. Third, I will sum up the relationship between these theses about modality, identity, names, and descriptions in this early literature by Smullyan, Fitch, and Marcus, on the one hand, and the much more fully developed theses of the so-called "new theory of reference" that emerged in the 70s, on the other. The historical survey begins with Quine's 1943 paper, 'Notes on Existence and Necessity,.8 It was there that Quine first set out his argument that ordinary objectual quantification into any opaque construction is unintelligible, and hence illegitimate. For Quine, a construction is opaque iff for some pair of coreferential singular terms, substitution of one for the other in some formula within the scope of the construction changes truth value. Quine thought that such failures of substitution show that occurrences of singular terms within the opaque construction are not, as he put it, purely designative. By this he meant that their contributions to the truth values of the opaque sentences in which they occur are not exhausted by their reference. Using these notions, we may reconstruct his argument as follows: (P 1)

Occurrences of singular terms in opaque constructions are never purely designative.

16

SCOTI SOAMES

(P2)

Bindable occurrences of objectual variables must be purely designative.

(Cl)

An occurrence of an objectual variable within an opaque construction cannot be bound by a quantifier outside the construction.

(P3)

Modal constructions are opaque.

(C2)

Objectual quantification into modal constructions is illegitimate.

This conclusion was, of course, violated in Ruth Marcus' systems of quantified modal logic. With this in mind, Quine published a short article in 1947, 'The Problem of Interpreting Modal Logic',9 in which he reiterated his earlier conclusion and explicitly mentioned Ruth Marcus as one to which it applies. 1o Although Quine'S arguments were mistaken, they were enormously influential, and they baffled large numbers of the profession for decades. Fortunately, thanks in large part to recent work by David Kaplan, II Ali Akhtar Kazmi,12 and Mark Richard,13 we can now clearly see where they went wrong. Quine'S premises 1 and 2 are unjustified, and indeed false. 14 I won't go into a detailed explanation of this except to reiterate a certain point made by David Kaplan. Kaplan observed in his 1986 paper, 'Opacity', that it follows from Quine's characterization of a purely designative occurrence of a singular term that some occurrences of such terms in opaque contexts must fail to be purely designative; but it does not follow that all occurrences of singular terms in such contexts fail to be purely designative. If there are different kinds of singular terms - e.g., Fregean definite descriptions, proper names, individual variables - then it may be that occurrences of certain of these kinds fail to be purely designative within a given type of opaque construction, while occurrences of other kinds of singular terms remain purely designative. For example, one may have an analysis in which occurrences of Fregean definite descriptions within modal constructions are not purely designative,15 even though occurrences of names and variables are purely designative. 16 Or, one may propose an analysis in which neither occurrences of names nor occurrences of descriptions in propositional attitude constructions are purely designative,17 even though occurrences of individual variables in such constructions are purely designative. As Mark Richard and Ali Akhtar Kazmi have made clear, one can even invent constructions into which objectual quantification makes sense despite the fact that occurrences of objectual variables within such constructions are not always purely designative. IS In sum, there simply is no logical or semantic problem about quantifying into opaque constructions in general. None of these errors in Quine'S argument depends on special features of modality, or the behavior of names in modal contexts. However, quantified modal logic was Quine'S primary target, and the argument itself tended to get confused (both by Quine and by others) with a different criticism that was special to modality, namely the (equally mistaken) view that quantified modal

REVISIONISM ABOUT REFERENCE

17

logic was committed to an objectual form of essentialism. Thus it is understandable that the initial responses to the argument focused on the special case of modality. Quine's 1947 paper attacking the legitimacy of quantifying into modal constructions was quickly followed by a pair of responses by Arthur Smullyan in 1947 and 1948, and a further response by Frederick Fitch in 1949. The crucial point common to these responses was a sharp distinction between definite descriptions and genuine names. Smullyan and Fitch maintained that coreferential names are substitutable in modal constructions without change in truth value, whereas codesignative descriptions in general are not. They also accepted a corollary of this point, namely the claim that where a and f3 are genuine names, the sentence a = f3 is, if true, necessarily true. They recognized, of course, that an analogous point does not hold for definite descriptions. In fact, both Smullyan and Fitch denied that definite descriptions are singular terms at all, analyzing them instead as incomplete symbols, along Russellian lines. 19 Their basic criticism of Quine was that his argument wrongly treated definite descriptions as if they were genuine names, despite the fact that the two are logically quite different. Indeed, with definite descriptions not being counted as singular terms at all, the modal logician is free to restrict the class of closed singular terms to genuine names, in which case modal constructions won't count as opaque in the sense needed in Quine'S argument. There is no question, then, that Smullyan and Fitch regarded genuine names as fundamentally different from definte descriptions. Moreover, they regarded the behavior of the two in modal constructions as showing this. Neither Smullyan nor Fitch had the fundamental semantic notion of the referent of a term at a world, let alone the concept of rigid designation - i.e., the concept of a term that refers to the same thing at all worlds. Nevertheless, their recognition of the logical properties of genuine names in modal constructions led them to treat names in the way that one would treat them if one explicitly recognized them to be rigid designators. In this sense, which is the sense that applies to Ruth Marcus in her 1961 article, they can be credited with an implicit recognition that genuine names are rigid designators. In his 1947 review of Quine, Smullyan adds to this the claim that coreferential names are synonymous, which strongly suggests that genuine names have no meaning apart from their referents, another view later endorsed by Marcus. This was the point on which Smullyan and Fitch were criticized by Alonzo Church in one of his typically insightful reviews in The Journal of Symbolic Logic. Reviewing Fitch's article in 1950, Church says that Fitch "holds (with Smullyan) that two proper names of the same individual must be synonymous. It would seem to the reviewer that, as ordinarily used, 'the Morning Star' and 'the Evening Star' cannot be taken to be proper names in this sense; for it is possible to understand the meaning of both phrases without knowing that the Morning Star and the Evening Star are the same planet. Indeed, for like reasons, it is hard to find any clear example ofa proper name in this sense".20 In this passage, Church uses

18

SCOTT SOAMES

the point about synonymy, and the meaning of proper names, to raise a critical issue about the discussions of Smullyan and Fitch. Both Smullyan and Fitch assumed that there is a class of singular terms - which I have been calling "genuine names" - that function exactly as what Russell called logically proper names, save for the fact that the referents of these names are not restricted to objects of direct Russellian acquaintance. Moreover, they seemed to suggest that at least some uses of ordinary proper names might qualify as names in this sense. However, they did not present any systematic examination of our use of ordinary proper names, nor did they produce a battery of arguments supporting their contentions. Because of this, it was inevitable that many readers of their discussions would retain substantial, and indeed unanswered, doubts about whether names as we ordinarily use them qualify as genuine names in their sense. Much the same can be said of Ruth Marcus' discussion of names in her 1961 article, "Modalities and Intensional Languages". As I indicated before, the discussion of names in the article is short, limited to just a few pages. The main themes of the article were first, that intensionality and extensionality come in degrees, and so there is no simple intensional/extensional distinction, and second, that Quine's attacks on modality in general, and quantifying into modal constructions in particular, are mistaken. The topic of names comes up, as it did with Smullyan and Fitch, in the course of answering Quine. In her discussion, Marcus reiterates the central points of Smullyan and Fitch namely that (i) coreferential names are substitutable in modal contexts without change in truth value whereas descriptions generally are not, (ii) identity statements involving proper names are necessary if true, (iii) names are not equivalent to descriptions, (iv) descriptions are not genuine singular terms, (v) proper names have no meanings over and above their referents, and (vi) coreferential names have the same semantic content, and are substitutable for one another in all nonquotational contexts. 21 It should be noted that although Marcus' discussion of these points is brief, it is more detailed and explicit than those of Smullyan and Fitch. For example, in discussing the necessity of identity statements involving genuine singular terms, she says that true statements of identity "must be tautologically true or analytically true".22 She also holds that only genuine singular terms can flank the identity sign, and that the terms involved in a true statement of identity must be intersubstitutable not just in modal constructions, but in every construction, including, presumably, attitude constructions such as those involving belief or knowledge. 23 The idea here, I take it, is the familiar Russellian view that since the meaning of a genuine singular term is its referent, any construction in which the meaning of the whole is a function of the meanings of its parts is one in which substitution of coreferential singular terms must preserve meaning, and hence truth value. Marcus treats names as genuine singular terms. In the text of her 1961 article, it certainly seems that when she speaks of

REVISIONISM ABOUT REFERENCE

19

names, Marcus means ordinary proper names. Nevertheless, essentially the same doubts that Church raised against Smullyan and Fitch arise against Marcus, and remain similarly unanswered. These doubts come up in a somewhat veiled way at one point in the discussion of her paper that occurred in 1962 at the Boston Colloquium for the Philosophy of Science. 24 Kripke (who was an undergraduate in the audience): Forgetting the example of numbers, and using your interpretation of quantification - there's nothing seriously wrong with it at all - does it not require that for any two names, 'N and 'B', of individuals, 'A = B' should be necessary, if true at all? And if 'N and 'B' are names of the same individual, that any necessary statement containing 'N should remain necessary if 'N is replaced by 'B'? Marcus: We might want to say that for the sake of clarity and ease of communication, it would be convenient if to each object there were attached a single name. But we can and we do attach more than one name to a single object. We are talking of proper names in the ideal sense, as tags, and not descriptions. Presumably, if a single object had more than one tag, there would be a way of finding out such as having recourse to a dictionary or some analogous inquiry which would resolve the question as to whether the two tags denote the same thing. If 'Evening Star' and 'Morning Star' are considered to be two proper names for Venus, then finding out that they name the same thing that 'Venus' names is different from finding out what is Venus' mass, or its orbit. It is perhaps admirably flexible, but also very confusing, to obliterate the distinction between such linguistic and properly empirical procedures. [my emphasisf5

Note the echo of Church here. 26 It is, I think, noteworthy that we find in this passage a hedged reference to "names in the ideal sense", and to the results of consulting a dictionary as opposed to invoking what Marcus calls "properly empirical procedures". The passage seems to suggest that it is to some degree an open question whether our ordinary use of proper names is one in which they are names in the "ideal sense" of Marcus' semantic theses. Altbougb I tbink ber intention in tbe paper was to take ordinary proper names to be names in tbe ideal sense, tbis cautious qualification seems to reflect an awareness of tbe continuing relevance of Cburcb-like doubts, and tbe lack of a battery of arguments sbowing tbat ordinary proper names really do have the semantic properties required by the Smullyan-Fitcb-Marcus response to Quine. So much for the historical background. We are now ready to examine Smith's contention that six fundamental theses of what he calls "the new theory of reference" were introduced by Ruth Marcus. In evaluating Smith's claims, I will put aside the qualification just noted and interpret Marcus, as well as Smullyan and Fitch, as suggesting that ordinary proper names are genuine names "in the ideal sense". Smith's first claim is that Marcus introduced the idea that proper names are not abbreviated or disguised descriptions, but rather are directly referential. This is false. Although Marcus did regard names as directly referential, and did not take them to be disguised descriptions, she did not introduce the idea; nor did she claim to. As we have seen, Smullyan and Fitch both invoked that idea in earlier responses to Quine that Marcus took herself to be repeating and elaborating. 27 ,28

20

seOTI SOAMES

Smith's second claim is more elusive. He introduces the claim by saying "A second idea that Marcus introduces is that we can single out a thing by a definite description, but this description serves only to single it out, not to be strongly equatable with a proper name of that thing". The idea that we can single out a thing by a definite description is, of course, not new, and was not introduced by Marcus. The idea that a description which applies to a thing is not equivalent to a name of that thing was already covered in Smith's first contention - namely that Marcus supposedly introduced the idea that names are not abbreviated or disguised descriptions but rather are directly referential. So what new point is supposed to be covered by Smith's second contention? What is the second idea that Marcus is supposed to have introduced? Smith provides a clue when he says, "This idea [namely the second idea allegedly introduced by Marcus] later became widely disseminated through Kripke's discussion of how reference-fixing descriptions are sometimes used to single out a thing as a bearer of a name, but that names are not disguised descriptions." Smith contrasts this means of reference fixing with the historical chain mechanism, describing the latter, and only the latter, as original with Kripke. Thus it appears that Smith's second contention is that Marcus introduced the idea that the reference of a name may be semantically fixed by a description, even though the description is not synonymous with the name. If this is his contention, it is false. Here is the quote from Marcus that Smith gives to support his point. It would also appear to be a precondition oflanguage [especially assigning names] that the singling

out of an entity as a thing is accompanied by many ... unique descriptions, for otherwise how would it be singled out? But to assign a proper name is different from giving a unique description. (Bracketed words above inserted into the text by Smith.)

The problem here is that there is no way to get from this brief remark to the doctrine that sometimes the referent of a name is semantically fixed by an associated definite description. To see this, one must be clear what the content of that doctrine really is. It is not just that sometimes, or as Marcus seems to think, always, when we have an individual in mind for which we have a name, or to which we want to assign a name, we also have one or more descriptions that we are ready to apply to the individual. Rather, a specific description must, as a matter of semantics, be linked to the name. This means that if the referent of a name is semantically fixed by a certain description, then being a competent speaker who understands the name will involve knowing that if the name has a referent at all, it must be one that satisfies the description. Similarly, if a name N has its referent semantically fixed by a description D, then one who understands the sentence "If N exists, then N is D" will know, without empirical investigation, that it expresses a truth. Finally, if one later finds that the description one associated with the name fails to pick out anything, or fails to pick out the individual one had in mind, then it will follow that the name failed to refer, or at least failed to refer to the individual one had in mind. None

REVISIONISM ABOUT REFERENCE

21

of this is remotely suggested in the quotation from Marcus. The reason it isn't is that Marcus didn't have the doctrine in mind, as can be seen by quoting the full Marcus text from the beginning of the relevant paragraph, without ellipsis, or insertion of Smith's own words. That any language must countenance some entities as things would appear to be a precondition for language. But this is not to say that experience is given to us a collection of things, for it would appear that there are cultural variations and accompanying linguistic variations as to what sorts of entities are so singled out. It would also appear to be a precondition of language that the singling out of an entity as a thing is accompanied by many - perhaps an indefinite number - of unique descriptions, for otherwise how would it be singled out? But to assign a thing a proper name is different from giving a unique description. 29

What immediately follows in Marcus' text is an example in which we theorists assign random numbers as names for the entities singled out as things by a given culture. 30 These numbers are not intended as examples of names whose referents are semantically fixed by descriptions. On the contrary, Marcus calls such names "tags" and says they have no meaning. Marcus' point in the passage was not to sketch any particular theory about how names get their reference, but simply to distinguish names from descriptions. What she was really saying was that despite the fact that recognizing something as a thing presupposes a readiness to apply descriptions to it, nevertheless we have a linguistic device, the proper name, which allows us to refer to a thing without describing it. There is nothing here about semantic mechanisms by which the referents of names are fixed. Smith's third contention is that Marcus introduced "the famous modal argument for the thesis that proper names are directly referential rather than disguised contingent descriptions". Now Marcus did hold that names and descriptions typically behave differently in modal constructions, and so cannot be regarded as equivalent. Moreover, it is reasonable to take this as a version of "the modal argument", later made famous by Kripke. However, it was not introduced by Marcus, since Fitch and Smullyan made the same point earlier. It seems to me that all three of these philosophers deserve a degree of recognition that they are often not given. However, one should not lose sight of certain factors that made Kripke's presentation of the modal argument so compelling, whereas the others had not been. One such factor was the context of Kripke's presentation of the modal argument in conjunction with a whole battery of other arguments and significant distinctions, all designed to undercut description theories of proper names. To take one example, many philosophers find it much easier to accept the claim that coreferential proper names are substitutable without change in truth value in modal constructions than to accept the claim that they are similarly substitutable in all constructions, including propositional attitude ascriptions. By linking the two claims, Marcus made objections to the latter seem like objections to the former; by separating them, Kripke makes the claim about the substitutivity of names in modal constructions more palatable. 31

22

SCOTT SOAMES

Another point to keep in mind is that Kripke's presentation of the modal argument was not limited to the behavior of names in modal constructions, but was explicitly concerned with the evaluation of simple sentences (free of modal operators) at alternative possible worlds. 32 The basic intuition tapped by Kripke's version of the modal argument was not only about the truth value of modal sentences, like Necessarily Aristotle was a philosopher, but also about the truth value of simple sentences, such as Aristotle was a philosopher, when evaluated at alternative worlds. This not only added depth and plausibility to the discussion but also rendered the modal argument immune to certain types of objection. 33 Smith's fourth contention has to do with the necessity of identity. He says: This modal argument goes back to Marcus' formal proof of the necessity of identity in her extension of S4 (Barcan, 1946; 1947), which is a fourth component she introduced into the New Theory of Reference. She showed that (T)(x = y)

~

necessarily x = y)

is a theorem of QS4, QS4 being her quantificational extension of Lewis' S4 .... Since identities are necessary, a failure of intersubstitutivity in modal contexts will show that a proper name does not express the relevant descriptive sense.

There are two problems here. First, the priorities are wrong. We do not accept the modal argument because we know antecedently that identities involving ordinary coreferential names are necessary. Rather, we recognize these identities as necessary because the modal argument convinces us that ordinary names are rigid designators. 34 Second, the version of the necessity of identity proven by Marcus involved only variables, and had nothing to do with proper names. Here it is crucial to distinguish the thesis (NI-l), derived by Marcus in her 1947 paper, from the theses (NI-2) and (NI-3). (NI-l)

(x)(y)(if x = y, then necessarily x = y)

(NI-2)

for all proper names a and b, if a = b is true then necessarily a = b is true.

(NI-3)

for all singular terms a and b (including singular definite descriptions), if a = b is true then necessarily a = b is true.

Theses (NI-l) and (NI-2) are true; whereas thesis (NI-3) is false, at least in formal languages in which the class of singular terms includes Fregean definite descriptions. What must be noticed is that the quantificational version (NI-I) of the necessity of identity, articulated and derived by Marcus, no more provides a direct route to the true thesis (NI-2) involving ordinary proper names than it provides a direct route to the false thesis (NI-3). It simply leaves

REVISIONISM ABOUT REFERENCE

23

the status of these further theses open. For this reason it is just plain wrong to cite Marcus' early systems of quantified modal logic as introducing the version - (NI-2) - of the necessity of identity that corresponds to the modal argument. I should mention that, historically, there was a strong tendency, shared not only by those who accepted NI-l, but also those who rejected it, as well as those, like Quine, who rejected quantifying-in entirely, to think that ifNI-l was true, then NI-2 and 3 must also be true. I suspect that one important reason for this was the equally widely shared, and equally mistaken conviction that the intelligibility of objectual quantification depended upon universal instantiation and existential generalization being universally truth-preserving, where genuine singular terms are involved. In the hands of Marcus, Smullyan and Fitch, this led to the view that the necessity of identity not only required coreferential names to be substitutable in modal contexts, but also prevented us from taking definite descriptions to be singular terms. In the hands of their opponents, who were more inclined to accept descriptions as singular terms, it was taken as refuting the necessity of identity. What is interesting here is the false presupposition shared by both sides of the debate, tying the intelligibility of quantification to familiar extensional rules of inference. It seems plausible that what ultimately lay behind this error was an overly proof-theoretic perspective, and insufficiently developed model-theoretic conception of intensional logic according to which it is the semantics of various intensional constructions, plus the usual Tarski-like clauses for quantifiers, that have priority and determine which rules of inference are valid. Now that we have achieved the appropriate semantic perspective, there is no longer any excuse for running NI -1, 2, and 3 together, or for thinking that the truth of one establishes the truth of the others. As for the thesis, (NI-I), involving variables, I want to acknowledge that there is a way of viewing it as a genuine contribution by Marcus to contemporary theories of reference. One can think of it as the material-mode counterpart of the semantic thesis that individual variables are rigid designators. Although Marcus herself did not explicitly characterize variables in this way, and many years later suggested a more indirect way of understanding quantification via substitution of proper names, her original thesis (NI-l) can be interpreted as providing the basis for characterizing variables as rigid designators. 35 This brings us to Smith's contention that a fifth contribution of Marcus to the so-called New Theory of Reference was the concept of rigid designation. I have just indicated that there is a sense in which Marcus can be viewed as providing a material-mode basis for taking variables to be rigid (relative, of course, to assignments of values). Moreover, she can be seen, following Smullyan and Fitch, as implicitly treating proper names as rigid. However none of these philosophers had the concept of a rigid designation. This concept presupposes the more general notion of the referent of a term at a world. Neither Smullyan, Fitch, nor Marcus in her papers in the 1940's, provided any semantic framework for modal systems, let alone one involving this notion.

24

scon SOAMES

Although an elementary semantic formalization is given by Marcus in her 1961 article, it did not invoke the notion of a referent of a singular term at a world, and did not make room for the full-blown concept of rigid designation as we now understand it. Rather, this semantic notion, though anticipated by Marcus, seems to be due to Kripke, and was present in his treatment of variables as rigid in his 1959, 'A Completeness Theorem in Modal Logic,.36 Smith's characterization of Kripke as contributing only the name 'rigid designation' for a concept discovered by Marcus is a grotesquely inaccurate caricature. Marcus did, of course, explicitly use the notion of rigid designation in her paper 'Essential Attribution', delivered at the December 1970 meetings of the APA. 37 Smith cites this paper, saying, "Marcus notes in her 1970 APA paper, 'Essential Attribution; presented at a symposium at which Kripke was one of the symposiasts, that "individual names don't alter their reference, except to the extent that in [respect ofl some worlds they may not refer at alf'.,,38 Why does Smith

feel that it is important to mention that Kripke was present at this symposium? Is he suggesting that Kripke's presence at the symposium supports the contention that he got the concept of rigid designation from Marcus? If so, why doesn't Smith also mention that the symposium occurred 11 months after Kripke had presented the three lectures of Naming and Necessity at Princeton University? Smith's final contention about the extent of Marcus' contributions to nonFregean theories of reference involves a pair of epistemological points he derives from the following one and one half sentence quote from Marcus' 1961 paper. You may describe Venus as the evening star, and I may describe Venus as the morning star, and we may both be surprised that, as an empirical fact, the same thing is being described. But it is not an empirical fact that (17) Venus I Venus. 39

Smith derives two conclusions from this brief passage. His first conclusion is that Marcus introduced the idea of a posteriori necessity into contemporary philosophy of language, since on her view the sentence Hesperus I Phosphorus is necessary, even though it is clearly a posteriori. Smith's second conclusion from the passage is that in it Marcus provided an epistemological argument against description theories of names to supplement the modal argument. Both of these claims are false. The first thing to do in showing this is to obtain an accurate quotation of the passage from Marcus. In presenting the passage, Smith breaks Marcus' final sentence in half, inserts a period immediately after her example sentence Venus I Venus, and omits, without any indication of ellipsis, the final half of her sentence. Here is the passage as it appears in Marcus.

REVISIONISM ABOUT REFERENCE

25

You may describe Venus as the evening star, and I may describe Venus as the morning star, and we may both be surprised that, as an empirical fact, the same thing is being described. But it is not an empirical fact that (17) Venus I Venus and if 'a' is another proper name for Venus that (18) Venus I a 40 [my emphasis]

What is Marcus saying here? She is saying that if a is any name for Venus, neither the sentence Venus = Venus nor the sentence Venus = a expresses "an empirical fact". If, as Smith maintains, Marcus is trying to make an epistemological point here, then by "an empirical fact" she must not mean "a contingent fact", for then there would be no epistemological point at issue. Rather, she must be taken to mean that these sentences do not express facts that are knowable only by empirical means. In other words they are knowable a priori. But if this is right, then true identity statements involving names are not characterized as examples of the necessary a posteriori, and Smith's contention that Marcus introduced this notion collapses. It should be noted that Marcus nowhere in the article says that there are true identity statements which are knowable only a posteriori. On the contrary, she says many things that strongly indicate the opposite. One remark that I cited earlier occurs in the paragraph immediately after the one just cited by Smith. There Marcus says" What I have been arguing is that to say truly of an identity (in the strongest sense of the word) that it is true, it must be tautologically true or analytically true".41 Thus, according to Marcus' true identity statements involving genuine names never express empirical facts, but rather are tautologically, or analytically true. Surely this is inconsistent with interpreting her as introducing the notion of necessary a posteriori identities. No author who was doing that would completely fail to use the notion a posteriori, while treating necessity, tautology, analyticity, and the notion of not expressing an empirical fact as if they were interchangeable. This point is reinforced by other things that Marcus says. For example, two pages before the passage cited by Smith, she says the following about the identity sentence alb (which is example 13 in her paper). Now if (13) is such a true identity, then a and b are the same thing. (13) doesn't say that a and bare two things that happen, through some accident, to be one. True, we are using two different names for that same thing, but we must be careful about use and mention. If, then, (13) is true, it must say the same thing as (14) ala. 42 [my emphasis]

The point to note here is that if a = a and a = b say the very same thing, then surely they express the same fact, in which case the fact expressed by one must be knowable a priori only if the fact expressed by the other is. Hence, on the view articulated by Marcus, true identity statements involving names are knowable a priori.

26

SCOTT SOAMES

Finally, you may recall that according to Marcus, coreferential proper names "must be intersubstitutable in every context".43 Thus, if prior to the relevant astronomical discovery Jones knew that Hesperus was identical with Hesperus, on the basis of reflection alone, then he also knew that Hesperus was identical with Phosphorus, by reflection alone - assuming that 'Hesperus' and 'Phosphorus' really are names in Marcus' sense. For all these reasons, it is as clear as anything can be that Marcus did not in her 1961 article embrace the notion of necessary a posteriori identities. What about Smith's contention that Marcus provided an epistemological argument against description theories of names to supplement the modal argument, and to rule out the possibility of modally rigid descriptive senses? Recall the passage cited. You may describe Venus as the evening star. and I may describe Venus as the morning star, and we may both be surprised that, as an empirical fact, the same thing is being described. But it is not an empirical fact that (I 7) Venus I Venus.

Smith claims to find in this passage an argument that where D is a description associated with a name N, the proposition expressed by N is D is not knowable a priori, contrary to the description theory. But there is no such argument. Moreover, once again what Smith says about the passage bears little relation to what is actually in it. Here is what Smith says: If 'Venus' expresses the modally stable sense expressible by 'whatever is actually the evening star and morning star', then the persons designated by 'you' and 'I' in the passage quoted from Marcus' article should be able to know a priori, simply by reflection upon the semantic content of the expressions 'Venus', 'the morning star' and 'the evening star' that Venus is both the morning star and the evening star. The fact that they cannot know this indicates that 'Venus' does not express the modally stable sense expressed by 'whatever is actually the evening star and morning star'.

Notice that in contrast to Marcus' passage, Smith reconstructs the example as one in which you and I both associate with 'Venus' the same description - the x: actually x is the morning star and x is the evening star. If we really did this, then, provided that we thought that we were successfully describing something, we certainly would not be surprised to find out that the morning star is the evening star (as Marcus says we are). Consequently, Smith's argument is his own invention; it is not in Marcus. In addition, the argument is no good. It assumes that if an expression r:t. means the same as the description the x: actually x is the morning star and x is the evening star, then, the claim expressed by r:t. = the x: actually x is a morning star and x is an evening star must be knowable a priori. But this is false, since that claim entails that there is something that is both the morning star and the evening star, it cannot be known a priori. Because of this the fact that it is not knowable a priori that Venus is both the morning star and the evening star does not establish that 'Venus' does not have the descriptive

REVISIONISM ABOUT REFERENCE

27

meaning under consideration. As a result, Smith's argument fails. 44 There is, however, a line of argument that can be constructed from elements in Marcus for the conclusion that neither the proposition that Hesperus is the evening star nor the proposition that Phosphorus is the morning star is knowable a priori. Ironically, however, the argument does no good for Smith, and cannot be used as an argument against descriptive senses. Recall the last line, omitted by Smith, of the passage cited from Marcus. "But it is not an empirical fact that Venus I Venus and if 'a' is another proper namefor Venus that Venus I a". If, as I have argued, Marcus is here expressing the view that identities between coreferential proper names do not express empirical facts, and so do not require empirical investigation to be known, then we can construct the following argument, which does not appear in her paper, on her behalf:

= Phosphorus.

(PI)

It is knowable a priori that Hesperus

(P2)

If it were also knowable a priori both that Hesperus is the (actual) evening star (if such exists) and that Phosphorus is the (actual) morning star (if such exists), then it would be knowable a priori that the (actual) evening star is the (actual) morning star (if there are such things).

(P3)

It is not knowable a priori that the (actual) evening star is the (actual)

morning star (if there are such things). (CI)

So it is not the case that it is knowable a priori both that Hesperus is the (actual) evening star (if such exists) and that Phosphorus is the (actual) morning star (if such exists).

(P4)

If one of these were knowable a priori, then the other would also be knowable a priori.

(C2)

Therefore neither is knowable a priori.

The first premise in this argument is, of course, incompatible with Smith's claim that for Marcus, the proposition that Hesperus is Phosphorus is an example of the necessary a posteriori. Thus, the argument is no good to him. However, given Marcus' treatment of proper names as Russellian logically proper names, and hence as supporting substitution without change of meaning in any non-quotational context, (PI) is plausible, as are the other steps in the argument. 45 Moreover, the conclusion is clearly incompatible with the description theorist's claim that the name 'Hesperus' has the same sense as 'the (actual) evening star', or any other plausible description that speakers might associate with the name. Why then isn't this precisely the epistemological argument against the

28

SCOTT SOAMES

description theory that we are looking for? The reason it isn't is that its first premise, namely that it is knowable a priori that Hesperus is Phosphorus, presupposes a prior rejection of the description theory. The basis for assuming this premise is simply a prior acceptance of Millianism. Given this, we can use the premise to draw out an epistemological consequence of Marcus' position, but we cannot use it in an argument against descriptivism. To get an epistemological argument against descriptivism one would have to argue directly, without appeal to any Millian premises, that one cannot know a priori that N is D, for relevant names N and descriptions D. Marcus does not do this. Thus, there is no epistemological argument in her paper which can be used to supplement the modal argument against the description theory. This concludes my critique of Smith's major contentions. There are a few other matters I would go into if there were time. 46 But since there is not, I will try to sum up. On the positive side, we can see in the writings of Marcus, Fitch and Smullyan, significant anticipations of some of the central theses of contemporary non-Fregean theories of reference. First, Marcus' quantified version of the necessity of identity anticipates the contemporary view that variables are rigid designators, with respect to assignments of values. Second, the claim, by Smullyan, Fitch, and Marcus, that substitution of coreferential names in modal constructions preserves truth value anticipates the view that ordinary proper names are rigid designators. Third, the use of this feature of names to discriminate them from some descriptions anticipates certain aspects of Kripke's modal argument against description theories. However, it is also important to notice how much of the development of contemporary theories of reference by Kripke and others is lacking from the earlier discussions of Marcus, Smullyan, and Fitch. First, we do not find Kripke's important distinction between the different ways that description theories can be understood - as theories of meaning vs theories of reference fixing. Second, apart from the modal argument, we do not find analogs to the battery of epistemological and semantic arguments against description theories that are given in Naming and Necessity. Third, there is no positive conception of how reference is determined - either by historical chain or by using descriptions to semantically fix reference. Fourth, there are no analogs of Kripke's doctrines of the necessary a posteriori and the contingent a priori. Fifth, the discussions of names are not set in the context of a wider theory of language, including, for example, natural kind terms. Sixth, there is no recognition that arguments stronger than those needed to establish rigidity are needed to support the claim that names are directly referentia1. 47 As a result, relatively uncontroversial claims, like the substitutability of coreferential names in modal constructions, were linked with highly controversial claims, like the substitutability of such names in all constructions, and thereby were rendered less persuaSIve. Don't get me wrong. I am not criticizing Marcus, Smullyan or Fitch for not having written Naming and Necessity. On the contrary, they should be praised

REVISIONISM ABOUT REFERENCE

29

for their prescient insight, despite the fact that they were not attempting anything so systematic or far-reaching. What I have tried to show is that providing them with proper credit does not result in a reassessment of the seminal role of Kripke and others as primary founders of contemporary nondescriptivist theories of reference. In fact, I think Smith does Kripke a grave injustice. Smith writes as if Kripke appropriated the major views expressed in Naming and Necessity from Marcus while denying her proper credit, and suggests that it is a scandal that the rest of the profession was thereby duped. We have not been duped; there was no misappropriation. Rather, what Smith has done is to mistakenly read many of Kripke's arguments and doctrines back into Marcus, and then to insinuate that Kripke is guilty of theft. This, it seems to me, is shameful. If there is any scandal here it is that such a carelessly and incompetently made accusation should have been given such credence. You may recall that, in hyping his accusations, Smith claimed to be correcting a misunderstanding in the recent history of philosophy "no less important than correcting the misunderstanding in a hypothetical situation where virtually all philosophers attributed the origin of the Theory ofForms to Plotinus". I hope that by now it is apparent how wildly exaggerated this remark really is. I also hope no one will be distracted by Smith's overheated rhetoric and irresponsible sowing of discord from the truly outstanding contributions of both Marcus and Kripke. Both have done work of great importance that should be appreciated and celebrated. They were never competitors in the past, and there is no need now to tear one down to elevate the other. 48 Department of Philosophy Princeton University

NOTES 1 Synthese, XIII, Dec. 1961,303-22. This paper is reprinted in Ruth Barcan Marcus, Modalities, (New York and Oxford: Oxford University Press), 1993. Citations from the paper will be from Modalities. 2 R.C. Barcan (Marcus), 'A Functional Calculus of First Order Based on Strict Implication', Journal of Symbolic Logic, 11, 1946, 1-16. Also, R.C. Barcan (Marcus), 'The Deduction Theorem in a Functional Calculus of First Order Based on Strict Implication', The Journal of Symbolic Logic, 11,1946,115-8. 3 R.C. Barcan (Marcus), 'The Identity of Individuals in a Strict Functional Calculus of Second Order', The Journal of Symbolic Logic, 12, 1947, 12-5. 4 In 'Modalities and Intensional Languages', pages 9 and 10, Marcus refers to theorem 2.32* -0(13, 1132) == (13, 1132)- of her 1947 paper as, in effect, establishing the necessity of identity. (The symbol 'I' is the (strict) identity predicate. The connective is strict equivalence, the terms are individual variables, and the theorem itself has the force of its universal closure.) Marcus derives this theorem in her quantified version ofS4 from her theorems 2.23 -0(13, 1m 132) == (13, 1132)- and 1.104* -OOA == OA. (The symbol '1m' stands for the material identity predicate. Material identity and strict identity are defined predicates that turn out to be provably strictly equivalent in Marcus' S4.) The point to note here is that Marcus' derivation of the necessity of identity depends on

30

SCOTT SOAMES

a controversial principle about the iterability of the necessity operator, which is not something that the thesis itself intuitively requires. (The proof also utilizes the Barcan and converse Barcan formulas. See below.) If variables are treated as rigid designators - with respect to assignments as in the now standard Kripke-style semantics for modal systems, the above version of the necessity of identity falls out automatically from the semantics without any special assumptions about the accessibility relation, or the corresponding axiomatic modal systems. In this connection, it is worth noting another theorem that Marcus refers to in 'Modalities and Intensional Languages' in her discussion of the necessity of identity. This is theorem 2.33* of her 1947 paper -(131 1m (32) == (131 I (32). This establishes the strict equivalence of Marcus' two forms of identity. (Theorem 2.32*, and hence the S4 premise guaranteeing the iterability of the necessity operator, is used in the proof of this.) The theorem itself is closely related to the necessity of identity. This can be seen as follows: First, we use the definition of 'I' (strict identity) to give the import of the theorem as (i) (i)

x 1m y

== (F)(Fx3 Fy)

Next we apply the definition of strict implication to give us (ii). (ii)

x 1m y

== (F)O(Fx:J Fy)

The combination of (ii) together with the second order Barcan formula - <):J¢A second order converse Barcan formula -:J¢<)A 3 <):J¢A - yields (iii). (iii)

x 1m y

3 :J¢<)A -

and the

== O(F)(Fx :J Fy)

Finally, we apply the definition of '1m' (material identity) to get (iv). (iv)

x 1m y

== Ox 1m y

Although this might be taken to be a version of the necessity of identity, as before the proof relies on controversial modal principles - including the second order Barcan and converse Barcan formulas. These principles are not intuitively required by the necessity of identity itself, and are dispensable if variables are taken as rigid designators in the semantics of all the modal systems. (I am indebted to Saul Kripke for a discussion of the material in this paragraph, especially the role of the second order Barcan and converse Barcan formulas.) The second order Barcan and converse Barcan formulas also seem to be involved in the proof of theorem 2.23 - 0(131 1m (32) == (131 1(32) - which Marcus uses in her proofs of both theorem 2.32* and 2.33*. Note, the definitions of material identity and strict identity are (F)(FX:J Fy) and (F)O(Fx :J Fy) respectively. Given the two second order Barcan formulas, the formulas (F) (Fx :J Fy) and (F) 0 (Fx :J Fy) are strictly equivalent, which establishes 2.23. Finally, it may be worth pointing out why the necessity of identity cannot be derived in the now familiar way in Marcus' systems without the help of some controversial supplementary principles like those we have been appealing to. We may begin by attempting a derivation from theorem 2.4 of her quantified extension of S2'

o

(i)

x Iy

3

(F)(Fx

3

(ii)

x Iy

3

(AZ(OX I z)x

(iii)

x I y 3 (Ox I x

(iv)

x I y :J (Ox I x :J Ox I y)

3

Fy) (Theorem 2.4 of Barcan 1947)

3

AZ(OX I z)y) (From (i) and 2.2 of Barcan 1947)

Ox I y) (From (ii) and 2.3 of Barcan 1947

Now all we need to establish is Ox I x, and we will have the desired result. But establishing this is problematic. Suppose we proceed as follows:

3

(v)

Fx

(vi)

(F) (Fx

Fx Axiom Schemata 2, 3, and 5 of the first Barcan 1946 paper)

(vii)

x I x (From (vi) by the definition of 'I')

3

Fx) (From (v) by universal generalization)

If S2 had necessitation we could derive O(x)(x I x). But S2 does not have necessitation. Moreover, even if we could get 0 (x) (x Ix), we would have to appeal to the converse Barcan formula to get the needed Ox I x, which has the force of (x)Ox I x.

REVISIONISM ABOUT REFERENCE

31

In light of this we might try a different derivation. This time we start with Marcus' theorem 2.17. x 1m y

(Az(Oxlmz)x

(iii)

3 3 x 1m y 3

(F)(Fx

x 1m y

(iv)

x 1m y

(Ox 1m X =:J Ox 1m y)

(v)

Fx

(i) (ii)

3

=:J

=:J

Fy) (Theorem 2.17 of Barcan 1947) =:J

AZ(OX 1m z)y) (From (i) and 2.2 of Barcan 1947)

Ox 1m X =:J Ox 1m y) (From (ii) and 2.3 of Barcan 1947)

Fx (Axiom Schemata 2, 3, and 5 of the first Barcan 1946 paper)

(vi)

O(Fx

(vii)

(F)O(Fx

=:J

Fx) (From (vi) by universal generalization)

(viii)

O(F)(Fx

=:J

Fx) (From (vii) plus the second order Barcan formula)

(xi)

Ox 1m X(From (viii) plus the definition of' 1m ')

(x)

x 1m y

=:J

=:J

Fx)«v) plus the definition of strict implication)

Ox 1m y (From (iv) and (ix»

As far as I can tell, this is the closest one can get to an unproblematic derivation of a version of the necessity of identity in Marcus' system. Although the derivation can be done in Marcus' quantified version of S2, it uses "material identity" rather than "strict identity", and, most importantly, it requires the second order Barcan formula. 5 Arthur Smullyan, review of w.v. Quine, "The problem of interpreting modal logic". Quine's paper appeared in The Journal of Symbolic Logic, 12, 1947,43-8. Smullyan's review appeared in The Journal of Symbolic Logic, 12, 1947, 139-41. 6 Arthur Smullyan, 'Modality and Description', The Journal of Symbolic Logic, 13,1948,31-7. 7 Frederic B. Fitch, 'The Problem of the Morning Star and the Evening Star', Philosophy of Science, 16,1949,137-41. 8 The Journal of Philosophy, Vol. XL, 1943, 113-27. The Journal of Symbolic Logic, Vol. 12, 1947,43-8. 10 In the 1947 article Quine gave a slightly different version of the argument, one which emphasized what he took to be the intimate connection between the logical content of objectual quantification and certain rules of inference - in particular existential generalization. We can reconstruct the basic idea as follows. (PI)a An occurrence of an objectual variable within a certain construction is bindable by a quantifier outside the construction only if existential generalization from that position is universally truth-preserving. (P2)a Existential generalization from the position of 'x' in a formula, (... x...), will be universally truth-preserving only if 'x' occupies a position which is transparent - only if for any pair a and (3 of coreferential singular terms the sentence, ( ... a ...), that results from substituting a for, 'x', is true iff the sentence, (... (3 ...), that results from substituting (3 for 'x', is true. (For suppose that this were not so. Suppose that for some pair a and (3 of coreferential singular terms the sentence, ( ... a ... ), is true whereas the sentence, (... (3 ...), is false. In such a situation, both the sentence [a = (3 & ( ... a .. .)] and the sentence [«(3 = (3 & ~( ... (3 ...)] are true. But then if existential generalization is universally truth preseving, 3x[x = (3 & ( ... x...)] and 3x[x = (3 & ~( ... x... )] must both be true, which is impossible.) Combining these premises with (P3), plus the definitions of opacity and pure designation, produces the same results as above. II 'Opacity', The Philosophy of WV Quine, Hahn and Schilpp (eds.), (LaSalle Illinois, Open Court), 1986,229-289. 12 'Quantification and Opacity', Linguistics and Philosophy, 10, 1987, 77-100. See also, his 'Some Remarks on Indiscernibility', Canadian Journal of Philosophy, Supplementary volume 18, 1992, 167-78. 13 'Quantification and Leibniz's Law', Philosophical Review, vol. 96, 1987, 555-78. 14 The same holds for premise PIa of the argument in footnote 10. 15 And do not support universally truth-preserving existential generalization. 16 And do support existential generalization.

32

SCOTT SOAMES

And do not support existential generalization. Richard, 'Quantification and Leibniz's Law'; Kazmi, 'Some Remarks on Indiscernibility'. 19 In addition, Smullyan explicitly maintained that existential generalization on a formula containing a modal operator is universally truth-preserving where names are involved, but fails to be so when the existentially generalized variable replaces a description rather than a name. 20 Alonzo Church, review of Frederic B. Fitch, 'The problem of the Morning Star and the Evening Star', The Journal of Symbolic Logic, 15, 1950, p. 63. 21 Point (vi) is explicit only in Marcus. 22 Page 12, in Modalities. 23 Seepages lOand 14-15. 24 'Discussion of the Paper of Ruth Marcus', Synthese, 14,1962,132-43. Reprinted in Modalities. 25 Pages 33 and 34 of Modalities. 26 It is significant that Marcus responds to a straightforward question about the necessity of identities involving names (and their intersubstitutability in modal constructions) with remarks about epistemology and synonymy. This reinforces the impression given by the text of her article that she thought of the view of names as mere tags (and hence as being synonymous if coreferential) as being inseparable from, and even the ground for, her views of the behavior of names in modal constructions. This linking of the modal status of names with controversial claims about their meaning and epistemology marks a very important contrast between Marcus' view and the view developed by K ripke in Naming and Necessity. 27 Marcus' discussion of proper names occurs in the section of her paper devoted to answering Quine's objections to quantifying-in. She begins her discussion by saying "The rebuttals are familiar and I will try to restate some of them". [Page lO in Modalities]. Moreover, in a very closely related paper, 'Extensionality', Mind, Vol. 59, 1960, pp. 55-62, Marcus briefly sketches a response to Quine's Morning Star-Evening Star example which is largely the same as the one she gives in 'Modalities and Intensional Languages'. The discussion in 'Extensionality' is prefaced by a footnote in which Marcus says that she is restating a point made by Fitch in his 1950 paper. 28 Moreover, Smith's quote from Marcus on this point contains massive ellipsis, and gives a misleading picture of her discussion. Here is his quote. 17

18

But to give a thing a proper name is differentfrom giving a unique description ... [An] identifying tag is a proper name of the thing ... This tag, a proper name, has no meaning. It simply tags. It is not strongly equatahle with any of the singular descriptions of the thing.

Here is the beginning of the passage quoted, without Smith's first ellipsis. But to assign a thing a proper name is different from giving a thing a unique description. For suppose we took an inventory of all the entities countenanced as things by some particular culture through its own language, with its own set of names and equatable singular descriptions, and suppose that number were finite (this assumption is for the sake of simplifying the exposition). And suppose we randomized as many whole numbers as we needed for a one-to-one correspondence, and thereby tagged each thing. This identifying tag is a proper name of the thing. [my emphasis] (p. II of Modalities)

Marcus says that this identifying tag - namely the number assigned to the object in our artificial, randomized assignment - is a proper name. Clearly it is not an example of an ordinary proper name. Rather, Marcus seems here to be illustrating the theoretical possibility that we, as theorists, might introduce certain elements as mere tags. The passage continues as follows: In taking our inventory we discovered that many of the entities countenanced as things hy that language-culture complex already had proper names, although in many cases a singular description may have been used. This tag, a proper name, has no meaning. It simply tags. It is not strongly equatable with any afthe singular descriptions of the thing ... (pp. II and 12 of Modalities)

What is Marcus referring to in this passage when she says "This tag, a proper name, has no meaning"? Is it to the numbers of our randomized assignment, which have just been called both tags and proper names two sentences back? Is it to proper names in the language of the culture we are describing? Or is it to both? My guess is that the reference is probably to both. But if it is to both, we are not given an argument for taking the two to be on a par.

REVISIONISM ABOUT REFERENCE

33

Page 11 in Modalities. See note 28. 31 I do not object here to the substance of Marcus' position, which is close to my own. Rather, I am merely pointing out one reason why the modal argument seemed more persuasive, and was more widely accepted, when presented by Kripke. 32 See Naming and Necessity, Harvard University Press, Cambridge, Mass., 1980, pp. 61-2. 33 Such as Dummett's objection that proper names can be taken to be equivalent to descriptions that always take wide scope over modal operators. See Dummett, Frege: Philosophy a/Language, Harper and Row, New York, 1973, pp. 112-116. 34 This is an important feature of Kripke's version of the modal argument that does not seem present in Marcus' version. Both Kripke and Marcus use modal properties of names to distinguish them from descriptions. However, the way they establish the modal properties of names is quite different. Kripke provides examples in which we evaluate simple sentences containing proper names (but no modal operators) at alternative possible worlds (counter factual circumstances). He uses these examples to show that the referent of a name at one world is the same as its referent at other worlds, and hence that names are rigid. It then follows that identities involving names are necessary if true, and that coreferential names are intersubstitutable in modal constructions. Marcus, on the other hand, seems to think that these modal properties of names somehow follow from the quantified version of the necessity of identity, either by itself or in conjunction with her controversial view that names are "mere tags". 35 However, this contribution must be understood as subject to the qualification in footnote 4. An explicit semantic system in which variables are treated as rigid designators is given in Saul Kripke, 'A Completeness Theorem in Modal Logic', The Journal of Symbolic Logic, Vol. 24, Number 1, 1959.1-14. (The paper was received by the journal in August of 1958.) 36 Ibid. 37 Journal of Philosophy, LXVIII, 1971, 187-202. This paper is reprinted in Modalities. 38 The passage quoted by Smith is taken from page 61 of Modalities. Two points are of interest here. First, at this time Marcus had no objection to characterizing proper names as rigid designators, even though she presumably did not think that in doing so she was "assimilating them to some descriptions". Second, in the passage from which this quote was taken, Marcus continues to express some ambivalence regarding the question of the extent to which names, as ordinarily used, are rigid designators. Thus, she ends the paragraph from which the quote was taken by saying, "For those that are quick to argue that ordinary names cannot a/ways be used in such a purely referential way, we can, in giving the interpretation. expand our lexicon to provide neutral, purely referential names where necessary". (p. 61 Marcus' emphasis) ]9 Page 12, Modalities. 40 Page 12, Modalities. 41 Page 12, Modalities. 42 Page 10, Modalities. The passage continues in a somewhat confusing way, with Marcus saying that therefore since (14) is "valid" (13) must be as well. She then adds that this was "precisely the import oImy theorem [of the necessity of identity]". Put aside that her theorem didn't involve names at all. Surely we would now see her claims about (13) and (14) saying the same thing, and being jointly valid, as going well beyond the claim that they are necessarily equivalent. Of course, if contrary to Smith's thesis - Marcus was not carefully distinguishing necessity from logical validity, analyticity, and aprioricity, then her tendency in the article to slip from one to another is quite understandable. 4] Page 10 of Modalities. 44 Smith would have avoided this particular problem, while keeping somewhat closer to the passage in Marcus, had he attempted to construct an argument ruling out the possibility that some name a for Venus has, for some particular person, the sense of 'the (actual) morning star' while some name (3 has, for some person, the sense of 'the (actual) evening star'. For example, consider the following: 29

30

(PI)

If the description theory is correct, and I semantically associate the name 'Venus' with the description 'the (actual) evening star', then I can know a priori the proposition I express by the sentence 'Venus is the (actual) evening star, if there is such a thing'.

(P2)

If the description theory is correct. and you semantically associate the name 'Venus' with the description 'the (actual) morning star', then you can know a priori the proposition you

34

SCOTT SOAMES

express by the sentence 'Venus is the (actual) morning star, if there is such a thing'. (CI)

Thus, if the description theory is correct, then both the proposition that Venus is the (actual) evening star, if there is such a thing and the proposition that Venus is the (actual) morning star, if there is such a thing must be knowable a priori. Hence, if the description theory is correct, we can know a priori that the (actual) evening star is the (actual) morning star, if there are such things.

(P3)

But one cannot know this a priori.

(C2)

Therefore, the description theory is incorrect.

Still, this argument is invalid. Conclusion (CI) does not follow from (PI) and (P2); it is simply false. According to the description theory if you and 1 semantically associate different descriptions with the name 'Venus' then the name has different senses for us. Because of this the proposition 1 express with the sentence Venus is the (actual) evening star is not the proposition you express by that sentence. Consequently, it would be correct for me to report, "I know a priori that Venus is the (actual) evening star, if there is such a thing, but I cannot know a priori that Venus is the (actual) morning star, if there is such a thing" while it would be correct for you to report the opposite. According to the description theory, there is no single interpretation of Venus is the (actual) evening star, if there is such a thing and Venus is the (actual) morning star, if there is such a thing according to which both are knowable a priori. Of course, the failure of this argument is not Marcus' responsibility; for there is no such argument in her paper. 45 P 2 could be questioned on the grounds that it assumes some sort of closure principle for propositions knowable apriori - e.g., if A, B, and C are knowable apriori, and D is a consequence of A, B, C, then D is knowable apriori. 1 am not endorsing such a closure principle, or even P 2. Rather, 1 am trying to make the strongest argument one can, based on Marcus' text, for the conclusion that the propositions that Hesperus is the evening star and Phosphorus is the morning star are not knowable apriori. 46 A case in point is Smith's uncomprehending criticism of the following remark of the undergraduate Saul Kripke in the 1962 'Discussion of the Paper of Ruth Marcus'. The tags are the 'essential' denoting phrases for individuals, but empirical descriptions are not and thus we look to statements containing 'tags', not descriptions, to ascertain the essential properties of individuals. Thus, the assumption of a distinction between 'names' and 'descriptions' is equivalent to essentialism. [my emphasis] (page 34, Modalities) Smith follows Marcus, 'A Backward Look at Quine's Animadversions on Modalities', reprinted in Modalities, 1993,226-227, in seeing Kripke's remark as attributing to Marcus the absurd view that it is essential property of an individual like Socrates that he was named 'Socrates' - i.e., the view that Socrates could not have existed without being named 'Socrates'. Smith then takes this as evidence that Kripke misunderstood Marcus. But it is Smith (and Marcus) who have misunderstood. Kripke was not criticizing Marcus for holding the view that individuals could not have had different names. Rather, Kripke was showing how the doctrine that there is a class of singular terms which are rigid designators collides with Quine's doctrine that even if de dicto necessity (in which the necessity operator is prefixed to a closed sentence) is accepted, de re necessity (in which we quantify into a modal context formed by prefixing the necessity operator to a formula containing a free variable) is illegitimate. Quine's idea was that even if we could make sense of a statement being necessary, we cannot make sense of an object having a property essentially, or necessarily (which is what we express by quantifying in). According to Quine we cannot make sense of an object necessarily having a certain property independently of how it is designated or described because for any such object 0 and property P there will be different (closed, singular) terms tl and t2 referring to 0 such that the sentences Necessarily t] is P and Necessarily t2 is P differ in truth value. Kripke's response (on behalf of Marcus) can be put in contemporary terms as follows: If names are rigid designators, and n is a name of 0, then the de re claim that 0 has the property P necessarily -(3x [x = n & DPx])- is equivalent to the de dicto claim that the statement Pn is necessary -(DPn). Since n refers to the same thing in all worlds, the de dicto claim is true iff 0 has the property P in every world - which is just what the de re claim says. Thus, if one grants the legitimacy of de dicto claims, and also recognizes the rigidity of names, one has to grant the legitimacy of de re,

REVISIONISM ABOUT REFERENCE

35

essentialist claims as well. Alternatively, if one insists (for whatever reason) that de re essentialist claims do not make sense, then one must insist that neither names, nor any other terms, can be rigid designators. In the discussion Kripke did not take a stand on which of these alternatives was correct, but was content to simply frame the issue between Marcus and Quine. (See page 35, Modalities.) 47 For a discussion of the difference between rigidity and direct reference see David Kaplan, 'Afterthoughts', Themesfrom Kaplan, Almog, Perry, and Wettstein, eds. (University Press Oxford, New York and Oxford), 1989,565-614. 48 Versions of this paper were read by Paul Benacerraf, Sarah Broadie, Fiona Cowie, AI Akhtar Kazmi, Mark Richard, Nathan Salmon, and Michael Thau. I would like to thank them for their helpful comments. In addition, I would like to thank Saul Kripke for a conversation about the material indicated in footnote 4, and Ruth Marcus for conversations about her views in 'Modalities and Intensional Languages'.

QUENTIN SMITH

MARCUS AND THE NEW THEORY OF REFERENCE: A REPLY TO SCOTT SOAMES

We are engaged in an inquiry in the history of philosophy. Specifically, we are inquiring into the historical origins of the New Theory of Reference. Relevant to this inquiry are arguments about the meaning of texts by Marcus, Kripke and others. I do not believe it is relevant or helpful to adopt the sort oflanguage that Soames uses in his reply, when he sees fit to call my work "careless, incompetent, grotesquely inaccurate, shameful, scandalous, and irresponsible". Philosophical disagreements are not solved by the disputants labelling each other's work with a variety of negative and emotive epithets; they are solved by presenting sound arguments, and I shall confine myself to presenting arguments in my reply to Soames. In my original paper I discussed seven ideas that Marcus introduced into the New Theory of Reference. Soames took issue with each of these seven claims, and I shall examine each of his seven criticisms in turn. (In my original paper, I talked of six ideas, since I included the modal and epistemic arguments for direct reference under one heading, but here I distinguish them for the sake of increased clarity.) But first I should note that Soames does not quote Marcus' original 1961 paper, which is the paper in dispute, but her revision of this paper, published in Modalities (1993), which differs at a number of places from the original 1961 article. I quoted only the 1961 article, but Soames, when he sometimes tries to show that my quotations are misleading or incomplete, does not in fact quote the 1961 article, but Marcus' 1993 revision of this article. Since this is an historical inquiry into the relation of Marcus' 1961 paper to Kripke's works of 1971 and 1972, it is her 1961 paper that is the proper object of examination. Before I examine Soames' specific criticisms of my seven arguments, I should make an observation about a general claim that Soames made. He said that Marcus could not possibly be the originator of the main ideas of the New Theory of Reference, since she presented her ideas in too few number of pages. Soames writes about my claim about Marcus: "One clue that this claim cannot be correct comes from the observation that proper statements and defenses of the various doctrines of the so called 'new theory' would not fit into such a small space" (p. 193, my italics). But surely the following argument is invalid: if an idea takes up only three pages in Marcus, but is spread with numerous 37 P. W. Humphreys et al. (eds.), The New Theory of Reference © Kluwer Academic Publishers 1998

38

QUENTIN SMITH

illustrations and repetitions over thirty pages in Kripke, that fact is sufficient to deprive Marcus ofpriority. By comparison, F.P. Ramsey is generally credited with priority for the Dutch book argument for justifying the axioms of probability on the personalist or subjectivist interpretations of the axioms. But this accreditation is based on exactly two sentences on page 182 of his essay "Truth and Probability" (Foundations of Mathematics, Routledge and Kegan Paul, 1931). De Finetti developed this argument much more elaborately several years later (over 30 pages in Fundamenta Mathematica, Vol. 17, pp. 298-329), but this does not erase the fact of Ramsey's priority. Indeed, de Finetti acknowledged that Ramsey originated the argument. Similarly, Marcus presented in a very concise fashion almost all the main ideas and arguments of the New Theory of Reference in her 1961 article; Kripke's extensive elaboration upon her ideas, with the introduction of numerous illustrations and a new terminology, does not count against Marcus' priority, as I shall now argue in detail. I. DIRECT REFERENCE AND DISGUISED DESCRIPTIONS

I argued that the first idea Marcus introduced into the New Theory of Reference is that proper names are directly referential and not equivalent to definite descriptions. Soames admits that Marcus did present the idea that proper names are directly referential and not equivalent to definite descriptions, but claims that Marcus' presentation of this idea was antedated by Fitch and Smullyan. Soames writes: Although Marcus did regard names as directly referential, and did not take them to be disguised descriptions, she did not introduce the idea; nor did she claim to. As we have seen, Smullyan and Fitch both invoked that idea in earlier responses to Quine that Marcus took herself to be repeating and elaborating. (pp. 198-9).

The textual evidence does not support Soames' claims about Fitch and Smullyan. In an article that Soames does not mention, Fitch's article 'Attribute and Class', published in 1950 in Philosophical Thought in France and United States (edited by Farber, with a preface written in 1949), Fitch explicitly attributes the idea that names and descriptions behave differently in modal and other contexts to his doctoral student Ruth Barcan Marcus (her name was then Ruth Barcan). Fitch writes: The system of modal logic developed by Ruth Barcan suggests that the simplest view is that no identities should be regarded as merely contingent and that identified entities should be everywhere intersubstitutable ... Furthermore, if entities X and Y have been identified with each other, it seems reasonable to suppose that the names of X and Y should also be everywhere intersubstitutable where they are being used as names. According to Church's view, on the other hand [which is distinct from Ruth Barcan's view] two names of the same thing might differ in sense and so not be intersubstitutable. (p. 252, edited Farber 1950, Buffalo.)

This passage confirms what Marcus told me in an interview, that she developed this view of names and descriptions in course of writing her thesis in 1943-45

MARCUS AND THE NEW THEORY OF REFERENCE

39

and explained them in conversation to her dissertation adviser Fitch. This is in keeping with the passage from Fitch I just quoted, where Fitch mentioned these ideas and attributes them to Marcus. Notice that he is not advancing these ideas as his own ideas; rather, he is explaining Marcus' ideas and contrasting them with Church's ideas as two alternate views of the behavior of names in intensional contexts. Fitch refers again to his indebtedness to conversations with Marcus in his paper on the 'Problem of Morning Star and Evening Star', where he says "The writer wishes to acknowledge his debt to Miss Barcan for some helpful discussions concerning the ideas of this paper". (Philosophy of Science, Vol. 16, 1949, p. 139, n. 6). Thus, Fitch received - and admitted receiving - the idea about the distinctive behavior of names from Marcus and it is false that Marcus repeated and elaborated on Fitch's original idea. Soames also claims that Smullyan earlier developed the idea that names are directly referential and are not disguised descriptions. Soames is also mistaken on this point. He refers to Smullyan's 1948 article on 'Modality and Description" in The Journal of Symbolic Logic and his 1947 review (in the same journal) of Quine's article on modal logic. The reference to the 1948 article on 'Modality and Description' is a red herring, since Smullyan there discusses the issue of whether the modal paradoxes suggested by Quine can be resolved by treating the relevant expressions, not as names, but as descriptions or class abstracts. For example on page 35 he says that: Our discussion of the argument, ABC, may seem to have depended, at certain crucial points, upon our decision, to interpret the expression, 'the number of the planets' as a definite description. What would have been the situation had the phrase in question been interpreted as a class abstract, as synonymous with 'x (x is equinumerous with the set of planets), for example?

Smullyan also goes on to show how the modal paradoxes can be avoided if the expression is interpreted as a class abstract. In this paper, Smullyan discusses only descriptions and class abstracts; names are not discussed at all, contrary to Soames' contention. Smullyan's 1947 review of Quine's paper, is, however, relevant, but even in this paper there is only one sentence about names - and even this one sentence Smullyan rejects as false. Furthermore, this sentence (even apart from the fact that Smullyan rejected it as false) does not even imply that names are directly referential and are not disguised descriptions. The sentence in question is on page 140. Smullyan writes: if 'Evening Star' and 'Morning Star' proper-name the same individual they are synonymous and B is false, where B is the statement: (B)

Evening Star is congruent with Evening Star and it is not necessary that Evening Star is congruent with Morning Star.

40

QUENTIN SMITH

Soames' gloss on this sentence is that "Smullyan adds to this claim that coreferential names are synonymous, which strongly suggests that genuine names have no meaning apart from their referents, another view later endorsed by Marcus" (p. 196). I think Soames is misinterpreting Smullyan's remark. If coreferential names are synonymous, that does not mean, imply or even suggest they are directly referential. And if coreferential names are synonymous in modal contexts, such as Smullyan's statement B, that still does not mean, imply or suggest they are directly referential. For it is perfectly consistent with the claims that they are indirectly referential, that they express a world-indexed sense, such as Linsky and Plantinga claimed. "Evening Star" and "Morning Star" could both express the sense "whatever is actually the planet Venus". Smullyan shows no awareness whatsoever of this distinction between direct reference and worldindexed senses, and it would be a mistake to read an awareness of this distinction, and an advocacy of direct reference rather than world-indexed sense, into his one sentence remark about synonymity. Further, Soames' suggestion that Smullyan advocated this idea about the expressions being synonymous names is false, since Smullyan in the very next paragraph goes on to reject the idea that "Evening Star" and "Morning Star" should be regarded as synonymous names in modal contexts. He instead says that they should be regarded as descriptions. Smullyan writes: "On the other hand, if, more naturally, we view "Evening Star" and "Morning Star" as abbreviations of descriptive phases, we find that A expresses an evidently impossible proposition" (p. 140, my italics), where A is another modal proposition. Thus, Soames'. claim that Marcus' theory that ordinary names are directly referential and not disguised descriptions was first espoused by Fitch and Smullyan is an incorrect claim. The idea is original with Marcus. (Of course, we are talking about the contemporary New Theory of Reference; as everyone knows, Mill held a direct reference theory in the 19th century.) We should also point out that some others claimed in passing that ordinary proper names are directly referential, such as Paul Fitzgerald, but they did not support their claims with the arguments about the nonequivalence of names and descriptions in modal and epistemic contexts that have since become part of the New Theory of Reference. 2. REFERENCE-FIXING DESCRIPTIONS

My second claim was that Marcus introduced the idea that entities are singled out by definite descriptions, but that assigning the entities a name is semantically different than describing them. Marcus' view is that singling out an entity by unique descriptions is a precondition oflanguage, and that singling out entities this way is the only way to single them out (Marcus, 1961: 309). It follows that when we assign a name to a thing that we have singled out, we can assign the name only on the basis of a unique description that we are using to

MARCUS AND THE NEW THEORY OF REFERENCE

41

single out the thing. Soames claims that this does not imply that a specific description is linked as a matter of semantics to the name, and thereby that Marcus did not originally state the idea of reference-fixing descriptions later stated by Kripke. But Soames here appears to misunderstand Marcus and the concept of a reference-fixing description. The crucial premises in Soames' argument are that (quoting from Soames' paper): ... if the referent of a name is semantically fixed by a certain description, then being a competent speaker who understands the name will involve knowing that if the name has a referent at all, it must be one that satisfies the description. Similarly, if a name N has its referent semantically fixed by a description D, then one who understands the sentence "If N exists, then N is D" will know, without empirical investigation, that it expresses a truth. Finally, if one later finds that the description one associated with the name fails to pick out anything, or fails to pick out the individual one has in mind, then it will follow that the name failed to refer, or at least failed to refer to the individual one had in mind. None of this is remotely suggested in the quotation from Marcus (pp. 199-200).

None of this is remotely suggested in the quote from Marcus since these ideas are neither Marcus', nor Kripke's, nor ideas that are consistent with the concept of a reference-fixing description in the New Theory of Reference. Soames is assuming that if a name N has its referent semantically fixed by a description D, then certain conditions must be met that violate the epistemic argument for direct reference that Marcus first put forth and Kripke, Salmon and others later repeated. According to the New Theory of Reference, a description may serve to originally fix the direct referent of a name, but the description is not everafter associated with the name by competent speakers of the language. The a priori association need be present only on the original reference-fixing occasion. Soames' second claim (viz., that if one later finds that the description one associated with the name fails to pick out what one had in mind, then the name will fail to refer) is inconsistent with the fact that descriptions can be used in a referential as well as an attributive way, as Keith Donnellan first pointed out. Even if we reject Donnellan's formulation of the notion of a referential definite description, as does Kripke ('Naming and Necessity', note 37, pp. 349), it is still the case that the reference of a name can be fixed by a description that fails to apply to the named object. Soames' contrary claim is directly contradicted by Kripke in footnote 34 of 'Naming and Necessity' (p. 348), where Kripke writes: "Following Donnellan's remarks on definite descriptions, we should add that in some cases, an object may be identified, and the reference of the name fixed, using a description which may turn out to be false of its object. The case where the reference of 'Phosphorus' is determined by 'morning star', which later turns out not to be a star, is an obvious example. In such cases, the description which fixes the reference clearly is in no sense known a priori to hold of the object ...". Thus, Soames' two conditions for a reference-fixing description are inconsistent with this notion as it is understood by Kripke, Marcus and other proponents of the New Theory of Reference. Further, he ignores the argument

42

QUENTIN SMITH

in Marcus' paper that names do have their reference fixed by descriptions. The argument is easily extracted from the theses advocated by Marcus. Marcus advances these theses: (1) (2) (3)

Names directly refer to entities. Whenever we refer to a single entity, the entity is singled out. Entities are singled out only by unique descriptions.

These theses logically imply, (4)

When we first assign a name to an entity, we do so by means of singling out the name's referent by a unique description.

Thus, I conclude that Soames' attempt to show that Marcus did not introduce the second idea into the New Theory of Reference is not successful. 3. THE MODAL ARGUMENT FOR DIRECT REFERENCE

Soames agrees with me that Kripke's modal argument for direct reference was earlier stated by Marcus in her 1961 paper and that Marcus should have been credited for this. But Soames erroneously claims that Fitch and Smullyan offered the modal argument before Marcus. I have already quoted the passage from Fitch where he says this argument was suggested to him by his student Marcus. As for Smullyan, he did not present the modal argument that proper names are directly referential. The nearest he gets to this argument is to mention (but then reject) the claim that if "Evening Star" and "Morning Star" are synonymous, and name the same individual, then it is false that it is not necessary that the Evening Star is congruent with the Morning Star. Several comments can be made about this sentence in Smullyan's review of Quine's article. First, Smullyan is not advancing the modal argument for direct reference. The modal argument is that names are co-referential in modal contexts and therefore are directly referential. But Smullyan instead says that names are co-referential in modal contexts and therefore are synonymous, which is a different argument altogether. Furthermore, as I have already pointed out, Smullyan does not endorse, but in fact rejects, the argument for synonymity he presents. After mentioning the idea that "Evening Star" and "Morning Star" are proper names and therefore synonymous in modal contexts, he goes on to reject this view. He says "On the other hand, if, more naturally, we view 'Evening Star' and 'Morning Star' as abbreviations of descriptive phrases ..." (p. 140, my italics). In sum, Smullyan did not endorse or even mention the modal argument for direct reference. It was only Marcus who argued that proper names in ordinary language are directly referential, and who used their behavior in modal contexts to show this. This idea is original with Marcus. I

MARCUS AND THE NEW THEORY OF REFERENCE

43

quoted her argument in my original paper and I will not quote her again: in any case, Soames does not dispute the fact that Marcus presented this argument. (I should add the qualification made in my original paper 'Marcus, Kripke, and the Origin of the New Theory of Reference', namely, that the modal argument shows only that proper names do not express contingent descriptive senses; it does not show that proper names do not express modally stable senses. The epistemic argument, discussed below, is required to refute the Linsky-Plantinga "modally stable sense" theory of proper names). {Material added in May, 1995:

My reply to Soames was written in response to a version of his paper which I received in mid-December, 1994, and Soames later added some new material to this version; the new version is the one he read at the APA colloquium and that is printed in this issue of Synthese. His new material consists of the long paragraph on page 202 beginning with "I should mention that, historically", the four sentences beginning with "There are two problems here" on page 201, footnotes 26, 34 and 36, and several minor wording changes and connections elsewhere in the paper. Here I shall partially respond to footnote 34, which is on the modal argument; in Section 8 I will respond to other aspects of footnote 34. In footnote 34, Soames argues that there "is an important feature of Kripke's version of the modal argument that does not seem present in Marcus' version. Both Kripke and Marcus use modal properties of names to distinguish them from descriptions" (p. 231, n. 34). Soames goes on to suggest that Kripke's version of the modal argument goes like this: (A)

(1) Ordinary proper names are rigid designators. Therefore, (2a) Identities involving ordinary proper names are necessary if true, and (2b) Co-referring ordinary proper names are intersubstitutable in modal contexts. Soames also suggests that Marcus' version of the modal argument goes like this: (B)

(3) (4)

The quantified version of the necessity of identity is true. Ordinary proper names are directly referential.

Therefore (Sa) Ordinary proper names are rigid designators,

44

QUENTIN SMITH

and (5b) Identities involving ordinary proper names are necessary if true, and (5c) Co-referring ordinary proper names are intersubstitutable in modal contexts. This added footnote indicates that Soames has not correctly identified the modal argument for direct reference. Neither argument (A) nor argument (B) are modal arguments for direct reference. The conclusion of the modal arguments for direct reference is proposition (4), which Soames makes a premise of argument (B) and which he omits altogether from argument (A). Soames' decision to use the phrase "modal arguments" to refer to (A) and (B) is idiosyncratic and does not conform to how this phrase is used by New Theorists of Reference. For example, Nathan Salmon writes that: A number of arguments have been advanced in favor of the central thesis of the theory of singular direct reference. Although the arguments are many and varied, most of them may be seen as falling under one of three main kinds: modal arguments, epistemological arguments, and semantic arguments ... The modal arguments are chiefly due to Kripke. [Salmon, Reference and Essence, Princeton University Press, 1981. pp. 23-24]

Salmon correctly observes that: there is a weakness in the modal arguments; they show only that names and indexical singular terms are not descriptional in terms of the simple sorts of properties that come readily to mind, properties like the authorship of a certain work ... But faced with the modal arguments, some descriptional theorists, such as Linsky [1977, p. 84] and Plantinga [1978] have moved to fancier descriptions employing modally indexed properties ... Thus the modal arguments seem ineffective against the thesis that proper names are descriptional in the Linsky way. [Ibid., pp. 26-27]

By distinguishing argument (A) from argument (B), and claiming (A) is present only in Kripke's writings and (B) present only in Marcus' writings, Soames is in effect changing the subject, which is whether Marcus first stated the modal argument that proper names are not descriptional but directly referential. And that Marcus first stated this argument is evident from my quotations in 'Marcus, Kripke, and the Origin of the New Theory of Reference', pp. 182-3, for these quotations show that she argued that co-referring proper names are intersubstitutable in modal contexts and therefore are not equivalent to contingent descriptions. (Of course, we need the caveat that the modal argument for direct reference only shows that ordinary proper names are not equivalent to contingent descriptions, as Salmon, Plantinga, Linsky and others have recognized.) End ofadded material}

MARCUS AND THE NEW THEORY OF REFERENCE

45

4. THE NECESSITY OF IDENTITY

The fourth argument I claimed Marcus introduced into the New Theory of Reference was the necessity of identity. Soames' reply is that in Marcus' 1946-47 papers, she proved that the necessity of identity involved only variables, and there is no direct route from this to the view that the theorem of the necessity of identity also applies to proper names. But Soames' reply is unsuccessful in two respects. First, he is incorrect in assuming that it is an obvious fact that there is no direct route from variables to proper names. There arguably is a formal equivalence between variables and proper names, an equivalence that has been recognized as early as 1939 with Quine's article, 'Designation and Existence' (Journal of Philosophy, 1939), where Quine writes: ... we might equivalently omit express mention of existential generalization and describe names simply as those constant expressions which replace variables and are replaced by variables according to the usual logical laws of quantification. (Vide, Readings in Philosophical Analysis, edited by Feigl and Sellars, 1949, p. 50.)

Thus, Soames is making a dubious claim when he says that it is not clear that his (NI-I) (about variables) goes to his (NI-2) (about names) rather than to his (NI-3)) (about any singular descriptions). But more importantly than this point about variables and proper names is the fact that Marcus herself applied the theorem of the necessity of identity to proper names in her 1961 paper. She takes great pains to make this application explicit in propositions (13) through (18) of her paper and in her section entitled "Semantic Constructions" (pp. 319-321). So it is simply false that this application was first made by Kripke in the 1970s. Material added in May, 1995: One of the additions Soames subsequently made to the paper he sent me in mid-December is a paragraph on page 202, the paragraph beginning with "I should mention that ..." and ending with " ... the truth of the others". In this new paragraph, Soames criticizes Quine, Marcus, Smullyan, Fitch and others for assuming that "the intelligibility of objectual quantification depended upon universal instantiation and existential generalization being universally truthpreserving, where genuine singular terms are involved" (p. 202). Soames states that this assumption is false and that this assumption was taken by Marcus and others as a reason to believe that if (NI-l) is true, then (NI-2) and (NI-3) are true, where the relevant theses are (NI-l) (x) (y) (if x = y), then necessarily x = y. (NI-2) For all proper names a and b, if a = b is true then necessarily a = b is true. (NI-3) For all singular terms a and b (including singular definite descriptions), if a = b is true then necessarily a = b is true. I think Soames' added paragraph about these three theses is misguided in several respects.

46

QUENTIN SMITH

First, the motive for his discussion of the relations among (NI-I), (NI-2)) and (NI-3) in his claim that I am "just plain wrong to cite Marcus' early systems of quantified modal logic as introducing the version - (NI-2) - of the necessity of identity that corresponds to the modal argument" (p. 202). However, I did not state that Marcus' early articles on quantified modal logic introduced the thesis (NI-2). Rather, I stated that her early articles introduced (NI-l) to the New Theory of Reference. (See pp. 183-4 of my 'Marcus, Kripke, and the Origin of the New Theory of Reference', where I stated that in her early articles Marcus presented a "formal proof of the necessity of identity in her extension of S4 [Barcan, 1946; 1947], which is a fourth component she introduced into the New Theory of Reference. She showed that (T)(xly) ~ o (xly) is a theorem ofQS4 ..."). Marcus introduced (NI-2) to the New Theory of Reference in her 1961 article, in propositions (13)-(18) and in the section on "Semantic Constructions". Second, Marcus does not argue that (NI-l) entails (NI-2) and (NI-3), and Soames makes no attempt to provide textual evidence that Marcus presents such an argument. I also do not argue that there is such an entailment. Soames' criticism of the belief that there is such an entailment seems to be a case of setting up a straw-woman (Marcus) and a straw-man (myself). Third, in this section (section #4, "The Necessity of Identity") of my reply to Soames, I was in part replying to a statement made in the paper I received from Soames in mid-December, 1994 (a statement that also appears in the version he read at the APA colloquium), viz. the statement that "the quantificational version (NI-l) of the necessity of identity, articulated and derived by Marcus, no more provides a direct route to the true thesis (N1-2) involving ordinary proper names than it provides a direct route to the false thesis (NI-3)" (p. 202). In his new paragraph, Soames takes the originally vague phrase "provides a direct route to" to have a precise meaning, namely, entails. If "provides a direct route to" is taken to mean entails, then of course the true thesis (NI-l) no more provides a direct route to the true thesis (N1-2) than it provides a direct route to the false thesis (NI-3). In order for (NI-l) to entail (NI-2), additional premises are needed, such as the premise that proper names do not express contingent descriptive senses and the premise that variables and proper names are formally equivalent in some contexts. A point I was making in section #4 of this paper was that Soames' originally vague claim that (NI-l) "no more provides a direct route to" (NI-2) than to (NI-3) is a dubious claim. It is dubious since it can be plausibly argued that there is a "direct route" from (NI-l) to (NI-2) in the sense that the true thesis (NI-2) can be logically derived from the true thesis (NI-l) in conjunction with some plausible supplementary premises about proper names and variables (e.g., that proper names are formally equivalent to variables in some contexts), whereas the false thesis (NI-3) cannot be logically derived from the true thesis (NI-l) in conjunction with supplementary premises. End ofadded material. }

MARCUS AND THE NEW THEORY OF REFERENCE

47

5. RIGID DESIGNATION

Another crucial idea is the concept of rigid designation. Soames allows that Marcus implicitly treated proper names as rigid designators. But Soames claims that in order to have the concept of a rigid designator, one needs a semantical framework for modal systems that features the notion of a referent of a singular term at a possible world, and he denies that Marcus had this notion. So ames seems to make a twofold error here. First, Marcus did present (in her 1961 paper) a semantical construction for a modal language that featured the notion of a referent of a singular term at a possible world. She presents a model-theoretic system that (to quote from her 1961 paper) "corresponds to the Leibnizian distinction between true in a possible world and true in all possible worlds" (p. 320). Specifically, she presents (again quoting from her 1961 paper): a language (L), with truth functional connectives, a modal operator (0), a finite number of individual constants, an infinite number of individual variables, one two-place predicate (R), quantification and the usual criteria for being well-formed. A domain (D) of individuals is then considered which are named by the constants of L. A model of L is defined as a class of ordered couples (possibly empty) of D. The members of a model are exactly those pairs between which R holds. To say therefore that the atomic sentence R(ala2) of L holds or is true in M, is to say that the ordered couple (b l, b2) is a member of M, where al and a2 are the names in L of b l and b2 ... If [a sentenceI A is 0 B then A is true in M if and only if B is true in some model M I. (Marcus, p. 319).

Here the models M, M j etc. are possible worlds, the individuals in the domain D are members of the worlds, and these individuals are referents of the individual constants. Clearly, we do have here a semantics for modal logic that features the notion of a referent of a singular term at a possible world. Kripke presented a largely similar semantics for modal logic in his 1963 paper 'Semantical Considerations on Modal Logic' with the main difference between Marcus' and Kripke's semantics being that Kripke (following Stig Kanger's 1957 work Provability in Logic [Stockholm]) allowed that the domain of individuals is not constant across models, whereas Marcus held the domain of individuals is constant across models. Marcus' models or worlds differ from one another in that her relation R holds between different individuals in different worlds. But this difference between Marcus and Kripke about the constancy of the domain of individuals is not relevant to the basic concept of rigid designation; rather, it is relevant to whether this concept has two species, weak rigid designation and strong rigid designation (as Kripke held) or whether this weak/strong distinction cannot be made. The relevant fact is that one can derive from both Marcus' seman tical system and Kripke's seman tical system the notion of a referent of a singular term at a possible world and the associated concept of a rigid designator. But it should be added that Soames' claim that we need a model-theoretic semantics with the notion of a possible world in order to have the notion of rigid designation is a dubious claim. The concept of rigid designation can be fully stated using only modal operators ("possibly", "necessarily") and sub-

48

QUENTIN SMITH

junctives ("might have been", etc.), without need of a semantics for modal logic and without the notion of a referent of a singular term at a possible world. Indeed, this is how Kripke states it in 'Naming and Necessity'. Kripke writes: One of the intuitive theses I will maintain in these talks is that names are rigid designators. Certainly they seem to satisfy the intuitive test mentioned above; although someone other than the U.S. President in 1970 might have been the U.S. President in 1970 (e.g. Humphrey might have), no one other than Nixon might have been Nixon .... For example, "the President of the U.S. in 1970" designates a certain man, Nixon; but someone else (e.g., Humphrey) might have been the President in 1970, and Nixon might not have; so this designator is not rigid. (Kripke, 1972, p. 270)

Here the notion of rigid designator is explained in terms of subjunctive expressions "might have been", "might not have been", and no mention of possible worlds is needed. This is how Marcus first introduced the notion in her 1961 article. Note that the distinctions made in the following passage from Marcus are logically equivalent to the distinctions made in the passage I just quoted from Kripke's 'Naming and Necessity'. Marcus writes: Let us now return to (10) and (15). [(10) is "The evening star eq the morning star" and (15) is "Scott is the author of WAVERLY"]. If they express a true identity, then"Scott" ought to be anywhere intersubstitutable for "the author of WAVERLY' in modal contexts, and similarly for "the morning star" and "the evening star". If they are not so universally intersubstitutable - that is, if our decision is that they are not simply proper names for the same thing; that they express an equivalence which is possibly false, e.g., someone else might have written WAVERLY, the star first seen in the evening might have been different from the star first seen in the morning - then they do not express identities (Marcus, 1961, p. 311)

This passage conveys the distinction between rigid and nonrigid designators that is also conveyed in the passage quoted from Kripke. If we add to this the fact that Marcus had a semantics for modal logic that featured the notion of a referent of a singular term in a world, then it is quite clear that Marcus was in full possession of the concept of rigid designation in 1961. Soames writes: "Smith's characterization of Kripke as contributing only the name 'rigid designation' for a concept discovered by Marcus is a grotesquely inaccurate caricature" (p. 203). I suggest that so far from being a grotesquely inaccurate caricature, it is the literal truth. 6. THE EPISTEMIC ARGUMENT FOR DIRECT REFERENCE

A further idea that Marcus introduced into the New Theory of Reference is the epistemological argument for direct reference. Consider this passage from Marcus' 1961 paper: You may describe Venus as the evening star, and I may describe Venus as the morning star, and we may both be surprised that, as an empirical fact, the same thing is being described, but it is not an empirical fact that (17)

Venus I Venus

MARCUS AND THE NEW THEORY OF REFERENCE

49

and if 'a' is another proper name for Venus that (18)

Venus I a. (p. 310).

Soames claims that if Marcus is making an epistemological point here, she cannot be using "empirical" to mean "contingent". But Soames' claim is not true. Even if we construe "empirical" as meaning "contingent", the epistemic argument for direct reference is still logically implied by this passage. The argument goes (1)

If "Venus" expresses the sense of "the evening star", then it is knowable a priori that Venus is the evening star.

(2)

This is not knowable a priori, since we may be surprised to learn that, as a matter of contingent fact, Venus is the evening star.

Therefore, (3)

"Venus" does not express the sense of "the evening star" (or any other descriptive sense) but is directly referential.

Soames concentrates on the word "empirical" in the passage from Marcus, but the epistemic argument for direct reference is sufficiently conveyed by Marcus' point that we may be surprised to discover that Venus is the evening star, which is something that we could not be surprised to discover if we knew this fact a priori by virtue of the meaning of the word "Venus". However, Marcus is in fact using "empirical" in its normal philosophical sense to mean a posteriori. Given this, it is even more clear that her passage implies the epistemic argument that proper names are directly referential. The argument goes: (4)

It is an a posteriori fact that Venus is the evening star.

(5)

If "Venus" has the sense of "the evening star", it would not be an a posteriori fact that Venus is the evening star.

Therefore (6)

"Venus" does not have the sense of "the evening star" (or any other descriptive sense, since this argument can be repeated for other descriptive senses).

Soames apparently does not see this argument in Marcus' paper. He writes: "To get an epistemological argument against descriptivism one would have to argue directly, without appeal to Millian premises, that one cannot know a priori that N is D, for relevant names N and descriptions D. Marcus does not do this" (p. 207). However, I have just shown that she did do this.

