Contingent A Priori Truths: Metaphysics, Semantics, Epistemology and Pragmatics (Synthese Library, 443) 3030866211, 9783030866211

Table of contents :
Acknowledgements
Introduction
References
Contents
1 The Starting Point: Kripke's ``Magic''
1.1 The Meter Case
1.2 The Neptune Case
1.3 Contingent A Priori Truths and Rigidity
1.4 Two Classical Objections
1.4.1 The Scope Ambiguity View
1.4.2 Rigidified Descriptivism
1.5 Two Gaps
References
2 Indexicals and Kaplan's Cases
2.1 Demonstratives and Pure Indexicals
2.2 Character and Content
2.3 Context of Utterance and Circumstance of Evaluation
2.4 Kaplan's Modal Argument
2.5 Proper Contexts and `I am Here Now'
2.6 Dthat Cases
2.7 Explanation: The Two Dimensions of Meaning
2.8 Necessary A Posteriori Truths
2.9 Some Partial Conclusions
References
3 Donnellan and the Acquaintance Requirement
3.1 Rigid Names and De Re Knowledge
3.2 The Acquaintance Requirement
3.3 Jeshion on Donnellan on Neptune
3.3.1 The De Re Principle and Donnellan's Problem
3.3.2 Donnellan's First Loophole
3.3.3 Donnellan's Second Loophole: PostulatingContingencies
3.4 Some Partial Conclusions
References
4 The Experience Requirement
4.1 A Dialogue Between Plantinga and Philonous
4.2 Salmon's Theories
4.2.1 Salmon I: The A Posteriori Only Theory
4.2.2 Salmon II: The Almost A Priori Theory
4.2.3 Salmon III: The Context Shift Theory
4.3 Soames and ``The Serpent's Egg'' of Two-DimensionalSemantics
4.3.1 Characters and Logical Truths
4.3.2 Characters and the Objects of Thought
4.3.3 De Re Propositional Attitudes
4.4 Some Partial Conclusions
References
5 Kripke's Reformulation of the Contingent A Priori
5.1 Elaborating On Donnellan's Distinction
5.2 Three Models of Descriptive Reference-Fixing
5.3 The Reply to Donnellan's Acquaintance Requirement
5.4 Contingent A Priori Truths as the Basis of OtherContingencies
5.5 Some Partial Conclusions
References
6 Evans and the Varieties of Contingency
6.1 Descriptive Names
6.2 Descriptive Names, Free Logic and Evans' Particular Strategy
6.3 Contingency: Deep and Superficial
6.4 Propositions, Cognitive Content and Evans' General Strategy
6.5 Contingency and Existence
6.6 Some Partial Conclusions
References
7 Two-Dimensionalism
7.1 Stalnaker's Assertion
7.1.1 The Point of Making an Assertion
7.1.2 The Two-Dimensional Semantics
7.1.3 The Diagonal Proposition and Kripke's Contingent A Priori
7.1.4 Rational Communication and the NecessaryA Posteriori
7.1.5 The Context-Relativity of Propositional Concepts
7.1.6 Problems With Belief Attributions
7.2 Davies and Humberstone's Alternative Notion of Necessity
7.3 Chalmers' Primary and Secondary Intensions
7.4 Primary/Secondary Intensions and Character/Content
7.5 Stalnaker's Later Retraction
7.6 Some Partial Conclusions
References
8 Some Other Cases
8.1 LDO Sentences
8.2 Williamson and The Believer
8.3 Best Explanations
8.4 The ``Most Unlikely Event''
8.5 Proper Contexts of Explanation
8.6 Some Partial Conclusions
References
9 Basic Tools: Elements of a Theory of Speech Acts
9.1 The Dimensions of a Speech Act
9.2 The Classes of Illocutionary Acts
9.3 The Creative Aspect of Declarative Acts
9.4 Performatives
9.5 Performatives and Truth
9.6 Performative Utterances as Declaratives
9.7 Institutional and Linguistic Facts
9.8 Illocutionary Commitment and Inconsistency
References
10 Stipulations as Performatives
10.1 Two Gaps
10.2 Stipulations and Illocutionary Acts
10.3 Horowitz
10.4 Contingency
10.5 Closing the Two Gaps
10.6 Some Partial Conclusions
References
11 One Ancestor: The Early Frege on Definitions
11.1 Assertion and the Assertion Sign
11.2 Definitions and the Definition Sign
11.3 Identities as Synthetic Judgements
11.4 Definitions: Turning the Synthetic Propositions Analytic
11.5 Some Partial Conclusions
References
12 Global Conclusions: The Varieties of Contingent A Priori Truths
References
Index

Citation preview

Synthese Library 443 Studies in Epistemology, Logic, Methodology, and Philosophy of Science

Marco Ruffino

Contingent A Priori Truths

Metaphysics, Semantics, Epistemology and Pragmatics

Synthese Library Studies in Epistemology, Logic, Methodology, and Philosophy of Science

Volume 443

Editor-in-Chief Otávio Bueno, Department of Philosophy, University of Miami, Coral Gables, USA Editorial Board Members Berit Brogaard, University of Miami, Coral Gables, USA Anjan Chakravartty, Department of Philosophy, University of Miami, Coral Gables, USA Steven French, University of Leeds, Leeds, UK Catarina Dutilh Novaes, VU Amsterdam, Amsterdam, The Netherlands Darrell P. Rowbottom, Department of Philosophy, Lingnan University, Tuen Mun, Hong Kong Emma Ruttkamp, Department of Philosophy, University of South Africa, Pretoria, South Africa Kristie Miller, Department of Philosophy, Centre for Time, University of Sydney, Sydney, Australia

The aim of Synthese Library is to provide a forum for the best current work in the methodology and philosophy of science and in epistemology, all broadly understood. A wide variety of different approaches have traditionally been represented in the Library, and every effort is made to maintain this variety, not for its own sake, but because we believe that there are many fruitful and illuminating approaches to the philosophy of science and related disciplines. Special attention is paid to methodological studies which illustrate the interplay of empirical and philosophical viewpoints and to contributions to the formal (logical, set-theoretical, mathematical, information-theoretical, decision-theoretical, etc.) methodology of empirical sciences. Likewise, the applications of logical methods to epistemology as well as philosophically and methodologically relevant studies in logic are strongly encouraged. The emphasis on logic will be tempered by interest in the psychological, historical, and sociological aspects of science. In addition to monographs Synthese Library publishes thematically unified anthologies and edited volumes with a well-defined topical focus inside the aim and scope of the book series. The contributions in the volumes are expected to be focused and structurally organized in accordance with the central theme(s), and should be tied together by an extensive editorial introduction or set of introductions if the volume is divided into parts. An extensive bibliography and index are mandatory.