50

QUENTIN SMITH

At the risk of belaboring the point, I would note that one premise of her argument is that it is an empirical or a posteriori fact that Venus is the evening star. This is not a Millian premise, where a Millian premise is one that states that names are directly referential and do not express senses. The second premise is that "if 'Venus' has the sense of 'the evening star', it would not be an a posteriori fact that Venus is the evening star". This is also not a Millian premise; it does not state that the word "Venus" is directly referential. The conclusion is that the word "Venus" does not have the sense of "the evening star"; this is a Millian thesis, but it is not a premise of the argument but the conclusion. In my paper, I gave a different illustration of the epistemic argument in order to show how the epistemic argument could be used to refute Plantinga's and Linsky's theory that names express modally stable senses, i.e., senses that include the actuality operator. My example was of the name "Venus" purportedly expressing the modally stable sense "whatever is actually the morning star and the evening star". Soames makes a puzzling comment about my example. He correctly states (on page 206) that the modally stable sense I mention, "whatever is actually the evening star and morning star", and the example I give, is not a reconstruction of Marcus' example, but is my own invention. But Soames seems to think that this statement counts as a criticism of my example, whereas it is clear from my paper that I make no claim to be analyzing Marcus' different example. Rather, I introduced my own example in order to show how "the epistemic argument for direct reference" can be used to refute the Linsky-Plantinga theory that names express modally stable senses. 1 7. A POSTERIORI NECESSITIES

I turn now to the idea of a posteriori necessities. Soames denies that the notion of a posteriori necessity is present in Marcus. Soames comments on the passage I just quoted from Marcus, which is worth quoting again to keep before our minds: You may describe Venus as the evening star, and I may describe Venus as the morning star, and we may both be surprised that, as an empirical fact, the same thing is being described. But it is not an empirical fact that (17)

Venus I Venus

and if "a" is another proper name for Venus that (18)

Venus I a. (Marcus, p. 310).

Soames comments that: If, as Smith maintains, Marcus is trying to make an epistemological point here, then by "an empirical fact" she must not mean "a contingent fact", for then there would be no epistemological point at issue. Rather, she must be taken to mean that these sentences do not express facts that are

MARCUS AND THE NEW THEORY OF REFERENCE

51

knowable only by empirical means. In other words they are knowable a priori. But if this is right, then true identity statements involving names are not characterized as examples of the necessary a posteriori, and Smith's contention that Marcus introduced this notion collapses. (p. 204)

Soames is here making an objectlmetalevel confusion that prevents him from understanding both Marcus' passage and the basic idea behind the necessary a posteriori. In the first sentence quoted, Marcus is saying that we may both be surprised that our two proper names of Venus, call them "a" and "b", are coreferring. She further states that it is an empirical fact that they are co-referring. In the first sentence, she is talking about metalevel facts about the reference of the two proper names. It is a contingent and a posteriori fact that the two proper names are co-referring. This is the case if the two proper names are homonyms (two uses of "Venus"), or whether one is "Hesperus" and the other "Phosphorus", or whether we call them "a" and "b". The reason this metalevel fact is contingent is that the two names might have had different referents. If your use of "Venus" is determined by the reference-fixing rule that "Venus" refers to the evening star, and my use of "Venus" is determined by the reference-fixing rule that "Venus" refers to the morning star, then it is a contingent fact that our two uses have the same referent, for it is not necessary that whatever has the property of being the evening star also has the property of being the morning star. Not only is this fact contingent, it is a posteriori, since it is only by observing the world that I can discover that one and the same body has these two properties and thereby is the referent of both uses of "Venus". Thus, the a posteriori feature of a posteriori necessary sentences pertains to the metalevel facts about the referents of the expressions in the sentence and thus to the metalevel facts about what is said by the sentence (i.e., what de re proposition is expressed by the sentence). Marcus uses the "what is said" terminology, and was the first to introduce this terminology (as Joseph Almog pointed out in his 1986 article 'Naming Without Necessity" in The Journal of Philosophy, Vol. 83, pp. 220, n. 8), but we can also use the terminology of the "de re proposition" or "singular proposition" expressed by a sentence. The necessary aspect of a posteriori necessary sentences pertains to the object level or, in other terms, to what is said by the sentences, the de re proposition expressed. In the first sentence in the quote, Marcus was talking about the metalevel, about our knowledge of the empirical fact that "the same thing is being described" by our two uses of the name "Venus". But in the second sentence in the quote, Marcus is talking about the object level, about what is said by her sentences. Her second sentence reads: "But it is not an empirical fact that: (17)

Venus I Venus

and if 'a' is another proper name for Venus that (18)

Venus I a".

52

QUENTIN SMITH

Note that Marcus is no longer talking about facts about what our words refer to; rather, she is talking about the identity of Venus with itself. This is what is said by the sentence, the de re proposition expressed. What is said by the sentence, that Venus is identical with Venus, is not an empirical or a posteriori fact. It is an a priori fact, since it is a mere tautology, an identity proposition. Thus, what is empirical is that two names (be they two uses of "Venus" or "Hesperus" and "Phosphorus") both refer to Venus. And what is not empirical is that Venus is identical with Venus. We have this correlation: the metalevel fact (mentioned by Marcus in her first sentence) is a posteriori and contingent, and the object level fact (mentioned by Marcus in her second sentence) is a priori and necessary. Generally speaking, what is said by an a posteriori necessary sentence may be a tautology or analytic truth that is known a priori. This is not only true in Marcus' theory of the a posteriori necessary, but also in Kripke's and Putnam's. For example, what is said by the sentence "Water is H 2 0" is the tautology H 2 0 is H 2 0, since the word "water" directly refers to H 20. This tautologous proposition is the object level fact. This tautological fact is known a priori; obviously, nobody learns a posteriori that H 20 is identical with H 20. The metalevel fact is the a posteriori fact; the metalevel fact is that the word "water" refers to H 2 0. This can only be known a posteriori, by observing the world to see what satisfies the reference-fixing description that governs the use of the word "water". The description that fixes the reference of the word "water" is the description "whatever is causally responsible for the observable properties of a thirst-quenching, clear, tasteless liquid". We learn a posteriori that H 20, rather than some other chemical structure, such as Putnam's XYZ, is what satisfies the reference-fixing rule of use of "water". Now compare this with Soames' remark on page 205: "Thus, according to Marcus true identity statements involving genuine names never express empirical facts, but rather are tautologically, or analytically true. Surely this is inconsistent with interpreting her as introducing the notion of the necessary a posteriori identities". This remark betrays Soames' object levellmetalevel confusion. What is said or expressed by true identity sentences involving genuine names is tautologically true, and what is said is known a priori. But that is perfectly consistent with knowing a posteriori the metalevel fact that the sentences in question express this tautological proposition rather than some other proposition. In order to know that they express this de re proposition, and not some other de re proposition, we need to observe the world to determine what satisfies the reference-fixing rules of the relevant words in the sentence. Soames' account of the notion of a posteriori necessity implies that it is logically impossible that there be any a posteriori necessities. For Soames claims that if what is said by a sentence is a priori, then the sentence cannot be an a posteriori necessary sentence. On this construal, the sentence "water is H 20" would count as an a priori necessity, since what is said by this sentence, that H 2 0 is H 2 0, is known a priori.

MARCUS AND THE NEW THEORY OF REFERENCE

53

A correct understanding of the a posteriori necessary would show that the identity sentence "water is H 20" is an a posteriori necessity because: (a) On the metalevel, it is an a posteriori fact that this sentence expresses the de re proposition H 2 0 is H 2 0 and not some other de re proposition, such as XYZ is H 2 0 (to borrow Putnam's example). (b) On the object level, what is expressed by this sentence, the de re proposition that H 2 0 is H 2 0, is a necessary fact, indeed, a tautology, and is known a priori. We may summarize this discussion by considering Soames' sentence: "on the view articulated by Marcus, true identity statements involving names are knowable a priori" (p. 205). A correct phrasing of Marcus' view would read instead "on the view articulated by Marcus, true identity statements involving names are knowable a priori on the object level but are knowable a posteriori on the metalevel". I believe a central reason why Soames did not recognize the argument for a posteriori necessary in Marcus' work is that he uses her absence of the current Kripkean terminology ("a posteriori necessities", etc.) as evidence that she did not have the concept of the a posteriori necessity. For example, Soames takes pains to emphasize that "It should be noted that Marcus nowhere in the article says that there are true identity statements which are knowable only a posteriori" (p. 204). Certainly, Marcus never called any of her arguments "the argument for a posteriori necessity" or the like. But the point, rather, is that these arguments are logically implied by what she does say, which can be put very succinctly. Soames admits that Marcus implicitly treated proper names as directly referential rigid designators. Given this admission, it is a short step to realize that this doctrine itself entails the doctrine of the a posteriori necessity. If "a" and "b" are directly and rigidly referential names of the same thing, it immediately follows (given Marcus' theorem of the necessity of identity) that "alb" states a necessary truth, but the fact that "a" and "b" are coreferential (and that "alb" thereby states a necessary truth) is known only a posteriori. On the object level, "alb" is necessary and a priori, but on the metalevel, it is contingent and a posteriori. We don't even need to argue about the meaning of Marcus' passage that I quoted, since the doctrine of the a posteriori necessity is logically entailed by the doctrines that Soames himself admits were espoused by Marcus. {New material added in May, 1995} One of the additions Soames made to the paper I received in mid-December is footnote 26, which analyses Marcus' response to one of Kripke's questions in the Discussion after her presentation of 'Modalities and Intensional Languages' on February 7, 1962 at the Boston Colloquium for the Philosophy of Science. Soames' new footnote pertains to his quotation (on page 198) from Marcus' Discussion in which she elaborates upon her doctrine of a posteriori necessities. In this quotation, Marcus links the epistemological status of coreferring names to their modal and semantic status, in a way that implies the

54

QUENTIN SMITH

doctrine of the a posteriori necessary. Soames' new footnote 26 contains the remark that this linkage marks "a very important contrast" between Marcus' theory and Kripke's. Part of the "very important contrast" that Soames alleges is based on his incorrect attribution to Marcus of the view that names are mere tags entails or "is the ground for" the thesis that coreferring names are intersubstitutable in modal contexts; Marcus does not hold this view since Marcus' theory is that further premises are needed for this entailment to go through, e.g., the premise that identities are necessary. But more germane to our discussion of a posteriori necessity is that in this new footnote Soames seems to find suggestive evidence of a contrast between Marcus and Kripke in that "Marcus responds to a straightforward question about the necessity of identities involving names (and their intersubstitutivity in modal constructions) with remarks about epistemology and synonymy" (p. 212, note 26). Soames' belief that Marcus' linkage of these issues marks a contrast between her and Kripke is indicative of his failure to grasp the fact that in her response she explains her argument for the a posteriori necessary, as I will now show in detail. In the 1962 Discussion, Kripke somewhat innocently or naively asks Marcus if her theory requires the thesis "that for any two names 'A' and 'B', of individuals, 'A = B' should be necessary, if true at all? (See "Discussion on the Paper of Ruth B. Marcus", Synthese, Vol. XIV, no. 2/3, September, 1962, pp. 141-42. This Discussion is reprinted with minor changes in Marcus' Modalities; see p. 33). Since it should have been obvious from Marcus' talk that this thesis was one of the major points of her talk, Marcus chooses to elaborate on her epistemological theory rather than simply respond that of course her theory requires this thesis, as she repeatedly emphasized in her talk. Marcus' elaboration includes the epistemological point about having recourse to a dictionary that Kripke later incorrectly represented in 'Identity and Necessity' and 'Naming and Necessity' - misrepresentations which played a major part (perhaps the major part) in preventing philosophers from recognizing that Marcus possessed the theory of the a posteriori necessary. Soames' footnote 26 pertains to the following passage from the 1962 'Discussion' in Synthese, p. 142 (Soames quotes this passage from Marcus' Modalities, pp. 33-34, which contain a few minor stylistic changes that are irrelevant to its meaning). My emphases are added. Marcus says: Presumably, if a single object had more than one tag, there would be a way of finding out such as having recourse to a dictionary or some analogous inquiry, which would resolve the question as to whether the two tags denote the same thing. If "Evening Star" and "Morning Star" are considered to be two proper names for Venus, then finding out that they name the same thing as "Venus" names is different fromjinding out what is Venus'mass, or its orbit. It is perhaps admirably flexible, but also very confusing to obliterate the distinction between such linguistic and properly empirical procedures.

Note several things about this famous remark and Kripke's even more famous

MARCUS AND THE NEW THEORY OF REFERENCE

55

criticisms of it in 'Identity and Necessity' (1971, pp. 142-43) and 'Naming and Necessity' (1972, p. 305). First, Marcus implies that having recourse to a dictionary is merely one example of finding out that the same object has more than one proper name. This is the significance of the "such as" that I italicized and the phrase "or some analogous inquiry". An analogous inquiry, but different from having recourse to a book, would be to learn from conversation with other people that the same object has more than one name. Marcus is correct in making these claims. One way we do find out that the same object has more than one name is by finding more than one name listed for the object in a dictionary. But notice that Kripke's representation of Marcus' remarks is misleading when he criticizes it in 'Identity and Necessity' (1971, pp. 142-43) and 'Naming and Necessity' (1972, p. 305). He writes in 'Identity and Necessity' (pp. 14243): "You have an object a and an object b with names 'John' and 'Joe'. Then, according to Mrs. Marcus, a dictionary should be able to tell you whether or not 'John' and 'Joe' are names of the same object. Of course, I do not know what ideal dictionaries should do, but ordinary proper names do not seem to satisfy this requirement". Kripke here represents Marcus' claim - which is that having recourse to a dictionary is merely one example of a way of finding out how we may find out that the same object has more than one name - as the claim that having recourse to a dictionary is a necessary condition (a "requirement") of finding out they have more than one name. This misrepresentation is compounded by Kripke deliberately choosing an example ("Joe" and "John") where we do not learn from a dictionary whether they are co-referring. In such an example, we would find out they are co-referring by "some analogous inquiry", such as asking somebody if Joe is also called "John". Kripke's second major misrepresentation of Marcus' theory appears in his next sentence, which is presented as a criticism of Marcus. Kripke writes: "You certainly can, in the case of ordinary proper names, make quite empirical discoveries that, let's say, Hesperus is Phosphorus, though we thought otherwise" (1971, p. 143). This sentence has misled countless readers into thinking that Marcus did not have the doctrine of the necessary a posteriori. Note that it mis-states Marcus' view; she does not deny that we can "in the case of ordinary proper names, make quite empirical discoveries [such as that] Hesperus is Phosphorus". What she implies in her remarks is that it is a different sort of empirical discovery than discovering that Venus has a certain mass. Marcus correctly observes that 'If Evening Star" and 'Morning Star' are considered to be two proper names for Venus, then finding out that they name the same thing as 'Venus' names is different from finding out what is Venus' mass, or its orbit" ('Discussion' in Synthese 14(2/3), 1962, p. 142). Certainly, looking through a telescope at Venus or bouncing electromagnetic rays off of Venus is a different sort of procedure from learning that it is a linguistic convention that "Hesperus" and "Venus" are both used to name the same object. As Marcus continues, "It is perhaps admirably flexible, but also very

56

QUENTIN SMITH

confusing to obliterate the distinction between such linguistic and properly empirical procedures" (my emphasis). It is clear from the context that by a "properly empirical procedure" Marcus means a scientific observation or an observation of a physical thing. And surely she is correct that there is some sort of distinction between the procedure of observing the physical properties of physical things and the procedure of learning about the references of names. But this distinction is far from implying that we do not learn a posteriori ("empirically", in a sense that is broader than the so-called "proper procedural" sense Marcus distinguishes in the "Discussion") that "Venus" and "Hesperus" name the same thing. As Marcus says in her 1961 paper, "You may describe Venus as the evening star and I may describe Venus as the morning star, and we may both be surprised that as an empirical/act, the same thing is being described" [1961, p. 310, my emphasis). Thus, it is Kripke who errs in representing Marcus' theory, not Marcus who errs in stating in her theory, as Kripke alleges in the following passage from 'Identity and Necessity': "I thus agree with Quine, that 'Hesperus is Phosphorus' is (or can be) an empirical discovery; with Marcus, that it is necessary. Both Quine and Marcus, according to the present standpoint, err in identifying the epistemological and metaphysical issues" ('Identity and Necessity', p. 154, n. 13). But the error is Kripke's, for "the present standpoint", that" 'Hesperus is Phosphorus' is (or can be) an empirical discovery" and state a necessary truth, is Marcus' standpoint. Soames writes in his paper that "Smith maintains that Saul Kripke learned these doctrines from her, initially misunderstood them, and, when he later straightened things out, mistakenly took the doctrines to be his own" (p. 191). I maintain the present instance is a case in point. (In a forthcoming paper, 'Direct, Rigid Designation and A Posteriori Necessity', I argue there are two very different senses of "a posteriori necessity" that have been confused in the post-Marcus literature. One sense pertains to identifications and I discuss this sense in William Lane Craig and Quentin Smith, Theism, Atheism and Big Bang Cosmology [Clarendon Press, Oxford, 1993], pp. 184-185, and the other pertains to laws of nature and is discussed in Quentin Smith, Language and Time [Oxford University Press, New York, 1993, pp. 203-204] and 'The Metaphysical Necessity of Natural Laws', forthcoming.) {End 0/ added material. } I conclude that seven of the main ideas and arguments of the New Theory of Reference are indeed present in Marcus' 1961 paper and that Kripke merely elaborated upon them and gave them a new terminology. Is it scandalous that this conclusion should have been given credence by the APA committee that selected the papers for this meeting, as Soames suggests? I suggest that it is not, but instead should be viewed the way the APA committee viewed it, namely, as an important issue about the history of recent analytic philosophy that deserves discussion. This contribution to the study of the history of philosophy

MARCUS AND THE NEW THEORY OF REFERENCE

57

is no more scandalous than is McCracken's recent book Malebranche and British Philosophy, in which McCraken shows (on pages 257ff.) that many of the allegedly original ideas in Hume's A Treatise of Human Nature were taken, without due acknowledgement, from Malebranche, and that this historical fact about Hume and Malebranche has gone unrecognized for centuries. This is simply one of the ways in which contributions to the study of the history of philosophy are made. Our concern here is not attacks or defenses of certain personalities or the doling out of credit and blame to our friends or colleagues, as Soames characterizes this discussion. Rather, our concern in the history of philosophy is the same as our concern in other areas of philosophy, a concern to find out the truth. {Material added in May, 1995: 8. CONCLUDING REMARKS: FOOTNOTE 34 AND THE INCONSISTENCY OF SOAMES' PAPER

In the section on the modal argument, I added some material showing that Soames' new footnote 34 misrepresents the modal argument for direct reference. Here I would like to make an additional response to this new footnote, that it reveals an internal inconsistency in Soames' paper. Footnote 34 shows that Soames concedes my general thesis that Marcus had many of the ideas of the New Theory of Reference prior to Kripke, even though she is not credited with them. However, this concession is inconsistent with the conclusion Soames draws in his paper. Soames' conclusion is worth comparing with his added footnote. His conclusion is that providing Marcus with proper credit: does not result in a reassessment of the seminal role of Kripke and others as primary founders of contemporary nondescriptivist theories of reference. In fact, I think Smith does Kripke a grave injustice. Smith writes as if Kripke appropriated the major views expressed in Naming and Necessity from Marcus while denying her proper credit, and suggests that it is a scandal that the rest of the profession was thereby duped. We have not been duped; there was no misappropriation. Rather, what Smith has done is to mistakenly read many of Kripke's arguments and doctrines back into Marcus, and then to insinuate that Kripke is guilty of theft. This, it seems to me, is shameful. If there is any scandal here it is that such a carelessly and incompetently made accusation should have been given such credence (pp. 208-9).

There are several problems with this passage, even apart from its "colorful" language and suggestion that the APA program committee behaved scandalously. One problem is that Soames mistakenly represents the conclusion of my paper as the claim that "Kripke is guilty of theft". In fact, the conclusion of my paper was that Kripke unconsciously assimilated Marcus' ideas (as opposed to consciously assimilating them and deliberately not crediting them to her, which is what a "guilty of theft" charge would imply). The "guilty of theft" interpretation of Kripke's relation to Marcus is not supported by the available evidence, for if Kripke wanted to "steal and conceal" he would not have discussed or referred to her 1961 paper at all, whereas he discusses it in both

58

QUENTIN SMITH

'Identity and Necessity' and 'Naming and Necessity'. The conclusion of my paper was instead that: a reasonable explanation of why Kripke did not attribute the central features of the "New Theory" to Marcus is that he originally misunderstood Marcus' New Theory of Reference. When he eventually understood it, after a year or two, the insight that came made it seem that the ideas were new. I suspect that such instances occur fairly frequently in the history of thought and art. ('Marcus, Kripke, and the Origin of the New Theory of Reference', p. 188)

Although several philosophers have privately disagreed in print or conversation with my conclusion and have argued on the basis of the evidence I presented that Kripke was guilty of plagiarism (I will leave them to make their own case publicly), I believe my contrary conclusion, that Kripke unconsciously assimilated ideas that were originally misunderstood, is better supported by the evidence. In addition to the fact that Kripke discusses and refers to Marcus' paper (which he would not do if he wanted to hide his main source), there is evidence he originally misunderstood many of Marcus' ideas that he later espoused and elaborated upon in the early 1970s. I mentioned in my first paper Kripke's 1962 misconstrual of Marcus' theory that is revealed in his attribution to her of the view that "the tags are the 'essential' denoting phrases for individuals", but empirical descriptions are not, and thus we look to statements containing 'tags', not descriptions, to ascertain the essential properties of individuals" (see page 142 of the 'Discussion' in Synthese and page 34 in Marcus'Modalities). (In footnote 46 of his paper, Soames denies that Kripke is here misunderstanding Marcus and denies that Kripke is saying that tags are essentially denoting phrases of the individuals they denote; Soames puts a different and rather fantastic construction on Kripke's remark, so fantastic that I will simply refer the reader to Soames' footnote 46 and let the reader decide for herself if there is any plausibility to Soames' claim that this construction is what Kripke meant by his remark.) Other remarks made by Kripke also suggest he did not adequately understand Marcus' theory. For example, he has to ask Marcus: "Now, do you admit the notion of 'identically the same' at all?" ('Discussion', p. l36; Modalities, p. 28), whereas it should have been evident to anyone who understood Marcus' talk that most of her talk consisted of her arguing for and explaining her notion of 'identically the same'. Marcus has to respond to Kripke's question that "I admit identity on the level of individuals certainly" (,Discussion', p. l36; Modalities, pp. 28-29). And later Kripke has to ask about her theory: "does it not require that for any two names, 'A' and 'B', of individuals, 'A = B' should be necessary, if true at all?" (,Discussion', pp. 141-142; Modalities, p. 33), whereas it would have been plainly obvious to someone who fully comprehended her talk that this was one of her main points. Further evidence that Kripke did not adequately understand Marcus' 1962 talk is that he himself says (in the 1980 Preface to Naming and Necessity, p. 5)

MARCUS AND THE NEW THEORY OF REFERENCE

59

that "Eventually I came to realize - this realization inaugurated the aforementioned work of 1963-64 - that the received presuppositions against the necessity of identities between ordinary names were incorrect". If Kripke fully understood Marcus' theory in 1962, why does he not say that he first realized at Marcus' 1962 talk that the received presuppositions against the necessity of identities between ordinary names were incorrect? For these reasons, among others, it seems to me that the "plagiarism" explanation of Kripke's relation to Marcus is false, and that the "unconscious assimilation of originally misunderstood ideas" explanation of the relation is correct. But we have not yet touched on the main problem with the passage I quoted from Soames' concluding remarks, namely, this his conclusion is inconsistent with his new footnote 34 and with the main body of his paper. Soames concludes that I "mistakenly read many of Kripke's arguments and doctrines back into Marcus" and that my general thesis that "Kripke appropriated the major views expressed in Naming and Necessity from Marcus" is an incompetently and carelessly made accusation. However, in footnote 34 and in the main body of his paper, Soames expressly concedes that many of the points made in my paper about Marcus' priority to Kripke are true. It will be instructive to quote footnote 34 in its entirety and to insert some comments in brackets and italics: This is an important feature of Kripke's version of the modal argument that does not seem present in Marcus' version. [Soames here concedes my point that Marcus did have the modal argument, in some version, prior to Kripke, even though he mis-states the modal argument in the way I described in section 3.] Both Kripke and Marcus use modal properties of names to distinguish them from descriptions. [Soames here concedes my point that Marcus had the distinction between names and descriptions, and an argument for this distinction, prior to Kripke.] However, the way they establish the modal properties of names is quite different. Kripke provides examples in which we evaluate simple sentences containing proper names (but no modal opeators) at alternative possible worlds (counterfactual circumstances). He uses these examples to show that the referent of a name at one world is the same as its referent at other worlds, and hence that names are rigid. It then follows that identities involving names are necessary if true, and that coreferential names are intersubstitutable in modal constructions. Marcus, on the other hand, seems to think that these modal properties of names somehow follow from the quantified version of the necessity of identity, either by itself or in conjunction with her controversial view that names are "mere tags". [Soames here concedes my four points that Marcus possessed before Kripke (1) the theory that coreferential names have the modal property of being intersubstitutable in modal constructions, (2) the theory that identities involving names are necessary if true, (3) the theory that names are rigid, (4) the theory that the referent of a name at one world is the same as its referent at another world.]

This new footnote is consonant with the main body of Soames' paper, in which these and other of my points about Marcus' priority are also conceded. For example, on page 197 Soames expressly concedes that in her 1961 paper Marcus makes the point "that (i) coreferential names are substitutable in modal contexts without change in truth value whereas descriptions are not, (ii) identity statements involving names are necessary if true, (iii) names are not equivalent to descriptions, (iv) descriptions are not genuine singular terms, (v) proper names have no meanings over and above their referents, and (vi) coreferential names have the same semantic content ...".

60

QUENTIN SMITH

Soames also concedes on page 196 that Marcus "can be credited with an implicit recognition that genuine names are rigid designators". (But Soames oddly seems to think that my quotation from Marcus' 1970 talk is meant to support my claim that Marcus had this notion before 1963-64, the years Kripke admitted he first came to understand this notion. See Soames, p. 203). Further, on page 200 Soames concedes both that Marcus had a "version of 'the modal argument' " and that for this reason she (and Smullyan and Fitch) "deserve a degree of recognition that they are often not given". Soames also allows that "Marcus' quantified version of the necessity of identity anticipates the contemporary view that variables are rigid designators" (p. 208). How are all these concessions consistent with Soames' conclusion that providing Marcus "with proper credit does NOT result in a reassessment ofthe seminal role of Kripke and others as primary founders of contemporary non-descriptivist theories of reference" (p. 208, my emphases)? If we eliminate the ideas that Soames concedes are present in Marcus' early writings from the contemporary non-descriptivist theories of reference, we do not even have a non-descriptivist theory of reference, let alone several philosophers who deserve instead of Marcus the title of a "primary founder" of contemporary non-descriptivist theories of reference. If Soames is to draw a conclusion that is consistent with footnote 34 and the main body of his paper, he should have concluded that the evidence I presented established that Marcus was a primary founder of the New Theory of Reference, even though Soames does not agree with all of the points I made about her primacy. This conclusion would at least enable Soames' paper to have the virtue of self-consistency, even if it would not thereby acquire the virtue of being entirely accurate. 2 Philosophy Department Western Michigan University

NOTES {May, 1995 addition: At the APA meeting, I crossed out my original version of the paragraph beginning "In my paper, I gave a different illustration ..." and did not read this paragraph at the APA colloquium, since the second half of the original version of this paragraph needed revision. I have now omitted the problematic parts of this paragraph and added three new sentences (beginning with "He correctly states [on page 206] that the modally stable sense ...".} 2 {May, 1995 addition: I am very much indebted to Arthur Falk for helpful and extensive email correspondence between December 16 and December 25,1994 about the origins of the New Theory of Reference. Arthur Falk's insights enabled the reply to Soames I read at the APA meeting to be much better than it would otherwise have been. I am indebted to William Valli cella for the reference to the Malebranche/Hume relation and to Robert Fogelin for calling my attention to an instructive analogy between the Marcus/Kripke relation regarding the New Theory of Reference and the Fogelin/Kripke relation regarding the Skeptical Paradox Interpretation of Wittgenstein's private language argument. See pages 241-246 of the second edition of Robert Foge1in's Wittgenstein (Routledge and Kegan Paul, London, 1987).}

MARCUS AND THE NEW THEORY OF REFERENCE

61

REFERENCES De Finetti: Fundamenta Mathematica, Vol. 17, pp. 298-329. Fitch, E: 1950, 'Attribute and Class', in Farber (ed.), Philosophical Thought in France and United States, Buffalo. Fitch, F: 1949, 'The Problem of the Morning Star and the Evening Star', Philosophy of Science, 16, 137-14l. Kanger, S.: 1957, Provability in Logic, Stockholm. Marcus, R.B. et al.: 1962, 'Discussion on the Paper of Ruth B. Marcus', Synthese, 14,132-143. McCracken, C.J.: 1983, Malebranche and British Philosophy, Oxford University Press. Ramsey, FP.: 1931, 'Truth and Probability', Foundations of Mathematics. Routledge and Kegan Paul. Quine, w.V.O.: 1949, 'Designation and Existence', in Feigl, H. and Sellars, W. (eds), Readings in Philosophical Analysis (New York: Appleton-Century-Crofts, Inc.), Reprinted from The Journal of Philosophy 36, 701-709. Smith, Q.: 1993, Language and Time. Oxford University Press, New York. Smith, Q.: 1995, 'Marcus, Kripke and the Origin of the New Theory of Reference', Synthese, 104, 179-189. Smith, Q.: (forthcoming), 'The Metaphysical Necessity of Natural Laws'. Smith, Q. and W.L. Craig.: 1993, Theism, Atheism and Big Bang Cosmology. Clarendon Press, Oxford. Soames, S.: 1995, 'Revisionism about Reference: A Reply to Smith'. Synthese, 104,191-216. Smullyan, A.: 1947, Review of Quine's 'The Problem of Interpreting Modal Logic', The Journal of Symbolic Logic, 12, 139-141. Smullyan, A.: 1948, 'Modality and Description', The Journal of Symbolic Logic, 13, 31-37.

PART II REPLIES

SCOTT SOAMES

MORE REVISIONISM ABOUT REFERENCE

At the Eastern Division Meetings of the American Philosophical Association in December of 1994, I replied to Quentin Smith's paper "Marcus, Kripke, and the Origin of the New Theory of Reference". After my comments, Smith was allowed to read a second paper, responding to me, which I had not had the opportunity to see. The three APA papers were later published in Synthese, with the result that Smith's comments on my paper continued to stand unchallenged. I take the opportunity to respond to them now. Since it would take volumes to respond to all of Smith's errors and misrepresentations, I will be selective, and concentrate on a few representative examples of his lax standards of scholarship and most blatant historical inaccuracies. I will take up substantive philosophical issues, beyond these discussed in my original reply, only where Smith's discussion might lead others astray, and where correction may produce some general enlightenment. Finally, some important issues, like Smith's treatment of the necessary aposteriori, and his interpretation of Marcus' famous "dictionary remark", will be left to the companion paper, "How Not to Write History of Philosophy: A Case Study" by my colleague, John Burgess.' I. QUOTATION

I begin with Smith's attempt to deflect criticism of his misleading use of quotation. On page 37* of his reply to me, he says the following: " ... I should note that Soames does not quote Marcus' original 1961 paper, which is the paper in dispute, but her revision of this paper, published in Modalities (1993), which differs at a number of places from the original 1961 article. I quoted only the 1961 article, but Soames, when he sometimes tries to show that my quotations are misleading or incomplete, does not in fact quote the 1961 article, but Marcus' 1993 revision of this article. Since this is an historical inquiry into the relation of Marcus' 1961 paper to Kripke's works of 1971 and 1972, it is her 1961 paper that is the proper object of examination." [my emphasis]

I grant Smith the point about the (1961) text being the definitive version. 2 However, I want to draw attention to two issues regarding the use he makes of this point. First, his remark suggests that my criticism of his misleading and "[Editors' note: Pagination references have been changed to refer to the reprinted paper in this volume.]

65 P. W. Humphreys et al. (eds.), The New Theory of Reference © Kluwer Academic Publishers 1998

66

SCOTT SOAMES

incomplete quotations from Marcus rest on my use of the (1993) text, and would lose its force if the (1961) text were consulted. This is false. There were four places where I called attention to the ways in which Smith's quotations differ from the corresponding passages in Marcus (1993): one of the Marcus passages is word for word identical in the (1961) and (1993) texts; one passage is identical in the two texts except for an insertion of the word 'that' to introduce a complement clause in the (1993) version; and in the remaining two cases the (1961) and (1993) texts differ only stylistically by a pair of philosophically irrelevant words, e.g., "to assign a thing a name" vs. "to give a thing a name". (See the appendix to this paper for details.) It is hard to understand how anyone who had actually compared the two versions of these four passages could have failed to recognize that the differences between the two versions are tiny and that nothing in my criticism of Smith's quotations is affected by them. Moreover, it does not seem too much to expect that anyone operating in good faith would have made such a comparison before making an issue of my quoting the reprint rather than the original. One can draw one's own conclusion about where this leaves Smith. Second, despite his self-righteous insistence on the principle that one should always quote from the original, Smith himself does not always do so. On page 6 of his original paper he gives the following quotation from Marcus. "Let us now return to (10) and (15). If they express a true identity, then 'Scott' ought to be anywhere intersubstitutable for 'the author of Waverly, in modal contexts, and similarly for 'the morning star' and 'the evening star'. If they are not so universally intersubstitutable - that is, if our decision is that they are not simply proper names for the same thing; that they express an equivalence which is possibly false, e.g., someone else might have written Waverly, the star first seen in the evening might have been different from the star first seen in the morning - then they are not identities. (1961 :311)" [my emphasis]

Although Smith cites page 311 of the (1961) text as the source of this passage, the (1961) text does not contain the words "in modal contexts". These words were added by Marcus and appear only in the (1993) text. 3 A further twist is added by the fact that there are a number of other inessential grammatical differences between the (1961) and (1993) versions of the passage. The material in Smith's quotation is an amalgam of the two. It is word for word identical with the (1961) passage, except for the insertion of the words 'in modal contexts', which come from the (1993) version. Put simply, this quote is manufactured. 4 Nor is this matter insignificant. It is clear from Marcus' paper that she regarded coreferential proper names as logically equivalent, and even synonymous. Hence, she thought they could be substituted for one another without change in truth value in any linguistic construction, including epistemic and propositional attitude constructions, (except for metalinguistic constructions about the names themselves) without change in truth value. But if that was her view, then, as I argued in my response, the grounds for her claims about the behavior of names in modal contexts were much different from, and more

MORE REVISIONISM ABOUT REFERENCE

67

questionable than, those of Kripke. The original passage from Marcus, without the insertion of the phrase "in modal contexts", supports this interpretation which is why Smith's inclusion of those words is so telling. 2. SMULLYAN AND FITCH

In my original reply I contested Smith's claim that Marcus introduced the idea that proper names are directly referential, and are not equivalent to descriptions. As Smith notes in his reply, I said: "Although Marcus did regard names as directly referential, and did not take them to be disguised descriptions, she did not introduce the idea; nor did she claim to. As we have seen, Smullyan and Fitch both invoked that idea in earlier responses to Quine that Marcus took herself to be repeating and elaborating."

Smith objects to this, claiming that Smullyan held no such view, and maintaining that Fitch got the idea from Marcus. Before showing that this is false, I want to remind the reader that the ultimate sources of these ideas were, of course, John Stuart Mill and, in our century, Bertrand Russell (in his doctrine ofiogically proper names). Smullyan, Fitch, and Marcus (as well as Arthur Prior) were all strongly influenced by Russell. They took themselves to be following important doctrines of his involving the distinction between names and descriptions, while dropping his requirement that logically proper names can refer only to objects of immediate, privileged, acquaintance (without responding to the substantive concerns that led Russell to this requirement). In particular, all of them believed that Quine'S argument against the intelligibility of modal logic, based on failure of substitution in modal contexts, could be answered by invoking the Russellian distinction between genuine names, on the one hand, and definite descriptions (which were not regarded as singular terms), on the other. The historical question at issue involves the priority of this response to Quine, and of the conception of names involved in that response. As I pointed out in my original reply, the brief discussion of names in Marcus (1961) occurs entirely within the context of replying to Quine's argument. After giving a rough sketch of the argument she begins her reply with an explicit disavowal that she is introducing new material. She says "The rebuttals are also familiar. Rather than tedious repetition, I will try to restate them more persuasively". 5 As it happens, she had dealt with the same question in a closely related paper, 'Extensionality" that appeared in Mind, in 1960. 6 Her response to Quine's argument is found in the antepenultimate paragraph of that paper, which begins: "The problem of the morning star and the evening star is resolved in an analogous way. For ... (12) The evening star equals the morning star is not unambiguous. If (12) involves proper names of individuals then 'the evening star' may replace

68

SCOTI SOAMES

'the morning star' without paradox in: (13) It is necessary that the evening star is the evening star

As I pointed out in my original reply, Marcus attributes the argument in this passage to Fitch. The attribution appears as a footnote, which is attached to the first sentence of the paragraph just quoted. It reads as follows: "The paragraph which follows restates a point made by EB. Fitch in [Fitch (1949))."

This point is indeed made in Fitch (1949). However, in footnote 4 of that paper Fitch notes his agreement with Smullyan, and in the main text of the paper he says: "Smullyan has pointed out that if the phrases (1) ["The Morning Star"] and (2) ["The Evening Star") are regarded as proper names of the same individual, then Quine's argument fails because (4) ["It is not necessary that the Evening Star is identical with the Morning Star"] would clearly be false.,,7

The relevant papers by Smullyan are (1947, 1948).8 Though the latter is better known, the former is more relevant. It is a short review of Quine'S original presentation, which follows it in the same year and the same volume of The Journal of Symbolic Logic. I don't believe that there is any other published response to Quine'S argument for which one could possibly claim priority. Despite this, in his reply to me Smith denies that in these papers Smullyan held that names are directly referential, and not equivalent to definite descriptions. Smith gives two reasons for this denial. His first reason is that Smullyan's discussion is allegedly too brief to indicate that he is ruling out the possibility that names may be disguised descriptions. This is false. As I pointed out in footnote 10 of my original reply, the basic idea behind Quine'S argument (extraneous complications aside) was that (i) existential generalization on a constant is a valid, truth-preserving, rule of inference, but (ii) this conflicts with the fact that advocates of modality must acknowledge that for some coreferential constants a and b, both a = b & [J a = a and b = b & ~ [J a = b are true. Together (i) and (ii) lead to the impossible conclusion that3xlx = b & [Ja = xl and 3 xlx = b & ~ [J a = xl are jointly true. (Bold face italics are used as corner quotes.) Smullyan's response to this argument was that if "by 'constant' is meant what is commonly understood by 'proper name"',9 then although (i) is true, (ii) is false, since when a and bare coreferential proper names, ~ [J a = b will be false (and hence [J a = b will be true); whereas if definite descriptions are allowed as constants, then (i) will be false, since existential generalization on singular descriptive phrases (out of modal constructions) will fail. 10 Evidently, there is in this a clear distinction, revealed by their different behavior in modal contexts, between definite descriptions and "what is commonly understood by 'proper name"'. The necessity of identity for names, but not descriptions, is implicitly endorsed. In addition, names, unlike descriptions, are implicitly treated by Smullyan as rigid designators, since from

MORE REVISIONISM ABOUT REFERENCE

69

the trivial truth a a = a one can derive 3x[O a = xl (which on the ordinary understanding can be true only if a refers to the same thing in every world). When Smullyan adds that "if 'Evening Star' and 'Morning Star' proper-name the same individual they are synonymous ..." this strongly suggests, as I noted in my original reply, that genuine names have no meaning apart from their referents. Hence it is reasonable to conclude that Smullyan has the main elements of the view of names advocated in Marcus (1961). II Smith's second reason for denying credit to Smullyan is that Smullyan says that in Quine's example "Morning Star" and "Evening Star" are more naturally viewed as abbreviations of descriptive phrases than as proper names. But this is beside the point. Smullyan's general thesis about the difference between the categories of names and descriptions is independent of his judgment about the naturalness of viewing a particular pair of expressions as used in a particular example as members of one category rather than the other. This is especially so since everyone from Smullyan to Marcus holds the example to be ambiguous. It is precisely because Marcus was not ready to categorically characterize "the morning star" and "the evening star" as names that her restatements of the Smullyan-Fitch response to Quine's argument are persistently hypothetical (as witnessed by the 'if -clause in the passage cited above from "Extensionality".) All of these points about Smullyan's response to Quine are obvious, and I am surprised that they should have to be repeated. It is worth noting in this connection that my understanding of Smullyan is absolutely standard, and was shared by participants in the debate, as well as by other leading figures in the development of modal logic. As already indicated, Marcus acknowledged that her reply to Quine restated the position of Fitch (1949), which attributed that very position to Smullyan. Nor was Fitch alone in this. As I noted in my original reply, Church's 1950 review of Fitch (1949) characterized him as agreeing with Smullyan. Additional confirmation comes from another pioneer in the development and philosophical application of intensional logics. In his (1967) article on modal logic in the Encyclopedia of Philosophy, Arthur Prior gives full credit to Marcus' contribution to quantified modal logic, and notes in particular the quantified version of the necessity of identity.12 Prior also cites Quine's objection, and credits Smullyan (not Marcus) with the response to Quine just outlined, including the distinction between definite descriptions, on the one hand, and expressions that "directly name" and "simply tag" objects, on the other. He also provides needed historical context by indicating the origin of these ideas in Russell's conception of logically proper names. I3 It is astounding to suggest that all of these contemporaries of one another should be fundamentally wrong in their understanding of Smullyan. Thus it is not surprising that when one carefully reviews Smullyan's texts, one finds that he did hold the views attributed to him. We now turn to Smith's treatment of Fitch, where he takes an entirely different approach. Here, Smith does not deny that Fitch (1949) contains the main points about the differences between names and descriptions, including

70

SCOTI SOAMES

their different behavior in modal contexts. Instead, he claims that Fitch learned these points from private communications with Marcus five years earlier, when she was his doctoral student. Smith does not claim that Fitch, too, unconsciously plagiarized Marcus; rather he claims that Fitch explicitly attributes the points to Marcus. On the face of it, this seems unlikely, since we have already seen that Marcus' first treatment of the issue attributes the points to Fitch (1949), which in turn contains a direct attribution of them to Smullyan - an attribution that Smith neglects to quote. Smith does, of course, quote other things from Fitch in an attempt to show that he really attributed the main points of his reply to Quine - i.e., the main points of Fitch (1949) - to Marcus. Examination of these texts provides another instructive lesson in Smith's technique. There are two relevant papers here - Fitch (1949), and Fitch (1950). Smith quotes a sentence from footnote 6 of Fitch (1949). In that footnote Fitch first refers back to footnote 3, then lists Marcus' three early technical papers on her formal system of quantified modal logic, and then includes a general, nonspecific acknowledgment to Barcan-Marcus for helpful conversations. This sentence, the second to last of the footnote, is quoted by Smith (on page 39). "The writer wishes to acknowledge his debt to Miss Barcan for some helpful discussions concerning the ideas of this paper."

Obviously, this is simply a polite thanks that implies nothing specific about anything that should be attributed to Marcus. However, the very next (and last) sentence of the footnote, which Smith chooses not to quote, provides further specificity: "In particular, she pointed out the importance of the equivalence of the two kinds of identity defined in her third paper."

The issue involving "the equivalence of the two kinds of identity" is a very minor technical one (about which Quine had become confused), concerning the difference between systems of quantified modal logic based on the modal system S2, and those based on S4. Fitch relegates discussion of this issue to footnote 3 (the one to which the footnote presently under discussion begins by referring). Clearly, there is nothing here to support Smith's contention that Fitch acknowledged Marcus as the source of the main point of his paper. Smith places more weight on a passage from Fitch (1950), "Attribute and Class". 14 "The system of modal logic developed by Ruth Barcan suggests that the simplest view is that no identities should be regarded as merely contingent and that identified entities should be everywhere intersubstitutable. (Indeed, no entity is correctly identifiable with any entity but itself, so permission of substitution of this sort is trivial anyway.) Furthermore, if entities X and Y have been identified with each other, it seems reasonable to suppose that the names of X and Y should also be everywhere intersubstitutable where they are being used as names. According to Church's view on the other hand two names of the same thing might differ in sense and so not be intersubstitutable." (The italicised material is omitted and marked by an ellipsis in Smith's quote.)

MORE REVISIONISM ABOUT REFERENCE

71

Smith takes this passage to show that "Fitch explicitly attributes the idea that names and descriptions behave differently in modal and other contexts to his doctoral student Ruth Barcan Marcus (her name was then Ruth Barcan)".15 The passage shows no such thing. First, it does not say that Marcus personally suggested anything to Fitch. It says that her early formal system suggested something. The exact words used by Fitch need mean no more than that something occurred to him which he judged to be in the general spirit of Marcus' formal system. Because his words can perfectly well be taken in this way, they would not have constituted an adequate acknowledgment if there had been an actual suggestion from Marcus personally. Second, what is said to have been suggested involves the claim that "identified entities should be everywhere intersubstitutable", which clearly involves a confusion of use and mention, since it is expressions, not the entities denoted by them, that are candidates for substitution. The confusion continues in the parenthetical sentence omitted by Smith, where the confused thing said to be suggested by Marcus' system is characterized as trivial. Third, in the next sentence of the passage Fitch indicates what he seems to take to be a further idea - namely that it is reasonable to suppose that names of the same entity should be everywhere intersubstitutable. So what we have is Fitch's claim (i) that Marcus' formal system suggests something (which he gives a confused account of and apparently regards as trivial), and (ii) that further reflection suggests the reasonable (but apparently non-trivial) claim that coreferential names are everywhere intersubstitutable. It must be emphasized here that the formal system of Marcus' early papers has no consequences whatever regarding the properties of names vs. descriptions; so if Fitch thought that it did suggest something about this (as I expect he did), then he was wrong. As I pointed out in my original reply, the language of that system contains no names, no individual constants, no function signs, no descriptive phrases; its only singular terms are individual variables. The system leaves entirely unanswered the question of which, if any, of the different categories of possible singular terms - individual constants, ordinary proper names, singular definite descriptions, functional expressions f( c), g( a, b), etc. should be allowed as instantial terms, to be intersubstituted with variables in rules of inference like existential generalization and universal instantiation, should the language be enriched to allowed such terms. 16 This is a point I made in my original reply, but which Smith has now muddied further in his second paper. Since there is a significant philosophical lesson here that may be of some general interest, I will return to it briefly below. For now it is enough to note that there is no way that the formal system of Marcus' early papers could have significant consequences about ordinary names and descriptions in natural language. The final point to be made about Fitch (1950) is that, as usual, Smith's quotation from it is misleading and incomplete. In the paragraph immediately following the passage cited by Smith, Fitch explicitly attributes the crucial

72

SCOTT SOAMES

distinction between names and descriptions, and the use of this distinction in responding to Quine's argument, not to Marcus, but to Smullyan. Here is the passage omitted by Smith. "This discussion of identity is well illustrated by the problem about the Morning Star and the Evening Star. Let us grant that the Morning Star is identical with the Evening Star, so that they have all the same attributes. Since the Morning Star is necessarily identical with the Morning Star, it must be the case that the Evening Star also has the attribute of being necessarily identical with the Morning Star. Hence we cannot assume that they merely happen to be identical without concluding that they must necessarily be identical, in the sense that their identity is a truth of logic rather than merely of astronomy. Quine regards this problem as indicating that modal logic must be rejected or at least severely restricted. But A.F. Smullyan has shown that there is no real difficulty ifthe phrase 'the Morning Star' and 'the Evening Star' are regarded either as proper names or as descriptive phrases in Russell's sense. His argument is essentially the same as that given above in discussing the phrase, "the author of Waverly". 17

The argument to which Fitch refers occurs on pp. 551-552, and is simply his restatement of Russell's treatment in "On Denoting" 18 of the example 'Scott is the author of Waverly': if both 'Scott' and 'the author of Waverly' are regarded as names, then the sentence is held to mean the same as 'Scott is Scott', and is characterized as a "truth of logic". On the other hand if 'the author of Waverly' is analyzed in accord with Russell's theory of descriptions, then the statement is not a genuine identity statement, "and so can be true without being true by logical necessity".

Fitch was right to characterize Smullyan's reply to Quine as essentially an application of familiar Russellian ideas about logically proper names vs. descriptions to a modern logical context. This response to Quine, first given by Smullyan, and later repeated by Fitch, and then Marcus, was a straightforward application of Russell, without consideration of the epistemic considerations that led him to require the referents of logically proper names to be objects of direct acquaintance. As a result, there was very little new ground broken. It is no wonder that Marcus should introduce the crucial discussion in her 1961 paper by referring to familial rebuttals, and their tedious repetition. The rebuttal was, by that time, very familiar - which is not to say that no important purpose was served by the vigorous revival and application of Russell. On the contrary, the points made were well taken, and they had a salutary effect. But priority for this response to Quine, with the accompanying revival of Russell's doctrines about descriptions vs. genuine names, belongs to Smullyan, not Marcus. Finally, it is important to appreciate that the lasting value of this bundle of doctrines, as applied to modal contexts, depends on the relevant modality being metaphysical necessity, in the sense of Kripke (1972), and not "analyticity", "logical necessity", "logical validity", "tautology" or "logical truth" - as disputants from Quine to Marcus habitually characterized it. Identity sentences involving coreferential names (whether they be ordinary proper names or individual constants in a logical system) are not analytic, logically necessary, logically valid, or logically true. 19 In addition, the argument in Kripke (1972)

MORE REVISIONISM ABOUT REFERENCE

73

for why identity sentences involving ordinary proper names are metaphysically necessary, if true, depends on considerations specific to that kind of necessity considerations which are absent from the relevant texts of Smullyan, Fitch, and Marcus. Certainly, Kripke's argument does not turn on any principle about names being "everywhere intersubstitutable where being used as names", which is something that is implicitly rejected in Naming and Necessity.20 So, by all means, let us recognize the value of the Smullyan response to Quine, and its later repetition by Fitch and Marcus; but let us not confuse it with the seminal advance of Kripke (1972). 3. THE NECESSITY OF IDENTITY, NAMES, AND RULES OF INFERENCE FOR THE QUANTIFIERS

In the previous section I repeated the point that, contrary to Smith, the formal system of Marcus (1947),21 which includes a proof of the necessity of identity (NI-l) (x) (y) (ifx = y, then 0 x = y), does not provide the basis for any thesis about names, and in particular does not provide the basis for (NI-2) For all proper names a and b, if a

= b is true, then Oa = b is true.

In his reply to me Smith disputes this, saying that " ... it can be plausibly argued that there is a "direct route" from (NI -I) to (NI -2) in the sense that the true thesis (NI-2) can be logically derived from the true thesis (NI-I) in conjunction with some plausible supplementary premises about proper names and variables (e.g., that proper names are formally equivalent to variables in some contexts) ..." (p. 46)

What Smith has in mind by the supplementary premises involving the ''formal equivalence between variables and proper names" is indicated in the following remark, in which he quotes from Quine. "There arguably is a formal equivalence between variables and proper names, an equivalence that has been recognized as early as 1939 with Quine's article, 'Designation and Existence' (Journal of Philosophy, 1939), where Quine writes: " ... we might equivalently omit express mention of existential generalization and describe names simply as those constant expressions which replace variables and are replaced by variables according to the usual logical laws of quantification."" (p. 45)

I take it that the idea here is this: The existential and universal quantifiers in modal and other logical languages are subject to the logical laws of existential generalization and universal instantiation. Where t is a proper instantial term, existential generalization is a universally truth-preserving rule that allows one to derive 3x (. .. x .. .) from (.'0 t.. .) and universal instantiation is a universally truth-preserving rule that allows one to derive ( ... t.. .) from (x) ( ... x .. .). The claim that proper names are proper instantial terms for existential generalization and universal instantiation is a standard, widely recognized, and

74

SCOTI SOAMES

independently plausible assumption. (See, for example, Quine (1939).) When this supplementary assumption is added to (NI-l), the resulting pair entails (NI-2). Since (NI-l) was a theorem of Marcus (1947), and since that theorem together with a widely accepted and independently plausible supplementary premise entails (NI-2), Marcus' formal system did in fact provide the basis for establishing the necessity of identity for proper names. Although this line of reasoning can be seductive, it is mistaken in subtle ways. Since the errors are important, instructive and I fear, widespread, I will try to correct them. The first point to remember is that whether the familiar rules for (objectual) quantifiers are truth preserving depends on the resources of the language, and in particular on the semantics of the structures quantified into. In an extensional language, existential generalization and universal instantiation are guaranteed to be truth preserving so long as the instantial terms are required to designate some individual. When one considers different kinds of intensional languages further restrictions on instantial terms are needed, and in some cases the constructions quantified into preclude any class of constant or logically compound singular terms from playing the role of instantial terms. In these languages, existential generalization and universal instantiation involving constant or compound singular terms are simply not valid rules of inference. A simplified example of this can be constructed if we let L be a standard first-order language, containing names as constant singular terms, to which the I-place formula-forming operator '0' has been added. To keep things simple we allow only formulas in which '0' is prefixed to formulas that do not already contain it. Thus there are no iterations to worry about, and no occurrence of the operator is ever within the scope of another. Intuitively, '0' is to be understood as "it is a logical truth that". More precisely, when '0' is prefixed to a (closed) sentence S the resulting sentence, 0 S, is true iff the sentence S is a logical truth (i.e., is true in all its interpretations); when '0' is prefixed to an open formula F in which certain variables occur free, the resulting formula, OF, is true relative to an assignment A of values to variables iff F is true relative to A in all interpretations of F. What about cases in which A assigns a variable free in F some object that doesn't exist in a particular interpretation I? Different decisions are possible here, but for our purposes we may stipulate the following: x = y is true in I relative to A iff the denotation of 'x' relative to A (i.e., the value A assigns to 'x') is the same as the denotation of 'y' relative to A (whether or not the object is a member of the domain of I); since the extensions of all other, non-logical, predicates in I are sets of n-tuples from the domain of I, atomic formulas not involving the identity predicate are true in I relative to A only if the denotations of all the relevant terms, relative to A, are members of the domain of 1. 22 In such a language (NI-I) is true, even though for any pair of distinct names a and b, 0 a = b is false, since there will always be interpretations in which the names are assigned different things. Therefore, (NI-2) is false statement about L. In

MORE REVISIONISM ABOUT REFERENCE

75

this language existential generalization and universal instantiation involving names are not universally truth preserving, and so do not qualify as rules of inference. 23 The point illustrated here is that when one is dealing with a new intensional operator that allows quantifying-in, one cannot simply assume without argument that the usual rules of inference governing the quantifiers in familar extensional systems will continue to apply. Thus, even though names are standardly chosen as instantial terms for these rules in extensional systems, one has no justification for extending this practice to a system containing a new intensional operator until one has explained the meaning of the operator, and shown that, with that meaning, the selection of names to play the role of instantial terms will guarantee that the rules are truth preserving. As we have seen, when '0' is given an interpretation corresponding to the informal characterizations found in the writings of Fitch, Marcus, and others, neither ordinary proper names, nor the constants in formal logical systems, can play that role. 24 Suppose, however, that we distinguish logical necessity (in the sense of standard logical truth) from the now familiar notion of metaphysical necessity, and interpret '0' as standing for the latter. Next we ask a question. What requirement must a class of singular terms satisfy if members of the class are to qualify as instantial terms in universally truth-preserving rules of existential generalization and universal instantiation? It will be sufficient to consider existential generalization (involving the standard objectual quantifier). The basic idea can be expressed as follows: Let a (... t ••.) be an arbitrary truth. Given the meaning of the modal operator, this will be true iff for every metaphysically possible world w, the referent of t in w satisfies ( ... x ...) in w. Now consider the sentence, 3x a (... x ...), which would be gotten by existential generalization. This sentence will be true iff there is some one object that satisifes ( ... x .•.) in every metaphysically possible world. Looking at these two truth conditions, one can easily see what requirement on t will guarantee that if the first of the truth conditions is satisfied, then the second must also be satisfied. There must be some one object which is the referent of t in all worlds - i.e., t must be a rigid designator. In general, when we interpret the modal operator in our language L as expressing metaphysical necessity, we can be sure that if there is a nonempty class of instantial terms for truth-preserving rules of existential generalization and universal instantiation, its members must be rigid designators. Moreover, because they are rigid, we know that for any two members t, and t2 of that class, if t1 = t2 is true, then Ot1 = t2 is true. Notice that what we have here is a generalized version - call it "(NI-2G)" - of (NI-2). (NI-2G)

For all members a and b of the class of proper instantial terms for truth preserving rules of existential generalization and universal instantiation, if a = b is true, then 0 a = b is true.

76

SCOTT SOAMES

The same point can be established in another way, using one part of the argument of Quine (1947).25 Let I be the class of instantial terms for a universally truth-preserving rule of existential generalization, and let a and b be any two coreferential members of that class - i.e., any two members such that a = b is true. It follows from the universally truth-preserving character of existential generalization that substitution of a for b in any sentence must preserve truth value. For suppose this were not so - i.e., suppose that some sentence, ( ... a••• j, were true whereas the corresponding sentence, ( ... b•..j, was false. In such a situation, both the sentence fa = b & ( •.• a••• )] and the sentence [(b = b & ~ ( •.• b...)] would be true. But then if existential generalization is universally truth preserving,3x(x = b & ( ... x ...)] and3x (x = b & ~ ( •.. x ...)] would both be true, which is impossible. Thus, for any two members a and b of I, if a = b is true, then a = a and a = b cannot differ in truth value. Since we know the former is true, the latter must be as well. The significance of this for my reply to Smith is obvious. In establishing (NI2G), what we have shown is that (NI-2) follows from the claim that proper names are the instantial terms for universally truth-preserving rules of existential generalization and universal instantiation in a modal system in which '0' stands for metaphysical necessity.26 But this claim is nothing other than the supplementary premise that Smith proposed to add to Marcus' (NI-I) so as to derive (NI-2). Of course, since (NI-2) follows from the supplementary premise alone, it follows from that premise plus (NI-l). Clearly, however, (NII) is irrelevant to this derivation. For this reason, it does not provide any basis for (NI-2). One cannot argue that Marcus' (1947) quantified version of the necessity of identity provided the basis for the necessity of identity involving proper names by pointing out that the conclusion about names could be drawn if only one adds to the Marcus theorem a supplementary premise which entails the conclusion all by itself. (Ditto for Smith's claim that the 1947 theorem provided the basis for "the modal argument".)27 Nor does it help to claim, as Smith does, that the important point is that "Marcus herself applied the theorem [(NI-1) J of the necessity of identity to proper names in her 1961 paper. She takes great pains to make this application explicit in propositions (13) through (18) ... " (p. 226). The important point is that there is no such thing as applying (NI-l) to proper names, in the sense of using it as a basis for deriving (NI-2). Any such derivation is bound to be question-begging in the sense of being forced to assume what is supposed to be proved. As I pointed out in my original reply, the way we establish the necessity of identity for proper names is not be deriving it from the quantified version, but rather, as Kripke did, by first establishing directly that proper names are rigid, and then deriving the necessity of identity for coreferential proper names as an immediate consequence. 28 Of course, this does not mean that Smith is wrong in characterizing Marcus as thinking that her theses about names are the direct results of her (1947) theorem. But if she did think this, then she, too, was confused. That she was

a

a

MORE REVISIONISM ABOUT REFERENCE

77

confused is indicated by the following passage from Marcus (1961). "Consider the claim that (13)alb is a true identity. Now if (13) is such a true identity, then a and b are the same thing. It doesn't say that a and b are two things which happen, through some accident, to be one. True, we are using two different names for that same thing, but we must be careful about use and mention. If, then, (13) is true, it must say the same thing as (14) a I a But (14) is surely a tautology, and so (13) must surely be a tautology as well. This is precisely the import of my theorem (8) [the (1947) quantified version of the necessity of identity]. We would therefore expect, indeed it would be a consequence of the truth of (13), that 'a' is replaceable by 'b' in any context except those which are about the names 'a' and 'b'. (my emphasis, p. 308).

4. WHAT IS "THE" MODAL ARGUMENT?

In his original paper Smith claims that Marcus introduced "the famous modal argument for the thesis that proper names are directly referential rather than disguised contingent descriptions." (p. 6). In reply to Smith I said that "Marcus did hold that names and descriptions typically behave differently in modal constructions, and so cannot be regarded as equivalent. Moreover, it is reasonable to take this as a version of "the modal argument" [scares quotes in the original], later made famous by Kripke." (p. 21). I then went on to point out, both in the main text and in footnotes, differences between Kripke's version of the argument and Marcus' version (which, as I indicated, was present in Smullyan and Fitch). In the reply he read at the APA, Smith begins section 3 by ignoring my qualification that Marcus had given "a version of "the modal argument" ", and simply asserts that I agree that "Kripke's modal argument for direct reference was earlier stated by Marcus ..." (p. 42). In the material added to that section later, he takes issue with my characterization of the different arguments offered by Marcus and Kripke. I see now that my formulation of the matter was insufficiently explicit; my time was severely limited, but I should have chosen a terminology that was harder to misunderstand, or misrepresent. Let me try again. I will call any argument that distinguishes the semantic content of names and descriptions on the basis of their behavior in modal constructions, or their use in describing alternative counterfactual situations, a modal argument. Understood in this sense, modal arguments have been offered both by Kripke and by Marcus (following Smullyan and Fitch). However, the modal argument offered by Kripke is different from the modal argument repeated by Marcus. The structure of Kripke's argument is given in footnote 34 of my original article.

78

SCOTI SOAMES

"Kripke provides examples in which we evaluate simple sentences containing proper names (but no modal operators) at alternative possible worlds (counterfactual circumstances). He uses these examples to show that the referent of a name at one world is the same as its referent at other worlds, and hence that names are rigid. It then follows that identities involving names are necessary if true, and that coreferential names are intersubtitutable in modal constructions." (p. 33)

Since contingent descriptions don't have these features, it follows that the semantic content of a name is not identical with the semantic content of any such description. In particular, for any name n and contingent description d, it is not the case that the referent of n with respect to an arbitrary world w is semantically determined to be whatever satisfies d with respect to w. The structure of Marcus' modal argument is more difficult to pin down. The first thing she has to say on the subject in her 1961 article is the passage quoted above at the end of the last section. In that passage she considers two true identity sentences containing arbitrary names - (J 3) alb and (14) ala. Regarding these sentences, she claims (i) that since (13) is true "it must say the same thing as (l 4)," (ii) that since "(14) is surely a tautology," "(13) must surely be a tautology," (iii) that (ii) "is precisely the import of my theorem" [the quantified version of the necessity of identity], and (iv) that "it would be a consequence of the truth of ( 13) that (:Z' is replaceable by 'b' in any context except those which are about the names (:Z' and 'b:" It is evident from this passage that one step in her argument distinguishing names from descriptions is that identity statements involving coreferential names are tautologies. Ignoring the issues raised by the identification of necessity with tautology, we may express this by saying that identity statements involving coreferential names are necessary. Where does this claim in the argument come from? In the passage Marcus tells us that it is "precisely the import" of her 1947 theorem. Surely what she is saying here is that the claim about names is a consequence either of the theorem itself, or of the theorem together with trivial supplementary assumptions that can be taken for granted without mention. This is just the confusion mentioned above in the previous section. Thus, if this was Marcus' route to the claim in her modal argument about necessary identities involving proper names, then her reasoning was unsound. 29 There is, however, another possible route to this claim that is suggested by some of her comments. Claims (i) and (iv) above indicate that she held the view that coreferential proper names are intersubstitutable without change in meaning, proposition expressed, or truth value in all contexts, except metalinguistic ones about the names themselves. Call this the universal substitutivity thesis. Given this thesis plus the necessity of (14), one can derive the necessity of (13) without appealing to Marcus' 1947 theorem at all. What then were her grounds for the universal substitutivity thesis? For this, we turn to the next passage in Marcus (1961). The paragraph immediately following the one we have just been discussing begins as follows:

MORE REVISIONISM ABOUT REFERENCE

79

"Now suppose we come upon a statement like (15) Scott is the author of Waverly and we have a decision to make. This decision cannot be made in a formal vacuum, but must depend to a considerable extent on some informal considerations as to what it is we are trying to say in (10) [(10) The evening star eq the morning star) and (15). If we decide that 'the evening star' and 'the morning star' are names for the same thing, and that 'Scott' and 'the author of Waverly' are names for the same thing, then they must be intersubstitutable in every context." (pp. 308-9)

Clearly, both the example and what is said about it, come directly from Russell. 3o Although no reference is cited, there was surely no intention to deceive, since Marcus could take it for granted that her audience would be well aware of the source of the Russellian doctrines she was repeating. In particular she could assume that they would understand that she was invoking Russell's doctrine that a genuine name has no meaning apart from its referent, that coreferential names are synonymous, and that they are universally intersubstitutable except in metalinguistic contexts. In the next two paragraphs Marcus explicitly states her view that proper names are simply tags of objects that have no meaning apart from their referents (adding that on occasion descriptive-phrases are used as names in this sense, as is sometimes indicated by the use of upper-case letters.) It is significant that no argument is given for this view. Rather it is simply taken as a premise, from which Marcus concludes, in the very next paragraph, that if S is truly an identity statement - by which she means one in which the terms flanking the identity predicate are used as genuine names - then, if S is true, "it must be tautologically true or analytically true". (pp. 310-11). In light of this, we may reconstruct one of the argumentative routes suggested by Marcus' comments as follows: We begin with the premise that proper names are mere tags, and have no meanings apart from their referents. 31 From this one infers the universal substitutivity thesis (of coreferential names). Given that (14) above is tautologous, analytic and ncessary, one concludes that (13) is tautologically, analytically, and hence necessarily, true. In general, identity statements involving coreferential proper names are necessary, if true. One problem with this argument involves the indiscriminate characterization of the modality involved. Even if it is granted that names have no meaning apart from their referents, it does not follow that identity sentences involving coreferential names are tautologies (or logical truths, in the standard sense); nor does it follow that they are analytic, if that implies that anyone who understands them has information from which it is possible to recognize their truth, solely on the basis of reflection. The conclusion about analyticity would follow if it were understood that names must refer to objects of immediate acquaintance in Russell's sense - objects about which Russell thought one could not be mistaken. But Marcus, of course, does not assume this. However, even if we put aside the point about the nature of the modality,

80

SCOTI SOAMES

there is another difficulty: whether or not the argument to this stage is sound depends on whether or not the initial premise about names being mere tags is true. This question is a difficult and controversial one that is still under active dispute today. But whatever one's judgment about its truth, it clearly is not something that can be accepted without argument. Thus, on this interpretation, Marcus' modal argument rests on an interesting, but unargued and highly controversial premise that goes well beyond anything asserted in Kripke's argument. Marcus draws her argument to a conclusion in the next paragraph, which Smith takes as indicating that contingent descriptions are not intersubstitutable in modal contexts: "Let us return now to (10) and (15). If they express a true identity, then 'Scott' ought to be anywhere intersubstitutable for 'the author of Waverly' [this is where Smith inserts the words "in modal contexts" from the 1993 text into the 1961 text] and similarly for 'the morning star' and 'the evening star'. If they are not so universally intersubstitutable - that is, if our decision is that they are not simply proper names for the same thing; that they express an equivalence which is possibly false, e.g., someone else might have written Waverly, the star first seen in the evening might have been different from the star first seen in the morning - then they are not identities." (p. 311)

Let us grant that here Marcus indicates that substitution of co-denoting descriptions in modal contexts does not always preserve truth. We can now reconstruct two different versions of her modal argument:

Version 1: (i)

The 1947 theorem, (NI-1), (plus perhaps some trivial extra premise about names)

(ii)

Hence, (NC-2) - the claim that identity sentences involving proper names are necessary if true. Similarly, coreferential proper names are intersubstitutable in modal contexts.

(iii) Identities involving co-denoting contingent descriptions are not in gen-

eral guaranteed to be necessary if true, and co-denoting descriptions are not intersubstitutable in modal contexts. (iv)

So names are not semantically equivalent to contingent descriptions.

Version 2: (i)

Proper names have no meanings apart from their referents; coreferential proper names are synonymous, and universally intersubstitutable, except in metalinguistic contexts. The rest of the argument is as in version 1.

MORE REVISIONISM ABOUT REFERENCE

81

There are several points to notice about these arguments. First, both are very different from Kripke's modal argument. Thus, although Marcus can be seen as offering a modal argument, she did not give Kripke's modal argument. Second, both versions of Marcus' argument suffer from an indiscriminate characterization of the relevant modality. When the modality is identified with tautology or analyticity, the second step in the argument is false. Third, even if one takes the modality to be metaphysical necessity, step (ii) in the argument is not securely established. On version 1, step (ii) does not follow from step (i) together with any non-question-begging supplementary principles. On version 2, step (ii) does follow from step (i), but step (i) is a highly controversial premise that requires, but is not given, an argument of its own. Finally, a word should be said about the phrase "directly referential". This terminology is not used either in Marcus (1961) or in Kripke (1972). The phrase does, however, have two related interpretations in the literature today. The weaker interpretation holds that a term is directly referential iff it is not the case that there is a descriptive condition c associated with it such that for all worlds w, the referent, or denotation, of the term is determined, as a matter of semantics, to be whatever satisfies c in w. Kripke's modal, and other, arguments are designed to show that ordinary proper names are directly referential in this sense. The second, and stronger, contemporary interpretation of the notion of direct reference holds that a term is directly referential if its semantic content that which it contributes to the proposition expressed by a sentence containing it - is simply its referent. Kripke nowhere argues that names are directly referential in this sense, and his treatment of the proposition that Hesperus is Hesperus as different from the proposition that Hesperus if Phosphorus can be seen as a denial that names are directly referential in this stronger sense. Marcus, by contrast, does hold that names are directly referential in the stronger sense. Since the stronger sense carries the weaker sense as an implicit consequence, Marcus can be seen as maintaining that proper names are directly referential in both senses. 5. CONCLUSION

I would like to finish up on the same positive note that I ended my original reply. In both of my responses the need to correct Smith's false and overblown claims have made it necessary to concentrate on the problematic aspects of Marcus' discussion, and to emphasize the large gap between it and Kripke's groundbreaking contribution. In this context there is a danger of losing sight of the positive aspects of the articles of Smullyan, Fitch, and Marcus. For this reason I would like to repeat the remarks I made on page 28 of my first response. "we can see in the writings of Marcus, Fitch and Smullyan significant anticipations of some of the central theses of contemporary non-Fregean theories of reference. First, Marcus' quantified version of the necessity of identity anticipates the contemporary view that variables are rigid designators,

82

SCOTT SOAMES

with respect to assignments of values. Second, the claim, by Smullyan, Fitch, and Marcus, that substitution of coreferential names in modal constructions preserves truth value anticipates the view that ordinary proper names are rigid designators. Third, the use of this feature of names to discriminate them from some descriptions anticipates certain aspects of Kripke's modal argument against description theories."

The recognition of those positive contributions is fully consistent with the simultaneous acknowledgment of Kripke, and others, as primary founders of contemporary nondescriptivist theories of reference. As I pointed out on page 28 of my original reply, it is clear that Smullyan, Fitch, and Marcus were not attempting anything as systematic or far-reaching as Kripke (1972); thus it is not surprising that in nearly all fundamental respects their positive contributions to nondescriptivist theories of reference fall far short of that of Kripke (1972). Unlike Smith, who feels free to speculate unfavorably about Kripke's conscious and unconscious thought processes, I will not advance a thesis about the mental state that prevents Smith from seeing this. APPENDIX SMITH'S SPURIOUS DEFENSE OF HIS QUOTATIONS FROM MARCUS

In my original reply to Smith, all my quotations from Marcus were from the reprinted version, Marcus (1993). I did not carefully attend to the mostly insignificant differences in the texts until Smith made a point of the issue in his reply to me. Instead I relied upon Marcus' assurances that the changes involved only "editorial corrections and nonsubstantive changes". (1993, page 3). I was further encouraged in this by Smith himself, who, in his original paper, introduces his main thesis as follows: "The fact of the matter is that the key ideas in the New Theory were developed by Ruth Barcan Marcus, in her writings in 1946-7 (1946: 1947) and especially in her 1961 article on "Modalities and Intensional Languages" (reprinted with small changes in (Marcus 1993))." (pp. 179-80).

Having now compared the differences in the texts, I think that although most of the differences are purely stylistic (as is illustrated below), a few are philosophically significant (as is illustrated above in the text by the passage in which "in modal contexts" is inserted in the 1993 version). Unfortunately for Smith, where there are significant differences, such as that one, the earlier version of the text is further from the now widely accepted Kripkean doctrines than the later version is. I now turn to Smith's attempt to deflect criticism of his quotations from Marcus by citing my use of her (1993) text. The first discussion of Smith's misleading quotes appears in footnote 28 of my reply. There I cite the following passage from Smith. "But to give a thing a proper name is different from giving a unique description ... [An] identifying tag is a proper name of the thing ....This tag, a proper name, has no meaning. It simply tags. It is not strongly equatable with any of the singular descriptions of the thing." (Smith, p. 5).

MORE REVISIONISM ABOUT REFERENCE

83

I then cite the passage from Marcus, without Smith's first ellipsis. There is only one philosophically irrelevant, stylistic difference between the (1961) text and the (1993) text. To obtain the (1961) text from the (1993) text, substitute the bracketed word for the italicised word immediately preceding it. "But to assign [give] a thing a proper name is different from giving a unique description. For suppose we took an inventory of all the entities countenanced as things by some particular culture through its own language, with its own set of names and equatable singular descriptions, and suppose that number were finite (this assumption is for the sake of simplifying the exposition). And supose we randomized as many whole numbers as we needed for a one-to-one correspondence, and thereby tagged each thing. This identifying tag is a proper name of the thing." [my emphasis] (p. II of Modalities, pp. 309-310 of Marcus 1961.)

The second criticism of Smith's handling of quotations, which is simply an extension of the first, also occurs in footnote 28. There I noted that the passage from Marcus just cited continues through Smith's second ellipsis as follows: [The passages are word for word identical in the (1961) and (1993) texts.] "In taking our inventory we discovered that many of the entities countenanced as things by that language--{;ulture complex already had proper names, although in many cases a singular description may have been used. This tag, a proper name, has no meaning. It simply tags. It is not strongly equatable with any of the singular descriptions of the thing ..." (pp. II and 12 of Modalities, Marcus (1961), p. 310)

The third place where I draw attention to the difference between Smith's quotations and the passages from Marcus occurs on pages 20-21. On page 20, I repeat Smith's citation: "It would also appear to be a precondition of language [especially assigning names] that the singling out of an entity as a thing is accompanied by many ... unique descriptions, for otherwise how would it be singled out? But to assign a proper name is different from giving a unique description." [Bracketed words above inserted into the text by Smith]

On page 21, I give the full Marcus text from the beginning of the relevant paragraph, without ellipsis, or insertion of Smith's own words. To obtain the (1961) text from the (1993) text I used in my reply, add the first bracketed expression to the text, and substitute the second bracketed expression for the underlined word immediately preceding it. "That any language must countenance some entities as things would appear to be a precondition for language. But this is not to say that experience is given to us as a collection of things, for it would appear that there are cultural variations and accompanying linguistic variations as to what sorts of entities are so singled out. It would also appear to be a precondition of language that the singling out of an entity as a thing is accompanied by many - perhaps an indefinite [or infinite] number - of unique descriptions, for otherwise how would it be singled out? But to assign [give] a thing a proper name is different from giving a unique description." [Modalities, p. 11, Marcus 1961, p.309]

Clearly, the tiny differences between the two versions of the Marcus text are irrelevant both to the philosophical issues at hand, and to the question of the fairness of Smith's citation of the passages. In this connection, it is worth noting that Smith himself seems to have had

84

SCOTT SOAMES

trouble living up to his professional strictures (in addition to the manufactured quote that is noted above in the text of this article). The scrupulous reader may notice that the final sentence of my repetition, on page 20, of Smith's citation of Marcus - "But to assign a proper name is different from giving a unique description" - is not identical with the final sentence of Smith's own citation, on page 5 of the published paper - "But to give a thing a proper name is different from giving a unique description". Did I misquote Smith's quotation of Marcus? No, my citation of Smith on page 20 is word for word identical with the citation given by Smith on page 6 of the typescript he submitted to the APA, and gave to me for the purpose of preparing my comments. However, in the published version of his paper he changed the last sentence of the quote to conform to the original (1961) text, without indicating that he had made any changes from his typescript. In the preface to the issue of Synthese in which the published debate appears, the editors claim that the published version of his original paper "is verbatim as delivered". I don't know whether or not that is so, but whether or not it is, the unacknowledged discrepancy provides further evidence (i) that in the production of his paper Smith was himself relying, at least at some points, on the (1993) text, and (ii) that at the time he wrongly suggested that differences between the two texts would dispel my criticism of his misleading and incomplete quotations he was familiar with the two versions of the relevant passages from Marcus and should have been aware that the differences between them were tiny and insignificant. The final, and most important, place where I unfavorably contrast Smith's quotation of Marcus with the actual passage from Marcus is on page 24. I first reproduce Smith's quotation. "You may describe Venus as the evening star, and I may describe Venus as the morning star, and we may both be surprised that, as an empirical fact, the same thing is being described. But it is not an empirical fact that (17) Venus I Venus:'

In criticizing this quotation, I say "In presenting the passage, Smith breaks Marcus' final sentence in half, inserts a period immediately after her example sentence Venus I Venus, and omits, without any indication of ellipsis, the final half of her sentence". (p. 24). I then reproduce the passage as it appears in Marcus (1993), which is identical word for word with the (1961) text, except for the addition of the complementizer "that", indicated in brackets in the following: "You may describe Venus as the evening star, and I may describe Venus as the morning star, and we may both be surprised that, as an empirical fact, the same thing is being described. But it is not an empirical fact that (17) Venus I Venus and if 'a' is another proper name for Venus [that) (18) Venus I a" [My emphasis: Modalities, p. 12, Marcus (1961), p. 310]

MORE REVISIONISM ABOUT REFERENCE

85

Clearly, in the face of my criticism of his handling of this quotation, Smith's appeal to differences between the two versions of the text is nothing more than obfuscation. These are all the places in my article where I unfavorably compared Smith's quotes from Marcus with the actual text. There are other places in my article where I quoted passages from the (1993) reprinting of Marcus' paper to support my criticisms of Smith; in none of these cases is there any difference of philosophical import between the (1961) and the (1993) versions of the text. Department of Philosophy Princeton University

NOTES I would like to acknowledge the invaluable assistance of John Burgess in helping me prepare this response. Since the time that he got drawn into this unpleasant controversy, we have had many opportunities to confer, share ideas, and encourage one another in the seemingly thankless task of setting the record straight. Were it not for his efforts, I am not sure that I would have had the stomach to go another round in this dreary affair. The present paper owes a great deal to him. Nevertheless, I alone am responsible for any mistakes. 2 As pointed out in Burgess, "Marcus, Kripke, and Names", Philosophical Studies, 84, 1996, 1-47, the Marcus text actually appeared in 1962. though it was dated 1961. (I will continue to use the 1961 date.) Marcus' text was also reprinted in Proceedings of the Boston Colloquium for the Philosophy of Science 1961/1962, MW. Wartofsky (cd.), (Reidel, Dordrecht), 1963,77-96. The text and page breaks in the 1963 reprinting are identical to those in the 1961 original, so that a passage occuring on page n of the 1961 version appears on page n - 226 of the 1963 reprinting. On at least one occasion (p. 8 of his original article), Smith gives without labeling it as such the (1963) page number (which is actually off by I: the passage occurs on page 84, not page 85) of a passage he quotes from Marcus (1961). 3 Thanks to A. Zabludowski for pointing this out to me. The same quote occurs again on page 48 of Smith's reply to me. Marcus 1961, p. 308. Vol. 69, pp. 55-62; at page 61. Frederic B. Fitch, "The Problem of the Morning Star and the Evening Star", Philosophy of Science, Vol. 16, no. 2, 1949, 137-41; at page 138. 8 Arthur Smullyan, Review of w.v. Quine "The Problem of Interpreting Modal Logic", Journal of Symbolic Logic, 12, (1947), 139-141, and 'Modality and Description', Journal of Symbolic Logic, 13, (1948),31-37. 9 Smullyan (1947), p. 140. 10 Smith's discussion of the brevity ofSmullyan's remarks on proper names is hopeless. On page 39 he says: "Smullyan's 1947 review of Quine's paper, is, however, relevant, but even in this paper there is only one sentence about names - and even this one sentence Smullyan rejects as false. Furthermore, this sentence (even apart from the fact that Smullyan rejected it as false) does not even imply that names are directly referential and are not disguised descriptions. The sentence in question is on page 140. Smullyan writes: "If 'Evening Star' and 'Morning Star' proper-name the same individual they are synonymous and B is false," where B is the statement: "Evening Star is congruent with Evening Star and it is not necessary that Evening Star is congruent with Morning Star." " Part of the muddle here is that, contrary to Smith, the sentence quoted from Smullyan that is about proper names is not the sentence that Smullyan rejects as false. The rejected sentence, when

86

SCOTT SOAMES

understood as containing proper names, is (B). The sentence about proper names is the one that says that (B), when understood this way, is false. Smullyan accepts this sentence. As it happens, this example is overly complicated because of extraneous factors introduced by the word "congruent". For our purposes that word can simply be read as "identical". The rejection of (B) then amounts to the rejection of its second conjunct, and hence to the assertion of the necessity of the identity statement Evening Star = Morning Star. Since the case is clearly intended to be general, what we have here is an implicit endorsement of the necessity of identity for proper names. In addition, Smullyan has more to say about names than this. As I note in the text, the assumptions of Quine's argument make crucial use of the notion of an expression being "a constant". Smullyan presents this formulation of Quine's assumptions, and then asks whether by "constant" Quine means "proper name" or whether he also includes "singular descriptive phrases". Smullyan goes on to argue that on either construal the argument fails. (Thus all the remarks about constants receive an interpretation in which they are about names.) In particular, immediately after asking how 'constant' is to be interpreted Smullyan says: "It is possible that by 'constant' is meant what is commonly understood by 'proper name: Under this interpretation it appears evident to this reviewer that the principle of existential generalization is true, However, we observe that if 'Evening Star' and 'Morning Star' proper-name the same individual they are synonymous and therefore B is false." (bold italic is my emphasis, p. 140). 11 One of the bizarre features of Smith's discussion of Smullyan is his claim that even if "coreferential names are synonymous in modal contexts" (p. 40), this is consistent with their being abbreviations of rigidified descriptions using the actuality operator, which is advocated much later in, for example, Leonard Linsky, Names and Descriptions, (Chicago and London: The University of Chicago Press), 1977. But Smullyan doesn't say that coreferential names are synonymous in modal contexts; he says they are synonymous - i.e., they have the same meaning. By phrasing it as he does, Smith invites the thought that Smullyan's view was simply that coreferential names are intersubtitutable in modal contexts (as are coreferential, rigidified descriptions). However, that was not the whole of Smullyan's view. Like the later Marcus, Smullyan thought that it is because coreferential names have the same meaning that they must be substitutable in modal contexts. This view is not consistent with a Linsky-like analysis of names as rigidified descriptions - since it is a crucial feature of that analysis that the descriptions corresponding to a pair of coreferential names may have different meanings. 12 He fails, however, to draw attention to the fact that this theorem in Marcus was proven in a quite unorthodox way, and depended on unnecessarily controversial assumptions. See footnote 4 of my original reply to Smith for details. 13 See pages 10 and II, volume 5, The Encyclopedia of Philosophy, Paul Edwards (editor in chief), (New York, Macmillan and The Free Press - London, Collier Macmillan), 1967. 14 Philosophic Thought in France and the United States, Marvin Farber (ed.), Buffalo, New York, University of Buffalo Publications in Philosophy), 1950,545-563. The following passage occurs on page 552 (erroneously cited by Smith as occurring on page 252.) 5 Smith page 38. 16 For a discussion of the variety of options for handling constants and other terms in systems of quantified modal logic see James W. Garson, "Quantification in Modal Logic", in D. Gabbay and F. Guenthner (cds), Handbook of Philosophical Logic, vol. 2 (1984), 294-308. 17 page 553. 18 B. Russell, "On Denoting". Mind, vol. 14, 1905,479-493. 19 This point is emphasized in section 5 of Burgess, "Marcus, Kripke, and Names", Philosophical Studies, 84, 1996, 1-47. I should add that the matter is complicated when the notion of analyticity is involved. If it is required in order for a sentence to be analytic that anyone who fully understands it should have a basis for coming to recognize it as true, solely on the basis of reflection, then identity sentences involving ordinary coreferential proper names are typically not analytic. On the other hand, if in order for a sentence to be analytic it is sufficient that it be transformable into a logical truth by substitution of expressions with the same meaning, then the view that identity sentences involving coreferential names are analytic is defensible, provided that it can be shown that the meaning of a name is referent. Since the question of how precisely to characterize the notion of analyticity is not here at issue, I will not explore the matter further. For purposes of this paper, I will simply stipulate that a sentence will here be considered analytic only if understanding it provides

MORE REVISIONISM ABOUT REFERENCE

87

one with a basis for coming to know that it is true, by means of reflection alone. In this sense true identity sentences involving ordinary proper names typically are not analytic. 20 And explicitly rejected in the last paragraph of the preface added in 1980. 21 R.C. Barcan (Marcus), "The Identity of Individuals in a Strict Functional Calculus of Second Order", Journal of Symbolic Logic, 12, 1947, 12-15. 22 For details of this system, see section XIV, of David Kaplan, "Opacity", in The Philosophy of wv. Quine, Lewis Edwin Hahn and Paul Arthur Schilpp (eds), (La Salle, Illinois, Open Court), 1986. 23 To see why existential generalization fails, note that Oa = a is true, but3x Ox = a is false. 24 One of the fascinating themes developed by John Burgess involves the mismatch between the historical development of the technical apparatus of modal logic and the informal interpretations of the modal operators offered by some of those responsible for those developments. See "Which modal logic is the right one?," Proceedings of the George Boolos Memorial Conference, in M. Detlefsen (ed.), forthcoming, and "Quinus ab Omni Nrevo Vindicatus," in Meaning, Reference and Truth, The Canadian Journal of Philosophy Supplementary Volume, 1997, A. Kazmi (ed.), forthcoming. 25 Quine, "The problem of interpreting modal logic", The Journal of Symbolic Logic, 12, 1947,43-8. 26 Actually, it follows from this together with the observation that a = a is true. Since everyone ~Jants this truth, I will suppress further reference to it. See pp. 6-7. 28 See note 34 of my original reply. 29 A further difficulty with this line of reasoning, in addition to those pointed out above, is that if Marcus' theses about names were derived from her (1947) theorem of the necessity of identity (perhaps with the help of supplementary premises), then her theses about names would rest on the assumptions she used in proving that theorem. But, as I pointed out in footnote 4 of my original reply, these assumptions included controversial principles - the S4 iterability of modal operators plus the second order Barcan and converse Barcan formulas - that neither the necessity of identity ~~uantified version) nor the relevant theses about names are standardly thought to rely on. See, "On Denoting", op. cit., chapter 16 of Introduction to Mathematical Philosophy, and The Philosophy of Logical Atomism, D. Pears (ed.), (La Salle Illinois: Open Court), 1985, at pp. 113--4. 31 This premise alone, together with the obvious thesis that definite descriptions are not mere tags, and do have meaning apart from their referents, would be enough to distinguish names from descriptions without appealing to modal notions at all. Indeed Marcus herself draws the conclusion that names and descriptions are not equivalent on pages 309 and 310, two paragraphs before the one cited by Smith and taken to indicate that descriptions do not share the behavior of names in modal contexts. Thus, though Marcus does give a modal argument in which the different behavior of names and descriptions in modal contexts shows the need to distinguish them, a crucial early premise of that argument provides a basis for distinguishing them even apart from modal considerations.

a

JOHN P. BURGESS

MARCUS, KRIPKE, AND NAMESl Great writers create their own predecessors - Borges

INTRODUCTION

Ruth Barcan Marcus, in several incidental passages in her later published writings, seems to imply a claim, which has subsequently been stated more explicitly by others,2 and given very wide publicity, that certain remarks on names in her colloquium talk "Modalities and Intensional Language" anticipate in an important but unacknowledged way Saul Kripke's discussion of names in his lecture series "Naming and Necessity". My aim in the present study is to assess the textual evidence bearing on such claims. Marcus' talk was presented early in 1962 at a legendary session of the Boston Colloquium for the Philosophy of Science, and was followed by a commentary by W. V. Quine, and a discussion between the two speakers and some members of their audience, mainly Kripke and Dagfinn Follesdal. Even though this was not a pre-arranged panel discussion - Kripke's opinion was invited by Quine on the spot, not in advance - it was recorded. All three components of the symposium were published in the journal Synthese, the talk and commentary in an issue bearing the date 1961 - apparently a spurious date, since according to Quine [1985, p. 35] the publication was subsequent to the colloquium - and an edited transcript of the discussion in an issue bearing the date 1962. Talk, commentary, and discussion were also published together in the official colloquium proceedings volume Wartofsky [1963], whose pagination I will use in giving quotations. A partial account of Kripke's then-unpublished views on naming was given in the latter part of Kaplan [1969]. Kripke's famous public lectures were given in early 1970, and a virtually verbatim transcript with some further footnotes and a section of addenda somewhat belatedly published in Davidson & Harman [1972]. A partial version of the same material had in the meantime appeared in Munitz [1971]. The reprinting of Kripke [1980] contains a substantial new preface. Marcus' well-known early work on axiomatic systems of modal logic, in Marcus [1947] and companion papers, is only indirectly relevant to the present study, since it was artificial rather than natural languages that were in question there, and even these did not contain constants, which are the closest symbolic 89 P. W. Humphreys et al. (eds.), The New Theory of Reference © Kluwer Academic Publishers 1998

90

JOHN P. BURGESS

counterpart of names (much as variables are the closest symbolic counterparts of pronouns). Part of the content of Marcus [1963] had appeared in the short paper Marcus [1960]. Marcus' next relevant work was a symposium presentation in late 1970, fairly promptly published as Marcus [1971]. The dates of the original oral presentations of Kripke [1972] and Marcus [1971] are clearly indicated in the published versions, and show that the order of original oral presentation was the reverse of the order of publication, with Kripke's lectures coming almost a year before Marcus' symposium contribution, so that chronological considerations exclude the possibility of influence of Marcus [1971] on Kripke [1972], a point I emphasize here because of the occurrence of misleading statements elsewhere. Marcus' well-known later work on belief statements is not directly relevant to the present study, but Marcus [1983], which is avowedly partially a response to Kripke [1979], will be compared briefly with Marcus [1963]. In citing Marcus' late works I will give the pagination in the reprinting in the more readily available volume of collected papers Marcus [1993], which is also the version I will quote. Some relevant items appear only in this collection. Kripke [1972] cites a wide range of literature, most of which falls into three groups: the classical writings of John Stuart Mill, Gottlob Frege, and Bertrand Russell; writings from late 1950s and very early 1960s by Peter Geach [1957], William Kneale [1962], John Searle [1958], Peter Strawson [1959], Paul Ziff [1960], and others; and later work by Hilary Putnam, David Kaplan, Keith Donnellan, and others, mainly on expressions other than proper names, which together with Kripke's own contemporary and partly overlapping, partly complementary work is known as the "new theory of reference", though it is now over a quarter-century old. (Appropriately enough, the use of a descriptive phrase to refer to a thing of which the description is untrue is one of the phenomena treated by the theory.) When I cite or quote Mill, Frege, or Russell, it will be in the versions of Ernest Nagel, Max Black, and David Pears, respectively: Mill [1950], Frege [1970], and Russell [1985]. I will only allude in passing to the writings from the five years preceding Marcus' talk, and will not have occasion to discuss new theorists other than Kripke. Such are the relevant texts. My readings have been influenced by discussions with my colleague Scott Soames (as well as by his written work, Soames [1995]). I will not itemize how far I follow and how far I depart from him, except to say that despite my indebtedness to him, he is not responsible for my ultimate conclusions. I do begin with one caveat that Soames has stressed. First some background. From the large claims that have been made for Marcus [1963], one might suppose it to have been a longer than average talk, entirely devoted to the topic of naming. In fact, naming is mentioned neither in the title of the talk nor in the introduction to the talk, where the overall purpose of the paper is identified as being the defense of modal logic against the criticisms of Quine [1961]; and only about four pages (82-85), or a little over half of one of the five sections of the paper, is devoted to the enunciation of

MARCUS, KRIPKE, AND NAMES

91

certain views about names and their application to the interpretation and defense of modal logic. For purposes of such application, Marcus sometimes needs or wants the assumption that everything has a "name". This is not true of names in the ordinary sense, but for purposes of the interpretation and defense of modal logic what matters is not so much that "names" with certain features should be in existence but that they should be capable of being introduced. Accordingly, Marcus invites us to contemplate the following imaginary scenario: (A) For suppose we took an inventory of all the entities countenanced as things by some particular culture ... And suppose we randomized as many whole numbers as we needed for a one-toone correspondence, and thereby tagged each thing.

When she continues, "This identifying tag is a proper name of the thing", what is being called a "name" is thus not a name in the ordinary sense, or even a linguistic expression, but a serial number. For comparison, here is a similar imaginary scenario from a later work, Marcus [1971, p. 61], where the purpose is the interpretation of formalized systems of modal logic and formal models for them: For those who are quick to argue that ordinary names cannot always be used in such a purely referential way, we can, in giving the interpretation, expand our lexicon to provide neutral, purely referential names where necessary.... Associated with individual constants are ordinary names, not singular descriptions. Where ordinary proper names are lacking, such nameless objects are first given "ordinary" names by a suitable convention for avoiding duplication of names. (A lexicon is kept.) Where more than one object has the same name, we distinguish them by a suitable convention. For symmetry we might add that, when one object has several names, we choose one as its standard name.

Now the caveat. As the foregoing quotations show, part of what Marcus assumes about "names" is clearly not intended to hold literally for names in the ordinary sense. This circumstance raises doubts as to how much of the rest of what Marcus says is intended to apply to names in the ordinary sense, in natural language. Such doubts are reinforced by the specific wording of some other passages (notably quotation (1) below, with its talk of "names in an ideal sense"). And as will become clearer in later sections of the present study, such doubts are futher reinforced by the general content of what is said about "names", and by its general similarity to earlier discussions of "names" in the literature in which names in the ordinary sense were unambiguously not what was meant. Also Marcus' lack of citations of the literature on names in natural language - the part of her talk on "names" contains no reference to the literature at all - would be most understandable if at the time of the colloquium she herself cared little about naming in the ordinary sense as a subject in its own right (as contrasted with how assumptions about "naming" in some sense could be pressed into service in the defense of modal logic). But unless Marcus intends most of what she says to apply to names in the ordinary sense, what she says could at best anticipate Kripke in some quite indirect way. In the body of this study I will assume that she does intend most

92

JOHN P. BURGESS

of what she says to apply to natural language, but I emphasize now at the outset that this is giving her the benefit of a real doubt. I. MILL VERSUS FREGE

A claim has recently been heard to the effect that the idea that names refer directly, or in other words, that they lack descriptive meaning, was "introduced into the new theory of reference" by Marcus. But neither this nor any of a number of other ideas to be reviewed in the present section could have been "introduced" by any speaker in 1962, whether in the strict sense of being originated, or in the loose sense of being revived after being forgotten. The true original sources for these ideas are discussed in Kripke [1972], as is the kind of later literature I will cite in this section. The omission of similar citations in Marcus [1963] was I think excusable in the context of the symposium, precisely on the grounds that the relevant sources were (or could reasonably have been presumed to be) familiar to her audience. But the recent occurrence of claims like the one alluded to in the preceding paragraph suggests that the absence of citations in Marcus [1963] has misled some later readers less knowledgeable than her original audience. Because in considering the alleged influence of Marcus [1963] on Kripke [1972] one must be careful to subtract off anything Kripke correctly credits to earlier sources than Marcus, I will begin by reviewing some of the pre-1962 literature, with apologies to readers to whom this background is already well known. The source for the idea that names have no descriptive meaning is correctly identified by Kripke as Mill's System of Logic (book I, chapter II, section 5; Mill [1950, p. 25]). Proper names are not connotative; they denote the individuals who are called by them, but they do not indicate or imply any attributes belonging to those individuals. ". Names are simply marks used to enable those individuals to be made subjects of discourse.

A similar view to Mill's is implicit in much though not all of practical lexicography, where it is the usual though not invariable custom to distinguish lexicon-type "dictionaries" (such as the first section of "partie Langue" of the Petit Larousse), containing only linguistic definitions giving the meanings of the words listed, from encyclopedia-type "dictionaries" (such as the second section or "partie Arts, Lettres, et Sciences" of the Petit Larousse), giving empirical identifying information about the things designated by the words listed. The practice is to exclude names from lexical dictionaries and include them only in encyclopedic dictionaries, which accords with a theory on which names do not have meaning. Opposed to Mill stands Frege ("Uber Sinn und Bedeutung", Frege [1970a]), who held that a name is an abbreviated description. One main objection proponents of Frege's descriptive view predictably raised against Mill's nondescriptive view is that a thing must be somehow specified before it can be

MARCUS, KRIPKE, AND NAMES

93

given a name, and that this can hardly be accomplished except through description (perhaps involving demonstratives and supplemented by ostension). In many cases the etymology of the name clearly suggests what the description must have been, as "Dartmouth Town" suggests "town standing at the mouth of the river Dart", or "Evening Star" suggests "first or brightest star of the evening". The response can only be that this description nonetheless does not remain permanently connected with the names as anything like a definition giving its meaning. Since Mill wrote before Frege, he did not confront the objection in exactly the form indicated, but he did give one argument supporting the response indicated: In may be said, indeed, that we must have had some reason for giving them those names rather than any others, and this is true, but the name, once given, is independent of the reason .... [A) town may have been named Dartmouth because it is situated at the mouth of the Dart. But it is no part of the significance ... of the word Dartmouth to be situated at the mouth of the Dart. If sand should choke up the mouth of the river or an earthquake change its course and remove it to a distance from the town, the name of the town would not necessarily be changed.

This is a primitive kind of counterfactual argument. A primitive kind of epistemological argument against the descriptive theory was already noted by its author. For one must credit to Frege himself the objection that one generally cannot find, for a given name a, a description t for which the identity statement a = t is discoverable to be true by any competent speaker of the language merely through linguistic reflection, as it ought to be if a is simply an abbreviation for t. This difficulty led Frege to modify his theory within a page or two of introducing it, to the extent of presenting it as a norm that in a scientific language for each name a there should be a description t of its bearer having the same meaning as a in the common language of all speakers, while presenting it as a fact that in natural languages for each name a there is for each speaker a description t of its bearer having the same meaning as a in the personal idiolect of that speaker. There is a long temporal gap between the mid-nineteenth and the midtwentieth century, and a large disciplinary gap between applied lexicography and theoretical philosophy, but one finds, even considering just the five years up to the time of Marcus' talk, and just the work of Marcus' fellow philosophers, the following items. In a talk at a 1960 international congress, later published in the proceedings volume as Kneale [1962], Mill on names is cited as a source the speaker expects to be familiar to his audience, and is then criticized - the citation, the expectation of familiarity, and the criticism all being typical of much of the literature of the period. The specific criticism or objection that description accompanying ostension would be needed to identify a thing being given a name is raised (and the opinion of unnamed linguists that names are not part of the language mentioned and criticized), in Geach [1957] (and follow-up work extending outside the time-period in question here, and collected later in Geach [1972]). The view that names have no meaning and are not part of the language is on the contrary endorsed, and the response to the

94

JOHN P. BURGESS

foregoing objection is made, that names may remain applicable while descriptions originally associated with them may become inapplicable, in Ziff [1960]. Epistemological objections against even Frege's modified view are raised, and further modification though not abandonment of the theory advocated, in Searle [1958] and Strawson [1959]. Thus clearly naming was under active discussion in the period preceding Marcus' talk, and participants in the discussion were in no need of "introduction" to the classical arguments and counter-arguments of Mill and Frege. Searle [1967] gives a good idea of the status of the question in the 1960s. The first passage that Marcus herself and others quote (see [1990a, p. 248n.]) when they wish to establish that she was an originator of the new theory of reference, reads as follows (Marcus [1963, pp. 83-84]): (B) But to give a thing a proper name is different from giving a unique description .... This identifying tag is a proper name of the thing .... This tag, a proper name, has no meaning. It simply tags. It is not strongly equatable with any of the singular descriptions of the thing, ...

(The ellipsis, not actually indicated in Marcus' self-quotation, is of some importance, since it is passage (A) quoted earlier that comes between the first and second sentences of (B), and this obviously affects the sense of what comes before and after.) To the extent that Marcus gives an argument or example to establish or illustrate the distinction between a name with a descriptive etymology and a genuine description, it is in the following nearby passage (p. 83): (C) One might even devise a criterion as to when a descriptive phrase is being used as a proper name. Suppose through some astronomical cataclysm, Venus was no longer the first star of the evening. If we continued to call it alternatively 'Evening Star' or 'the evening star' then this would be a measure of the conversion of the descriptive phrase into a proper name.

Clearly what we find in (B) and (C) are no innovations, but simply a restatement of a thesis and argument of Mill - compare in particular Marcus' "cataclysm" with Mill's "earthquake" - and moreover of a thesis and argument of Mill's that were familiar to knowledgeable specialists circa 1962. There is one difference between Mill and Marcus, in that he considers a phrase with a descriptive etymology being used as a proper name from the start, while she considers such a phrase being used first as a genuine description, then as a proper name (suggesting in a passage I have not quoted that capitalization may mark the change, though conceding that usage is not systematic in this regard). The converse phenomenon of transformation of names to descriptions was familiar to linguists (the stock example being the conversion of "Caesar", family name of the first few Roman emperors, into a title meaning "emperor"), but I myself know of no earlier source for Marcus' observations on transformation of descriptions into names. However, original or not, those observations have played no significant role in subsequent discussions of names such as Kripke's. What quotations (B) and (C) do show is that Marcus endorsed a view like

MARCUS, KRIPKE, AND NAMES

95

Mill's at a time when there were few others - though as has been seen, it cannot be said that there were no others - who did so. But in philosophy, personal endorsements count, or should count, for little in comparison to impersonal arguments. Whether Marcus "introduced" or "originated" cogent arguments beyond those in the classics remains to be considered. But first some aspects of Marcus' view that do not have antecedents in Mill must be noted. 2. RUSSELL

Important aspects of Marcus' view are connected with "Leibniz' Law", the principle (X) of the indiscernibility of identicals, and its corollary, the principle (Y) of the necessity of identity: (X)

For all P: If x

(Y)

If x

=

y, the P(x) if and only if pry)

= y, then necessarily x = y

Formalized symbolic counterparts of (X) and (Y) had figured already in Marcus' early work on axiomatic systems of modal logic, but it is going beyond anything in her own early work, as well as beyond Mill, when Marcus in her colloquium talk maintains that (X) and (Y) hold when names are substituted for the variables. Formulations in her own words are the following (p. 82, p. 84): (D) ... [I)t would be a consequence of the truth of (13) [alb, the identity of a and b) that 'a' is replaceable by 'b' in any context except those which are about the names 'a' and 'b'. (E) ... What I have been arguing for the past several minutes is, that to say of an identity (in the strongest sense of the word) that it is true, it must be tautologically true or analytically true.

But to say that such aspects of Marcus' view have no antecedent in Mill is not to say that they have no antecedents at all. John Bacon [1965] for one has found some of what Marcus says reminiscent of Russell - and well he might. Specifically, antecedents for Marcus' view may be sought in Russell's 1918 Philosophy of Logical Atomism, lecture VI (and his 1919 Introduction to Mathematical Philosophy, chapter XVI). For consider what Russell [1985, p. 113] says about the character of true identity statements involving names: If you were to try to substitute [in 'Scott is --') any name whatever, say 'c', so that the proposition becomes 'Scott is c', then if 'c' is a name for anybody who is not Scott, that proposition would become false, while if, on the other hand, 'c' is a name for Scott, then the proposition will become simply a tautology. It is at once obvious that if 'c' were 'Scott' itself, 'Scott is Scott' is just a tautology. But if you take any other name which is just a name for Scott, then if the name is being used as a name and not as a description, the proposition will still be a tautology. For the name itself is merely a means of pointing to the thing, and does not occur in what you are asserting, so that if one thing has two names, you make exactly the same assertion whichever of the two names you use, provided they are really names and not truncated descriptions.

96

JOHN P. BURGESS

Compare this with what Marcus says in her colloquium talk on the same topic (p.82): (F) Consider the claim that (13) alb

is a true identity. Now if (13) is such a true identity, then a and b are the same thing. It doesn't say that a and b are two things which happen, through some accident, to be one. True, we are using two different names for that same thing, but we must be careful about use and mention. If, then, (13) is true, it must say the same thing as (14) ala.

But surely (14) is a tautology, and so (13) must surely be a tautology as well. That is precisely the import of my theorem (8) [the strict equivalence of xly and 0 (xly)].

Consider next how Russell applies the doctine to distinguishing naming from describing (p. 114): So there are only two alternatives. If 'c' is a name, the proposition 'Scott is c' is either false or tautologous. But the proposition 'Scott is the author of Waverly' is neither, and therefore is not the same as any proposition of the form 'Scott is c', where 'c' is a name. This is another way of illustrating the fact that a description is quite a different thing from a name. [By contrast] ... if you substitute another name in place of 'Scott' which is also a name of the same individual, say, 'Sir Walter' then if 'Scott' and 'Sir Walter' are being used as names and not as descriptions, your proposition is strictly a tautology.

Compare how Marcus applies her doctrine to the same task (p. 84): (G) But my point is only to distinguish tagging from describing, proper names from descriptions. You may describe Venus as the evening star and I may describe Venus as the morning star, and we may both be surprised that, as an empirical fact, the same thing is being described. But it is not an empirical fact that (17) Venus I Venus and if 'a' is another proper name of Venus [that] (18) Venus I a.

(Passage (H) below is also relevant in this connection.) The extent to which the logical atomist school anticipated Marcus' approach becomes clearer on reading some comments of Arthur Prior [1967, chapter VIII, §4] on the history of the idea of "the necessity of identity". For as the foregoing quotations already show, this idea does have a history going back to Russell, and could no more have been introduced by a speaker in 1962 than the idea that names have no descriptive meaning. In Ludwig Wittgenstein and FP. Ramsey one finds not only endorsement of this idea of Russell's, but also two further, related ideas. First, there is the thought that an existential quantification 3xP(x) should count as true just in case at least one instance P(a) for

MARCUS, KRIPKE, AND NAMES

97

some name a is true - the ancestor of the "substitutional interpretation" adopted in the next section of Marcus' talk after the one on names. Second, there is the thought that it is tautologous that there are as many things and just the things that there are - the "constant domain" adopted in the formal modeltheoretic construction in the final section of Marcus' talk. As Prior happens to mention, the time interval is about ten years between the last of Ramsey and the first of Marcus. (For more on logical atomism and modal logic, see Cocchiarella [1984].) The Russellian material, like the Millian material considered in the preceding section, was hardly forgotten in the 1960s. Prior [l967a] mentions "Russell ian logically proper names" casually, as something he expects will be familiar to his readers, and Russell's Logical Atomism and Wittgenstein's Tractatus are the first modern works cited in Searle [1967]. The work of avowed Russellians like Arthur Smullyan provided the immediate background for Marcus' talk, as explained in the next section. As with Mill, so with Russell, Marcus' failure to cite this background literature at the colloquium is excusable precisely because it could have been presumed familiar to her audience. But as with Mill, so with Russell, the absence of citations seems to have misled some later readers less knowledgeable than the original colloquium audience. (In the case of Russell, unlike that of Mill, Marcus does cite other aspects of his work in her talk, and discusses some of the aspects most relevant to her thoughts on naming in some of her later publications. This is also to a degree true of Ramsey and Smullyan.) What the logical atomists had not considered was formal systems of modal logic, and what Prior correctly attributes to Marcus is the development of a formal system of modal logic with quantification and identity, and a formal derivation of a law of necessity of identity within the system. But clearly if Marcus is to have made an original contribution beyond Russell and the logical atomists to the theory of names as a topic in its own right (as opposed to the kind of contribution involved in her purely formal work, or in applying a preexisting theory of names to the defense and interpretation of formal systems of modal logic) she must have meant by what she said something substantially different from what Russell meant by what he said in very similar-looking passages. What Russell meant turns out to be a somewhat subtle question, which will be taken up in the next section. But first some review is needed of Quine's critique of modal logic, to which Marcus' talk was avowedly a response. 3. QUINE VERSUS SMULLYAN

There can be no question here of a full discussion of the evolution and reception of Quine's critique of modal logic, for which see Linsky [1969, 1971a] and especially F0llesdal [1986]. Background assumptions of Quine's critique are that there is not much point to modal logic unless one is going to go

98

JOHN P. BURGESS

beyond sentential logic to consider predicate logic - Quine like Prior correctly cites Marcus as a "pioneer" in so doing - and that there is not much point to modal predicate logic unless one is going to go beyond de dicto modality, in which modalities occur only outside quantifiers in application to closed formulas, as in D3P(x), to consider de re modality, in which modalities occur inside quantifiers in application to open formulas, as in 3xD (Px). Now taking the quantifier as a genuine quantifier, 3xDP(x) holds just in case there is some thing such that DP(x) holds of it, which in turn holds if and only if there is something such that P(x) holds necessarily of it. To make sense of de re modality, one needs to make sense of an open P(x) holding necessarily rather than contingently of a thing, or in another terminology, of expressing an essential rather than an accidental feature of the thing. It is indicated in the antepenultimate paragraph of Quine [1961] that the critique there is mainly directed against the conception adopted by c.1. Lewis, Rudolf Carnap, and other early writers on modal logic, according to which the necessity sign D stands for logical or analytical truth. The notion of an open formula holding necessarily rather than contingently of a thing seems problematic on such a conception because the notion of logical truth or validity is usually explained in logic textbooks only as it applies to closed formulas, as in de dicto modality (and similarly for analytic truth). The only obvious strategy for trying to make sense of the application of the notion of logical or analytic truth to open formulas, as in de re modality, would be to attempt to reduce de re modality to de dicto modality by defining the open P(x) to hold logically or analytically of a thing if and only if the closed P(t) holds logically or analytically, where t is a term designating the thing. But here one would face difficulties. First, there is what might be called the problem of anonymity. There may be no term t designating the thing. If this problem is waived, at least for the sake of argument, then second there is what might be called the problem of aliases: There may be two terms t and s both designating the thing such that P(t) holds logically or analytically but P(s) does not. Considering just the simplest case where P(x) is x = t, so that P(t) is t = t and holds logically and analytically, P(s) is s = t and may not hold logically or analytically. Quine supposes m0dal logicians will be driven by the problem of aliases to restrict the class of terms allowed, adopting "a frankly inegalitarian attitude towards the various ways of specifying" the same thing, and the difficulty Quine sees is simply that how and why some terms should be privileged over others has not been adequately explained or justified. This might be called the problem of arbitrariness. As examples of the problem of aliases, Quine would submit that the following are all true (on the conception of necessity just indicated): (1)

The evening star is the morning star.

(2)

Necessarily the morning star is the morning star.

MARCUS, KRIPKE, AND NAMES

99

(3) (4)

It is not the case that necessarily the evening star is the morning star.

The author of Ivanhoe is the author of Waverly.

(5) (6)

Necessarily the author of Waverly is the author of Waverly. It is not the case that necessarily the author of Ivanhoe is the author of Waverly.

(7) (8) (9)

The number of planets is the square of three. Necessarily the square of three is the square of three. It is not the case that necessarily the number of planets is the square of three.

An example of the inegalitarianism of the kind Quine had in mind is provided by some systems of "intensional arithmetic". (See Shapiro [1985].) These in effect take the descriptions of zero, one, two, and so on, as the least natural number, the next-to-Ieast, the next-to-next-to-Ieast, and so on, to be privileged descriptions of these numbers. What follows logically or mathematically from its privileged description counts as holding necessarily or essentially of a number, while what follows logically or mathematically only from some non-privileged description counts as holding only contingently or accidentally. It is harder to think of plausible candidates for privileged descriptions in the literary and astronomical examples. Marcus indicates in Marcus [1963, p. 82] that part of what she says has to say in response to all this should be familiar from unspecified earlier discussions in the literature, of which she says that she "will try to restate them more persuasively". In Marcus [1993a] she identifies the source with which she was presupposing familiarity as Smullyan [1948]. Smullyan does not really address what commentators like Linsky or F01lesdal consider Quine's general challenge, that of making sense of de re modality. Rather, in Smullyan [1947, 1948] he interprets Quine's presentation of cases like (1)-(3), (4)-(6), or (7)-(9) as a presentation of alleged counter-examples to laws like (X) and (Y) of the preceding section. (It is (7)-(9) and (X) that he considers explicitly, but his point obviously generalizes.) Smullyan's claim is that these are not genuine counter-examples, because one does not get genuine instances of the laws if one substitutes for the variables in them descriptions like "the morning star" or "the author of Waverly" or "the number of planets". Smullyan's claim is that an item like (1) or (4) or (7), which Quine and others would call an identity statement between descriptions, is not a genuine identity statement in the strictest sense, but rather is something that should be analyzed as having some other logical form (either by the method of Russell's famous theory of descriptions or in some other method). In addition to this negative claim that laws like (X) and (Y) should not be expected to hold with descriptions in place of the variables, Smullyan suggests in passing the

100

JOHN P. BURGESS

positive claim that these laws should be expected to hold with names in place of the variables, but does not elaborate. A good summary is given in Prior [1967a, p. 10]: [I]t may be argued (as it has been by Arthur Smullyan) that Leibniz' law holds only for x's and y's that directly name whatever they do name, not for x's and y's that pick out what they name merely as the thing, whatever it is, that answers to some definite description.

A very similar line is followed by Fitch [1949] and Marcus [1960], but the positive claims about names (as opposed to the negative claims about descriptions) are not extensively elaborated until Marcus [1963]. Marcus emphasizes (in line with her views on the possibility of an expression passing back and forth over time between the category of descriptions and the category of names) that there may be an ambiguity as to whether an expression like "the morning star" or "Morning Star" is being used as a description or a name. In addition to the method of imagining whether the expression would still be used for the thing if it were no longer true of it as a description - the method of passage (C) - she suggests another for distinguishing names from descriptions. She insists in passage (G) that if "the morning star" and "the evening star" are being used as descriptions, then indeed it may be a surprising empirical discovery that the morning star is the evening star, while if "Morning Star" or "Evening Star" or some other expression a is being used as a genuine proper name, then it is not an empirical fact that Venus is a. So if something like (1)-(3) appears to be a counter-example to a law like (Y), according to Marcus that is simply an indication that the expressions involved are descriptions and not genuine proper names. To give Marcus' own words (p. 85): (H) Let us now return to (10) ["the evening star eq the morning star"] and (15) ["Scott is the author of Waverly"]' If they express true identities [between names], then 'Scott' ought to be anywhere intersubstitutable for 'the author of Waverly' and similarly for 'the morning star' and 'the evening star'. If they are not so universally intersubstitutable [then] our decision is that they are not simply proper names for the same thing ...

Marcus herself, as already mentioned, described her discussion as an attempt at a more persuasive restatement of Smullyan's point, and thus her remarks were not presented at the time as an introduction of striking novelties. Nor do they seem to have been taken as such even by writers as highly sympathetic to Marcus as Prior, the writer who immortalized her name by attaching it to the "Barcan formula". Or at least, he does not find it needful to mention Marcus alongside Smullyan in this connection in his concise account Prior [1967a], or to include Marcus [1963] on his select list of recommended readings. Quine also gave no indication that he thought of himself as introducing striking novelties in his rebuttal to Marcus. On the contrary, he presents himself as somewhat wearily going one more time over familiar material. What he says in response to the specific point of Marcus just discussed is largely

MARCUS, KRIPKE, AND NAMES

101

anticipated by the earlier response in Church [1950] to Smullyan and Fitch, where it is insisted that names in the ordinary sense do not have the feature that true identity statements involving them are always analytic or a priori, and that there seem to be no class of expressions of natural language with this feature. For the most basic point of Quine's rebuttal is just that his astronomical example can easily be reformulated with names rather than descriptions (on the same conception of necessity as in the earlier examples): (1 *) Hesperus is Phosphorus.

(2*) Necessarily Phosphorus is Phosphorus. (3*) It is not the case that necessarily Hesperus is Phosphorus. The original formulation in his commentary, Quine [1963, p. 101], is worth quoting: We may tag the planet Venus, some fine evening, with the proper name 'Hesperus'. We may tag the planet Venus, some fine morning, with the proper name 'Phosphorus'. When at last we discover that we have tagged the same planet twice, our discovery is empirical. And not because the names were descriptions.

If Quine is right about such examples, then if there is to be a class of "tags" such that true identity statements involving them are logically or analytically true, it cannot just be the class of all ordinary names, but will have to be a privileged subclass, not containing both "Phosphorus" and "Hesperus" - and it does seem arbitrary to privilege one of the two over the other. What Quine says about the astronomical case can be adapted to the literary case. For we may hear on one occasion of the famous author, Scott, and on another occasion of the bankrupt baronet, Sir Walter, and we will be unable to discover by mere reflection that it has been the same person spoken of both times. The assertion that Scott is Sir Walter can then be informative, as the assertion that Scott is Scott cannot be, and so these two are not the same assertion, and what Russell said in the passages quoted in the preceding section is thus false. Or rather, it is false if taken literally. But it was not meant to be taken literally. Russell's view of ordinary names like "Scott" - or "Romulus", the example Russell himself uses in the pages immediately preceding the passages quoted - was that they are "truncated descriptions". In the passages quoted he was expounding the features not of ordinary names but of extraordinary "logically proper names", merely pretending for the sake of illustration that the ordinary names "Scott" and "Sir Walter" were such "logically proper names". Not only did Russell not take it for a fact that names in natural language function as "logically proper names", but also he did not even propose it as a norm that names in scientific language should function as "logically proper

102

JOHN P. BURGESS

names". For not only did he not think "logically proper names" for ordinary things were already in existence, but also he did not even think they were capable of being introduced. His reason for thinking that ordinary objects could not bear "logically proper names" was that ordinary things can be presented in different ways on different occasions, so that names associated with the thing on the two occasions cannot always be recognized by mere reflection to name the same thing - precisely the feature of ordinary things on which the foregoing examples turn. Such examples go back to even before Russell: Quine's use of an astronomical example is a nod towards Frege, who had used such examples, and had pointed out that asteroids and comets provide a continuing supply of cases where hard empirical work is needed to determine whether a body sighted and given a name on one occasion and a body sighted and given another name on another occasion are or are not one and the same. (This hard work is simply assumed to have been somehow successfully completed when one assumes as in quotation (A) that one has completed an "inventory" of the entire contents of heaven and earth.) For Frege, the main argument in favor of the descriptive theory was precisely that if each name is assumed to have a descriptive content, and distinct names to have distinct descriptive contents, then it can be explained how the identity statement linking two distinct names for the same thing can be intelligibly doubted, and substantially informative. It may be mentioned that while at the colloquium Kripke thought Marcus had "a perfectly valid point of view" - he says this immediately after passages (I) and (J) quoted below - he later concluded "Quine is right". In Kripke [1972] all of Frege, Russell, and Marcus are criticized for maintaining, each in his or her own way, the false dichotomy according to which the only alternative to the view that names are abbreviated descriptions would be the view that names so absolutely immediately indicate their bearers that it can always be determined by mere reflection whether two names designate the same thing. Quine, Kripke, and more recent reviewers such as Bacon [1995] all interpret Marcus in essentially the same way on this point (though Bacon takes this as evidence that Marcus cannot be speaking of names in the ordinary sense, but rather must be proposing a drastic regimentation of ordinary language, an interpretation 1 set aside for the sake of argument in the introduction to this study). While Quine in the passage just quoted asserts that even if names are not descriptions, true identity statements involving names can be empirical, Kripke [1972] attempts to explain how this can be so, and thereby answer what for Frege and Russell has been the main argument for the descriptive theory. Assuming Marcus meant what she said to apply to ordinary names, then what she meant did in that respect differ from what Russell meant. But if Marcus differed from Russell only in this respect - if Marcus were just a kind of fundamentalist, taking literally words not meant literally when originally uttered by Russell - then her originality would in an obvious sense be minimal, and more importantly, the atronomical and literary examples would show her

MARCUS, KRIPKE, AND NAMES

103

to be making a straightforward mistake, not a positive contribution. If she is to have made an original, positive contribution, she must differ from Russell in more respects than one. How she must differ from Russell if she is more specifically to have made an original, positive contribution anticipating Kripke needs to be considered more closely. 4. KRIPKE

A claim has lately been heard to the effect that "almost all the main ideas and arguments" of the new theory are to be found presented in very concise form in Marcus [1963], anything added in Kripke [1972] being mere elaboration, illustration, or simple repetition. But this is quite false. There are many items in Kripke, to say nothing of the other new theorists he cites, that are patently or admittedly not in Marcus. In considering the alleged influence of Marcus [1963] on Kripke [1972], one must be careful to subtract off everything that is in Kripke that is thus not in Marcus. A partial listing of such items follows. First off, there is an extensive if informal survey of the literature. This covers different versions and variants of the descriptive theory, and - since no one argument or example suffices against all versions and variants - different types of counter-arguments and counter-examples already in the literature. The more original contributions beyond the existing literature begin with newer examples and arguments. These include counterfactual arguments to be considered later, as well as epistemological arguments from ignorance, according to which names are often used with the speaker having in mind no uniquely identifying description, and epistemological arguments from error, according to which names are often used with the speakers having in mind an incorrect one. Besides negative arguments against the descriptive theory, there is a positive alternative theory, the historical chain account, on which a name is initially bestowed on a thing specified by description, and on each subsequent occasion is used with the intention of continuing to refer to what it has been being used to refer to, so that the circumstances of the earliest uses may soon be forgotten, though the name still succeeds in designating what it does only through a chain of usage extending back to the initial bestowal of the name. (As a result, though names do not have descriptive content, different names for the same thing, since they involve different chains of usage, still indicate their common bearer in different ways, the historical details of which may be unknown to current users of the names, explaining why such names cannot be recognized by mere reflection to have the same bearer.) And besides this material on proper names there is a generalization to common nouns, which establishes the connection with the work of other new theorists, who mainly considered common nouns, demonstrative pronouns, and definite descriptions rather than proper names. In addition to the theory proper, there are applications thereof to problems of analytic metaphysics. As already mentioned, there is no survey and indeed no citation of the

104

JOHN P. BURGESS

literature in the part of Marcus [1963] about names. As for the epistemological argumentation, Marcus herself [1993, p. 211] attributes it to Kripke. The index in Marcus [1993, p. 258] contains several listings under "causal (historicalchain) account of reference (proper names)", and these all lead to passages where Marcus attributes the historical theory to Kripke and/or others; and though she sometimes seems to neglect differences between Kripke's version and others, she never claims to have had any version of her own. As to Kripke's generalization to common nouns, to say nothing of the related work of other new theorists mentioned by Kripke, notably Putnam, these again are patently not in Marcus, who considers only proper names. The applications involve this generalization and thus are also not in Marcus. Opinions inevitably will differ as to the comparative importance of different components of the new theory, but the importance of those I have just identified seems undeniable. Marcus herself seems to concede the importance of the survey; and certainly she concedes the importance of the epistemological argumentation, which she describes at the place cited as "interesting and decisive". The importance of the positive account is certainly widely recognized, so much so that "historical" theory or "causal-historical" theory is a common alternate label for the new theory as a whole, and the importance of the generalization to other expressions besides names also seems to be recognized. A major scientific or philosophical theory is a complex whose components are more significant for being connected with each other than they would be considered in isolation. That is part of the reason why the methodological approach that considers components in isolation from their context and looks to see if someone, somewhere, sometime had once said something not unlike, either as a passing remark, or as a component of an overall very different and less successful theory, produces such bad historiography. Certainly the new theory of reference does not consist of a half-dozen isolated, discrete ideas, but rather of an intricately interconnected network of theses and arguments. The significance of any particular argument against the descriptive theory is enhanced by being conjoined with other arguments like the older ones surveyed and the newer ones introduced by Kripke. All these negative arguments are given added significance by the presentation of a positive alternative theory. And all this discussion of the - after all, rather highly specialized - topic of proper names is given added significance by the generalization to other classes of expressions. And indeed all the discussion of the - after all, still somewhat specialized - topic of reference is given added significance by its applications. It remains to consider counterfactual arguments in Kripke [1972]. These are based not like Mill's arguments on considerations of how we would speak in a counterfactual situation, but rather on considerations of how we do speak of a counterfactual situation, the central notion being "rigidity". In speaking of a counterfactual situation, we use "Dartmouth Town" to designate the same thing we use it to designate when speaking of the actual situation, but "the

MARCUS, KRIPKE, AND NAMES

105

town standing at the mouth of the Dart" to designate whatever town if any would in the counterfactual situation stand at the mount of the Dart, rather than the town that does in the actual situation stand at the mouth of the Dart In Kripke's terminology, names are used "rigidly", descriptions as a general rule "non-rigidly", (Exceptions to the general rule include descriptions involving special adverbial modifiers, such as "the town actually standing at the mouth of the Dart".) A corollary is that true identity statements involving names are counterfactually stable in the sense that they could not have failed, no matter what If it is true that Ahasuerus is Xerxes, then Ahasuerus (alias Xerxes) could not have been anyone other than Xerxes (alias Ahasuerus). This follows from rigidity: Two names that are so used that they designate the same thing when speaking of the actual situation are so used that they designate the same thing when speaking of any counterfactual situation, because each is used when speaking of the counterfactual situation to designate the same thing it is used to designate when speaking of the actual situation. (By contrast, true identity statements between a name and a description, or between two descriptions, generally are not thus stable, with the same exceptions as for rigidity. Ahasuerus alias Xerxes needn't have been the Persian king deceived by Haman, or the Persian king defeated by Themistocles.) A further corollary is that true identity statements involving names are a posteriori necessities. They are "necessary" in the sense just indicated of being counterfactually stable, but they are a posteriori in the sense of not being discoverable by armchair methods, as the astronomical and other examples of Frege, Russell, and Quine from the preceding section show. These are only the simplest examples: The class of "necessities" neither is contained in nor contains the class of a priori implications of true identity statements involving names, since there are also more substantial examples of a posteriori necessity, the most famous being, "water is a compound of hydrogen and oxygen", and examples of the inverse phenomenon of a priori contingencies, the most famous being, "the standard meter rod is one meter long". There patently is no explicit discussion of rigidity - no explicit discussion of cross-comparison between actual and counterfactual situations, or between one counterfactual situation and another - in Marcus. Moreover, the index to Marcus [1963, p. 264] contains several listings for "rigid designators", all leading to passages where Marcus objects to Kripke's emphasis on rigidity. But there is a distracting side-issue connected with the model-theoretic construction in the last section of Marcus' talk. A model for a formal language with modalities is a set of models for the language without the modalities, and Marcus treats constants of the language as designating the same thing in each model in the set If the models are thought of as formal counterparts of counterfactual situations, and the constants as formal counterparts of proper names, then this treatment of constants is a formal counterpart of treating names as rigid. But it is noted in

106

JOHN P. BURGESS

Cocchiarella [1984], and indeed is well known, and even is acknowledged by Marcus [1963, p. 96n], that this kind of model-theoretic construction goes back to Carnap. So if the rigidity argument were attributable to an author merely on the strength of using such a model-theoretic construction, it would have to be regarded as having originated not with Kripke and not with Marcus but with Carnap (and it would have to be said that Kripke arrived at the rigidity argument not in "Naming and Necessity' but in his much earlier work on model theory for modal logic, which had already been announced though not yet fully published at the time of the colloquium, and was cited in the discussion by Marcus as an alternative to Carnap's approach that she could have used instead). But this is absurd: What matters so far as attribution of the rigidity argument is concerned is not that one implicitly treats symbols in an artificial formal language as rigid, but that one explicitly emphasizes that names in unformalized natural language are rigid. There can be no question of Marcus having arrived at Kripke's stability argument, which is to say of her having arrived by Kripke's route at the doctrine that true identity statements involving names are counterfactually stable, since in Kripke's argument the stability property is a corollary of an explicitly emphasized rigidity property. There can be no question of Marcus having arrived by Kripke's route at recognition of the phenomenon of a posteriori necessity, either, for the same reason, even apart from the fact that the more substantial kind of examples of the phenomenon, and examples of the inverse phenomenon, which both involve expressions other than names, are patently not in Marcus. These observations leave open that Marcus may have arrived by some different but cogent argument at the conclusions that true identity statements involving names are counterfactually stable, and that they are examples of a posteriori necessity. Thus it is a question of Marcus' perhaps anticipating, in a variant version and by a different argument, and disconnected from the context that gives them a considerable part of their importance, one or two of Kripke's significant conclusions. But did she anticipate even this much? 5. MARCUS

No one denies that Marcus [1963] argues, in passages already quoted, roughly as follows: Suppose a = b is a true identity, where a and b are names. Then a and b should be intersubtitutable in all pertinent contexts. But necessarily a = a, so subtituting, necessarily a = b. Nothing analogous holds generally for descriptions. (To be quite precise, Soames does question that the foregoing line of thought should be called an argument for distinguishing names from descriptions, since the prior assumption that names differ from descriptions by designating their bearers immediately seems to be what underlies the intersubtitutability claim. But let me waive this objection.)

MARCUS, KRIPKE, AND NAMES

107

The first interpretive issue is what is meant by "necessary" here. On the one hand, if "counterfactually stable", as clearly distinguished from other notions, is what is meant, then Marcus' argument does partially anticipate Kripke to the extent discussed at the end of the preceding section. On the other hand, if what is meant is "discoverable to be true by linguistic analysis or other armchair methods", then the argument is unsound; while if what is meant is sometimes the one, sometimes the other, with no clear distinction being maintained between the two, then the argument is equivocal or confused. Likewise the argument is unsatisfactory if what is meant is that true identity statements involving names hold "necessarily" in every sense of "necessity" (logical, physical, deontic, doxastic, and so on) considered in the literature on modal logic (or alluded to by Marcus in the preamble to her talk). To put the matter another way, an argument of the above type after all occurs already long before Marcus in Russell, and to get from the relevant bits of Russell to the relevant bits of Kripke, two changes must be made simultaneously, from Russell's "logically proper names" to ordinary names, and from Russell's "tautology" to counterfactual stability. Making just the first change without the second does not get one half-way to Kripke in any interesting sense, but merely changes views of Russell that are subtle if debatable into views that are straightforwardly false, and that Russell himself recognized to be so. Thus Marcus can be credited with a version or variant of the stability argument, and an original and positive contribution beyond Russell, only if she can also be credited with some version or variant of the distinction between "necessity" in the sense of counterfactual stability and other notions of "necessity" adopted by Lewis, Carnap, and much of the early literature on modal logic - only if she can be credited with recognizing that possibility in the "metaphysical" sense of what (is or isn't but) potentially could have been the case is not to be analyzed as or conflated with possibility in the "logical" sense of what it is not logically or analytically self-contradictory to assert or assume actually is the case. A second interpretive issue is what is meant by "pertinent contexts". Regardless of what sense of "modal" is attributed to Marcus, if her grounds for taking names designating the same thing to be intersubtitutable in "modal" contexts are simply that this is a special case of their being intersubstitutable in arbitrary contexts, then her argument is not sound, and certainly is not an anticipation of Kripke. For according to Kripke, as according to Quine and Russell and Frege before him, ordinary names designating the same thing simply are not intersubstitutable in all contexts. In particular they are not intersubstitutable in epistemological contexts or in epistemic contexts. "Ahasuerus" and "Xerxes" are not intersubstitutable in the context, "It is discoverable by linguistic analysis or other armchair methods that x is the same person as Xerxes", or in the context, "All early twentiethcentury scholars knew that x is the same person as Xerxes". (Indeed, if the participants at the 1962 colloquium had turned to the latest (1959) edition of

108

JOHN P. BURGESS

the Petit Larousse, they would have found it still undecided as to whether Ahasuerus [Assuerus] was Darius, Xerxes, or Artaxerxes.) So even if Marcus has the right sense of "possibility" and 'necessity", she is reaching the right conclusion for very wrong reasons, and hardly making a positive contribution, unless she also has the right range of "pertinent contexts". (Recall that in Kripke everything is a corollary of rigidity, which is a notion pertaining specifically to counterfactual contexts from the beginning.) No one denies that an exception for epistemic contexts is made in later writings of Marcus, and that a distinction between being discoverable a priori to be true, and expressing a necessary state of affairs, is made in still later writings. The question is, how much of this represents development and evolution in Marcus' views, subsequent and perhaps partially in response to Kripke, and how much was already present in Marcus [1963]? Taking up the first interpretive question, the strongest evidence that counterfactual stability was not being distinguished from logical validity, linguistic analyticity, epistemological aprioricity, and related notions in Marcus [1963] is the negative evidence: the evidence of what is not said. Marcus clearly, distinctly, explicitly, and repeatedly insists on Mill's distinction between naming and describing, but says nothing clear, distinct, or explicit about there being any equally important but much less familiar distinction of her own between the truth of a sentence being discoverable by linguistic analysis or other armchair methods on the one hand, and the fact or state of affairs expressed by a sentence being counterfactually necessary on the other hand. Marcus' silence about any such distinction is the more eloquent given the context of debate with Quine. For consider what was Quine's objection in his commentary: that "Hesperus" cannot be known to designate the same thing as "Phosphorus", and hence that "Hesperus is Phosphorus" cannot be known to be true, simply by linguistic analysis or other armchair procedures, but only through empirical investigations. Manifestly, in offering this observation as an objection to Marcus, he is assuming that Marcus means by "necessary" something implying "discoverable to be true by non-empirical methods". If she had in mind some distinction between notions of this kind and the notion of what "could not have been otherwise". Quine's interpretation would be a major misunderstanding of her position, and one would expect her to clarify her position by explaining the distinction she is making in the post-colloquium discussion, in which she is given the first word. But she does not and instead begins by defending a minor example, not pertaining to names. Moreover, on anything like the way of making the distinction to be found in Marcus' later writings (to be discussed in the next section), Quine'S observation, when presented as an objection to Marcus' position, would involve a straightforward confusion between the question of the (epistemological) status of a sentence (or "sentence-like proposition") and the question of the (metaphysical) status of the fact (or "state of affairs") it expresses. It would be a simple instance of confusing expression and expressed, or mention and use -

MARCUS, KRIPKE, AND NAMES

109

the very confusion of which he so often has accused modal logicians in general, and of which he accuses Marcus in particular in his commentary, in connection with the minor example alluded to above. Given how very concerned Marcus shows herself to be in the colloquium talk with Quine's charge that modal logic rests on a confusion of use and mention - it is the very first thing she mentions in her talk, as well as the first topic she takes up in the post-colloquium discussion, and she remains preoccupied with it even in the prefatory note Marcus [1993a] to the the reprinting decades later - if she had really held at the time a view according to which QUine's objections to modal logic rest on a kind of confusion of use and mention on Quine's part, it is virtually inconceivable that she would not have announced this supreme irony with much sounding of the trumpet and beating of the drum, which would have made the colloquium an even more legendary affair that it has become. In addition to what Marcus does not say, there is also what she does say. Just consider the quotations already produced from the colloquium talk. For instance, she does not just say in (G) if "Morning Star" or "Evening Star" or some other expression a is being used as a genuine proper name, then it is not an empirical fact that Venus is a. Rather, she adds in the almost immediately following passage (E), than an identity in the strict sense, an identity statement involving names, such as "Venus is a", is a tautologically or analytically true. More generally, she uses the following expressions interchangeably in the quoted passages and throughout her talk:

() ()

necessary is tautologically true is analytically true

()

it is not an empirical fact that 0 is valid

( )

IS

(This last item is not in Marcus [1963], but it is what is said in Marcus [1993al to have been meant by "tautology", and what is actually substituted for "tautology" at several places in the reprinting.) This is not the diction one would expect from someone who was making a clear distinction of the kind under discussion, who was distinguishing necessity in the sense of a metaphysical feature of facts from epistemological features of sentences like aprioricity. Several of the items have a distinctly epistemological ring, and Marcus seems regularly to use "empirical" as other philosophers use "a posteriori", which Latinism Marcus herself avoids. Several of the items have a distinctly syntactic ring also, and "valid" is a technical term of art in logical theory that definitely applies to sentences, not states of affairs. The point is not that Marcus does not anticipate Kripkean terminology for

110

JOHN P. BURGESS

marking the relevant distinction, for she does not fully adopt it even in her later writings where she is uncontroversially making a distinction of the kind in question. The point is that she does not anticipate even the Marcusian terminology of her later papers, where for instance "necessary" and "contingent" and are applied only to states of affairs, and "true" and "false" only to sentences. Or rather, the point is that she does not maintain any terminological distinction at all. Besides what Marcus says in her talk, there is what she says in the postcolloquium discussion. (And it should be noted that, as it appears in the published version, the remarks there can be no mere slips of the tongue, since as reported in Marcus [1993c], the transcript of the tape was circulated to the participants for editing before publication.) The most important statement in the post-colloquium discussion is perhaps the answer from Marcus to the following question from Kripke (p. 115): ... does [your view] not require that for any two names 'A; and 'B', of individuals, 'A = B' should be necessary, if true at all? And if 'A; and 'B' are names of the same individual, that any necessary statement containing 'A; should remain necessary if 'A' is replaced by 'B'?

Her answer turns out to be - apart from an "oh, no!" interjected during a subsequent remark to Quine's - Marcus' last word in the discussion, and appropriately so, since it is no minor aside but her response to the only direct request for confirmation of her doctrine on a controversial point. In this response to a question that provides a golden opportunity to elaborate on how her sense of "necessity" differs (if it does) from the sense assumed by Lewis, Carnap, Quine, or others, she first says the following (p. 115): (I)

We might want to say that for the sake of clarity and ease of communication that it would be convenient if to each object there were attached a single name. But we can and do attach more than one name to a single object. We are here talking of proper names in the ideal sense, as tags and not descriptions. Presumably, if a single object had more than one tag, there would be a way of finding out such as having recourse to a dictionary or some analogous mode of inquiry, which would resolve the question as to whether the two tags denote the same thing.

This response is entirely in harmony with the assertion in passage (E) that true identity statements involving names are analytically true, which it is very difficult to read as saying anything other than that their truth can be discovered by linguistic analysis or reflection on meaning. This response made a considerable impression on Kripke, who subsequently always cited Marcus' view as the view that "the truth-values of identity statements between tags are ascertainable merely by recourse to a dictionary". (Though there are all sorts of works with "dictionary" in their title, philosophers often use "what belongs in a dictionary versus what belongs in an encyclopedia" as a colloquial way of putting "analytic versus synthetic", and Kripke clearly took Marcus to be referring to a dictionary of lexical rather than encyclopedic typel Marcus goes on to conclude her contributions to the colloquium with the

MARCUS, KRIPKE, AND NAMES

111

following formulation: (J)

If 'Evening Star' and 'Morning Star' are considered to be two proper names for Venus, then finding out that they name the same thing as 'Venus' names is different from finding out what is Venus' mass or its orbit. It is perhaps admirably flexible, but also very confusing to obliterate the distinction between such linguistic and properly empirical procedures.

It is well nigh impossible to read this as saying anything other than that the kinds of inquiry that would be involved in finding out that "Hesperus" names the same planet as "Venus", and hence that "Hesperus is Venus" is true, are "linguistic", and very different from the kinds of procedures that would be involved in finding out what is the mass of Venus, which are "empirical". This simply is not a formulation that would be offered by a person who was clear on the distinction between counterfactual necessity and epistemological aprioricity - unless she for some reason wished to conceal her view and give a contrary impression, in which case she could hardly make a more effective choice of words. Taking up the second interpretive question, again the strongest evidence that Marcus took names designating the same thing to be intersubstitutable in arbitrary contexts is negative, and consists in her failure to state explicitly an exception even for epistemic contexts; but there is also positive evidence. Consider the repeated emphasis on intersubstitutability anywhere, on universal intersubstitutability in passage (H). And consider quotation (D), the only passage making an explicit exception to intersubstitutability, an exception for contexts about the names as names, as words. (For instance, "Xerxes" and "Ahasuerus" are not intersubstitutable in the contexts "x was called something like that in Greek" or "x was called something like that in Hebrew".) Surely many would think that an exception for such contexts would go without saying. Perhaps some would think that an exception for epistemic contexts would go without saying also. But who would think that the former exception needs to be stated explicitly, while the latter can go without saying? Marcus has not simply forgotten about the existence of epistemic contexts, since she was discussing them just a few pages earlier, and will be discussing them again just a few pages later. For if one reads on beyond the discussion of names, one finds Marcus discussing weaker equivalence relations than identity - which as already mentioned for Marcus as for Smullyan applies only to names, not descriptions - that apply to other expressions than names. And one finds her saying that the weakest such equivalence relations do not guarantee intersubstitutability in modal contexts, that stronger ones do guarantee intersubstitutability in modal but not in epistemic contexts, that there may be ones strong enough to guarantee intersubstitutability even in epistemic contexts, and adding (p.87): (K) But they too would be short of identity, for there are surely contexts in which substitutions allowed by such stronger equivalences, would convert a truth into a falsehood.

112

JOHN P. BURGESS

Surely a contrast is implied with identity (between names), which never converts a truth into a falsehood in any context. (And this is entirely in harmony with the general Russellian background. As Prior [1967a, p. 11] says, there is no reason to be concerned about what kinds of "contexts" one is considering if one is operating with terms assumed to have the features of "logically proper names".) Indeed, if one reads the whole paper, one will find that this is actually its main theme, and the topic indicated in its title, as it was the main theme and the topic indicated in the title of the earlier paper Marcus [1960]: Weaker equivalences (between expressions other than names) only guarantee intersubstitutability in a restricted range of contexts, whereas there is no such limitation on identity (between names); and if the range of contexts admitted in a language is restricted to such a degree that some equivalence weaker than identity guarantees intersubstitutability in all of them, then to that degree the language "extensional" rather than "intensional". What has been at issue thus far is the nature of Marcus' contribution in her colloquium talk. As already partially indicated, no one denies that in later writings Marcus unambiguously concerns herself with natural language and the role therein of names in the ordinary sense, with her own conception of a kind of necessity distinct from aprioricity, and above all with subtleties about knowledge and belief statements, in connection with which she takes a distinctive position. Though detailed consideration of this later work is beyond the scope of the present study, some passages are worth noting briefly for their contrast with what she says in her colloquium talk. 6. MARCUS VERSUS MARCUS

To begin with, in Marcus [1971] an exception to intersubstitutability for epistemic contexts is hinted immediately after the passage quoted in the introduction to this study, and then explicitly stated in a footnote a bit later (p. 6In): We are excluding here epistemic contexts along with stronger obliquity in the formation of predicates. For example, 'John' and 'Jill' may both be replaced in 'John might have married Jill', but only 'John' may be replaced in 'John knew Jill left town' or 'John wishes Jill would marry him',

What is significant about this footnote for the present study is the conspicuous absence of anything even remotely like it from Marcus [1963]. In Marcus [1983], the positing of a realm of "states of affairs" (distinguished from sentences or "sentence-like propositions") and the postulation that sentences differing only by using different names for the same thing express the same "state of affairs" are in effect presupposed, and a distinction is then drawn between a sentence's being discoverable non-empirically to be true and a sentence's expressing a necessary state of affairs: It can only be discovered empirically that "Hesperus is Phosphorus" is true, but the state of affairs of

MARCUS, KRIPKE, AND NAMES

113

expressed by "Hesperus is Phosphorus" is the same as that expressed by "Phosphorus is Phosphorus", and is necessary. This is certainly not Kripke's doctrine of a posteriori necessity, which does not depend on positing a realm of "states of affairs" or postulating special properties for such entities, but which does include many examples showing that the class of necessities neither contains nor is contained in the class of a priori implications of true identities involving names. Indeed, this is arguably not literally a doctrine of a posteriori necessity at all, since there seems to be no single item that is both a posteriori and necessary. It is the sentence (or associated "sentence-like proposition") that is known a posteriori, but the fact (or "state of affairs") that is necessary. Yet this may at least be said to be something very like a doctrine of a posteriori necessity: Something of the flavor of the theory is given by the following passage (Marcus [1983, p. 152]): Nor can the predicament be explained as a conceptual confusion to be resolved by a standard lexicon. Proper names once given are fixed in their semantic values by the language viewed as an historical institution evolving over time. But it is not arbitrary what, once fixed, those values are. It may not be known to or believed by an entire community of speakers of a language in a given historical time slice, that two proper names have the same semantic value, but they may have the same value nevertheless. An encyclopedia would simply be mistaken if it claimed the contrary. On the assumption that the Hesperus-Phosphorus story is accurate, an encyclopedia, published before certain astronomical discoveries, might claim that Hesperus was different from Phosphorus and be endorsed by universal assent to the sentence 'Hesperus is different from Phosphorus', but that claim would be mistaken.

But again this is just the sort of statement that is very conspicuously absent from Marcus [1963]; and indeed it is difficult to read the foregoing as anything but a direct contradiction or retraction, twenty years later, and ten years after the publication of Kripke [1972], of the doctrine advanced in quotations (I) and (J) in response to Kripke's request for clarification of Marcus' position in 1962. (Marcus goes on to say: "In fact, I shall urge, it cannot be believed that Hesperus is different from Phosphorus, appearances or encyclopedias to the contrary". This last assertion is explained and/or justified by saying that one cannot enter into the "belief-relation" with the impossible "state of affairs" of Hesperus alias Phosphorus not being the same as itself. I am not sure whether, in view of this last assertion, Marcus still thinks it needful to make the exception for epistemic contexts made in Marcus [1971], or would now defend the view that even Racine "knew" that Ahasuerus is Xerxes, despite his asserting in his Esther that Ahasuerus was Xerxes' father Darius. But surely it would be too far-fetched to attempt to explain the absence of explicit exceptions for epistemic contexts in 1962 by supposing that at that time she already held the 1983 doctrine, from which she subsequently deviated in 1971, only to return to it again by 1983. Of course the doctrine in Kripke [1972] does not depend on adopting any such controversial views about belief or knowledge.) Among her latest writings, Marcus [1990] is of interest in a different way

114

JOHN P. BURGESS

from the texts considered so far in this section. In that paper Marcus applies an intriguing historiographical methodology to Quine [1953/6l/80]. Passages that are changed from one version to the next are made the focus of attention, as being exceptionally revealing as to Quine's position at one time and its differences from his position at a later time. This inevitably invites attention to the changes that Marcus herself makes when she reprints her own works, and in particular to the changes that occur in the crucial pages 82-85 of Marcus [1963] when these reappear as pages 10-13 in Marcus [1993b]. The one change Marcus herself mentions in Marcus [1993a] is the one I have already mentioned in passing, the occasional change of the obsolescent (and characteristically, though not exclusively, Russellian or logical atomist) term "tautology" to "valid" at several places, notably passage (F). Beyond this, Marcus says that she makes only "nonsubstantial" changes, but I take this to mean no more than that she makes only changes that· she considers to be merely clarification of what, according to her present recollection, she had in mind when she spoke. It should be needless to say that in an historical study the accuracy of such recollections cannot be taken for granted without corroboration, and much more importantly, that in a study of a speaker's alleged influence on a member of her audience it must be recognized that what .he speaker had in mind was available to her audience only through her words. Thus changes that might from some other standpoint seem insubstantial may seem more significant from the standpoint of an historical study of alleged influence. The kinds of changes Marcus makes are indeed less extensive than the changes Quine makes, and in a passage like (G) are so trivial that it makes no difference whether one reads the 1963 or the 1993 version. However, the changes in some other passages need to be noted, since my interpretation has put some weight on their wording in the original, and since some of them have sometimes been quoted in other than their original form elsewhere. Two changes that seem to be possibly significant from the standpoint of the present study are the deletion from quotation (D) of the exception to the principle of intersubstitutability for contexts about names as names, whose presence I suggested made more conspicuous the absence of any similar exception for epistemic contexts; and the insertion in quotation (H) of a restriction to modal contexts, so that instead of saying as originally that names designating the same thing "ought to be anywhere intersubstitutable", an assertion whose unrestricted character I suggested was significant, the second sentence of the passage now says they "ought to be anywhere intersubstitutable in modal contexts". That Marcus considers such changes "nonsubstantial" I take to be indirect testimony as to her present recollection of what she meant by what she said thirty-odd years ago, but I will defer consideration of her autobiographical testimonies - for there are other, more direct ones - until the next section. Where Marcus would wish to make a change in Marcus et al. [1963] that she herself acknowledges to be substantial, she does not alter the text, but rather

MARCUS, KRIPKE, AND NAMES

115

appends a footnote [1993d]. This is attached in quotation (I) to the word "dictionary". This footnote is described in Marcus [1993c] as providing both "clarification and correction". It reads as follows: Since such entries are usually described as "nonlexical", the dictionary here is functioning as an encyclopedia. But this whole passage needs clarification and emendation. As indicated in the earlier text, discovering that Hesperus and Phosphorus have the same path is an empirical discovery that entails that they are identical and hence that 'Hesperus' and 'Phosphorus' name the same thing. But the identity, once given, is necessary.

I find this, despite its being identified as partly "correction" or "emendation", potentially misleading about "the earlier text", inasmuch as it might well be read as suggesting that there is some unspecified place in the original talk or discussion where Marcus has said of the names "Hesperus" and "Phosphorus" that the discovery that they name the same thing is empirical. But there is no such place. Marcus does say of the descriptions "the evening star" and "the morning star" that the discovery that they describe the same thing is empirical, but she says this in quotation (G) as part of an account of how descriptions are supposed to differ from names. And of course it is said by Quine that the discovery that the names "Hesperus" and "Phosphorus" name the same thing is empirical, but this is said in his commentary, which of course is not reproduced in Marcus [1993], and is said there as an objection to Marcus. In this section I have been considering Marcus' later substantive writings on names, nece~sity, belief, and related topics. But as already indicated one needs to consider also her later autobiographical writings. 7. MARCUS VERSUS KRIPKE

Marcus [1990a], a companion piece to Marcus [1983], contains an autobiographical footnote (p. 248n) mentioning Kripke's remarks in the 1962 discussion and adding: [We] can see in those 1961 [sic] remarks Kripke's move towards his theory of "rigid designators".

This directly contradicts Kripke's own account, in which his views on "rigid designators" and other aspects of the theory of names are dated only from 1963-64, not 1961-62, and in which Marcus' talk is not cited as a significant positive influence. The autobiographical reminiscences Marcus [1990] also contains a passage (pp. 226-227) in which mention of Kripke's remarks in the 1962 discussion is immediately followed by the assertion: My claims about the special semantical role of proper names reemerged more widely in the early 1970s in theories of direct reference.

This contradicts the account of most commentators, according to which the claims about proper names incorporated in what is variously called the "direct" or "new" or "causal" theory of reference of the early 1970s are attributed mainly to Kripke, and not to Marcus. These somewhat oblique formulations

116

JOHN P. BURGESS

are the most explicit suggestions in Marcus' published writings of a relationship between Kripke [1972] and Marcus [1963] that is not acknowledged by Kripke and not recognized by most commentators. In both these incidental autobiographical passages, Marcus claims that Kripke misunderstood her position at the colloquium in some general way, and more specifically she claims that he misunderstood her remark about a dictionary (to which he alludes in characterizing Marcus' position in Kripke [1972]). Marcus seems to attach considerable importance to the more general claim. (The suggestion seems to be that Kripke initially misunderstood Marcus, only a year of two later in 1963 or 1964 arrived at a position similar to what she had had in mind already in 1961 or 1962, but then many not have recognized that the position at which he had thus belatedly arrived was what she had had in mind all along. This would make Kripke's alleged appropriation of Marcus' idea largely unconscious, and that is perhaps why the matter seems important to Marcus. But she is not explicit.) But the more general claim seems to me largely irrelevant for purposes of the present study insofar as it goes beyond the more specific claim. For if Kripke is right in agreeing with Quine in taking Marcus to be thinking of the "necessity" of identity in terms of what is discoverable to be true by linguistic inquiry as opposed to empirical procedures, then it hardly matters for present purposes whether he was wrong on some other exegetico-hermeneutical point, though I do not think he was. As for the more specific claim about the "dictionary" remark, Marcus returns to it again and again (though always concentrating on the word "dictionary" in quotation (I) or in Kripke's summary of her view, rather than the continuation of the passage in quotation (J)). Her most explicit testimony as to her present recollection of what she meant is to be found in a footnote Marcus [1990e] added to Marcus [1990] in the reprinting: In response to the query about determining who or what was the reference of a proper name, I said one might look in a dictionary. I had in mind something like a biographical dictionary or those parts of a dictionary so headed, or like the Oxford Classical Dictionary, which provides information about the objects named. The direct-reference theory for proper names is bolstered by the fact that linguists generally do not take proper names as lexical items at all. The absence of "content" makes them recalcitrant to "definition".

Straight off, the first clause of the first sentence in the quoted passage shows the unreliability of Marcus' memory. For her remark about the dictionary was not in response to a "query about determining who or what was the reference of a proper name". Kripke's question, which has already been quoted, was simply whether it was her view that true identities between names are necessary. It was Marcus' answer that introduced epistemological issues about "determining" yet another clue that for her "necessity" was an epistemological notion. This initial slip should put the reader on guard when considering the recollection that she meant encyclopedia as opposed to lexicon by "dictionary". Two points must be made about this recollection. First, the recollection is difficult to reconcile with the wording of the full text

MARCUS, KRIPKE, AND NAMES

117

of Marcus' remarks (including (J) as well as (I)), The actual passage in Marcus et al [1963] reads like a straightforward fallacy of equivocation between lexicon-type "dictionaries", which would contain only the results of linguistic inquiry, and encyclopedia-type "dictionaries", which would contain information about alternate names of the same thing - or between lexical entries and encyclopedic entries in a "dictionary" that does not separate the two types of entry - overlooking that the lexicons would not contain proper names, and that the encyclopedias would equally contain empirical information of other sorts, for instance about orbits. For example, if the participants in the colloquium had consulted the Petit Larousse, they would have found no entry for Venus in the first or lexical section, but the following entry in the second or encyclopedic section: VENUS, la deuxieme des planetes qui gravitent autour du Solei!. Son orbit se trouve entre celie de Mercure et celie de la Terre. On l'appelle aussi Vesper ete, communement, etoile du Berger.

One simply does not get a plausible or even a coherent reading if one takes Marcus to have had encyclopedias, consciously distinguished from lexicons, continuously in mind throughout. (Just try crossing out "dictionary", writing in "encyclopedia", and reading the result, not stopping at the end of (I), but going on to (1).) In addition there is indirect evidence. One finds Marcus in the passage from Marcus [1971] quoted in the introduction to this study still speaking of adding names to a "lexicon" and keeping a "lexicon" of names. Perhaps "dictionary" can mean "encyclopedia as opposed to lexicon", but surely "lexicon" itself cannot. (The whole passage from Marcus [1971] seems to me to combine the thoughts of quotations (A) and (I), still overlooking or assuming away the fact that if every asteroid or comet ever sighted is to be given a standard name, it will be needful to settle indefinitely many empirical astronomical problems about the identity or otherwise of bodies sighted on different occasions.) Only after she has, in the passage from Marcus [1983] quoted earlier, dismissed "lexicons" and "encyclopedias'" alike as irrelevant, do we find her distinguishing in print between the two types of reference works (or the two types of entries, lexical and encyclopedic, in a "dictionary" that does not separate the two). For what seem to be Marcus' earliest allusions to the view of unnamed linguists that names to not belong in a "dictionary" in the sense of "lexicon as opposed to encyclopedia", accompanied by the lone citation of Ziff in her collected papers, occurs in Marcus [1986, p. 203n]. Second, the recollection is ultimately irrelevant. It would be relevant if all Marcus had said had been what Kripke says by way of summary of her view. But of course that is only Kripke's summary, and she actually said much more, and all that she said points in the same direction. Serious historians tend not to place much reliance on autobiographical reminiscences from decades after the event, not because of doubts that autobiographers sincerely report what they remember, but because of the

118

JOHN P. BURGESS

notorious falliability of human memory, and especially its well-documented tendency to confuse hindsight with foresight. Marcus' footnote from the early 1990s provide no serious reason to doubt that in the early 1960s she held the doctrine that commentators from Quine onwards have attributed to her, and rightly criticized. CONCLUSION

To summarize, in no sense is it true that Marcus introduced the notion of direct reference, or originated the idea that identity is necessary. Marcus [1963] endorses the very well known, but at the time not very popular, view of Mill that names have no descriptive meaning, as well as some only slightly less well known views of Russell, including the view that true identity statements involving "names" are "tautologies"; and she applies the views endorsed to the interpretation and defense of formalized systems of modal logic, elaborating on early applications of this kind by Smullyan. An exaggerated impression of Marcus' originality seems to have been created in the minds of some recent readers by the absence of any citations of Mill or of (the relevant parts of the work of) Russell or Smullyan in her talk, but what made such citations dispensable in the context of the 1962 colloquium was precisely the familiarity of these views to Marcus' audience at the time. Marcus leaves some ambiguity as to how far what she says about "names" is really intended to apply to names in the ordinary sense, in natural language Mill's views had been intended to be thus applicable, while Russell's views had not been - and if the features that logical atomism attributes to Russellian "logically proper names" were attributed to names in the ordinary sense, the resulting doctrine would be, as Russell recognized, straightforwardly false. Such falsehood could be avoided only if substantial adjustments were made in the sense of "necessity" involved, and substantial restrictions were imposed on the "universal" intersubstitutability of names designating the same thing; but there are no indications of such adjustments or restrictions at the places where one would expect to find them, or indeed anywhere, in the text of Marcus [1963] and Marcus et at. [1963], while there are many counter-indications. Indications of the adjustments and restrictions appear rather in Marcus' writings from ten or twenty years later, after and at least partly in response to Kripke (and partly appear also in alterations, not all signalled in Marcus [1993a], to the text of Marcus [1963] in the reprinting in m [1993]); but their presence in these later places only makes their absence from Marcus [1963] and Marcus et al. [1963] the more conspicuous. Even with such adjustments and restrictions as appear later, Marcus' argument would remain quite significantly different from its closest counterpart in Kripke [1972]. In any case the larger part of Kripke [1972] is without even a remote counterpart in Marcus [1963]. In no sense is it true that Marcus anticipated the main ideas and arguments of the new theory of reference.

MARCUS, KRIPKE, AND NAMES

119

As for any suggestion of unacknowledged dependence of Kripke [1972] on Marcus [1963], that such damaging claims have been broadcast to an audience extending well beyond specialist circles, without careful prior examination of whether they have any foundation in fact, is regrettable. 4 For an examination of the textual evidence, such as I have reported here, shows these claims to be entirely unfounded. APPENDIX: MARCUS VERSUS KRIPKE, CONTINUED

In the body of this study I have set aside Marcus' claims that Kripke was guilty of some general misunderstanding of her position at the 1962 colloquium, concentrating only on the claim that he specifically misunderstood her "dictionary" remark. The more general claim may deserve closer attention since Marcus herself seems to attach great importance to it. At issue is what Kripke meant by a half-dozen or so remarks to Quine about Marcus at the very end of the post-colloquium discussion (pp. 115-116), which remarks according to Marcus show a misunderstanding of her position. These remarks concerned "essentialism", and a complicating factor is that this term was differently understood by Quine and by Marcus. (Kripke, in addressing Quine, naturally used "essentialism" in Quine's and not Marcus' sense.) For while Marcus understands the question whether modal logic is committed to "essentialism" to be question of whether formal laws of a certain type are formally derivable, Quine repeatedly and emphatically asserts that on his understanding "essentialism" is not such a formal feature. What Quine objects to as arbitrary "essentialism" is an inegalitarian treatment of different ways of specifying the same thing, according to which some features of a thing are to be counted as essential and others as accidental to it, even though the latter follows just as analytically from some ways of specifying the thing as the former do from others. This much said by way of background, perhaps the best way to proceed will be to give an explication de texle, clarifying what Kripke meant by his remarks, before considering Marcus' complaint that she was misunderstood. Kripke's series of remarks pursues a single idea from beginning to end. He begins by double-checking that he has correctly understood Marcus' position by asking her whether (X) and (Y) below - he does not seem to think it needful to ask about (Z) - are part of her view of the nature of "tags" a, b, c, ... : (X)

If a and b designate the same thing, then for all pertinent P(x) , P(a) holds necessarily if and only if P(b) holds necessarily.

(Y)

If a and b designate the same thing, then the identity statement involving them, a = b, holds necessarily.

(Z)

For each thing there is (or could be introduced without disrupting (X) or (Y» at least one a designating it.

120

JOHN P. BURGESS

Kripke's own formulation of the question has already been quoted. This is the question to which Marcus responds with the "dictionary" formulation. Kripke, after saying that Marcus seems to him to have "a perfect valid point of view", expects Quine to say that the assumption of "tags" with such properties - among which he cites (Y) specifically - is tantamount to the arbitrary inegalitarianism Quine finds objectionable: It seems to me the only thing Professor Quine would be able to say ... is that the assumption of such a distinction between tags and empirical descriptions, such that the truth-values of identity statements between tags (but not between descriptions) are ascertainable merely by recourse to a dictionary, amounts to essentialism itself.

Quine had been thinking of a special class of descriptions as being the privileged terms used in reducing de re to de dicto modality, but non-descriptive "tags" with the above properties could also be used to settle which open formulas or sentences hold necessarily of a thing: The tags are the "essential" denoting phrases for individuals, but empirical descriptions are not, and thus we look to statements containing "tags", not descriptions, to ascertain the essential properties of individuals.

Since Quine does not at once agree, going off on a tangent, Kripke reformulates the point that if "tags" with such properties are assumed - among which he cites (Z) and (X) specifically this time - then de re modality can be reduced to de dicto modality, according to a definition like the following: (l)

P(x) holds necessarily of a thing if and only if for some a designating it, P(a) holds necessarily.

In Kripke's own formulation: Suppose the assumption in question is right - that every object is associated with a tag, which is either unique or unique up to the fact that substituting one for the other does not change necessities - is that correct? Now then granted this, why not read "there exists an x such that necessarily p of x" as ... "there exists an object x with a name a such that p of a is analytic". Once we have this notion of name, it seems unexceptionable.

As Quine finally does seem to agree, assuming "tags" with such features is tantamount to the arbitrary inegalitarianism he finds objectionable: "Such an assumption of names is equivalent to essentialism". The moderator called the session to a close on this note, before there was time to elaborate, but the reader will readily verify that assumptions (X) and (Z) are just what are needed to establish the equivalence of (1) with: (2)

P(x) holds necessarily of a thing if and only if for any a designating it, P(a) holds necessarily.

and to give various other laws one would expect. (For instance, a conjunction holds necessarily of a thing if and only if each conjunct does.)

MARCUS, KRIPKE, AND NAMES

121

One of the two main passages where Marcus complains of a general misunderstanding of her position by Kripke follows a quotation of the objection from Quine. It reads as follows (Marcus [1990, pp. 226-227]): So, even on the matter of supposing as I did that there were directly referring proper names, it appeared that for Quine the trouble also came down to essentialism, since it suggested that things have their proper names necessarily. During the discussion that ensued after Quine's comments Kripke reinforced Quine'S view with his remark that "such an assumption of names is equivalent to essentialism". But that was not my claim. Socrates might have been named Euthyphro: he would not thereby be Euthyphro.

Now Quine's objection has been quoted in the body the present study, and it says nothing that even remotely suggests that what he thinks is wrong with Marcus' theory is that it implies that having been called "Socrates" is an essential feature of Socrates. Rather, what he thought wrong with Marcus' theory was that it implies that "Hesperus is Phosphorus" is an a priori truth. So Marcus has obviously misunderstood Quine. And in charging that Kripke misunderstood her position as having the consequence that having been called "Socrates" is an essential feature of Socrates, Marcus has equally misunderstood Kripke. In order to see this, for the sake of illustrating the operation of definition (1) or (2), and just for the space of this paragraph, suppose that "Jupiter" is a "tag" for Jupiter, and consider the following assertions about the fifth planet: (3)

"x is the largest planet" holds necessarily of it.

(4)

"x is Jupiter" holds necessarily of it. "x was called 'Jupiter'" holds necessarily of it.

(5)

According to the definition, (3)-(5) above have the same truth values as (6)-(8) below: (6) (7) (8)

"Jupiter is the largest planet" holds necessarily. "Jupiter is Jupiter" holds necessarily. "Jupiter was called 'Jupiter'" holds necessarily.

Presumably (6) is false, and (7) is true, while (8) is false again, given that the planet was called "Zeus" by Greek, and something else again by Babylonian, Egyptian, Indian, Chinese, Mayan, and other astonomers. It follows that (3) is false, and (4) is true, while (5) is false. In another terminology, it is an accidental feature of Jupiter to be the largest planet, and it is an essential feature of Jupiter to be Jupiter, while it is an accidental feature of Jupiter to have been called "Jupiter". Thus the doctrine Kripke and Quine were contemplating involves the view that "tags are the 'essential' denoting phrases for individuals", in the sense that

122

JOHN P. BURGESS

"we look to statements containing 'tags' to ascertain the essential properties of individuals", using them to reduce de re to de dicto modality - not in the sense of the claim that a tag is a phrase it is an essential property of an individual to be denoted by. The other of the two main passages where Marcus complains of a general misunderstanding of her position by Kripke immediately follows Marcus' selfquotation (B) described in the body of this study. It reads as follows (Marcus [1990a, p. 248n]): It should be noted that, on this view, proper names are not assimilated to descriptions, even "rigid" descriptions. Kripke, in [Kripke, [1972)] classifies proper names as "rigid designators" along with rigid descriptions, thereby obscuring the difference in semantical relationship between a proper name and the object named as compared with the relationship between a rigid description and the object described. Kripke in [Marcus et a/. [1963)] interpreted my views as including the position that "the tags are the essential denoting phrases for individuals". That was not part of my account ...

Other commentators have of course found Kripke anything but obscure on the difference between how a description succeeds in designating what it does (by saying something true about it) and how a name does so (through an historical chain of usage). Just what misunderstanding Kripke is being accused of in this passage is rather hard to make out, since Marcus does not explicitly tell us what she thinks Kripke meant by the formulation she quotes. But curiously the misunderstanding of which Kripke is being accused in this passage does not seem to be the same misunderstanding of which he was accused in the other passage. Perhaps the complaint is that Kripke took the feature that one can "look to statements containing 'tags' to ascertain the essential properties of individuals" or use tags to reduce de re to de dicta modality to constitute the whole of Marcus' conception of a "tag", so that there would be no distinction between "tags" and any exceptional descriptions that could also be used in this way. But clearly Kripke insists on this particular aspect of Marcus' theory simply because it is the aspect that is relevant to the concerns of Quine, whom he is addressing. Marcus' apparent failure to grasp this point is most likely a consequence of her failure to grasp - despite the efforst of Kripke, F011esdal, and Quine himself during the post-colloquium discussion - what was the point of Quine's examples of the failure of intersubstitutability, whose force she herself says (in a passage to which she calls attention in Marcus [1993a]) she has never appreciated. Department of Philosophy Princeton University

123

MARCUS, KRIPKE, AND NAMES NOTES

Reprinted with very minor changes from Philosophical Studies. 1996,84, 1-47. Notably Smith [1995, 1995a]. The recent claims quoted in the first paragraph of §1 and the first paragraph of §4 below both come from Smith [1995a, p. 128). The parenthetical suggestion in the second paragraph of §7 below is made explicitly in the last couple of pages of Smith [1995]. 3 At an editor's suggestion, let me note that at this point I am only describing Kripke's impression of passage (I). I will return to passage (I) later, and consider Marcus' complaints in recent years that Kripke's impression was a misimpression. Perhaps I should add that the impression Kripke retained in later years of this passage (I) from the Boston colloquium will have been colored by what Marcus said six months later at a Helsinki conference, as quoted in my "How Not to Write History of Philosophy: A Case Study". 4 Soames [1995) not inappropriately uses a stronger word.

2

REFERENCES Bacon, J.: 1995, review of Marcus [1993), Journal of Symbolic Logic, 60, 1005-1009. Barrett, R.B. and R.F. Gibson.: 1990, Perspectives on Quine. Blackwells, Oxford. Church, A.: 1950, review of Fitch [1949), Journal of Symbolic Logic, 15, 63. Cocchiarella, N.: 1984, 'Philosophical Perspectives on Quantification in Tense and Modal Logic', in Gabbay and Guenthner [1983-89), 2,309-353. Davidson, D. and Harman, G.: 1972, Semantics of Natural Language. Reidel, Dordrecht. Davidson, D. and Hintikka, J.: 1972, Words and Objections: Essays on the Work of Quine. Reidel, Dordrecht. Fitch, F.: 1949, 'The Problem of the Morning Star and the Evening Star', Philosophy of Science, 16, 137-141. F0llesdal, D.: 1986, "Essentialism and Reference", in Hahn and Schilpp, 97-113. Frege, G.: 1970, (trans. and eds., P. Geach and M. Black), Translations from the Philosophical Writings of Gottlob Frege. Basil Blackwell, Oxford. Frege, G.: 1970a, (trans. M. Black), 'On Sense and Reference', in Frege [1970), 56-78. Gabbay, D. and Guenthner, F.: 1983-89, Handbook of Philosophical Logic, 4 vols. Reidel, Dordrecht. Geach, P.: 1957, Mental Acts. Routledge and Kegan Paul, London. Geach, P.: 1972, Logic Matters. University of California Press, Berkeley. Hahn, L.E. and Schilpp, P.A.: 1986, The Philosophy of Quine. Open Court, La Salle. Kaplan, D.: 1969, 'Quantifying In', in Davidson and Hintikka [1969), 178-214; also in Linsky [1971), 112-144. Kneale, W: 1962, 'Modality, De Dicto and De Re', in Nagel et al. [1962), 622-633. Kripke, S.A.: 1971, 'Identity and Necessity' in Munitz [1971), 135-164. Kripke, S.A.: 1972, 'Naming and Necessity: Lectures Given to the Princeton University Philosophy Colloquium, January 1970', in Davidson and Harman [1972), 253-325 and 763-769. Kripke, S.A.: 1979, 'A Puzzle About Belief, in Margalit [1979), 234--283. Kripke, S.A.: 1980, Revision of Kripke [1972) with a new preface (Cambridge: Harvard University Press). Linsky, L.: 1969, 'Essentialism, Reference, and Modality', Journal of Philosophy, 20, 687-700; also in Linsky [1971), 88-100. Linsky, L.: 1971, Reference and Modality. Oxford University Press, Oxford. Linsky, L.: 1971a, Introduction to Linsky [1971]. Marcus, R.B.: 1947, 'Identity and Individuals in a Strict Functional Calculus of First Order', Journal of Symbolic Logic, 12, 3-23. Marcus, R.B.: 1960, 'Extensionality', Mind, 69, 55-62, reprinted in Linsky [1971), 44-51. Marcus, R.B.: 1963, 'Modalities and Intensional Languages', in Wartofsky [1963),77-86, reprinted with changes in Marcus [1993), 3-23.

w.v.

w.v.