More information about this series at http://www.springer.com/series/6607

Marco Ruffino

Contingent A Priori Truths Metaphysics, Semantics, Epistemology and Pragmatics

Marco Ruffino Department of Philosophy University of Campinas (UNICAMP) Campinas, S˜ao Paulo, Brazil

ISSN 0166-6991 ISSN 2542-8292 (electronic) Synthese Library ISBN 978-3-030-86621-1 ISBN 978-3-030-86622-8 (eBook) https://doi.org/10.1007/978-3-030-86622-8 © Springer Nature Switzerland AG 2022 This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission or information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology now known or hereafter developed. The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication does not imply, even in the absence of a specific statement, that such names are exempt from the relevant protective laws and regulations and therefore free for general use. The publisher, the authors, and the editors are safe to assume that the advice and information in this book are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or the editors give a warranty, expressed or implied, with respect to the material contained herein or for any errors or omissions that may have been made. The publisher remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

This Springer imprint is published by the registered company Springer Nature Switzerland AG The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland

For Daniela, Ana Luísa and Júlia, my three little stars

Acknowledgements

This book grew out of my interest for the fascinating phenomenon discovered by Kripke in Naming and Necessity (1980) that there might be truths knowable A Priori that are, nevertheless, contingent. At some point, it came to my attention that the kind of cases discussed by Kripke are the product of stipulations, and stipulations are not, from the perspective of illocutionary acts, ordinary assertions. Along the years, I had the fortune to discuss some of the ideas here developed and get critical background from Ernie Lepore, Peter Ludlow, François Recanati, Dirk Greimann, Célia Teixeira, Eros Corazza, Philip Atkins, Ludovic Soutif, André Leclerc, Eleonora Orlando, Genoveva Martí, Manuel García-Carpintero, José Zalabardo, Santiago Echeverri, Sílvio Pinto, Max Fernández de Castro, Max Kölbel, Camilo Vergara, Filipe Martone, Max Freund, Rafael Albiero, Ricardo Santos, Galen Strawson, Jeferson Santos, Thainá Demartini, Iago Batistela, Giorgio Venturi, Emiliano Boccardi, and the now late Maite Ezcurdia and Daniel Vanderveken, among many others. They helped to shape my proposal and to correct many mistakes. (The remaining mistakes are all my own contribution.) I warmly thank all of them. Matheus Valente made a very careful and competent revision of a final draft of the entire manuscript. Arthur Pagani and Gilson Olegario helped with many technical details related to LATEX. Thanks are also due to Otávio Bueno, the editor of the Synthese Library, for his support and patience. My greatest debt, however, is to my wife, Patrícia, for her unconditional love and encouragement through the years. Some of the material presented here appeared in part in earlier publications: • Chapter 3 appeared in part in “Descriptive Reference Fixing and Epistemic Privileges” in Aufklärung 8:123–132 (2021). • Chapter 4 appeared in part in “The Contingent A Priori and De Re Knowledge” in Carlo Penco, Massimiliano Vignolo, Valeria Ottonelli, and Cristina Amoretti (eds). Proceedings of the 4th Latin Meeting in Analytic Philosophy. Genoa: CEUR-WS.org, 2007, pp. 45–58. • Chapter 6 appeared in part in “Superficially and Deeply Contingent A Priori Truths”, Croatian Journal of Philosophy 16 (2):247–266 (2016). vii

viii

Acknowledgements

• Chapter 10 is a slightly modified version of “Contingent A Priori Truths and Performatives”, Synthese 198 (Suppl 22):S5593–S5613 (2021). I thank the editors of the above journals and collections for their kind permission to use some of the printed material here. Research for this book was supported by Grant 2018/17011-9 from the FAPESP and Grant 428084/2018-4 from the CNPq (Brazil).

Introduction

It has been nearly fifty years since Kripke gave the famous series of lectures later published under the title Naming and Necessity (1980). Perhaps the most surprising theses defended in the lectures (or at least the consequences that attracted most attention and sparked most discussion in the literature) were two consequences of Kripke’s considerations about the rigidity and directly referential nature of names. The first is that there might be truths that are necessary but, nevertheless, can be known only a posteriori, i.e. only through some sort of empirical investigation. The prototypical examples are identities involving only ordinary names such as ‘Hesperus is Phosphorus’ and identities involving natural kind terms such as ‘Water is H2 O’, but we can find examples that are not identities, such as ‘This table is composed of molecules’. This thesis is particularly shocking in view of a long and established philosophical tradition of considering apriority an intrinsic aspect of necessary truths. Being necessary, they must be true independently of how the world is and, hence, no empirical investigation concerning any particular feature of the actual world that distinguishes it from other possible worlds would seem to be relevant to convince us of their truth. Kripke’s discovery that things are not so straightforward, and that some necessary truths not only can but must (given our epistemic limitations) be known a posteriori, triggered a revision of our perspective on identities in general, and on basic scientific laws in particular. One vivid illustration of Kripke’s impact on this front is the advent of the so-called two-dimensional semantics—a research program whose main motivation (at least in the writings of leading authors such as Stalnaker, Jackson and Chalmers) is the development of a conceptual and technical framework to accommodate Kripke’s cases of necessary a posteriori truths. The second thesis is that there might be truths that are contingent and, nevertheless, knowable a priori because they are the product of linguistic stipulations. Although it is as revolutionary and shocking as the first thesis, it has been the subject of considerably less attention in the literature and, as I see it, the consequences of the existence of this kind of truth have not yet been fully appreciated. Most of the intense discussion in the literature that followed Kripke’s lectures treats contingent a priori and necessary a posteriori truths as two faces of the same coin or, ix

x

Introduction

to put it differently, two manifestations of the same basic phenomenon.1 However, although most philosophers (at least those impressed by Kripke’s ideas concerning the rigidity of proper names and the distinction between epistemic and alethic modalities) are convinced by the examples of necessary a posteriori truths, there is a tendency of skepticism regarding the thesis and examples of contingent a priori truths, and a widespread feeling that these cases are artificial, built only for the purpose of theoretical discussion.2 Shortly after Kripke’s lectures, there was the appearance of Kaplan’s seminal work (1977) on the semantics of demonstratives and pure indexicals with analogous theses for sentences containing these kinds of expressions. Sentences such as ‘I am here now’ can be known to generate true propositions in any context of utterance, although the propositions must be contingent. From that point on, one issue that dominated the literature concerning Kripke’s cases of contingent a priori truths was whether all cases of such truths are due to the presence of (explicit or implicit) indexicals, so that the whole phenomenon could be seen as part of the supposedly less problematic phenomenon of indexicality.3 The many reactions to Kripke’s examples of (and to the principles that, according to him, are behind) contingent a priori truths are themselves worth studying because they reveal different approaches to many important issues, such as the nature of stipulative reference-fixing, rigid reference, singular thought, de re a priori knowledge, and so on. While I shall review different attempts of dealing with Kripke’s cases of contingent a priori truths, the main purpose of this book is to develop my own perspective regarding his examples in the first place and, more broadly, on the phenomenon of contingent a priori truths as a whole. The perspective that I defend does not see Kripke’s cases as resulting from explicit or hidden