124

JOHN P. BURGESS

Marcus, R.B.: 1971, 'Essential Attribution', Journal of Philosophy, 68, 187-202, also in Marcus [1993],53-73. Marcus, R.B.: 1983, 'Rationality and Believing the Impossible', Journal of Philosophy, 80, 321-338, in Marcus [1993], 143-16l. Marcus, R.B.: 1986, 'Possibilia and Possible Worlds', Grazer philosophische Studien, 25/26, 107133, in Marcus [1993],189-213. Marcus, R.B.: 1990, 'A Backward Look at Quine's Animadversions on Modal Logic', in Barrett and Gibson [1990], 23-243, reprinted with an added footnote in Marcus [1993], 233-255. Marcus, R.B.: 1990a, 'Some Revisionary Proposals about Belief and Believing', Philosophy and Phenomenological Research, 50 (supplement), 133-153, also in Marcus [1993], 233-255. Marcus, R.B.: 1993, Modalities. Oxford University Press, Oxford. Marcus, R.B.: 1993a, Prefatory note to revised version of Marcus [1963], in Marcus [1993], 3-4. Marcus, R.B.: 1993b, Revised version of Marcus [1963], in Marcus [1993], 5-23. Marcus, R.B.: 1993c, Prefatory note to reprinting of Marcus et al. [1963], in Marcus [1993], 24. Marcus, R.B.: 1993d, Footnote added to reprinting of Marcus et al. [1963], in Marcus [1993], 34. Marcus, R.B.: 1993e, Footnote added to reprinting of Marcus [1990], in Marcus [1993], 226-227. Marcus, R.B., Quine, Wv., Kripke, S., McCarthy, 1. and D. Follesdal.: 1963, Discussion, in Wartofsky [1963], 105-116, reprinted with an added footnote in Marcus [1993], 24-35. Margalit, A.: 1979, Meaning and Use. Reidel, Dordrecht. Mill, IS.: 1950, John Stuart Mill's Philosophy of Scientific Methods. Hafner, New York. Munitz, M.: 1971, Identity and Individuation. New York University Press, New York. Nagel, E., Suppes, P. and A. Tarski.: 1962, Methodology and Philosophy of Science: Proceedings of the 1960 International Congress. Stanford University Press, Stanford. Prior, A.N.: 1967, Past, Present and Future. Clarendon, Oxford. Prior, A.N.: 1967a, 'Logic, Modal', in Weiss [1967], 5, 5-12. Quine, Wv.: 1947, 'The Problems of Interpreting Modal Logic', Journal of Symbolic Logic, 12, 4348. Quine, wv.: 1953/61180, From a Logical Point of View. Harvard University Press, Cambridge. Three editions. Quine, W.Y.: 1961, 'Reference and Modality', a second version, in second edition of Quine [19531 61180],130-159; also in Linsky [1971],17-34. Quine, Wv.: 1986, 'Autobiography', in Hahn and Schilpp [1986],1-46. Russell, B.: 1985, The Philosophy of Logical Atomism. Open Court, La Salle. Searle, IR.: 1958, 'Proper Names', Mind, 67,166-173. Searle, J.R.: 1967, 'Proper Names and Descriptions', in Weiss [1967], 6, 487-491. Shapiro, S.: 1985, Intensional Mathematics. North Holland, Amsterdam. Smith, Q.: 1995, 'Marcus, Kripke, and the Origin of the New Theory of Reference'. Synthese, 104, 179-189. Smith, Q.: 1995a, 'Marcus and the New Theory of Reference: A Reply to Scott So ames'. Synthese, 104,217-244. Smullyan, A.: 1947, review of Quine [1947], Journal of Symbolic Logic, 12, 139-141. Smullyan, A.: 1948, 'Modality and Description', Journal of Symbolic Logic, 13, 31-37. Soames, S.: 1995, 'Revisionism about Reference: A Reply to Smith', Synthese, 104, 191-216. Strawson, P.: 1959, Individuals. Methuen, London. Wartofsky, M.W: 1963, Proceedings of the Boston Colloquium for the Philosophy of Science 19611 1961. Reidel, Dordrecht. Weiss, P.: 1967, The Encyclopedia of Philosophy. MacMillan, New York. Ziff, P.: 1960, Semantic Analysis. Cornell University Press, Ithaca.

JOHN P BURGESS

HOW NOT TO WRITE HISTORY OF PHILOSOPHY: A CASE STUDY

INTRODUCTION

Did Saul Kripke in his 1970 lecture series on naming and necessity merely repeat, with elaborations but without acknowledgments, a theory that had been propounded to him by Ruth Barcan Marcus in a 1962 colloquium talk on modalities and intensional languages? That is what seems to be insinuated in footnotes in recent publications by Marcus, and what was explicitly alleged by Quentin Smith in a notorious abstract on the program for the December, 1994, convention of the American Philosophical Association (APA). This damaging accusation or innuendo has been elaborated in two papers, Smith (1995, 1995a). Partial responses have also been given, but only partial responses. The APA permitted Scott Soames to make a response Soames (1995) to Smith's first paper, but imposed limits of length and permitted no similar response to Smith's second paper. Burgess (1996) replies to Marcus' recent publications, but does not deal with Smith. All four of the papers just cited are reprinted in the present collection, and Smith continues to produce new material on the topic. The editors of the present collection invited Soames to submit a further response to Smith. But given schedule constraints it seemed best to him to divide the task of preparing such a reply with me. The present note is my installment. My indebtedness to Soames, acknowledged already in my first paper on the topic, is even greater in the case of this one, but nonetheless the conclusions I draw here are my responsibility and not his. METHODOLOGY

I begin by listing ten reasons why Smith's original paper should never have

been taken seriously as a contribution to historical scholarship, ten ways in which the paper conspicuously violates elementary rules of historiographical methodology. He attempts to reduce a complex philosophical theory to a list of a half-dozen discrete points. His whole presentation is organized around the attempt to reduce the "new" theory of reference to such a list. Since the importance of a complex theory consists at least as much in the interconnections it

125 P. W. Humphreys et al. (eds.), The New Theory of Reference © Kluwer Academic Publishers 1998

126

JOHN P. BURGESS

establishes among its novel elements, and between these and elements it retains from the previous literature, this listing procedure would be an error even if the list produced were a reasonably accurate enumeration of the novelties of the "new" theory. But it is very far from being so, on account of three further methodological errors. He attempts to show that the work of a later writer was anticipated by an earlier writer without seriously examining the work of the later writer to see what points it contains. He simply asserts that his list contains almost all the

main ideas and arguments of the new theory of reference, without seriously examining the contents of Kripke (1972), let alone of the work of other theorists there cited. The result is that he omits from his list important novelties of the "new" theory, treating them as minor elaborations or ignoring them altogether, even in cases where the importance and novelty of the ideas in question has been conceded by Marcus herself in the past, as in Marcus (1978). (A notable example is the historical chain account of the reference of a name to its bearer.) The question of the relation of Marcus to Kripke cannot emerge clearly from Smith's account, because the views of Kripke do not emerge clearly. He attempts to show that a writer introduced important points merely by attempting to show that she endorsed them, without seriously examining the earlier literature to see whether she was merely repeating them. The result is

that he includes on his list items that were not novelties in the "new" theory, and that go back to sources earlier than Marcus, even in cases where the original sources are extensively discussed in Kripke (1972). (A notable example is the idea that proper names lack descriptive meaning, which had been defended by writers from Mill to Ziff.) If one does examine the previous literature, and in particular if one follows up the explicit acknowledgments in Marcus (1960, 1963b) to Fitch (1949, 1950) and in turn the explicit acknowledgments of Fitch (1949, 1950) to Smullyan (1947, 1948), it emerges that the most relevant earlier writers are Russell, with his theory of ideal names, and especially Russellians such as Arthur Smullyan and Frederic Fitch who were less cautious than Russell himself about applying the theory to ordinary names. But the relationship of Marcus to the Russellian tradition cannot emerge clearly from Smith's account because that tradition itself is not seriously examined. He indicates the points whose history he is supposed to be examining by highly ambiguous labels. Especially ambiguous are the labels "direct reference"

and "necessity of identity". The question is not whether doctrines to which these labels might be applied occur prior to Kripke. For there are doctrines that could be called "direct reference" and "necessity of identity" in

HOW NOT TO WRITE HISTORY OF PHILOSOPHY

127

Russellians from Marcus' contemporary Arthur Prior (1963, 1967) right back to F. P. Ramsey and Russell himself. The crucial question is rather whether Marcus merely repeats such Russellian doctrines, or whether she repudiates those doctrines (and in the latter case, if she does so in a way that anticipates later Kripkean doctrines). A concise account of the needed background on the content of Kripke's theory and the contrast with the Russellian theory is supplied in the early sections of my previous paper, and I will not repeat it here, except to include a summary in tabular form. No such background is provided by Smith, and hence the crucial question cannot emerge clearly from his account. Table Russel/ian

Kripkean

Reference of name to its bearer ...

... is absolutely immediate: the meaning of a name is its bearer.

... is not mediated by descriptive meaning, but is mediated by an historical chain of usage.

(2) Using one rather than the other of two names for a thing ...

... does not change what is meant or what proposition is expressed.

[No such claim and indeed no use of the notion of proposition is made on this theory.)

(3) Two names for the same thing are intersubstitutable ...

... in all contexts (except those of quotation).

... in extensional and counterfactual contexts.

(4) "Necessity ofIdentity": True identities involving names are ...

... necessary in any and every sense, including that of being tautologous, analytic, and apriori.

... necessary only in a counterfactual sense, not in the sense of being tautologous, analytic, or apriori .

(5) The necessity of identity in the indicated sense(s) is ...

... an immediate corollary of the intersubstitutability of two names for the same thing in all contexts.

... the conclusion of a separate argument involving crosscomparison of the behavior of the same name in extensional and in counterfactual contexts.

(1) "Direct Reference":

As a result of the four methodological errors just enumerated, Smith approaches the texts asking the wrong question. He simply asks whether there are passages in Marcus (1963) that could be read as endorsements of doctrines to which the labels "direct reference" and "necessity of identity" and so on could be applied. Then in interpreting the passages he selects he commits a series of six further methodological errors.

128

JOHN P. BURGESS

He simply assumes that a writer of decades ago would not have talked herself into any doctrine that would seem obviously wrong to us today. He thus

prejudges the question, since our sense today of what is obvious has been influenced by Kripke, and since the Russellians of decades ago did talk themselves into doctrines that to us today would seem obviously wrong (at least if, where there is ambiguity in their writings, we take them to be speaking of names in the ordinary rather than some ideal sense, as Smith takes Marcus to be speaking, even though there is ambiguity in her writings). This prejudgment betrays an unhistorical way of thinking, since every historian knows that even the greatest philosphers of the past often talked themselves into positions that to later philosophers, in many cases including even their own later selves, have seemed obviously wrong. He ignores context when proposing readings of the texts he quotes. This includes not only ignoring general historical context and the works of other writers of the period, but also ignoring the context of the particular work Marcus (1963) he is interpreting. He omits from quotations words that alter the sense of the words remaining. He not only ignores the global context of the work as a whole, but text immediately preceding or immediately following the passages he quotes. Indeed in some cases he omits words within those passages. And in at least one case his omissions are not clearly indicated by a suspension "" .". He places uncritical reliance on an author's current recollections of what she meant by what she wrote decades ago. Though there is no acknowledgment

to Marcus in Smith's first paper, on comparison with her writings it seems obvious that she is the source of many of Smith's claims; and with his second paper he begins to become more open about the extent to which he is relying on private communications from Marcus. He seems to accept without question everything she now says about what she meant by what she said decades ago, especially where it is attempted to explain away what seem clear endorsements of Russellian as opposed to Kripkean positions. This procedure again betrays an unhistorical way of thinking. For every historian know how easy it is for memory to slip over the course of the decades from "What I didn't think then but came to think later was"." to "What I should have said was"." to "What I really meant but failed to say was"." to "What I said, though not quite as clearly as I should have, was "." to an unqualified "What I said was. ,,". He imports into his readings ofquotations ideas not in the text or surrounding context. In addition to ideas originating from the Marcus of decades later,

he imports ideas of his own invention. And he fails to recognize that, given

HOW NOT TO WRITE HISTORY OF PHILOSOPHY

129

a choice between a reading departing significantly from the text, and an alternative reading avoiding importation of extraneous ideas, the burden of proof is on proponents of the former, not the latter. He accompanies his quotations with tendentious glosses. Thus he not only imports alien ideas into the passages he considers, but surrounds them with alien words that alter their meaning. Indeed, in some cases he interpolates such words within the passage. And in at least one case he does so without clearly indicating his insertions by brackets "[ ]".

Mill's famous discussion of names in his System of Logic includes a footnote (n.5) saying of one of his critics: It is well to be now and then reminded to how great a length perverse misquotation (for, strange as it appears, I do not believe that the writer is dishonest) can sometimes go.

Mill warns readers that even when they see volume and page cited, and see "the apparent guarantee of inverted commas", they should be careful "not to give implicit credence ... without verifying the reference". Mill's warning remains as timely as ever, and should especially be kept in mind when reading Smith. It would not be feasible to indicate all instances of the foregoing errors in the large and growing corpus of Smith's writings, nor should it be needful to do so. A collection of a few well-selected examples of errors of omission and errors of commission should suffice to warn readers about the sort of thing they need to be on guard against. Such a collection is what Soames and I aim to provide. Though we discuss specifically only Smith's first and second papers as already published, a reader who takes warning from our examples should be safe against being misled by anything further Smith has written or may write, and we leave the detection of the kind of faults listed above in his later writings as an "exercise to the reader". Leaving the issue of the accuracy of quotations to Soames, since it has already been broached in his previous exchange with Smith, my installment will include two examples illustrating the tendencies to read things into the text on the one hand, and to ignore context on the other. EXAMPLE A

There is a technique worthy of a place in Schopenhauer's Art of Controversy deployed at several places in Smith's writings, where he encounters some doctrine to which a catchy label has been applied, that has played an important role in recent philosophy of language, and that he cannot with even the most superficial show of plausibility attribute to Marcus. He first of all locates some other doctrine, one that has played a negligible role in recent philosophy of language, that he thinks can be so attributed to Marcus. He then invents a "genus" of which the important doctrine and the negligible one are claimed to be "species". He next transfers the catchy label so that it no longer applies to

130

JOHN P. BURGESS

the important doctrine, but rather applies to this "genus". He then claims that as discoverer of the first "species" of the "genus" Marcus should be considered discoverer of the "genus" itself. He thus is' in the end able after all to say that Marcus was the discoverer of the doctrine to which the catchy label applies. By deploying this technique he is in particular able to attribute to Marcus the discovery of "aposteriori necessity". In his first paper he claimed Marcus to have been the discoverer of "aposteriori necessity", a phenomenon whose discovery has generally been attributed to Kripke; but in his second paper it emerges that he understands by "aposteriori necessity" a different species of phenomenon from the one discovered by Kripke. Kripke's view - the view that prior to Smith has generally been understood by the phrase "the doctrine of aposteriori necessity" - is the view according to which there is some one item, exemplified by: (1)

Hesperus is Phosphorus

of which two things may be said. On the one hand, the honorific "necessity" may be applied to it, provided this honorific is understood in the proper sense, the "metaphysical" sense. On the other hand, such honorifics as "tautologous" and "analytic" and "non-empirical" and "a priori" may not be applied to it. Now in Marcus (1963), unlike Kripke (1972), the honorifics "necessary" and "analytic" and "non-empirical" and so forth are used interchangeably throughout. What Smith calls "aposteriori necessity" and attributes to Marcus is, nearly enough, the following view: (2)

(1) says something to which the honorifics may be applied, something that may be called necessary (and analytic and apriori).

(3)

(2) says something to which the honorifics may not be applied, something that must be called (contingent and) synthetic and aposteriori.

This were better called the doctrine of the contingently necessary, or the empirically a priori, or the synthetically analytic. But what the view should be called is a side-tissue. The key issue is whether this view, call it what you will, was endorsed in Marcus (1963). (I do not dispute that something rather like it is endorsed in some papers of Marcus responding to Kripke and dating from the 1980s.) The passage Smith cites in this connection reads as follows, the underlined portion being as much of it as Smith quoted in his first paper: But my point is only to distinguish tagging from describing, proper names from descriptions. You may describe Venus as the evening star and I may describe Venus as the morning star, and we may both be surprised that, as an empirical fact, the same thing is being described. But it is not an empirical fact that

HOW NOT TO WRITE HISTORY OF PHILOSOPHY

131

(17) Venus I Venus and if 'a' is another proper name of Venus [that] (18) Venus I a . ... What I have been arguing for the past several minutes is, that to say of an identity (in the strongest sense of the word) that it is true, it must be tautologically true or analytically true. [Underlining added.]

My interpretation can perhaps best be conveyed by a free paraphrase: My point is to distinguish genuine proper names from mere definite descriptions. Taking the usual planet as our example, consider first descriptions. Suppose you give a description of it in the words "the last star of the morning" and T give a description of it in the words "the first star of the evening". That the thing answering to your description and the thing answering to my description are the same, that the first star of the evening is the last star of the morning, is empirical and aposteriori, and perhaps quite surprising to both of us. Consider by contrast names, say the usual name "Venus" and the alternative name "Aphrodite". That the thing answering to the former name and the thing answering to the latter name are the same, that Venus is Aphrodite, is no more empirical or aposteriori than that the thing answering to one of these names is the same as itself, that Venus is Venus. A true equivalence involving descriptions, such as "the first star of the evening is the last star of the morning", is synthetically true, but a true identity involving names, such as "Venus is Aphrodite", is analytically true.

Thus on my reading the passage states standard Russellian doctrine (and closely parallels things said by Russell himself quoted in my first paper). Smith has not indicated his interpretation of the passage- or rather, of the snippet from it he quotes - by offering a paraphrase, so I will do so on his behalf. I urge to reader to double-check against (the last few pages of) Smith's second paper, to verify that the paraphrase corresponds to the interpretation Smith propounds: My point is to distinguish the question whether the proposition expressed by a given sentence has such-and-such a status from the question whether the proposition expressed by the sentence "the proposition expressed by the given sentence has such-and-such a status" has such-and-such a status. Take the usual planet as our example. Suppose you introduce a new proper name for it, identifying the thing to which the name is to be applied with the words "Let the last star of the morning be named ...", and suppose I introduce a new proper name for it, identifying the thing to which the name is to be applied with the words "Let the first star of the evening be named ...". And suppose by coincidence the new name you introduce and the new name I introduce happen to be spelled and pronounced just like each other and just like the usual English name "Venus". Then the proposition expressed by the sentence "Venus [your word] is Venus [my word]" is necessary, since it is just the proposition that the planet is self-identical. But the proposition expressed by the sentence "the proposition expressed by the sentence 'Venus [your word] is Venus [my word], is necessary", or the proposition expressed by the sentence "the name 'Venus [your word]' and the name 'Venus [my word]' refer to the same thing", is synthetic and aposteriori, and perhaps surprising to both of us.

A conspicuous difference between my interpretation and Smith's is that I take Marcus' displayed item (17) to contain two occurrences of the usual English name "Venus", while Smith takes it to contain one occurrence each of two names, one hypothetically newly introduced by you, the other hypotheti-

132

JOHN P. BURGESS

cally newly introduced by me, which names happen coincidentally to be homographic and homophonic. Now Marcus says a lot in her talk about the potential for a descriptive phrase such as "the great dissenter" to be converted into a proper name such as "Great Dissenter", but she says very little, and nothing in the near vicinity of the quoted passage, about the introduction of new names ex nihilo. And while I have not perhaps read the whole Marcusian corpus, I have read a dozen and more of her works without once encountering a passage where she says anything about homonyms, and certainly she says nothing explicit about them anywhere in the paper from which the above quotation is taken. It seems to me about as plain as anything can be that the source of the idea of homonymous newly-introduced names here is Smith's imagination, not Marcus' text. He is importing ideas neither in the text nor its context. But that is not all that he is doing. He is omitting words from his quotation. It is easy to see why Smith prefers not to quote or - when obliged by Soames' reply to quote it in his second paper, not to discuss - the displayed item (18). For it is a mystery why if, as is the case on his reading, (17) is already about two different names for the same thing, another example about two different names for the same thing is wanted in (18). And there is a greater mystery: Which of the two names "Venus", yours or mine, is supposed to be occurring in (18)? EXAMPLE B

While I consider the foregoing an important example, there is another on which I personally am willing to stake the whole issue. But first some background. The common noun "planetoid" has the same meaning as "asteroid". If one has a full mastery of English, including these two particular items of vocabulary, then one can determine simply by linguistic reflection that "planetoid" has the same meaning and (hence) reference as "asteroid". If one has less than full mastery of English, one can engage in the kind of linguistic research required to get new words into one's vocabulary, or one can consult the results of others' linguistic research as reported in a lexicon or glossary. In any case, one need not engage in astronomical research. One need not even consult an encyclopedia or almanac for reports of others' astronomical research. This is what philosophers of the period had in mind when distinguishing "linguistic" from "empirical", or in more traditional terminology, analytic from synthetic or apriori from aposteriori. And this is the distinction that Quine, the commentator on Marcus' paper at the colloquium, had famously attacked in "Two Dogmas of Empiricism" and Word and Object. Now for the example. The proper name "Hesperus" has the same reference as the proper name "Phosphorus". If one is unable to conceive of any more than just the two options, that either the meaning of a name must be taken to be the same as that of some description, or else that the meaning of a name must be taken to be just its referent, and if one joins the whole series of writers from

HOW NOT TO WRITE HISTORY OF PHILOSOPHY

133

Mill to Ziff in rejecting the first option, then one will have to join Smullyan and Fitch in adopting the second option. But then one will be committed to the conclusion that "Hesperus" and "Phosphorus" have the same meaning, which would imply on the general understanding of "linguistic" versus "empirical" prevailing among philosophers of the period that one should be able to find out that "Hesperus" and "Phosphorus" have the same reference by linguistic as opposed to empirical means. This is a conclusion that Kripke, who emphatically repudiated it, took Marcus to be drawing explicitly in her crucial last speech in Marcus (1963a), and in particular in the notorious dictionary remark: Presumably, if a single object had more than one tag, there would be a way of finding out such as having recourse to a dictionary or some analogous mode of inquiry, which would resolve the question as to whether the two tags denote the same thing.

Even readers who lacked background knowledge of the Russellian tradition and imagined that Marcus (1963) was advancing unprecedented doctrines, and who lacked background knowledge of the contents of Kripke (1972) and did not imagine that Kripke addressed any questions not already addressed by Marcus, would have to recognize that if anything like Kripke's reading of the dictionary remark is correct, then Marcus' doctrine was very different from his doctrine on a very central question. And all this is certainly recognized by Smith, who accordingly has devoted enormous exegetical efforts (beginning already in his first paper) to disputing Kripke's reading. Smith is especially insistent in claiming that in the remark Marcus meant by "dictionary" one of encyclopedic rather than lexical type. Smith is also insistent in claiming that in any case she only meant that one way one might find out whether two names name the same thing would be by consulting a dictionary. (The first claim surely derives from Marcus; the second may perhaps be his own contribution.) The remark means no more, Smith claims, than that if you are wondering whether "Hesperus" names the same thing as "Phosphorus", one way you might be able to find out would be by consulting an encyclopedia. Now it must be conceded that, considering just this one sentence, if it really were (as Smith wrongly characterizes it) a minor aside, and especially if it occurrred in isolation and were the only thing Marcus said on the topic, Smith's exegetical efforts could not be confidently rejected. But the sentence does not occur in isolation; it occurs in a definite context. The most important methodological error in Smith's exegesis of this remark is his obliviousness to contextual considerations. First, there is the immediate context. On the one hand, there is what came immediately before, the question to which the passage was an answer. Kripke asks whether it isn't Marcus' view that if "Hesperus" and "Phosphorus" are names, and "Hesperus is Phosphorus" is true, then "Hesperus is Phosphorus" is necessary. Marcus answers, according to Smith, that if you were wondering whether "Hesperus" names the same thing as "Phosphorus", you might be able

134

JOHN P. BURGESS

to find out by consulting an encyclopedia, and then again you might not. Clearly Smith's interpretation succeeds in making Marcus' remark perfectly innocuous only at the expense of making it perfectly inane and perfectly irrelevant to the question to which it was supposed to be the answer.

On the other hand, there are the two sentences that come immediately after: If 'Evening Star' and 'Morning Star' are considered to be two proper names for Venus, then finding out that they name the same thing as 'Venus' names is different from finding out what is Venus' mass or its orbit. It is perhaps admirably flexible, but also very confusing to obliterate the distinction between such linguistic and properly empirical procedures.

I read each of these as a further endorsement of the idea I already found in the first sentence. I read the last sentence as alluding to Quine's most famous paper, "Two Dogmas of Empiricism", upholding the linguistic/empirical distinction criticized by Quine in that paper; and placing such procedures as would be involved in finding out that "Hesperus" and "Phosphorus" name the same thing firmly on the linguistic side as opposed to the empirical side. The mixture of compliment and criticism in the words "admirably flexible, but also very confusing" from this last sentence in Marcus' remarks in the postcolloquium discussion matches the similar mixture in the words "exhilarating, but unnecessary" from the last sentence in Marcus' colloquium talk proper, which explicitly alludes to Quine's most famous book, Word and Object. In any case, what we have before us is a passage in which there occur several sentences, each such that significant exegetical effort is required to keep it from coming out as an expression of the doctrine of analyticity and aprioricity of true identities involving names, with no counter-balancing sentence clearly enunciating any opposing doctrine. Further, there is the context of the colloquium as a whole. What we have before us is only one of several passages each such that significant exegetical effort is required to keep it from coming out as an expression of the doctrine of the analyticity and aprioricity of true identities involving names, with no counter-balancing passage clearly enunciating any opposing doctrine. Yet further, the textual situation is the same for the doctrine of the universal substitutability of co-referential names in all contexts except those of quotation, of which the doctrine of the analyticity and aprioricity of true identities involving names is an immediate corollary or special case. (The other relevant passages are quoted in my previous paper.) Finally, there is the context of Marcus' other papers of the same period. Six months after her talk at the Boston colloquium, Marcus presented a paper at a Helsinki conference, generalizing some of what she had said earlier about such equations as "Morning Star = Evening Star" to such equations as "{Mercury, Morning Star, Earth, Mars} = {Mercury, Evening Star, Earth, Mars}". Naturally, some of the material in her earlier discussion Marcus (1963, 1963a) had to be reviewed in her later paper Marcus (1 963b), so that the latter contains a number of passages that plainly parallel or paraphrase passages in the former. In particular, there is a parallel (p.132) to the dictionary passage.

HOW NOT TO WRITE HISTORY OF PHILOSOPHY

135

The wording of the parallel passage makes three things clear about her earlier dictionary remark. First, Marcus was not just saying that one needs to investigate linguistic usage in addition to planetary motions in order to determine that "Hesperus" and "Phosphorus" name the same thing: She was saying one doesn't need to investigate planetary motions. Second, the "analogy" Marcus had in mind was between consulting dictionary entries in the case of a natural language and consulting meaning postulates in the case of a formal language, a paradigmatically linguistic as opposed to empirical procedure. Third, the "dictionary" Marcus had in mind was a glossary containing only purely linguistic information, and not an almanac containing also miscellaneous empirical information, a lexicon and not an encyclopedia. The parallel passage is one that does not reappear in the heavily revised version of Marcus (1963b), dating from long after Kripke (1972) and virtually amounting to a different paper, that appears under a different title in Marcus (1993). And the passage is one that Smith somehow omits to quote and to take into consideration in interpreting the dictionary remark. So let me close by quoting its key assertions and leaving the reader to ponder their implications. First: And to discover that we have alternative proper names for the same object we turn to a lexicon, or, in the case of a formal language, to the meaning postulates.

and again: To resolve the puzzle, one doesn't investigate the planets, but the accompanying lexicon.

Department of Philosophy Princeton University

REFERENCES Burgess, 1.P.: 1996, 'Marcus, Kripke, and Names', Philosophical Studies, 84, 1-47. Fitch, F.: 1949, 'The Problem of the Morning Star and the Evening Star', Philosophy of Science, 16, 137-141. Fitch, F.: 1950, 'Attribute and Class', in M. Farber (ed), Philosophic Thought in France and the United States. University Press, Buffalo, 640-{i47. Kripke, S.A.: 1972, 'Naming and Necessity: Lectures Given to the Princeton University Philosophy Colloquium, January, 1970', in D. Davidson and G. Harman (eds), Semantics of Natural Language. Reidel, Dordrecht, 253-355 and 763-769. Marcus, R.B.: 1960, 'Extensionality', Mind, 69, 55-62. Marcus, R.B.: 1963, 'Modalities and Intensional Languages', in M.W. Wartofsky (ed), Proceedings of the Boston Colloquium for the Philosophy of Science 196111962. Reidel, Dordrecht, 77-96. Reprinted in R.B. Marcus, Modalities. Oxford University Press, Oxford, 1993,3-23. Marcus, R.B.: 1963a, Remarks in R. Barcan Marcus, W.V. Quine, S. Kripke, 1. McCarthy, and D. F0llesdal, 'Discussion', in M.W. Wartofsky (ed), Proceedings of the Boston Colloquium for the Philosophy of Science 196111961. Reidel, Dordrecht, 1963, 105-116. Reprinted in R. Barcan Marcus, Modalities. University Press, Oxford, 1993,324-35.

136

JOHN P. BURGESS

Marcus, R.B.: 1963b, Classes and Attributes in Extended Modal Systems', Acta Philosophica Fennica, 16, 123-136. Marcus, R.B.: 1978, review of L. Linsky Names and Descriptions. University of Chicago Press, Chicago, 1977. Philosophical Review, 87, 497-504. Prior, A.N.: 1963, 'Is the Concept of Referential Opacity Really Necessary?', Acta Philosophical Fennica, 16, 189-199. Prior, A.N.: 1967, 'Logic, Modal', in P. Weiss (ed), The Encyclopedia of Philosophy. MacMillan, New York, 5, 5-12. Smith, Q.: 1995, 'Marcus, Kripke, and the Origin of the New Theory of Reference', Synthese, 104, 179-189. Smith, Q.: 1995a, 'Marcus and the New Theory of Reference: A Reply to Scott So ames', Synthese, 104,217-244. Smullyan, A.: 1947, review of w.v. Quine 'The Problem of Interpreting Modal Logic', Journal of Symbolic Logic, 12, 43-48, 139-141. Smullyan, A.: 'Modality and Description', Journal of Symbolic Logic, 13, 31-37. Soames, S.: 1995, 'Revisionism about Reference: A Reply to Smith', Synthese, 104, 191-216.

QUENTIN SMITH

DIRECT, RIGID DESIGNATION AND A POSTERIORI NECESSITY: A HISTORY AND CRITIQUE

INTRODUCTION

This paper aims to accomplish two interdependent goals, one historical and one philosophical. First, I aim to present a partial history of the theories of direct, causal or rigid reference as they appear in the New Theory of Reference, an account additional to that presented in [Smith, 1995a, 1995b]. Second, I aim to critically assess and further develop these theories. The historical part of the paper mostly centers on the largely unknown contributions made by Peter Geach and Ruth Barcan Marcus (that is, contributions additional to the ones discussed in my [1995a] and [1995b]. One of my main theses is that Geach [1969], not Donnellan [1970], originated the causal or "historical chain" theory of names. I also argue that Marcus is not correct when she recently writes [1993: xiii-xiv] about her earlier article, [1961], that in this article "proper names are not assimilated to what later came to be called 'rigid designators' by Saul Kripke ...". I shall argue that in her [1961], Marcus does assimilate her proper names to rigid designators. I shall also argue that she does not correctly characterize the relation of her referential definite descriptions in her [1961] to those of Donnellan [1966]. In the course of this argument, I shall put forth the view that the standard history of the New Theory of Reference is mistaken in attributing the origin of the referential! attributive distinction regarding definite descriptions to Donnellan. The most important philosophical (as distinct from historical) aspects of the paper lie in the development and evaluation of two different definitions of rigid designation, what I call the "direct reference definition" and the "world definition". The "world definition" is commonly associated with Kripke and the "direct reference definition" with Kaplan. I shall argue that only the definition based on Kripke's early work [1971; 1972] is sound.! DIVISION ONE: DIRECT REFERENCE AND CAUSAL REFERENCE

1. The referential use of definite descriptions in Marcus

One inaccuracy in the "standard" history of the New Theory of Reference's arguments against the Frege-Russell descriptional theory is that the first 137 P. W. Humphreys et al. (eds.), The New Theory of Reference © Kluwer Academic Publishers 1998

138

QUENTIN SMITH

landmark publication appeared with Keith Donnellan's "Reference and Definite Descriptions" [1966]. Two other inaccuracies are that the first articulations of the theory of causal reference appeared in Donnellan's "Proper Names and Identifying Descriptions" [1970] and Saul Kripke's "Naming and Necessity" [1972], and that the first presentation of the concept of rigid designation appeared in Kripke's "Identity and Necessity" [1971]. The "standard" view is well represented by one of the recent historical studies of these theories, Tyler Burge's "Philosophy of Language and Mind: 1950-1990" [Burge, 1992]. Burge writes about "a completely different picture of reference" that emerged with Donnellan's 1966 article: "In 1966 Donnellan pointed out that there is a use even of definite descriptions in which their meaning - the conditions laid down by the definite descriptions - does not fix the referent" [1932: 23]. Burge is mistaken, since this fact about definite descriptions was first pointed in 1961 by Marcus, as I shall show shortly. Burge proceeds to claim that "the decisive further move was made in 1970 by Kripke and Donnellan, independently of one another", namely, that the referent of a name is usually fixed by a historical "chain of uses of the name" [1992: 24]. But in fact this "decisive further move" was first made by Peter Geach in his "The Perils of Pauline" [1969], as I shall show below. Burge then identifies Kripke's essays as the first "account of names [in terms of] a theory of necessity. He counted names as 'rigid designators' - expressions that maintained a certain constancy of reference through variation in the possible worlds by reference to which modal sentences might be evaluated" [1992: 25]. However, these ideas about names and rigid designators were not presented first by Kripke, but earlier presented by Plantinga [1967], F011esdal [1961] and Marcus [1961]. Furthermore, a theory of rigid names was presented independently of Kripke's and in the same year by Kaplan (in a talk, "Dthat", in the Fall of 1970 at Stanford University). I shall discuss Plantinga's [1967] and F011esdal's [1961] in my [1998b] and shall focus on Geach, Marcus, Donnellan, Kripke and Kaplan in this paper. Let us begin with the idea that Tyler Burge claimed first introduced this new picture of reference, the idea that uses of some definite descriptions are referential rather than attributive. The virtually universal belief is that the distinction between attributive and referential uses of definite descriptions is original with Donnellan [1966]. But Donnellan at best only discovered one species of this distinction. Another species and the summum genus, so to speak, was discovered by Marcus in her [1961 ]. Donnellan argued that some definite descriptions in a given natural language can be used at some times by some individuals in a referential way, and can be used at other times by these or other individuals in an attributive way. For example, I may use "the man in the corner who is drinking a martini" in a referential way to directly refer to that man (regardless of whether or not he is drinking a martini) and at a later occasion I may use it attributively to refer to

DIRECT, RIGID DESIGNATION AND A POSTERIORI NECESSITY

139

whatever x has the property of being the only man in the corner who is drinking martini. But Marcus earlier introduced into the New Theory of Reference another species of the referential/attributive distinction and thereby introduced the genus (whoever introduces the first species ipso Jacto introduces the genus). Marcus' species is the conversion of an attributively used definite description into a referentially used definite description that occurs over a time and among the users of a given natural language as a whole. For example, "the evening star" may at first be used in our culture in an attributive way to refer to whatever has the property of being the first heavenly body to be visible in the evening. But at a later time, this definite description may come to be used by language-users in a purely referential way, to refer to that individual, regardless of whether or not it has the property of being the first heavenly body visible in the evening. In these cases, the referentially used definite description comes to be used as a directly referential name of the object. Using the expression "tag" for what later came to be called a "directly referential name", Marcus writes: In fact it often happens, in a growing, changing language, that a descriptive phrase comes to be used as a proper name - an identifying tag - and the descriptive meaning is lost or ignored. Sometimes we use certain devices such as capitalization and dropping the definite article, to indicate the change in use. "The evening star" becomes "Evening Star", "the morning star" becomes "Morning Star", and they may come to be used as names for the same thing .... One might even devise a criterion as to when a descriptive phrase is being used as a proper name. Suppose through some astronomical cataclysm, Venus was no longer the first star of the evening. If we continued to call it alternatively "Evening Star" or "the evening star" then this would be a measure of the conversion of the descriptive phrase into a proper name [1961: 309].

In this passage we find the idea, later to become central to The New Theory of Reference, that the reference of a locution is not routed through a descriptive sense that the locution expresses; the reference is direct. 2 Marcus herself comments on her distinction between referential and attributive uses of definite descriptions. In her 1981 article "A Proposed Solution to a Puzzle About Belief" she writes that this distinction was present in her 1961 article: I also note that a description can come to be used purely referentially and say on p. 309, 'in fact it often happens, in a growing, changing language, that a descriptive phrase comes to be used as a proper name - an identifying tag - and the descriptive meaning is lost or ignored.' I note that sometimes we use certain devices such as capitalization and dropping the definite article to mark the shift in use (a foreshadowing of Donnellan's 'Reference and Definite Descriptions' Philosophical Review 75 [1966]: 284--304) [Marcus, 1981: 509, n. 2].

Marcus correctly characterizes the sort of referential descriptions that are discussed in her [1961], but it is not strictly correct to say that her referential descriptions foreshadow the sort of referential descriptions in Donnellan's [1966]. In fact, if this passage is taken to suggest that Marcus' distinction is the same species as Donnellan's, it is false (there is no prior discovery of his species). But if we interpret the "foreshadowing" relation in a very broad sense, then we can read the passage as correctly implying that Marcus first discovered

140

QUENTIN SMITH

the genus "referential use of a definite description". The relevant genus/species distinction is demonstrably present in this referential/attributive distinction. The genus is the referential use of a definite description (the use of it to refer directly to something, not necessarily the object that meets the descriptive conditions belonging to the description). One species has for its specific differentia: using referentially a definite description D constantly and by all language users, such that D used to be normally used attributively by earlier generations of language users. A second species has for its specific differentia: using referentially a definite description D on some occasions by some people, such that on many other occasions, many people also use D in an attributive way. However, I believe in the end we must reject this species/genus distinction, since Donnellan's alleged species is not a genuine species. Kripke has made a plausible case in "Speaker's Reference and Semantic Reference" ([ 1977, reprinted in [Martinich, 1985]) that Donnellan's species is not a genuine semantic case, since his examples can all be accounted for by the speaker's reference/semantic reference distinction. Kripke writes: "Suppose I come upon Smith foully murdered. The condition of Smith's body moves me to say, 'Smith's murderer is insane'." [Martinich, 1985: 251]. This would be an attributive use of "Smith's murderer"; but if someone uttered the same sentence "while observing the wild behavior of the defendent at the dock" [1985: 251], this definite description would be used referentially to refer to the person who, in fact, is sitting in the dock, regardless of whether or not he has the property of being Smith's murderer. Kripke notes that the distinction between the two cases is not due to a semantic ambiguity of the definite description, but is a pragmatic distinction dependent on the speech acts. Given this, Donnellan's distinction does not count against Russell's theory that the semantic content of definite descriptions is captured by treating them as referring in a descriptional or attributive way. It does not amount to a "direct reference" semantics, but is consistent with the traditional descriptivist theory of the semantic reference of expressions. This interpretation of Donnellan's article is consistent with a remark that Donnellan himself makes. He writes: " ... whether the description is used referentially or attributively ... [is not due to] ... an ambiguity in the meaning of the words; it does not appear to be semantically ambiguous. (Perhaps we could say that the sentence is pragmatically ambiguous: the distinction between roles that the description plays is a function of the speaker's intentions.)" [1966]; reprinted in [Martinich, 1985: 236-248, see p. 244]. Accordingly, if Kripke is right (and I think he is) Donnellan's theory about the "referential use" of definite descriptions is not a new semantic theory since it is not a semantic theory; it retains the old Frege-Russell semantic theory and adds to it a new pragmatic theory about definite descriptions. By contrast, the change of "the little corporal" (as attributively used) to "the little corporal" used as a directly referential name of Napoleon is a semantic distinction

DIRECT, RIGID DESIGNATION AND A POSTERIORI NECESSITY

141

(Marcus' example). This expression first has the semantic content expressed by "the person who has the property of being the one and only little corporal [in the regiment x at time t]" and later has a new semantic content, namely, the content that consists solely of its direct referent, Napoleon. Thus, there are not two semantic species of "referential uses of definite descriptions", Marcus' species and Donnellan's species, but only one. Only with Marcus' distinction do we have a semantic theory of the direct reference of some uses of definite descriptions. 2. Geach's causal theory of reference

Following Tyler Burge's order of presentation of the ideas in the New Theory of Reference, the second major step is supplying the idea of referential uses of expressions with a "mechanism" that explains how they directly refer. It is questionable if this second step is necessary, since Marcus, Kaplan, Salmon, Almog, Wettstein and other proponents of the direct reference theory take the direct reference relation as primitive. Nonetheless, Kripke, Donnellan, Devitt and Putnam believe it important and the causal theory of reference they propounded is often associated with one of the "new" ideas of the New Theory of Reference. But did this new idea really come from the standard list of characters, as Tyler Burge and others claim? I have elsewhere said that "the most serious and widespread error in the history of recent analytic philosophy" is the belief that Kripke, not Marcus, originated the theory of rigid designation, the modal argument for direct reference, etc. Another candidate for "the most serious and widespread error" is that Donnellan and Kripke originated the causal theory of reference, also known as the historical chain theory of reference. The remark by Tyler Burge that I quoted (that the second major step, after Donnellan's distinction, is the Donnellan-Kripke causal theory of reference in 1970-72) is similar to (literally) hundreds of other such remarks that can be found in journals and books. Nathan Salmon's Reference and Essence develops a number of new and significant ideas and in addition presents what is perhaps the best and even the "classic" presentation of the standard history of the New Theory of Reference; my work builds upon and refines his achievement, rather than rejects it. Indeed, I based part of my history on Salmon's [1981] in my [1995a] and [1995b]. But one problem with Salmon's work is that he repeatedly refers to the causal theory as Donnellan's and Kripke's; e.g., "Donnellan [1972] and [1974] and Kripke [1972] provide accounts of proper name denotation by means of such historical chains of communication." [1981: 31]. Even Michael Devitt's important study, "Against Direct Reference" [1989], which presents one of the most accurate histories of the New Theory of Reference (e.g., he recognizes the origin of the theory of rigid designation in Marcus), attributes the causal theory to Donnellan and Kripke: Devitt writes: "The Causal Theory:

142

QUENTIN SMITH

This theory is the one that really is 'new'. It was discovered by Kripke [1980] and Donnellan [1972] ..." [1989: 210]. (Of course, Devitt is referring to the 1980 reprint of Kripke's [1972]; he is not thinking Kripke first published his theory in 1980.) Tyler Burge, Nathan Salmon and Michael Devitt are to be credited with a piece of historical knowledge many other philosophers lack: Many people mistakenly think that Kripke discovered this theory. These three authors are careful to co-attribute it to Donnellan and Kripke. But are Salmon's and Devitt's sentences correct in referring to the [1972] reprinting of Donnellan's [1970], giving it the same date as Kripke's [1972]? David Kaplan [1989b: 602, n. 86] writes more exactly that the notion of a historical chain of acquisition by which a name is passed from user to user: "first appears in print in Keith Donnellan's 'Proper Names and Identifying Descriptions", Synthese 21 (1970): 335-58; reprinted in Semantics of Natural Language ed. D. Davidson and G. Harman (D. Reidel Publishing Company 1972). It then appears in Kripke's Naming and Necessity, which coincidentally, was first published in the same collection in which Donnellan's article is reprinted (Semantics of Natural Language). Kripke notes, 'the historical acquisition picture of naming advocated here is apparently very similar to views of Keith Donnellan' (addenda to Naming and Necessity, p. 164)."

(Kaplan's reference is to [Kripke, 1980], but his addendum also appears in the original text [Kripke, 1972: 769]. Kaplan is right about Donnellan's theory appearing in print before Kripke's, but is wrong about this historical chain theory "first appearing in print" with Donnellan's work. Donnellan was ten months too late: Peter Geach's article, first presenting the causal theory of reference, appeared in "The Perils of Pauline", The Review of Metaphysics Vol. XXIII, No.2, December 1969, pp. 287-300, and Donnellan's appeared in October, 1970. Thus, contrary to the prevalent consensus of opinion, the historical chain theory of reference of names is neither "Kripke's theory" nor "Donnellan's theory", nor "the Donnellan-Kripke theory". It is "the Geach theory" of reference. Geach writes: I do indeed think that for the use of a word as a proper name there must be in the first instance be someone acquainted with the object named. But language is an institution, a tradition; and the use of a given name for a given object, like other features of language, can be handed on from one generation to another; the acquaintance required for the use of a proper name may be mediate, not immediate. Plato knew Socrates, and Aristotle knew Plato, and Theophrastus knew Aristotle, and so on in apostolic succession down to our own times; that is why we can legitimately use 'Socrates' as a name the way we do. It is not our knowledge of this chain that validates our use, but the existence of such a chain; just as according to Catholic doctrine a man is a true bishop if there is in fact a chain of consecrations going back to the Apostles, not if we know that there is. When a serious doubt arises (as happens for a well-known use of the word 'Arthur') whether the chain does reach right up to the object named, our right to use the name is questionable, just on that account. But a right may obtain even when it is open to question .... I introduced the use of the proper name 'Pauline' by way of the definite description 'the one and only girl Geach dreamed of on N-night'; this might give rise to the idea that the name is an abbreviation for the description. This would be wrong. [1969: 288-89].

DIRECT, RIGID DESIGNATION AND A POSTERIORI NECESSITY

143

I refer the reader to the rest of Geach's article for the other aspects of his theory, but instead of comparing its details with Donnellan's [1970] and Kripke's [1972] (see my [1998b] for a comparison), I should note that Michael Devitt has some reason to be listed with Geach, Donnellan and Kripke as one of the early developers of this theory. Devitt developed the most comprehensive causal theory of reference in his unpublished 1972 doctoral dissertation at Harvard, The Semantics of Proper Names: A Causal Theory of Reference and in a series of publications in the 1970s, 1980s and 1990s, beginning with his [1974]. In addition to the historical chain theory of reference, we see in Geach's 1969 article another principal idea of "The New Theory of Reference": the distinction between a directly referential name and a reference-fixing description ("I introduced the use of the proper name 'Pauline' by way of the definite description 'the one and only girl Geach dreamed of on N-Night' "), a distinction later made famous by Kripke [1972]. Apart from Geach, H.P. Grice also had the theory of a reference-fixing description for a directly referential name earlier than those to whom it is commonly attributed. Grice writes [1969: 144]: A name Cl may be introduced either so as to be inflexibly tied, as regards the truth-value of utterances containing it, to a given definite description 8, or so as to be not so tied (8 being univocally employed); so the difference between the two ways of introducing Cl may reasonably be regarded as involving a difference of sense or meaning for Cl; a sense in which Cl may be said to be equivalent to a definite description and a sense in which it may not. It is, then, not arbitrary so to design Q [the language] that its individual constants are to be regarded as representing, among other linguistic items, names with one of their possible kinds of meaning, namely that in which a name is not equivalent to a definite description.

Did Geach, Donnellan and Kripke develop the causal theory of reference independently of each other or were there relations of influence among two or more of them? The scanty published evidence seems to suggest an independent development. In Kripke's addendum to "Naming and Necessity" that Kaplan quoted (see my above quotation from Kaplan), Kripke mentions a similarity (but no relations of influence) between Donnellan's statement of the theory and his own statement of the theory. There is a sentence about Geach that is not present in the relevant footnote in the original version of "Naming and Necessity" [1972: 342: n. 2] but which is added to the 1980 reprinting [Kripke, 1980: 23, n. 2]. This is the footnote where Kripke mentions the various philosophers who held similar views and in the 1980 reprinting he added the sentence: "I also recall the influence of early conversations with Albritton and with Peter Geach on the essentiality of origins". [1980: 23, n. 2]. This may suggest the Geach-Kripke interchanges did not involve the causal theory of reference (but only the theory of the essentiality of origins). Geach does not in his publications discuss the relation of his theory to Kripke's or Donnellan's. Donnellan writes about the relation of his theory to Kripke's: "I believe that

144

QUENTIN SMITH

Saul Kripke has a very similar position, at least insofar as denial of the prevalent theories go. And, indeed, I think I may owe one of my counterexamples to him through a second hand source (although I did not understand the relevance until much later)." [1970: 357]. Donnellan indicates in a later footnote the second hand source is Rogers Albritton [Donnellan, 1970: 358, n. 18]. Neither Donnellan nor Kripke mention Geach's earlier statement of the causal theory of reference. Given this limited evidence, it appears reasonable to conclude that the three accounts of the causal theory were developed independently. If the three accounts of the theory (by Geach, Donnellan and Kripke) were developed independently and roughly simultaneously, then we should refer to it either as Geach's theory or as the Geach-Donnellan-Kripke theory (analogously to the Armstrong-Dretske-Tooley theory that laws are relations among universals, which was developed independently and roughly simultaneously by these three philosophers). Given these facts, virtually every single mention of the causal theory of reference in the philosophical literature for the past 25 years or so is false, inasmuch as it is mentioned as the Donnellan and/or Kripke theory. The same holds for the Geach-Grice-Kripke doctrine of reference-fixing descriptions, which is implicit in Marcus [1961] but explicit in these three writers. For purposes of clarification of my account of the causal theory, we may examine whether the recent characterization of Kripke's presentation of this theory in [Burgess, 1996] is accurate. Burgess writes that Kripke has "a positive alternative theory, the historical chain account, on which a name is bestowed on a thing specified by description, and on each subsequent occasion is used with the intention of continuing to refer to what it used to refer to, so that the circumstances of the earliest uses may soon be forgotten, though the name still succeeds in designating what it does only through a chain of usage extending back to the initial bestowal of the name" [1996: 20, my emphasis]. But this is not Kripke's theory; Kripke repeatedly insisted in [1972] that in most cases of bestowing a name, the name is not bestowed on a thing specified by a description. For example, Kripke writes that in some cases the name is bestowed by a description but that in general this picture of how a name is bestowed fails. The name is usually bestowed by ostension: "There may be some cases where the description picture is true, where some man really gives a name by going into the privacy of his room and saying that the referent is to be the unique thing with certain identifying properties ... Or he points to a star and says, That is to be Alpha Centauri'. So he can really make himself this ceremony: 'By 'Alpha Centauri' I shall mean the star right over there with such and such coordinates'. But in general this picture fails ... A rough statement of a [causal] theory might be the following: An initial baptism takes place. Here the object may be named by ostension, or the reference of the name may befixed by a description." [1972: 301-302, my italics].

Kripke's theory (like Geach's and Donnellan's) is that in some cases the name is bestowed by means of a description, but that in general it is bestowed by ostension.

DIRECT, RIGID DESIGNATION AND A POSTERIORI NECESSITY

145

As I indicated, the second published account of the causal theory of reference appeared in Donnellan's October 1970 paper, "Proper Names and Identifying Descriptions". He offers an example that is supposed to show that reference is determined by a historical chain rather than by descriptions. Suppose that Aristotle's and Herodotus' uses of "Thales" are connected by a historical chain to a man who never held that "all is water" and did not possess other identifying properties normally ascribed to him (e.g., falling into a well while looking at the stars). Suppose this man really was a well-digger with a reputation for saying wise things and who once exclaimed "I wish everything were water so I wouldn't have to dig these damn wells". Suppose further that there was some other ancient Greek hermit, unknown to Aristotle and Herodotus, who did hold that all is water and who fell into a well while looking at the stars. Now there is an historical chain connecting the well-digger to Aristotle's and Herodotus' use of the name "Thales" and an historical chain connecting Aristotle's and Herodotus' use of "Thales" to our use. According to Donnellan, Geach and Kripke, this shows that the person we refer to by "Thales" is the well-digger and not the philosopher. Are Geach, Donnellan and Kripke right? How shall we evaluate the "semantic argument for direct reference", as Salmon calls it? Nathan Salmon calls the semantic argument "the strongest and most persuasive of the three kinds of argument for the primary thesis of the direct reference theory" [1981: 29], the other two kinds being the modal and epistemic arguments. It seems to me this is not the case, for the semantic argument is inconclusive due to the fact that it does not elicit the linguistic intuition that its proponents believe it elicits. The example regarding Thales does not elicit the linguistic intuition that the well-digger is the referent of our use of "Thales"; nor does it elicit the contrary intuition that the hermit philosopher is the referent. Rather, it elicits the intuition that the reference of "Thales" is ambiguous. Our linguistic intuitions are best expressed as a reflection of this ambiguity: "I thought that 'Thales' referred to the philosopher who held that all is water, but I also thought that 'Thales' referred to the person that Aristotle's and Herodotus' use of 'Thales' was connected to by a historical chain; now the situation is confused and in the future it needs to be made clear as to whom we are referring by 'Thales'." This ambiguity is consistent with the descriptional theory of reference, for the descriptivists may explain the ambiguity by the fact that the descriptions pick out two different people; the descriptive sense the philosopher who said all is water refers to the hermit unknown to Aristotle, but the descriptive sense the person to whom Aristotle's use of "Thales" is causally connected refers to the welldigger. Likewise, in Geach's example, he makes a plausible case only as long as he makes only one description of Pauline inaccurate. But if most of the important descriptions are inaccurate, then it seems that a good explanation of why "communication" [1969: 289] breaks down is that the semantic content of the name is not unambiguously determined.

146

QUENTIN SMITH

Thus, I do not believe the "semantic argument for direct reference" is conclusive. As I argued in [1995b], the modal argument for direct reference is not conclusive either, since it does not refute the rigid descriptivist theories. And I argue in [1997] that the epistemic argument for direct reference is unsound. If this is indeed the case, the result would seem to be that the evidence underdetermines the direct reference and rigid descriptivist theories. DIVISION TWO: RIGID DESIGNATION

3. Kaplan on the origins of the rigid designation theory

Following Tyler Burge's order of presentation of ideas, the next important idea is supposed to be the interconnection of names and modality. Tyler Burge follows the standard history and claims that Kripke originated this idea; Burge writes that Kripke's essays are the first "account of names [involving] a theory of necessity. He counted names 'rigid designators' - expressions that maintained a certain constancy of reference through variation in the possible worlds by reference to which modal sentences might be evaluated" [1992: 25]. The origin of very few theories has been so widely misunderstood as the origin of the theory of rigid designation. As is well known, the theory that names are rigid designators was presented by Kripke in his January, 1970 lectures at Princeton University, published as [1972]. It is less well-known that (i) a theory that names, indexicals and other expressions are rigid designators was developed independently and presented by Kaplan in a talk, "Dthat", in the Fall of 1970 at Stanford University (see Kaplan, [1989a: 487], (ii) a theory of the rigid/nonrigid distinction was presented by Plantinga in seminars given in 1963 and later at Wayne State University, stated in final form in a section of a manuscript written in 1965, submitted to a publisher in 1966 and published in [1967: 177-183]3 and (iii) this distinction was presented earlier by Hintikka in an August, 1962 talk attended by Marcus, Montague, Prior, Kripke, Geach and others (published as [1963]; see Marcus [1993: 89] for the date of his talk), (iv) this distinction was also presented by F0llesdal [1961; 1967] and (v) presented by Marcus in [1961]. I discuss Plantinga's, Hintikka's and F011esdal's theories in [1998b] and shall concentrate here on the theories of Kaplan, Kripke and Marcus. Most philosophers think Kripke originated this theory in his 1971 and 1972 articles, which are transcriptions of talks he gave in January 1970 at Princeton University [1972] and later that year at New York University [1971]. Although David Kaplan's works are almost as widely known and discussed as Kripke's, it is rarely if ever remarked that Kaplan independently developed a theory of rigid designation and presented his theory in the same year, 1970, in a Fall talk at Stanford University. (See my quotes below from [Kaplan, 1989a: 487]. Kaplan's 1970 talk, later published as "Dthat" in 1978 (Syntax and Semantics, Vol. 9, ed. P. Cole [New York: Academic Press]), laid out the apparatus of

DIRECT, RIGID DESIGNATION AND A POSTERIORI NECESSITY

147

singular propositions, direct reference, possible world semantics, etc., and the argument that proper names, indexicals and other locutions are rigid designators. Kaplan refers to his 1970 talk as his 1978 publication "Dthat", which is reason to think there are no differences between the two. In Kaplan's own words, he says: "In fall 1970, I wrote for a conference at Stanford, a paper 'Dthat'." [1989a: 487]. He appends to this sentence a footnote reading: "David Kaplan, 'Dthat', in Syntax and Semantics, vol. 9, ed. P. Cole (New York: Academic Press, 1978); reprinted in Martinich, op. cit.") [Kaplan, 1989a: 487, n. 5]. Readers of Kaplan's "Dthat" will see that he argues that proper names, as well as indexicals, are rigid designators. (But Kaplan did not use the phrase "rigid designators'.) The reason people do not co-credit Kaplan and Kripke with the rigid designation theory is that Kaplan's article was not published until 1978 and Kripke's talks were published years before in [1971] and [1972]. Kripke gave his lectures on "Naming and Necessity" from January 20-29, 1970, and Kaplan gave his lecture on "Dthat" in the Fall of 1970, but Kaplan had not read a transcript of Kripke's talk until after his own talk on "Dthat". Kaplan explains this and other facts about his 1970 theory in the following sentences from [1989a: 487]; my inserted comment is in italics and brackets. "In fall 1970, I wrote, for a conference at Stanford, a paper 'Dhat'. Using Donnellan's ideas as a starting point, I tried to develop the contrast between Fregean semantics and the semantics of direct reference ... In Spring 1971, I gave a series of lectures at Princeton on the semantics of direct reference. By this time I had seen a transcript of Naming and Necessity and I tried to relate some of my ideas to Kripke's. [In a/ootnote to this sentence, Kaplan writes]: "Although the central ideas of my theory had been worked out before I became familiar with Naming and Necessity, I have enthusiastically adopted the 'analytical apparatus' and some of the terminology of that brilliant work." [Kaplan: 487., n. 6]

Given these facts, it seems that, at the very least, the rigid designation theory should be called the Kripke-Kaplan theory. In my [l995a], I mistakenly included Kaplan among those who had not recognized Marcus' earlier statement of the rigid designation theory and origination of the basic ideas of the New Theory of Reference. First of all, I misinterpreted his sentence that Kripke uses the term "rigid designator" in connection with Kripke's "controversial, though, I believe, correct claim that proper names ... are rigid designators" [Kaplan: 1989a: 492]. In my [1995a] I interpreted this sentence as Kaplan attributing the origin of the theory of rigid designation to Kripke. But (as I now realize) this sentence does not mean, imply or suggest that Kripke originated this idea. It merely states that Kaplan endorses this claim, a claim that Kaplan himself independently developed at least as early as 1970 and prior to his reading "Naming and Necessity". Second, I mistakenly said that in his published articles that Kaplan did not attribute the theory of rigid proper names to Marcus. In Kaplan's [1986], he does just this. In Kaplan's discussion of modal contexts in his [1986], he says that when we quantify into modal contexts, we can use as surrogates for values

148

QUENTIN SMITH

of variables "proxy names (closed singular terms)" and that either Marcus' tags or Kripke's rigid designators can serve as these modally stable closed singular terms. Kaplan continues: "Something like the intuitive idea of a tag [Marcus, 1961] or a rigid designator [Kripke, 1971, 1972] may guide out choice of proxy names." [1986: 253]. Third, Kaplan wrote a summary of Marcus' contributions to contemporary philosophy that includes a recognition of her priority in founding the New Theory of Reference. An excerpt from this appears on the back cover of the paperback edition of Marcus' Modalities. The excerpted parts on the book cover read: "Marcus is a brilliant, original, learned, tenacious, and productive scholar, many of whose early out-of-fashion ideas have now come into fashion and have largely swept the competition away ... The topics to which Marcus has contributed are high on the agenda of philosophers throughout the world". Apart from the mistaken account of Kaplan's views I gave in [1995a], I did not realize in my [1995a] and [1995b] that Michael Devitt had stated in his [1989: 210] that Marcus originated the theory ofrigid designation. This was a major omission from footnote 1 of my [1995a], since I there attempted to list the philosophers who recognized to the greatest extent Marcus' contributions, and it turns out that (apart from Kaplan), Devitt has recognized this to the greatest extent. Devitt wrote in [1989: 220): THE RIGID DESIGNATION THEORY The Rigid Designation theory has much more claim to be considered "new" though it is, in effect, to be found in Ruth Barcan Marcus (1961). Kripke is famous for urging the theory (1971,1980).

I am grateful to Devitt for pointing this out to me in 1995, although he now reports (private communication of January 1997) ~hat he has worked out a new and different theory of rigid designation and a new history. In his new theory [Devitt, forthcoming], Devitt distinguishes two main parts of the rigid designation theory, the rigid reference in modal contexts (which he argues was originated by Marcus) and the rigid reference in simple sentences (which he argues was originated by Kripke). Hintikka and Sandu [1995] have also recently pointed out that Marcus discovered the rigid designation theory, e.g., after noting that Marcus developed the necessity of identity thesis about variables in 1947, they say; "This Marcus did years before she formed the rigid designation idea in 1961." [1995: 271]. Recently, these facts are becoming more widely known, as is evinced by the entry for Marcus in the 1996 Encyclopedia of Philosophy: Supplement, which reads in part: "She also proposed that ordinary proper names are contentless tags [Marcus, 1961]. In so doing, Marcus rejected earlier 'descriptivist' accounts, often associated with Gottlob Frege and Bertrand Russell, and laid the cornerstone of the so-called new theory of reference later elaborated by Saul Kripke, Keith Donnellan, David Kaplan, and others." [Raffman and Schumm, 1996: 322].

DIRECT, RIGID DESIGNATION AND A POSTERIORI NECESSITY

149

Given this and what Burgess has called the "very wide publicity" [1996: 1) that has been given to the claim that Marcus' [1961) included the idea that names are directly referential rigid designators, we should say that by the present time (say, May 7, 1997) there is no longer a widespread belief that this theory was originated by Kripke. But there does remain one thing, namely, to present a more detailed argument than is present in the above-mentioned works (including my [1995a) and [1995b)) that Marcus' [1961) includes the concept of rigid designation. This is especially important in the light of the fact that Marcus, in her own recent interpretations of her [1961), denies that her proper names are rigid designators. She states that in her 1961 article "proper names are not assimilated to what later came to be called 'rigid designators' by Saul Kripke ..." [Marcus, 1993: xiii-xiv). Furthermore, the main argument for the thesis that Marcus' names are rigid designators, presented in Devitt's [1989), is unsound, as I argue in [Smith, 1997). Devitt also no longer believes his [1989) argument is sound (private communication), but for different reasons than the ones I present in [Smith, 1997). It appears, then, if we are able to have an accurate history of the concept of a rigid designator, a considerable amount of work remains to be done. In the following, I will endeavor to present an argument that Marcus' 1961 names are in fact rigid designators and that Marcus' own (recent) interpretation of her [1961) is mistaken. After I do this, I will compare Marcus' theory with the two most well-known theories of rigid designators, Kripke's and Kaplan's. I will also, in the sections on the world-definition and direct-reference definition of rigid designation, develop or make explicit the rigid designation theory beyond what is in the texts of Marcus, Kripke and Kaplan, and argue that only one of these definitions is correct, the one based on Kripke's work. This requires three tasks, two historical and one philosophical. First, I shall remove an obstacle created in some people's mind that prevents them from seeing the notion of rigid names in Marcus. (This is not the obstacle that prevents her from seeing that her names are rigid; I discuss her argument later.) I am now talking about the "real doubt" that some people have about whether her tags or names are meant to belong to natural languages or instead to an artificial language. If she is not talking about natural languages, then she is not introducing a main idea of the New Theory of Reference, that names in natural languages are direct rigid designators. Second, I will show that in Marcus' [1961), "necessity" is not an epistemological notion, but a logical/metaphysical notion. If "necessity" is defined to mean a priori, as in Ayer and others, then rigidity (which involves the notions of possibility and necessity) is an epistemological notion and has little relevance to rigidity defined in terms of possible worlds, which is how this notion is defined in the New Theory of Reference. I shall show Marcus defines necessity and rigidity in terms of possible worlds. Then I will construct a "complete theory" of rigidity in a technical sense I shall explain. This complete theory involves giving two definitions, with several

150

QUENTIN SMITH

elements in each definition, and comparing the two definitions to determine which is the most satisfactory. This "complete theory" will be the most substantive philosophical part of this paper, whereas the other sections belong in large part to the history of philosophy. 4. The discussion of natural and artificial languages in Marcus' [ 1961 J

Marcus introduced the notion that names in natural languages are rigid designators only if she discussed natural languages. Is there a "real doubt" that she is discussing names in natural languages, as Scott Soames [1995] and John Burgess [1996] have maintained? At some points of her 1961 paper, Marcus is talking about artificial languages (e.g., in the entire section called "Semantic Construction"), and at other points she is talking about natural languages, such as the following passage, which is a continuous text in Marcus' [1961: 309-10]: "In fact it often happens, in a growing, changing language, that a descriptive phrase comes to be used as a proper name - an identifying tag - and the descriptive meaning is lost or ignored. Sometimes we use certain devices such as capitalization and dropping the definite article, to indicate change in use. "The evening star" becomes "Evening Star", and they may come to be used as names for the same thing. Singular descriptions such as 'the little corporal', 'the Prince of Denmark', 'the sage of Concord', or 'the great dissenter' are as we know often used as alternative proper names of Napoleon, Hamlet, Thoreau and Oliver Wendell Holmes. One might even devise a criterion as to when a descriptive phrase is being used as a proper name. Suppose through some astronomical cataclysm, Venus was no longer the first star of the evening. If we continued to call it alternatively 'Evening Star' or 'the evening star' then this would be a measure of the conversion of the descriptive phrase into a proper name. If, however, we would then regard (10) [The evening star eq the morning star] as false, this would indicate that 'the evening star' was not used as an alternative proper name of Venus. We might mention in passing that although the conversion of descriptions into proper names appears to be asymmetric, we do find proper names used in singular descriptions of something other than the thing named, as in the statement 'Mao Tse-tung is the Stalin of China', where one intends to assert a similarity between the entities named. That any language must countenance some entities as things would appear to be a precondition for language. But this is not to say that experience is given to us as a collection of things, for it would appear that there are cultural variations and accompanying linguistic variation as to what sorts of entities are so singled out. It would also appear to be a precondition of language that the singling out of an entity as a thing is accompanied by many - and perhaps an indefinite or infinite number of unique descriptions, for otherwise how would it be singled out? But to give a thing a proper name is different than giving a unique description. For suppose we took an inventory of all the entities countenanced as things by some particular culture through its own language, with its own set of names and equitable singular descriptions, and suppose that number were finite (this assumption is for the sake of simplifying the exposition). And suppose we randomized as many whole numbers as we needed for a one-to-one correspondence, and thereby tagged each thing. This identifying tag is a proper name of the thing. In taking our inventory we discovered that many of the entities countenanced as things by that language-culture complex already had proper names, although in many cases a singular description may have been used. This tag, a proper name, has no meaning. It simply tags. It is not strongly equitable with any of the singular descriptions of the thing ..."

Several comments about this passage are in order. (a) Obviously, if any philosophy of language is about natural language, this passage exhibits such a

DIRECT, RIGID DESIGNATION AND A POSTERIORI NECESSITY

151

philosophy. (b) Obviously, the example of randomized whole numbers being used as tags is used to illustrate the idea that tags or proper names in natural languages have no meaning at all and differ in this respect from descriptions. (c) The main evidence that Soames and Burgess have presented as "real doubts" that Marcus is talking about names in natural language is quoting the sentence about "randomizing whole numbers" out of context. Here is how Burgess [1996] produces this doubt, where (A) contains a quote from Marcus: "Accordingly, Marcus invites us to contemplate the following imaginary scenario: (A) For suppose we took an inventory of all the entities countenanced as things by some particular culture ... And suppose we randomized as many whole numbers as we needed for a one-to-one correspondence, and thereby tagged each thing. When she continues, 'This identifying tag is a proper name of the thing', what is being called a 'name' is thus not a name in the ordinary sense, or even a linguistic expression, but a serial number." [Burgess, 1996: 3-4, my emphasis]

By these textual manipulations, Burgess has managed to create a doubt in the mind of some of his readers who are unfamiliar with Marcus [1961] about whether or not Marcus' theory of tags is a theory of proper names in ordinary language. Note that: (i) Burgess omits all the preceding sentences where it is evident that she is talking about natural language, and replaces them by his misleading introduction to the quote, where he says "Marcus invites us to contemplate the following imaginary scenario". Note also (ii) the ellipses in Burgess' quote, which consist of omitted material (present in my above, long quotation) that show she is talking about natural language, and (iii) Burgess misleadingly concludes about this passage that Marcus' serial numbers were not introduced as illustrations of the fact that names in natural languages have no meaning, and instead presents this illustration as all that she means by a name - a name is just a serial number in a non-natural language. I conclude, then, that Burgess and Soames have failed in their attempt to establish a "real doubt" as to whether Marcus is talking about names in natural languages. 5. Necessity in Marcus' 1961-1962 theory

What does Marcus mean by "necessity"? Does she hold that "necessary" means knowable by an analysis of its linguistic meaning alone? Or does she mean obtains in all possible worlds? There is no controversy about whether she characterizes certain classes of statements, e.g., identity statements involving names, as "tautologies", "analytic truths", "logical truths", "logically necessary truths", "necessary truths" and "not contingent truths". The question, rather, is what she means by these terms. Since her main discussion is about quantified modal logic, it is reasonable to assume that by a "logical truth" or a "logically necessary truth" she is talking

152

QUENTIN SMITH

about "logic" in the sense pertinent to quantified modal logic (her semantic construction on pp. 319ff. is in terms of QS4). In fact, she begins her discussion of identity statements by saying precisely that: "In the light of the previous remarks I would like to turn specifically to the criticisms raised against extended modal systems in connection with identity and substitution. In particular, I will refer to my extension of Lewis' S4 which consisted ofintroducing quantification in the usual manner and the addition of the axiom ... " [she then states a version of the Barcan formula].

It is uncontroversial that she takes identity statements such as "Venus is

Phosphorus" to be logical truths (in the sense pertinent to QS4), so we will move on. In her discussion of theorems of QS4 and its "logical truths" on pages 307-308, she uses "tautology" to mean a logical truth; e.g., she says the fact that "alb" is a "tautology" "is precisely the import of my theorem (8)" [ 1961: 308], where theorem (8) is a theorem of QS4 stating: "(8)

(xly)

~

D(xly)

where '0' is the modal symbol for logical necessity". [1961: 308]. The use of "tautology" for theorems of quantified modal logic is not new with Marcus; in Kripke's [1959] he called the logical truths ofS5 tautologies. In this article, Kripke says a statement true in every possible world is a "universal logical validity' [1959: 3] and adds: "Using these observations it is easy to prove that for formulas of S5 our notion of tautology coincides with our notion of universal validity". [1959: II] "... All axioms of S5 are tautologies" [1959: 12]. A formula "A is said to be universally valid if and only if A is valid in every non-empty domain". [1959: 2]. The point here is not that Kripke's semantics for modal logic and conception of tautologies are the same as Marcus'; they are not. Rather, the point is that the use of the word "tautology" does not ipso facto mean that one is not talking about statements true in all possible worlds and does not mean one is writing in the tradition, say, of A.J. Ayer. Marcus also calls "alb" or "Venus is Hesperus" an "analytic" truth; she writes: "What I have been arguing in the past several minutes is, that to say of an identity (in the strongest sense of the word) that it is true, it must be tautologically or analytically true". The question now is: what does Marcus mean by "analytic"? The word "analytic" is standardly used in many different senses, sometimes in a linguistic sense to mean a sentence that is transformable into a logically true sentence by substitution of synonyms for synomyns, sometimes in an epistemic sense to be knowable a priori, by reflection on linguistic meaning alone, sometimes in a conceptual sense to mean a thought where one concept contains another concept as a part, and sometimes in a logical sense to mean a logically necessary truth. In what sense does Marcus use it? The preceding quotations suggest that "analytic" means tautological and that tautological means logically necessary (in the sense of QS4). Is there any further evidence that she means "analytic" in the sense of (modal) logical necessity?

DIRECT, RIGID DESIGNATION AND A POSTERIORI NECESSITY

153

In the subsequent discussion of her paper [Marcus et al. 1962: 140, my emphases] she distinguishes between a narrow and broad sense of "analytic": "Necessary attributes would correspond to analytic functions in the broader sense of analytic. These may be thought of as a kind of essential attribute, although necessary attribute is better here. For these are attributes which belong necessarily to every object in the domain, whereas the usual meaning of essentialism is more restricted. Attributes like mathematician and cyclist do not correspond to analytic functions". It seems here that the broader sense of "analytic" is used in a modal logical sense, rather than an epistemic, linguistic or conceptual sense; a broadly analytic function corresponds to an attribute that belongs necessarily to every object in the domain. This broad sense of "analytic" has metaphysical import, for it is defined in terms of every object in the domain. "Domain" is used here in the sense of her earlier "semantic construction", where all possible worlds are said to include only objects that belong to the actual world. The "domain of objects" is all actually existing objects. A broadly analytic function corresponds to an attribute that belongs necessarily to every object in every possible world. This use of traditional terminology to introduce new ideas shows why it is easy to misunderstand Marcus' theory, as Kripke, Soames and Burgess have done. (But Kaplan [1986], Devitt [1989], Hintikka and Sandu [1995] and others seem to have understood the main points of her theory.) Marcus is trying to express a new idea by adding "broader" to a term that is traditionally used in other senses, just as Plantinga [1970] adds "broader" to the traditionally used "logically necessary" to express a new idea, namely, non-trivial essences, such as being prime, being a number, and world-indexed properties such as being snub-nosed in Kronos, where "Kronos" is a name for the actual world. Plantinga pointed out [1970] that the concept of a broadly logical necessity belongs to metaphysics, not epistemology, and argues that there are logical necessities that are known a posteriori. The modal-metaphysical import of Marcus' use of "analytic", "tautologial" and "logically necessary" is further evidenced by her definition of a logically true sentence. She offers her definition of logical truth in terms of her semantic construction, which she says "corresponds to the Leibnizian distinction between true in a possible world and true in all possible worlds" [1961: 320, my emphasis]. As some other moda110gicians do at this time, she uses "possible world" interchangeably with "model", so that "every model" is synonymous with "every possible world". Here is her definition of a logically true sentence: "A logically true sentence is one which would be true in every model". [1961: 319]. She also talks about counterfactual situations in her discussions of possibility and necessity, further evincing her metaphysical understanding of these notions. John Burgess stated to the contrary that Marcus' [1961] does not contain a single explicit discussion of a counterfactual comparison (and thus

154

QUENTIN SMITH

cannot contain the concept of a rigid designator): "there patently is no explicit discussion of rigidity - no explicit discussion of cross-comparison between actual and counterfactual situations, or between one counterfactual situation and another - in Marcus" [1996: 27, my emphases]. In fact, there are many explicit discussions of cross-comparisons between actual and counter factual situations in Marcus [1961], such as the following, where Marcus is talking about the criteria for deciding whether "evening star/morning star" and "Scott/author of Waverley" are being used as names or contingent descriptions: "... if our decision is that they are not simply proper names for the same thing; that they express an equivalence which is possibly false, e.g., someone else might have written Waverley, the star first seen in the evening might have been different from the star first seen in the morning - then they are not identities". [1961: 311, my emphases]. Other explicit discussions of crosscomparisons between actual and counterfactual comparisons appear on pages 308,309,311,312,313,315,316,317, and 318 of her [1961]. Burgess and Soames believe that by "necessary" Marcus means "discoverable by reflection on linguistic meaning alone". Thus, necessarily true identity statements involving names are (according to Burgess' and Soames' interpretation of Marcus) identity statements whose truth can be determined simply by reflecting on their meaning. But this cannot be what Marcus meant, if only for the reason that she denies that names have meaning. "This tag, a proper name, has no meaning. It simply tags". [Marcus, 1961: 310]. In order for "Hesperus is Phosphorus" to be known to be true by reflection on its linguistic meaning alone, "Hesperus" and "Phosphorus" would have to express concepts or descriptive senses, or some other sort of meanings, grasped by reflection. But with Marcus' theory, we have no such concepts or "graspable meanings" and we have merely meaningless syntactical entities (the tags), the identity sign, and astronomical bodies millions of miles away - there is nothing here to be reflectively grasped by contemplating the sentence in my "armchair", to use Burgess' word [1996: 25]. Soames and Burgess also ascribe to Marcus the view that the name-object relation is a priori, e.g., that it is an a priori truth that the public language name "Hesperus" (rather than "Kansas" or "Beethoven", for example) refers to Venus. The consequence they draw from this is that "on the view articulated by Marcus, true identity statements involving names are knowable a priori" [Soames, 1995: 205]. This is how Kripke understood Marcus, but Kripke mis-understood and misquoted [1971: 142-3; 1972: 305] her remark in [Marcus et aI., 1962: 142]. (See Smith [1995b: 236-37]; the misquotation appears in [Kripke, 1971: 142, n. 7]). Can we use Kripke's mis-quotation of Marcus as evidence that Marcus held the name-object relation to be a priori? Burgess asserts this is evidence in his [1996]. It is not obvious, however, that if x says "F" and y quotes her as saying "not-F", that counts as evidence that x did not say "F" but instead said "notF".

DIRECT, RIGID DESIGNATION AND A POSTERIORI NECESSITY

155

Apart from this issue, how is this theory of an "a priori name-object relation" consistent with Marcus' remarks that objects can be named ostensively and a posteriori, by inspection? Marcus writes: "For we can on inspection name the objects to which it applies just as one can on inspection name the members of the class designated by [Venus, Neptune, Mars]". [1963: 131]. The "it" refers to an abstract that is instantiated by replacing its individual constants by names we assign to objects on inspection. The abstract is "a((aIul) v (aIu2) v ... (aIun)) where 'I' names the identity relation and u 1 , U2 ... un are individual constants". [1963: 131]. If the name-object relation is a priori, how could there be inspections in which we name objects? How could one "on inspection name the members of the class designated by [Venus, Neptune, Mars]"? [1963: 131, my italics]. If objects had their names a priori, then it would be impossible to name them on inspection, for any attempt to name an object on inspection would result in the realization that the object already has a name, prior (a priori) to that inspection and any other possible inspection. Marcus' notes that "there are cultural variations and accompanying linguistic variations as to what sorts of entities are so singled out ... [S]uppose we took an inventory of all the entities countenanced as things by some particular culture through its own language, with its own set of names and equitable singular descriptions ..." [1962: 309-310]. The mere fact that the inventory of named entities varies from culture to culture implies that empirical investigation is required to learn a given culture's inventory of named entities. One has to learn through observations or records of linguistic practices the culture's inventory of named entities, and, in addition, the very development and compilation of the inventory by members of the culture requires observations of the linguistic practices of the people in that culture. The creators or compilers of that culture's dictionaries or lexicons are observers of that culture's linguistic practices. Once names have acquired an established use in a language, their usage may be recorded and the recording may be consulted to find out if two names have been assigned to the same object [Marcus et aI., 1962: 142]. One way we do find out that the same object has more than one name is to find more than one name listed for the object in a dictionary. For example, Webster's New World Dictionary (2nd ed., 1974, p. 256) has this entry: "Cic-er-o, (Marcus Tullius) 106-43 B.C.; Roman statesman, orator, and philosopher". Merriam Webster's (1941, p. 466) has for an entry: "Hes'per-us, n. The evening star; Venus". Kripke's theory that recordings of established uses, such as are found in dictionaries or lexicons, represent a priori knowledge has obscured the common sense fact that a paradigmatic way in which we answer empirical questions about the references of names is through reading a culture's lexicons or recordings of how people in that culture use names. We "inspect" the objects to which the culture has assigned names only in atypical cases where we suspect the culture has made a mistake about whether

156

QUENTIN SMITH

two names are co-referring or not. This atypical sort of a posteriori knowledge about identity statements involving names was first dicussed by P1antinga [1970]. Marcus' discussion is only about the typical cases where we assume that the culture's use of names is accurate and their lexicons give reliable empirical knowledge. In these typical cases, we consult the dictionaries rather than engage in a scientific observation of (say) "Venus' mass, or its orbit" [Marcus et al., 1962, p. 142]. Such cases as the case of the Babylonians whose lexicon or established use of names is inaccurate are resolved by resorting to the "naming inspections", to see where or how people went wrong in assigning names to objects (see [Plantinga, 1970]). If we combine Marcus' theory with Plantinga's, then we have a theory of the typical and atypical ways in which we learn a posteriori about the truth of necessary identity statements involving names. The theory (which Kripke seems to presuppose) that all questions of coreference are resolved in the manner in which they are in fact resolved when the lexicons are assumed to be unreliable fails to take into account the standard or typical case, which is when we assume our lexicons are reliable and that consulting them (rather than the objects themselves) is sufficient to answer our questions. (Perhaps it goes without saying that we need to use selected cases of famous names in order to talk about consulting dictionaries, since "analogous inquiries" are required for names that are not famous.) Marcus uses the empirical fact that names, but not contingent descriptions, are listed in lexicons as part of her argument that aggregates of objects are necessarily identical. Her novel argument about aggregates in [1963] has this structure: the relevant aggregates are necessarily identical since (i) an aggregate can be designated by a set of names, (ii) the identity sign is flanked only by names, (iii) identities are necessary, (iv) lexicons list names but not contingent descriptions, and (v) questions of identity about aggregates (that are designated by different sets of names) can be answered by consulting the cross-listing of co-referring names in a lexicon. Marcus' theory has not been widely appreciated as a consequence of Kripke's misinterpretation of her theory in terms of his false theory of dictionary knowledge as a priori knowledge. What is dictionary knowledge? David Gurlank and the others who engaged in the empirical research required to determine how English words are used, and to record this usage in Webster's New World Dictionary [Gurlank, 1974], put such entries in the dictionary as: "Hesperus. the evening star, esp. Venus" [Gurlank, 1974: 658]. I omitted the pronunciation symbols and Greek root from the entry]. If somebody asks me, "Is Venus identical with Hesperus"? and I understand the question and respond that I don't know the answer, but then proceed to look up the name in the dictionary, I will find the entry on page 658 for "Hesperus" and then learn these two names are co-referring. How is this different from being asked the question, "Is gold identical with the atomic element 79"?, understanding the question, and then proceeding to find out the

DIRECT, RIGID DESIGNATION AND A POSTERIORI NECESSITY

157

answer by reading a chemistry textbook? Kripke believes the former procedure is a case of a priori knowledge and the latter a case of a posteriori knowledge. Kripke either has an incorrect theory of the a priori/a posteriori distinction or else does not understand correctly the nature of dictionaries. Even if, counterlegally, dictionaries do contain a priori truths about nameobjects relations, looking up entries in a dictionary would still be a posteriori knowledge. For (as Kripke notes) "one can learn a mathematical truth a posteriori by consulting a computing machine, or even by asking a mathematician ... All the cases of the necessary a posteriori advocated in the text [of (Kripke, 1972)] have the special character attributed to mathematical statements ..." [1972: 765]. The exception Kripke makes for consulting dictionaries is simply a mistake. It is worth noting that the only place in the philosophical literature where the bizarre thesis that "the ordinary name-object relations reported in dictionaries are a priori" is even mentioned is in the writings of Kripke [1971; 1972; also see his remarks in Marcus et al., 1962], Burgess [1996] and my rebutals in [1995a; 1995b] and the present paper. As I indicated in [1995b], we do not find in Marcus' early writings an explicit discussion of the traditional thesis that the necessary and the a priori are identical or coextensive. Marcus' theories of direct reference, rigid designation, the necessary of identity between names, the modal argument for direct reference, etc., are explicit in her early writings, but as I maintained [1995b: 235], the argument that necessary identities can be a posteriori is implied by (but not explicitly stated in) her theory. The traditional thesis that "necessity = a priori" was not explicitly addressed and rejected until Plantinga's October, 1969 talk at Cornell University,3 later published as [P1antinga, 1970]. It may well be that a later theory that explicitly discusses and rejects the "necessary = a priori" thesis is more historically important than an earlier theory that merely implies this rejection. But this is not the issue here; the point is merely that Kripke, Soames and Burgess are mistaken in ascribing to Marcus the explicit or implicit endorsement of the theory that the ordinary name-object relation is a priori and that "Hesperus is Venus" is necessary in the sense that it is knowable by reflection on its meaning alone. This and the various other considerations I have discussed in this section indicate that Marcus understood "necessity" in a logical-metaphysical sense, not in a linguistic, epistemic or conceptual sense. This fact is further evinced if we examine the claim that Marcus is not a New Theorist of Reference, and is not writing in the tradition of Kaplan, Kripke, Putnam, Donnellan, etc., but is instead a Russellian who is merely repeating Russell's 1918 doctrines in her 1961 article. 6. Russell and Marcus

One reason both Burgess and Soames do not understand or correctly represent

158

QUENTIN SMITH

Marcus is that they insist that Marcus is repeating Russell's theory. For example, Soames writes that Marcus treats proper names as Russellian logically proper names: "However, given Marcus' treatment of names as Russellian logically proper names, and hence as supporting substitution without change of meaning in any non-quotational context, (PI ['It is knowable a priori that Hesperus = Phosphorus') is plausible, as are the other steps in the argument" [1995: 207]. And Burgess writes: "Marcus (1961) endorses the very well known, but at the time not very popular, view of Mill that names have no descriptive meaning, as well as some only slightly less well known views of Russell, including the view that true identity statements involving 'names' are 'tautologies' ... " [1996: 39).