1 For

example, Evans’ (1979) influential discussion considers what he calls deep and superficial contingencies as more or less the same as superficial and deep necessity, respectively. Davies, in the spirit of Evans, characterizes the contingent a priori and the necessary a posteriori as “mirrorimage puzzles” (Davies, 2004, p. 83). 2 For example, this is the view expressed in Evans (1977); Evans’ own reconstruction of the cases of contingent a priori truths is based on the notion of descriptive name, which is a highly artificial kind of name and, as Evans himself recognizes, very rare in ordinary language. Contrasting the cases of necessary a posteriori with contingent a priori truths, Stalnaker says that In Naming and Necessity Saul Kripke presented some striking examples that convinced many philosophers that there are truths that are both necessary and a posteriori and also truths that are both contingent and a priori. The classic examples of the former are identity statements containing proper names (Hesperus=Phosphorus) and statements about the nature of natural kinds (Gold has atomic number 79). Realistic examples of the second kind of statement are harder to come by—perhaps there are none—but once one sees the idea, it is easy to construct artificial examples. (Stalnaker, 2001, p. 141) 3 Many authors tend to be more comfortable with Kaplan’s cases of indexical sentences that produce true propositions in any context. This is so probably because an important part of Kaplan’s theory of indexicals, namely, the distinction between content (i.e. the proposition expressed) and character, has already become the received view, so that apriority and contingency are seen as belonging to distinct dimensions.

Introduction

xi

indexicality, but from the illocutionary aspects of the linguistic acts involved in their creation. It also does not see contingent a priori truths as the “mirror image” of necessary a posteriori one. From this perspective, a posteriori necessity and some (i.e. the most interesting) instances of a priori contingency are seen as separable phenomena, with different roots. Indeed, looking at contingent a priori truths as the result of what Searle and Vanderveken call declarative illocutionary acts, an entirely new perspective on Kripke’s cases opens up where they are neither disguised cases of indexicality nor mirror images of necessary a posteriori truths. There is, or so I shall argue, a more interesting way of looking at such cases. According to my proposal, contingent a priori truths play a fundamental role in the constitution of a framework in which scientific identities may be discovered. Hence, they are far from being a mere artificial curiosity. Moreover, they evince a particular kind of power of language in creating the basis of science by means of stipulations. On the perspective I will develop, the phenomenon of contingent a priori truths is not restricted to some exotic and highly artificial cases derived from the introduction of special proper names in ideal circumstances. Once the mechanism behind the production of such truths is clearly recognized, one cannot but agree that they are a much more widespread phenomenon essential to our cognitive lives. They are fundamental not only in ordinary language, but also in science and civil institutions. Differently from most of the literature on the topic (which, as I said, tends to focus on the issues of rigidity of names versus non-rigidity of ordinary definite descriptions), my own approach focuses on the illocutionary aspects of the speech acts that create such truths. These acts are not (and cannot be) just ordinary assertions and, hence, they do not describe any fact existing independently of the act itself. This has some interesting consequences (or so I shall argue) and helps us in solving some of the inconveniences that critics have pointed out in Kripke’s notion. In particular, this will provide a way of mending some gaps in Kripke’s discussion, as I will point out in Chap. 1. The structure of this book is the following. I start with a more or less standard presentation of Kripke’s classic cases of contingent a priori truths, i.e. the Meter Case and the Neptune Case (so that the text can also be useful for a reader not familiar with the discussion) and point out what I take to be two fundamental gaps left open in his own account of these cases. Kripke’s discussion focuses solely on the perspective of the baptizer, and it is only for the baptizer that he sees the possibility of a priori knowledge of some contingent truths. But, this seems to leave out cases in which we would tend to regard non-stipulators as also having a priori knowledge in virtue of some stipulation. It seems intuitive that we all know some things a priori not because we are the stipulators, but because we accept some institutions or accept the authority of someone who, e.g. fixes some standard of measurement. So, in a more liberal epistemology, we all know some things a priori, provided that we are members of a community that accepts a standard of measurement. Nobody needs to actually go to Paris to measure the standard meter stick in order to know that it is exactly one meter long. Chapter 2 is another more or less standard exposition of Kaplan’s analogous conclusions for some sentences containing pure indexicals and demonstratives. Next, I review the most important perspectives (most of them

xii

Introduction

critical) on these cases previously presented in the literature and eventually point out some reasons for being less than satisfied with each of them. Chapter 3 discusses Donnellan’s acquaintance requirement for de re belief, the basis of his criticism of Kripke’s cases. Chapter 4 discusses a popular family of objections based on the claim that there can be no a priori knowledge in Kripke’s cases because some sort of perceptual contact between the baptizer and the object of the baptism is required. Although this criticism is advanced by a number of authors, I concentrate on the formulations of Plantinga, Salmon and Soames. Chapter 5 presents and discusses Kripke’s own later (1986) reformulation of the contingent a priori in a series of lectures given after Naming and Necessity. Chapter 6 is dedicated to the exposition and discussion of Evans’ distinction between two notions of contingency, which was, as we know, introduced as a way of dealing with what he sees as a “puzzle” represented by Kripke’s cases. Chapter 7 discusses the prospects of yet another contemporary theory (or, better said, family of theories) developed in response to Kripke’s cases, namely, the two-dimensional semantics. From Chap. 9 onwards, I develop my own approach. I first outline some basic elements of the theory of illocutionary acts on which it is based. The most popular taxonomy of illocutionary acts is the one derived from Austin’s (1962) seminal work and later developed by Searle (1969, 1979a) and Searle and Vanderveken (1985). There are, nowadays, other alternative taxonomies (e.g., Roberts, 2018), but since they do not radically depart from Searle and Vanderveken’s one, I will follow their taxonomy (and the method employed in grounding it). Chapter 10 presents my own view on the subject, based on the application of the theory of illocutionary acts to the very act of introducing contingent a priori truths. It has escaped notice to most of the literature concerned with Kripke’s cases that they are generated by linguistic acts that are fundamentally different from normal assertions.4 I will explain why this way of seeing things can fill the gaps pointed out in Chap. 1 and provide a more natural way of seeing this kind of truth as parts of our normal cognitive life. Finally, in Chap. 11 I present Frege’s early view on definitions as an ancestor of the perspective developed in Chaps. 9 and 10. I do not claim that Frege was aware of everything I shall be saying or that he meant it the way I present; but, there is a remarkable resemblance between his early views on the act of definitions and Kripke’s contingent a priori truths under my perspective.

References Austin, J. (1962). How to do Things With Words. Oxford: Clarendon Press. Davies, M. (2004). Reference, contingency, and the two-dimensional framework. Philosophical Studies, 118, 1–2. Reprinted in M. García-Carpintero, & J. Macià, J. (Eds). (2006). Twodimensional semantics (pp. 141–175, pp. 83–131). Oxford: Oxford University Press.

4 Two

exceptions are Jeshion (2002) and Horowitz (1983), although they reach very different conclusions from the ones I shall be arguing for.

Introduction

xiii

Evans, G. (1977). The contingent a priori and rigid designators. Midwest Studies in Philosophy, 2(1), 12–27. Horowitz, T. (1983). Stipulation and epistemological privilege. Philosophical Studies: An International Journal for Philosophy in the Analytic Tradition, 44(3), 305–318. Jeshion, R. (2002). Acquaintanceless De Re belief. In J. Campbell, M. O’Rourke, & D. Shier (Eds.), Meaning and truth. Investigations in philosophical semantics (pp. 53–78). Oxford: Oxford University Press. Kaplan, D. (1977). Demonstratives. An essay on the semantics, logic, metaphysics and epistemology of demonstratives and other indexicals. In J. Almog, J. Perry, & H. Wettstein (Eds.), Themes from Kaplan (pp. 481–563). Oxford: Oxford University Press. Kripke, S. (1980). Naming and necessity. Cambridge, MA: Harvard University Press. Kripke, S. (1986). Rigid designation and the contingent a priori: The meter stick revisited. Exxon distinguished lectures at Notre Dame University. Unpublished. Roberts, C. (2018). Speech acts in discourse context. In D. Fogal, D. Harris, & M. Moss, (Eds.), New work in speech acts (pp. 317–359). New York: Oxford University Press. Searle, J. (1969). Speech acts: An essay in the philosophy of language. Cambridge: Cambridge University Press. Searle, J. (1979a). A taxonomy of illocutionary acts. In J. Searle (Ed.), Expression and meaning: Studies in the theory of speech acts (pp. 1–29). Cambridge: Cambridge University Press. Searle, J., & Vanderveken, D. (1985). Foundations of illocutionary logic. Cambridge: Cambridge University Press. Stalnaker, R. (2001). On considering a possible world as actual. In Aristotelian society supplementary volume 1 (Vol. 75, pp. 141–156).

Contents

1

The Starting Point: Kripke’s “Magic” . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.1 The Meter Case . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.2 The Neptune Case . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.3 Contingent A Priori Truths and Rigidity . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.4 Two Classical Objections . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.4.1 The Scope Ambiguity View . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.4.2 Rigidified Descriptivism . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.5 Two Gaps . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

1 3 5 6 8 8 13 14 16

2

Indexicals and Kaplan’s Cases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.1 Demonstratives and Pure Indexicals. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.2 Character and Content . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.3 Context of Utterance and Circumstance of Evaluation . . . . . . . . . . . . . 2.4 Kaplan’s Modal Argument . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.5 Proper Contexts and ‘I am Here Now’ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.6 Dthat Cases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.7 Explanation: The Two Dimensions of Meaning . . . . . . . . . . . . . . . . . . . . 2.8 Necessary A Posteriori Truths . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.9 Some Partial Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

17 17 19 21 23 24 26 28 30 31 32

3

Donnellan and the Acquaintance Requirement . . . . . . . . . . . . . . . . . . . . . . . . . 3.1 Rigid Names and De Re Knowledge . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.2 The Acquaintance Requirement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.3 Jeshion on Donnellan on Neptune . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.3.1 The De Re Principle and Donnellan’s Problem . . . . . . . . . . . . 3.3.2 Donnellan’s First Loophole. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.3.3 Donnellan’s Second Loophole: Postulating Contingencies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.4 Some Partial Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

35 36 37 40 42 42 45 48 49 xv

xvi

4

5

Contents

The Experience Requirement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.1 A Dialogue Between Plantinga and Philonous. . . . . . . . . . . . . . . . . . . . . . 4.2 Salmon’s Theories . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.2.1 Salmon I: The A Posteriori Only Theory . . . . . . . . . . . . . . . . . . . 4.2.2 Salmon II: The Almost A Priori Theory . . . . . . . . . . . . . . . . . . . . 4.2.3 Salmon III: The Context Shift Theory. . . . . . . . . . . . . . . . . . . . . . . 4.3 Soames and “The Serpent’s Egg” of Two-Dimensional Semantics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.3.1 Characters and Logical Truths. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.3.2 Characters and the Objects of Thought . . . . . . . . . . . . . . . . . . . . . 4.3.3 De Re Propositional Attitudes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.4 Some Partial Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Kripke’s Reformulation of the Contingent A Priori . . . . . . . . . . . . . . . . . . . . 5.1 Elaborating On Donnellan’s Distinction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.2 Three Models of Descriptive Reference-Fixing . . . . . . . . . . . . . . . . . . . . 5.3 The Reply to Donnellan’s Acquaintance Requirement . . . . . . . . . . . . 5.4 Contingent A Priori Truths as the Basis of Other Contingencies. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.5 Some Partial Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

51 51 53 53 57 63 65 66 67 70 74 75 77 77 79 81 85 86 87

6

Evans and the Varieties of Contingency. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89 6.1 Descriptive Names . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90 6.2 Descriptive Names, Free Logic and Evans’ Particular Strategy. . . . 93 6.3 Contingency: Deep and Superficial. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95 6.4 Propositions, Cognitive Content and Evans’ General Strategy . . . . 97 6.5 Contingency and Existence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100 6.6 Some Partial Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105