Burgess' reference to Mill has some truth to it, but his and Soames' claim that Marcus was repeating or endorsing Russell's theory if false. Burgess and Soames claim at several points that what Marcus meant by "tautology", "identity", "names", "necessity", etc., is the same thing that Russell meant. This is demonstrably untrue. First, Russell's tautologies are theorems in extensional logic, specifically, first order propositional calculus or first order predicate calculus; but Marcus' tautologous statements are theorems in intensional logic, specifically, quantified modal logic (she uses QS4 in her 1961 semantic construction). Secondly, Marcus uses "identity" to express a relation different than strict equivalence, material equivalence, and other weaker sorts of equivalence relations, but Russell did not use "identity" to express this strong relation. For Russell, the identity sign can be flanked by a name and a definite description, but a key idea in Marcus' philosophy is that the identity sign can be flanked only by names. Russell used the word "identity" to express what Marcus calls an equivalence relation weaker than identity. Furthermore, for Russell, identities can be contingent, but for Marcus, identities are necessary. Russell writes: "That illustrates how 'the author of Waverley' is quite a different thing from a name. You can prove this point very clearly by formal arguments. In 'Scott is the author of Waverley', the 'is' of course expresses identity, i.e., the entity whose name is Scott is identical with the author of Waverley" [1918: 245]. Also see Russell [1918: 247]. Here Russell is using "Scott" as an artificial example of a name, since in ordinary usage (according to Russell) it is an abbreviated description and in reality names are demonstratives such as "this" and "that". But Marcus denied that the "is" in "Scott is the author of Waverley" expresses the identity relation and she denies that identity statements can be contingently true. Thirdly, by "is necessarily true" Russell means the temporal notion, is always true; for Russell, "necessity" means what is always true in the actual world, not the modal notion of what is true in all possible worlds. (This is Russell's doctrine in his 1918 essay on the "Philosophy of Logical Atomism", which Soames and Burgess take Marcus to be repeating.) But Marcus does not use "necesssary truth" in a temporal sense but in the contemporary modal sense to

DIRECT, RIGID DESIGNATION AND A POSTERIORI NECESSITY

159

mean truth in all possible worlds. Russell says that "one may call a propositional function necessary, when it is always true" [1918: 231, my emphasis], but Marcus rejects this temporal definition of necessity and accepts a possible world definition; a necessarily true sentence is one that is "true in all possible worlds" [1961: 320]; she uses "model" as a synonym for "world" and writes in addition: "A logically true sentence is one which would be true in every model". [1961: 319]. Fourth, Russellian logically proper names are names in a private language for my fleeting sense data with which I have an immediate acquaintance. "The only words one does use as names in the logical sense are words like 'this' or 'that'. One can use 'this' as a name to stand for a particular with which one is acquainted at the moment ... [A proper name] seldom means the same thing two moments running and does not mean the same thing to the speaker and to the hearer". [1918: 210]. Indeed, logically proper names are necessarily private; in a logically proper language "all the names that it would use would be private to the speaker and could not enter in the language of another speaker" [1918: 198]. Russell's names (a) are in a private language, (b) refer only to a person's introspected sense data, (c) seldom mean the same thing in two successive moments, (d) are demonstrative words, such as "this" and "that" and (e) are not intersubstitutable in (metaphysical) modal contexts, since for Russell there are no such contexts (for Russell, contexts involving "necessarily", "possibly" and the like are temporal contexts). But Marcus' names (a) are in a public language, (b) refer to ordinary things in the universe, such as Venus or Walter Scott, (c) usually have a constant reference across long periods of time, (d) are not demonstratives such as "this" and "that", but are what are ordinarily called "names", e.g., "Scott", "Venus", etc., and (e) are intersubstitutable in (metaphysical) modal contexts. Burgess claims that Marcus merely repeats the "views of Russell, including the view that true identity statements involving 'names' are 'tautologies' ..." [1996: 39]. But Russell's theory that "a true identity statement involving names is necessarily true" is the theory that the true private language statement about my sense data, "this is this'; is always true and expresses a kind of identity relation that can obtain contingently. Marcus' theory that "a true identity statement involving names is necessarily true" is the theory that a true public language statement about publically observable things in the universe, such as "Venus is Hesperus'; is true in every possible world, where "is" expresses her strong relation I of identity (which can never obtain contingently) and where the statement is a "tautology" in the sense of being a theorem of quantified S4 (rather than of the extensional logic of Whitehead's and Russell's Principia Mathematica). The claim that Marcus' [1961] is a "repetition and endorsement" of Russell's 1918 theory is as mistaken as any claim about the history of philosophy can possibly be. Soames and Burgess occasionally recognize (what one would hope to be) the obvious fact that Marcus' names are not Russellian logically proper names, but

160

QUENTIN SMITH

they continue to confiate the concepts. This is evident in Soames' discussion of a thesis (PI) he mis-attributes to Marcus: (PI) It is knowable a priori that Hesperus = Phosphorus.

Soames says about (PI): the premise "that it is knowable a priori that Hesperus is Phosphorus, presupposes a prior rejection of the description theory. The basis for assuming this premise is simply a prior acceptance of Millianism". [1995: 207]. But this is a mistake; the Millian theory of names ia not a prior basis for accepting, and does not imply, the premise that it is knowable a priori that Hesperus is Phosphorus. Kripke, Kaplan, Donnellan, Salmon, Wettstein and others are Millians and they reject the idea that Millianism is a basis for holding that Hesperus is Phosphorus is knowable a priori. Indeed, the fact that Millianism does not imply that such identity statements are a priori is standard orthodoxy among New Theorists of Reference, so one wonders how Soames went wrong here. Reading elsewhere on this page [Soames, 1995: 207], it appears that Soames is confiating Millian names with Russellian logically proper names: "However, given Marcus' treatment of proper names as Russellian logically proper names, and hence as supporting substitution without change of meaning in any non-quotational context, PI ["It is knowable a priori that Hesperus = Phosphorus"] is plausible, as are the other steps in the argument". Contra Soames, identity statements involving Russellian logically proper names ("this" and "that", names of my private sense data) are a priori, but identity statements involving Millian proper names are not a priori. The attribution of both theories to Marcus can only lead to confusion. (The Russell-Mill confiation also underlies Soames' claim that Smullyan originated all the ideas about names that appear in Marcus [1961] and that it was Smullyan, not Marcus, who is the unacknowledged founder of the basic ideas of the New Theory of Reference (direct reference, the modal argument for direct reference, the rigidity of names, the thesis that identity statements involving names are necessary if true, etc.). Smullyan is a Russellian, but Marcus is not, and SmUllyan did not hold any of the above-mentioned ideas, as I argue in my [1998b].) 7. The Marcus-Kripke theory that necessity is logical-metaphysical necessity

Soames and Burgess think that by "necessary truth" Marcus means the epistemological notion of being true a priori rather than the metaphysical notion of truth in all possible worlds. They argue that since Kripke means by "necessary truth" a non-logical and metaphysical notion of truth in all possible worlds, Kripke made the decisive advance into the post-1970 New Theory of Reference. Soames and Burgess identify the transition from logical modalities to metaphysical modalities as the (alleged) major advance made by Kripke from

DIRECT, RIGID DESIGNATION AND A POSTERIORI NECESSITY

161

Marcus' position. For example, Burgess writes: ... Marcus can be credited with a version or variant of the stability argument [made by Kripke] ... only if she can be credited with recognizing that possibility in the 'metaphysical' sense of what (is or isn't but) potentially could have been the case is not to be analyzed as or conflated with possibility in the 'logical' sense of what it is not logically or analytically self-contradictory to assert or assume actually is the case." [Burgess, 1996: 25]

The viewpoint of Burgess and Soames is erroneous since both Marcus and Kripke identify possibility in the so-called "metaphysical" sense of what could have been the case with logical possibility. I have already argued above that Marcus used "tautology, logically true, broadly analytic", etc., in quantified modal logic senses that are meant to have metaphysical import about what must be the case, what is the case in all possible worlds. Let us repeat a couple of sentences from Marcus I have already quoted: "... the rough outline above [her semantic construction for QS4] corresponds to the Leibnizian distinction between true in a possible world and true in all possible worlds ..." [1961: 320, my italics]. She calls a possible world a model and defines a logically true sentence as follows: "A logically true sentence is one which would be true in every modef'. [1961: 319, my italics]. A necessary truth, for Marcus, is defined in modal terms; it is not defined as a priori truth. She does not write instead: "A logically true sentence is one which can be known to be true independently of all experience". Marcus likewise defines an individual's necessary possession of a property, not in terms of it being true a priori that the individual possesses the property, but in terms of it being true that the individual possesses the property in every possible world (model). "To say of thing a that it necessarily has a property 'P(D(<pa)) is to say that 'Pa is true in every model. Self-identity would be such a property". [Marcus et aI., 1962: 133].

This identification of a logically necessary truth with a truth in every possible world is also made by Kripke. Kripke's [1971] and [1972] reiterate Plantinga's [1970] thesis that the metaphysical distinction between necessity and possibility is not equivalent to the epistemological distinction between a priori and a posteriori. This metaphysical vs. epistemological distinction is made in Kripke's work, but Soames and Burgess are quite mistaken when they say Kripke also makes a metaphysical vs. logical distinction. In fact, Kripke identified metaphysical necessity with logical necessity. This can be seen from numerous passages, a small number of which I will quote. First, consider that on the page immediately after the page where Kripke says the distinction between necessity and possibility belongs to metaphysics and not epistemology [1971: 150], he starts talking about logical possibility and necessity (which he takes to be the same thing as metaphysical possibility and necessity). He talks about "performing an infinite number of acts like looking through each number one by one. A vast philosophical literature has been written on this: Some have declared it logically impossible; others that it is logically possible; and some do not know. The main point is that it is not trivial that just because such a statement is necessary it can be known a priori" [1971: 151; my emphasis]. This shows that Kripke's point is to distinguish between a

162

QUENTIN SMITH

prioricity and logical necessity, not (contra Soames and Burgess) to distinguish between metaphysical necessity and logical necessity. Examples from Kripke's [1972] make this even more clear; at one place he writes that "it is (metaphysically) possible that there should have been animals that resembled tigers but were not tigers" [1972: 330; my emphasis; the parentheses are in the original text]. But he immediately proceeds to say he is talking about logical possibility. He says "the notion of a posteriori necessary truth may still be somewhat puzzling" and represents an objector to his theory as asking "What then can you mean when you say that such eventualities are impossible"? [1972: 331]. Kripke formulates the objection to his theory, and his answer to this objection, in terms of logical possibility. He formulates the objection as follows: "it really is logically possible that gold should have turned out to be a compound, and this table might really have turned out not be made of wood, let alone of a given particular block of wood" [1972: 332, my emphasis]. Kripke answers this objection as follows: "In the case of some necessary a posteriori truths, however, we can say that under appropriate qualitatively identical evidential situations, an appropriate corresponding qualitative statement might have been false. The loose and inaccurate statement that gold might have turned out to be a compound should be replaced (roughly) by the statement that it is logically possible that there should have been a compound with all the properties originally known to hold of gold" [1972: 333; my emphasis]. For yet another example, note that Kripke later phrases his argument about the essence of pain and brain states in terms of logical possibilities; e.g., he writes "it would seem logically possible that B could exist without any sensation with which it might plausibly be identified" [1972: 335; my emphasis]. For a final example I will give (there are many others), I will note Kripke's famous discussion of the necessity of origins. His discussion of the necessity of origins is framed in terms of "contradictions"; Kripke says there is "no contradiction" [1972: 312, my emphasis] in the announcement that the thing we thought to be the Queen is in fact an automaton, but distinguishes this noncontradictory statement from the (allegedly) contradictory statement that the Queen was born of different parents from the parents from whom she actually came. In his [1971] and [1972], Kripke consistently makes these equations: logical necessity = metaphysical necessity = obtaining in all possible worlds. He does not maintain that there is a distinction between logical necessity and metaphysical necessity. In summary, So ames and Burgess are mistaken in holding that Kripke made a major advancement over Marcus by distinguishing logical from metaphysical necessity and they are mistaken in holding that Marcus failed to have the concept of necessary truth as truth in all possible worlds. Both Marcus and Kripke identified logical with metaphysical necessity and both defined logically necessary truth as truth in all possible worlds.

DIRECT, RIGID DESIGNATION AND A POSTERIORI NECESSITY

163

The following sections will further evince that Marcus is not a RussellianMillian, but a Kripkean-Millian. But first I will rebut Marcus' recent thesis that she is not a Kripkean-Millian, i.e., that her "proper names are not assimilated to what later came to be called 'rigid designators' by Saul Kripke, although they share some features with rigid designators" [Marcus, 1993: xiiixiv]. Marcus seems to hold this view since she believes Kripke "... classifies proper names as 'rigid designators' along with rigid descriptions, thereby obscuring the difference in semantic relationship between a proper name and the object named as compared with the relationship between a rigid description and the object described". [1993: 248, n. 19]. I believe she is wrong if only for the reasons that Kripke denied that proper names refer descriptively and that Kripke held names are connected with their referents by an historical chain, whereas rigid descriptions are not. Since Kripke's alleged "assimilation" is the only reason Marcus offers for her denial that her tags are rigid designators, I think it is reasonable to conclude that her denial is based on an incorrect interpretation of Kripke's theory and that her denial does nothing to impugn the extensive textual evidence that her tags are rigid designators. 4 8. Necessary identifications and non-trivial essential attributions in Marcus [1961J

A distinction that is present in both Marcus' [1961] and Kripke's [1972] is the distinction between necessary identifications and non-trivial essential attributions. "alb" is an identification, whereas (to use Marcus' symbolism), a E Dr may make a non-trivial essential attribution (it attributes the property of being necessarily r to the individual a). Marcus introduces this symbolism in conjunction with her distinction between necessary properties that belong to some objects and necessary properties that belong to every object in the domain of objects in all possible worlds. She presents the following two definitions to capture this distinction in item (57) of her [1961]: (57)

x E Dr

=df

D(x

E

r), f- Dr

=df

(x)(x

E

Dr)

The first definition translates as "x instantiates the necessary attribute r =df necessarily, x instantiates the attribute r". The second definition translates as "r is a logically necessary attribute r =df for every x , x instantiates the necessary attribute r". Marcus makes this definitional distinction for the purposes of distinguishing between properties that necessarily belong to each thing (such as self-identity), which is captured in the second definiton, and properties that necessarily belong to some things (such as being rational), which is captured in the first definition. Her point here is to distinguish "necessary properties" in the sense of non-trivial essences, such as being rational (one of the examples she is using in discussing Quine's argument about being rational, two-legged, a cyclist and a mathematician) from "necessary properties" in the sense of

164

QUENTIN SMITH

properties that of logical necessity belong to each thing. Quine argued that the semantics of quantified modal logic is committed to non-trivial essences, but Marcus counter-argues that this semantics is committed only to logically necessary properties possessed by each thing, such as self-identity. In these definitions, Marcus is also introducing the notion of "de re necessity" in the sense of necessary properties of things, as distinct from "de dicto necessity", which pertains to statements or propositions having the property of being necessarily true. For some individual x, xED r; this means x has a necessary property r, such as being rational. By contrast D(x E r), the definitional equivalent of xED r, states that it is necessarily true that (x E r), where "necessity" is an operator that takes a "dictum" as its operand. Thus, in Marcus' symbolism, a necessary identification, such as "Necessarily, Venus is Hesperus" appears as Dalb, whereas a necessary attribution of a nontrivial essence, such as "Socrates is necessarily rational", appears as a E Dr. In Marcus' 1961 theory, the first statement is defined as meaning it is true in all possible worlds that Venus is Hesperus and the second statement is defined as meaning Socrates instantiates rationality in each possible world. (Marcus has a constant domain in her semantics for modal logic, so it would be redundant to say "Socrates instantiates rationality in each possible world in which he exists".) I have presented the necessary identity in a de dicto formulation and the necessary attribution in a de re formulation, but both necessities can be given either a de dicto or a de re formulation in Marcus' [1961] theory. In my [1995b] I explained the theory of necessary a posteriori identities in terms of object level facts (which are necessary and a priori) and metalevel facts about the relation of statements to the fact stated by them (the metalevel fact being contingent and a posteriori). However, there is a third ingredient in the theory of necessary a posteriori identities that needs to be mentioned. The statement itself, "Hesperus is Venus", is necessarily true and is a posteriori. (More exactly, it is necessarily true given the actual interpretation of these names, viz., that "Hesperus" and "Venus" both refer to Venus.) On Marcus' theory, statements of the form alb possess both the property of being necessary and (at least in some cases) the property of being a posteriori. Thus, we have one and the same item (the statement) that is necessary and a posteriori. Marcus' introduction and defense of de re necessities marks the first departure in the literature on modal logic from the orthodox view of Carnap, Quine, C.I. Lewis, Kanger and others, namely, that the notion of de re necessity is nonsensical and that "necessity resides only in our language, not in things themselves". This is one of the facts that has gone unrecognized in standard histories of modal logic and the New Theory of Reference. It should now be evident that Marcus' [1961] develops ideas that belong to the theoretical framework of Kripke's and Kaplan's 1970s writings and that her [1961] does not consist of repetitions of Russell's 1918 doctrines. Accordingly, I will now pass to comparing, developing and evaluating the theories of rigid designation that can be found in the works of Marcus, Kripke and Kaplan. In

DIRECT, RIGID DESIGNATION AND A POSTERIORI NECESSITY

165

the next section, we will concentrate on Marcus' theory of rigidity and show how it can be used to develop a "complete concept" of a rigid designator. 9. The "complete concept" of a rigid designator

Neither Marcus nor Kripke nor Kaplan talked about (what I shall call) the complete concept of a rigid designation, so here I am making something more explicit than what is present in any of their texts. The complete concept of rigid designation contains an informal part and a formal part. The informal part belongs to the philosophy of language and is an intuitive explication of the difference between rigid and nonrigid designators in ordinary language. The formal part is a definition of rigid designation in terms of a semantics for modal logic, and embodies the formal notion of the referent of a singular term at a model or possible world. Somebody has the complete concept of a rigid designation if and only if she has both the informal and formal parts. There are two ways to explicate formally the notion of rigidity, associated with Kripke and Kaplan respectively, and they are both present in Marcus' 1961 work (but are not explicitly formulated in any of their works). I shall argue in a later section that the definition associated with Kripke is accurate but the definition associated with Kaplan is not. But before I engage in a discussion of these formal explications, I shall illustrate the informal explication of the rigid/nonrigid distinction in Kripke's and Marcus' writings. The informal explication does not require a model-theoretic semantics, the notion of a possible world, or the concept of a singular proposition. The concept of rigid designation can be stated in an informal way using only modal prefixes or adverbs ("possibly", "necessarily") and subjunctives ("might have been", etc.). In the following few quotations, I am repeating my [l995b], but this is necessary to introduce my more complete account here. Kripke conveys the distinction between rigid and nonrigid designation in "Naming and Necessity" in this informal way. Kripke writes: One of the intuitive theses I will maintain in these talks is that names are rigid designators. Certainly they seem to satisfy the intuitive test mentioned above; although someone other than the U.S. President in 1970 might have been the U.S. President in 1970 (e.g., Humphrey might have), no one other than Nixon might have been Nixon ... For example, 'the President of the U.S. in 1970' designates a certain man, Nixon; but someone else (e.g., Humphrey) might have been the President in 1970, and Nixon might not have; so this designator is not rigid [1972: 270].

Here the notion of rigid designators is explained in terms of the subjunctive expressions "might have been", "might not have been", and no mention of possible worlds or singular propositions is needed. This is also how Marcus introduced the notion in her 1961 article. The rigid/nonrigid distinction between names and contingent definite descriptions is present in the following passage, but with a different terminology:

166

QUENTIN SMITH

Let us return now to (10) and (15). [(10) is "The evening star eq the morning star" and (15) is "Scott is the author of Waverley"]. If they express a true identity, then "Scott" ought to be anywhere intersubstitutable for "the author of Waverley", and similarly for "the morning star" and "the evening star". If they are not so universally intersubstitutable - that is, if our decision is that they are not simply proper names for the same thing; that they express an equivalence which is possibly false, e.g., someone else might have written Waverley, the star first seen in the evening might have been different from the star first seen in the morning - then they are not identities. [1961: 311]. 5

I shall use the conceptual materials in Marcus' [1961] to formulate the formal way of explicating the concept of rigid designation. The formal explication is based on the semantics for modal logic and includes notions of individual constants or variables, modal operators, predicates, models (worlds) and related ideas. The formal explication uses these notions to define rigidity in terms of the referent of a singular term at a model or possible world. Let us see how this explication is based on Marcus' 1961 model-theoretic construction that "corresponds to the Leibnizian distinction between true in a possible world and true in all possible worlds" [1961: 320]. Her construction embodies the simplifying assumption that there is only one two-place predicate (R). She presents: a language (L), with truth functional connectives, a modal operator (<»), a finite number of individual constants, an infinite number of individual variables, one two-place predicate (R), quantification and the usual criteria for being well-formed. A domain (D) of individuals is then considered which are named by the constants of L. A model of L is defined as a class of ordered couples (possibly empty) of D. The members of a model are exactly those pairs between which R holds. To say therefore that the atomic sentence R(al a2) of L holds or is true in M, is to say that the ordered couple (b l . b2) is a member of M, where al and a2 are the names in L of bl and b 2. If a sentence A of L is of the form ~ B, A is true in M if and only if B is not true in M. If A is of the form BI. B2 then A is true in M ifand only if both BI and B2 are true in M. If A is of the form (3x)B, then A is true in M if and only if at least one substitution instance of B is true (holds) in M. If A is <)B then A is true in M if and only if B is true in some model MI. We see that a true sentence of L is defined relative to a model and a domain of individuals. A logically true sentence is one which would be true in every model. [1961: 319].

Here the models M, Ml, etc. are possible worlds, the individuals in the domain D exist in each world, and these individuals are referents of the individual constants ai, a2, etc. In a later article, Marcus said of her 1961 theory that "in my own sketch of a semantics for modal logic, the domains of individuals assigned to alternative worlds were coextensive. Given that one of the worlds is the actual world, no entities are spawned that are not in this world and no entities of this world are absent in others". [Marcus, 1993: 195]. Marcus uses the phrase "member of a model" or "member of a world" to refer to any pair of individuals between which R holds; thus the empty world contains no "members" in the sense that it contains no pair of individuals between which R holds. If one prefers, one could say that the ordered couples are members of a model whereas the individuals in the domain D are present or exist in a model (the

DIRECT, RIGID DESIGNATION AND A POSTERIORI NECESSITY

167

"present/absent" terminology conforms to Marcus' later phraseology in [1993: 195]. The "present in" or "exists in" terminology may be explained in part in terms of the fact that the quantifiers in each model range over the domain D, which is required for her derivation of the Barcan formula. Since the domain D is fixed, Marcus' worlds differ from one another in that her relation R holds between different individuals in different worlds (or does not hold between any individuals in her empty world). For example, Alice and Ed exist in one world M as related by the relationship of friendship and exist in another world M I , but not as related by this relation. This semantics for modal logic contains the notion of a referent of a singular term at a possible world and a dyadic rule of interpretation that connects an individual constant a to an individual b in the domain D. This may be expressed as the rule V (a) = (b), where a is the individual constant and b the individual in the domain. This dyadic rule of interpretation entails the triadic rule of interpretation, V (a, M) = b, which assigns to a the referent b in the model M. This entailment holds since Marcus' domain of individuals is constant across her models. The triadic rule of interpretation is the formal semantical notion of a referent of an individual constant at a possible world. This formal semantic notion of rigid designators (be they individual variables or individual constants) is also present in earlier work by Hintikka [1961; 1963], Kripke [1959; 1963] and others. But it does not seem that this idea can be traced back earlier than the late 1950s. Marcus' writings in the 1940s are the first to display the idea that variables have the formal features of rigid designators. Her thesis [1947] of the necessity of identity for variables, (x)(y) (xly) == D(xly), is the object-language counterpart of the metalanguage statement that individual variables are rigid designators in the formal sense. However, Marcus' [1947] does not include a semantics for modal logic and such a semantics is necessary to have the formal semantical notion of rigidity. But what of Carnap? Marcus notes in her 1961 article that her semantical construction is a variant of Carnap's in his Meaning and Necessity [1947], which is related to Carnap's modal semantics in his "Modalities and Quantification" [1946]. However, a crucial difference between Marcus' semantics and Carnap's is that Carnap's does not contain the formal or informal explication of a rigid designator. Carnap's proper names are not directly referential but have intensions (individual concepts) as well as extensions and in modal contexts Carnap's names do not refer to their extensions (the relevant individuals in the various possible worlds) but to their intensions. A necessary condition of a name being rigid is that it has the same referent in both extensional and modal contexts, and Carnap's names, unlike Marcus', do not meet this condition. In fact, Carnap explicitly rejects the theory of what he calls "the name-relation" [1947: 98], which implies certain theses associated with the theory of rigid designation, most notably, the thesis that co-referring names are intersubstitutable in modal contexts.

168

QUENTIN SMITH

Carnap also does not have a formal conception of the rigidity of individual variables, for an analogous reason. Carnap writes: "the values of individual variables in a modal system are not individuals but individual concepts" [1946: 37-38]. In order for individual variables in a modal system to be rigid, their values need to be individuals, not individual concepts. 6 Accordingly, it seems that Marcus' [1961] was the first work to contain the complete concept of a rigid designator, since it included both the informal explanation in terms of natural languages and the formal semantical concept of a referent of a singular term at a possible world. In the next section, I shall use her modal semantics to construct the two different ways of defining the formal semantical concept of a referent of a singular term at a possible world, the direct-reference definition and the world-definition. 10. The direct-reference definition of rigidity versus the world-definition of rigidity

The direct-reference definition of a rigid designator is famously associated with Kaplan and he is its main contemporary champion. The direct-reference definition is suggested by Kaplan's concept of an obstinately rigid designator. A recent presentation of this concept is made in his 1989 "Afterthoughts" [Kaplan, 1989b]. Kaplan writes: If the individual is loaded into the proposition (to serve as the propositional component) before the proposition begins its round-the-worlds journey, it is hardly surprising that the proposition manages to find that same individual at all of its stops, even those in which the individual has no prior, native presence. The proposition conducted no search for a native who meets propositional specifications; it simply 'discovered' what it had carried in. In this way we achieve rigid designation. Indeed, we achieve the characteristic. direct reference, form of rigid designation, in which it is irrelevant whether the individual exists in the world at which the proposition is evaluated. [Kaplan, 1989b: 571].

This way of explicating the concept of rigid designation logically implies that characteristic rigid designation involves direct reference, since only in cases of direct reference does the relevant expression (e.g., a use of a name or indexical) introduce the entity into the proposition itself. In cases of descriptional reference, where reference is routed through sense, rigid designation is not achieved through loading the referent into the proposition. For example, the modally stable sense expressed by "the largest actual galaxy" does not introduce a galaxy into the proposition, but rather a sense, and this sense denotes whatever galaxy is the largest galaxy in a (where "a" is a name of the actual world) in each possible world in which this galaxy exists. Marcus' [1961] semantics provides the conceptual materials for a directreference definition of rigidity. The rigid designation of an individual constant a is achieved by loading an individual b from the domain D into the proposition expressed by the sentence (or, in Marcus' terminology, loading the individual into "what is said" [the singular proposition] by the "statement" [the sentenceD.

DIRECT, RIGID DESIGNATION AND A POSTERIORI NECESSITY

169

This Kaplanian distinction between the statement and what is said by the statement was first made by Marcus, as Kaplan and his colleague Joseph Almog wish to remind readers. Almog [1986: 220, n. 8] pointed out that Marcus was the first to introduce the terminology whereby we distinguish the "statement" from "what is said" by the statement, a distinction that was later used to coincide with the distinction between the sentence-token (or use of a sentence) and the singular proposition expressed by (the semantic content of) the sentence-token. Almog writes: "When I say 'Quine is a philosopher', a particular individual, Quine, becomes the subject constituent of the proposition I express .. , 'Quine' names a flesh-and-blood individual: Quine. Thus a proposition with an objectional constituent is generated, what Kaplan calls a singular proposition. This is what I say, when I say 'Quine is a philosopher' ". [1986: 220]. At this point, Almog has a footnote [1986: 220, n. 8] which reads: In modern times, I think Ruth Marcus was the first to see this, in her "Modalities and Intensional Language", Synthese, XIII, 4 (December, 1961): 303-330. She was the first to use the language "what is said" on which (see her p. 308) "a = a" and "a = b" say the same thing, when a is b. She also introduces the old Millian terminology of a name being a simple tag.

Using Marcus' initial idea and Kaplan's extensive development of it, we may formulate the direct-reference explication of the rigidity of individual constants as follows. We take an individual constant at and assign to it a tagged individual (a direct referent) b l from the domain D, and we do this prior to our "round the worlds" trip through the different models M" M 2 , etc. This is reflected in the fact that the triadic rule of interpretation V (a, M) = b is derivative from the dyadic rule V (a) = b in Marcus' semantic construction. When I utter a name, I directly dip into domain D and load the relevant individual into what is said (the singular proposition expressed) by my statement containing the tag. The rigidity of the tag is guaranteed by the actual assignment of b to the tag as the tag's direct referent, without needing to check whether the tag's descriptive conditions are met at this or that world (since the tag has no relevant descriptive conditions). But we cannot proceed here without noting a lacuna that appears in the theories being discussed. I am talking at length about directly and rigidly referential expressions and the associated singular propositions that constitute "what is said". I have said much about the linguistic expressions, but what, exactly, is a singular proposition? "It is a proposition that includes a thing as one of its constituents, and may be represented as an ordered n-tuple". Such oneliners are practically all that has been said in the literature about the parts and structure of singular propositions. Indeed, nothing beyond a few remarks have ever been offered about the parts and structure of general propositions. In an attempt to fill this lacuna, I developed a detailed theory of the parts and structure of both singular and general propositions in Part Two of Language and Time [1993a]. Alan Sidelle [1995: 680-681] seems to be one of the few who has recognized this lacuna and the need to fill it, e.g., he writes about Part Two of

170

QUENTIN SMITH

Language and Time: "Smith has some interesting discussions here both of the relation between sentences and the propositions they express, and of some semantic data that make his views here much more worthy of serious attention ..." [Side11e, 1995: 680]. Graham Nerlich presents a very interesting critical discussion of my theory of propositions in Robin Le Poidevin (ed.), Questions of Time and Tense (Oxford: Clarendon Press, 1998). Here I shall simply note what seems to me to be a major lacuna in the presently discussed accounts of rigiditythe lack of a detailed theory of singular propositions - and concentrate on formulating definitions of rigid singular terms. I believe this can be done with reasonable precision even if we bracket the question about the precise nature of a singular proposition. Philosophers today operate with an intuitive notion of a singular proposition - amounting roughly to what Kaplan has said about them in his writings - and this is sufficient for my present purposes. Using Marcus' ideas about tags and individual constants as our basis, we can formulate definitions that can apply to any expression that rigidly designates some individual. I shall refer to tokens rather than types, since at least in the case of indexicals we need to ascribe rigidity only to the various tokens of the word-type ("I", "now", "here", etc.). In my definition, I use "a" but this need not be an individual constant; it may be, for example, a referential use of a definite description. The direct-reference definition is: (DR) A token a is a rigid designator of an individual b if and only if (i) a directly refers to b in the actual world M, (ii) a directly refers to b at each possible world MJ, M 2 , etc., in which b exists, and (iii) a directly refers to b even at worlds M 3 , M 4 , etc., in which b does not exist. Condition (iii) was first formulated by Kaplan and is purely redundant in Marcus' [1961] semantics, since she held a constant domain theory; there are no worlds at which b does not exist. When Marcus is discussing a varying domain semantics in her [1971], she rejects condition (iii) and instead advocates a condition that belongs to the world-definition (see below), namely, that in respect of the worlds in which b does not exist, a does not designate anything. In "Essential Attribution", Marcus writes that "individual names don't alter their reference, except to the extent that in some worlds they may not refer at all" [1971: 194]. Condition (iii) will prove to be one of the two crucial items in my subsequent argument that the world-definition is the only acceptable definition of rigidity. The world-definition of the rigid designation of tokens includes the condition Marcus added in her 1971 article. The world-definition is: (W) A token a is a rigid designator of an individual b if and only if (i) a refers to an individual b in

the actual world M, (ii) a refers to b at each world M j, M 2 , etc. in which b exists and (iii) a does not refer to anything at worlds M 3 , M4 etc. in which b does not exist.

Apart from the difference between the conditions (iii) in the two definitions, the crucial difference between the direct-reference definition and the world-definition is that the world-definition does not imply that the token is directly

DIRECT, RIGID DESIGNATION AND A POSTERIORI NECESSITY

171

referential. (W) says the token "refers" but does not require that this reference is direct. The next question is: which is the correct definition? 11. Kripke versus Kaplan: what is the correct definition of a rigid designator?

A clarification of some of the differences between the world-definition and the direct-reference definition of rigidity may be obtained by examining an interesting exchange of letters between Kripke and Kaplan. There is a discussion between Kripke and Kaplan about whether or not Kripke held the world-explication in his 1971-1972 writings. Kripke argues against Kaplan's suggestion that he (Kripke) held the world-explication and claims that he (Kripke) remained neutral between the world- and direct-reference explications. But this is not quite precise, since Kripke and Kaplan do not specifically address the world- and direct-reference explications of rigid designation, but instead the logically related issue of whether a proper name designates nothing in respect of a world W if the bearer of the name does not exist in W. The issue is whether the definition of a rigid designator should include the idea that a rigid designator has no referent in respect of worlds where the object does not exist.

Does "Socrates" have a referent in worlds where Socrates does not exist? According to the world-explication, the answer would have to be negative, since rigidity is defined in terms of the locution picking out the object in each possible world in which the object exists. If the object does not exist in a certain world, the designator does not pick out anything at that world. Kripke presents this explication in his [1971: 146]: ... when I use the notion of rigid designator, I do not imply that the object referred to necessarily exists. All I mean is that in any possible world where the object in question does exist, in any situation where the object would exist, we use the designator in question to designate that object. In a situation where the object does not exist, then we should say that the designator has no referent and that the object in question so designated does not exist.

This allows for the distinction between weak and strong rigid designators; weak rigid designators designate objects that exist in some but not all possible worlds, and strong rigid designators designate objects that exist in all possible worlds (see Kripke [1972: 269-270]). By contrast, the direct-reference explication implies that a rigid designator designates the object even in the possible worlds where the object does not exist. The individual is directly designated and thereby becomes a component of the singular proposition expressed by the sentence, and thus does not need to exist in the world at which the proposition is evaluated. However, in a letter to Kaplan (quoted in [Kaplan, 1989b: 569-570]), Kripke disputes Kaplan's suggestion that he held the world-explication in his 1971 and 1972 papers. Kripke states he held this formulation: "a designator d of an object x is rigid, if it designates x with respect to all possible worlds, and never designates an object other than x with respect to any possible world" (quoted

172

QUENTIN SMITH

by Kaplan in [1989b: 569]). This formulation is consistent with the directreference-explication, since it implies that d designates x with respect to all worlds, and not merely with respect to all the possible worlds in which x exists. Kripke, however, seems to be mistaken here since, as Kaplan points out, Kripke explicitly stated in [1971] that "In a situation where the object does not exist, then we should say that the [rigid] designator has no referent and that the object in question so designated does not exist" [Kripke, 1971: 146]. In his letter to Kaplan, Kripke claims in regard to this sentence (transcribed from his 1970 talk at New York University), that "It is also possible, I think, that the sentence is mistranscribed from the tape of the talk. A simple change of 'and' to 'or' in the sentence would make it entirely consistent [with the formulation he offered to Kaplan in his letter] .... The corrected version would read better if 'so' were changed to 'though' (an easy mistake in the transcription of an oral presentation)" (See [Kaplan, 1989b: 570, n. 8]. However, it appears that Kripke's interpretation of his 1971-72 theory is implausible for three reasons. First, Kripke states merely that it is possible that there is a mistranscription of his oral talk, not the stronger claim that the evidence shows there actually is one. (Since the oral tape still exists, presumably the question of mis-transcription could be definitively answered by checking the oral tape.) Second and more importantly, even if we make the changes Kripke suggests, we still have the clause "In a situation where the object does not exist, then we should say that the [rigid] designator has no referent", which Kripke would have no motive for saying if Kripke actually believed the negation of this clause. Third, Kripke suggests the same view in his other 1970 talk, "Naming and Necessity", e.g., in his remark "a designator rigidly designates a certain object if it designates that object wherever the object exists; if, in addition, the object is a necessary existent, the designator can be called strongly rigid" [1972: 270]. This is the world-explication of rigid designation, with its associated distinction between strong and weak rigid designators. Nonetheless, Kripke did espouse the Kaplanian thesis in his 1980 Preface to Naming and Necessity, where he writes: "a proper name rigidly designates its referent even when we speak of counterfactual situations where that referent would not have existed" [1980: 21, n. 21]. Accordingly, we can allow that Kripke held the "Kripkean" theory in 1971-72, but in 1980 endorsed the "Kaplanian" theory. This discussion of the Kripke-Kaplan relation helps to clarify the difference between the world-definition and the direct-reference definition of rigidity. We shall now address the issue of which definition is preferable. It seems clear to me that Kripke's 1971-72 theory is preferable. There are characteristic cases of rigid designation where the designator does not directly refer to its referent, and were the referent is not a component of the proposition expressed by the sentence. For example, "the only number that immediately succeeds the number eleven", "the maximally excellent being", and "the actual

DIRECT, RIGID DESIGNATION AND A POSTERIORI NECESSITY

173

inventor of bifocals" all rigidly designate (if they designate at all), and none of these expressions directly refer to an object. Rather, they introduce a descriptive sense into the proposition expressed by the sentence, and the reference is routed through this sense; the descriptive sense is satisfied only at the worlds where the object exists, and is not satisifed by anything at worlds where the object does not exist (thereby depriving the expression of a referent). Among these three examples, only "the actual inventor of bifocals" is a weak rigid designator (the other two are either strong rigid designators or nondesignators); this definite description does not refer to anything in respect of those worlds where Ben Franklin does not exist. Given these counterexamples, it is clear that the direct-reference definition (DR) of rigid designation is false. Kaplan does briefly note that some rigid designators are not directly referential [1989b: 571, n. 12]. But Kaplan does not attempt to explain how his recognition of this fact is consistent with his explication of characteristic rigid designation. Is he trying to suggest that descriptional rigidity is "uncharacteristic" rigidity? Such a thesis would run contrary to the structure of natural languages, which characteristically exhibit descriptional rigidity as well as nondescriptional rigidity. It would seem that the recognition of characteristic rigid descriptions should motivate one to adopt the world-explication. The world-definition is preferable since it accommodates the various rigid designators that are descriptional rather than directly referential. Although versions of the direct-reference definition (DR) are now becoming more popular among philosophers than versions of the world-definition (W), I suggest that we ought to return to Kripke's 1971-72 world-definition. CONCLUSION

The main philosophical conclusion of this paper is that the world-definition of rigid designation is preferable to the direct-reference definition. Many historical theses were advanced in this paper. The claims about Marcus' priority in certain areas have been given "very wide publicity", as Burgess notes, so I would like to emphasise here my historical account of the origin of the causal theory of reference in Peter Geach's [1969]. I think the fact that Peter Geach originated the causal (historical chain) theory of reference, but has not been credited for his achievement, is almost of equal importance as the neglect of Marcus' 1961 theory of direct, rigid designation. For over 25 years now, the theory that Geach originated has been attributed to Donnellan and Kripke, and Geach's pioneering article has been almost completely ignored. Ifhistories of philosophy ought to include Marcus' priority with regard to the direct rigid designation theory and other theories, they equally ought to include Geach's priority with the historical chain theory of reference. 7 Philosophy Department Western Michigan University

174

QUENTIN SMITH NOTES

This paper presents a different and more developed account of these matters than my two earlier papers [Smith 1995a; 1995b]. I do not now agree with all the doctrines in these two earlier papers. For instance, I claimed in [1995a] that Kripke originated the historical chain theory of reference, but in the present paper I argue that Peter Geach originated this theory, and I now disagree with Marcus' present interpretation of Kripke's 1962 remark about tags and denoting phrases (see my [1998b]), There are other differences between the present paper and my [1995a; 1995b], but I will not pause to note them all; it suffices to say that I currently subscribe only to the views in [1995a; 1995b] that are consistent with the views of the present paper. As I explain below, the theme of this paper is structured in accordance with Tyler Burge's way of presenting a history of the New Theory of Reference. This paper is not intended as a response to [Soames, 1995] or [Burgess, 1996], although I do discuss some parts of their papers that are relevant to the theme of the present paper. Burgess and Soames view themselves as involved in a "personal feud" rather than a philosophical debate, as they themselves acknowledge [Soames, 1995: 191] [Burgess, 1996: 40, 45, n. 2], and some have commented that beginning with Burgess [1996] their writings appear to have taken on the character of a "vendetta". Because of this "feud" atmosphere of their papers, it has been suggested to me by several people that their writings should be ignored. However, I should like to point out that the fact that they seem to be written as "feud" or "vendetta" pieces does not entail that they are of no philosophical value or interest. Their papers contain some interesting arguments and their arguments can be abstracted from the emotional and personal context in which they are set and treated as philosophical arguments that may be address sed in a philosophical context. In the present paper, I treat in this manner their arguments that are relevant to the philosophical issues I discuss, but shall not respond to their arguments ad hominem. 2 It almost goes without saying that numerous others - Plato, Augustine, Ockam, Mill, Peirce, Wittgenstein, Williams, Ziff, etc., etc. -held direct reference theories prior to Marcus. Some of these facts, especially the commonplace about Mill, are too obvious to be worth repeating. The question is who originated this idea in the context of the New Theorists of Reference arguments against the Frege-Russell descriptional theory ofordinary names. As I said in [I 995b]: "Thus, Soames' claim that Marcus' theory that ordinary names are directly referential and not disguised descriptions was first espoused by Fitch and Smullyan is an incorrect claim. The idea is original with Marcus. (Of course, we are talking about the contemporary New Theory of Reference; as everyone knows, Mill held a direct reference theory in the 19th century.) We should also point out that some others claimed in passing that ordinary proper names are directly referential, such as Paul Fitzgerald, but they did not support their claims with the arguments about the nonequivalence of names and descriptions in modal and epistemic contexts that have since become part of the New Theory of Reference." [1995b: 221; parentheses are in the original text]. I am grateful to Alvin Plantinga for answering my requests about information regarding the background of some of his published ideas. He emphasized to me that many of the ideas grew out of discussions with Robert Sleigh, Ed Gettier, Richard Cartwright, Hector-Neri-Castenada, and others at Wayne State University at least as early as 1963 and thus that the ideas should not be credited to him alone. 4 At this juncture, it is worth noting that my interpretation of Marcus' [1961] differs in more respects than one from Marcus' recent interpretation of her [1961]. In addition to rejecting the idea that her [1961] names are rigid designators, Marcus [1993: 211] attributes the origin of the epistemic argument for direct reference to Kripke [1972], not to her [1961], whereas I argue in [1995b] it is present in her [1961]. In my [1995a: 187] I quoted and endorsed Marcus' interpretation [1993: 226-227] of Kripke's 1962 remark about tags being the essential denoting phrases for individuals; but I now argue at length that her interpretation is mistaken [Smith, 1998b]. In my [1995a: 187-188] I quoted and tacitly endorsed her recent theory [Marcus, 1993: 34, n. I] about the relationship between dictionaries and encyclopedias, but I now think her theory mistaken. (But I think her earlier theory of dictionaries in her [1961] and [1963] is correct.) Some of the other differences are explained in the text, e.g., regarding the interpretation of the similarity and differences between Donnellan's and her "referential uses of definite descriptions". The fact that Marcus disagrees with some of my theses may count as some evidence against these theses, but I

DIRECT, RIGID DESIGNATION AND A POSTERIORI NECESSITY

175

believe the textual evidence and arguments I provide evidentially outweigh the fact of her disagreements. 5 The printed text of this quoted passage in my [1995b: 230] and [1995a: 183] is erroneous since it includes "in modal contexts". This phrase appears in one of the several revisions of Marcus' paper, viz., her [1993], but not in the original text [1961] or in her first revision [1974]. This is relevant since neither Marcus [1961] nor Kripke [1971; 1972] denied that names are intersubstitutable in epistemic contexts, but only Marcus explicitly affirmed the universal intersubstitutivity thesis. I also misquoted Marcus' [1961] throughout my [1995a; 1995b] by mis-spelling Walter Scott's book Waverley as "Waverly". Neither Soames [1995] nor Burgess [1996] discuss or appear to be aware of Marcus' [1974] revision of her [1961]. I indicated in [1995b] that Soames [1995] did not appear to be aware that [Marcus, 1993], the text from which Soames quoted, was a revision of her [1961], but I showed the revisions were too minor to affect his arguments about her [1961], which I argued to be unsound for other reasons. But the revisions in [Marcus, 1974] are major and this 1974 revision is relevant to substantive questions of interpretation. This is because in Marcus' [1974], the material relating to the epistemic argument for direct reference and the a posteriori feature of necessary identities is completely removed and replaced by a very different passage that could have been used to seemingly support the contention she is not making an epistemological point in this part of her paper. See Marcus [1974: 844]. In reality, what this substantial 1974 revision shows is that Marcus revised her [1961] in 1974 in order to make a point different than the one she wanted to make in 1961. In [1993] she revised the raper again (back nearly to its original form) so that she is again making her 1961 point. Nino Cocchiarella [1984: 318] argues that individual constants are "rigid designators" in Carnap's 1946 semantics in the sense that (3x)D(a = x) is valid in Carnap's semantics. However, two problems appear with the claim that Carnap held that individual constants are rigid designators. First, Carnap argues in his related semantics in [1947] that proper names of individuals designate their intensions in modal contexts, and individual constants are the formal counterparts of proper names. Second, Carnap says in [1946: 37-38]: "the values of individual variables in a modal system are not individuals but individual concepts"; also see [Carnap, 1947: 181). Carnap does say that individual constants are "L-determinate and that there is a one-one correlation between the individuals and the individual constants" [1947: 181], but the two above-mentioned problems suggest that Carnap's individual constants are not rigid designators. 71 am grateful to Alvin Plantinga for providing me with requested information about the background of some of his publications. I am also grateful to David Kaplan for discussions that led to a more accurate account of his theory in sections 10 and 11, but this does not imply Kaplan agrees with or endorses my interpretation of his theory (let alone my "critique"). I also thank Nino Cocchiarella for his communications about a posteriori metaphysical necessity, which improved my understanding; however, Cocchiarella should not be understood as agreeing with or endorsing the theory presented in this paper. I am indebted to Dagfinn F0llesdal for providing requested information about his early ideas. I am also indebted to private communications with Michael Devitt about his ideas, but no assumptions should be made about his agreement with my views. I should like to thank three referees for Synthese for helpful comments on earlier drafts. I want to especially thank one of the three referees, who wrote a very insightful five-page single-space report on the first draft, which motivated a substantial revision of the paper. This same referee then generously wrote another very insightful eight-page single-spaced report on my revised draft, which motivated a further substantial revision. Although I did not agree with all of this referee's arguments, even the arguments with which I disagreed stimulated me to develop ideas I would not otherwise have developed. (The paper subsequently underwent several more substantial revisions and some parts of this paper that benefitted from the referee's comments were omitted.)

REFERENCES Almog, Joseph: 1986, 'Naming without Necessity', The Journal of Philosophy, 83, 210-24. Burge, Tyler: 1992, 'Philosophy of Language and Mind: 1950-1990', The Philosophical Review, 101(1), 3-51.

176

QUENTIN SMITH

Burgess, Iohn: 1996, 'Marcus, Kripke, and Names', Philosophical Studies, 84, 1-47. Carnap, Rudolph: 1947, Meaning and Necessity. University of Chicago Press, Chicago. Carnap, Rudolph: 1946, 'Modalities and Quantification', Journal of Symbolic Logic, 11, 33-64. Church, A.: 1946, 'A Formulation of the Logic of Sense and Denotation' (abstract), Journal of Symbolic Logic, 11, 31. Church, A.: 1943, 'Review of Quine', Journal of Symbolic Logic, 8, 45-47. Cocchiarella, Nino: 1984, 'Philosophical Perspectives on Quantification in Tense and Modal Logic', in D. Gabby and F. Guenther, eds, Handbook of Philosophical Logic, Volume 2: Extensions of Classical Logic. D. R~idel Publishing Co., Dordrecht, pp. 309-353. Devitt, Michael: (forthcoming). 'Reference' in Routledge Encyclopedia of Philosophy. Routledge, London. Devitt, Michael: 1989, 'Against Direct Reference', in P. French, ed., Midwest Studies in Philosophy, Volume XIV: Contemporary Perspectives in the Philosophy of Longuage II. University of Notre Dame Press, Notre Dame, 206-240. Devitt, Michael: 1981, Designation. Oxford University Press, New York. Devitt, Michael: 1974, 'Singular Terms', Journal of Philosophy, 71, 183-205. Devitt, Michael: 1972, The Semantics of Proper Names: A Causal Theory. PhD Thesis, Harvard. Donnellan, Keith: 1974, 'Speaking of Nothing', The Philosophical Review, 83,3-32. Donnellan, Keith: 1972, Reprint of 'Proper Names and Identifying Descriptions', (1970) in D.Davidson and G. Harmon, Semantics of Natural Longuages. D.Reidel, Dordrecht. Donnellan, Keith: 1970, 'Proper Names and Identifying Descriptions', Synthese, 21 (3/4), 335-358. Donnellan, Keith: 1966, 'Reference and Definite Descriptions', The Philosophical Review, 75, 281304. F0llesdal, Dagfinn: 1986, 'Essentialism and Reference', in Hahan and Schilpp, eds., The Philosophy of w.v. Quine. Open Court, La Salle. F0llesdal, Dagfinn: 1967, 'Knowledge, Identity, and Existence', Theoria, 33, 1-27. F0llesdal, Dagfinn: 1961, Referential Opacity and Modal Logic. Harvard University dissertation, unpublished. Geach, P.T.: 1969, The Perils of Pauline', Review of Metaphysics, 23, 287-300. Gurlank: 1974, Webster's New World Dictionary. William Collings and World Publishing Co., New York. Grice, H.P.: 1969, 'Vacuous Names', in Donald Davidson and Iaakko Hintikka, eds., Words and Objections. D. Reidel, Dordrecht, pp. 118-145. Hintikka, Jaakko: 1963, The Modes of Modality', Acta Philosophica Fennica, 16, 65-79. Hintikka, Iaakko: 1961, 'Modality and Quantification', Theoria, 27, 119-128. Hintikka, Iaakko: 1957a, 'Quantifiers in Deontic Logic', Societa Scientariarum Fennica, Commentationes Human Litteraarum, 23, 3-23. Hintikka, Iaakko: 1957b, 'Modality as Referential Multiplicity', Ajatus, 20, 49--{j4. Hintikka, Iaakko and Sandu, Gabriel: 1995, The Fallacies of the New Theory of Reference', Synthese, 104(2), 245-283. Holt, Jim: 1996, 'Whose Idea is it, Anyway?', Lingua Franca, February 1996 issue: 29-39. Kaplan, David: 1989a, 'Demonstratives', in 1. Almog et aI., eds., Themes from Kaplan. Oxford University Press, New York. Kaplan, David: 1989b, 'Afterthoughts', in 1. Almog, I. Perry and H. Wettstein, eds., Themesfrom Kaplan. Oxford University Press, New York, pp. 565-614. Kaplan, David: 1986, 'Opacity', in E. Hahn and P. Schilpp, eds., The Philosophy of w.v.o. Open Court Publishers, La Salle, Illinois. Kaplan, David: 1978, 'Dthat', in Peter Cole, ed., Syntax and Semantics, Volume 9. Academic Press, New York, pp. 221-253. Kanger, Stig: 1957, Provability in Logic. Stockholm. Kripke, Saul: 1980, Naming and Necessity. Harvard University Press, Cambridge, Mass. Kripke, Saul: 1977, 'Speaker's Reference and Semantic Reference', in French et aI., eds., Contemporary Perspectives in the Philosophy of Language. University of Minnesota Press,

DIRECT, RIGID DESIGNATION AND A POSTERIORI NECESSITY

177

Minneapolis. Reprinted in Martinich, ed., The Philosophy of Language. Oxford University Press, New York. Kripke, Saul: 1972, 'Naming and Necessity', in D. Davidson and G. Harman, eds., Semantics of Natural Language. Reidel, Dordrecht, pp. 253-355. Kripke, Saul: 1971, 'Identity and Necessity', in M. Munitz, ed., Identity and Individuation. New York University Press, New York. Kripke, Saul: 1963a: 'Semantical Analysis of Modal Logic I: Normal Propositional Calculi', Zeitschrififur Mathematische Logik, 9,67 -96. Kripke, Saul: 1963b, 'Semantical Considerations on Modal Logic, Acta Philosophica Fennica, 16, 83-94. Kripke, Saul: 1959, 'A Completeness Theorem in Modal Logic', Journal of Symbolic Logic, 24, 1-14. Linsky, Leonard: 1977, Names and Descriptions. University of Chicago Press, Chicago. Marcus, Ruth Barcan: 1993, Modalities. Oxford University Press, New York. Marcus, Ruth Barcan: 1981, 'A Proposed Solution to a Puzzle About Belief', Midwest Studies in Philosophy, 6, 501-510. Marcus, Ruth Barean: 1978, 'Nominalism and the Substitutional Quantifier', Monist, 61, 351-362. Marcus, Ruth Barean: 1974, 'Modalities and Intensional Languages' (revised version of[I96I]), in F. Zabeeh, E. Klempke and A. Jacobson, eds., Readings in Semantics. University of Illinois Press, Urbana. Marcus, Ruth Barean: 1971, 'Essential Attribution', Journal of Philosophy, 7, 187-202. Marcus, Ruth Barean: 1967, 'Essentialism in Modal Logic, Nolis, 1,90-97. Marcus, Ruth Barcan: 1963, 'Classes and Attributes in Extended Modal Systems', Proceedings of a Colloquium on Modal and Many-Valued Logic. Acta Philosophica Fennica, 123-136. Marcus, Ruth Barcan: 1961, 'Modalities in Intensional Language', Synthese, 13, 303-322. [Marcus] Barcan, R.: 1947, 'The Identity of Individuals in a Strict Functional Calculus of First Order', Journal of Symbolic Logic, 12, 12-15. [Marcus] Barcan, R.: 1946, 'A Functional Calculus of First Order Based on Strict Implication', Journal of Symbolic Logic, 11,1-16. Marcus, Ruth Barcan et al.: 1962, 'Discussion of the Paper of Ruth B. Marcus', Synthese, 14, 132143. Martinich, A.: 1985, The Philosophy of Language. Oxford University Press, New York. Plantinga, Alvin: 1978, The Boethian Compromise', American Philosophical Quarterly, 15, 129138. Plantinga, Alvin: 1974, The Nature of Necessity. Clarendon Press, Oxford. Plantinga, Alvin: 1970, 'World and Essence', The Philosophical Review, 79, 461-492. Plantinga, Alvin: 1969, 'De Re Et De Dicto', Nous, 3, 235-258. Plantinga, Alvin: 1967, God and Other Minds. Cornell University Press, Cornell. Quine, W.v.O.: 1949, 'Designation and Existence', The Journal of Philosophy, 36 (1939), 701-709. Quine, W.v.O.: 1947, The Problem ofInterpreting Modal Logic', The Journal of Symbolic Logic, 12, 43-48. Quine, w.V.O.: 1943, 'Notes on Existence and Necessity', The Journal o/Philosophy, 40, 113-127. Quine, w.V.O.: 1941, 'Whitehead and Modern Logic', in P. Schi1pp, ed., The Philosophy of Alfred North Whitehead. Tudor Publishing Co., New York. Raffman, D. and Schumm, G.: 1996, 'Marcus, Ruth Barcan', in D. Borchert, ed., The Encyclopedia of Philosophy Supplement. Simon & Schuster Macmillan, New York, pp. 322-323. Russell, Bertrand: 1919, Introduction to Mathematical Philosophy. George Allen & Unwin Ltd, London. Russell, Bertrand: 1918, 'The Philosophy of Logical Atomism', in R. Marsh, ed., Logic and Knowledge. Capricorn Books, New York, 1971. Russell, Bertrand: 1905, 'On Denoting', in R. Marsh, ed., Logic and Knowledge. Capricorn Books Ltd, New York. Russell, Bertrand: 1903, Principles of Mathematics. The Norton Library, New York. Salmon, Nathan: 1989, 'The Logic of What Might Have Been', The Philosophical Review, 98, 3-34.

178

QUENTIN SMITH

Salmon, Nathan: 1986, Frege's Puzzle. MIT, Cambridge, Mass. Salmon, Nathan: 1981, Reference and Essence. Princeton University Press. Sidelle, Alan: 1995, Review of Quentin Smith's Language and Time, The Review of Metaphysics, 48, 679-681. Sidelle, Alan: 1989, Necessity, Essence and Individuation. Cornell University Press, Ithaca. Smith, Quentin: 1998b, 'A More Comprehensive History of the New Theory of Reference', this volume. Smith, Quentin: 1997, Ethical and Religious Thought in Analytic Philosophy of Language. Yale University Press, New Haven. Smith, Quentin: 1995a, 'Marcus, Kripke, and the Origin of the New Theory of Reference, Synthese, 104, 179-189. Smith, Quentin: 1995b, 'Marcus and the New Theory of Reference: A Reply to Scott Soames', Synthese, 104,217-244. Smith, Quentin: 1994, 'Temporal Indexicals', in. L. Nathan Oaklander and Quentin Smith, eds., The New Theory of Time. Yale University Press, New Haven. Smith, Quentin: 1993a, Language and Time. Oxford University Press, New York. Smith, Quentin: 1991, 'The New Theory of Reference Entails Absolute Time and Space', Philosophy of Science, 58, 411-16. Smith, Quentin: 1989, 'The Multiple Uses of Indexicals', Synthese, 78,167-91. Smith, Quentin: 1988, 'Tensed States of Affairs and Possible Worlds', Grazer Philosophische Studien, 31,225-35. Smith, Quentin: 1986, The Felt Meanings of the World: A Metaphysics of Feeling. Purdue University Press, West Lafayette. Smith, Quentin and Oaklander, L. Nathan: 1995, Time, Change and Freedom. Routledge, London. Smith, Quentin and Craig, William Lane: 1993b, Theism, Atheism and Big Bang Cosmology. Clarendon Press, Oxford. Scott, Soames: 1995, 'Revisionism about Reference', Synthese, 104, 191-216. Smullyan, Arthur: 1947, Review of Quine'S 'The Problem of Interpreting Modal Logic', The Journal of Symbolic Logic, 12, 139-141. Smullyan, Arthur: 1948, 'Modality and Description', The Journal of Symbolic Logic, 13, 31-37. Wettstein, Howard: 1986, 'Has Semantics Rested on a Mistake?', The Journal of Philosophy, 83, 185-209. Yablo, Stephen: 1993, 'Is Conceivability a Guide to Possibility?" Philosophy and Phenomenological Research, 53, 1-42.

PART III HISTORICAL ORIGINS

DAGFINN F0LLESDAL

REFERENTIAL OPACITY AND MODAL LOGIC*

SINGULAR TERMS AND SINGULAR INFERENCE

§ 16. Definite descriptions in modal logic

In § 12 we found that the formula (1)

(x)(y)(x = y.:::J N(x = y))

is derivable in any semantically complete system of quantified modal logic with identity. In § 13 this result was used to establish that Carnap's system of modal logic in Meaning and Necessity is semantically incomplete if '=' is construed as an identity sign (Theorem III). We mentioned there that if S is a semantically complete system of quantified modal logic with identity, in which descriptions are treated with Fregean, not the Russellian way, as singular terms which in particular have access to identity contexts and may be used to instantiate universal quantifications, S contains among its theorems: (2)

r-(m)'P = (w.)'lj;.:::J N[(uy)'P = (m)'lj;r

Now (2) is an undesirable result. Let us consider our example from § 11 again, with two possible worlds, and two objects a and b: Possible, nonactual world

Fa.-Ga.-Ha.Ka

-Fb. Gb. Hb.-Kb

Actual world

Fa.-Ga.Ha.-Ka

-Fb.Gb.-Hb.-Kb

This example satisfies the standard axioms and rules of inference of quantification theory and identity theory, supplemented by e.g., S5 or any weaker system of modal logic. But the example does not permit a theory of descriptions treating descriptions as names, since in the example we would then have e.g., (3)

(1x)Hx = (1X)Fx. -N[(1X)Hx = (1x)Fx]

which contradicts (2). *Editor's note (1997): This paper comprises the entirety of §§ 16-19 of the author's Ph.D. thesis, submitted to Harvard University on April 3, 1961.]

181 P. W. Humphreys et al. (eds.), The New Theory of Reference © Kluwer Academic Publishers 1998

182

DAGFINN F0LLESDAL

Not all philosophers, however, regard (2) as undesirable. Thus, discussing descriptions in modal logic, Feys writes in "Les systemes formalises des modalites aristoteliciennes" (1950): Mais n'y a-t-il pas deux sortes d'identite, i et I? Et n'y a-t-il donc pas deux sortes de descriptions, de sorte que ce qui fait un objet unique dans un sens ne fait pas un objet unique dans l'autre? Pour reprendre un exemple autour duquel a ete ecrite une imposante serie d'artic1es, "I'Etoile du Soir" et "1'Etoile du Matin" se trouve "materiellement" etre Ie meme astre, Venus; elles ont donc, en fait, les memes attributs. Mais ces memes attributs n'appartiennent pas a l'Etoile du Soir et a l'Etoile du Matin en vertu d'une necessite logique. Donc, du point de vue de I'identite i, les deux etoiles ne font qu'un; du point de vue de I'identite I elles sont deux. De meme Monsieur x appraissant en qualite de a et Monsieur x apparaissant en qualite de b pourront devoir etre identifies du point de vue de l'identite i et distingues du point de vue de I'identite 1. Cette situation paradoxale existe en effet pour les systemes plus faibles que S4,1 et, pour cette raison comme pour d'autres, ces systemes meriteraient d'etre etudies plus amplement. Mais dans les systemes sur lesquels l'attention s'est jusqui'ici principalement portee, S4 et S5, la question ne se pose pas, puisque i et I s'identifient. (pp. 507-8).

But there are other, harder-to-digest consequences of treating descriptions as proper names in modal logic. Let S be an interpreted system of modal logic which contains the axioms and rules of Godel's basic system, supplemented by a standard set of axioms and rules for quantification theory and identity theory, and a standard theory of descriptions? in which descriptions are defined the way Carnap defines them in Meaning and Necessity: (4)

'00

(1X)( . . x)oo'

for '(Ey)(x)[( .. x .. =c. x = y). ooyoo]v. -(Ey)(x)( .. x.. =c. x = y). ooa * 00'

Meaning and Necessity, 8-2. a * is some arbitrary individual, selected as a common description for all descriptions in which' .. x.. ' is true of no object or of more than one object.) Then one may prove that if among the objects of our universe there is one, Z, which is necessarily distinct from a *, then any true sentence of the system S is necessarily (necessarily necessarily, etc.) true. In the proof we will find it convenient to make use of the theorem: (5)

'Ncp. N1/!.::) N(cp.1/!)'

which, as proved by Lewis,3 holds already in Sl, and afortiori in Godel's basic system. The proof that modal distinctions collapse in S under the above conditions runs as follows:

183

REFERENTIAL OPACITY AND MODAL LOGIC

*(1')

(1X)(X = z.p) = Z.Z # a*

*(2')

(Ey)[(x)(x = z. p.::. x = y). y = z]v. -(Ey)(x)(x = z. p.::. x = y).a* = z

(I') by def. (4) above

*(3')

(Ey)[(x)(x = z. p.::. x = y). y = z]

(1')(2')

*(4')

(x)(x = z. p.::. x = y)

(3') y

*(5')

y = z. p. ::. y = y

(4')

*(6')

p

(5')

(7')

(1X)(Z = z.p) = z.z # a*.:l p

*(6')

(8')

N((1X)(X=Z.p)=z.z #a*.:lp)

(7') RL

*(9')

p. N(z # a*)

*(10')

x = z.p.:: x = z

(9')

*(11')

(x)(x = z.p.::.x = z)

(lO')x

*( 12')

(x)(x = z. p.::. x = z). z = z

(11')

*( 13')

(Ey)((x)(x = z. p.::. x = y). y = z)

(12')

*(14')

(Ey)((x)(x = z. p.::. x = y). y = z)v. -(Ey)(x)(x = z. p.:: x = y). a* = z

(13')

*(15')

(1X) (x = z.p) = z

(14') by def. (4) above

*(16')

(x)(y)(x = y.:l N(x = y)

Theorem I or II

*(17')

(Y)((1X)(X = Z. p) = y.:l N((1X)(X = Z. p) = y))

(16')

*(18')

(1X)(X = Z. p) = z.:l N((1X)(X = Z. p) = z)

(17')

*(19')

N((1X)(X = Z. p) = z). N(z # a*).:l N((1X)(X = z.p) = z.z # a*)

Theorem 5 above

*(20')

N((1X)(x = z.p) = z.z # a*.:l p) :l. N((1X)(x = z.p) = z.z # a*) :l Np

A.2

*(21') (22')

Np

(8') (9') (15') (18') (19') (20')

p. N(z # a*).:l Np

*(21')

So if we pick for our z any object necessarily distinct from a*, then any true sentence turns out to be necessarily true. In the above proof, 'p' may represent any sentence. Thus we may conclude not only

184

DAGFINN F0LLESDAL

p. N(z

=f. a*). :::> Np

but also Np. N(z =f. a*). NNp NNp. N(z =f. a*). :::> NNNp

etc. By truth-functional logic we then have p. N(z

(6)

=f. a*).:::>

N(l) ... N(m)P

for any natural number m. 4 In the system got by supplementing Godel's basic system with "Brouwer's axiom" '( r.p ::) NMr.p)'" we may similarly, by Theorem XV, derive the stronger (6')

p. z

=f. a*. :::>

N(l) ... N(m)P

for any natural number m. In a Russellian theory of descriptions, the condition 'N(x i- a*)' (or 'z could be dropped altogether, and we might conclude simply

i- a*')

p :::> N(l) ... N(m)P

(6")

for any natural number m. 4 So we may conclude: (XVI)

Let 3 be an interpreted system of modal logic got by supplementing G6del's basic system (or a stronger system, e.g., S4 or S5) by a standard set of axioms and rules for quantification theory with identity. Then if S contains a standard theory of descriptions (i.e., a theory of descriptions in which any well-formed sentence of the system with 'x' as the only free variable is permitted to take the place of 'Fx' in '(1X)Fx', then

1)

If descriptions are treated as names along the lines of Carnap's Meaning and Necessity, and if 'N(z of a*)' is true for some z, or if "Brouwer's axiom" ''P :::) NM'P~ is a theorem of S and 'z of a*' is true, then modal distinctions collapse in S.

2)

If a Russellian theory of descriptions is used, and if, as is usual in extensional logic, (1Q)'P is treated as a name when '(Ei3)(a)('P =.(3 = a)~is true, then modal distinctions collapse in S.

Similar awkward consequences follow, of course, in modal systems containing a theory of classes in which class abstracts are defined with the help of descriptions of the type above. Disastrous consequences follow also if one treats class abstracts as names, adopts the following identity condition for classes: (7)

a

= {3. =. (x) (x

E a. =. x E {3)

and allows substitutivity of concretion, viz. that

185

REFERENTIAL OPACITY AND MODAL LOGIC

(8)

'xEz( ... z ... )'and' ... x .. .'

may replace each other in all contexts. When Godel's basic system or a stronger system of sentential modal logic is supplemented by such a system of class theory, modal distinctions collapse unconditionally, as is seen from the following proof:

= x.p) = x(x = x)

*(1')

x(x

*(2')

(x)(x E x(x

*(3')

x E x(x

*(4')

x = x. p. =. x

*(5')

p

(7')

N(x(x

*(9')

x

=x

= x. p) = x(x = x). ~ p

x(x

p

= x. p). =. x E x(x = x)

(1') by (7) (2') (3') by (8) (4')

(6')

*(8')

= x. p). =. x E x(x = x))

= x.p) = x(x = x). ~ p)

= x.p.=.x = x = x.p). =.x E x(x = x)

* (5')

(6')RL

(8')

*(10')

x E x(x

*(11')

(x)(x E x(x

*(12')

x(x

*(13')

(x)(y)(x

*(14')

(y)(x(x

*(15')

x(x

*(16')

N(x(x N(x(x

*(17')

Np

(7') (12') (15') (16')

p~Np

* (17')

( 18')

= x. p). =. x E x(x = x))

= x.p) = x(x = x) = y. ~ N(x = y))

= x. p) = y. ~ N(x(x = x. p) = y))

= x. p) = x(x = x). ~ N(x(x = x. p) = x(x = x)) = x. p) = x(x = x). ~ p) ~. = x. p) = x(x = x)) ~ Np

(9') or (8) (10') x (11') by (7) Theorem I (or II)

(13') (14') GOdel's axiom A.2

By reasoning parallel to that which led to (6) above, we arrive at: (9)

p ~ N(l) ... N(m)P

We may therefore conclude: (XVII) Let S be a system of modal logic got by supplementing GOdel's basic system (or a stronger system, e.g., S4 or S5) by a system of class theory which contains a standard theory of identity (so that, in particular, the identity relation is substitutive with respect to all

186

DAGFINN F0LLESDAL contexts expressible in S), and such that classes are identical if they have the same members. Then if in S class abstracts are treated as names and substitutivity of concretion is allowed, modal distinctions collapse in S.

Among the proposed systems of modal logic, the second order functional calculi of Miss Barcan and Bayart permit class abstraction, but do not treat class abstracts as names. (Their class abstracts cannot be used to instantiate universal generalizations, and do not have access to identity contexts.) Carnap's system S2 in Meaning and Necessity is the only proposed system of modal logic which incorporates a theory of descriptions. In S2 descriptions may be used to instantiate universal generalizations and have access to '=:' contexts and to '~' contexts. The disastrous result XVI above is, however, avoided in S2 regardless of whether we construe the identity relation as '=:' or '~'. If we were to consider '=:' as an identity sign, then as we saw in § 13, S2 is incomplete since (1)

(x)(y)(x =: y.

~

N(x =: y))

is not among its theorem. But (1) is presupposed in the proof of theorem XVI (line 16». (1) would have been derivable in S2 if '=:' had been universally substitutive. So S2 is saved from a collapse of modal distinctions simply because identity of extension looks this most characteristic feature of an identity relation. What Carnap calls "identity of intension" behaves, however, as we saw in § 13 exactly the wayan identity relation is supposed to do. Nevertheless the disastrous result XVI is avoided in S2 when we use '~' for the '='. For in the proof of XVI, the description '1x)(x~z. p)' occurs in several of the lines. And since 'x~z' is short for 'N(x =: z)' (M&N 39-6), descriptions of this type are not permitted by Carnap, who on p. 184 of Meaning and Necessity states that "in order to avoid certain complications, which cannot be explained here, it seems advisable to admit in S2 only descriptions which do not contain 'N'''. Since in any proof of theorem XVI we must be able to prove that 'p' is true if and only if the entity described in a description is necessarily distinct from a *, it seems inevitable that at least one description in the proof contains the sign '~', i.e., an 'N'. So although S2 contains a theory of descriptions, S2 is saved from a collapse of modal distinctions by the circumstances that its theory of descriptions is not a standard one. (Cf. footnote 2 of this section.)

REFERENTIAL OPACITY AND MODAL LOGIC

187

17. Singular terms and the individuation of our objects

In § 13 we observed that in Carnap's system S2 the variables take as their values individual concepts, not individuals, properties, not classes, and propositions, not truth values (Theorem IV). One might be tempted to think that an intensional ontology is a price we have to pay for quantifying into modal contexts. But this is not so. The variables and quantifiers by themselves do not require us to individuate our entities finer than they are individuated in extensional logic. The trouble comes in with our definite singular terms. If, e.g., the descriptions 'the Morning Star' and 'the Evening Star' are among our singular terms, and if in some world which is possible with respect to the actual world these two descriptions refer to distinct entities, then by (1)

(x)(y)(x

=

y.:J N(x

=

y))

they have to refer to distinct entities also in our actual world. And the material object to which they both refer in non-modal logic therefore has to be expurgated from our universe of discourse. In order to avoid difficulties of this type, the only kind of descriptions which we may treat like singular terms are those descriptions ~ (1(}:)'P ~ in which 'P satisfies the condition: (2)

~(E,B)N(a)('P=.=,B)~

or in general, if we admit iterated modalities:

where the variables range over only the objects we want to admit in our universe of discourse, i.e., extensions. Abstracts, whether defined by descriptions or directly, are subject to parallel restrictions. Evidently words like 'Hesperus' and 'Phosphorus' give rise to similar difficulties. A natural way out of these difficulties would be to treat such name-like words just the way we would treat descriptions. We replace them in context using contextual definitions of the Russellian type, insisting with Quine on the "primacy of predicates" (Methods of Logic, p. 218). Instead of saying that Hesperus and Phosphorus are identical we might, e.g., say that there is one and only one object which Hesperizes and one and only one object which Phosphorizes, and these objects are identical. This solution leads us to regard a word as a proper name of an object only if it refers to this one and the same object in all possible worlds. This does not seem unnatural. Neither does it seem preposterous to assume as we just did, that if a name-like word does not stick to one and the same object in all possible worlds, the word contains some descriptive elements. 5 Frege held that "every grammatically well-formed expression representing a proper name always has a sense". 6 And Quine, by insisting on the primacy of predicates, 7 eliminates all

188

DAGFINN F0LLESDAL

definite singular terms along the above lines. Russell, in "Knowledge by acquaintance" held that 'I' and 'this' are the only two words "which do not assign a property to an object, but merely and solely name it".8 In a footnote added in 1917 Russell excluded 'I' from this list. The only remaining word 'this' will hardly be used as a definite singular term in any system of logic, modal or non-modal. The name-like words which disturbed Russell were the ones which fail to refer to one and only one object in our actual world. The ones which disturb us in quantified modal logic are the ones which fail to refer to one and the same object in all possible worlds. And the solution outlined above strongly resembles Russell's. Russell permitted a description '(1OO)t.p ~ to be treated as a singular term if and only if the following condition is satisfied (3)

'(E,6)(oo)(
In modal logic, we had to require: (2)

'(E,6)N(oo)(

= ,6)~

or rather the stronger (2'). If and only if this condition is satisfied can '(100) t.p ~ be treated as a singular term. If we want to keep extensional objects in our universe of discourse in a system of quantified modal logic three main approaches are now open to us: (1)

We eliminate all definite singular terms, e.g., by insisting on the primacy of predicates and replacing them in context as illustrated above in connection with the Hesperus and Phosphorus example. ('Fa' is replaced by '(Ey)[(x)(x a's =. x = y). Fy]" etc. How much of the context we should include in 'F', i.e., the scope of the "description", would be a problem on a par with other problems raised by logical paraphrase.)

(2)

We permit some singular terms to appear in well-formed expressions of our modal system, viz. those expressions p, which satisfy the condition '(E,6)N(oo)(oop,'s =.00 = ,6)~ or, if we permit iterated modalities, the stronger condition 'N(l) ... N(m) (E,6)N(1) ... N(n) (oo)(oop,'s =. oo=,6)~. Among these singular terms there may well be descriptions and class abstracts. For descriptions the condition just stated is equivalent to condition (2') above.

(3)

We use a full theory of descriptions. Descriptions are eliminated by standard Russellian contextual definitions. Thus, e.g., '(1OO)t.p =,6~ expands into '(oo)(t.p=.oo=,6)~ etc. We do, however, not treat a description' (1OO)t.p ~ as a name unless the supporting lemma '(E,6)N(a)(

is at hand.

= ,6)~ 9

REFERENTIAL OPACITY AND MODAL LOGIC

189

Thus one of our laws for substitution of a description for a singular term becomes: If 'l/J I is like 'l/J except for containing free occurrences of '( 1O~)
x'

Class abstracts are treated along similar lines. All name-like expressions p, which do not satisfy the condition' (E,B)N(a)(a p,'s ==. a = (3)' are treated as covert descriptions or class abstracts. Each of these three ways of handling singular terms has its virtues and vices. Approach 2) comes close to Hilbert and Bernays' treatment of descriptions JO and shares its main drawback, the difficulties we get in determining whether a given string of marks is a well-formed formula. I I In most formal systems one wants an effective procedure for determining well-formedness. Hilbert and Bernays' system and approach 2) above lack such a procedure. When, unlike Hilbert and Bernays, one admits factual sentences into one's system, then the notion of well-formedness may even come to depend on questions of facts. That is, however, happily avoided in our case, due to the 'N' in the condition

190

DAGFINN F0LLESDAL

r(E,B)N(a)(cp =.a = ,B)'. Another drawback of approach 2) is that the category of singular terms comes to vary not only as we pass from extensional logic to modal logic, but also as we pass from e.g., the logical modalities to other types of modalities. Thus, as we shall see in § 23, in systems or epistemic modalities the operator 'N' in (2) and (2') and in the corresponding conditions in 2) must be replaced by a 'V' ('it is verified that' or 'it is known that'). Approach 3), which follows Russellian lines, does not share the former of these two defects. On approach 3) we may easily get a simple, effective procedure for determining well-formedness. The difficulty relating to our category of singular terms will, however, remain with us if we decide to regard as covert descriptions only those name-like expressions J-L which fail to satisfy r (E,B)N(a) (aJ-L's =. a = ,B)'. This difficulty may, however, be overcome if we treat all name-like expressions, also those which satisfy this condition, as covert descriptions. Another difficulty relating to approach 3) is that our laws of substitution and universal instantiation for descriptions come to vary as we pass from an extensional logic to modal logic, and from one type of modalities to another. The main drawback of approach 3) seems, however, to be that we get difficulties with the universal substitutivity of identity. Thus, e.g., 'the Morning Star = the Evening Star' may be true and yet 'the Morning Star' not universally substitutable for 'the Evening Star'. These difficulties are best avoided e.g., by saying that an '=' flanked by a description is not an identity sign. It should, however, be noticed that when such a statement is expanded, an ordinary identity sign is used in the expression. Thus e.g., r(1a)cp = ,B', where the '=' is not an identity sign, goes into r (a) (cp =. a = ,B)' where the '=' is an identity sign. This difficulty concerning substitutivity of identity could be avoided if we based our theory of descriptions on a set of contextual definitions different from that of Russell. Instead of like Russell letting e.g., 'G(1x)Fx' expand into '(Ey)((x)(Fx =.x = y).Gy)', we could let it expand into '(Ey)(N(x) (Fx =. x = y). Gy)'. Then any two expressions J-L and 1/ for which r J-L = 1/' is true could be substituted for each other in all contexts. On the other hand, a statement like 'the Morning Star = the Evening Star' would then be false. Approach 1) avoids all the difficulties mentioned above. Its drawbacks for practical purposes are, however, obvious: The system becomes cumbersome, lacking the convenience afforded by a theory of descriptions and class abstraction. However, nothing prevents one who chooses alternative I) from developing a system of shorthand notations and proofs, utilizing descriptions and class abstracts as one would do on Approach 3). This shorthand notation would not be a part of the formal system itself, but would only be an adjunct to it, added in order to facilitate the use of the system. Approach 1), supplemented by such a shorthand notation, seems to be the most recommendable approach. It combines the theoretical virtues of ap-

REFERENTIAL OPACITY AND MODAL LOGIC

191

proach 1) with the practical advantages of approach 3). 18. Substitutivity of identity and other types of inference turning on singular terms

While inference in sentential modal logic does not turn on singular terms, the advent of quantification theory and identity theory introduces three types of inference turning on singular terms, viz. universal instantiation, existential generalization, and substitution. In the preceding section we found that in modal logic an expression J.L can be expected to behave like a singular term if and only if it satisfies the condition (1)

'N(I) ... N(m) (E,8)N(I) ... N(n) (a) (ap,'s =. a = ,8)'

for all natural numbers m and n. (The condition stated in the outline of approach 2) of the preceding section.) A description '(1a)cp', in particular, can be expected to behave like a singular term if and only if (2)

'(N(I) ... N(m) (E,8)N(l) ... N(n) (a)(cp

=. a =

,8),

is satisfied «(2') of the preceding section). Two other views on these three types of inference in modal logic have, however, been propounded. One is that of Smullyan, who as mentioned in footnote 2 of § 7 holds that all three types of inference work in modal logic if descriptions are contextually defined and one is judicious in choosing the scope of the descriptions. Another rather widespread view is that these three principles of inference, in particular substitutivity of identity, do not work and should not be expected to work in modal contexts. In this section these two views will be examined. Smullyan, in his review of Quine'S "The problem of interpreting modal logic" and in the article "Modality and description" argues that one is free to use the full apparatus of Russellian descriptions in modal logic, and that all the laws laid down for descriptions in Principia Mathematica work also in modal logic. One only has to be judicious in choosing the scopes of the descriptions, and one must in particular observe that in modal contexts the scope of a description' (1a)cp' matters also when the condition

is satisfied. Also contextually defined class abstracts can be freely admitted in modal logic if one, contrary to custom in extensional logic, uses prefixes to indicate their scope. Many others, including Miss Barcan and Fitch, have expressed that they are largely in agreement with this view. 12 But Wilson's counter-example,'3 quoted in footnote 2 of § 7 above, shows that something is wrong with it. The general difficulty illustrated by Wilson's example is that if we substitute a description

192

DAGFINN F0LLESDAL

for a description the scope is already selected, and there is no question of judiciously choosing the scope: (1X)(WX) = (1x)(Mx) N{[(1X)(WX)S(1X)(WX) == [(1X)(WX)]S(1X)(WX)} :. N{[(1X)(WX)S(1X)(WX) == [(1x)(Mx)]S(1X)(Mx)} Similar counter-examples may be constructed also for existential generalization, using e.g., the 'H' from our example concerning two possible worlds and two objects a and b: 14 N[(1X)Hx]«1X)Hx = (1x)Hx) :. (Ey)N[(1X)Hx](y = (1x)(Hx)

Here the premiss is true, the conclusion is false. A counter-example similar to these two examples is in fact supplied by Smullyan himself on p. 37 of his article. Writing 'xC Ax)' for '9', 'x(Bx)' for 'the number of planets', and 'f for 'is less than 10', Smullyan gets: x(Ax) = x(Bx) N([x(Ax)] f x(Ax)) N([x(Bx)] f x(Bx))

(the scope symbols are put in by me for emphasis. Smullyan leaves them out on the convention that the scope of the abstract is understood to be the shortest formula containing the abstract unless otherwise indicated.) Smullyan remarks that here the conclusion cannot be deduced from the premisses. But he adds that "if in the second premiss and the conclusion the class abstract is given maximum scope, the conclusion would be valid". He then concludes from this example that "the modal paradoxes arise out of neglect of the circumstance that in modal contexts the scopes of incomplete symbols, such as abstracts or descriptions, affect the truth values of those contexts" (p. 37). Smullyan's defense against the other two counter-examples mentioned above would probably be similar: He would observe e.g., in the case of Wilson's example that if in the second premiss and the conclusion the descriptions are given the maximum scope, one would get the following valid argument: (1X)(WX) = (1x)(Mx) [(1X)(WX)] [(1X) (Wx)] N{S(1X)(WX) == S(1X)(WX)} :. [(1X)(WX)] [(1x)(Mx)] N{S(1X)(WX) == S(1x)(Mx)}

That this weaker conclusion is obtainable was, however, observed by Wilson himself. And Smullyan's defense is a poor one. For the claim of Smullyan's paper can then not be that "the unrestricted use of modal operators in connection with statements and matrices embedded in the framework of a logical system such as Principia Mathematica does not involve a violation of

REFERENTIAL OPACITY AND MODAL LOGIC

193

Leibniz's principle".15 His claim dwindles apparently to the far weaker: In A system of the type just described, any argument utilizing Leibniz's principle, however fallacious it may be, can be turned into a correct argument if we judiciously change the scopes in the premisses and the conclusion (or judiciously choose them, if not all scopes in the arguments have been selected beforehand). In Smullyan's article there is also another claim which seems untenable. Discussing class abstracts, Smullyan proposes as axioms of abstraction and extensionality respectively (4)

N(Ea)(
and (5)

N(x E a =x x E (3. :::) a

=

(3)

observing that if in place of (4) he had stipulated (6)

(Ea)N(
paradoxes would arise, among them that no true formal equivalences are contingent (Smullyan's article, pp. 36-37). Now (6) corresponds to the condition for descriptions we arrived at in the previous section, viz. (7)

'(E(3)N(a)(
Condition (4) is weaker; the corresponding condition for descriptions (8)

'N(E(3)(a)(

would be satisfied by expressions like '(1x)Hx', where 'H' is the 'H' of our example concerning the two objects a and b in the two possible worlds (footnote 3 above). In the preceding section we found that unless a description satisfies (7) it cannot be expected to behave like a singular term. And examples showing that (8) is too weak a condition are easily forthcoming. Thus in our example concerning our two possible worlds, the premisses of the following argument are both true, and the conclusion false: (1x)Hx = (1x)Fx N[(1X)Fx] F (1x)Fx :. N[(1X)Hx] F (1x)Hx If in the second premiss and the conclusion large scope had been used, the inference would have been valid. This incidentally shows that scope matters also when the condition (8) is satisfied and hence give an independent proof that there are descriptions which satisfy (8) but do not behave as singular terms. (A description behaves as a singular term only if no conventions

194

DAGFINN F0LLESDAL

regarding scope are needed.) One may feel that our model concerning two possible worlds is a rather artificial one, especially since in it 'H' is true of one and only one object in each possible world, but of different objects. Such qualms do not affect the validity of the above refutation of Smullyan, since our model with two possible worlds satisfies all the axioms and rules of even the strongest system of modal logic proposed, viz. S5, and still the premisses in our example come out true and the conclusion false. There seem, anyway, to be predicates which behave like 'H' also in less artificial models. The predicate 'is the number of planets" seems to be a case. For although this predicate is true of one number, 9, in our actual world and of other numbers in other possible worlds, it is presumably true of one and only one natural number in each possible world (0 being a natural number). The number of material objects of any specified type would apparently be equally good candidates for our 'H'. Miss Barcan, in her review of Smullyan's article, and Fitch, in "The problem of the Morning Star and the Evening Star", point out that if Smullyan replaces his condition (5) with (9)

'N(x E axx E (3)

::J.

a=

(3~

(6) may take the place of (4) without paradoxes. 16 Miss Barcan seems to prefer this alternative, since (5) easily permits her to derive that if two classes have the same members then they necessarily have the same members. 17 But Fitch apparently prefers the first alternative, stating on p. 140 of his paper that "in modal logic the condition for the interchangeability of scopes of '(lx)fx' is 'NE!(lx)fx)'''. Smullyan, Miss Barcan, and Fitch hoped to overcome the difficulties connected with substitutivity of identity in modal contexts by using contextually defined descriptions and manipulating the scope symbols. Another, apparently less complicated way out would be to abandon Leibniz's principle and rest contented with a principle of restricted substitutivity of identity, e.g., substitutivity in extensional contexts only. Carnap's treatment of what he calls "identity of extension" seems to be an attempt in this direction. But the reasons against restricting the substitutivity of identity are many. One is standard usage. A relation in a deductive system is usually called an identity relation if and only if it is reflexive and substitutive with respect to all contexts expressible in the system. 18 Another related reason, is our intuitions concerning identity, the feeling many have, that "if identity does not mean universal interchangeability, then I do not really understand identity at all".19 A third reason may be had from the discussion of identity of individuals in § § 12 and 13. We there found that unless we took the identity relation to be universally substitutive, we were apparently unable to derive

REFERENTIAL OPACITY AND MODAL LOGIC

(10)

(x)(y)(x

195

= y.:::J N(x = y))

let alone the stronger (10')

N(I)N(2) ... N(m)(x)(y)(x = y.:::J N{l)N(2) ... N(n) (x = y))

And in § 12 we found (10) and (10') to be true in any interpreted system of modal logic, so that according to Corollary 1.3, any system of quantified modal logic with identity in which (10) is not derivable is semantically incomplete. Also logical considerations concerning completeness therefore demand universal substitutivity of identity. It should be noticed that the reasoning leading up to Thesis 1 and its corollaries, among them Corollary 1.3, at no point presupposed anything about the substitutivity of identity. What concerned us was exclusively the interpretation of the quantifiers. Therefore, observing that e.g., the expressions 'the Morning Star' and 'the Evening Star' cannot be substituted for each other in all modal contexts, let us not jump to the conclusion that the identity relation is not universally substitutive. Let us rather cease to consider expressions like these as names. as indicated in the preceding section. Or, if we insist that they are names, let us then admit, with Frege, that in modal contexts we talk not about material objects, but about intensions: concepts, attributes, and propositions. CHAPTER VI. EXAMINATION OF THE DIFFICULTIES

19. Examination of the difficulties surveyed in § 10

With the results from the preceding discussion at hand, let us now examine the difficulties relating to quantification into modal contexts which were surveyed in § 10. We will identify the arguments by the numbers (1-9) which we used in § 10. I.