7

Two-Dimensionalism . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7.1 Stalnaker’s Assertion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7.1.1 The Point of Making an Assertion . . . . . . . . . . . . . . . . . . . . . . . . . . 7.1.2 The Two-Dimensional Semantics . . . . . . . . . . . . . . . . . . . . . . . . . . . 7.1.3 The Diagonal Proposition and Kripke’s Contingent A Priori . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7.1.4 Rational Communication and the Necessary A Posteriori . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7.1.5 The Context-Relativity of Propositional Concepts . . . . . . . . . 7.1.6 Problems With Belief Attributions . . . . . . . . . . . . . . . . . . . . . . . . . . 7.2 Davies and Humberstone’s Alternative Notion of Necessity . . . . . . . 7.3 Chalmers’ Primary and Secondary Intensions . . . . . . . . . . . . . . . . . . . . . . 7.4 Primary/Secondary Intensions and Character/Content . . . . . . . . . . . . . 7.5 Stalnaker’s Later Retraction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

107 108 109 110 112 115 119 120 122 126 131 133

Contents

xvii

7.6 Some Partial Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138 8

Some Other Cases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8.1 LDO Sentences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8.2 Williamson and The Believer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8.3 Best Explanations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8.4 The “Most Unlikely Event” . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8.5 Proper Contexts of Explanation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8.6 Some Partial Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

141 141 143 148 150 152 154 154

9

Basic Tools: Elements of a Theory of Speech Acts . . . . . . . . . . . . . . . . . . . . . . 9.1 The Dimensions of a Speech Act . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9.2 The Classes of Illocutionary Acts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9.3 The Creative Aspect of Declarative Acts. . . . . . . . . . . . . . . . . . . . . . . . . . . . 9.4 Performatives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9.5 Performatives and Truth . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9.6 Performative Utterances as Declaratives . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9.7 Institutional and Linguistic Facts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9.8 Illocutionary Commitment and Inconsistency . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

157 157 159 161 163 166 168 170 174 175

10

Stipulations as Performatives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10.1 Two Gaps . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10.2 Stipulations and Illocutionary Acts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10.3 Horowitz . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10.4 Contingency . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10.5 Closing the Two Gaps . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10.6 Some Partial Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

179 179 180 188 189 192 196 197

11

One Ancestor: The Early Frege on Definitions . . . . . . . . . . . . . . . . . . . . . . . . . . 11.1 Assertion and the Assertion Sign . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11.2 Definitions and the Definition Sign. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11.3 Identities as Synthetic Judgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11.4 Definitions: Turning the Synthetic Propositions Analytic . . . . . . . . . . 11.5 Some Partial Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

199 199 201 202 206 208 209

12

Global Conclusions: The Varieties of Contingent A Priori Truths . . . . 211 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 214

Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 215

Chapter 1

The Starting Point: Kripke’s “Magic”

One of the most interesting results of the so-called “direct reference theories” is the distinction between epistemic and metaphysical modalities. This goes against an old established tradition in philosophy of considering necessary and a priori truths on the one hand, and contingent and a posteriori truths on the other hand, to be on a par. One of Kripke’s lessons in the first lecture of Naming and Necessity (1980) is that the notions of apriority and necessity are not intensionally equivalent since the latter is a metaphysical notion, while the former, having to do with the justification of beliefs, is epistemic. But although these notions are not intensionally equivalent, it could turn out that they are extensionally equivalent, i.e., that all a priori truths are necessary and vice-versa. However, Kripke makes the further point that, if one accepts the thesis that names in natural language are rigid designators (and there are independent, intuitive arguments for this thesis), then the notions of apriority and of necessity are not even extensionally equivalent, i.e., there are cases of necessary truths that can be known only a posteriori, as well as cases of contingent truths that can be known a priori. In this chapter I review the essential aspects of Kripke’s theses and examples of necessary a posteriori and contingent a priori truths. Naturally, I shall concentrate on the contingent a priori cases. Kripke’s text first suggests an example involving Goldbach’s Conjecture1 (pp. 35–6) as a case in which the parallel between apriority and necessity might be seen as collapsing. Presumably, since there is so far neither a proof nor a disproof of the Conjecture, we cannot say that anyone has a priori knowledge of it because, as a rule, knowledge in mathematics requires proof as justification. But since it is a mathematical conjecture, if it is true, it is a necessarily true, and if it is false, it is necessarily false. Now suppose that a very powerful computer could tell us that an incredibly large number is an exception to Goldbach’s conjecture, a number so large that it surpasses the human capacity of effectively doing or checking all the

1 I.e.,

the conjecture that, given any even number n, there are two prime numbers n1 and n2 such that n = n1 + n2 .

© Springer Nature Switzerland AG 2022 M. Ruffino, Contingent A Priori Truths, Synthese Library 443, https://doi.org/10.1007/978-3-030-86622-8_1

1

2

1 The Starting Point: Kripke’s “Magic”

computational steps. In this case, since our confidence that the computer yields the right result is based on a number of hypotheses from physics, electrical engineering, etc., our knowledge that Goldbach’s conjecture is false would have to be considered a posteriori, although we have a necessary falsity here (or a necessary truth, if we take its negation). Although Naming and Necessity is meant to be a discussion of the consequences of the thesis that proper names are rigid designators, this first example does not seem to depend on any particular properties of proper names, being solely based on the nature of mathematical knowledge and computer-based results. This example is less than fully convincing as it stands and can, at best, work as a sort of pressure for the clarification of the notion of a priori. A more convincing case is that of identity statements. If ordinary proper names like ‘Hesperus’ and ‘Phosphorus’ (which were introduced at different times in history as names of the planet Venus) are both seen as rigid designators and directly referential (and part of the task of the first Lecture of Naming and Necessity is to show that this is the perspective that better fits our intuitions as speakers), then an identity-statement like ‘Hesperus is Phosphorus’ expresses the proposition that the objects referred to by the first and by the second names are one and the same. That is to say, if true, the statement says that the planet Venus is self-identical, being therefore a necessary truth. If false, on the other hand, it says of two different objects that they are the same, which is necessarily false. However, this is hardly something that we could know a priori since, as Frege (1892) pointed out when he introduced this kind of example in the contemporary philosophical culture, it was a remarkable discovery that the celestial body named ‘Hesperus’ is the same as the one named ‘Phosphorus’, and one that came as the result of astronomical observation. How could we account for the aposteriority of this statement, given that it is necessary? One explanation is that there are other possible worlds, phenomenologically indistinguishable from ours, in which the celestial body named ‘Hesperus’ is the brightest one visible in the early evening and has the same appearance as our Hesperus, the celestial body named ‘Phosphorus’ is the brightest one visible in the early morning and has the same appearance as our Phosphorus, but in which it turns out that these are not the same. Hence, some empirical research is needed in order to determine which of the phenomenologically equivalent worlds is the actual one. It is crucial to these examples that the names flanking the identity sign are directly referential, and not the abbreviation of non-rigid definite descriptions. If the names involved were abbreviations of such descriptions, then the identity would have the form D.D.1 = D.D.2 . Hence, if D.D.1 and D.D.2 are different descriptions, the statement is not necessary but contingent, since D.D.1 and D.D.2 could capture different objects in different possible worlds. If the descriptions are the same, we have a necessary truth, but it is also a priori. Other examples that convinced many philosophers come from the consideration of properties that are in someway essential to an individual but which cannot be known to be possessed by this individual without some sort of empirical research. Kripke presents the following illustrative case (p. 47): a great amount of scientific investigation is needed in order to discover, e.g., that a certain table is composed of molecules. But given that the table is composed of molecules, there could not be a