The statement (1)

(Ex)N(x > 7)

was found to be disturbing, because apparently 9, but not the number of planets, is one of the numbers whose existence is affirmed in (1). If, however, we cease to regard the expression 'the number of planets' as a name, then we may well accept the fact that (2)

N(9 > 7)

is true and nevertheless (3)

N (the number of planets> 7

false. For on neither of the three approaches in § 17 would

196

DAGFINN F0LLESDAL

(4)

9 = the number of planets

be an identity sentence. If we take approach 3), (4) would be well-formed and true, but the '=' would not be an identity sign. Ifwe choose approaches 1) or 2), (4) would not be well-formed; we should rather write (4')

(Ey) «x) (the planets are x in number

=. x = y). y = 9)

and (3) would correspondingly become (3')

N (Ey) «x) (the planets are x in number

=. x =

y). y > 7)

(Since mentioned in the preceding section, there is in each possible world exactly one natural number of which 'the planets are x in number' is true, the clause '(x) (the planets are x in number =. x = y)' in (4') and (3') could be shortened to 'the planets are y in number'.) 2. The observation that if substitutivity of identity breaks down in modal contexts, then we cannot quantify into them, has been found to be correct. In § 12 we saw that in any interpreted system of quantified modal logic with identity, (5)

N(l) ... N(m)(x)(y)(x = y.

~

N(l) ... N(n) (x = y))

is true. In § 17, we found that (5) would be satisfied if and only if we require that all of our singular terms keep their references in all possible worlds. But this requirement also removes the cause of the apparent breakdown of substitutivitity of identity, which was that two singular terms could be coreferential in one possible world but not in another. If we want to quantify into modal contexts we must hence require that substitutivity of identity holds in them. Something therefore has to be done with statements like (6)

N (the number of planets> 7)

which in § 7 led us to conclude tentatively that substitutivity of identity breaks down in modal contexts. Church's proposal, that in modal contexts 'the number of planets' does not refer to the object referred to by '9' is a way out. We leave our stock of singular terms untouched, and individuate the entities of our universe of discourse finely enough to secure that all our singular terms have one and only one reference which they may keep in all possible worlds, i.e., we quantify over intensions. Another way out is sketched in § 17; we quantify over extensions, but restrict our stock of definite singular terms, either by eliminating them completely (approach I)) or by keeping only those of them which refer to the same extensional object in all possible worlds (approaches 2) and 3)).

REFERENTIAL OPACITY AND MODAL LOGIC

197

3. The unusual character which the development of an intensional language adequate to general purposes would otherwise have to assume, is an argument in favor of restricting the stock of singular terms rather than extruding extensional objects from the range of our variables. The sentence (7)

The number of planets is a power of three

may be kept unchanged if we choose approach 3), while if we take approaches 1) or 2) it goes into e.g., (7')

The planets number a power of three

The sentence (8)

The wives of two of the directors are deaf

may remain unchanged on all three approaches. 4. Quine's argument relating to congruence shows that if 'Morning Star' and 'Evening Star' are among our singular terms in quantified modal logic with identity, our ontology is intensional. In order to avoid an intensional ontology one must cease to regard the two expressions as singular terms and rather treat them e.g., as 'Hesperus' and 'Phosphorus' were treated in § 17. Quine's 'C' is then no longer the relation of identity, but corresponds rather to a relation of coextensiveness between predicates true of one and only one object. 5.

Although (9)

(x) (y)(x = y.::l N(x = y))

is provable in Miss Barcan's modal calculi, the values of her variables do not therefore have to be intensions. As we saw in § 17, (9) does not require extensional objects to be extruded from one's universe of discourse provided one's stock of singular terms is appropriately restricted. It goes without saying that even with an appropriately restricted stock of singular terms, we are free to commit ourselves to an intensional ontology if we so wish. We may, e.g., affirm statements like '(x) (x is an intensional object)'. In this respect, modal logic is similar to extensional logic. The point of § 17 and the observation above concerning Miss Barcan's systems is that quantification into modal contexts and results like (9) above do not compel us to extrude extensional objects from our universe of discourse. 6.

It is right that in modal logic the universal closure of (10)

N(x = y'X + y'X + y'X =f- y'X. ::l. x

> 7)

198

DAGFINN F0LLESDAL

is true, while the universal closure of (11)

N (there are exactly x planets :l. x> 7)

is false. This shows, as was mentioned in § § 6 and 7, that coextensional open sentences cannot be interchanged in modal contexts salva voritate as they can in extensional contexts. It also indicates that the modal operator 'N' is rather like the expression 'is valid' or 'is logically true' except that the latter two attach to names of sentences, while 'N' attaches to sentences (§ 6). But that coextensional general terms be interchangeable in a context is not required in order that we shall be able to quantify into the context. If we want to quantify into modal contexts, we must, however, require that if we admit open sentences like (12)

N(x> 7)

(13)

N (if there is life on the Evening Star then there is life on x)

or

into our system of modal logic, fulfillment of them by the objects over which we quantify must make sense. If, as indicated in § 17, we expurgate from our stock of singular terms those expressions f1, which do not satisfy the requirement '(E,8)N(a)(af1,'s =. a = ,8), then 'the number of planets', 'the Evening Star', and 'the Morning Star" all disappear from our vocabulary of singular terms. On all the three approaches outlined in § 17, the open sentence (12) above may be kept unchanged and become true of e.g., the number 9, false of e.g., 5. The open sentence (13) may be kept unchanged when we use approach 3), but becomes unacceptable if we take approaches 1) and 2). Instead we might, e.g., write (13')

N (there is one and only one star which shines conspicuously bright in the evenings, and if there is life on this star then there is life on x).

7. If in order to overcome the difficulties confronting quantified modal logic one requires that any two conditions which uniquely determine an object in our universe of discourse be analytically equivalent, then, as Quine points out in From a Logical Point of View, (14)

(x)(y)(x = y.:l N(x = y))

will be true. Since, however, as we say in § 12, we have to accept (14) anyway, and since Quine in Word and Object derives more disastrous consequences from the above requirement, let us postpone the examination of it to part 9 of this section.

199

REFERENTIAL OPACITY AND MODAL LOGIC

8. The three results at which Quine arrives in his discussion of quantified modal logic in "Three grades of modal involvement" have all been arrived at again in this thesis. In § 12 we found that (14)

(x)(y)(x = y.:J N(x = y))

and even the stronger (15)

N(1) ... N(m) (x = y.:J N(1) ... N(n) (x = y))

is true in every interpreted system of quantified modal logic with identity. (14) and (15) were true, not only because standard usage requires an identity relation to be universally substitutive, but because interpretation of the quantifiers was found to be impossible unless (15) and (14) are true. Quine's remark in "Three grades of modal involvement" (p. 81) that interference in the contextual definition of singular terms even when their objects exist might be a way out of the difficulties, has been followed up in this thesis. The three approaches sketched at the end of § 17 are all based on manipulation with the singular terms. Thirdly, Aristotelian essentialism is unavoidable in quantified modal logic. But if the modal operator 'N' itself makes sense, then, as we say in part 6 of this section, open sentences with an 'N' prefixed make sense too if we restrict our stock of singular terms appropriately. To make sense of Aristotelian essentialism and to make sense of open sentences with an 'N' prefixed are one and the same problem, and a solution to the one is a solution to the other. Much of this thesis, especially § 17, is therefore indirectly a discussion of Aristotelian essentialism. 9. In Word and Object, Quine finds that if conditions which uniquely determine the same object in our universe of discourse are required to be analytically equivalent, then modal distinctions collapse. In § 16 modal distinctions were found to collapse if descriptions (or class abstracts) are treated as names and in addition some other rather general conditions are satisfied (Theorems XVI and XVII). We therefore had to introduce a rather severe restriction on the type of descriptions we could permit to be treated as names: ~(E,8)N(a)(ip =.a = ,8)~ had to be true in order that ~ (7a)) ip ~ could be treated as a name. If the above restrictions are adhered to, and if our modal system is based on Godel's basic system or a stronger system, then one may easily prove that any two conditions which are allowed to occur in descriptions which are treated as names are analytically equivalent if they uniquely determine the same object. For let 'Fx' and 'Gx' represent two such conditions. Then we have: *(1')

Fxo:=.x=y:Gxo:=.x=y

*(2')

Fx 0:= Gx

( 1')

200

DAGFINN F0LLESDAL

(3')

Fx ==. x = y : Gx ==. x = y :::). Fx == Gx

*(2')

(4')

N(Fx ==. x = y : Gx ==. x = y :::). Fx == Gx)

(3') RL

*(5')

(Ey)(N(x)(Fx ==. x = y). N(x)(Gx ==. x = y))

by hypothesis

*(6')

N(x)(Fx ==. x = y). N(x)(Gx ==. x = y)

(5') y

*(7')

(x)N(Fx ==. x = y). (x)N(Gx ==. x = y)

(6') by (2) of § 15

*(8')

(x)N(Fx ==.x = y: Gx ==.x = y)

(7') by (5) of § 16

*(9')

N(Fx ==. x = y : Gx ==. x = y)

(8')

N(Fx ==.x = y: Gx ==.x = y :::).Fx == Gx)::). N(Fx ==.x = y: Gx ==.x = y)::) N(Fx == Gx)

Godel's axiom A.2

*(11')

N(Fx == Gx)

(4') (9') (10')

*(12')

(x)N(Fx == Gx)

(II') x

*(10')

In fact, any such condition turns out to be analytically or necessarily true of the object it determines. And if Brouwer's axiom ~ tp ::) NMtp ~ is a theorem of the system, then one may also prove that any such condition is necessarily false of all other objects in our universe of discourse. 2o The requirement that two conditions 'Fx' and 'Gx' which uniquely determine the same object in our universe of discourse are analytically equivalent, is as we saw in § 17 (approach 3), all that is needed in order that the descriptions '(1X)Fx' and '(1X)GX) may be substituted for each other in modal contexts. But as observed in § 17, if the description' (1X) Fx' shall behave in all respects like a singular term, such that it e.g., may be used to instantiate universal quantifications, then 'Fx' has to satisfy the stronger requirement '(Ey)N(x)(Fx =. x = y)'. Open sentences which do not occur in descriptions which are substituted for each other or in other ways treated like proper names do, however, not have to be analytically equivalent even if they uniquely determine the same object. We may therefore conclude that if open sentences occur in descriptions which are substituted for each other, they have to be analytically equivalent. But open sentences which do not occur in such descriptions do not have to be analytically equivalent, even if they uniquely determine the same object. For as mentioned in part 6 of this section, neither quantification theory nor identity theory requires coextensional sentences to be interchangeable. What we may learn from the argument in Word and Object is therefore that in quantified modal logic one cannot use a standard theory of descriptions. If one does, modal distinctions collapse. Department of Philosophy University of Oslo

REFERENTIAL OPACITY AND MODAL LOGIC

201

NOTES This slip of Feys' is apparently due to too hasty reading of Miss Barcan's "The identity of individuals in a strict functional calculus of second order" (1947), to which he refers. As we saw in § 13, Miss Barcan proved the two kinds of identity to be equivalent already in S2, and her proof holds also in S l. 2 As mentioned already in footnote 1 of § 13 a theory of identity will be considered standard only if, in particular, the identity relation is universally substitutive. A theory of descriptions will be said to be standard only if, in particular, any well-formed sentence with 'x' as the only free variable is rermitted to take the place of 'Fx' in '(7x)Fx'. Lewis and Langford, Symbolic Logic, theorem 19.81. 4 This observation, which simplifies and strengthens my original result, is due to Professor Quine. This attitude towards proper names is not unlike that of Wilson, e.g., in The Concept of Language. 6 "Ueber Sinn und Bedeutung", quoted from Black's translation in Philosophical Writings of Gottlob Frege, p. 58. 7 Methods of Logic, p. 218. P. 224 of the reprint of "Knowledge by acquaintance" in Mysticism and Logic. Here and in what follows we will for brevity state the requirements only for uniterated modalities. The general requirements for iterated modalities may easily be got from these by replacing the 'N' in this and the following formulae by 'N(1)'" N(n)' and by prefixing 'N(I)'" N(m)' to the expression (E,6) wherever it occurs. to Hilbert and Bernays, Grundlagen der Mathematik, I, pp. 383 If. II Cf. e.g., Carnap, Meaning and Necessity, pp. 33-34. 12 Miss Barcan in her review of "Modality and description" and Fitch in his article "The problem of the Morning Star and the Evening Star", pp. 138, footnote 4. 13 Wilson, The Concept of Language, p. 34. 14 The example concerning the two possible worlds was as follows: possible, non -actual world

Fa. -Ga. -Ha. Ka

-Fb. Gb. Hb. -Kb

actual world

Fa. -Ga. Ha. -Ka

-Fb. Gb. -Hb. -Kb

Smullyan "Modality and description", p. 34. Miss Barcan's review of "Modality and description", p. 150, Fitch, "The problem of the Morning Star and the Evening Star", p. 138, footnote 4. 17 Miss Barcan's review of "Modality and description", p. 43. 18 See e.g., A.A. Fraenkel "The relation of equality in deductive systems" and Quine's review of this article in The Journal of Symbolic Logic, 14,1949, p. 130. 19 Wilson, The Concept of Language, p. 39. 20 For let 'Fx' represent such a condition. Then one has: IS

16

(I) (2) (3) (4)

* (5)

*(6) *(7) *(8) *(9) *(10)

*(11) *(12)

y=y N(y=y) Fy =. y = y ::::l: y = y. :::l Fy N(Fy =. Y = Y ::::l: y = y. :::l Fy) (Ey)N(x)(Fx =. x = y) N(x)(Fx =.x = y) (x)N(Fx =. x = y) N(Fy =.y = y) N(Fy =.y = y ::::l: y = y.:::l Fy) :::l. N(Fy =. Y = y) :::l N(y = y.:::l Fy) N(y = y.:::l Fy) :::l. N(y = y) :::l NFy NFy (Ex)NFx

(l)RL (3)RL by hypothesis (5) y (6) by (2) of § 15 (7) by Godel's axion A.2 by GOdel's axiom A.2 (II) (II)

And if Brouwer's axiom'

202 (I) (2) *(3) *(4) *(5) *(6) **(7) **(8) **(9) **(10) **(11 )

*(12) '(13)

'(14)

DAGFINN F0LLESDAL Fx =. x = y ::J: x # y.:J -Fx N(Fx =.x = y ::J: x # y.:J -Fx) (Ey)N(x)(Fx =.x = y) N(x)(Fx =. x = y) (x)N(Fx =. x = y) N(Fx =.x = y) x#y x # y.:J N(x # y) N(Fx =. x = y ::J: x # y.:J -Fx) :J. N(Fx =. x = y) :J N(x # y.:J -Fx) N(x # y.:J -Fx) :J. N(x # y) :J N - Fx N -Fx x # y.:J N - Fx (x)(x # y.:J N - Fx) (Ey)(x)(x # y.:J N - Fx)

(2) RL by hypothesis (3) y (4) by (2) of § 15 (5) Theorem XV (§ 15) by Godel's axion A.2 by Godel's axiom A.2 (2) (6) (7) (8) (9) (10)

'(II)

(12) x (13)

STEN LINDSTROM

AN EXPOSITION AND DEVELOPMENT OF KANGER'S EARLY SEMANTICS FOR MODAL LOGIC*

I. INTRODUCTION

Stig Kanger - born of Swedish parents in China in 1924 - was professor of Theoretical Philosophy at Uppsa1a University from 1968 until his death in 1988. He received his Ph.D. from Stockholm University in 1957 under the supervision of Anders Wedberg. Kanger's dissertation, Provability in Logic, was remarkably short, only 47 pages, but also very rich in new ideas and results. By combining Gentzen-style techniques with a model theory ala Tarski, Kanger obtained new and simplified proofs of central metalogical results of classical predicate logic: G6del's completeness theorem, L6wenheim-Skolem's theorem and Gentzen's Hauptsatz. The part that had the greatest impact, however, was the 15 pages devoted to modal logic. There Kanger developed a new semantic interpretation for quantified modal logic which had a close family resemblance to semantic theories that were developed around the same time by laakko Hintikka, Richard Montague and Saul Kripke (independently of each other and independently of Kanger). Although his work did not receive much attention at the time, it is generally accepted nowadays that Kanger played a crucial role in the development of model-theoretic semantics for modal logic. The precise nature of his contribution is, however, not so widely appreciated. It is sometimes said that Kanger was the true originator of the so-called possible worlds or Kripke semantics for modal logic. Thus, Dagfinn F011esdal, in his address to the Stig Kanger Memorial Symposium in Uppsala August 1991, proposed the label "KangerKripke-semantics" for the possible worlds approach.! *Earlier versions of this paper have been presented at the UgLLI-workshop on Philosophical Logic at Uppsala University, March 18, 1995, at the Swedish national conference in philosophy, "Filosofidagarna", Umea, June 7-9, 1995, and discussed with friends and colleagues. A shorter version was also published as Lindstrom (1996). I am especially indebted to Joseph Almog, Lennart Aqvist, Thorild Dahlquist, Bengt Hansson, Risto Hilpinen, Jaakko Hintikka, Paul Needham, Peter Pagin, Wlodek Rabinowicz, Krister Segerberg and Rysiek Sliwinski for their very helpful comments and advice. A fellowship at the Swedish Collegium for Advanced Studies in the Social Sciences (SCASSS) during the Autumn of 1996 provided an excellent research environment for finishing the paper.

203 P. W. Humphreys et al. (eds.), The New Theory of Reference © Kluwer Academic Publishers 1998

204

STEN LINDSTROM

Such an assimilation of Kanger-type semantics to standard possible worlds semantics could, however, be misleading. Kanger's formal semantics for modal logic does not, at least explicitly, utilize the notion of a possible world. Nor is there in his early works from 1957 any mention or discussion of possibilia (possible worlds, counterfactual states of affairs, possible individuals). I do not think this is an accident. Kanger's goal was to generalize and extend the standard Tarski-style definition of truth in a model for first-order predicate languages to languages of quantified modal logic. He wanted to do so without introducing any new primitives in the metalanguage. In (1957b), he explicitly mentions it as an advantage of his approach that it does not presuppose any "intensional" entities like Fregean senses, meanings or intensions. Although he does not discuss the matter, one gets the decided impression that Kanger's ontology is no more hospitable toward possibilia than it is toward intensional entities. There are, I shall argue, important differences, both of a conceptual and a technical nature, between Kanger's approach to modal semantics and the possible worlds approach. In this connection, one should distinguish between: (i) possible worlds semantics proper: a particular type of model-theoretic semantics for modal logic; and (ii) the possible worlds interpretation of modal concepts. Accordingly, we may pose the two questions: "Is Kanger's semantics a kind of possible worlds semantics?" and "Is Kanger's interpretation of modal notions a 'possible worlds interpretation'?". In this paper I shall argue that both questions should be answered in the negative. Kanger semantics differs from standard possible worlds semantics in many ways. As we shall see, the notions of logical truth and logical consequence are defined differently for the two kinds of semantics. Moreover, the underlying intuitions about modality are different. 2. SEMANTICS FOR QUANTIFIED MODAL LOGIC: FROM CARNAP TO KRIPKE

The proof theoretic study of quantified modal logic was pioneered by Ruth Barcan Marcus (1946a, 1946b, 1947) and Rudolf Carnap (1946, 1947) who were the first to formulate axiomatic systems that combined quantification theory with (S5- and S4-type) modal logic. The attempts to interpret quantified modal logic by means of formal semantic methods also began with Carnap (1946, 1947), where he presented a semantics for logical necessity based on Leibniz's old idea that a proposition is necessarily true if and only if it is true in all possible worlds. In his formal semantics, Carnap used syntactic entities state-descriptions - as representatives of possible worlds. Suppose that we are considering a first-order predicate language C with a countably infinite set of individual constants, an infinite sequence of n-ary predicate letters for each n ~ 1, but without function symbols or the identity symbol. In addition to Boolean connectives and quantifiers, the language C contains the modal

KANGER'S EARLY SEMANTICS FOR MODAL LOGIC

205

operator 0 for logical necessity. A state-description for C is simply a set of (closed) atomic sentences of £,.2 Carnap (1947), p. 9, writes "... the statedescriptions represent Leibniz's possible worlds or Wittgenstein's possible states of affairs". Relative to a state-description S, the notion of truth for sentences ¢ of C is defined as follows (we write S F ¢ for ¢ being true in S): (1)

SFP(al, ... ,an)iffP(al, ... ,an)ES.

(2)

SF' iffS

(3)

SF(->1j;)iffS~orSF1j;

(4)

SF ' iff, for every individual constant c, SF (cjx).

~

.

Here, ¢(c/x) is the result of substituting c for every free occurrence ofx in ¢. Finally, Carnap gave the following truth clause for the operator 0: (5)

S F 0 iff, for every state-description S', s' F .

That is, the modal formula "it is (logically) necessary that ¢" is true in a statedescription S if and only if ¢ is true in every state-description S'. Carnap defines logical truth as truth in all state-descriptions (we write F ¢ for ¢ being logically true). Hence, (6)

S F O iff F .

(7)

S F .O iff

(8)

F O, if F ; and F .O, otherwise.

~

.

It is easy to verify that 0 satisfies the usual laws of the system S5, together with

the so-called Barcan formula and its converse, and the rule of necessitation, (K)

F O( -> 1j;) -> O( -> 1j;).

(T)

F O -> .

(S5)

F ,O -> O'O.

(Ba)

F '(x)

->

O'(x).

(CBa) F O'(x) -> '(x). (Nee)

If F , then F D.

Let QC(C) be the set of all logically true sentences, according to the above semantics, in the language C. Theorem. The set QC(C) is not recursively enumerable, so there is no formal system with this set as its theorems. Proof Suppose the set QC(C) is recursively enumerable. Let ¢ be any sentence of the non-modal fragment Co of C. Then, we have:

206

STEN LINDSTROM

(i)

Either

(ii)

If

(iii)

If

1= Dc/> or 1= ,Dc/>.

1= Dc/>, then 1= c/>. 1= ,Dc/>, then ~ c/>.

Since, by assumption, there is an effective enumeration of the logically true sentences of C, we can effectively decide which of F D¢ or F ...,D¢ holds. But, this means in virtue of (ii) and (iii) that we can effectively decide which of F ¢ or ~ ¢ that holds. By Godel's completeness theorem for the predicate calculus, for any sentence ¢ of Co, F ¢ holds if and only if ¢ is a theorem of the firstorder predicate calculus. But this is contrary to Church's theorem according to which the pure predicate calculus is undecidable (cf. Kleene 1967, § 45). Q.E.D.

The next step in the development of a viable semantics for quantified modal logic was taken by Stig Kanger in his dissertation Provability in Logic (1957a). Kanger's ambition was to provide a language of quantified modal logic with a model-theoretic semantics ala Tarski. For that purpose Kanger introduced the notions of a primary valuation and a system. A primary valuation for a language C of quantified modal logic is a function v which for every non-empty domain D assigns an appropriate extension in D to every individual constant, individual variable, and predicate constant in C. A system is an ordered pair S =< D, v>, where D is a (non-empty) domain and v is a primary valuation. Notice that v does not only assign extensions to symbols relative to the designated domain D, but relative to all domains simultaneously. In Kanger's dissertation there is to be found, for the first time in print, a semantics for modal operators in terms of so-called accessibility relations: each modal operator D is associated with an accessibility relation Ro between systems in terms of which the semantic evaluation clause for D is spelled out: S 1= Dc/> iff for every system S' such that SRoS', S'

1= c/>.

That is, D ¢ is true in the system S if and only if ¢ is true in every system S' that is Ro-accessible from S. One particular modal operator that Kanger introduces is one he calls logical necessity and which he provides with the following semantic clause: S 1= [Jc/> iff for every system S', S'

1= c/>.

Thus, R!!J is the universal relation between systems. Kanger points out that by imposing certain formal requirements on the accessibility relation, like reflexivity, symmetry, transivitity, etc., one can make the operator satisfy corresponding well-known axioms of modal logic. In this way, the introduction of accessibility relations made it possible to apply semantic and model-theoretic methods to the study of a variety of modal

KANGER'S EARLY SEMANTICS FOR MODAL LOGIC

207

notions other than logical necessity. Although Kanger was the first to publish, other researchers, among them Hintikka and Montague, also came up with the idea of utilizing accessibility relations in the semantics of modal notions. One source of inspiration for Kanger's use of accessibility relations in modal logic was no doubt the work of Jonsson and Tarski (1951) on representation theorems for Boolean algebras with operators. 3 Jonsson and Tarski define operators 0 on arbitrary subsets X of a set U in terms of binary relations R ~ U x U in the following way: oX = {x

E

U : 3y

E

X(yRx)},

that is oX is the image of X under R. They also point to correspondence between properties of 0 and properties of R. Among other things, they prove a representation theorem for so-called closure algebras that, via the TarskiLindenbaum construction, yields the completeness theorem for propositional S4 with respect to Kripke models with a reflexive and transitive accessibility relation. However, Jonsson and Tarski do not say anything about the relevance of their work to modal logic. Perhaps they considered the connection too obvious or of too little importance to mention it. A semantic approach to first order modal predicate logic that has a close resemblance to Kanger's was developed by Montague (1960).4 Like Kanger, Montague starts out from the standard model-theoretic semantics for nonmodal first-order languages and extends it to languages with modal operators. He defines an interpretation for an ordinary first-order predicate language £ to be a triple S =< D, I, g >, where (i) D is an empty set (the domain); (ii) I is a function that assigns appropriate denotations in D to the non-logical constants (predicate symbols and individual constants) of £;5 and (iii) a function g that assigns values in D to the individual variables of £. For each non-logical constant or variable X, let SeX) be the semantic value (i.e., denotation for nonlogical constants and value for variables) of X in the interpretation S. Then the notion of truth relative S is defined as follows: (1)

S F P(t\, ... ,tn ) iff < S(t\), ... ,S(tn )

(2)

S F (t\

(3)

SF'¢> iff S ~ ¢>.

(4)

SF (¢>

(5)

S F Vx¢> iff for every object a E D, S(a/x) F¢>'

=

->

t2) iff S(t\)

'l/!) iffS

~

=

>E

S(P).

S(t2)'

¢> or SF 'l/!.

Here, Sea/x) is the interpretation which is exactly like S, except for assigning the object a to the variable x as its value. Montague now asks the same question as Kanger: How can this definition of the truth-relation F be generalized to first-order languages with modal operators? As we recall, Kanger solved the problem by modifying the notion

208

STEN LINDSTROM

of an interpretation: a Kanger-type interpretation (what is called 'a system') assigns denotations to the non-logical constants and values to the variables not only for one single domain (the 'actual' one) but for all domains. Montague's approach is simpler than Kanger's: he keeps the notion of an interpretation S of first-order logic intact, and just adds semantic evaluation clauses for the modal operators. As in the Kanger semantics, each modal operator D is associated with an accessibility relation RD. Now, however accessibility relations are relations between interpretations S =< D, I, g > of the underlying non-modal first-order language. The semantic clause corresponding to the operator D, with associated accessibility relation R o , is: (6)

SF D¢ iff for every interpretation S' such that SRoS', S' F ¢.

Montague observes that for each variable x the universal quantifier Vx, that binds the variable x, can be viewed as a modal operator with the accessibility relation R\;ix defined by the condition:

< D, I,g > Rvx < D', I',g' > iffD every variable y different from x.

=

D', I

=

I' and g(y)

=

g'(y) for

Given this definition, (5) becomes a special case of (6). Montague associates with the operator [jJ of logical necessity the accessibility relation R[jJ defined by:

< D,I,g >R!D< D',I',g' > iff D = D' and g = g'. Thus, his semantic clause for (l)

[jJ

becomes:

< D, I, g > F I!J¢ iff for every I' defined over D, < D, I', g > F ¢.

This semantic clause should be compared with Kanger's stricter condition: (2)

S FI!J¢ iff for every system S', s'

F ¢.

The difference between (1) and (2) corresponds to a difference between two different conceptions of logical truth: Tarski's (1936) conception and the modern model-theoretic one. According to Tarski (1936), truth and logical truth are properties that primarily apply to interpreted formal languages. An interpreted first-order language L comes with a domain of discourse D and an interpretation function I that gives the denotations in D of the non-logical constants of L. Hence, the (absolute) notion of truth is well-defined for such a language. A formula ¢ of L is true relative to an assignment g of values in L to the variables of L if < D, I, g >F ¢. ¢ is true, simpliciter, if it is true relative to every assignment, that is if it is true in the intended model < D, I >.6 Now, according to Tarski (1936), a sentence (closed formula) ¢ of L is logically true if it is true and its truth is invariant with respect to all reinterpretations of its nonlogical constants relative to the given domain of discourse D. That is, ¢ is

209

KANGER'S EARLY SEMANTICS FOR MODAL LOGIC

logically true if and only if, for every interpretation < D, I', g' > with the given domain D, < D, I', g' >F ¢. If ¢ is a sentence (closed formula), then its truthvalue is independent of the assignment of values to the variables. Hence, Tarski's (1936) definition yields for sentences ¢ of £: ¢ is logically true iff for every I' defined over D, < D, I' ,g > p ¢,

where D is the domain of discourse of £ and g is any assignment of values in D to the variables. Hence, for closed formulas of the underlying first-order language, Montague's truth-clause (1) above coincides with Tarski's (1936) definition of logical truth. It also accords well with Carnap's definition of logical truth as truth in all state descriptions, provided that the language contains names for all the objects in a fixed domain of discourse. Tarski's (1936) definition yields the result that all true sentences of an interpreted language that do not contain any non-logical symbols are logically true. So, if the domain of discourse is infinite (and provided that = is taken as a logical constant), then all the sentences of the form: (n)

:JXl ... :JXn (Xl =1= X2 1\ ... 1\ Xl =1= Xn 1\ X2 =1= X3 1\ ...

. . . 1\ X2 =1= Xn 1\ ... 1\ Xn-l =1= Xn)

n

~

2, are logically true. So according to Montague (1960), the sentences:

([]n)

[]

:JXl ... :JXn(XI =1= X2 1\ . .. 1\ Xl =1= Xn 1\ X2 =1= X3 1\ ...

.. . 1\ X2 =1= Xn 1\ .. . 1\ Xn-l =1= Xn)

are true for an interpreted language with an infinite domain. Notice, however, that according to Montague (1960), it is not the case that:

P []:JXl ... :JXn(XI =1= X2 1\ . .. 1\ Xl

-1=

Xn 1\ X2 =1= X3 1\ ... 1\ Xn-l =1= xn),

that is the sentences ([iJn) are not logically true. This is so, since the definition of logical truth that Montague adopts in the metalanguage is the standard modeltheoretic one of truth in all interpretations, not just truth in all interpretations with a given domain. In conclusion, Montague's and Kanger's respective truth clauses for [iJ yield, for modality-free sentences ¢, []¢ is true iff ¢ is logically true in the sense of Tarski (1936) []¢ is true iff ¢ is logically true in the model-theoretic sense.

(Montague) (Kanger)

There are other possibilities for interpreting logical necessity within Montague's framework, one interesting alternative being: (3)

< D, I, g >p []¢ iff for every model < D', I' > such that ~ D', < D',I',g >p ¢.

D

Notice that on Montague's interpretation (1) of [iJ, both the Barcanformula:

210 (Ba)

STEN LINDSTROM

'v'xi!J¢>ex) ->l!J'v'x¢>ex)

and its converse: (CBa) i!J'v'x¢>ex) -> 'v'xi!J¢>ex)

are logically true. However, on the interpretation (3), (CBa) still holds, but the Barcan formula fails. Kanger's interpretation (2), finally, makes both (Ba) and (CBa) fail. As a matter of fact, on the interpretation (2) the following is a logical truth: i!J¢>(XI' ... ,xu)

+--+

I!J'v'XI ... 'v'Xn¢>(XI, ... , xn).

Hence, if we write, for example,

the initial quantifier does not really "bind" the free variables

Xl

in the formula

In other words, Kanger's interpretation does not allow us to "quantify in" past the modal operator w. The interpretations (1) and (3), on the other hand, give coherent sense to "quantifying in". For example, given interpretation (3), the following sentence is clearly meaningful. As a matter of fact, it is easily seen to be false in any domain with two elements: 3x3y(x oF y /\ 'v'Z i!J(z = x V Z = y)) -> 3x3y(x oF y/\ i!J 'v'z(z = x V Z = y)).

The semantical frameworks of Carnap (1947), Kanger (1957a, 1957b) and Montague (1960) - although important developments in themselves, were still not what we nowadays call possible worlds semantics. Something was still missing: an ingredient that was provided by laakko Hintikka and Saul Kripke. 7 In order to see what was missing, let us just make a seemingly small change in Carnap's semantics. Instead of considering the set of all state-descriptions, we shall only consider those state-descriptions that belong to some arbitrary non-empty set C. Let us call such a set C of state-descriptions a Carnap frame. The notion of truth in a state-description now becomes relativized to a Carnap frame. Thus for any SEC, (1)

S FC P(al,.··, an) iffP(al, ... , an) E S.

(2)

S FC .¢> iff S ~ C¢>.

(3)

S FC (¢> -> 'Ij;) iffS

(4)

S FC 'v'x¢> iff, for every individual constant c, S FC ¢>(c/x).

(5)

S FC

~c

¢> or S FC 'Ij;.

D ¢> iff, for every state-descripition Sf in C, Sf FC ¢>.

KANGER'S EARLY SEMANTICS FOR MODAL LOGIC

211

Observe how the relativization of the definition to C changes the meaning of D. Instead of D¢ meaning that ¢ is logically true, it now means that ¢ is universally true relative to the given set of state descriptions. The metatheoretic

definition of logical truth must be changed accordingly: ¢ is logically true 0= ¢) if and only if, for every C and every SEC, S Fc¢. It is not difficult to axiomatize the quantified modal logic that corresponds to this semantics. As a matter of fact, Hintikka's and Kripke's innovation is essential for proving completeness results. The set C represents the space of possible words: not every state-description may correspond to a genuine possibility. The last step to possible worlds semantics, taken by Kripke, is to let the possible worlds be explicitly represented in the semantics by an index set W. Hence, by replacing the nonempty set of state-descriptions by an indexed family C = {Sw}weW of statedescriptions and utilizing accessibility relations between indices (possible worlds), we get something rather similar to possible worlds semantics. Sentences are now evaluated at indices rather than at state-descriptions: FC P(al, ... , an) iff(al, ... , an) E Sw.

(1)

W

(2)

W

(3)

W

Fc ...,
(4)

W

FC Vx¢ iff, for every individual constant c,

(5)

W

FC D¢ iff, for every u E W such that wRu, u FC ¢.

W

Fc~. W

FC
Finally, we get a full-fledged possible worlds semantics if we replace the use of state-descriptions by an indexed family of domains {Ow} WEW together with an interpretation function I that assigns, for each possible world w E W, an appropriate semantic value to the non-logical symbols of L. In Carnap's, Kanger's and Montague's early theories, the space of possibilities is represented by one comprehensive collection containing all state descriptions, systems or models, respectively. Hence, every state description, system, or model is thought of as representing a genuine possibility. Hintikka, Kripke and modern possible worlds semantics are instead working with semantic interpretations in which the space of possibilities is represented by an arbitrary non-empty set K of model sets (in the case of Hintikka) or possible worlds (Kripke). Following Hintikka's (1980, 1989) terminology, one may say that the early theories of Carnap, Kanger, and Montague were considering standard interpretations only, where one quantifies over what is, in some formal sense, all the possibilities. In the possible worlds approach, one also considers non-standard interpretations, where arbitrary non-empty sets of possibilities are considered. 8 The consideration of interpretations (model structures) that are non-standard in this sense - in combination with the use of accessibility relations between worlds in each interpretation - made it possible for Kripke (1959, 1963b and 1965) to prove completeness theorems for various systems of

212

STEN LINDSTROM

propositional and quantified modal logic (T, B, S4, etc.). We end this section by presenting a version of Kripke's (1963a) semantics for modal predicate logic with identity, where the notion of a possible world is an explicit ingredient of the semantic theory. A (Kripke) model structure for a language C of first-order modal predicate logic is a quintuple S =< W, D, R, E, Wo > where, (i) W is a non-empty set (of possible worlds); (ii) D is a non-empty set (of possible objects); (iii) R ~ W x W is the accessibility relation of the structure; (iv) E is a function which assigns to each w E Wa subset Ew of D (intuitively Ew is the set of objects that exist in w); and (v) Wo is a designated element of W (the actual world). It is required that D = UwEWEw, i.e., that every possible object exists in at least one world. A Kripkemodelis an ordered pair 9Jl =< S, I >, where S =< W, D, R, E, Wo > is a model structure and I is a function satisfying: (vi) for each w E W, I assigns to each n-ary predicate constant P of C a subset I(w, P) of Dn; and (vii) I assigns to each individual constant c of C an element I(c) ED. A model9Jl of the form < S, I > is said to be based on the model structure S. Observe that I(w, P) is not necessarily a subset of (Ew i.e., the extension of P in w may involve objects that do not exist in w. Moreover, individual constants are treated as rigid designators, i.e., they are assigned denotations in a world-independent way. An assignment in 9Jl is a function g which assigns to each variable x an element g(x) in D. For any term t in C, we define 9Jl (t, g) to be get) if t is a variable; and I(t) if t is an individual constant. With these notions in place, we can define what it means for a formula ¢ to be true in a world w with respect to the model9Jl and the assignment g (in 9Jl) (in symbols, w FlIJl ¢[g]):

t,

(1)

w pm P(t" ... , tn)[g) iff dUl(t" g), ... , 9Jl(tn, g) >E I(w, P).

(2)

wpm (t, = tn)[g) iff9Jl (t"g) = 9Jl(t2,g).

(3)

w pm -,¢ [g) iffw lim ¢ [g).

(4)

wpm (¢

(5)

wpm Vx¢ [g) iff, for every a

(6)

wpm D¢ [g) iff, for every u E W such that wRu, u pm ¢ [g).

--->

'IjJ) [g) iffw lim ¢ [g) or wpm 'IjJ [g). E

Ew, wpm ¢ [g(a/x»).

We say that ¢ is true with respect to the model 9Jl and the assignment g (in symbols FlIJl ¢[g]), iff ¢ is true at the actual world Wo with respect to 9Jl and g. ¢ is true in the model9Jl (in symbols, FlIJl ¢), if for every assignment g, FlIJl ¢ [g]. ¢ is true in a model structure S (FS ¢) if it is true in every model based on S. Let K be a class of model structures. We say that ¢ is K-valid if ¢ is true in every S E K. We are especially interested in model structures where R is the universal relation in W, i.e., in which:

KANGER'S EARLY SEMANTICS FOR MODAL LOGIC

(6)

w Fall

o¢ iff, for every U

E

213

W, U F9.lI ¢.

Let QS5= be the class of such structures. As Kripke (l963a) showed, neither the Barcan formula nor its converse is (QS5=)-valid. It should also be pointed out the Kripke's semantics validates the Law of Identity, (L =) Vx(x = x),

as well as the principle of Indiscernibility of Identicals, (I =)

VxVy[x = y --> (¢(xjz)

-->

¢(yjz»],

applicable without restrictions also to modal contexts ¢(z). From these principles, together with the rule of Necessitation it is easy to infer: (0 =) VxVy(x = y --> O(x = y»

(Necessity of Identity)

(0 ""') VxVy(x "'" y --> O(x "'" y».

(Necessity of Distinctness)

The latter principles are controversial. Hintikka has repeatedly stressed that they are unacceptable for propositional attitude constructions like "John knows that ...", George the IV believes that ...", etc. 9 On the other hand, Kripke's formal semantics fits well with his conception of metaphysical necessity - as it is expounded in Kripke (1980) - and the possible worlds metaphysics that goes with it. In section 6.1, I shall argue for the thesis that (QS5=) is the (first-order) quantified modal logic corresponding to Kripke's conception of metaphysical necessity. And I shall write IMl rather than D when the necessity operator is interpreted as metaphysical necessity. The dual operator, metaphysical possibility, is written as ~ that is,
According to the dominant intuition underlying Kanger's work, modal operators of the object language, "It is logically necessary that", "It is settheoretically necessary that", etc., are obtained "from above" as it were, by "reflecting" corresponding predicates of the metalanguage: "truth in all interpretations", "truth in every interpretation that is normal with respect to the set-theoretic vocabulary", etc. According to this intuition, it is (analytically) necessary that all bachelors are unmarried, just because the sentence "All bachelors are unmarried" is an analytic truth. Writing llil for "it is analytically necessary that ¢", where ¢ is a sentence of the object language, we have: (l)

IHl ¢ is true iff ¢ is an analytic truth.

By reiativizing this schema to a semantic interpretation S, we get:

214 (2)

STEN LINDSTROM

INI

The basic idea can be expressed in the form of a reflection principle: (R)

SpO
where 0 is an operator of the object language and P is a metalinguistic predicate that may be satisfied by a formula ¢ and an interpretation S.1O By itself, the schema (R) only gives "truth-conditions" for formulas containing o. But if we think of (R) as explaining the meaning of the operator 0 in terms of the metalinguistic predicate P, we have what may be called a metalinguistic interpretation of O. Kanger's interpretation of modal operators is usually metalinguistic in this sense. Consider, for example, his introduction of the operator I1J of logical necessity. The underlying intuition is expressed by the schema: (3)

ID

In virtue of the standard model-theoretic definition of "logical truth" as "truth in every interpretation of the object language", we get: (4)

ID

(4) corresponds to the following clause of Kanger semantics: 1D

Analytic truth is explicated by Kanger in the following way: (5)

valuation v iff

This corresponds to the following semantic clause for the operator 1liI: (6)

1NI

The possible worlds approach, by contrast, is based on the intuition that there is a multitude of ways in which the world could have been different. On this conception, necessity is not primarily a metalinguistic notion: the necessary truths are rather the truths that would have remained true had any of these possibilities been realized. It is necessarily true that 2 + 2 = 4 just because, for any of the ways ("possible worlds") in which the world could have been different, had the world been different in that way, the proposition that 2 + 2 = 4 would still have been true. These two intuitions about the source of necessity are clearly distinct and indeed, as we shall see, give rise to very different formal semantic theories. To further clarify the relationship between Kanger's semantics for modal

KANGER'S EARLY SEMANTICS FOR MODAL LOGIC

215

logic and possible worlds semantics, we may make use of a terminology due to Etchemendy (1990) and distinguish between interpretational and representational semantics. Intuitively, the truth-value of a sentence is determined by two factors: (i) the interpretation of the symbols in £; and (ii) the Jacts (the way the world is). We may keep the world fixed but let the interpretation of the symbols vary. This is the basic idea behind the standard definitions of logical truth and logical consequence in the model-theoretic tradition after Tarski (1936). Etchemendy calls it interpretational semantics. On this approach, set-theoretic models represent ways of reinterpreting the language. Or, we may keep the interpretation of the linguistic units fixed, but consider what would have happened to the truth-value of the sentence had the facts been different. A semantics based on the latter idea Etchemendy calls representational. On this approach set-theoretic models represent ways in which the world could have been different (possible worlds). The classical example of representational semantics is, of course, Carnap's semantics in terms of state descriptions. We can use either of the two ideas, the representational or the interpretational, to interpret modal operators. For example, (i) and (ii) below are alternative ways of interpreting a necessity operator: (i)

tiJ is true (in the actual domain relative to the intended interpretation) iff cp is true in every domain relative to every interpretation of the (non-logical) symbols of cp.

(ii)

1M! cp is true (in the actual world relative to the intended interpretation) iff cp is true and would have remained true (relative to the intended interpretation of the language) even if the world had been different in any of the ways in which it (metaphysically) could have been different.

The definition of [D is an example of the interpretational approach: it corresponds to Kanger's explication of logical necessity. The definition of IMI. corresponding to Kripke's notion of metaphysical necessity, is an example of the representational approach. Both ideas playa role in Kanger's semantics, but the first one is clearly the one that dominates his thinking. The second idea turns up, in a relatively mild form, in his semantics for analytic necessity: it is analytically true that (in the actual domain and on the intended interpretation of object language) iff is true in every domain. In order for this definition to make intuitive sense, we must allow domains that contain not only actually existing objects but also merely possible objects, i.e., objects that do not exist but might have existedor at least representatives of such object. Otherwise, all true universal generalizations of the form 'v'x(Fx - t Gx) would on the proposed definition be analytically true. We could not distinguish between the analytically true sentence "All bachelors are unmarried" and the presumably synthetic truth "All bachelors are less than 150 years old". Kanger expresses the idea behind his definition of analyticity: "Our definition of analyticity may be regarded as

216

STEN LINDSTROM

an explication ... of the idea that an analytic proposition is a proposition that is true in every possible universe." II Actual truth is defined with respect to the domain that Kanger describes as "the class of all 'real' individuals".12 We see that alternative domains of individuals play somewhat the same role in Kanger semantics as possible worlds in standard possible worlds semantics. Kripke's notion of metaphysical necessity - truth in every metaphysically possible world - is, however, foreign to Kanger's way of thinking. Modal operators are viewed as projections of metalinguistic concepts rather than as expressing genuine properties of propositions. 4. KANGER'S SEMANTICS FOR FIRST-ORDER QUANTIFIED MODAL LOGIC

We shall now take a closer look at Kanger's semantics for quantified modal logic. We consider a first-order predicate language C with identity and a family of unary modal operators {Di : i E I}. A domain is a non-empty set D. A primary valuation (for C) is a function v, which, given any domain D, assigns: (i) an element of D to each term t of C; (ii) a set of ordered n-tuples of elements in D to each n-place predicate constant P in C. We may, whenever convenient, think of a valuation as consisting of two components: (i) an interpretation I that assigns denotations I(D, a) to the nonlogical constants of C relative to domains D; and (ii) an assignment g that assigns, for each domain D and each variable x, a value g (D, x) to x in D. Of course, v = lug. For any I and D, we define ID to be the unary function defined by letting Io(a) = I(D, a). Vo and go are defined analogously. A system is an ordered pair S =< D, v >, where D is a domain and v is a valuation. When we want to emphasize the distinction between the interpretation I and assignment g, we also write a system as: S =< D, I, g >. Kanger defines the notion of a formula ¢ being true in a system S =< D, v> (in symbols, S F ¢) in the following way:'3 (1)

S

(2)

S ~ P(t\, ... , t n ) iff < v(D, t\), ... , v(D, tn ) > E v(D, P).

(3)

S ~.L

(4)

S~(r,b-41/J)iffS~r,borS~1/J.

(5)

< D, I, g >~ Vxr,b iff < D, I, g >p r,b, for each g' such that g'

(6)

< D, I, g >~ 3xr,b iff < D, I, g' >p r,b, for some g' such that g'

(7)

for every operator 0, S ~ Or,b iff \;IS', if SRo S', then S' ~ r,b.

~

(t\ = t2) iffv(D, tJ) = v(D, t2).

=x g. =x g.

Explanation: g' is like g except possibly at x (also written, g' =x g) if and only if, for each domain D and each variable yother than x, g'(D, y) = g(D, y). In the

KANGER'S EARLY SEMANTICS FOR MODAL LOGIC

217

above definition, Ro is a binary relation between systems that is associated with the modal operator O. Ro is the accessibility relation associated with the operator O. Among the modal operators in £, there are two designated ones llil ("analytic necessity") and [j] ("logical necessity") with the semantic clauses: < D, l,g >p Iffiq., iff for every domain D',< D',I,g >p q.,. < D, I, g >p [Jq., iff for every system S, S p q.,. We define a Kanger model for £ to be an ordered pair Wl =< D, I> of a domain and an interpretation. Notice that for any D and any I, the structure < D,lo > is a model for the non-modal fragment £0 of £. Hence, an interpretation I can also be represented as a family {< D, 10 > }OED of firstorder models indexed by the collection V of all domains. Notice also that this family contains exactly one model for each domain D. A Kanger model can then be written as an ordered pair of the form Wl =< D, { < D, 10 > }OED >. A system, finally, can be represented as an ordered pair < Wl, g > of a Kanger model Wl and an assignment g. We say that formula ¢ is true in a Kanger model Wl (in symbols, Wl po ¢), if it is true in every system S of the form < Wl, g >. If g and g' are assignments that agree for each domain on all the variables that occur free in ¢, then < Wl,g >po ¢ iff < Wl,g' >po ¢. Hence, a sentence (closed formula) is true in a Kanger model Wl = {< D, 10 > bED iff it is true in some system < Wl, g >. A formula ¢ is said to be valid (logically true) if it is true in every system < D, v >. A formula ¢ is a logical consequence of a set r offormulas (in symbols, r po ¢) if ¢ is true in every system in which all the formulas in r are true. Kanger defines a proposition as an ordered pair < ¢, I > of a sentence (closed formula) ¢ and an interpretation 1. 14 An n-ary predicate is defined as an ordered pair < P, I >, where P is an n-ary predicate constant and I is an interpretation. A name is a pair < c, I >, of a constant c and an interpretation I. The proposition < ¢, I> is said to be true in the domain D if < D, I >po ¢. < ¢, I > is called analytic if it is true in every domain D. Notice that for any domain D, (I)

is true in D iff < q." I> is analytic.

(2)

< [Jq." I > is true in D iff q., is logically true.

The notion of (non-relativized) truth is explained as followsY Let Do be the set of all "real" individuals. Let 10 be the interpretation which, for every domain D, assigns to the non-logical symbols the denotations that they have according to the intended interpretation of £.16 A proposition < ¢, I > is true if it is true in the domain Do. A sentence ¢ is true if the proposition < ¢,Io > is true. ¢ is said to be analytic if < ¢,Io > is analytic. Notice that llil ¢ is true iff ¢ is analytic. Similarly, [j]¢ is true iff ¢ is logically true.

218

STEN LINDSTROM

Let us say that a modal operator 0 is ontological if there exists a relation R such that for every domain D, every valuation v and every formula ¢:17

< D, v >p DqJ iff for every D', ifDRD', then < D', v >p qJ. In (1970) Kanger presents a version of his semantics where he considers ontological operators only. That is, every operator 0 is associated with a relation Ro between domains rather than between systems. In (1972) the ontological approach has become his official semantics for modal logic. 5. THE TREATMENT OF "QUANTIFYING IN" IN KANGER SEMANTICS

In Provability in Logic (1957a) and in "The morning star paradox" (1957b) Kanger gives the following evaluation clause for the universal quantifier: (i)

< D, I, g >p VXqJ iff < D, I, g' >p qJ for every assignment g such that g =x g.

According to this clause, the formula Vx¢ is true in the system < D, I, g > if and only if ¢ is true in every system < D, I, g' > which is exactly like < D, I, g > except, possibly, for the values it assigns to the variable x in the various domains. In (1957c) and (l957d) he considers the following alternative clause: (ii)

< D, I, g >p VXqJ iff < D, I, g' >p qJ for every g' such that (i) g' =x g; and (ii) g'(D',x) = g(D',x) for all domains D' that are distinct from D.

That is, the formula Vx¢ is true in the system < D, I, g > if and only if ¢ is true in every system < D, I, g' > which is exactly like < D, I, g > except possibly, for the value it assigns to the variable x in the domain D. Kanger (1957c) suggests the following informal readings of these two alternatives: (x)qJ is true iff qJ is true for every interpretation ofx

(Ux)¢ is true iff qJ is true whatever x may denote, writing (x) and (Ux) for the quantifiers with the first and the second evaluation clause, respectively. If ¢ is a non-modal formula, then, of course, (x)qJ

<-->

(Ux)qJ

is valid. However, (Ux) is really an odd quantifier, which does not even validate: (Ux) ~Fx

~

(Uy)

~Fy.

Consider, namely, a system S =< D, I, g > such that Fx is true in some domain D' different from D, and Fy is false in every domain. Suppose also that the extension of F in D is the empty set. Then, ~ Fx and --, ~ Fy are both true in S. By the semantic clause for (Ux), (Ux) ~ Fx will also be true in S. But, --, ~ Fy is

219

KANGER'S EARLY SEMANTICS FOR MODAL LOGIC

true for every assignment g satisfying the conditions (i) g' =y g; and (ii) for all domains D' that are distinct from D, g'(D', y) = g(D', y). Hence, (Uy) ~ Fy is false in S. That is, substitutivity of alphabetic variants fails for the operator (Ux). For this reason, I shall not consider it further. So, when I speak of Kanger's interpretation of the quantifiers, I shall understand the treatment in (l957a) and (1957b), according to which the universal quantifier \:I has the semantic clause (i) and the existential quantifier has the dual clause: (iii)

< D, I, g >F= ¢ iff < D, I, It >F= ¢ for some g' such that g'

=x

g.

In order to get a clearer understanding of Kanger's treatment of quantification, I shall speak of selection functions that pick out from each domain an element of that domain as individual concepts. To be more precise, an individual concept, in this sense, is a a function f, with the collection of all domains as its range, such that for every domain D, f(D) E D. We can think of a system S =< D, I, g > as assigning to each individual constant c the individual concept {< D, I(D, c) >: D is a domain} and to each variable x the individual concept {< D, g(D, x) >: D is a domain}. The formula P(t\, ... , t n) is true in S =< D, I, g > if and only if the individual concepts designated by t\, ... , tn pick out objects in the domain D that stand in the relation I(D, P) to each other. The identity symbol designates the relation of coincidence between individual concepts (at the "actual" domain D). That is t\ = t2 is true in a system S =< D, I, g > if and only if the individual concepts designated by t\ and t2, respectively, pick out one and the same object in the domain D of S. Kanger's quantifier \:Ix, with the semantic clause (i), can now be thought of as an objectual quantifier that ranges not over the "individuals" in the "actual" domain D, but over the (constant) domain of all individual concepts. That is, \:Ix¢ is true in a system < D, I, g > if and only if ¢ is true in every system which is exactly like < D, I, g > except, possibly, for the individual concept that it assigns to the variable x. Note, that interpreted in this way, the range of the quantifiers \:Ix and :3x is independent not only of the domain D but also of the system S: the range of the quantifiers \:Ix and :3x is fixed, once and for all, to be the collection of absolutely all individual concepts. While formulas of the form t\ = t2 express coincidence, identity between individual concepts is expressed by formulas of the form l1il(t\ = t2)' Writing x == y for l1il(x = y), the following principles are valid: (L ==)

\fx(x == x)

(I ==)

\fx\fy(x == y

--+

(¢(x/z)

--+

¢(y/z))),

that is, == satisfies the formal laws of the identity relation. None of the sentences:

220

STEN LINDSTROM

saying that the "actual" domain has at least n elements (for n 2: 2), is logically true according to Kanger's semantics. In contrast all sentences of the form: ::3xl··· ::3xn(·(Xl == X2) /\ ... /\ '(X2 == X3) /\ ... /\ .(X2 == xn) /\ ... /\ '(Xn-l == xn)),

are logically true. Intuitively, these sentences say that there are, for each n, at least n individual concepts. Consider now Kanger's (l957b) discussion of the so-called Morning Star paradox. The paradox arises from the following premises: (1)

(2) (3)

~ (Hesperus = Hesperus) Phosphorus = Hesperus • ~ (Phosphorus = Hesperus),

where, "Phosphorus" and "Hesperus" are two proper names (individual constants) and Iillis to be read "it is analytically necessary that". We assume that "Phosphorus" is used by the language community as a name for a certain bright heavenly object visible in the morning and that "Hesperus" is used for some bright heavenly object visible in the evening. Unbeknown to the community, however, these objects are one and the same, namely, the planet Venus. "Hesperus = Hesperus" being an instance of the Law of Identity is clearly an analytic truth. It follows that the premise (1) is true. (2) is true, as a matter of fact. "Phosphorus = Hesperus" is obviously not an analytic truth, "Phosphorus" and "Hesperus" being two different names with quite distinct uses. So, (3) is true. However, using sentential logic together with the following laws of predicate logic: (UI)

Vxq;

(I =)

VxVy(x = y ---+ (q;(x/z)

---+

q;(t/x),

(Universal Instantiation) ---+

q;(y/z))),

(Indiscernibility of Identicals)

we can infer from (2) and (3): (4)

•

~

(Hesperus = Hesperus).

But (1) and (4) contradict each other, so something must have gone wrong in this argument. According to Kanger's diagnosis of the Morning Star paradox it is (I =) that is at fault. Given Kanger's semantic clause (i) for the universal quantifier and his semantic treatment of individual terms, (UI) is valid but (I =) is not. Only the following restricted version of (I =) is valid: (I =')

VxVy(x = y ---+ (q;(x/z) ---+ q;(y/z))), provided that no free occurrence of z in q; is within the scope of a modal operator.

With this change in the underlying logic, (4) can no longer be inferred from (2) and (3).

KANGER'S EARLY SEMANTICS FOR MODAL LOGIC

221

As Kanger himself points out, there are still some difficulties left. From (UI) and the equivalence 3x¢ ...... -,\lx-,¢, we get: (EG)

¢(t/x)

--->

3x¢.

(Existential Generalization)

But, in view of (1 )-(3) and the Law ofIdentity, the following sentences are true: (5)

Phosphorus = Hesperus 1\

(6)

Hesperus = Hesperus 1\ IHI (Hesperus = Hesperus).

-,

IHI (Phosphorus = Hesperus).

So, it follows that: (7)

3x(x = Hesperus 1\

(8)

3x(x =Hesperus 1\ IHI (x = Hesperus».

-,

IHI (x = Hesperus».

Although unintuitive, this result is perfectly compatible with the interpretation of the quantifiers as ranging over individual concepts and of the identity symbol as designating coincidence between individual concepts. According to this interpretation, (7) and (8) mean: (7')

There is an individual concept x which actually coincides with the individual concept Hesperus but does not do so by analytical necessity.

(8')

There is an individual concept x which not only happens to coincide with the individual concept Hesperus but does so by analytic necessity.

As Quine (1947) was the first to point out, however, (7) and (8) are incompatible with interpreting \Ix and 3x as objectual quantifiers meaning "for all objects x (in the domain D)" and "for at least one object x (in D)" and letting the identity sign stand for genuine identity between objects (in D). Because, under this interpretation, (7) and (8) have the readings: (7")

There is an object x (in the actual domain D) which is identical with Hesperus and which is not necessarily identical with Hesperus.

(8")

There is an object x (in the actual domain D) which is identical with Hesperus and which is necessarily identical with Hesperus.

meaning that one and the same object, Hesperus, both is and is not necessarily identical with Hesperus, which is absurd. So Kanger's semantics for quantified modal logic is incompatible with interpreting the quantifiers as ranging over actually existing individuals (as opposed to individual concepts) and at the same time interpreting = as identity between individuals.

222

STEN LINDSTROM

In Kanger's semantics there are no means of identifying individuals from one domain to another. In particular, the truth-values of formulas will not be affected if we make all the domains disjoint, by systematically replacing every domain D by the set: {< D, a >: a ED}.

In other words, set-theoretic relations between domains like inclusion, overlap and disjointness, have no semantic significance. Suppose we make the claim: (9)

Something is such that it is the number of planets but might not have been so.

It seems reasonable to formalize this claim in quantified modal logic as: (10)

3x(Px /\ -.DPx).

We cannot use any of the Kanger's quantifiers for this purpose, however. Suppose, namely, that: g(D,x) E I(D,P),DRoD',D f:. D',g(D',x)

~

I(D',P).

Intuitively this means that one thing is the number of planets in the domain D and one thing or another is not the number of planets in the modal alternative D' to D. From this, we should not be able to conclude (10). But on any of Kanger's interpretations of the universal quantifier, (10) follows. So his approach does not allow us to express the claim that one and the same object has a given property in one domain and lacks that property in another domain. Now, we might ask how we could repair Kanger's semantics in order to allow for genuine quantification over individuals. There are many possibilities. One that is particularly straightforward technically is to adapt Kripke's (1963a) treatment of quantification to Kanger's approach. This means that we modify the notion of an assignment g in such a way that an individual variable x is assigned an object g(x) in a domain-independent way. That is, we make two changes with respect to Kanger's notion of an assignment: (i) the value g(D, x) of an individual variable x in a domain D is no longer required to be a member ofD; (ii) for all domains D and D', we require that g(D,x) = g(D',x). After these changes are made, an assignment simply becomes a function g that assigns to each variable x an object g(x). We then adopt the following evaluation clauses for the universal and existential quantifiers:

< D, I, g > PVxrj; iff < D, I, g' > p rj;, for every g such that (i) g is like g except possibly at x; and (ii) g'(x) ED. < D, I, g >p 3xrj; iff < D, I, g' >p rj;, for some g such that (i) g' is like g except possibly at x; and (ii) g'(x) ED.

223

KANGER'S EARLY SEMANTICS FOR MODAL LOGIC

With these clauses our previous objection to Kanger's approach that it could not make correct sense of sentences like 3x(Px 1\ 'DPx) seems to have been met. For this sentence to be true at < D, I, g >, there needs to be an element a E D such that: a

E

lCD, P) and for some D', DRoD' and a

~

I(D', Pl.

We have been able to express the claim that one and the same individual has the property P in domain D and lacks that property in some domain D' that is possible relative to D. Let us now see how the modified semantics might handle the Morning Star paradox. In this semantics, = is interpreted as genuine identity between objects. Accordingly the logical principles for = are the expected ones: (L =) Vx(x = x).

(I =)

(Law of Identity)

VxVy(x = y --> (¢(xjz)

-->

¢(yjz)).

(Indiscernibility of Identicals)

However, instead of (UI) we have: (UI')

Vx¢ 1\ :Jx(x = y)

-->

¢(y Ix).

Now, how should we handle individual constants within the modified Kanger semantics? An intuitively appealing approach is to assign denotations to constants in a domain-dependent way as before, but not require the denotation I(D,c) of a constant c relative to a domain D to be a member of D. With this treatment of individual constants, we cannot infer from (I =) to: (11)

Phosphorus = Hesperus --> (Iffi (Phosphorus = Phosphorus) Iffi (Phosphorus = Hesperus)),

-->

unless the following requirements are met: 3x !ill (x

=

Phosphorus),

3x!ill (x

=

Hesperus).

But these conditions hold, only if: 3x !ill (x

=

Phosphorus),

!ill (Phosphorus = Hesperus).

The last of these conditions contradicts (3), so it cannot be assumed. It would, presumably, hold only if "Phosphorus" and "Hesperus" were synonymous. Hence, we cannot infer (11) from (I =). We can also verify, directly, that the modified semantics does not allow the inference from (2) and (3) to (4). So the Morning Star paradox, in the form that Kanger presented it, is resolved. Let us say that a modal operator 0 is a constant assignment operator, if there exists a binary relation Ro between Kanger models < D, I >, such that for every assignment g,

224

STEN LINDSTROM

< D, I, g > F D4> if and only if for every Kanger model < D', I' > such that < D, I > Ro < D', I' >, < D', I', g > F 4>. Intuitively, constant assignment operators do not affect the values that are assigned to free variables within their scopes. In contrast, Kanger's operator of logical necessity, which is not a constant assignment operator, binds all free variables within its scope. Hence, quantifying in, past this quantifier, does not make sense. As Kaplan (1986) has shown, however, we can interpret logical necessity in a way that admits of quantifying in. Within the present framework, Kaplan's treatment of logical necessity amounts to the following semantic clause: < D, I, g > F [jJ 4> if and only if for every Kanger model < D', I' >, < D', 1', g >F 4>. Although, we do have for this operator, Barcan Marcus' (1947) Necessity of Identity principle: (D =) VxVy(x = y ->

[jJ

(x = y)),

we, of course, do not have: (12)

Phosphorus

=

Hesperus

->[jJ

(Phosphorus

=

Hesperus)

and this, we think, is as it should be. It is, of course, very doubtful whether Kanger would have approved of these changes to his semantics. 6. THE SET-DOMAIN VERSUS THE CLASS-DOMAIN SEMANTICS

Kanger presents his semantic theory within an informal set-theoretic framework, where all sets are treated on a par, as genuine objects that can be members of other sets. 18 This theory - what we might call naive Kanger semantics - is, however, threatened by paradoxes: given normal assumptions about sets the theory is inconsistent. For example, primary valuations are themselves non-empty sets, that is, domains. It follows that for any valuation v, v itself belongs to the first argument domain of v. In consequence, for an individual constant c, < < v, c >, v(v, c) >E v. This is contrary to the ordinary assumption of set theory that sets are well-founded. This consequence may not seem especially serious from Kanger's point of view, since in section 5.6 of [1957a] he briefly discusses the introduction of non-well-founded sets in set theory. However, there are more serious problems. Consider any valuation v. Its first argument domain is the collection U of all domains. But in order for v to be a set, the collection U must also be a set. Let r+(U) be the set of all non-empty subsets of U. Then, by Cantor's theorem, card (r-t (U)) > card (U). But this

KANGER'S EARLY SEMANTICS FOR MODAL LOGIC

225

contradicts the fact that p+(U) <:;:; u. Thus, given normal set-theoretic assumptions the collections U and v cannot be sets. But this is in apparent contradiction to Kanger's free use of the valuations v in various set-theoretic constructions, for instance, when he defines the notions of a system, a proposition, truth in a system, logical truth, etc., or when he introduces accessibility relations between systems. We are going to consider two alternative ways of modifying Kanger's approach in such a way that the threat of paradox is avoided: the set-domain approach and the class-domain approach. 6.1. Kanger set-domain semantics

A natural way of modifying naive Kanger semantics is to assume, as before, that the domains are arbitrary non-empty sets but that the valuation functions are proper classes. The elements of a domain are, of course, either individuals (urelements) or sets. A (primary) valuation v is now defined in the same way as before, except that it is taken to be a functional (proper) class rather than a functional set. The same holds for the notions of an interpretation I and an assignment g. The models < D, In> for the non-modal base language £0 are sets. An interpretation is a proper class {< D, In > }nEv of such models (relational structures) indexed by the class 'D of all non-empty sets. A Kanger (set-domain) model consists of a domain D and an interpretation 1= {< D, In > }DEV. Such a model (we use the notation (D, I) for it) cannot be a set nor, in fact, a class, since it is intuitively an ordered pair, one of the components of which is a proper class. Although, we cannot define it in the usual systems of set-theory (Zermelo-Fraenkel, G6del-Bernays-von Neumann, Kelley-Morse), it makes intuitive sense to speak of the ordered pair (D, I). An assignment g is now a functional class which for any domain D and any variable x assigns a value g(D, x) in D to x. A system is an ordered pair S = (Wl, g) consisting of a Kanger model Wl and an assignment g. The accessibility relation Ro that is associated with a modal operator 0 of £ is a collection of ordered pairs (S, S') of systems. Or, in the case of ontological operators: a collection of ordered pairs of domains. The following notions are defined exactly as before: (i) S F ¢ and Wl F ¢; (ii) a formula ¢ being valid or logically true (written as F ¢); (iii) logical consequence, r F ¢ where r is a set of formulas and ¢ a formula. The Kanger set-domain semantics presupposes a strong metatheory in which one can speak not only of the usual cumulative hierarchy of sets over a set of individuals, but also of classes that may contain sets and individuals, collections that may contain such classes, collections that may contain collections, and so on. Only the individuals and sets are regarded as genuine objects, while classes and collections are thought of as essentially predicative in nature. As a proper formal language for formalizing the metatheory, we think of a language of

226

STEN LINDSTROM

simple-type theory, where the individual variables (i.e., variables of type L) range over the elements of the cumulative hierarchy and the predicate variables of type (a\, ... ,an) range over arbitrary n-ary relations whose i'th domain for I :::; i :::; n consists of the class of all entities of type ai. So for instance, the variables of type (L) range over arbitrary classes of objects (urelements and sets). Variables of type ((t), (L)) range over arbitrary relations between classes of objects, etc. A metalanguage of this kind should be appropriate also to formalize the next version of Kanger semantics. 6.2. Kanger class-domain semantics

This semantics differ from the previous one in the following respects: A domain is now defined to be a non-empty class of elements of the cumulative set hierarchy, i.e., we also allow domains to be proper classes. A valuation is a function v, which given any domain 0, assigns appropriate denotations over 0 to the non-logical constants (propositional constants, individual constants and predicate constants) and individual variables of C. We now allow the denotation of an n-place predicate constant P of C to be a class of n-tuples of elements in O. The notions of an interpretation and an assignment are adjusted accordingly. Hence, we allow models (0,1 0 ) for the base language Co, where o is a class and 10 assigns appropriate classes to the non-logical predicate constants of C. Thus, Kanger models now have the form 9Jl = (0, {(O, Io)}o~:v)' 6.3. Comparing the two approaches

We might now ask how validity with respect to the class-domain semantics is related to validity with respect to the set-domain semantics. If the language C has sufficient expressive capacity, neither implies the other. Let, for example, C' be the language which is obtained from our language C by adding the generalized quantifier (::labs infX) as a new logical constant. 19 The intuitive reading of (::labsinfX)rj> is for absolutely infinitely many x, c/>(x),

which means that the class of all objects a that satisfy rj>(x) is a proper class. The conception of the absolutely infinite is due to Cantor. Intuitively, a class A is absolutely infinite if it does not have exactly Ii elements for any cardinal number Ii (compare, A being infinite if it does not have exactly n members for any natural number n). Consider now the sentence (::labs infX)(X = x). This sentence is true in a domain 0 iff the domain is a proper class. According to the class-domain semantics, there are domains that are proper classes. So, according to this semantics the following sentence:

KANGER'S EARLY SEMANTICS FOR MODAL LOGIC

(1)

~ (::JabsinfX)(X

227

= x)

is true in e\'.ery domain (and for every valuation v). Hence, it is logically true according to the Kanger class-domain semantics. On the other hand, according to the Kanger set-domain semantics the sentence (:Jabs infX) (x = x) is false in every domain. Hence, according to this semantics, it is instead (2)

-, ~ (::Jabs infX)(X

= x)

that is a logical truth. Here we have a dramatic difference between the two semantic theories. Clearly, it is the class-domain semantics that yields the intuitively correct result in this case. Instead of adding the "artificial" quantifier (:Jabs infX) to 1:, we could instead have assumed that I: contained a modal operator D with the semantic clause: (0, v)

F 0

Then, we would have for the dual operator 0, (0, v)

F o¢ iff ::JD'

0' is a proper class and (D', v)

F
The class-domain semantics would then pronounce: (3)

o\l'x(x = x)

logically true, but according to the set-domain semantics it would instead by its negation: (4)

D::Jx(x

I' x)

that is logically true. Once again, the set-domain semantics gets the wrong result by arbitrarily excluding interpretations that are intuitively legitimate. 7. LOGICAL VERSUS METAPHYSICAL NECESSITY

7.1. On the adequacy oj Kripke's logic (QS5=) as the logic oJmetaphysical necessity

According to the metaphysical picture of modal reality inspired by Kripke's Naming and Necessity (1980), there is a space W of possible worlds in which the actual world Wo is just one of the worlds. There is also a collection D of possible objects and there are properties that the possible objects can have and relationships that they can have to each other (we call these properties and relationships attributes). For each possible world w, n-place attribute A, and possible objects al, ... ,an, it is determinatcly either true or false that A(al' ... ,an) holds in w. For each world w there is also the collection Ew of all the objects that exist in that world. Presumably, every possible object exists in at least one possible

228

STEN LINDSTROM

world. Also, certain objects like pure sets exist in all possible worlds. A reasonable existence principle for sets is that a set exists in a world w if and only if all its elements exist in that world. Given that sets are objects and that the pure sets exist in all possible worlds, it follows that the collection Ew of all individuals that exist in a world w is always a proper class rather than a set. Consider now the first-order modal language C, with identity, and with lMJas its only modal operator. In terms of the metaphysical picture described above, we can describe the intended interpretation of C and define the notion: "truth in a world w relative to an assignment g", for formulas of C. The intended interpretation consists of the following ingredients: (i) the class W of all metaphysically possible worlds; (ii) the class D of all possible objects; (iii) for each w E W, a class Ew ~ D of objects existing in the world w; (iv) the n-ary predicate symbols of C designate n-ary attributes and the individual constants of C designate possible objects; (v) the attributes have extensions relative to possible worlds (the extension of an n-ary attribute relative to a world w is a (possibly proper) class of n-tuples of possible objects). A sentence ¢ of the interpreted language C is true if it is true at the actual world (w. r. t. the intended interpretation) relative to every assignment. Now, it may very well be that the collections of possible worlds, possible individuals, and individuals existing in particular worlds do not form sets. In that case, the intended interpretation of C does not correspond to a Kripke model structure. We cannot, then, directly conclude from a sentence ¢ being true to it being true in some model structure. As long as we are considering the language of first-order modal logic, however, we can still make this inference in an indirect way.20 By the completeness theorem for Kripke's system of quantified S5, there exists a formal system QS5= which is (strongly) complete with respect to the set of all (QS5=)-structures. 21 Hence, for any sentence ¢ of C: ¢ is true => ¢ is (QS5=)-consistent (by the intuitive soundness of QS5= with respect to the intended interpretation) => ¢ is (QS5=)satisfiable (by the completeness theorem for QS5=).

It follows that ¢ is (QS5=)-valid => .¢ is not (QS5=)-satisfiable => .¢ is not true => ¢ is true.

Now, consider the language C with the intended interpretation given above. We have a notion of truth for C, but what could it mean for a sentence of C to be logically true. The notion of the intended interpretation for C suggests a notion of an interpretation for C. An interpretation for C is just like a Kripke model 9Jl =< S, I > based on a model structure S =< W, D, R, E, Wo > except that the collections W, D, R, Ew for w E W, are not required to be sets but are allowed to be proper classes. (We are of course only considering interpretations

KANGER'S EARLY SEMANTICS FOR MODAL LOGIC

229

where R = D x D). Hence, the intended interpretation becomes one of the interpretations. Let us call a sentence 1> of C supervalid if it is true in every interpretation of C and valid if it is true in every (QS5=)-model (that is, if it is (QS5=)-valid)?2 We now make the following conjecture: The logical truths of the language [. of metaphysical necessity are precisely those sentences of [. that are supervalid.

Adapting an argument due to Kreisel (1969), we can prove that supervalidity coincides with validity for the language C. The argument goes as follows: Since Kripke (QS5=)-models are interpretations, we have: (I)

if rp is supervalid, then

rp is valid.

The completeness theorem for the system (QS5=) yields: (2)

if rp is valid, then rp is (QS5=)-provable.

However, the system (QS5=) is intuitively sound with respect to supervalidity. That is the axioms are easily seen to be supervalid and the only rule of inference, modus ponens, preserves supervalidity. Hence: (3)

if rp is (QS5=)-provable, then

rp is supervalid.

(2) together with (3) yield: (4)

if rp is valid, then

rp is supervalid.

Hence, the notions of validity and supervalidity are co extensional for the language C. From this together with the conjecture, we conclude that (QS5=) is the first-order logic of metaphysical necessity.23 7.2. Logical necessity

An (interpreted) sentence 1> is metaphysically necessary if it is true in every possible world. It is logically necessary if it is true for every domain and every interpretation of its non-logical symbols. Given a certain conception of modal reality, I have argued that Kripke's (1963a) semantics for quantified S5 adequately captures the logic of metaphysical necessity. This means that the logic of metaphysical necessity is relatively meager. Although there are, on the Kripkean metaphysical picture, a wealth of metaphysically necessary truths, only a few of them are also logically necessary. For example, if the axioms of Zermelo-Fraenkel set theory are true, they are presumably true in all possible worlds, and hence metaphysically necessary. But they are not truths of logic, not even of the logic of metaphysical necessity.

230

STEN LINDSTROM

Consider now the sentences saying that there are at least n(n ?: 1) individuals: (n)

i- X2 /\ ... /\ Xl i- Xn /\ X2 i- X3 /\ ... i- Xn /\ ... /\ Xn-l i- Xn).

:JXl .•• :JXn (Xl . .. /\ X2

Each of these sentences is presumably metaphysically necessary. So for each positive n, the following is a truth of metaphysics: (1M] n)

i- X2 /\ ... /\ XI i- Xn /\ X2 i- X3 /\ ... i- Xn /\ ... /\ Xn-I i- xn).

1M] :JXl ... :JXn(Xl . .. /\ X2

It is, of course, not a logical truth. We do not have for any n ?: 1,24

F (QS5=) 1M] :JXl •.. :JXn(Xl i- x2 /\ ... /\ XI i- xn /\ x2 i- x3 /\ ... /\ X2

i- xn /\ ... /\ xn-I i- xn).

Nor do we have for any n,

F (QS5=) /\ X2

~:JXI ... :Jxn (XI i- x2 /\ ... /\ Xl i- Xn /\ X2 i- X3 /\ ... i- Xn /\ ... /\ Xn-I i- Xn).

In sharp contrast to this, Kanger's semantics for logical necessity validates every instance of

F ¢:JXI ... :JXn(XI i- X2/\···/\ XI i- Xn /\ X2 i- X3/\··· /\ X2

i- Xn /\ ... /\ Xn-l i- xn).