1.1 The Meter Case

3

situation in which it were not composed of molecules. That is to say, to be composed of molecules is a necessary property of this table, but we cannot know a priori that it has this property. Hence, if it is true that the table is composed of molecules, it is necessarily true (and, if false, it is necessarily false), but this truth can only be known a posteriori. Notice that, contrary to the case of Goldbach’s Conjecture, this is a case of a necessary truth that can only be known a posteriori, at least by minds such as ours. Kripke’s examples of necessary a posteriori truths are deep, convincing, and gave rise to several interesting revisions in semantics and in our philosophical understanding of science. Regardless of that, we shall leave those interesting cases aside since our focus in this book will be on the (apparently dual) cases of contingent a priori truths. Let us start with two classical examples presented in Naming and Necessity. They both come from the combination of three theses defended by Kripke: (i) Ordinary names (such as ‘Aristotle’) are rigid designators, while ordinary definite descriptions with the same reference (such as ‘the author of the Nicomachean Ethics’) are (normally) non-rigid; (ii) Names are not equivalent in meaning to definite descriptions (or clusters of definite descriptions), being therefore directly referential;2 (iii) Although definite descriptions do not give the meaning of proper names, some name might eventually have its reference fixed by a description.3 Once we understand how Kripke’s classic cases combine these theses, we can produce any number of similar examples based on them.

1.1 The Meter Case Kripke’s first classic illustrative example concerns knowledge of the length of the standard meter stick. The example is borrowed from Wittgenstein (1953, §50), who had introduced it with a different purpose (i.e., showing that statements that establish standards of measurement play a special role in a language game and, hence, are neither true nor false). What we call ‘meter’ was once conventionally established to

2 This

is, as we shall see later in this chapter, a consequence of the first thesis. thesis that Kripke argues for in Naming and Necessity is that, contrary to what Frege and Russell might have though, there is a fundamental difference between the task of fixing the reference of a name (which ordinary definite descriptions or cluster of descriptions can do) and the task of providing a meaning for the name (which definite descriptions or cluster of descriptions cannot do, at least for ordinary names in natural language). In the Fregean framework, senses of proper names do both tasks at once. 3 Another

4

1 The Starting Point: Kripke’s “Magic”

be the length of a particular stick S made of platinum-iridium at a certain time t0 .4 In other words, the term ‘one meter’ (or simply ‘meter’) was defined as the length of S at t0 . If this is so, there is something rather curious about the modal status of the sentence (M) The length of S at t0 is one meter What we have on the left side is a non-rigid definite description (‘the length of S at t0 ’), which means that, although in the actual world it refers to one meter, it could refer to other lengths because S could be longer or shorter than it actually is. On the other hand, ‘meter’ is the proper name of a certain abstract length and, as such, it is referentially rigid, i.e., it refers to the same length in every possible world. When we describe counterfactual situations such as in ‘the same S could have been three meters long’, or ‘the same S could have been just one-tenth of a meter’, we always refer to the same length by ‘meter’. Hence, since ‘The length of S at t0 ’ and ‘meter’ have different modal behaviors, (M) is contingently true. However, given that ‘meter’ was stipulated as being the length of S at t0 , one (i.e., the stipulator) can know (M) without any further appeal to the relevant experience, i.e., without any measurement. This is why Kripke concludes that we can know (M) a priori, although it is only contingently true: What then, is the epistemological status of the statement ‘Stick S is one meter long at t0 ’, for someone who has fixed the metric system by reference to stick? It would seem that he knows it a priori. For if he used stick S to fix the reference of the term ‘one meter’, then as a result of this kind of ‘definition’ (which is not an abbreviative or synonymous definition), he knows automatically, without further investigation, that S is one meter long. On the other hand, even if S is used as the standard of a meter, the metaphysical status of ‘S is one meter long’ will be that of a contingent statement, provided that ‘one meter’ is regarded as a rigid designator: under appropriate stresses and strains, heatings or coolings, S would have had a length other than one meter even at t0 . (1980, p. 56)

The fact that the truth of (M) results from a stipulation, being therefore knowable a priori, generates the illusion that it is also necessary, but its modal status does not follow its epistemic status. It is worth insisting on some crucial features of the example. First, (M) has the form of an identity between two singular terms referring to the same entity (in this case, an abstract length).5 Second, the name ‘meter’ is rigid, while the description ‘the length of S at t0 ’ is non-rigid. If both were rigid designators, the identity would be necessarily true or necessarily false. Third, the definite description appearing

4 Later

the convention changed; nowadays the meter is defined as the length of the distance 1 of a second. 299, 792, 458 5 At this point I am assuming, with Kripke, that definite descriptions are singular terms and ignoring alternative views such as Russell’s and Evans’. Although this and the next examples in particular are identities, there are other examples of contingent a priori truths that are not of this form, as we will later see. travelled by light in a vacuum in

1.2 The Neptune Case

5

in one side of the identity is the one fixing the reference of the name appearing in the other side of the identity. If the description were not playing this reference-fixing role, then the truth of the identity could not be known a priori since some (empirical, if the description is non-rigid) investigation would have to be made to check whether the referents are the same. Finally, the description is only used to fix the reference of the name, but does not give its meaning. If the description had been introduced as a synonym of the name, then the identity would have expressed a necessary truth instead of a contingent one.

1.2 The Neptune Case The second classical example introduced by Kripke explores the historical episode of the discovery (around 1846) of the planet Neptune by the French astronomer Urbain Le Verrier, made even before Johann Gottfried Galle actually observed it at the Berlin Observatory for the first time guided by Le Verrier’s calculations. Le Verrier predicted the existence of Neptune (and the position in which it could be seen) based on some observed perturbations in Uranus’ orbits. In Kripke’s reconstruction of the episode, Le Verrier used the description ‘the planet causing the perturbations in Uranus’ orbits’ to fix the reference of the name ‘Neptune’. According to theses (i)–(iii) above, ‘Neptune’ is rigid, while ‘the planet causing the perturbations in Uranus’ orbits’ is not. And although the description fixes the reference of the name, it is not equivalent to it in meaning. Because he so fixed the reference of ‘Neptune’, there was something that Le Verrier knew a priori about Neptune (without the need of actual observation of it). As Kripke puts it, If Leverrier indeed gave the name ‘Neptune’ to the planet before it was even seen then he fixed the reference of ‘Neptune’ by means of the description just mentioned. At that time he was unable to see the planet even through a telescope. At this stage, an a priori material equivalence held between the statements ‘Neptune exists’ and ‘some one planet perturbing the orbit of such and such other planets exists in such and such a position’ and also such statements as ‘if such and such perturbations are caused by a planet, they are caused by Neptune’ had the status of a priori truths. Nevertheless, they were not necessary truths, since ‘Neptune’ was introduced as a name rigidly designating a certain planet. Leverrier could well have believed that if Neptune had been knocked off its course one million years earlier, it would have caused no such perturbations and even that some other object might have caused the perturbations in its place. (1980, p. 79, n. 33)

We could represent what Le Verrier knew, according to Kripke’s reconstruction, in the following schematic way (where ‘φ’ abbreviates the property of being a planet causing the perturbations in Uranus’ orbits): (N) (∃!x)φx → (Neptune = the x φx) This sentence has a conditional form. The antecedent is an existential clause because, although Le Verrier’s hypothesis of Neptune’s existence was based on solid

6

1 The Starting Point: Kripke’s “Magic”

astronomical evidence, it could be false, i.e., it could be the case that there is no planet causing the perturbations after all, or it could be that more than one planet was responsible for the perturbations. In each case, there would not be one and only one object having the property of being a planet causing the perturbations in Uranus’ orbits. If the antecedent of (N) is true, i.e., if there is one and only one planet satisfying φ, and given that Le Verrier fixed the reference of ‘Neptune’ by using the corresponding definite description, then the consequent is true. If the antecedent is false, (i.e., if there were no such planet or more than one) the conditional is also true, or so Kripke seems to think.6 In either case, (N) is true, and Le Verrier is in a position to know it simply in virtue of having fixed the reference of ‘Neptune’ by means of the definite description ‘the planet causing the perturbations in Uranus’ orbits’. On the other hand, (N) is contingent since it could well be false: there could be a possible world in which the antecedent is true (i.e., in which there is a unique planet causing the perturbations in Uranus’ orbits) and, nevertheless, that planet is not Neptune, and the consequent if false.

1.3 Contingent A Priori Truths and Rigidity As we saw, Kripke’s two classical cases above have the form N = D.D.

6 This is not

clear, however, as Evans (1979b) points out. If the antecedent is false, it is not obvious whether (N) expresses a proposition because, since ‘the planet causing the perturbations in Uranus’ orbits’ has no reference, the reference of ‘Neptune’ has not been fixed. This is the reason for Evans’ requiring that Kripke’s example be formulated in a free logic, for otherwise there would be no guarantee that (N) expresses a proposition which could itself be an object of knowledge. It is illustrative to contrast the case of Neptune’s discovery with that of the hypothetical postulation of another planet by the same Le Verrier: in 1859 he proposed to call ‘Vulcan’ a planet until then unobserved that would presumably be responsible for some observed perturbations in Mercury’s orbits. But later it was discovered that there is no such planet, and the perturbations in Mercury’s orbits are actually explained by the General Theory of Relativity as due to the distortion of space in the vicinity of the Sun. Hence, there is no single planet causing the perturbations in Mercury’s orbits and the description ‘planet causing the perturbations in Mercury’s orbits’ which is the one that Le Verrier used to fix the reference of ‘Vulcan’ actually has no reference. It follows that, despite Le Verrier’s act of baptism, ‘Vulcan’ has no reference either, being rather like a fictional name. In this case, as in the Neptune case, we could so represent what Le Verrier thought was a truth (where ‘ψ’ abbreviates the description ‘planet causing the perturbations in Mercury’s orbits’): (V) (∃!x)ψx → (Vulcan = the x ψx)

Since there is no single object satisfying ψx, the antecedent is false and (V) would be true. However, because of this same fact, the name ‘Vulcan’ has no reference either and, as Evans points out, it is unclear whether (V) expresses a proposition at all. We will come back to this point in Chap. 6.

1.3 Contingent A Priori Truths and Rigidity

7

where N is a rigid name, and D.D. is a non-rigid description that is used to fix the reference of N. Only under these conditions can we claim that the identity is both knowable a priori, because it results from a stipulation, and contingent, because it is true in the actual world, but could be false in other possible worlds. Kripke offers several arguments throughout Naming and Necessity for the claim that ordinary proper names do not have the same content as the descriptions eventually used for their reference-fixing (or as any descriptions whatsoever), the most important of them known in the literature as the Modal Argument. The argument assumes as premise that ordinary proper names are semantically rigid. This is not based on any sophisticated or formalized theory about proper names, but simply on the strong intuition that, whenever they are employed in the description of counterfactual situations, they always tend to hold on to their original reference. E.g., if we try to describe a counterfactual situation by saying: Imagine that Aristotle had never done any philosophy and had, instead, been an ordinary salesman, and that the author of the Nicomachean Ethics had been the most dishonest sophist in Athens; even so the book would have been extremely valuable for moral philosophy. It seems clear that the name ‘Aristotle’ in this description still refers to our Aristotle (the great philosopher who actually wrote the Nicomachean Ethics), despite the fact that we are projecting him in a counterfactual situation in which he lacks most of the properties for which we know him in the actual world. The same does not happen with the description ‘the author of the Nicomachean Ethics’, which we also use to refer to our Aristotle in the actual world but is not committed to him in the described situation, and refers to whoever satisfies the property of being the author of the Nicomachean Ethics in that situation. Given the premise that proper names are rigid in this sense, the Modal Argument proceeds by a kind of reductio of the thesis that a name is equivalent in meaning to some definite description. Suppose that ‘Aristotle’ has the same meaning as a description like ‘the author of the Nicomachean Ethics’. In this case, ‘Aristotle’ and ‘the author of the Nicomachean Ethics’ would have exactly the same modal behaviour, and there could not be a possible world in which ‘Aristotle is the author of the Nicomachean Ethics’ is false. But unless we assume a metaphysical view of Aristotle that regards all of his properties as necessary, it is perfectly conceivable that this could be false. (Our example above seems like a description of a situation that could perfectly well have been the case, metaphysically speaking.) So, if we assume that names are rigid (which seems intuitively correct), the additional assumption (many times referred to as descriptivism in the literature) (D) ‘Aristotle’ is equivalent in meaning to ‘the author of Nicomachean Ethics’ leads to the conclusion that (A) Aristotle is the author of the Nicomachean Ethics is equivalent to