This is as I think it should be. It is a logical truth that it is logically possible that there are at least n objects. When comparing Kanger's semantics for modal logic with Kripke's we come to the conclusion that the former (at least in its class-domain version) is adequate for the notion oflogical necessity, while the latter adequately captures a form of metaphysical necessity. Neither semantics can handle adequately the notion that is captured by the other. To devise a semantics that can treat both notions is a challenge that still remains to be met. As we have seen, Kanger's model-theoretic semantics for quantified modal logic differs in many respects from modern possible worlds semantics. However, it raises sufficiently many questions both of a technical and of a philosophical kind to motivate an interest that is not merely historical. Department of Philosophy Umea University

KANGER'S EARLY SEMANTICS FOR MODAL LOGIC

231

NOTES At the 9th International Congress of Logic, Methodology and Philosophy of Science. See F Iilllesdal (1994). 2 Actually Carnap's state descriptions are sets of literals (i.e., either atomic sentences or negated atomic sentences) that contain for each atomic sentence either it or its negation. However, for our purposes we may identify a state description with the set of atomic sentences that it contains. Also, in order to make things simple, I am not discussing here Carnap's treatment of identity statements. 3 On p. 39 in (1957a) Kanger makes an explicit reference to Jonsson and Tarski (1951). 4 Montague (1960) writes: "The present paper was delivered before the Annual Spring Conference in Philosophy at the University of California, Los Angeles, in May, 1955. It contains no results of any great technical interest; I therefore did not initially plan to publish it. But some closely analogous, though not identical, ideas have recently been announced by Kanger [(I 957b)], [(l957c)] and by Kripke in [(1959)]. In view of this fact, together with the possibility of stimulating further research, it now seems not wholly inappropriate to publish my early contribution." 5 We are not going to consider languages that contain function symbols. 6 Here we ignore the possibility of D not being a set but a proper class and I not assigning sets but proper classes as extensions to the predicate symbols of C. If this were the case, then the intended interpretation of C would not be a model in the formal sense of model theory. Of course, there are interpreted first-order languages whose intended interpretations are not models in the formal sense, the first-order language of set theory, with the proper class V of all sets as its domain and the proper class {< x, y >: x is member of y} as the interpretation of 'E'. This opens up the possibility for a sentence of an interpreted formal language of being true although it may be false in all models in the sense of model theory. This possibility is precluded for first-order languages, by the LowenheimSkolem theorem: the truth of implies its consistency (by the intuitive soundness of first-order logic), which in turn, by the Lowenheim-Skolem theorem, implies having a (countable) model. But for formal languages that are able to express such notions as 'there is a proper class of x's such that ' the implication: (I)

if is true, then is true in some model (which is built up from sets).

fails. Let, for example, be the sentence "there is a proper class of x's such that x = x". This example is due to McGee (1992). 7 Cf. Kripke (1959, 1963) and Hintikka (1957a, 1957b, 1961). 8 For the standard-non-standard distinction, see also Cocchiarella (1975). 9 See, for example, Hintikka (1969). 10 We leave it open, for the time being, exactly what is meant by a (semantic) interpretation. II Kanger (1970), p. 49. 12 Kanger (1957b), p. 4. Cf. also Kanger (1970), p. 50. 13 Kanger uses the notation T(D, v, f= [g] and he speaks of the operation Twhich, for every domain D, every primary valuation v and every sentence , assigns one of the truth-values 0 or I to as the secondary valuation for C. 14 Actually he uses valuations here, but since the assignment of values to the variables is immaterial in this context it is more natural to work with interpretations. 15 Cf. Kanger (l957b), p. 4. 16 Kanger (1957b) uses the terminology "standard usage of C" instead of "intended interpretation ofe". 17 Here our terminology differs slightly from Kanger's. Our ontological operators correspond to what Kanger calls purely ontological operators. Cf. Kanger (1957a) p. 34. 18 In his definitions Kanger speaks of "classes' rather than "sets", but this terminological difference is inessential, since he does not make any distinctions within the category of all classes but rather treats all classes that he speaks of as genuine objects that can be members of other classes. 19 I have taken the quantifier (3 a bsinfX) from McGee (1992), where he uses it to show that there are interpreted formal languages for which the equivalence: (M) is logically true iff is true in every model (in the standard model-theoretic sense of "model" according to which models are sets).

232

STEN LINDSTROM

fails. He considers the language of set-theory with (:labs infX) added to it. The sentence (:labsinfX)(X = x) is then an example of a true sentence which is not true in any model (whose domain is a set). So if (M) were correct then ,(:labs infX) (x = X) would be an example of a false but logically true sentence. But there are of course no such sentences, so the equivalence (M) cannot hold in general. 20 If we are considering languages that are sufficiently strong in expressive power, then Kripke's model-theoretic semantics is not sufficient to capture the notions of metaphysical necessity and possibility. Consider, for example, the sentence: (*)

IMI (:labs infX)(X = X)

This sentence is presumably true in the intended interpretation. However, there is no Kripke model structure where it is true. 21 To be exact, we let QS5= by the system of free (i.e., 'free' of existential assumptions) modal predicate logic which is defined as follows. Axioms: (I) Any substitution instance of a theorem of propositional S5. (2) 'Ix¢! 1\ :ly(t = y) - t ¢!(t/x), provided that t is an individual constant or a variable that is free for x in ¢!. (3) Vx(¢! -t,p) - t ('Ix¢! - t Vx,p). (4) 'Ix¢! <-> ¢!, provided x is not free in ¢!. (5) Vx:ly(y = x). (6) t = t. (7) t = tf - t (eb(t/x) - t ¢!(t' Ix)~, provided that t is an individual constant or a variable that is free for x in ¢!. Deduction rules: (MP) If ~ eb and ~ ¢! - t ,p, then ~,p. (Nec) If ~ ¢!, then ~ D¢!. (UG) If ~ eb, then ~ 'Ix¢!. Cf. Garson (1984) and Hughes and Cresswell (1996), chap. 16-17, where the this and similar systems are formulated and proved to be complete with respect to Kripke's (l963a) semantics (these are the systems that Garson refer to as

gIR).

The term "supervalidity" is due to Boolos (1985). The concept itself goes back to Kreisel (I 969). Here, we have, of course, presupposed Kripke's picture of metaphysical reality. Given another picture, for example that of Lewis (1985), we get a different logic of metaphysical necessity (but still a form of quantified S5). 24 Kripke's (1963) semantics allows the domains of quantification to be empty. 2

23

REFERENCES Barcan, (Marcus), R.: 1946a, 'A Functional Calculus of First Order Based on Strict Implication', The Journal of Symbolic Logic, 11, 1-16. Barcan, (Marcus), R.: 1946b, 'The Deduction Theorem in a Functional Calculus of First Order Based on Strict Implication', The Journal of Symbolic Logic, 11,115-118. Barcan, (Marcus), R.: 1947, 'The Identity of Individuals in a Strict Functional Calculus of Second Order', The Journal of Symbolic Logic, 12, 12-15. Boolos, G.: 1985, 'Nominalist Platonism', The Philosophical Review, XCIV(3), 327-344. Carnap, R.: 1946, 'Modalities and Quantification', The Journal of Symbolic Logic, 11, 33-64. Carnap, R.: 1947, Meaning and Necessity: A Study in Semantics and Modal Logic. University of Chicago Press, Chicago. Second edition with supplements, 1956. Cocchiarella, N.: 1975, 'On the Primary and Secondary Semantics of Logical Necessity', Journal of Philosophical Logic, 4, 13-27. Etchemendy, 1.: 1990, The Concept of Logical Consequence. Harvard University Press, Cambridge, Massachusetts. Follesdal, D.: 1994, 'Stig Kanger in Memoriam', in D. Prawitz and Westerstilhl, D. (eds), Logic, Methodology and Philosophy of Science, IX, 885-888. Elsevier, Amsterdam. Garson, 1.w.: 1984, 'Quantification in Modal Logic', in D. Gabbay and Guenthener, F. (eds), Handbook of Philosophical Logic, II, 249-307. D. Reidel, Dordrecht. Hintikka,1.: 1957a, 'Quantifiers in Deontic Logic', Societas Scientiarum Fennica, Commentationes humanarium lifterarum, 23(4). Hintikka, 1.: I 957b, 'Modality as Referential Multiplicity', Ajatus, 20, 49-64. Hintikka, 1.: 1961, 'Modality and Quantification', Theoria, 27, 110-128. Hintikka, 1.: 1969, 'Semantics for Propositional Attitudes', in 1.W. Davies et al. (eds), Philosophical Logic, 21-45. D. Reidel, Dordrecht. Reprinted in 1. Hintikka, Modelsfor Modalities. D. Reidel,

KANGER'S EARLY SEMANTICS FOR MODAL LOGIC

233

Dordrecht, 1969; and in L. Linsky (ed), Reference and Modality. Oxford University Press, London, 1971. Hintikka, J.: 1980, 'Standard vs. Nonstandard Logic: Higher-Order, Modal, and First-Order Logics', in E. Agazzi (ed), Modern Logic - A Survey. D. Reidel, Dordrecht, 283-296. Hintikka, 1.: 1989, 'Is Alethic Modal Logic Possible?', in 1. Hintikka and M.B. Hintikka (eds), The Logic of Epistemology and the Epistemology of Logic. Kluwer, Dordrecht. Hughes, G.E. and M.1. Cresswell.: 1996, A New Introduction to Modal Logic. Routledge, London and New York. Jonsson, B. and A. Tarski.: 1951, 'Boolean Algebras with Operators', American Journal of Mathematics, 73, 891-939; 74, 127-162. Kanger, S.: 1957a, Provability in Logic, Acta Universitatis Stockholmiensis, Stockholm Studies in Philosophy 1, Almqvist and Wiksell, Stockholm. Kanger, S.: 1957b, 'The Morning Star Paradox', Theoria, 23, 1-11. Kanger, S.: 1957c, 'A Note on Quantification and Modalities', Theoria, 23,133-134. Kanger, S.: 1957d, 'On the Characterization of Modalities', Theoria, 23,152-155. Kanger, S.: 1970, New Foundationsfor Ethical Theory, in Hilpinen (ed.), Deontic Logic: Introductory and Systematic Readings, Reidel, 36-58. Earlier mimeographed version: New Foundations for Ethical Theory, Part 1, Stockholm, 1957. Kanger, S.: 19972, 'Law and Logic', Theoria, 38, 105-132. Kaplan, D.: 1986, 'Opacity', in Hahn and Schilpp (eds), The Philosophy of wv. Quine, The Library of Living Philosophers, Vol. XVIII, Open Court, La Salle, Illinois. Kleene, S.c.: 1967, Mathematical Logic, John Wiley & Sons Inc., New York. Kreisel, G.: 1969, 'Informal Rigour and Completeness Proofs', in I. Lakatos (ed), Problems in the Philosophy of Mathematics. North-Holland, Amsterdam. Kripke, S.: 1959, 'A Completeness Theorem in Modal Logic', The Journal of Symbolic Logic, 24, 114. Kripke, S.: 1963a, 'Semantical Considerations on Modal Logic', Acta Philosophica Fennica, fasc. 16, Helsinki, 83-94. Kripke, S.: 1963b, 'Semantical Analysis of Modal Logic I, Normal Propositional Calculi', Zeitschriftfur Mathematische Logic und Grundlagen der Mathematik, 9, 67-96. Kripke, S.: 1965, 'Semantical Analysis of Modal Logic II, Non-Normal Propositional Calculi', in I.W. Addison, L. Henkin and A. Tarski (eds), The Theory of Models. North Holland, Amsterdam, 206-220. Kripke, S.: 1980, Naming and Necessity. Basil Blackwell, Oxford. Lewis, D.: 1985, On the Plurality of Worlds. Basil Blackwell, Oxford. Lindstrom, S.: 1996, 'Modality Without Worlds: Kanger's Early Semantics for Modal Logic', in Odds and Ends, Philosophical Essays Dedicated to Wlodek Rabinowicz on the Occasion of His Fiftieth Birthday, Uppsala Philosophical Studies, 45, Department of Philosophy, Uppsala University. McGee, V.: 1992, 'Two Problems with Tarski's Theory of Consequence', Proceedings of the Aristotelian Society, new series, 92, 273-292. Montague, R.: 1960, 'Logical Necessity, Physical Necessity, Ethics and Quantifiers', Inquiry, 4, 259269. Reprinted in R. Thomason (ed), Formal Philosophy: Selected Papers of Richard Montague. Yale University Press, New Haven and London, 1974. Quine, W. Y.: 1947, 'The Problem of Interpreting Modal Logic', The Journal of Symbolic Logic, 12, 43-48. Tarski, A.: 1936, 'Uber den Begriff der logischen Folgerung', Actes du Congres International de Philosophie Scientifique, 7, I-II. (English translation: 'On the Concept of Logical Consequence', 409--420 in Logic, Semantics, Metamathematics, second edition, Hackett Indianapolis, 1983).

QUENTIN SMITH

A MORE COMPREHENSIVE HISTORY OF THE NEW THEORY OF REFERENCE

1. INTRODUCTION

This essay continues the effort of some previous essays to present a more accurate history of the origins of the New Theory of Reference. The "New Theory of Reference" is typically used by philosophers in different ways; for example, sometimes it is used to refer to a theory that essentially includes the historical chain theory of reference. But the consequence of this is that Kaplan, Perry, Salmon, Marcus, Wettstein, Almog and many others commonly identified with this theory are excluded. The "New Theory of Reference", which is used as a name or referentially used definite description, can be used to refer to many different ideas and in this essay I use it in the following narrow and broad senses. If the New Theory of Reference is defined in a narrow sense so that it includes all or virtually all of the philosophers who are standardly associated with this theory, then we may say someone is a New Theorist of Reference if and only if he or she presented an argument against the Frege-Russellian descriptivism that was the orthodox theory in the 1950s and early 1960s and replaced this descriptivism by a historical chain or direct reference theory (and perhaps an associated doctrine of singular propositions). There are many other ideas commonly associated with the New Theory of Reference, even though all of the New Theorists in the narrow sense do not subscribe to all of these ideas. These ideas include the notions of rigid designation, a posteriori necessity and metaphysical necessity, as well as other ideas. In this paper, I shall study contributions to the "New Theory of Reference" in the broad sense that includes these associated ideas. Someone contributes to or in some respect is a New Theorist of Reference in the broad sense if and only if she or he rejected some of the "Frege-Russell tradition" that was orthodoxy in the 1950s and early 1960s and argued for two or more of the following positions: (i) the idea that names are not disguised, contingent definite descriptions but are directly referential or refer by means of a historical chain, (ii) the idea that names are rigid designators and contingent definite descriptions are nonrigid designators, (iii) the distinction between logical and metaphysical necessity and (iv) the distinction between the necessary / contingent contrast and the a priori/a posteriori contrast.

235

P. W. Humphreys et al. (eds.), The New Theory of Reference © Kluwer Academic Publishers 1998

236

QUENTIN SMITH

This system of classification enables us to consider the highly relevant ideas of philosophers who sometimes are not associated with the New Theory of Reference, such as Plantinga, the early Hintikka, Cocchiarella, F011esdal and others. A large number of later contributors may also be included, such as Graeme Forbes [1985; 1989] and Stephen Yablo [1992], and most of the philosophers of religion who write in the Plantingian tradition. Indeed, this classification would imply (correctly) that the New Theory of Reference is so widespread that a large number, if not most, analytic philosophers in the 1990s can be considered as contributing or belonging in some respects to this philosophical tradition. The use of this broad sense of "New Theory of Reference" is combined with talk of someone contributing to the New Theory or being in some respect a New Theorist of Reference, as distinct from the unqualified statement that someone is a New Theorist of Reference. For example, a philosopher may argue there are a posteriori necessities and metaphysical necessities and yet hold that names are disguised, contingent descriptions. It would be natural to say that such a philosopher is not a New Theorist of Reference in the narrow sense, but nonetheless made some contributions to the New Theory of Reference understood in the broader sense. The basic problem with the standard history of the New Theory of Reference (in the broad sense I am discussing in this essay) is the failure to make an adequate distinction between originators and developers of the basic ideas. The first wave of developers (mainly Kaplan, Donnellan, Kripke and Putnam) are mistakenly thought to be the originators of the ideas, who in fact include mainly Marcus, F0llesdal, Hintikka, Plantinga and Geach. These two groups need to be distinguished from the second wave of developers, who include Michael Devitt, Nathan Salmon, Howard Wettstein, John Perry, Joseph Almog, and many others. In previous essays, I largely concentrated on two of the originators (Marcus and Geach) and three of the members of the first wave of developers (Kripke, Donnellan and Kaplan), but in the present essay I will discuss more or less equally several philosophers who belong to the class of originators - Marcus, Plantinga, Hintikka, Geach, F0llesdal and some others. I will also discuss Kripke's main unknown contribution, his 1962 theory of descriptionally rigid proper names, and Cocchiarella's contribution to the logical/metaphysical distinction. But first we need to understand what it means to "originate" an idea in the New Theory of Reference. 2. DIRECT REFERENCE, SINGULAR PROPOSITIONS, A POSTERIORI NECESSITY, ETC., IN MOORE'S 1899 ESSAY "THE NATURE OF JUDGMENT"

One might say that British analytic philosophy began in August 1898 in a letter that Moore wrote to Desmond MacCarthy, in which Moore announced: "I have arrived at a perfectly staggering doctrine ...". [Moore, 1898]. This (eds.),

A MORE COMPREHENSIVE HISTORY

237

staggering doctrine was an early version of some of the basic ideas in the New Theory of Reference that emerged in the late 1950s, 1960s and has been widely prevalent since the 1970s. This doctrine was first published in Moore's 1899 essay, "The Nature of Judgement", the first bona fide publication of British analytic philosophy, in which he laid out his theory of direct reference, singular propositions that include concrete objects as parts, a posteriori necessities, and other ideas that (in one version or another) later formed the foundation of the New Theory of Reference. The fact that the first publication in AngloAmerican analytic philosophy included an early version of many ideas in the New Theory of Reference is not widely known and it is worth explaining some of Moore's basic notions. Moore considers reality to be composed of (mind-independent) concepts; Moore uses "concept" to mean a universal; he holds a bundle theory of objects and that we directly refer to these objects (bundles of concepts) and their parts in our language. The concrete universe of stones, flowers and humans is made up of concepts and nothing more. Some sentences express propositions that consist of these concepts; these propositions are the analogues of what we today call "singular propositions" or "de re propositions". Two main aims of Moore's article is to reject the descriptivist theory of reference of Bradley and others and replace it with a theory of direct reference and singular propositions, and to reject the Kantian tradition that held all logical necessities are a priori and to replace this with a theory that some logical necessities are a posteriori. Some passages convey Moore's theory of direct reference and singular propositions: "When, therefore, I say 'This rose is red', I am not attributing part of the content of my idea to the rose, nor yet attributing part of the content of my ideas of rose and red together to some third subject. What I am asserting is a specific connexion of certain concepts forming the total concept [the concrete thing] 'rose' with the concepts 'this' and 'now' and 'red'; and the judgment is true if such a connexion is existent ... If the judgment is false, that is not because my ideas do not correspond to reality, but because such a conjunction of concepts is not to be found among existents ... A proposition is composed not of words, nor yet of thoughts, but of concepts ... All that exists is thus composed of concepts ... It seems necessary, then, to regard the world as formed of concepts" [1899: 179-182]. Moore also has a theory of a posteriori logical necessities. Moore considers existential propositions, e.g., Red exists, to be a posteriori and yet necessary. He argues that they are not a priori but nonetheless have the essential mark that Kant assigned to a priori propositions, namely, necessity. An existential and empirical proposition asserts that a simple concept exists in time. These propositions are a posteriori and "perception is to be regarded philosophically as the cognition of an existential proposition" [1899: 183]. Furthermore, "a simple concept cannot be known as one which could exist in time, except on the ground that it has so existed, is existing, or will exist. But we have now to point (eds.),

238

QUENTIN SMITH

out that even existential propositions have the essential mark which Kant assigns to a priori propositions - that they are absolutely necessity". [1899: 189, my italics]. If I assert that Red exists, I mean Red exists now, and "this connexion of red and existence with the moment of time I mean by 'now', would seem to be as necessary as any other connexion whatever. If it is true [which is discoverable by sense perception], it is necessarily true, and if false, necessarily false. If it is true, its contradictory is as fully impossible as the contradictory of2 + 2 = 4". [1899: 190, my emphases]. Accordingly, in the case of some propositions "there would seem no doubt that we mean by it [the proposition] to assert an absolute necessity; but between what precise concepts the necessary relation, of which we are certain, holds, we must leave to experience to discover". [1899: 188]. Since a theory of direct reference, singular propositions and a posteriori necessities appeared in the very first publication of Anglo-American analytic philosophy, the "excitement" expressed by some philosophers over the apparent novelty of ideas of this sort when they become popular in the 1970s may seem to some historians of philosophy to be a trifle ironic or misplaced. The specific versions of the 1960s-1990s theories of direct reference, singular propositions, a posteriori necessities, etc., differed from Moore's [1899], but the general ideas are significantly similar. Some philosophers correctly trace the general idea of directly referential ordinary names and singular propositions back to Russell's [1903], where a somewhat different and considerably more developed account is given than in Moore's [1899]; but the first statement of ideas of this general sort appears in Moore's earlier work, whose influence Russell acknowledged [1903: xviii]. However, in Russell's [1903] we already see the eclipse of some of Moore's anticipations of the New Theory of Reference, such as Moore's theory that there are a posteriori necessities, and in this respect as well Moore's [1899] may be seen as the most important precursor to the later works of the New Theorists of Reference. In 1903 Moore expanded on these ideas to include the distinction among physical necessity, metaphysical necessity and logical necessity. In Moore's [1903: 29] he says the existence of a natural part is a necessary condition for the existence of the good which is constituted by the whole. It is not "merely a natural or causal necessity" [1903: 29]. And yet the relation is synthetic, not analytic, and hence is not a logical necessity [1903: 7, 33]. The synthetic necessities are about what can or "cannot conceivably exist" [1903: 29]; they are metaphysical necessities, and the word "conceivably" is used in Kripke's [1959], Plantinga's [1970], Putnam's [1975] and Yablo's [1992] sense, as a less misleading term than "modal intuitions", which achieved a widespread use in the 1970s. This early appearance of ideas that would be revived in various forms in the late 1950s and 1960s was soon overshadowed by the appearance of Russell's 1905 article "On Denoting". This article initiated the eventual downfall of the "early" or "Moorian" version of the New Theory of Reference and its (eds.),

A MORE COMPREHENSIVE HISTORY

239

replacement by (i) a descriptional theory of ordinary names, (ii) a restriction of directly referential names to demonstrative words in my private language for my sense data (foreshadowed in the second paragraph of "On Denoting"), (iii) an equation of necessity with the linguistically analytic and a priori, (iv) the theory that general rather than singular propositions are stated by sentences with ordinary names, (v) and the elimination of a "metaphysical necessity" that was broader than logical necessity and narrower than physical necessity. All or most of these ideas (or some variant upon them) became the orthodox theory up until the 1960s. This orthodoxy is often referred to as "the Frege-Russell tradition", which is an acceptable practice as long as this phrase is used as a name or referentially used definite description. (It needs to be kept in mind that Frege and Russell did not both hold all these doctrines and Frege's and Russell's theories are incompatible on some major points.) An early exception to this orthodoxy is Wittgenstein's [1922]. Wittgenstein [1922] anticipated the New Theory of Reference to a more significant extent than other logical atomists. He held that names directly refer to objects, that elementary sentences express singular propositions, and that names are rigid designators in that they refer to an object as it exists in each possible world (or state of affairs). Wittgenstein argued that these semantic facts are implied by the underlying structure of any logically possible language (disguised by what is used as "ordinary language"). The above considerations suggest that a point needs to be made about what "introducing an idea into the New Theory of Reference" means. This does not imply being original in the absolute sense, i.e., being the first person to discover or entertain or believe or write about the idea. It has been repeated many times since the early 1970s that the direct reference theory of proper names can be traced back to Mill, although it is rarely noted that this theory can also be found in Hume, Ockham, Plato and Parmenides, to name but a few others. Likewise, the nonequivalence of the a priori/a posteriori and the necessity/ contingent distinction is no more "new" than is the direct reference theory. This distinction was widespread in the scholastic period, with Aquinas and others arguing that some necessary truths (e.g., that God exists) are a posteriori. Some of Aristotle's metaphysical necessities, statements about natural kinds, are a posteriori and Plato argued that certain necessary relations between Forms are a posteriori. Virtually none of the main ideas in the New Theory of Reference are "original" in this absolute sense. Being "original" in the relative sense of introducing an idea or argument into the New Theory of Reference means this: being the first to put forth an argument against the "Frege-Russell orthodoxy" that remained predominant up until at least the 1950s and replacing a part of this tradition with one of the ideas or arguments that are associated with the New Theory of Reference (even if the idea or argument was previously presented by Parmenides, Aristotle, Locke, Aquinas, or others). Accordingly, the question we are addressing in the history of philosophy is this: What ideas and arguments were put forth that led (eds.),

240

QUENTIN SMITH

to the demise of the Frege-Russell tradition and the emergence of the New Theory of Reference in the late 1950s and 1960s (which then led to the New Theory's elaboration and wide acceptance in the 1970s)? 3. QUANTIFYING INTO MODAL CONTEXTS: 1918-1960

The New Theory of Reference is a theory about natural language, but an account of its origins requires us to discuss theories of logical languages, specifically, the languages used in interpretations of quantified modal logic. A syntactics for propositional modal logic appeared in c.l. Lewis' [1918] and a syntactics for quantified modal logic first appeared in 1946 with Marcus' [1946] and (two months later) in Carnap's [1946]. An interpretation or semantics for modal logic first appeared in Carnap's [1946] and [1947]. The key idea of Carnap's semantics is that the interpretation of "OA" was based on the Leibnizian idea that necessary truth is truth in all possible worlds (what Carnap called "state descriptions'); thus "OA" means A is true in all possible worlds (in all state descriptions). A state description is a maximally consistent set of atomic sentences. It is a logically consistent set W of sentences, such that for each atomic sentence S, either S is a member of the set W or the negation of S is a member of the set W. The route from logical languages to natural language followed a path started by Quine in 1941, in a footnote to an article he contributed to Schilpp's The Philosophy of Whitehead. Here Quine first stated his famous "paradox" about quantified modal logic. It is not widely known that Quine first started this paradox in 1941, four years before modal logic was first quantified by Marcus. This paradox is often traced back to a later writing of Quine, his [1947] or sometimes his [1943]. The paradox as it originally appears in Quine's [1941: 142, n. 26] comprises two sentences offootnote 26: "c.1. Lewis and C.H. Langford (Symbolic Logic, New York, 1932), e.g., use a non-truth-functional operator '~' to express logical possibility. Thus the statements: ~

(the number of planets in the solar system < 7)

~

(9

< 7)

would be judged as true and false respectively, despite the fact that they are interconvertible by interchanging the terms '9' and 'the number of planets in the solar system', both of which designate the same object."

If it were not for this footnote (which led to Quine's later restatements of the paradox) and the responses to this paradox, it is doubtful that the New Theory of Reference would have been developed. The traditional form of the paradox in virtually all later discussions is taken from Quine's [1947: 47] formulation, where "C" means congruence (which may be the relation of identity, but Quine does not wish to prejudge that):

(eds.),

A MORE COMPREHENSIVE HISTORY

241

"Morning Star C Evening Star· D (Morning Star C Morning Star).

Therefore, according to (ii) ['An existential quantification holds if there is a constant whose substitution for the variable of quantification would render the matrix true']: (I) (:Jx) (x C Evening Star· D (x C Morning Star)).

But also: Evening Star C Evening Star . ~ D (Evening Star C Morning Star),

so that, by (ii) (2) (:Jx) (x C Evening Star . ~ D (x C Morning Star))."

Since (1) and (2) are mutual contraries, we have a paradox. Before we consider the first relevant response, Smullyan's in 1947, one other idea needs to be introduced which proved crucial for the New Theory of Reference. In his [1943] Quine discusses but rejects a principle of universal intersubstitutivity: "Given a true statement of identity, one of its two terms may be substituted for the other in any true statement and the result will be true". [1943: 113]. This statement may appear to be natural or plausible, given the indiscernability of identicals. If two individuals x and yare identical, they are indiscernible; this may suggest the idea that singular terms for x and y should be intersubstitutable in any context, including modal, temporal, deontic, epistemic and other intensional contexts. But Quine argues that singular terms are not intersubstitutable in modal and epistemic contexts. The idea that singular terms, most notably names, are always "purely designative" was associated with this universal intersubstitutivity thesis. Quine also rejects this idea: "Failure of substitutivity [in modal and epistemic contexts] 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 of the name". [1943: 114] Quine holds that names have a meaning ("criterion of application") in addition to their designative feature, and that the meaning, not just the designation, is relevant to modal and epistemic contexts. By rejecting the universal intersubstitutivity thesis and the associated thesis that names do not have a meaning but instead are purely designative, Quine has to mention these theses and make an issue about whether they are plausible, and his discussion of these theses was a motivating factor in the future discussions that led to the New Theory of Reference. (The impression should not be given that the conception of the universal intersubstitutivity thesis is "original" with Quine. This thesis has been regularly propounded or discussed throughout the history of philosophy, and can be traced at least as far back as Aristotle's Topics vii, I. The point is that Quine drew attention to it in a way that proved influential. The association of an intersubstitutivity thesis with the "purely designative" thesis became a central element of the New Theory of Reference that was not present in Moore's early theory.) (eds.),

242

QUENTIN SMITH

The theory Quine rejected (or rather, one largely similar to it) is also mentioned and rejected four years later, in Carnap's Meaning and Necessity; Carnap calls the theory "the name-relation method". Carnap writes [1947: 98] that the name-relation method is based on three principles (the numbers in brackets are my insertions) and is meant to apply to an artificial or logical language: " [I] Every expression used as a name (in a certain context) is the name of exactly one entity; we call it the nominatum of the expression ... [2] A sentence is about (deals with. includes in its subject matter) the nominata of the names occurring in it. [3] The principle of interchangeability (or substitutivity). This principle occurs in either of two forms: [3a] If two expressions name the same entity ... the two expressions are interchangeable (everywhere). [3b] If an identity sentence ['... is identical with - - -'] is true, then the two argument expressions ' .. .' and' - - -' are interchangeable (everywhere)."

Carnap rejects this name-relation method for several reasons, including the reason that expressions that name the same entity are not always intersubstitutable in modal contexts. The second relevant event in 1947 is Smullyan's innovative use of the "purely designative" and "substitutivity" theses in modal contexts to respond to Quine's paradox. Smullyan is responding to Quine's [1947], in which Quine's point is that "when modal logic is extended (as by Miss Barcan) to include quantification theory ... serious obstacles to interpretation are encountered". [Quine, 1947: 43-48]. Smullyan suggests two ways for interpreting quantified modal logic, each of which is able to resolve Quine's paradox; but only the first way is pertinent to the New Theory of Reference. This solution appears in the third sentence in this passage: "It is possible that by 'constant' is meant what is commonly understood by 'proper name'. Under this interpretation it appears evident to this reviewer that the principle of existential generalization is true. However, we observe that if 'Evening Star' and 'Morning Star' proper-name the same individual they are synonymous and therefore B is false." [Smullyan, 1947: 140]

B is the premise: Evening Star is congruent with Evening Star . rv D Evening Star is congruent with Morning Star. Smullyan's theory has been subjected to many mis-interpretations in recent years, including one of my previous discussions of his theory,l and consequently it is worth quoting enough material from his brief review (which is two and one-half pages in length) to provide sufficient textual evidence for a correct interpretation. In the abovequoted phrase, "what is commonly understood by 'proper name"', Smullyan is not referring to ordinary proper names, but to what philosophers of logic of that time commonly understood as proper names, i.e., proper names in Russell's sense (Russell's logically proper names, "this" and "that", which directly refer to my sense data). Smullyan characterizes his response to Quine's charge of contradiction as a response that employs Russell's theory of descriptions and proper names: "We have seen that in terms of Russell's theory ofdescriptions and proper names this contradiction can be avoided". [1947: 141;

(eds.),

A MORE COMPREHENSIVE HISTORY

243

my italics]. Smullyan's employment of Russell's theory of logically proper names (as distinct from Mill's theory of ordinary proper names) in his first response to Quine was later noted by Prior [1967: 10-11] and others. Smullyan's second and preferred way of responding to Quine, in terms of Russell's theory of descriptions, is that the necessity of identity thesis for individuals is false, which is an implication of the following sentence, the crux of his second response: "For if it is not necessary that the morning star exists then it is not necessary that the morning star is self-congruent" [1947: 140]. This is a rejection of one of the premises of Quine's paradox, viz., "0 (Morning Star is congruent with Morning Star)". Smullyan is talking about artificial (formal or logical) languages (specifically, about artificial languages suitable for interpreting quantified modal logic), rather than natural or ordinary language. For this reason, Russell's argument that in ordinary language, expressions such as "Evening Star" or "Hesperus" are not proper names but truncated descriptions is not pertinent to the discussion. In his artificial language, Smullyan is not required to use words in the way they are ordinarily used. But Smullyan does remark that the treatment of "Evening Star" or "Hesperus" as a logically proper name in modal contexts does not accord with ordinary language and therefore that this particular solution to Quine's paradox is disadvantaged relative to his second solution. He prefers the solution in terms of Russell's theory of the primary and secondary occurrences of definite descriptions, which accords more with ordinary language. Thus, Smullyan concludes his review by saying: "In the judgment of the reviewer, the complications to which Carnap and Church resort underscore certain advantages obtained by retaining Russell's treatment of descriptions with its associated doctrine concerning the primary and secondary occurrence of descriptive phases. This doctrine, as Carnap admits [1947: 140), has the advantage of being in close accord with ordinary usage. It also permits a logical theory which unifies the theory of quantification with that of modality in a manner which Quine believes impossible. However, in defense of Quine's skepticism, in the reviewer's opinion, there is not a scintilla of evidence." [1947: 141).

At the risk of belaboring this point, Smullyan is noting that one of his ways of resolving Quine's paradox in an artificial language has, as an added advantage to its logical validity, the feature of "being in close accord with ordinary use". Being in accordance with ordinary language is not a necessary condition o((but merely an added advantage for) a solution to Quine's paradox, since Smullyan takes the discussion to be about logical language used in an interpretation of quantified modal logic, not about ordinary language. Smullyan is also saying that this added advantage belongs to the resolution in terms of primary and secondary occurrences of descriptive phases, not to the resolution in terms of logically proper names, which reflects Smullyan's belief that names in ordinary languages are not directly referential. (Smullyan reaffirms his preference for a descriptional theory in his [1948]). Here we are still a long way from the Millian theory of ordinary proper names, the necessity of identity for Millian names, the intersubstitutivity of Millian names in modal contexts, the rigidity of (eds.),

244

QUENTIN SMITH

Millian names, etc., that first appear in Marcus' [1961]. The next pertinent year in the route to the New Theory of Reference is 1949, which features Frederick Fitch's "The Paradox of the Morning Star and Evening Star". Fitch, like Smullyan, is talking about a logical language; his aim is to show that "the modal logician is free to deal directly with actual individuals and to employ the relation of identity between them". [1949: 138]. Fitch is criticizing Quine's argument that if we are to use quantified modal logic we must follow Carnap and let our variables range over individual concepts. (Quine rejected individual concepts, but had argued the modal logician is committed to them.) Fitch restates Smullyan's two arguments, including the argument employing Russellian logically proper names, to solve Quine's paradox. Fitch adds a third argument, based on using Marcus' 1946 theorem of the necessity of identity and considerations of scope differences for descriptive phrases. Although this third argument is the main point of the article, the significance of the article for "the route to the New Theory of Reference" is that we find a second person (Marcus' dissertation advisor) endorsing the validity of using Russellian proper names in modal contexts in an artificial language to respond to Quine's paradox. Fitch's 1950 article on "Attribute and Class" is more important since we find him going beyond Smullyan and mentioning the full universal intersubstitutivity thesis that Quine and Carnap discussed and rejected. (Smullyan mentioned only intersubstitutivity in modal contexts.) Fitch first writes about the import of Marcus' 1947 article on "The Identity of Individuals in a Strict Functional Calculus of First Order" for the identity of entities: "the system of modal logic developed by Ruth Barcan suggests that the simplest view is that no identities should be regarded as merely contingent and that identified entities should be everywhere intersubstitutable. (Indeed, no entity is correctly identifiable with any entity but itself, so permission of substitution of this sort is trivial anyway". [1950: 552]. Fitch mixes ontological with semantic theses in these two sentences, but proceeds to formulate a specifically semantic thesis: "Furthermore, if entities X and Y have been identified with each other, it seems reasonable to suppose that the names of X and Y should also be everywhere intersubstitutable". [1950: 552]. Fitch, however, no more endorses this view that Smullyan endorses the thesis about Russellian names in modal contexts. Fitch immediately adds: "According to Church's view, on the other hand, two names of the same thing might differ in sense and so not be intersubstitutable". [1950: 552]. Fitch later mentions Carnap's view, and Smullyan's two responses to Quine's paradox in terms of Russellian proper names and descriptions, but Fitch does not commit himself to anyone of these views (the universal intersubstitutivity thesis, Smullyan's two responses to Quine, Church's view, and Carnap's view.) Fitch's general point is the same point made in his [1949] and in Smullyan's [1947] and [1948], namely, that there are many ways to respond to Quine's paradox about quantified logic and to construct a valid logic of this sort. (eds.),

A MORE COMPREHENSIVE HISTORY

245

As least three steps need to be made to get from Smullyan's and Fitch's papers to the New Theory of Reference. First, somebody needs to endorse, and not merely mention as one of the many possible solutions to Quine's paradox, the thesis that names are directly referential and intersubstitutable in modal contexts. Second, somebody must start talking about natural language and ordinary names, and not merely about names in an artificial language used for purposes of interpreting quantified modal logic. Third, the names need to be conceived as Millian names (in the broad and familiar contemporary sense) rather than as Russellian logically proper names. The first step (endorsing the direct reference and substitutivity thesis) is taken in 1960 in an article by Marcus, "Extensionality" [1960] and at greater length by F0llesdal in his 1961 Harvard dissertation on Referential Opacity and Modal Logic. (For my chronology to be accurate, I should point out that between Fitch's [1950] and Marcus' [1960] there appeared Kanger's [1957b] and Hintikka's [1957a; 1957b], which contained more extensive and logically original solutions to Quine's paradox than Smullyan's and Fitch's relatively brief responses. I shall discuss Kanger's and Hintikka's work separately when I discuss the origination of the concept of a metaphysical [as distinct from logical] semantics for modal logic.) In Marcus' [1960], she continues in the tradition of talking about artificial or logical languages that are suitable for interpreting quantified modal logic. She sums up her paper as follows: "I have tried in this brief paper, to characterize the theory of extensionality, and to show that logical systems are more or less extensional" [1960: 62]. In a footnote [1960: 61, n. 2] Marcus sees herself as continuing in the Fitch (and thus Smullyan) tradition of stating that one of the ways to resolve Quine's paradox is to treat the relevant expressions as proper names, in which case there is intersubstitutivity. She writes about the sentence "It is necessary that the evening star is the evening star" and she says that if it "involves proper names of individuals then 'the evening star' may replace 'the morning star' without paradox ..." [1960: 61]. In Marcus' [1960], we do not see the reservations expressed by Smullyan and Fitch about the response to Quine that involves treating "evening star" as a proper name, and we do not find here any sympathy with Smullyan's and Fitch's suggestion that a solution in terms of definite descriptions may be preferable because of greater conformity with ordinary usage. Marcus views the "proper name" solution as provably valid, given her theorems about identity, indiscernability, weaker equivalences, and about the intersubstitutivity of expressions; she argues that the "proper name" solution follows from her theorems. But Marcus in this article has her attention only on the interpretation of logical systems, and is not concerned to make any claims about ordinary usage. Further, we have as yet no indication that she has in mind Millian names as distinct from the Russellian names that Fitch and Smullyan used, although the reference to Fitch suggests that she does not at this time see herself as departing from the theory of names employed in Fitch's and (eds.),

246

QUENTIN SMITH

Smullyan's articles. The "Big Leap" into the New Theory of Reference is made next year, in her "Modalities and Intensional Languages" [1961]. But first we will consider Dagfinn F011esdal's 1961 theory, which is the most fully developed formal analogue (a theory of the language used in interpretations of quantified modal logic) to the theory of natural language in the New Theory of Reference. 4. F0LLESDAL'S 1961 THEORY OF DIRECTLY REFERENTIAL RIGID DESIGNATORS

Smullyan [1947], Fitch [1949; 1950] and Marcus [1960] discussed some formal analogues in the language of quantified modal logic to the natural language counterparts in the New Theory of Reference. By 1961, the philosopher who presented the most extensive formal analogue to the New Theory of Reference was Dagfinn F011esdal in his 1961 Harvard doctoral dissertation, Referential Opacity and Modal Logic. This was later privately published with minor additions in [F011esdal, 1966]. F0llesdal originates or develops the theory that genuine singular terms are rigid, the universal intersubstitutivity thesis, the necessity of identity thesis, the difference between rigid names and nonrigid definite descriptions, the difference between rigid definite descriptions and rigid, non-descriptional names, the modally oriented characterization of proper names in terms of the notion of possible worlds, the notion of a weak rigid designator, the endorsement of essentialism, and other theses that place him squarely in the conceptual context of the New Theory of Reference rather than of that of the Frege-Russell tradition that circumscribed the limits of Smullyan's and Fitch's discussion. F011esdal's original solution to Quine'S paradox is to reject the assumption that singular and general terms have the same semantics and to argue that genuine singular terms are characterized by the fact that they exhibit rigidity. F011esdal's "genuine singular terms" include variables, pronouns, and "genuine names" in the broad sense, which include both proper names and necessary descriptions. Genuine singular terms are defined as having a stable or rigid reference in terms of the notions of the referential transparency and extensional opacity of a construction. Constructions are referentially transparent in that whatever is said to be true of an object is true of the object regardless of how it is referred to. They are extensionally opaque in that we cannot substitute coextensional general terms or sentences for one another in modal contexts. In modal contexts, genuine singular terms refer to the same object in each possible world (or each possible world in which the object exists) and are intersubstitutable. F0llesdal presents three possible solutions to Quine's paradox, depending on which terms in our language are rigid, only variables and pronouns, or also genuine names. The solution involving "genuine names" includes both rigid definite descriptions and rigid names that have no descriptive context. (eds.),

A MORE COMPREHENSIVE HISTORY

247

F011esdal defends at greater length than previous writers (specifically Marcus [1960)) the universal intersubstitutivity thesis and the strong identity thesis.

F011esdal holds that the identity sign cannot be flanked by a contingent description but only by a genuine singular term (a rigid designator). F011esdal sees the universal intersubstitutivity thesis as belonging to the meaning of identity and as a necessary premise in the derivation of the necessity of identity. [1966: 111-112]. (As I mentioned, F011esdal privately reprinted his 1961 dissertation in [1966] with an added preface and small additions to sections with which I am not here discussing or quoting; I quote and refer to only the parts of his [1966] that are unchanged from his [1961)). F011esdal is also the first to present a defense of the universal substitutivity thesis in both epistemic and modal contexts. (For example, Marcus endorsed the universal substitutivity thesis, but defended it only against objections pertaining to modal contexts in her [1960], [1961] and [1963].) F011esdal argues that the unrestricted substitutivity of identity in modal contexts amounts to essentialism: "if an attribute is necessary of an object, it is necessary of an object regardless of the way in which the object is referred to". [1966: 120]. F011esdal here is referring to Quine'S idea that an attribute is necessary of an object only relative to a certain way of describing it, and that denying this is tantamount to affirming essentialism. At the time of this dissertation (1961), the discussions by Marcus [1961; 1967; 1971], Parsons [1969] and others of the distinctions between trivial and nontrivial essentialism had not yet been published and F 011esdal has in mind only nontrivial essentialism in his dissertation. F011esdal characterized proper names in his 1961 Harvard dissertation as follows: "This solution [the second solution to Quine's paradox] leads us to regard a word as a proper name of an object only if it refers to this one and the same object in all possible worlds". [1966: 96-97]. F011esdallater discusses the problem of possible worlds in which the object does not exist and introduced what later became called [Kripke, 1972] "weak rigid designators". F011esdal formulates the notion of a weak rigid designator in [1966: 124-134]. In this discussion, F011esdal becomes the first to propose a theory ofa varying domain semantics combined with a theory that names are rigid designators, which is the first time the conceptual distinction between strong and weak rigid designation is made. A weak rigid designator designates the same object in every world in which that object exists, and designates nothing in worlds in which that object does not exist. A strong rigid designator designates the same object in every world and the object exists in every world. But note that F011esdal's distinction is made in the context of discussing the language of quantified modal logic, and Kripke's distinction is instead made in the context of analyzing ordinary usage [Kripke, 1971; 1972]. Quine was F011esdal's dissertation advisor and was a professor at Harvard University when F011esdal was a graduate student and Kripke an undergraduate student at this university. Quine has maintained that F011esdal's (eds.),

248

QUENTIN SMITH

concept of genuine names in the broad sense is the same as Kripke's concept of rigid designators, even as late as [1994]. Quine writes in [1981: 118] about Kripke's rigid designators and F0llesdal's genuine names and genuine singular terms: "a term thus qualified is what F0llesdal called a genuine name and Kripke has called a rigid designator. It is a term such that (Ex)D(x = a), that is, something is necessarily, where 'a' stands for the term .... A rigid designator differs from others in that it picks out its object by its essential traits". In Quine's [1994: 148] he writes: " ... Hence Dagfinn FeJllesdal's genuine singular terms, or Saul Kripke's rigid designators. These are the terms that obey substitutivity of identity even in modal contexts ... As we might say in a modal spirit, these are the terms that name their objects necessarily". Quine appears to be basically correct in his characterization. For example, Kripke [1971: 145] writes: "What do I mean by 'rigid designator'? I mean a term that designates the same object in all possible worlds". F0llesdal says the same of his genuine singular terms. For example, he writes in 1961 of some of his genuine singular terms, proper names, that we "regard a word as a proper name of an object only if it refers to this one and the same object in all possible worlds. This does not seem unnatural'. [F011esdal, 1966: 97]. Although F011esdal's statements (e.g., "this does not seem unnatural") suggest that he believes his theory also applies to ordinary language, he does not engage in the ordinary language analysis of Kripke's [1971; 1972] and has his attention on the language of quantified modal logic. This is the most important difference between the two theories that Quine does not mention. Furthermore, Quine is mistaken in believing that Kripke's rigid designators and F011esdal's genuine singular terms always refer in a descriptional way, i.e., that such an expression always "picks out its objects by its essential traits". As is well-known, Kripke held that proper names are not descriptionally rigid in his [1971] and [1972]. And F011esdal writes in his 1961 dissertation that if a name-like word is nonrigid, it is descriptional, and that proper names do not contain descriptional elements: we regard "a word as a proper name of an object only if it refers to this one and the same object in all possible worlds. This does not seem unnatural. Neither does it seem preposterous to assume as we just did, that if a name-like word does not stick to one and the same object in all possible worlds, the word contains some descriptive element'. [1966: 97]. F011esdal's and Kripke's proper names refer nondescriptivally to the same object in all possible worlds. They differ in that F011esdal does not subscribe to the historical chain theory of reference [Geach, 1969], whereas Kripke does subscribe to this theory, which is the main difference between the two theories if they are both taken to characterize natural language. The idea of directly referential rigid names appears again in F011esdal's [1967] recapitulation of some of the basic ideas in his 1961 dissertation, and he writes about epistemically possible worlds that: "The fact that the expressions 'the man who comes towards me' and 'Coriscus' change their reference from world to world in this manner, should perhaps be taken as evidence that they (eds.),

A MORE COMPREHENSIVE HISTORY

249

contain some descriptive element, and that they should therefore not be regarded as genuine names. The only descriptions that should be regarded as genuine names are those which keep the same descriptum in every epistemically possible world". [1967: 11-12]. In 1961, F011esdal regards "a word as a proper name of an object only if it refers to this one and the same object in all possible worlds". [F011esdal, 1966: 96-96]. F 011esdal regards this as a necessary condition of a proper name, not as a sufficient condition, since the word in addition needs to be directly referential. A proper name does "not assign a property" to its referent and serves "merely and solely to name it" [1966: 97]. If "a name-like word does not stick to one and the same object in all possible worlds, the word contains some descriptive element" [1966: 97] and thus is not a proper name. According to Almog [1986] and Marcus [1993: 212, 248, n. 19], Kripke [1971; 1972] regards F011esdal's necessary condition ("refers to this one and the same object in all possible worlds") as a sufficient condition to be a proper name. Both Almog and Marcus criticize this approach since it fails to distinguish directly referential names from rigid definite descriptions. (Almog and Marcus only refer to Kripke in their discussion, not to F011esdal.) However, it seems Almog and Marcus are not correct on this matter, since Kripke held that names are connected to their referents by an historical chain and that they do not refer to the object descriptively; these two criteria distinguish names from rigid descriptions. According to Kripke, the condition ("refers to this one and the same object in all possible worlds") is merely a necessary condition to be a proper name. F011esdal interacted with the other philosophers who at this early time were discussing rigidity - Kripke, Marcus, Hintikka and others - but the lines of influence are not easy to trace. We do know that F011esdal [1961] and Marcus [1960] refer to the Smullyan and Fitch responses to Quine as their background material; F011esdal [1961] also expresses indebtedness to Hintikka's [1957a; 1957b] and extensively discusses Marcus' 1940 writings on quantified modal logic and the necessity of identity, as well as Carnap's [1946; 1947], and Quine's [1941; 1943; 1947]. Hintikka [1963: 71-72] refers to his earlier [1962: 138-158], F01lesdal [1961] and Quine's writings as the background for Hintikka's discussion of rigid names/nonrigid descriptions [1963], presented at a conference in August, 1962 at Helsinki attended by Marcus, Kripke and others [Marcus, 1993: 89]. F011esdal indicated [1966: 43, n. 1] he was indebted to conversations with Kripke (his fellow student at Harvard in the early 1960s) for Kripke's explanation to him (in Spring 1961) of Kripke's theory of iterated modalities. Kripke and F011esdal had extensive discussions in 1960 and 1961 (see [F011esdal, 1966: pp. v and 43, n. 1; 1994: 888, n. 7]); in their discussions, Kripke did not communicate to F011esdal (in 1961 or earlier) any ideas about the distinctions among directly referential names that were weak rigid designators, nonrigid definite descriptions, rigid definite descriptions or the idea about the necessity of identity between directly referential rigid names (Kripke (eds.),

250

QUENTIN SMITH

[1980: 3-5]; FlIlllesdal [1966: pp. v, and 43, n. 1]). Kripke at this time (19601961) had not developed and did not endorse a theory of the directly referential, rigid name/nonrigid description contrast or the necessity of identity between rigid names (see [Kripke, 1980: 3-5]. According to Kripke's account of his influences, it may appear that FlIlllesdal's 1961 theory, Marcus' 1961 theory, and the earlier writings of Fitch and Smullyan had no influence on his development in 1963-64 of the rigid name/nonrigid definite description theory, the direct reference theory, the theory of the necessity of identity between names, and related theories (see Kripke [1980: 3-5] and [1972: 342343, n. 2]. FlIlllesdal, Fitch and Smullyan are not mentioned in [Kripke, 1971; 1972]. This account may be natural inasmuch as [Kripke, 1971; 1972] is only about ordinary usage and FlIlllesdal, Fitch and Smullyan are concerned primarily with the language used in an interpretation of quantified modal logic; by comparison, in Kripke's work on quantified modal logic (e.g., [1963b, n. 1] he refers to Hintikka's influence and similar views. Marcus [1961] is in significant part about ordinary usage and her theory of ordinary language is mentioned in Kripke [1971; 1972], but Marcus is said to have had no influence (see Kripke's remark quoted in [Holt, 1996: 36]; also see an alternative hypothesis formulated in [Smith, 1995a; 1995b]). Only Rogers Albritton is specifically named in [Kripke, 1972: 342, n. 2] as having an influence on Kripke's [1972]: "Albritton called the problems of necessity and a prioricity in natural kinds to my attention ..." [1972: 342, n. 2]. Another sentence, "I also recall the influence of early conversations with Albritton and with Peter Geach on the essentiality of origins", is added to the footnote in [1980: 23, n. 2]. Putnam, Donnellan, Chastain, Slote and some "philosophers mentioned in the text" [1972: 342, n. 2] are mentioned as "independently" [1972: 342, n. 2] expressing some similar views. There are also the following two sentences in the footnote [1972: 342, n. 2]: "The apology in the text still stands; I am aware that the list in this footnote is far from comprehensive. I make no attempt to enumerate those friends and students whose stimulating conversations have helped me". As this section indicates, FlIlllesdal made a number of largely unrecognized contributions to the New Theory of Reference in [1961]. To summarize a few, he integrates the universal intersubstitutivity thesis and the thesis of the necessity of identity, holds that only rigid designators flank the identity sign, holds that proper names are weak rigid designators and are not disguised descriptions, and defends the universal intersubstitutivity thesis against objections based on both epistemic and modal contexts. Earlier [Marcus, 1960] and independently [Marcus, 1960; 1961], Marcus made some partly similar points (and additional points). F0llesdal had extensively studied Marcus' 1940s writings, but not her [1960; 1961] before finishing his dissertation [F0llesdal, 1961: 54-59]. Marcus did not read F011esdal's dissertation until after it was privately printed in 1966 [Marcus, 1993: 231]. One of the crucial differences (among many) between Marcus (eds.),

A MORE COMPREHENSIVE HISTORY

251

[1961] and F01lesdal [1961] is that Marcus explicitly intends her theory to be about ordinary usage. F011esdal remains in the tradition of Smullyan, Fitch and Marcus [1960] of centering his discussion on the language of quantified modal logic. If F011esdal's theory had been explicitly developed as a description of ordinary language in his [1961], then he and Marcus would have both (independently) made the breakthrough to the New Theory of Reference in 1961. 5. THE BREAKTHROUGH TO THE NEW THEORY OF REFERENCE: MARCUS, 1961

The striking feature about Marcus' [1961] is that she restates the "proper name" solution to Quine's modal paradox and then proceeds to show its conformity to ordinary language and that the relevant "proper names" are ordinary names in the Millian sense and exhibit necessity of identity. She further argues that Millian names are stable in modal contexts and are rigid designators. She later in the article provides a possible world semantics for quantified S4, which includes rigid individual constants, and thereby provides a formal backing to her theory of rigid names in natural language. The New Theory of Reference makes its first appearance in this article. In addition to the ideas mentioned, other ideas belonging to the New Theory of Reference also first appear in her [1961]. The relevant quotes and discussion appear in my [1998a; 1995b; 1995a] and I refer the reader to these three essays. There is a historical and logical connection between the theses discussed by Quine and Carnap in the 1940s and Marcus' first statement of the New Theory of Reference. Marcus applies to natural language something analogous to Quine'S "universal intersubstitutivity thesis" and the idea that names are always "purely designative", and also something relevantly analogous to Carnap's "name-relation method". Quine and Carnap rejected this position for both natural and artificial languages, but Marcus adopts a version of this position for both logical and natural languages. We have seen how these theses were discussed by Quine, Carnap, Smullyan, Fitch, F011esdal and Marcus [1960] in the interpretation of quantified modal logic; Marcus' [1961] stands out by virtue of her arguments that these theses apply to ordinary language. In her [1961] Marcus only addresses the objections to intersubstitutivity in modal contexts in natural and artificial languages. She did not address the objections to intersubstitutivity in epistemic contexts and does not develop a theory of epistemic contexts, although her direct reference theory of names would seem to imply the sort of theory of epistemic contexts that was later developed by New Theorists (including herself) in the 1980s. More importantly, F011esdal developed a relevant theory of epistemic contexts in his [1961; 1967] in which he defended the universal intersubstitutivity thesis for both modal and epistemic logic.

(eds.),

252

QUENTIN SMITH

Marcus discusses F"lllesdal's [1966] in an April 1988 talk at Washington University in St. Louis (see Marcus [1993: 215-232]. She characterizes F0llesdal's "second solution" to Quine's paradox as follows: "The second solution seems to accommodate what Kripke [1971] later called 'rigid designators', which include proper names as well as some descriptions." [1993: 231]. Thus, both Marcus and Quine [1981; 1994] believe F0llesdal's "genuine singular terms" express the same concept as Kripke's "rigid designators". I have discussed the merits of this belief in the earlier section on F0llesdaI. Marcus also believes F0llesdal [1966] did not hold the theory that proper names are directly referential and she refers to F011esdal's [1986] as F011esdal's endorsement of the direct reference theory. Marcus is correct inasmuch as the directly referential character of proper names is only very briefly discussed in F011esda1 [1966] and is not the focus of his discussion, as it is in Marcus' [1961]. But F011esdal does hold proper names are directly referential in [1966: 96-97], with the only caveat being that F0llesdal did not present his theory as an analysis of ordinary language but as a theory of quantified modal logic. 6. KRIPKE'S 1962 THEORY OF DESCRIPTIONALLY RIGID NAMES

After Marcus' [1961], the next relevant development in the theory of ordinary names appeared in the February, 1962 discussion of Marcus' [1961] among Kripke, F01lesdal, Quine, Marcus and others. Kripke did not have a theory of rigid names in 1961 (see Kripke [1980: 3-5], but we can see sketches ofa theory of descriptionally rigid names in Kripke's 1962 remarks in [Marcus et aI., 1962]. This counts as one of Kripke's original and important contributions to the New Theory of Reference, for Marcus [1961] and F011esdal [1961] held that rigid proper names (in ordinary or logical language) are directly referential. The theory of descriptionally rigid proper names is standardly said to be originated by Linsky in [1977] and Plantinga in [1978], but I think the following texts show that Kripke has priority in this regard. On February 7, 1962, we see some relevant ideas discussed by Quine, Kripke, F01lesdal, Marcus and others at the Discussion following Marcus' talk [Marcus et aI., 1962]. At this time, Kripke held a Fregean-Russellian descriptional theory of proper names and he appears to have misunderstood Marcus' theory in a fruitful way by trying to assimilate it to this paradigm. First, we should note how Kripke (in 1980) described the theory he held in 1962. He recounts that: "the ideas in Naming and Necessity evolved in the early sixties - most of the views were formulated in about 1963-64 .... Eventually I came to realize - this realization inaugurated the aforementioned work of 1963--64 - that the received presuppositions against the necessity of identities between ordinary names were incorrect, that the natural intuition that the names of ordinary language are rigid designators can in fact be upheld .... Thus at this stage I rejected the conventional description theory as an account of meaning.... Let me not pay inadequate tribute to the power of the then prevailing complex of ideas, emanating from Frege and from Russell, that I then abandoned ....

(eds.),

A MORE COMPREHENSIVE HISTORY

253

Although I, with others, always felt some strain in this edifice ["the description theory of proper names"), it took some time to get free of its seductive power." [1980: 3-5).

As Kripke indicates, he "abandoned" the Frege-Russell "description theory of proper names" in 1963-64. This account accords well with his remarks on February 7, 1962, which reveal him trying to understand Marcus' theory of proper names in terms of the descriptional theory of proper names he then held. I will quote the entire, continuous text [Marcus et aI., pp. 142-143] that is relevant to his theory. (The printed transcript is a heavily edited version of the original remarks; all the discussants approved of the edited version prior to its publication.) I will put the quoted passages in italics, to separate them clearly from my extensive interpolations regarding Kripke's and Quine's remarks. My interpolations are placed in brackets. We begin with Marcus' "dictionary remark": "Presumably, if a single object had more than one tag, there would be a way of finding out such as having recourse to a dictionary or some analogous inquiry, which would resolve the question as to whether the two tags denote the same thing. If "Evening Star" and "Morning Star" are considered to be two proper names for Venus, then finding out that they name the same thing as "Venus' names is different from finding out what is Venus' mass, or its orbit. It is perhaps admirably flexible, but also very confusing to obliterate the distinction between such linguistic and properly empirical procedures."

[Kripke responds:] "That seems to me like a perfectly valid point of view': [Here Kripke is agreeing with what he takes to be Marcus' theory. Kripke continues:] "It seems to me the only thing Professor Quine would be able to say and therefore what he must say, I hope, is that the assumption of a distinction between tags and empirical descriptions, such that the truth-values of identity statements between tags (but not between descriptions) are ascertainable merely by recourse to a dictionary, amounts to essentialism itself'. Here Kripke misinterprets what Marcus just said; Marcus said recourse to a dictionary is an example of a way of finding out if two names are co-referring, and Kripke mis-states this as saying that the truth-values of identity statements between tags are always ascertainable "merely by recourse to a dictionary". In his [1971: 142-143] and [1972: 305] Kripke quoted or restated his own remark here about a dictionary, stated that he had in mind "ideal dictionaries", attributed this remark to Marcus, and thereby mistook Marcus to be saying that identity statements between names are a priori. This is the theory of Kripke held in February 1962 (as we shall see), but it is not Marcus'. A second point to make about this sentence of Kripke's (the last italicized sentence I quoted) is that it reveals the Frege-Russellian descriptional theory of proper names he held in February 1962; he regards the theory arrived at via his (mis)- interpretation of Marcus as a "perfectly valid point of view" that implies that identity statements between names are known a priori merely by recourse to a (an ideal) dictionary. The reason he thinks this implies essentialism is stated in his next sentence:] "The tags are the 'essential' denoting phrases for individuals, but empirical descriptions are not, and thus we look to statements containing 'tags; not descriptions, to ascertain the essential properties of indivi(eds.),

254

QUENTIN SMITH

duals. Thus the assumption of a distinction between 'names' and 'descriptions' is equivalent to essentialism':

[Marcus (see [Marcus, 1993: 142]) and my first essay on this topic [Smith, 1995a], contain a misinterpretation of this sentence. Marcus' interpretation of this sentence (which I endorsed in my [1995aD is that it implies that an individual, say Socrates, has the essential property of being named "Socrates". But this is not what Kripke is saying. Rather, Kripke is saying here what Quine and F011esdal- participants in this 1962 discussion - took Kripke to be saying. Consider this later remark by Quine: " ... Hence Dagfinn F",J1esdal's genuine singular terms, or Saul Kripke's rigid designators. These are the terms that obey substitutivity of identity even in modal contexts. These are the terms, also, that support inferences by existential generalization, even in modal contexts: other terms do not. As we might say in a modal spirit, these are the terms that name their objects necessarily. They name them on the score of essential traits, not accidental ones." [Quine, 1994: 148].

Here "name their objects necessarily" does not mean the objects have the same name in each possible world (contra Marcus [1993: 142]) and contra my earlier self [Smith, 1995a: 187]. It means what Quine explains it to mean in his next sentence: "They name them on the score of essential traits, not accidental ones". [Quine, 1994: 148]. The phrase "naming or referring to an object necessarily or essentially" was the sort of language used in the early 1960s to talk about rigid designation. Thus in Hintikka's August, 1962 talk at a conference with Marcus and Kripke, he presents his theory of rigid designators in these terms; he writes of a singular term: "But referring to it [the object] in all these alternatives ["possible worlds"] is tantamount to referring to it necessarily". [1963: 73]. Kripke's remark (the last sentence in italics I quoted) shows his assimilation of Marcus' theory of the distinction between proper names and contingent descriptions to his own Frege-Russell descriptional theory of names. Kripke believes that Marcus' distinction implies that names refer to objects rigidly via identifying them in terms of their essential properties. Using Frege's language (which is not Kripke's), Kripke's theory is that the sense of names includes the essential properties of the object referred to. Using Russell's language (which is not Kripke's), Kripke's theory is that names are truncated definite descriptions that describe objects in terms of their essential properties. Let us continue with Quine's response to Kripke in the 1962 discussion, which is a rejection of Kripke's theory: "Professor Quine: My answer is that this kind of consideration is not relevant to the problem of essentialism because one doesn't ever need descriptions or proper names. If you have notations consisting of simply propositional functions (that is to say predicates) and quantifiable variables and truth functions, the whole problem remains. The distinction between proper names and descriptions is a red-herring. So are the tags. (Marcus: Oh, no.)"

[Here Quine seems to be getting at the point that if we interpret quantification objectually, we do not need to phrase the problem of essentialism in terms of a special class of rigid singular terms, proper names. It can be phrased in (eds.),

A MORE COMPREHENSIVE HISTORY

255

terms of quantifiable variables, propositional functions and truth functions. F011esdal [1961] argued that quantifiable variables are rigid singular terms and F011esdal might agree that the "problem" of essentialism remains even without considering Marcus' distinction between proper names and descriptions. Quine continues:] "All it is a question of open sentences which uniquely determine. We can get this trouble every time as I proved with my completely general argument of p in conjunction with ¢x where x can be as finely discriminated an intension as one pleases - and in this there's no singular term at all except the quantifiable variables or pronouns themselves. This was my answer to Smullyan years ago, and it seems to me the answer now. Mr Kripke: Yes, but you have to allow the writer what she herselfsays, you see, rather than arguingfrom the point of view ofyour own interpretation of the quantifiers." [Kripke here is saying that for purposes of evaluation of Marcus' talk we should allow Marcus' substitutional interpretation of quantification, and not assume Quine's objectual interpretation ("the quantificational sense of quantification", as Quine puts it).] "Professor Quine: But that changes the subject, doesn't it? I think there are many ways you can interpret modal logic. I think it's been done. Prior has tried it in terms of time and one thing and another. I think any consistent system can be found an intelligible interpretation. What I've been talking about is quantifying, in the quantificational sense of quantification, into modal contexts in a modal sense of modality. Mr Kripke: Suppose the assumption in question is right - that every object is associated with a tag, which is either unique or unique up to the fact that substituting one for the other does not change necessities - is that correct? Now then granted this, why not read 'there exists an x such that necessarily p ofx'as (put in an ontological way ifyou like) 'there exists an object x with a name a such thatp ofa is analytic: Once we have this notion of a name, it seems unexceptionable. Professor Quine: It's not very far from the thing I was urging about certain ways of specifying these objects being by essential attributes and that's the role that you're making your attributes play. Mr Kripke: So, as I was saying, such an assumption of names is equivalent to essentialism." [Kripke is reaffirming his position that names denote objects by way of specifying their essences and that 'a is p' is analytic if "a' is a name and p an essence of a. At this point he is prepared to accept the conventionally accepted equivalence between the analytic, the a priori, the necessary and the essential. He does not appreciate Marcus' point that "the kind of uses to which logical modalities are put have nothing to do with essential properties in the old ontological sense. The introduction of physical modalities would bring us closer to this sort of essentialism". (Marcus et aI., p. 141). At this time, only Marcus distinguishes between non-trivial essential attributions (e.g., 'Cantor is a mathematician') and what is logical and analytic.] "Professor Cohen: I think this is a goodfriendly note on which to stop."

The transcript of the discl,lssion ends here. All of the italicized sentences form a continuous text in [Marcus et aI., 1962, pp. 142-143]. Kripke maintains that if we assume Marcus' tag theory of names, then we have esentialism since her tags (allegedly) involve designating objects in terms of their essential attributes. Kripke misinterprets Marcus' naming relations to objects as ways of specifying these objects by essential attributes and Kripke holds that sentences with names and essential predicates are analytic. Quine rejects this theory but Kripke finds it to be "a perfectly valid point of view" and holds that "once we have this notion of a name, it seems unexceptional". (eds.),

256

QUENTIN SMITH

[Marcus et aI., pp. 142-143]. F0llesdal, who had developed his own theory of modally stable and non-descriptional proper names a year earlier, did not misunderstand Marcus. See [F0llesdal, 1986: 105]. It seems that Kripke was under the influence or "seductive power" (as Kripke later put it [1980: 5]) of the then prevalent descriptional theory of names and thereby did not fully comprehend Marcus' theory that names pick out their objects non-descriptionally. More important than this misunderstanding is the positive theory that resulted, namely, Kripke's origination of the theory that names are de scriptionally rigid. Kripke's remarks in this 1962 discussion suggest that his 1980 recollection of when he first came to believe that rigid names are directly referential is better supported by the textual evidence than his 1996 recollection. In [1980: 5] Kripke says "the ideas in Naming and Necessity evolved in the early sixties most of the views were formulated in about 1963-64 .... Eventually I came to realize - this realization inaugurated the aforementioned work of 1963-64 that the received presuppositions against the necessity of identities between ordinary names were incorrect, that the natural intuition that the names of ordinary language are rigid designators can in fact be upheld. Part of the effort to make this clear involved the distinction between using a description to give a meaning and using it to fix a reference. Thus at this stage [in 1963-64] I rejected the conventional description theory as an account of meaning .... Let me not pay inadequate tribute to the power of the then prevailing complex of ideas, emanating from Frege and from Russell, that I then [in 1963-64] abandoned .... Although I, with others, always felt some strain in this edifice ["the description theory of proper names"], it took some time to get free of its seductive power". [1980: 3-5]. Over fifteen years later (see Holt, 1996: 36] Kripke seems to recollect that he was already in grasp of Marcus' "direct reference" theory of rigidity before Marcus' February 1962 talk: "Ruth [Marcus] said in her 1962 talk that proper names were not synonymous with descriptions. A subset of the ideas I later developed were present there in a sketchy way, but there was a real paucity of argumentation on natural language. Almost everything she was saying was already familiar to me at the time. I knew about Mill's theory of names and Russell's theory of logically proper names, and I hope that, having worked on the semantics of modal logic, I could have seen the consequences of such a position for modal logic myself". As Kripke acknowledges, it is hard to remember exactly what beliefs one held thirty years ago, and I think is safer to rely on textual evidence. I have argued that the 1980 recollection (in which Kripke says he first realized the Millian consequences of modal logic for ordinary names in 1963-64) is confirmed by the 1962 text I quoted and thus that an historical account of this period should accord with the 1980 recollection in this respect. If Kripke's 1980 recollection is correct and my interpretation of the 1962 discussion among him, Quine, F0llesdal and Marcus is correct, then Marcus' (eds.),

A MORE COMPREHENSIVE HISTORY

257

later interpretation [1993: 226-27] of Kripke's 1962 remarks is mistaken. We should instead conclude that in 1962 Kripke held the theory that names are descriptionally rigid. In at least one respect, this reverses the standard history of the New Theory of Reference; the standard history attributes the origin of the descriptionally rigid theory of names to Plantinga and the origin of the argument for the nonequivalence of the necessary/contingent and a priori/a posteriori distinction to Kripke. In fact, Kripke originated (at least in a sketchy form) the theory of names as descriptionally rigid and Plantinga originated the argument for the nonequivalence of the necessary/contingent and a priori/a posteriori distinction (see section 9). But there are still more ways in which the standard histories of contemporary philosophy need to be re-examined. 7. METAPHYSICAL POSSIBILITY IN HINTIKKA: 1957-63

The New Theory of Reference is a theory of natural language that includes the ideas that expressions in natural languages are informal analogues to certain terms in a semantics for quantified modal logic. The ideas of a rigid designator, of a statement that is true in all possible worlds, and of the distinction between logical and metaphysical possibility, belong to the New Theory of Reference and are analogues of certain notions in modal logic. What is metaphysical possibility? In this section, I understand this notion to include two ideas, one being the idea of what really might have been the case and the second being the idea of a material or nonformal component that belongs to the possibility operator (; in modal logic. The first idea involved in the notion of metaphysical possibility or necessity is that what might exist or what must exist is a distinct notion from what is logically possible or logically necessary. The primitive metaphysical notions of what might have been or could have existed (de re possibility) or what might have been true (de dicto possibility) are distinct notions from the notion of what does not involve a logical contradiction. Likewise, the ideas of what must be or what must be true are distinct ideas from the idea of something whose negation or denial involves a logical contradiction. If "real possibility", possibility in the sense of what might have been the case, is divorced from logical possibility, how are we to reason about it? What theory of axioms, rules of inference, etc., will capture its patterns of reasoning? The answer that many New Theorists of Reference adopt is that we employ the semantics for metaphysical necessity in one of the systems S5, S4, M, etc., of modal logic. This metaphysical semantics is based on the idea that the necessity and possibility operators in modal logic have a material or nonformal content, and in this respect differ from the formal modal operators in the semantics for logical necessity and possibility. There is an essential connection between the semantics of metaphysical necessity and the theory of rigid designators in natural language. A crucial question for the New Theory of Reference is: what sort of semantics for modal (eds.),

258

QUENTIN SMITH

logic is "the right one" for the individual constants that are the formal counterparts of rigid names in natural language? This semantics cannot be a semantics for logical necessity, such as Carnap's [1946; 1947]; the semantics of logical modalities cannot accomodate the notion that some singular terms are rigid designators. Suppose for purposes of reductio that a name is the natural language counterpart of an individual constant in a semantics for logical modalities. Consider the use of "Socrates" as a name of the Greek philosopher, Socrates; we may consider "Socrates" in this usage as a hononym. If names are counterparts of individual constants in a semantics for logical modalities, then the hononym "Socrates" would refer to Socrates as he exists in all logically possible worlds, and in many of these worlds he is a typewriter or a number or the color red, and such worlds are not the right sort of worlds for rigid designators in natural languages. For example, the sentence "Socrates might have been the color red" is inconsistent with the rules of use of names in natural languages (given that "Socrates" is the hononym I mentioned). If "Socrates" is a different hononym and is used by fabric sellers as a name of a special shade of red, then the sentence is consistent with the rules of use of names in natural languages. Names in natural languages are related to metaphysically possible worlds. The reason is that rigid designators go hand in hand with the notion of non-trivial essences (even if the designators are directly referential). When we refer to the Greek philosopher Socrates by the rigid name "Socrates", we refer to Socrates in each possible world in which he exists, but this means each metaphysically possible world, for we restrict the worlds at which "Socrates" has a referent to the worlds in which Socrates has his non-trivial essences, such as being human. A rule of use of "Socrates" and each other ordinary name is that it is used to refer to an object only insofar as it exists in metaphysically possible worlds. Hintikka's [1957a; 1957b; 1961; 1963] first offered the conceptual materials to formulate (but he himself did not formulate) this rule of use of names in terms of a semantics for modal logic, for he introduced metaphysically possible worlds as the relevant worlds for rigid designators. Hintikka introduced the idea that the possibility and necessity operators have a nonformal content in the semantics for modal logic, but did not explicitly relate this content to nontrivial essences. (It is worth emphasizing that the theory of rigid ordinary names, metaphysically necessary worlds, non-trivial essences, the necessity of identity, and other ideas that constitute the New Theory of Reference is different than the theory that Hintikka concentrated on developing, especially in his later writings. For references and a criticism of the Marcus~Kripke version of the New Theory of Reference, see Hintikka and Sandu [1995]. I shall here concentrate on Hintikka's 1957~1963 writings on the semantics for metaphysically possibility and necessity.) It is now known by some logicians that Kanger's [1957] is the first publication introducing into modal logic the idea that features of the alternativeness relation (also called the "accessibility" or "relative possibility" rei a(eds.),

A MORE COMPREHENSIVE HISTORY

259

tion), features such as reflexivity, symmetry and transitivity, can give us semantics for different modal logics, e.g., S4, S5 and Fey's calculus T [Kanger, 1957: 40; the relevant discussion begins on page 33]. Montague [1960] also gave an early and independent presentation of some partly similar ideas about S5. Montague presented his ideas in a May, 1955 conference at the University of California at Los Angeles [Montague, 1974: 71] and Kanger presented his ideas in Spring 1955 in a course on logic at the University of Stockholm (see [F011esdal, 1994: 886]). Hintikka independently introduced the accessibility or alternativeness relation in the same year as Kanger's publication (Hintikka, 1957a; 1957b]; Hintikka first introduced the idea of metaphysical necessity in [1957a; 1957b] and Hintikka was the first to develop (in a manuscript written prior to his 1957 articles; see [Hintikka, 1957a: 10] and the first to publish (in [1961]) a metaphysical semantics for modal logic in terms of the reflexive, transitive, etc., features of the alternativeness relation. The metaphysical alternativeness relation was introduced in Hintikka's [1957a; 1957b] and its features of reflexivity, transitivity, etc., were used to develop semantics for the systems M, S4, S5, etc., in [1961]. In his 1957 essay on "Quantifiers in Deontic Logic", which is the earliest system of quantified deontic logic, Hintikka writes: "My treatment [of deontic logic] derives from an earlier treatment of quantification theory along the same lines as well as from a similar (unpublished) theory of modal logic. Most of the formal considerations will turn out to be special cases of this new general theory of modal logic I have developed'. [1957a: 10]. (Presumably, Hintikka's unpublished theory of modal logic is the theory published in his [1961; 1963.) Hintikka first introduces in [1957a] the modal metaphysical notion of alternativeness and its deontic counterpart, copermissability. The metaphysical notion of alternativeness involves the idea that states of affairs or possible worlds are not possible absolutely (as Carnap [1946] and others assumed) but are possible relative to a certain world W, such that Wand the worlds possible relative to W constitute a subset of the set of all logically possible worlds. Necessary truth is not truth in all logically possible worlds (as Carnap [1946] held), but is truth in a world Wand in all the alternatives to W, such that Wand the alternative worlds are members of a subset of the set of all logically possible worlds. 2 (The subset may be an improper subset, in which case it would be the set of all logically possible worlds; but this is not required by the meaning of the possibility and necessity operators.) This introduces content into the (otherwise formal) interpretations of the box and diamond. The content enables a subset of the set of all the formally (logically) possible worlds to be demarcated from other subsets of the set of all formally possible worlds. What is necessarily true is not determined by what is true by virtue of its form, but what is true by virtue of a certain content. Other logicians had discussed relevantly similar nonlogical necessities, such as physical necessity, but they did not develop a semantics for these nonlogical necessities but characterized them in terms of logical necessity (eds.),

260

QUENTIN SMITH

and a conjunction of physical laws or (laws of some other sort). The characterization of "metaphysical necessity" in terms of logical necessity and a conjunction of laws of some sort is (at least implicitly) the sense in which Moore [1903] and Wittgenstein [1922] had the concept of "metaphysical necessity". Hintikka's second 1957 article, "Modality as Referential Modality", outlines the second system presented of quantified epistemic logic (later more fully given in [Hintikka, 1962]), the first being Von Wright's [1951]). But this article is also in large part about non-epistemic modal logic. One way the idea of a semantics for metaphysical modalities appears in Hintikka's [1957b] involves his explanation of how proofs can be carried out in modal logic. In the following passages, he makes some suggestions about the alternativeness relation in (non-epistemic) modal logic. He has indicated [Hintikka, 1955] how the notion of a model can be replaced by the notion of a model set of logical formulae, and Hintikka adds the idea: "It turns out that an intuitive and powerful theory of modal logic can be based on these notions [of a model or a model set of logical formulae]. The main novelty is that we have to consider several interrelated models (or model sets). They correspond to the different situations we want to consider in modal logic, and they are interconnected, in the first place, by a rule saying (roughly) that whatever is necessarily true in the actual state of affairs must be (simply) true in all the alternative states of affairs. It turns out that it suffices, for the interpretation of each given proof carried out by means of modal logic, to consider only a finite number of alternative models (model sets). Of course, no finite number wiJI suffice for the interpretation ofal1 the proofs" [1957b: 61-62].

Let us see exactly how this notion of metaphysical necessity differs from the notion of logical necessity given in Kanger's [1957] and Carnap [1946; 1947]. Kanger discusses models rather than model sets or state descriptions. In general terms, a model consists of a domain D of individuals and an interpretation of a language L in terms of that domain. Individual variables in L are assigned to individuals in D and predicates in L are assigned to n-tuples of individuals in D. By contrast, a model set is a set of logical formulae that partially describe a possible world; in this respect, model sets are somewhat analogous to Carnap's [1946; 1947] state descriptions, which Carnap says are complete descriptions of a possible world. In Kanger's version of a model-theoretic semantics for logical modalities, < D, V> is a system where D is a domain of individuals and V a primary valuation for a language L of quantified modal logic. V is a function which for every domain D' assigns an extension in D' to each individual variable, individual constant and predicate constant in the language L. Logical necessity is characterized as follows: Dp is true in the system S if and only if p is true in every system S' (where each system is an alternative to each other system). (See [Lindstrom, 1998] for a more detailed discussion of Kanger's theory.) This differs from Carnap's [1946; 1947] theory of logical necessity in that Carnap discussed state descriptions rather than models and said a statement is logically necessary if and only if it is true in all state descriptions; Carnap did not have (eds.),

A MORE COMPREHENSIVE HISTORY

261

the notion of accessibility or alternativeness and thus did not specify that logical necessity requires each state description to be an alternative to each other state description. It is easy to see how Hintikka's metaphysical theory differs from Kanger's and Carnap's theories of logical necessity. For Hintikka in [1957a; 1957b], a model set is a set of logical formulae that partially describe a possible world. Necessity is truth in a model set, such that Op is true in the model set M if and only if p is true in M and in every model set that is a real alternative to M. But not every model set need be a real alternative to M and thus 0 is not required to quantify over all model sets. Op requires merely that p is true in a subset of the set of all logically possible model sets. This imports material content into 0 and thus 0 now means a non formal, metaphysical necessity. The second appearance of metaphysical necessity is in Kripke's [1959a], where Kripke uses models rather than model sets. Kripke uses a metaphysical semantics to provide a completeness proof for quantified S5 and this proof requires a metaphysical rather than logical semantics. (This requirement was first demonstrated by Cocchiarella [1975b]; in this work Cocchiarella also first demonstrated that the logical ["primary"] semantics for quantified modal logic is incomplete ~ an issue left unresolved in Carnap [1946].) Kanger [1957] introduced models into theories of modal logic but Kripke [1959a] is the first to use metaphysical models. In Kripke's [1959a], he mentions a complete assignment for any formula A in a domain D of individuals. A is any formula that contains free individual variables, propositional variables and n-adic predicate variables. A model of A in the domain D is a pair (G, K), where G is a complete assignment for A in D and K is a set of complete assignments for A in D, such that G is a member of K and all members of K differ from G only in the assignments for the propositional and predicate variables in A. In Kripke's semantics for necessity, necessity is characterized in terms of some set K of complete assignments, where K is not required to be the one and only set C of all complete assignments. This implies that necessary truth does not mean truth in all logically possible worlds, but truth in all the members of some subset of the set of all logically possible worlds. (The subset may be an improper subset of the set C of all complete assignments, i.e., C itself, but the characterization of necessity does not require that the subset be this improper subset.) The first step in the metaphysical semantics for modal logic occurred in 1957 with Hintikka's [1957a; 1957b] and the second in Kripke's [1959a]. The next significant step is Hintikka's 1961 article on "Modality and Quantification". This 1961 article was the first article to present a metaphysical theory of possible worlds characterized in terms of the reflexive, transitive, symmetrical, equivalence and connected features that can be possessed by the alternativeness relation. Tn other words, Hintikka [1961] provided the first metaphysical semantics for standard systems of modal logic, e.g., the systems M, S4 and S5, that used the reflexive, transitive or symmetrical properties that can be (eds.),

262

QUENTIN SMITH

possessed by the alternativeness relation. (Kanger's [1957] semantics for several systems was a logical semantics and Kripke's metaphysical semantics for these and other systems appeared in [1963b], but with a [l959b] abstract.) Hintikka proceeds by "saying that a model set is the formal counterpart to a partial description of a possible state of affairs (of a 'possible world'). It is, however, large enough a description to make sure that the state of affairs in question is really possible. For it is natural to say that a set of sentences is satisfiable if and only if it can be imbedded in a (partial or exhaustive) description of a possible state of affairs; and this is just what we demonstrated if model sets are interpreted as such descriptions" [1961: 122]. In modal logic, we do not consider just one model set at a time (e.g., one partially describing the actual world), but a set of model sets. Hintikka calls such sets of sets model systems. More fully, a model system < U, R > is a couple where U is a set of model sets each of which satisfies a condition (stated informally) that if necessarily p is true in a model set (partial description of a possible world), then p is true in that model set. The relation R of alternativeness satisfies the condition that if possibly p is true in a model set U that belongs to a model system S, there is an alternative set v in S in which p is true. The relation of alternativeness also satisfies the condition that if necessarily p is true in a model set u that belongs to the model system S, then p is true in every model set in S that is an alternative to u. (These last two conditions are also satisfied by the model system S.) But not every model set in the model system S to which u belongs need be an alternative to u. Given these conditions, we have a metaphysical modal semantics for the Von Wright system of modal logic M. If we add that the relation of alternativeness is transitive, we have a metaphysical modal semantics for S4, and if symmetry is also added, we have a semantics for S5. But in each of these cases the model system, in which is included each alternative to a given model set, is not required to include all logically possible model sets. The necessity of a formula p is p's truth in all the alternative model sets in a model system, and thus necessity is not characterized as a logical necessity. Further or different conditions need to be used to characterize the necessity of a formula if we adopt a varying domain semantics. In this case, if an individual exists in the actual world and necessarily has a property F, it does not follow that the individual exists in every alternative world and has F in these worlds. Rather, it follows that if the individual exists in some alternative world, then the individual possesses F in that world. This requires the following condition be added to the characterization of metaphysical necessity. In Hintikka's logic, the presence of a free variable in a formula p that belongs to a model set u is the formal counterpart to the existence of its value in the possible world partially described by u. Accordingly, if Dp, it does not follow that p is true in every model set that is an alternative to u. We cannot transfer a formula p from a model set u to an alternative model set v unless the values of the variables in p exist in the possible world partially described by v. If Dp, (eds.),

A MORE COMPREHENSIVE HISTORY

263

then p is true only in alternative model sets where the free individual variables in p are contained in formulae in these alternative model sets. See {CN. *} in Hintikka [1961: 125]. How does this notion of metaphysical necessity connect with the notion of a rigid designator? The theory of nonrigid singular terms (typically, contingent definite descriptions) and rigid individual constants or names is developed in Hintikka's most well-known early article on modal logic, his "The Modes of Modality" [1963] presented at an August 1962 conference in Helsinki in which Marcus, Kripke, Montague, Prior and Geach participated (see Marcus [1993: 89]). It is worth quoting the passage in which the "rigid individual constant or name versus nonrigid definition description" distinction is made, since this is the most explicit appearance of this distinction after Marcus' February, 1962 talk. This distinction appears in Hintikka's example of the contingent description "the number of planets" as contrasted with the modally stable name "9". Hintikka refers to his [1962: 138-158], F011esdal [1961] and Quine's many writings on the topic as background material. Hintikka writes: "Why do some terms fail in modal contexts to have the kind of unique reference which is a prerequisite for being a substitution-value of a bound variable? An answer is implicit in our method of dealing with modal logic. Why does the term "the number of planets" in (i) ["the number of planets is nine but it is possible that it should be larger than ten"] fail to specify a well-defined individual? Obviously, because in the different states of affairs which we consider possible when we assert (i) it will refer to different numbers. (In the actual state of affairs it refers to 9, but we are also implicitly considering other states of affairs in which it refers to larger numbers.) This at once suggests an answer to the question as to when a singular term (say a) really specifies a well-defined individual and therefore qualifies as an admissable substitution-value of the bound variables. It does so ifand only ifit refers to one and the same individual not only in the actual world (or, more generally, in whatever possible worlds we are considering) but also in all the alternative worlds which could have been realized instead of it; in other words, if and only if there is an individual to which it refers in all the alternative worlds as well. But referring to it in all these alternatives is tantamount to referring to it necessarily. Hence (Ex)N(x = a) formulates a necessary and sufficient condition for the term a to refer to a well-defined individual in the sense that critics of quantified modal logic have been driving at, exactly as I suggested."

(Ex)N(x = a) is the material mode definition of a rigid designator. Note the two ways of using "a". In "the term a" it refers to a rigid designator and in "(Ex)N(x = a)" it refers to the individual that is the rigid designatum of the term. This dual usage has become standard in discussions of modal logic. Well-defined reference is a stable reference. It is a broader category than (non-epistemic) modally stable reference, for Hintikka notes that well-defined reference breaks down for "even proper names" [1963: 73] in epistemic contexts. (Unlike Marcus' [1961] and F011esdal's [1961] theories, Hintikka's theory rejects the universal intersubstitutivity thesis.) Hintikka's point is that an individual constant or proper name may have a stable reference in modal contexts (and thus be a "rigid designator") but fail to have a stable reference in epistemic contexts. Contingent definite descriptions, by contrast, are always unstable in modal contexts. The above history of the metaphysical semantics for modal logic ill accords (eds.),

264

QUENTIN SMITH

with the standard history of metaphysical semantics and some defence is needed. According to the standard history, a modal semantics for logical necessity was originally developed by Carnap in 1946 and 1946 (which is true), and the modal semantics for metaphysical necessity was first developed in Kripke's 1963 work on modal logic (which is false). One of the most precise and best statements of the standard history of metaphysical necessity appears in Joseph Almog's [1986]. He writes: "But where did metaphysical possibility and necessity come from? It came from Kripke's 1963 paper" [1986: 217]. Almog's argument for this is that in Kripke's [1963a: 69], he characterizes worlds as primitive "points" that "may include a world, a time, an agent, a spatial location, or what have you" [Almog, 1986: 218], and does not characterize them in terms of complete noncontradictory assignments of extensions to a given language. (According to Almog, Kripke [1963b] introduced into modal logic the idea of varying domains; but varying domains are present in Hintikka [1957a; 1957b; 1961; 1963] and Kanger [1957].) Almog writes about Carnap's "... logically possible worlds. Really, they were 'state descriptions' or, later, models, but still language-bound, i.e., complete assignments of extensions to a given language. On this conception, possibility obeys a maximum principle: any noncontradictory assignment is possible. This view was not confined to Carnap. It was also the view of Stig Kanger, Jaakko Hintikka, Richard Montague (of that period) - and the 1959 Kripke. Indeed, Kripke had at that time nothing more than 'complete assignments' and the modality he worked with was definitely logical possibility". [Almog, 1986: 217]. I shall confine myself to some of the mis-interpretations of Hintikka and Kripke that appear in this passage. First, Hintikka was the first person to introduce, in [1957a; 1957b; 1961; 1963], the idea that possibility does not obey a maximum principle, i.e., it is not the case that any noncontradictory assignment is possible. Hintikka's notion of possibility requires merely that, for any given world, only some noncontradictory assignments are possible (relative to that world). More precisely (and using Hintikka's terminology), a model system is a set of model sets and a model system need not include all model sets. If a formula is possibly true in a model set U, this requires only that the formula is true in another model set v that both belongs to the same model system as U and that is an alternative to u. For example, if we suppose u partially describes the actual world, we (as members of the actual world) may say that tnere are many noncontradictory formulae that are not possible, since these formulae either are not (i) elements in any maximal model set in which u is embedded and that maximally describes the actual world, or are not (ii) elements. of a, model set that is an alternative to u and that belongs to the same modef system as u. This characterization of possibility is based on Hintikka's condition (C.M.*) in [1961: 123] and [1963: 67]. As Hintikka intuitively puts it, "we have assumed that not every possible world (say P) is really an alternative to a given possible world (say Q) in the sense that P could have been realized instead of Q". [1963: 67]. Thus, Hintikka (eds.),

A MORE COMPREHENSIVE HISTORY

265

rejects the Carnapian idea that possibility obeys a maximum principle. A second respect in which Almog's interpretation of Hintikka is mistaken is that Almog thinks that Hintikka's possible worlds are linguistic items, i.e., are descriptions. This is not so. Hintikka's linguistic items are descriptions of a possible world, they are not the possible worlds themselves. A model set (m.s.) is a description: a "m. s. may be thought of as a partial description of a possible state of affairs or a possible course of events (,possible world'). Although partial, these descriptions are large enough to show that the described state of affairs are really possible ..." [1963: 66]. Also see [1961: 122]. Each model set can be embedded in a maximal model set and "each maximal model set is an extended state-description" [1961: 121]; it is not the possible world described by the extended state-description. A third respect in which Almog's interpretation is inaccurate is that Hintikka does not work with models, but with model sets. A fourth respect in which Almog's theory is problematic is that the definition of a world as a "primitive point", rather than as a model (or as what is described by a maximal model set), is a sufficient but unnecessary condition for the possibility and necessary operators to have nonformal, metaphysical content. Another sufficient condition for this nonformal content is that the box or necessity operator be interpreted in terms of all the members of a set of logically possible worlds, rather in terms of all the members of the one and only set of all logically possible worlds. In the specifics of Hintikka's logic, this nonformal content appears (among other places) in the condition (C.N+) [1961: 123; 1963: 67], which says that if a formula is necessarily true in a model set U, then the formula is true in each model set that is an alternative to U in a model system to which u belongs. Kripke first published the notion of metaphysical necessity in [1959a], not (as Almog maintains) in 1963. In the specifics of Kripke's semantics [1959a], the necessity operator is interpreted in terms of a set K of complete assignments, not the one and only set C of all complete assignments. Kripke [1959a] gave a metaphysical semantics for Lewis' quantified S5 but not based on the alternativeness relation. Kripke's metaphysical semantics that was based on features of the alternativeness relation appeared in Kripke [1963a; 1963b]. Kripke had developed the basic ideas of his 1963 semantics at least as early as Spring 1961 (see [F011esdal, 1966: vD. He also developed a metaphysical semantics for S2 and S3 at least as early as 1961 (see Hintikka, 1961: 124). Kripke's [1959b] abstract for his [1963b] may plausibly be read as implying the metaphysical alternativeness relation with its reflexive, etc., features, but it is too brief to draw definite conclusions about its implications. Kripke says in the first footnote to his [1963b] that Hintikka's and Kanger's theories have the closest points of contact with his own theory and that his treatment of quantification draws some inspiration from Hintikka's and Prior's methods. But Kripke's idea of a metaphysical alternativeness relation plausibly seems to have been formulated independently. F011esdal mentioned to Kripke in 1961 that Hintik(eds.),

266

QUENTIN SMITH

ka [1957a; 1957b] and Kanger [1957] contained the notion of the a1ternativeness relation (private communication from F011esda1, April 23, 1997). Kripke read [Kanger, 1957], which contained the notion of a logical alternativeness relation, after January 1958 and before the end of 1960; there is a letter from Kripke to Kanger, dated January 24, 1958, in which Kripke asks Kanger for a copy of his [1957] (see [F011esdal, 1994: 887-888, n. 61], and Kripke reported to F011esdal in 1960 that he studied Kanger but that his study of Kanger was not "thorough" (see [F011esdal, 1994: 888, n. 7]). The alternativeness relation does not appear in Kripke's paper [1959a], but may plausibly be read as implied by his abstract [1959b]. The paper [1959a] was received by The Journal of Symbolic Logic in August, 1958 and the abstract [1959b] was received by this journal in October, 1959. This may suggest that in August, 1958 Kripke did not have the notion of the alternativeness relation, but in October 1959, almost two years after his letter requesting Kanger's [1957], Kripke was in possession of this notion. In his abstract [1959b] Kripke does not mention Kanger but does mention similar work by Hintikka: "(For systems based on 84, 85, and M, similar work has been done independently and at an earlier date by K.J.J. Hintikka.)" [Kripke, 1958b: 324]. Given all of the above-mentioned facts, it may be conjectured that Kripke acquired the notion of the alternativeness relation some time after the summer of 1958 (when he submitted his [1959a]) and before the fall of 1959 (when he submitted his [1959b]). If Kripke's notion of the alternativeness relation was acquired from earlier work, it seems more likely it was acquired from Hintikka's than Kanger's work. This is suggested by the fact that neither Hintikka nor Kanger are mentioned in [Kripke, 1959a] and only Hintikka is mentioned in [1959b] and is said to have done similar work. However, there is no good reason to think that Kripke did not acquire this notion independently. F011esdal, who had extensive discussions with his fellow student Kripke in 1960-1961, is in a position to reliably report that "Saul Kripke got the idea [of the alternativeness relation] independently [of Kanger and Hintikka] and extended it (in 1959) so as to make it applicable to the Lewis systems S2 and S3. (Cf. his "8emantical analysis of modal logic (abstract) [1959] ..." [F011esdal, 1966: 43, n. 1]. Given all of this evidence, the most probable hypothesis is that Kripke and Hintikka came up with the idea of a metaphysical alternativeness relation independently, but that Hintikka developed the theory first (prior to his 1957 publications) and published it first [Hintikka 1957a; 1957b]. Kanger [1957] and Montague [1960], independently of each other, came up with the idea of a logical, but not metaphysical, alternativeness relation, which they originated in 1955. 8. COCCHIARELLA AND THE SECONDARY SEMANTICS FOR LOGICAL NECESSITY

The notion of metaphysical necessity includes two ideas, the primitive idea of what must be and the idea that the necessity operator is nonformal or includes (eds.),

A MORE COMPREHENSIVE HISTORY

267

a material content. The idea of a nonformal necessity and possibility that is present in Kripke's writings on modal logic was first demonstrated to be different from the notion of logical necessity and possibility in Nino Cocchiarella's 1973 talk at the University of North Carolina at Chapel Hill (published as [1975a]). 3 Especially see Cocchiarella [1975b], and also see Hintikka's 1977 talk in Rome [Hintikka, 1980] and Hintikka's [1982]. This demonstration was made in terms of models, rather than model sets, and involved showing the equivalence of some early writings on modal logic to a primary semantics for logical necessity and the equivalence of some other writings to a secondary semantics for logical necessity. Note that it is not requisite to demonstrate this difference within one's modal logic in order to present a metaphysical or logical semantics. Some authors in the 1940s, 1950s and 1960s worked with logical modalities (e.g., Carnap [1946], Kanger [1957], Montague [1960], Beth [1960], Kaplan [1964]; but Montague [1963] and Kaplan [1978a; 1978b; 1989a] later used metaphysical modalities). Other writings on modal logic worked instead with metaphysical modalities (Hintikka [1957a; 1957b; 1961; 1963] and Kripke [1959a; 1963a; 1963b]; but in [Hintikka, 1980; 1982] and later works Hintikka discusses logical modalities). As I indicated, Nino Cocchiarella [1975a; 1975b] was the first to show that metaphysically possible worlds defined in terms of models relate to the distinction between the secondary semantics for logical necessity and the primary semantics for logical necessity. Cocchiarella argued this distinction involves the notion of a model (e.g., as in Kripke's [1959a)) and Cocchiarella did not discuss model sets or model systems (the notions Hintikka used). (Hintikka also discussed his related distinction, between standard and nonstandard logics, in terms of models rather than model sets [Hintikka, 1980; 1982].) Cocchiarella showed that Kripke's [1959a] is equivalent to a secondary semantics for logical necessity and thereby is a semantics for metaphysical necessity. Roughly speaking, this means that "all possible worlds" in Kripke's semantics are a subset of the set of all logically possible worlds and thus that we permit "modal operators to range (in their semantic clauses) over arbitrary non-empty subsets of the set of all the possible worlds (models) based upon the given universe of objects and the set of predicates in question ... the exclusion of some of the worlds (models) ofa logical space, imports material conditions into the semantics of modal operators". [Cocchiarella, 1975b: 13]. Cocchiarella points out that Carnap [1946], Kanger [1957] and Montague [1960] developed a primary semantics for logical necessity, but that since the early 1960s most modal logicians concentrated on the secondary semantics. (But this is not to say these logicians were aware of this primary/secondary distinction; Cocchiarella was the first to note this distinction in 1973 and this distinction is not often noted today.) Cocchiarella does not say that there is a primitive idea of what must be that is included in the notion of metaphysical necessity and in this respect (among others) his account differs from the one T am presenting. One significant (eds.),

268

QUENTIN SMITH

difference is that Cocchiarella made his metaphysical/logical distinction in terms of models and thus his distinction does not correspond to the metaphysical/logical distinction I made, where it is indifferent whether models or model sets are used. A model set is a set of logical formulae but a model is a structure such as < D, R >, where D is a non-empty set (a domain of objects) and R an interpretation of the relevant terms in the language with respect to D. This difference can be explained more exactly in Cocchiarella's terms. Cocchiarella's primary semantics includes a satisfaction clause for necessity that covers all the models for a domain D. Cocchiarella calls them L-models in [1984: 311], where L = the set of predicate constants that are provided an interpretation for the language in question. An L-model is a structure of the form < D, R >, where the domain D is a set of objects and R is an interpretation of the predicate constants in the language. For each predicate F in L, R(F) is a set of n-tuples of members of D. Accordingly, the clause for the necessity operator interprets this operator as ranging over all the L-models having D as their domain, such that the predicates in the language are assigned all possible reinterpretations (extensions) drawn from the domain D. (See clause 6 in Cocchiarella's [1984: 312] for a more technical formulation. His primary semantics can also be applied to a varying domain [1984: 321-323]. Cocchiarella's secondary semantics includes a satisfaction clause for necessity that is "cut-down" so as to cover, not all the L-models having D as their domain, but all the L-models in a set K of L-models having D as their domain, where K need not include all the L-models based on the domain D. Since the meaning of the satisfaction clause for necessity does not include all the Lmodels for D, the necessity is not a logical necessity but a necessity with some nonformal content, a metaphysical necessity. (The satisfaction clause is stated in [1984: 315]). Cocchiarella argues that this secondary semantics is equivalent to the semantics in Kripke [1959a]. Hintikka's notion of a model set does not allow the contrast Cocchiarella made between the primary and secondary semantics for logical necessity. This contrast is most naturally seen in a theory of models, and for this reason Hintikka's later discussion [1980; 1982] of Kripke's and others' metaphysical semantics in terms of the primary/secondary semantics distinction used the notion ofa model. However, I believe Hintikka's early semantics (1957-1963) was a secondary semantics and that one may formulate the primary/secondary distinction using only the concepts in Hintikka's early writings. Each model set can be embedded in a maximal model set and a maximal model set is an extended state description [1961: 121]. Although Hintikka does not say this, I think we can take all the maximal model sets with respect to a language Land characterize logical necessity in terms of the set of all the maximal model sets with respect to L. This would pertain to the primary semantics for logical necessity and the secondary semantics for logical necessity would be a "cut down" in the sense that necessity is instead analyzed in terms of all the members of a subset of the set of all the maximal model sets with respect to L. (eds.),

A MORE COMPREHENSIVE HISTORY

269

But we must not think that the distinction between the primary and secondary semantics must be formulated within Hintikka's modal logic in order for it to be the case that his semantics for modal logic is a secondary semantics. Even if Hintikka's early modal logic (1957-1963) does not permit the formulation of a primary semantics for logical necessity, it is still the case that his early modal logic is a secondary semantics. More generally, we must not think that his semantics must be a secondary semantics (if a secondary semantics is held to essentially involve models) in order for his semantics to be a metaphysical semantics. It suffices that the necessary operator in his modal logic is not analyzed in terms of the one and only set of all logical possibilities, and we have seen in our above discussion that the necessary operator in his semantics does not quantify over all logical possibilities but a subset of them. Modal logicians in the 1950s and 1960s were not aware of such metaphysical/logical distinctions. At this time, neither Hintikka nor Kripke noted this distinction (even though their semantics were metaphysical rather than logical semantics). The main element in this distinction (insofar as it pertains to a primary and secondary semantics involving models) was first noted by Cocchiarella in 1973, as I indicated above. Cocchiarella's dissertation [1966], which was finished in 1965, discussed the problem of the nature of the material content in secondary semantics. In his dissertation, he showed how this content can be explained if we adopt a temporal interpretation of modality, where the alternativeness relation is interpreted in terms of temporal relations, such as a causal signalling relation between different local times (as part of Einstein's Special Theory of Relativity). (A.N. Prior discussed a temporal interpretation of modal notions, but not in terms of the primary/secondary semantics distinction.) Cocchiarella subsequently developed a logic of natural kinds with the view in mind of interpreting the material content in terms of essences (see Cocchiarella [1987] for references). Hintikka first discussed the primary/secondary semantics distinction (or rather, a relevantly similar "standard/nonstandard" distinction) in a 1977 talk in Rome, later published as his [1980]. (For the reference to the Rome talk, see Hintikka's [1982: 89]. Note that the "early Hintikka semantics" I analyzed is different than Hintikka's later semantics. In Hintikka's later writings on modal logic, e.g., his [1980; 1982; 1986], one of his aims is to show that the secondary or "nonstandard" semantics for logical modalities is not the proper semantics for logical modalities. Hintikka has also raised serious doubts about whether an adequate semantics for logical modalities can be developed (e.g., [Hintikka, 1982]). Furthermore, Hintikka now rejects the notion of a metaphysical necessity inasmuch as this is understood as something narrower than logical necessity and broader than physical necessity [Hintikka and Sandu, 1995]. Contrary to what many writers say, Kripke has evinced no awareness of a logical/metaphysical distinction in his writings. Even Cocchiarella [1984: 317] is not strictly accurate when he writes: "Kripke himself, it should be noted, (eds.),

270

QUENTIN SMITH

speaks of the necessity of his semantics not as a formal or logical necessity, but as a metaphysical necessity (cp. Kripke [1971: ISO]". It is true that Kripke speaks of his necessity as a metaphysical necessity. However, Kripke explicitly and emphatically identified logical with metaphysical necessity, as I demonstrated at length in my [1998a] with extensive quotations from Kripke's [1971] and [1972]. I think Cocchiarella's demonstration that Kripke's semantics is not a semantics for logical necessity is sound, but Kripke at no time has expressed an awareness of the distinction or an affirmation that his semantics is a secondary semantics for logical necessity. I have suggested that the nonformal element in the interpretation of the box and diamond should be understood in terms of the non-trivial essences of objects. Many New Theorists of Reference, however, are in a position similar to Kripke in not taking a stance about (or mentioning) the primary and secondary semantics for logical necessity and its relation to non-trivial essentialism. Marcus [1961; 1967; 1971] has argued that neither a Carnapian semantics nor a Kripkean semantics for modal logic includes any non-trivial essential attributions among its theorems (but without mentioning a primary/ secondary semantics distinction). At the opposite end, Plantinga [1974: 241-S1] has argued that the semantics for modal logic (what Plantinga calls "applied semantics") that Marcus considers as well as Kripke's semantics include nontrivial essential attributions among their theorems (but without mentioning a primary/secondary semantics distinction). Notwithstanding this difference, both Marcus [1961; 1967; 1971] and Plantinga [1967; 1969; 1970; 1974] believe objects have non-trivial essences. What this goes to show is the fundamental importance of Cocchiarella's [197Sa; 1975b] for the New Theory of Reference, for in these publications we see for the first time the recognition and argument that systems of modal logic that include non-trivial essentialism (or more precisely, a nonformal content in the systems using models) in their semantic characterizations of possibility and necessity are secondary semantics oflogical necessity. It also shows that we can interpret the New Theorists of Reference in this way but this does not imply that this is how all (or even most) New Theorists of Reference interpret their own work. It is not requisite for New Theorists to interpret their own work in this way in order for this interpretation to be true. (The evaluation of Marcus' theory that neither Carnapian nor Kripkean semantics include non-trivial essential attributions as theorems, and the evaluation of Plantinga's theory that the "applied semantics" of both Marcusian and Kripkean semantics include non-trivial essential attributions, is too large a task to undertake here. It suffices to note that there is no uniform agreement of New Theorists of Reference about the relation of non-trivial essentialism to modal logic.) Cocchiarella himself writes of the secondary semantics for logical necessity that there is a difficulty or unsolved problem with its "objective, as opposed to its merely formal, significance" [1984: 310]. This is tantamount to the problem of the meaning or truth of non-trivial essentialism. (eds.),

A MORE COMPREHENSIVE HISTORY

271

Cocchiarella [1984] believes that the introduction of a secondary semantics for logical necessity, which he believes was first made in Kripke's 1[959a], provided the semantical framework for the later theory that some "logical necessities" (in the sense of the secondary semantics) are knowable a posteriori. Cocchiarella shares with others the belief that the theory of a posteriori necessities was first published in [Kripke, 1971; 1972]. I shall now leave the issue of the primary/secondary semantics distinction, a topic in modal logic, and turn to the issue of the distinction between necessity and a prioricity, which is a topic in the disciplines of "epistemology" and "metaphysics". 9. PLANTINGA'S DISTINCTION BETWEEN THE EPISTEMIC NOTIONS OF A PRIORI/A POSTERIORI AND THE METAPHYSICAL NOTIONS OF NECESSITY ICONTINGENCY

Who first made (in the context of the New Theory of Reference) the explicit distinction between the epistemic notion of a priori and the metaphysical notion of necessity? (This distinction is logically implied by Marcus [1961; 1963], as I argue in [Smith 1995a; 1995b; 1998a] but we are here interested in the first explicit statement of this distinction.) It is a "commonplace" that this explicit distinction was first published in Kripke's [1971] and [1972]. This distinction is now even called "Kripke's Thesis" by some and an a posteriori necessity is called by some a "Kripkean necessity". For example, D.M. Armstrong talks as follows about a question involving a necessary a posteriori truth: "it may be a question to be decided a posteriori to the extent that it can be decided. But it is not a contingent matter. It is what might be called a Kripkean necessity". [Armstrong, 1989: 67]. Are these commonplaces about the history of contemporary analytic philosophy accurate? Would it be more accurate to call an a posteriori necessity a Plantingean necessity, and the relevant thesis Plantinga's Thesis, if one is interested in attributing an idea to its originator? (This concerns the context of the New Theory of Reference; the theory is already present in Moore [1899] and can be traced back to Aristotle and Plato. Perhaps we should call the thesis "Moore's thesis" or "Plato's thesis".) First let us quote from Kripke's [1971]; Kripke said we should "distinguish between the notions of a posteriori and a priori truth on the one hand, and contingent and necessary truth on the other hand" [1971: 152-53]. Kripke wrote in a well-known passage that "Hesperus is Phosphorus" is both empirical and necessary: "I thus agree with Quine, that 'Hesperus is Phosphorus' is (or can be) an empirical discovery; with Marcus, that it is necessary. Both Quine and Marcus, according to the present standpoint, err in identifying the epistemological and metaphysical issues" (1971: 154, n. 13). Kripke's [1971] and [1972] are held by some to have "astonished" the philosophical community with the discovery of this distinction, but what is astonishing, however, is not this alleged discovery but the pervasiveness of the (eds.),

272

QUENTIN SMITH

oversight that Plantinga earlier published this same distinction in an article that was hard to miss ("World and Essence", in the October, 1970 issue of The Philosophical Review).

Kripke presented his lectures on "Naming and Necessity" at Princeton University in January 1970 and Plantinga presented his paper (in final draft form) on "World and Essence" in October 1969 at Cornell University. The ideas in the final draft were worked out by Plantinga in 1968~69, when he was a fellow in the Center for Advanced Study in the Behavioral Sciences in Palo Alto.4 Here is Plantinga's statement (the third sentence, "And hence ..." states the traditional view he is refuting): "Consider the well-known facts that Cicero is identical with Tully and that Hesperus is the very same thing as Phosphorus. Do not these facts respectively represent (for many of us, at least) historical and astronomical discoveries? And hence are not the counterfactuals Hesperus and Phosphorus are entities distinct and Tully is diverse from Cicero, though counterfacts indeed, contingently counterfactual? [Plantinga says the answer to this last question should be NO; they are necessarily false. He says the traditional thesis that they are contingently counterfactual is false, for the following reason (in this passage, "Kronos" is the name of the actual world):] "The argument here implicit [that if a truth is a posteriori, it is contingent] takes for granted that the discovery of necessary truth is not the proper business of the historian and astronomer. But this is at best dubious. I discover that Ephialtes was a traiter; I know that it is Kronos that is actual; accordingly, I also discover that Kronos includes the state of affairs consisting in Ephialtes' being a traitor. This last, of course, is necessarily true; but couldn't a historian (qua, as they say, historian) discover it, too? It is hard to believe that historians and astronomers are subject to a general prohibition against the discovery of necessary truth. Their views, if properly come by, are a posteriori; that they are also contingent does not follow." ]1970: 480-481].

Plantinga proceeds to discuss a posteriori necessary identities, such as Hesperus is Phosphorus. "Exactly what was it that the ancient Babylonians discovered? Was it that the planet Hesperus has the property of being identical with Phosphorus? ... [Before the discovery] the Babylonians probably believed what can be expressed by pointing in the evening to the western sky, to Venus, and saying 'This is not identical with' (long pause) 'that' (pointing to the eastern sky, to Venus, the following morning). If so, then they believed of Hesperus and Phosphorus-identity that the latter does not characterize the former; since Phosphorus-identity is the same property as Hesperus-identity, they believed of Hesperus-identity that it does not characterize Hesperus. ... The quality of their intellectual life was improved by the Discovery in that they no longer believed of Hesperus that it lacked the property of Hesperus-identity.... Still, this is at best a partial account of what they discovered. For they also believed that there is a heavenly body that appears first in the evening, and another, distinct from the first, that disappears last in the morning. This is a contingent proposition; and part of what they discovered is that it is false .... And of course this is a contingent fact; there are possible worlds in which the thing [Venus] that in fact has the distinction of satisfying both sets of criteria [appearing first in the evening and disappearing last in the morning] satisifes only one or neither." [1970: 481-482].

Plantinga includes identities as well as nontrivial essential attributions to be among the necessities that are a posteriori. It is perhaps symbolic of the neglect of Plantinga's original contributions to the New Theory of Reference that his own later [1974: 81~87l account of a (eds.),

A MORE COMPREHENSIVE HISTORY

273

posteriori necessities is widely regarded as a restatement and defense of "Kripke's Thesis" (e.g., Wong [1996: 53]. "Kripke's Thesis" is that identity statements such as "Hesperus is Phosphorus" are necessary a posteriori. Wong notes "Plantinga makes the same claim as Kripke about the modal status of 'Hesperus is Phosphorus' " [1996: n. 24]. But Wong is not aware that Plantinga made this claim prior to Kripke in Plantinga's "World and Essence"; Wong regards Plantinga's claim (later stated in [Plantinga, 1974: 81-87] as a defence of Kripke's statement of this Thesis [Kripke, 1972], which is supposed to be where Plantinga and everybody else got the Thesis. Thus Wong says "I will now outline a fairly common way of defending Kripke's [sic] Thesis, as suggested by Alvin Plantinga [1974: 81-87]" [Wong, 1996: see his discussion of his sentence numbered (6)]. Plantinga's [1970] appears nowhere in Wong's very extensive bibliography. The virtually universal belief that Plantinga's statement of the Thesis in [1970] does not exist and that the Thesis was first stated in Kripke [1971, 1972] is one of the main beliefs that needs to be changed if an accurate account of the genesis of the ideas associated with the New Theory of Reference is to replace the "standard account" that remains prevalent in some quarters. Plantinga's original presentation of the theory of a posteriori necessities takes place in the context of a theory of metaphysical necessity. Plantinga has in mind metaphysically necessary truth when he says that in some cases people can make an empirical "discovery of necessary truth. Their views, if properly come by, are a posteriori; that they are also contingent does not follow". [1970: 481]. To see this, note first of all Plantinga's discussion of nontrivial essential properties. Plantinga writes: "Are there any nontrivial essential properties? Certainly; the number six has the properties of being an integer, being a number, and being an abundant number". [1970: 465]. Socrates "also has essentially some properties not had by everything: being a non-number and being possibly conscious are examples". [1970: 473]. Further, the correct way of explaining Socrates' trivial and nontrivial essences' is "to explain Socrates' essence and essential properties by means of properties he has in every world in which he exists" [1970: 477]. Plantinga is here talking about metaphysically possible worlds ("metaphysical possibility" is here used interchangeably with "broadly logical possibility"; by the time of his [1974], Plantinga uses almost exclusively the latter phrase). Here is the first mention of metaphysical possibilities in the articles or books published in the 1970s: "Recall that a possible world is a state of affairs that could have obtained if it does not. Here 'could have' expresses, broadly speaking, logical or metaphysical possibility.... If a state of affairs S is possible in at least one world W, then S is possible in every world. This principle may be false where it is causal or natural possibility that is at stake; for logical or metaphysical possibility, it seems clearly true." [1970: 475).

There is another and related feature of Plantinga's philosophy of language (eds.),

274

QUENTIN SMITH

and metaphysics that is worth noting; Plantinga's [1967] was the first publication after Marcus' [1961] and [Marcus et aI., 1962] to argue that ordinary names are rigid designators and (contingent) definite descriptions are nonrigid designators. (Hintikka [1963] and F0llesdal [1961; 1967] were talking only tangentially about ordinary names.) In "De Re Et De Dicto", Plantinga says [1969: 257]: "Perhaps the notion ofa proper name itself involves essentialism; perhaps an analysis or philosophical account of the nature of proper names essentially involves essentialist ideas". What Plantinga means here appears on pages 176ff. of his God and Other Minds, published in 1967; the relevant section of this book was written in 1965 and the ideas were presented in seminars at Wayne State University between 1963 and 1967 (after Planting a left Wayne State University for Calvin College).4 Planting a asks us to consider: "( 1) An object x has a property P essentially if and only if x has P and the statement x lacks P is necessarily false." [1967: 176].

Plantinga notes problems will arise if we substitute definite descriptions for the individual variable, but will not arise in the case of names: "Difficulties arise if we generalize to (1); but the instantiation of (1) for Socrates may seem harmless enough, as it will in any case where a proper name replaces the variable 'x' in (1)." [1967: 177]. Plantinga adds that we need rigid designators of properties (e.g., "whiteness") and not non-rigid designators of properties (e.g., "the property I am thinking of') if (1) is to be satisfactory. (Here again I am using the terminology, "rigid/nonrigid' that has become an established part of philosophical vocabulary, and whose meaning is not tied to the details of anyone philosopher's theory. I think a more accurate phrase is "modally stable signifier" but it facilitates communication to stay with established terminology.) Plantinga notes an analogy between rigid/nonrigid designators of properties and rigid/nonrigid designators of individuals: "Still, expressions like 'whiteness', 'masculinity', 'mean temperedness', and the like, differ from expressions like 'Socrates' least important property', 'the property I am thinking of, 'the property mentioned on page 37', and so on, in pretty much the way that proper names of individuals differ from definite descriptions of them". [1967: 178]. P1antinga already holds at this time the theory of nontrivial essences and he uses his theory of essences to indicate how names are rigid designators but definite descriptions are not. Given the rigidity of names for individuals and properties, then (1) is satisfactory. This requirement is embodied in Plantinga's conditions involving rigid names and nonrigid descriptions: "(3) x has P necessarily if and only if x has P and the proposition x lacks P is necessarily false (where the domain of the variable 'x' is unlimited but its substituend set contains only proper names, and

(eds.),

A MORE COMPREHENSIVE HISTORY

275

where the domain of the variable 'P' is the set of properties and its substituend set contains no definite descriptions or expressions definitionally equivalent to definite descriptions)" [1967: 179].

The idea is that (to use my example) "Scott lacks rationality" is necessarily false since the reference of the name "Scott" is rigid in modal contexts. But the definite description, "the biped discussed most frequently by Russell" is nonrigid in modal contexts. Thus, we cannot substitute the definite description of Scott for the variable x in (3) if we want to determine whether Scott has rationality necessarily; it is possible that the biped discussed most frequently by Russell is not Scott. Plantinga extends his criticism of the traditional descriptivist theory of names in his October 1969 talk at Cornell University, "World and Essence". This talk, published as [1970], contains a lengthy criticism of the modal thesis that belongs to the descriptivist theory of names, in particular, Searle's cluster theory, which is also the main target of Kripke's attack in Kripke's January 1970 talk, "Naming and Necessity", published as [1972]. Plantinga argued at length that it is false that "if x exists, x has most of the properties commonly attributed to it". Kripke quotes Searle on page 287 of [Kripke, 1972]: "it is a necessary fact that Aristotle has the logical sum, inclusive disjunction, of properties commonly attributed to him". Kripke comments: "This is what is not so. It just is not, in any intuitive sense of necessity, a necessary truth that Aristotle had the properties commonly attributed to him". [1972: 287]. Plantinga's earlier criticism of Searle's theory is similar. For instance, Plantinga writes such things as: "Searle is wrong, I believe, in thinking the disjunction of the Sj [the properties commonly attributed to Socrates, is] essential to Socrates". [1970: 473]. "Socrates could have been born ten years earlier and in Thebas, let us say, instead of Athens. Furthermore, he could have been a carpenter all his life instead of a philosopher. He could have lived in Macedonia and never even visited Athens." [1970: 473]. In this paper, Plantinga distinguishes two uses of proper names, one in which they are rigid and one in which they are not; he here differs from his [1967], which suggests that names only have a rigid use. Plantinga made several other early contributions to the New Theory of Reference that are not included in the standard history, e.g., there is an explicitly descriptional theory of rigidity Inonrigidity in Plantinga's 1969 article "De Re Et De Dicto"; this paper which was given in a seminar at the University of Michigan in Fall, 1967 and later in a summer 1968 conference that was mentioned in Quine's brief autobiographical sketch in Schilpp's [1986: 37]. The ideas were worked out in a series of seminars given at Wayne State University in 1963-67. 4 By a descriptional theory of rigidity I do not mean a theory that simply notes that some definite descriptions, such as "the number that immediately follows number two" are rigid designators. Rather, Plantinga developed a theory of rigid descriptions that (arguably) could account for all cases of rigidity, even proper names, so that an appeal to direct reference is otiose. In Plantinga's 1969 article, he makes a distinction between definite (eds.),

276

QUENTIN SMITH

descriptions that are used de re or de dicto. Plantinga considers the definite description (using his numbering): (32) It is possible that the number of apostles should have been prime.

Plantinga writes: "Now (32) can be read de dicto, in which case we may put it more explicitly as (32a) The proposition, the number of apostles is prime, is possible;

It may also be read de re, that is, as (32b) The number that numbers the apostles (that is, the number that as things in fact stand numbers the apostles) could have been prime." [1969: 244]

Plantinga's "de re reading" of the description involves the description designating rigidly by virtue of its descriptive conditions. Plantinga uses the phrase "as things in fact stand" to express a rigid descriptive condition that belongs to the "de re reading" of the description "the number of apostles'. This descriptive condition is that "the number of apostles" refers to whatever number is the number of the apostles in the actual world. The description condition is a world-indexed condition that picks out a certain possible world, the actual world. The standard history of the New Theory of Reference has Plantinga making an appearance merely in his [1978], as a "co-originator" of the rigid descriptional theory of names, along with Linsky [1977]. I mistakenly endorsed this part of the standard history in my [1995a] and [l995b]. As I briefly suggested earlier, the standard history has some of the contributions of Kripke and Plantinga reversed. The theory of rigid descriptional names is first sketched in part by Kripke in 1962, but Plantinga had published before Kripke's [1971] and [1972] the distinction between rigid names and nonrigid descriptions, the extensive modal argument against Searle's "cluster of descriptions" theory of reference, and the theory that the necessary/contingent distinction is a metaphysical distinction and is nonequivalent to the epistemological distinction between the a priori and the a posteriori. 10. GEACH, DONNELLAN, KRIPKE AND THE "HISTORICAL CHAIN" THEORY OF REFERENCE

The reason why Peter Geach [1969] is not recognized or credited for originally publishing the "historical chain" theory of reference is one of the most perplexing mysteries of the standard history of contemporary analytic philosophy. A review of the literature suggests the nearly universal consensus of opinion is that Donnellan [1970] or Kripke [1972] developed the theory and that Geach's presentation of the theory does not exist. There is no debatable question about textual interpretation, since Geach's presentation of the theory [1969] is virtually identical with Kripke's presentation [1972] and to a slightly lesser extent with Donnellan's presentation [1970]. Both Kripke [1972: 769] and (eds.),

A MORE COMPREHENSIVE HISTORY

277

Donnellan [1970: 357] refer to each other's presentation of the theory, but neither refer to Geach's presentation and Geach is absent from subsequent discussions of the theory and accounts of its origin. (Kripke added a sentence to footnote 2 of his [1972] in [1980: 23, n. 2], reading: "I also recall the influence of early conversations with Albritton and with Peter Geach on the essentiality of origins". The "essentiality of origins" is a different topic than the historical chain theory of reference.) In this section, I need only quote the parallel presentations. Geach writes: [1969: 288-89] "I do indeed think that for the use of a word as a proper name there must be in the first instance be someone acquainted with the object named. But language is an institution, a tradition; and the use of a given name for a given object, like other features of language, can be handed on from one generation to another; the acquaintance required for the use of a proper name may be mediate, not immediate. Plato knew Socrates, and Aristotle knew Plato, and Theophrastus knew Aristotle, and so on in apostolic succession down to our own times; that is why we can legitimately use 'Socrates' as a name the way we do. It is not our knowledge of this chain that validates our use, but the existence of such a chain; just as according to Catholic doctrine a man is a true bishop if there is in fact a chain of consecrations going back to the Apostles, not if we know that there is. When a serious doubt arises (as happens for a well-known use of the word 'Arthur') whether the chain does reach right up to the object named, our right to use the name is questionable, just on that account. But a right may obtain even when it is open to question .... I introduced the use of the proper name 'Pauline' by way of the definite description 'the one and only girl Geach dreamed of on N-Night'; this might give rise to the idea that the name is an abbreviation for the description. This would be wrong."

Geach's theses can be compared with Kripke's [1972: 298-300]: "Someone, let's say, a baby, is born; his parents call him by a certain name. They talk about him to their friends. Other people meet him. Through various sorts of talk the name is spread from link to link as ifby a chain .... A certain passage of communication reaching ultimately to the man himself [Richard Feynman] does reach the speaker. He is then referring to Feynman even though he can't identify him uniquely.... So he doesn't have to know these things [descriptions]. but, instead a chain of communication going back to Feynman himself has been established, by virtue of his membership in a community which passed the name on from link to link .... On our view, it is not how the speaker thinks he got the reference, but the actual chain of communication, which is relevant. ... Obviously the name is passed on from link to link. But of course not every sort of causal chain reaching from me to a certain man will do for me to make a reference. There may be a causal chain from our use of the term 'Santa Claus' to a certain historical saint. but still the children, when they use this, by this time probably do not refer to that saint."

Kripke's theory is more similar to Geach's than is Donnellan's, but the basic ideas are also present in Donnellan [1970]. Donnellan writes [1970: 352-353] "In general, our use of proper names for persons in history (and also those we are not personally acquainted with) is parasitic on uses of the names of other people - in conversation, written records, etc.... The history behind the use of a name may not be known to the individual using it. ... Yet, in such cases the history is of central importance to the question of whether a name in a particular use has a referent and, if so, what it is. The words of others, in conversations, books and documents can ... distort our view of what we are naming. But at the same time it can, to one who knows the facts, provide the means of uncovering the referent, if there is one.

(eds.),

278

QUENTIN SMITH

The role of this history leading up to a present use of a name has almost always been neglected by those who accept the principle of identifying descriptions . ... That is, if we neglect the fact that there is a history behind our use of the name 'Thales' or 'Aristotle' and concentrate only upon the descriptions we would supply about their life, their works and deeds, it is possible that our descriptions are substantially wrong without the consequence being that we have not been referring to any existent person."

Despite these virtually identical passages, the historical chain theory of names has been attributed almost universally to Donnellan and Kripke for the past 25 years or so, with Geach being eclipsed from the scene. I I. CONCLUSION

The basic ideas of the New Theory of Reference had been published or presented in lectures before 1970, mainly by Marcus, F0llesdal, Hintikka, Plantinga and Geach. What remained was to develop and extend the concepts formulated by these five thinkers. The development had two waves: the first involving Kaplan (who extended the ideas to indexicals, among other contributions), Kripke and Putnam (who extended the ideas to natural kind terms, among other contributions), and Donnellan (who concentrated mostly on the causal theory of reference, but also made other contributions). Kaplan, Kripke, Donnellan and Putnam (who appear in the standard history as the four main originators of the New Theory of Reference) are in fact four of the main figures in the first wave of developers of this theory. There is a fifth figure in the first wave of developers who does not appear in the standard history, namely, Nino Cocchiarella, whose explanation of the distinction between metaphysical and logical necessity in terms of primary/secondary semantics is crucial to understanding the New Theory of Reference but has gone unrecognized by most New Theorists. The second wave of developers, headed by such philosophers as Michael Devitt and Nathan Salmon, covers a large number, if not most, of contemporary philosophers. What is know by the referentially used description, "The New Theory of Reference", has already become one of the major movements in the history of analytic philosophy, comparable to logical positivism and ordinary language philosophy. Given the extent and importance of this movement, it is not unreasonable to think that the time is ripe for more philosophers to engage in the research necessary to construct a complete history of the origins and development of this movement. 5 Philosophy Department Western Michigan University

(eds.),

A MORE COMPREHENSIVE HISTORY

279

NOTES My mis-interpretation appears in [Smith, 1995b]. My mistake lay in not realizing that Smullyan was talking about an artificial language and in not realizing that his first solution to Quine's paradox involved using a Russellian theory of logically proper names. I am not aware of any other recent publications in which Smullyan's theory is interpreted accurately. 2 I am aware of the problems with the idea that there is a set of all possible worlds, but will continue to talk of sets since this is the language used in the 1950s and 1960s. 3 I thank Nino Cocchiarella for providing me with the information about the origin of his ~1975a].

I am grateful to Alvin Plantinga for answering my questions about the background of some of his publications. Plantinga indicates that (according to his best recollection) some of these ideas were circulating around Wayne State University before he left in 1963 (in discussions that included, besides Plantinga, Robert Sleigh, Ed Gettier, Richard Cartwright, Hector-Neri Castaneda, George Nakhnikan and [a bit later] Keith Lehrer) and Plantinga emphasized that these other named philosophers deserve as much credit as he for the general gist of the ideas I am attributing to Plantinga. 5 I thank two referees for their helpful criticisms of an earlier draft of this paper, which enabled improvements to be made. My main intellectual debt is to the writings by, and private communications with, Nino Cocchiarella, whose responses to my questions about his theory of the primary/secondary semantics for logical necessity has proved helpful in my account of the history of modal logic. My interpretation of Kripke's [1959a] derives from Cocchiarella's interpretation and is similar on most points. I have also been influenced by Lindstrom's [1998] and Hintikka's [1980; 1982]. I differ from these three authors on a number of points but my main difference from them is my interpretation of Hintikka's 1957-1963 writings and the particular way in which I have formulated the metaphysicalliogical distinction (but my basic idea of a nonformallformal interpretation of the box and diamond comes from Cocchiarella). I thank Alvin Plantinga for providing me with background information about the genesis and dates of origin of some of his publications (see footnote 4). I also thank Dagfinn F0llesdal for clarifying some of his ideas for me, but he is not aware of all the particulars of my interpretation of his ideas and I am not sure if he would agree with all of them. REFERENCES

Almog, Joseph: 1986, 'Naming Without Necessity', The Journal of Philosophy, 83, 210-242. Armstrong, D.M.: 1989, A Combinatorial Theory of Possibility. Cambridge University Press, Cambridge. Bealer, George: 1987, 'The Limits of Scientific Essentialism', Philosophical Perspectives I: Metaphysics, J. Tomberlin, ed. Ridgeview Publishing Co., Atascadero, California, pp. 289-365. Beth, E.W.: 1960, 'Extension and Intension', Synthese, 12, 375-379. Burge, Tyler: 1992, 'Philosophy of Language and Mind: 1950-1990', The Philosophical Review, 101, 3-51. Carnap, Rudolph: 1963, Reply to Critics, in P. Schilpp, ed., The Philosophy of RudolfCarnap. Open Court Press, La Salle. Carnap, Rudolph: 1947, Meaning and Necessity. University of Chicago Press, Chicago. Carnap, Rudolph: 1946, 'Modalities and Quantification', Journal of Symbolic Logic, 11, 33-64. Church, A.: 1946, 'A Formulation of the Logic of Sense and Denotation' (abstract), Journal of Symbolic Logic, 11, 31. Church, A.: 1943, 'Review of Quine', Journal of Symbolic Logic, 8, 45-47. Cocchiarella, Nino: 1987, Logical Studies in Early Analytic Philosophy. Ohio State University Press, Columbus. Cocchiarella, Nino: 1984, 'Philosophical Perspectives on Quantification in Tense and Modal Logic', in Gabby, D. and Guenther, F., eds. Handbook of Philosophical Logic, Volume Two: Extensions of Classical Logic. D. Reidel Publishing Co., Dordrecht, pp. 309-353.

(eds.),

280

QUENTIN SMITH

Cocchiarella, Nino: 1975a, 'Logical Atomism, Nominalism, and Modal Logic', Synthese, 31, 23-62. Cocchiarella, Nino: 1975b, 'On the Primary and Secondary Semantics of Logical Necessity', Journal of Philosophical Logic, 4, 13-27. Cocchiarella, Nino: 1966, Tense Logic: A Study of Temporal Reference. UCLA Dissertation. Devitt, Michael: (forthcoming). 'Reference', in Routledge Encyclopedia of Philosophy. Routledge, London. Devitt, Michael: 1989, 'Against Direct Reference', in P. French et aI., eds., MidWest Studies in Philosophy, Volume XIV: Contemporary Perspectives in the Philosophy of Language II. University of Notre Dame Press, Notre Dame, pp. 206-240. Devitt, Michael: 1981, Designation. Oxford University Press, New York. Devitt, Michael: 1974, 'Singular Terms', Journal of Philosophy, 71, 183-205. Devitt, Michael: 1972, The Semantics of Proper Names: A Causal Theory. PhD Thesis, Harvard. Donnellan, Keith: 1974, 'Speaking of Nothing', The Philosophical Review, 83, 3-32. Donnellan, Keith: 1972, 'Proper Names and Identifying Descriptions' (reprint of [1970], in Davidson, D. and Harman, G., eds., Semantics of Natural Language. D. Reidel, Dordrecht. Donnellan, Keith: 1970, 'Proper Names and Identifying Descriptions', Synthese, 21(3-4), 335-358. Donnellan, Keith: 1966, 'Reference and Definite Descriptions', The Philosophical Review, 75, 281304. Fitch, Frederick: 1950, 'Attribute and Class', in Marvin Farber, ed., Philosophical Thought in France and the United States. University of Buffalo Publications in Philosophy, Buffalo. Fitch, Frederick: 1949, 'The Problem of the Morning Star and the Evening Star', Philosophy of Science, 16, 137-141. Fl1lllesdal, Dagfinn: 1994, 'Stig Kanger in Memorial', in D. Prawitz et aI., eds., Logic, Methodology and Philosophy of Science IX. Elsevier Science B.V. Fl1lllesdal, Dagfinn: 1986, 'Essentialism and Reference', in Hahan and Schilpp, eds., The Philosophy of wv. Quine. Open Court, La Salle. Fl1lllesdal, Dagfinn: 1967, 'Knowledge, Identity, and Existence', Theoria, 33, 1-27. Fl1lllesdal, Dagfinn: 1966, Referential Opacity and Modal Logic (a private printing and minor revision of Fl1lllesdal's [1961]). Fl1lllesdal, Dagfinn: 1961, Referential Opacity and Modal Logic. Harvard University dissertation, unpublished. Forbes, Graeme: 1989, Languages of Possibility. Basil Blackwell Inc., Oxford. Forbes, Graeme: 1985, The Metaphysics of Modality. Oxford University Press, Oxford. Geach, P.T.: 1969, 'The Perils of Pauline', Review of Metaphysics, 23, 287-300. Grice, H.P.: 1969, 'Vacuous Names' in Donald Davidson and Iaakko Hintikka, eds., Words and Objections. D. Reidel, Dordrecht, pp. 118-145. Hintikka, Iaakko: 1986, 'Who's Who' in Hahan and Schilpp, eds., The Philosophy of wv. Quine. Open Court, La Salle. Hintikka,Iaakko: 1982, 'Is Alethic Modal Logic Possible?', Acta Philosophica Fennica, 35, 89-105. Hintikka,Iaakko: 1980, 'Standard vs. Nonstandard Logic: Higher Order, Modal, and First-Order Logics', in E. Agazzi, ed., Modern Logic: A Survey. D. Reidel, Dordrecht. Hintikka,Iaakko: 1963, 'The Modes of Modality', Acta Philosophica Fennica, 16, 65-79. Hintikka,Iaakko: 1962, Knowledge and Belief Cornell University Press, Ithaca. Hintikka,Iaakko: 1961, 'Modality and Quantification', Theoria, 27,119-128. Hintikka,Iaakko: 1957a, 'Quantifiers in Deontic Logic', Societa Scientariarum Fennica, Commentationes Human Litterarum, 23, 3-23. Hintikka,Iaakko: 1957b, 'Modality as Referential Multiplicity', Ajatus, 20, 49-64. Hintikka, laakko: 1955, 'Form and Content in Quantification Theory', Acta Philosophica Fennica, 8,7-55. Hintikka, laakko and Sandu, Gabriel: 1995, 'The Fallacies of the New Theory of Reference', Synthese, 104, 245-283. Kaplan, David: 1989a, 'Demonstratives', in 1. Almog, ed., Themesfrom Kaplan. Oxford University Press, New York.

(eds.),

A MORE COMPREHENSIVE HISTORY

281

Kaplan, David: 1989b, 'Afterthoughts' in 1. Almog et aI., eds., Themes from Kaplan. Oxford University Press, New York. Kaplan, David: 1986, 'Opacity' in E. Hahn and P. Schilpp, eds., The Philosophy of wv. 0. Quine. Open Court Publishers, La Salle, Illinois. Kaplan, David: 1978a, 'Dthat', in Peter Cole, ed., Syntax and Semantics, Volume 9. Academic Press, New York, pp. 221-253. Kaplan, David: 1978b, 'On the Logic of Demonstratives', Journal of Philosophical Logic, 8, 81-98. Kaplan, David: 1974, UCLA dissertation (unpublished). Holt, Jim: 1996, 'Whose Idea is it, Anyway?', Lingua Franca, February 1996 issue, pp. 29-39. Kanger, Stig: 1957, Provability in Logic. Stockholm. Kanger, Stig: 1957a: 'The Morning Star Paradox', Theoria, 27, I-II. Kripke, Saul: 1980, Naming and Necessity. Harvard University Press, Cambridge, Mass. Kripke, Saul: 1977, 'Speaker's Reference and Semantic Reference', in P. French et aI., eds., Contemporary Perspectives in the Philosophy of Language. University of Minnesota Press, Minneapolis. Reprinted in Martinich, ed., The Philosophy of Language. Oxford University Press, New York, 1985. Kripke, Saul: 1972, 'Naming and Necessity', in D. Davidson and G. Harman, eds., Semantics of Natural Language. D. Reidel, Dordrecht, pp. 253-355. Kripke, Saul: 1971, 'Identity and Necessity', in M. Munitz, ed., Identity and Individuation. New York University Press, New York. Kripke, Saul: 1963a, 'Semantical Analysis of Modal Logic I: Normal Propositional Calculi', Zeitschriftfur Mathematische Logik, 9, 67-96. Kripke, Saul: 1963b, 'Semantical Considerations on Modal Logic', Acta Philosophica Fennica, 16, 93-94. Kripke, Saul: 1959a, 'A Completeness Theorem in Modal Logic', Journal of Symbolic Logic, 24, 114. Kripke, Saul: 1959b, 'Semantical Analysis of Modal Logic' (abstract), Journal of Symbolic Logic, 24,323-324 Lewis, C.I.: 1918, A Survey of Symbolic Logic. University of California Press. Linsky, Leonard: 1977, Names and Descriptions. University of Chicago Press, Chicago. Lindstrom, Sten: 1998, 'An Exposition and Development of Kanger's Early Semantics for Modal Logic', this volume. Marcus, Ruth Barcan: 1993, Modalities. Oxford University Press, New York. Marcus, Ruth Barcan: 1981, 'A Proposed Solution to a Puzzle About Belief', Midwest Studies in Philosophy. French et aI., eds. Vol. 6, 501-510. Marcus, Ruth Barcan: 1971, 'Essential Attribution', Journal of Philosophy, 7, 187-202. Marcus, Ruth Barcan: 1967, 'Essentialism in Modal Logic', Nous, 1, 90-97. Marcus, Ruth Barcan: 1963, 'Classes and Attributes in Extended Modal Systems', Proceedings of a Colloquium on Modal and Many-Valued Logic. Acta Philosophica Fennica, 16, 123-136. Marcus, Ruth Barcan: 1961, 'Modalities in Intensional Language', Synthese, 13, 303-322. [Marcus] Barcan, R.: 1947, 'The Identity of Individuals in a Strict Functional Calculus of First Order', Journal of Symbolic Logic, 12, 12-15. [Marcus] Barcan, R.: 1946, 'A Functional Calculus of First Order Based on Strict Implication', Journal of Symbolic Logic, 11,1-16. Marcus, Ruth Barcan et al.: 1962, 'Discussion of the Paper of Ruth B. Marcus', Synthese, 14, 132143. Martinich, A.: 1985, The Philosophy of Language. Oxford University Press, New York. Montague, Richard: 1974, Formal Philosophy. Yale University Press, New Haven. Montague, Richard: 1963, 'Syntactical Treatments of Modality, with Corollaries on Reflexion Principles and Finite Axiomatizability', Acta Philosophica Fennica, 16,153-167. Montague, Richard: 1960, 'Logical Necessity, Physical Necessity, Ethics and Quantifiers', Inquiry,

4,259-269. Moore, G.E.: 1903, Principia Ethica. Cambridge University Press, Cambridge.

(eds.),

282

QUENTIN SMITH

Moore, G.E.: 1899, 'The Nature of Judgement', Mind,8, 176-193. Moore, G.E.: 1898, Letter to Desmond McCarthy; quoted from Thomas Baldwin's 'Moore's Rejection of Idealism', in Richard Rorty, lB. Schneedwind and Quentin Skinner, eds., Philosophy in History. Cambridge University Press, Cambridge. Parsons, Terence: 1969, 'Essentialism and Quantified Modal Logic', The Philosophical Review, 78, 35-52. Plantinga, Alvin: 1978, 'The Boethian Compromise', American Philosophical Quarterly, 15, 129138. Plantinga, Alvin: 1974, The Nature of Necessity. Clarendon Press, Oxford. Plantinga, Alvin: 1970, 'World and Essence', The Philosophical Review, 79, 461-492. Plantinga, Alvin: 1969, 'De Re Et De Dicto', Nous, 3, pp. 235-258. Plantinga, Alvin: 1967, God and Other Minds. Cornell University Press, Ithaca. Prior, A.N.: 1967, 'Modal Logic' in Paul Edwards, ed., Encyclopedia of Philosophy, Volumes 5 and 6. Macmillan Publishing Co., New York. Putnam, Hilary: 1975, 'The Meaning of 'Meaning", in H Putnam Collected Papers. Cambridge University Press, Cambridge. Quine, W.V.O.: 1994, 'Promoting Extensionality', Synthese, 98, 143-151. Quine, W.V.O.: 1981, Theories and Things, Harvard University Press, Cambridge. Quine, w.v.o.: 1947, 'The Problem of Interpreting Modal Logic', The Journal of Symbolic Logic, 12, 43-48. Quine, W.v.o.: 1943, 'Notes on Existence and Necessity', The Journal of Philosophy, 40, 113-127. Quine, W.v.O.: 1941, 'Whitehead and Modern Logic', in P. Schilpp, ed., The Philosophy of Alfred North Whitehead. Tudor Publishing Co., New York. Quine, W.v.O.: 1939, 'Designation and Existence', The Journal of Philosophy, 36, 701-709. Russell, Bertrand: 1919, Introduction to Mathematical Philosophy. George Allen & Unwin Ltd., London. Russell, Bertrand: 1918, 'The Philosophy of Logical Atomism' in R. Marsh, ed., Logic and Knowledge. Capricorn Books, New York, 1971. Russell, Bertrand: 1905, 'On Denoting', in R. Marsh, ed., Logic and Knowledge. Capricorn Books, New York, 1971. Russell, Bertrand: 1903, Principles of Mathematics. The Norton Library, New York. Salmon, Nathan: 1989, 'The Logic of What Might Have Been', The Philosophical Review, 98,3-34. Salmon, Nathan: 1986, Frege's Puzzle. MIT Press, Cambridge, Mass. Salmon, Nathan: 1981, Reference and Essence. Princeton University Press, Princeton. Smith, Quentin: 1998a, 'Direct, Rigid Designation and A Posteriori Necessity: A History of Critique', this volume. Smith, Quentin: 1997, Ethical and Religious Thought in Analytic Philosophy of Language. Yale University Press, New Haven. Smith, Quentin: 1995a, 'Marcus, Kripke, and the Origin of the New Theory of Reference', Synthese, 104, 179-189. Smith, Quentin: 1995b, 'Marcus and the New Theory of Reference: A Reply to Scott Soames', Synthese, 104, 217-244. Smith, Quentin: 1994, 'Temporal Indexicals', in L. Nathan Oaklander and Quentin Smith, eds., The New Theory of Time. Yale University Press, New Haven. Smith, Quentin: 1993, Language and Time. Oxford University Press, New York. Smith, Quentin: 1991, 'The New Theory of Reference Entails Absolute Time and Space', Philosophy of Science, 58, 411-16. Smith, Quentin: 1989, 'The Multiple Uses oflndexicals', Synthese, 78, 167-91. Smith, Quentin: 1988, 'Tensed States of Affairs and Possible Worlds', Grazer Philosophische Studien, 31,225-35. Smith, Quentin: 1986, The Felt Meanings of the World: A Metaphysics of Feeling. Purdue University Press, West Lafeyette. Smith, Quentin and Oaklander, L. Nathan: 1995, Time, Change and Freedom. Routledge, London.

(eds.),

A MORE COMPREHENSIVE HISTORY

283

Smith, Quentin and Craig, William Lane: 1993, Theism, Atheism and Big Bang Cosmology. Clarendon Press, Oxford. Smullyan, Arthur: 1948, 'Modality and Description', The Journal of Symbolic Logic, 13, 31-37. Smullyan, Arthur: 1947, Review of Quine's 'The Problem of Interpreting Modal Logic', The Journal of Symbolic Logic, 12, 139-141. Wettstein, Howard: 1991, Has Semantics Rested on a Mistake and Other Essays. Stanford University Press, Stanford. Wettstein, Howard: 1986, 'Has Semantics Rested on a Mistake?', The Journal of Philosophy, 83, 185-209. Von Wright, Richard: 1951, An Essay in Modal Logic. North-Holland Publishing Co., Amsterdam. Wittgenstein, Ludwig: 1922, Tractatus Logicus-Philosophicus. Translated by D. Pears and B. McGuiness. Kegan Paul, Trench, Trubner and Co., London, 1961. Wong, Kai-Yee: 1996, 'Sentence-Relativity and the Necessary A Posteriori', Philosophical Studies, 83, 53-91. Yablo, Stephen: 1993, 'Is Conceivability a Guide to Possibility?" Philosophy and Phenomenological Research, 53, 1-42.

(eds.),

NAME INDEX

Albritton, Rogers 143-4, 250, 277 Almog, Joseph 3, 4, 11, 51, 141, 169, 235-6, 249,264-5 Aquinas, St Thomas 239 Aristotle 145, 199,239,241,271 Armstrong, David 144,271 Augustine 174 Ayer, Alfred Jules 149, 152 Bacon, John 95, 102 Bayart, 186 Bemays, Paul 189 Beth, E.w. 267 Black, Max 90 Boolos, George 232 Bradley, F.H. 237 Braun, David 4 Brouwer, L.E.J. 184,200,201 Burge, Tyler 138, 141, 142, 146, 174 Burgess, John x, 65, 85-87, 144, 149, 150,151, 153-155, 157-163, 173-175 Cantor, G. 226 Camap, Rudolf 98, 106, 107, 110, 165, 167, 168,175,181,182,184,186,187,194,2045, 209-11, 215, 231, 240, 242-4, 249, 251, 258-61, 264-5, 267, 270 Cartwright, Richard 175,279 Casteneda, Hector-Neri 175, 279 Chastain, Charles 250 Church, Alonzo 15, 17, 19, 39, 69, 100, 196, 243-4 Cocchiarella, Nino 97, 105, 175,236, 261, 26771,278-9 Cohen, Robert S. ix, xii Craig, William Lane 56 Creswell, Max 232 De Finetti, Bruno 38 Devitt, Michael 141-42, 148-49, 153, 175,236, 278 Donnellan, Keith xii, 3-5, 8,11,14,41,90,137145, 147-48, 158, 160, 174, 175,236,250, 277-8

285

Dretske, Fred 144 Etchemendy, John 215 Falk, Arthur 60 Feys, 182,201,259 Fitch, Frederick 15, 17, 19,21, 23, 28, 38-40, 42,45, 60, 67-73, 75, 77, 81, 82, 100, 126, 132, 174, 191, 194,201,244-6,250-1 Fitzgerald, Paul 40, 174 Fogelin, Robert xii, 60-61 Follesdal, Dagfinn x, xii, xiii, 89, 97, 99, 122, 138, 146, 175, 203, 236, 245, 246-52, 254-7, 263,266,274,278-9 Forbes, Graeme 236 Fraenkel, A.A. 201 Frege, Gottlob 4, 5, 90, 92-94, 102, 105, 107, 137, 140, 147, 148, 187, 195,235,239,2524,256 Garson, James W. 86, 232 Geach, Peter 90, 93, 137, 138, 142-146, 174, 236,248,250,263,277-8 Gentzen, G. 203 Gettier, Edmund 175,279 Godel, Kurt 182, 184, 185, 199,203 Grice, H. Paul 143-4 Gurlank, David 156 Herodotus 145 Hilbert, David 189 Hintikka, Jaakko x, xii, 146, 148, 153, 167,203, 207, 210-11, 213, 236, 245, 249-50, 254, 258-69,274,278-9 Holt, James 250, 256 Hughes, G.E. 232 Hume, David 57, 239 Jonsson, B. 207, 231 Kanger, Stig x-xi, 47, 165, 203-233, 245, 25962, 264, 266-7 Kant, Immanuel 237-8 Kaplan, David 3-5, 11, 14, 16,87, 89, 90, 137, 138, 141-43, 146-49, 149, 153, 158, 160, 165,168-73,175,224,235-6,267,278 Kneale, William 90, 93

286

NAME INDEX

Kreisel, Georg 229,232 Kripke, Saul passim, but see especially vii-xiii, 3-12,13-35,37-61,66,72,73,76,77,80,82, 89-123, 125-135, 137, 138, 140-78,203-233, 236,238,247-50,252-7,261-79 Langford, C.H. 201, 240 Lehrer, Keith 279 Leibniz, Gottfried Wilhelm 47, 95, 100, 153, 166,192,194,204-5,240 Lewis, David 232 Lewis, Clarence Irving 7, 98, 107, 110, 152, 165,182,201,240,265 Lindstrom, Sten x-xi, 279 Linsky, Leonard 7,8,40,43,44, 50, 86, 99, 252, 276 Locke, John 239 MacCarthy, Desmond 236 Malebranche, 57 Marcus, Ruth Barcanpassim, but see especially vii-xiii, 3-12, 13-35, 37-61, 65-87, 89-123, 125-135,137-39,141,144,146-78,186,191, 194, 197,201,204,224,235-6,240,244-7, 249-57,263,270-2,274,278 McCracken, C.l 57 McDowell, John 11 McGee, Van 231 Mill, John Stuart 28, 40, 50, 67, 90, 92-95, 97, 104, 108, 118, 126, 129, 132, 158, 160, 163, 169,174,239,243,245,251,256 Montague, Richard x, 146, 203, 207-11, 231, 259,263-4,266-7 Moore, George Edward 236-8,241,260,271 Nagel, Ernest 90 Nakhnikan, George 279 Nerlich, Graham 170 Ockham, William 174, 239 Parmenides 239 Parsons, Terence 247 Pears, David 90 Peirce, Charles Saunders 174 Perry, John 3, 4, 5, 235-6 Plantinga, Alvin 7, 8,40, 43, 44, 50, 138, 146, 153,156,157,161,175,236,238,252,257, 270-6, 278-9 Plato 174,239,271 P10tinus 5 Prior, Arthur 67,69,96,97, 100, 112, 127, 146, 243,255,263,266,269

Putnam, Hilary xii, 3,4,5,52,53,90, 104, 141, 158,236,238,250,278 Quine, Willard van Orman viii-xiii, 3, 14, 15, 17,18,23,38,39,45,56,67-73,75,85,86, 89,90,97-102,105, 107-1l0, 114, 116, 1l9122, 132, 134, 164, 165, 187, 191, 197-199, 201,221,240-9,251-7,263,271,276 Ramsey, Frank Plumpton 38, 96, 97, 127 Recanati, F. 4 Richard, Mark 16 Russell, Bertrand 5, 67, 72, 79, 86, 90, 95-97, 99,101-103,105,107,112,114, 1l8, 126, 127, 131, 137, 140, 148, 158-160, 163, 184, 188,190,235,238-9,242-5,252-4,256 Salmon, Nathan 3, 4, 6, 9,11,41,44,141,142, 145, 160, 235, 236, 278 Sandu, Gabriel 148, 153,258,270 Schopenhauer, Arthur 129 Searle, John 90, 94, 97, 275-6 Shapiro, S. 99 Sidelle, Alan 170 Sleigh, Robert 175, 279 Siote, Michael 250 Smith, Quentin vii-xiii, 13-35,48,51,56,65-87, 123, 125-135, 137 Smullyan, Arthur 15, 17-19,21,23,28,38-40, 42,45,60,67-73,77,81,82,85,86,97,99, 100, Ill, 118, 126, 132, 160, 174, 191-94, 241-2,243-6,249-51,255,279 Soames, Scott vii-xiii, 3, 4, 37-61, 90,106, 123, 125, 129, 132, 150, 151, 153-155, 157, 163, 174,175 Strawson, Peter 90, 94 Tarski, Alfred 204,206,207-9,231 Thales 145 Tooley, Michael 144 Vallicella, William 60 Von Wright, Georg Henrik 260, 262 Wedberg, Anders 203 Wettstein, Howard 3, 4, 5, 11, 141, 160, 235-6 Whitehead, Alfred North 160 Wilson, Neil x, 191, 192,201 Wittgenstein, Ludwig 60, 96, 97, 174,205, 239, 260 Wong, Kai-Yee 273 Yablo, Stephen 236, 238 Zabludowski, A. 85 Ziff, Paul 90, 94, 126, 132, 174

SUBJECT INDEX

a posteriori 7, 49-50, 51, 131-2, 155-7,235, 239. See also necessity a posteriori a priori 7, 8, 26, 27, 49, 52, 101, 109, 111-2, 127,130-2,134,153-7,160-2,235,237,239, 253, 271. See also contingent a priori absolutely infinite 226 abstracts 187 accessibility relation 206-8,211-12,217,225, 259. See also alternativeness relation accidental vs essential. See essential properties acquaintance 142 aliases, problem of 98 alternativeness relation 259-60, 262, 265-6, 269. See also accessibility relation American Philosophical Association vii, xi analytic truth 8, 18,25,52,72,79,81,86,95, 98, 101, 109, 110, 127, 130-2, 134, 152-3, 213-5,217,220 anonymity, problem of 98 arbitrariness, problem of 98 assignment 212, 216-8,222,225,228 assignment operator, constant 223-4 Barcan formula 30, 87, 205, 209-10, 213 belief statements, see knowledge statements Brouwer's axiom 184,201 capitalization 94 Carnap frame 210 causal theory ofreference vii, 6, 9, 20, 103-4, 115, 125, 127, 137-8, 141-5, 173-4,235,248, 277-8 Church's theorem 206 classes, identity condition for 184 class abstracts 39, 184, 186, 188-90, 193 closed formula 98 coextensiveness 197-8, 200 coincidence relation 219, 221 common nouns 103-4 completeness theorem, Godel's 203, 206 completeness theorems for modal logics 181, 195,211,228 concept 237

concept, individual 168, 187,219-21,244 congruent 6, 39,42,240 constant, individua147, 86, 89, 143, 155, 166-7, 169,258 constant domain (in model theory) 97, 170 context 6, 107-8, 111-2, 127. See also modal context contingent. See necessity, necessary truth contingent a priori vii, 28, 105 contingent vs empirica125, 49,51 counterfactual argument (against descriptive theory) 93,103,104-7. See also modal argument Cornell University 157, 272, 275 counterfactual contexts 77, 108 counterfactual situation 154 counterfactual stability 105-7. See also necessity, metaphysical definition 93 definition, contextual 187-8 demonstrative pronouns 103, 239 descriptions passim, but see especially 7, 38, 90, 94,96,99, 100, 101, 106, 122, 130-2, 138, 142, 144, 181, 182, 184, 186, 189-91, 199-200 description, attributively used 8,41, 137-41 description, contingent 8, 43, 44, 46, 80, 156, 236 descriptions, disguised 3,5,6,38-40,67, 189 description, empirical 10 description, Fregean 16 description, identifying 4 description, modally stable 8 description, necessary 7 description, reference-fixing 6, 41,52, 143-4 description, referentially used 8, 41, 137-41 description, rigidified 86 description, rigid 122, 276 descriptions, Russell's theory of 99, 184, 191, 242-3 description, singular 5, 45, 150-1 description, unique 5, 40 descriptive meaning 92

287

288

SUBJECT INDEX

descriptive sense, contingent (of names) 6 descriptive theory of names 92, 93, 95,101, 102, 103,275 designator, non-rigid 8, 48, 105,274 designator, rigid. See rigid designator dictionary. See also lexicon, encyclopedia 10, 19,54-5,92,110,115-7,119-20,133-4,1557, 175,253 dictionary, biographical 10 direct reference vii, xii, 3-12, 19,35,38-9,42, 43,50,68,81,92,115,121,126-7,137,139, 141,143,168,173,235,237-9,245,256 discovery, non-empirical 112 domain 47,206-8,208-9,214,216,217-8,222-6 empirical, see a priori encyclopedia, encyclopedic dictionary entry. See also dictionary, lexicon 10, 92, 110, 113, 115-7,132-3,135 epistemic contexts, see knowledge statement, context epistemological argument (against descriptive theory) 7, 24, 27-8, 37,43-4,48-50,93, 1034, 146 equivalence, strict 6-7 essence 162-4, 258, 270 essential properties 10, 98,121-2,153,163-4, 254-5, 270, 273 essentialism vii, x, 3-4,10,14,17,34,119-21, 143,153, 199,247,253-5,270-71 equivalence relation 6 existential generalization 73-5, 86, 191-2,221 extensional opacity 246 extensional vs intensional, see intensional extensions 187,206,212,228 fact. See state of affairs fact, empirical 7-8, 24, 25, 26, 49-50, 56,96, 100, 109 fact, contingent 25, 49, 51 Gentzen's Haupstatz 203 genus 8, 129-30, 139-40 historical chain theory. See causal theory of reference identity vii, 6-7, 48, 58, 70,100,106,111-3, 131, 134, 152, 181-2, 190-1, 194-5,200,213, 219-221,223,253 identity, necessity of 3,7,9, 10, 11, 13, 14, 17, 18,19,22-23,28,29-31,38,43-6,54,59,60, 69,70,72,73-77,78-9,80,87,95,105,116,

126-7,156,163-4,213,224,243,250,256, 272-3 identity of extension 186, 194 identity of intension 184 identity statement 99,101,102,105-7,110, 118-20, 152, 158-9, 160 identity, transworld 3 incompleteness, semantic 195 indexicals 8, 44,146-7,188 indiscernibility 6, 13 Indiscernibility ofIdenticals 95, 100, 192,213, 220,223 intensional entities 187, 197,204 intensional 18, 39, 74-5, 112, 197,204. See also context intensional arithmetic 99 interpretation 207-9, 216-8, 225, 228-9 interpretation, metalinguistic 214 interpretational semantics 215 inventory 91, 102 K-validity 212 Kanger model 217, 223-6 Kanger semantics 204,208,214-15,218,223-6, 230. See also semantics, set domain; semantics, class domain Kanger-Kripke semantics 203 knowledge statement 18, 66, 107, 108, 111-4, 251,264 Kripke models 207, 212 Kripke semantics 203, 230, 232 language, formal with modalities 105 language, ordinary vs artificial 14, 150-1, 2425, 248, 250-1, 279 Leibniz' Law. See Indiscernibility of Identicals lexicography 92 lexicon, lexical dictionary entry, see also dictionary, encyclopedia 91, 92, 110,113, 1167, 132-3, 135, 155-6 linguistic vs empirical. See analytic truth, a priori logical atomism 95,96,97, 114 logical consequence 204, 215, 217, 225 logical truth 72, 98, 101, 152, 154, 161,204-6, 208-11,214-15,217,219-20,227-8,230 logicaUy proper names 14-15, 17,27,67,69,72, 97, 101-2, 107, 112, 118, 158-60,242-3. See also tag, names in an ideal sense, proper name Lowenheim-Skolem theorem 203, 231 meaning postulates 135

SUBJECT INDEX modal argument 6-7, 21-22, 28, 33, 37,42-4, 46,57-80,77-81, 146,276. See also counterfactual argument modal context 3, 6, 8, 16, 17, 18,23,38,40,42, 44,59,66,77,107, 111, 114, 147, 159, 196-8, 210,218,224,255 modal distinctions, collapse of 182, 184-6, 199200

modal logic x, 9, 38, 89, 91, 95, 97, 109, 118, 203-4, 206-7 modal logic, quantified 13, 17, 152, 158, 181, 195-8,200,203-4,206,211-13,216,221-2, 240, 242-3, 246 modal logic, semantically complete 181 modal operator 47-8,75, 166,205-7,210,2137,225, 227-8, 267-8 modal paradoxes 39 modal system, Godel's 182, 184-5 modal system, QS5= 213, 227-9 modal system SI 182 modal system S2 70,186-7,266 modal systems S4, QS4 7, 13,22,29,70,87, 152,158,160,182,184-5,204,207 modal system S5 152, 181, 182, 184-5, 194,204 modal test 7 modalities, iterated 187, 201 modality, de dicto 34, 98, 120-2, 164, 257 modality, de re 34,51-2,98,99, 120-2, 164, 257 modality, epistemic 190 modality, logical 190, 255 modality, metaphysicaL See necessity, metaphysical model 47, 105, 154, 159, 166-7,209,211-12, 217,231,260-1,265,267-8 model sets 211, 260-5, 267-9 model structures 211-12, 228 model theory 23,47,97, 105-6, 166,203 Morning Star paradox 220, 223. See also identity, substitutivity names, naming passim, but see especially xi, 311, 13-35,37-61,73-137, 155, 159, 165-6, 181, 189-90,195,199,239 names, coreferential 157 names, in an ideal sense 91, 1l0, 126, 128. See also tag, logically proper names names, logically proper. See logically proper names names, proxy 148 name-relation method 241 natural kind 3, 8, 9 natural kind terms vii necessary truth 9, 109, 110, 159, 161,204,214, 259

289

necessity 98-99,101,106-7,110,112,130-1, 152,154,205,214,235,239,271 necessity, a posteriori 3, 8, 11,24-8,50-56, 1056,113,130,157,162,164,235,236-9,271-3 necessity, analytic 213,215,217,220-1 necessity, de dicto. See modality, de dicto necessity, de reo See modality, de re necessity, logical xi, 72, 75, 107, 152-3, 161-3, 204-9,214-17,224,229-30,257-8,260-1, 264,267-70 necessity, metaphysical, see also counterfactual stability xi, 72, 75, 81,107,130,161-3,213, 215,227,229-30,232,235,238,257-63,26470,273-4 necessity, metaphysical feature of facts 109 necessity, senses of (metaphysical vs other) 107, III, 127, 158,238 necessity, synthetic 9, 11 necessity of origins thesis 9, 162 New York University 172 New Theory of Reference passim but see especially 3-11,14-15,37-61,90,92,103-4,1256,137,148,160,235-283 nonlexical, see dictionary ontological operator 216-7, 225 ontology, intensionaL See intensional entities opaque construction 15-17 open formulas or sentences 98 ostension 93, 144, 155 possibilia 204, 212,215,227-8 possibility. See necessity possible world 8, 47-8,153-4,159,164; 166-7, 181,193-4,203-5,211-13,215-6,227-8,240, 259,265 primary valuation 206, 214, 216, 224-5 Princeton University vii, 146-7,272 procedure, empirical 19, 55-6 procedure, linguistic 55-6 proper name x, 3-8, 14, 16, 17, 18, 19,38-9,43, 45,46-7,59-60,66,68,76,81,82,137,139, 142, 147, 149-50, 163, 182, 187,200,201, 242-3, 247-9, 253. See also logically proper name; tag; names, in an ideal sense property, modally indexed 44 proposition 112-3,217 proposition, de re 237 proposition, general 170, 239 propositional attitude 66 purely designative occurrence 15-16 quantification theory 182,191,199-200,208, 218-19,222 quantifier, general 226

290

SUBJECT INDEX

quantifiers, substitutional interpretation of 96, 255 quantifying in. See modal context quantification, objectual16, 23, 31, 255 reference, causal theory of or historical chain theory. See causal theory of reference referential transparency 246 reflection principle 214 regimentation 102 representation theorems 207 representational semantics 215 rigid designation, complete concept 165 rigid designation, direct reference definition 137,168,171-3 rigid designation, world definition 137, 170-3 rigid designator vi, x, 3-4, 8, 10, 17,22,23,28, 33,43,47-8,60,68,75, 105, 115, 122, 137-8, 146-9,163,167-70,212,235,239,248,252, 254,258,263,274,276 rigid designator, obstinate 168 rigid designator, strong 47,171-3,247 rigid designator, weak 47, 171-3,247 rigid names, descriptionally 252, 257, 276 rigid singular terms 170, 255, 263 rigidity 4,35, 104-6, 108, 154, 165, 168,249, 256 rigidity argument 106 scope (of description) 188, 191-3 semantic argument (for direct reference) 28, 44, 145-6 semantic content 7, 77-8, 81, 141,145 semantic value 207 semantics 4, 41, 140-1 semantics, class domain 224, 226-7. See also Kanger semantics semantics, modal x-xi, 166-7,218,240,259 semantics, model-theoretic 48, 203-4, 206-7, 230,260 semantics, possible world 9,204,210-11,215, 230 semantics, primary 261, 267-9 semantics, secondary 267-71 semantics, set domain 224-7. See also Kanger semantics sense 7, 143, 169, 187 sense, descriptive 3, 7 sense, Fregean 4 sense, modally stable 7-8, 26, 43, 50 singular proposition 3-4, 169-70, 237-8

singular term 16, 45, 60, 181, 188, 190-1, 193, 196-200,246-8,252 singular term, purely designative 16, 17 species vs genus, see genus speech act 140 standard interpretation (of modal logic) 211 Stanford University 138, 146-7 state descriptions 204-5,209-11,215,231,240, 260 state of affairs I 08-II 0, 112-3 substitutability salva veritate, universal intersubstitutability of co-referential names 6-8, 14,17,18,22,26,28,38,43-4,48,54,59,66, 70-71,73,78-80,100,106,107,111-2,114, 118,122,127,134,152,158,168,185,190-1, 194-6, 199,241-2,244-5,247,251,263 substitutivity of concretion 184, 186 substitutivity, of description for description 189 substitutivity, of description for singular term 189 supervalidity 229 synonymy, of names 17,40,42,66,68,80,86 synthetic vs analytic. See analytic truth synthetic a posteriori 9, 131 system 206-8,211,216-9,225 tag, identifying 5, 6,10,19,21,32,54,58-9,69, 79-80,82-83,90,94,96, 101, 110, 119-22, 130,133,139,148-51,154,163,169-70,2535. See also logically proper name, names, in an ideal sense tautology, tautological truth 18, 25, 52, 72, 789,81,95,96, 107, 109, 114, 118, 127, 130-1, 152, 158, 160-1 truncated description. See descriptive theory of names truth, analytic. See analytic truth truth in a mode1204, 207-8 truth, in various senses 203-233 truth, logically necessary. See logical truth universal substitutivity thesis. See substitutivity universal instantiation 73-4, 189, 191, 220 universal truth 211 University of Michigan 276 use vs mention 108-9 validity 98,109,114,152,217,225-6,229 variable, individual 16, 23, 28, 46, 47, 95 Wayne State University 146, 175,276 well-formedness (of sentences) 189-90 world-indexed sense 40

The New Theory of Reference: Kripke, Marcus, and Its Origins

Read more

Ancient Philosophy of the Self (The New Synthese Historical Library)

Read more

Ancient Philosophy of the Self (The New Synthese Historical Library)

Read more

Ancient Philosophy of the Self (The New Synthese Historical Library)

Read more

Lebesgue's Theory of Integration: Its Origins and Development

Read more

Lebesgue's Theory of Integration: Its Origins and Development

Read more

Lebesgue's Theory of Integration: its Origins and Development

Read more

Lebesgue's Theory of Integration: Its Origins and Development

Read more

Proof Theory: History and Philosophical Significance (Synthese Library)

Read more

Proof Theory: History and Philosophical Significance (Synthese Library, 292)

Read more

Saul Kripke

Read more

Early Buddhism and Its Origins

Read more

(Over)Interpreting Wittgenstein (Synthese Library)

Read more

Tense Logic (Synthese Library, 111)

Read more

Tense Logic (Synthese Library, 111)

Read more

Formal Ontology and Conceptual Realism (Synthese Library)

Read more

Attitudes and Changing Contexts (Synthese Library, 332)

Read more

The Reference Library

Read more

The Academic Library: Its Context, Its Purpose, and Its Operation

Read more

The Pomegranate Vase: Its Origins and Continuity

Read more

Kant's Idealism: New Interpretations of a Controversial Doctrine (The New Synthese Historical Library)

Read more

Wittgenstein's Legacy, Kripke, and Cavell

Read more

The Theory of Evolution and Its Impact

Read more

Scepticism and the Causal Theory of Reference

Read more

Coherence, Reference, and the Theory of Grammar

Read more

Scepticism and the Causal Theory of Reference

Read more

The Logic of Essentialism: An Interpretation of Aristotle's Modal Syllogistic (The New Synthese Historical Library)

Read more

Reference, Truth and Conceptual Schemes: A Defense of Internal Realism (Synthese Library, 296)

Read more

Flora of the Hudson Bay Lowland and Its Postglacial Origins

Read more

The New Heat Theorem,: Its Foundations in Theory and Experiment

Read more

Recommend Documents

The New Theory of Reference: Kripke, Marcus, and Its Origins

Ancient Philosophy of the Self (The New Synthese Historical Library)

Ancient Philosophy of the Self The New Synthese Historical Library Texts and Studies in the History of Philosophy VOL...

Ancient Philosophy of the Self (The New Synthese Historical Library)

Ancient Philosophy of the Self The New Synthese Historical Library Texts and Studies in the History of Philosophy VOL...

Ancient Philosophy of the Self (The New Synthese Historical Library)

Ancient Philosophy of the Self The New Synthese Historical Library Texts and Studies in the History of Philosophy VOL...

Lebesgue's Theory of Integration: Its Origins and Development

LEBESGUE'S THEORY of INTEGRATION LEBESGUE'S THEORY o INTEGRATION ']ts Origins and Development THOMAS HAWKINS AMS ...

Lebesgue's Theory of Integration: Its Origins and Development

LEBESGUE'S THEORY Of INTEGRATION LEBESGUE'S THEORY of INTEGRATION its Origins and Development THOMAS HAWKINS AMS...

Lebesgue's Theory of Integration: its Origins and Development

Lebesgue's Theory of Integration: Its Origins and Development

Proof Theory: History and Philosophical Significance (Synthese Library)

Proof Theory: History and Philosophical Significance (Synthese Library, 292)