8

1 The Starting Point: Kripke’s “Magic”

(A*) The author of the Nicomachean Ethics is the author of the Nicomachean Ethics which is necessarily true. But this clashes with our modal intuitions since we regard (A) as only contingently true. Alternatively, (D) leads to the conclusion that (A**) ♦(Aristotle is not the author of the Nicomachean Ethics) is equivalent to (A***) ♦(The author of the Nicomachean Ethics is not the author of the Nicomachean Ethics). But (A**) is intuitively true (unless we adopt the strong essentialist metaphysics mentioned above), while (A***) is necessarily false (since it says of a contradiction that it is possible). So we must reject (D). Or, perhaps in more careful way, we should say that the three claims: that names are rigid, that (A) is contingent, and (D) are jointly inconsistent, so one of them must be discharged. Since the first two claims correspond closely to our ordinary intuitions and the third is the product of a theory (roughly, the Fregean theory), Kripke chooses to keep what is intuitive, and to discharge what is the product of a theory.7

1.4 Two Classical Objections The Modal Argument seems like a strong evidence that names do not have the same content as (ordinary) definite descriptions and, hence, must be directly referential. However, some philosophers do not take it to be as decisive as Kripke thinks. There are two classical objections raised in the literature. Although I think that they were both successfully refuted (e.g., by Salmon (1981), Soames (1998), and Kripke (1980) himself), in this context, I shall only briefly present the objections (mainly because the first is at the center of Donnellan’s worries, as we’ll discuss in Chap. 3) and some replies, without a deeper discussion. As I shall mention in the last section of this chapter, even if we assume that Kripke is right and descriptivism (both in its original version, and in the modified versions assumed in the objections) is untenable, there still are some fundamental gaps in his view on contingent a priori truths.

1.4.1 The Scope Ambiguity View The first classical objection was originally formulated by Dummett (1973, pp. 113–5) and later endorsed by Donnellan (1977, pp. 14–6). Dummett challenges 7 Basically the same argument applies, mutatis mutandis, if names are seen as equivalent not with a single description but with a cluster of descriptions, in the way proposed in Searle (1958).

1.4 Two Classical Objections

9

the claim that names introduced through reference-fixing by means of non-rigid definite descriptions are rigid in Kripke’s sense of not being synonymous with those descriptions.8 If a referential name is not synonymous with any definite description then it is directly referential and its semantic content is simply the object referred. Dummett’s criticism is based on the idea that a definite description might take different scopes when the sentence in which it occurs contains a modal operator.9 He remarks that (A***) has in fact two possible readings, depending on how we take the scope of the first definite description ‘the author of the Nicomachean Ethics’ (which supposedly, according to (D), is the one equivalent to the name ‘Aristotle’). The first reading takes the description as having narrow scope (i.e., as being within the scope of the modal operator), something like (A∗∗∗ n ) ♦(the author of the Nicomachean Ethics is not the author of the Nicomachean Ethics) (which looks similar to the original (A***), but here there is an indication that the first description has narrow scope). In the second reading, however, the description is taken as having wide scope (with the modal operator falling within it), and (A***) is understood as (A∗∗∗ w ) The author of the Nicomachean Ethics is such that ♦(he is not the author of the Nicomachean Ethics) which says, of the author of the Nicomachean Ethics in the actual world (i.e., Aristotle), that he could have failed to be the author of the Nicomachean Ethics, which is true. The conclusion of Kripke’s Modal Argument depends on the ∗∗∗ additional assumption that (A***) is to be read as (A∗∗∗ n ) and not as (Aw ). But, so the challenge goes, the additional assumption is not granted. Conversely, Dummett suggests that Frege’s theory of proper names, which he takes as paradigmatically descriptive, can be seen as having the implicit convention that a definite description, whenever abbreviating a proper name, is to be taken as having the widest possible scope in modal sentences (1973, p. 115, p. 128). Dummett’s concern about ambiguity of scope and how it affects Kripke’s Modal Argument had an impact on Donnellan’s discussion in a famous paper (1977) about the relation between rigidity and contingent a priori truths (which will be discussed in more detail in Chap. 3). This is somewhat ironic, since Donnellan, along with Kripke, also champions the idea that proper names are rigid designators. But the fact

8 Dummett also makes a more general point. He thinks that the very fact that Kripke has arrived at examples of contingent truths that are knowable a priori shows that there is something wrong with the idea of rigid designation and the Modal Argument (1973, p. 121), for such combination of modalities should be seen as absurd from the outset. Dummett’s point clearly begs the question against Kripke. 9 Considerations of this kind were first brought up by Russell (1905) in analyzing different interpretations of sentences containing definite descriptions and the negation operator (such as ‘The current king of France is not bald’) or epistemic operators (such as ‘Charles V wanted to know whether Scott is the author of Waverley’).

10

1 The Starting Point: Kripke’s “Magic”

is that he takes Dummett’s concerns as a serious obstacle for the Modal Argument since someone bent on defending the descriptivist interpretation of names may always appeal to the ambiguity of (A***) (Donnellan calls this strategy “evasion” (p. 15)), as Dummett does, to block Kripke’s conclusion. If the intuitive reading of (A**) does not count as evidence for the rigidity of ‘Aristotle’, can there be any such evidence? There certainly is nothing in the syntax of the name that indicates that it should be taken rigidly, for the name can always be seen as syntactically ambiguous. Donnellan thinks that only the intention of the one introducing the name (in our case, the name ‘Aristotle’) could determine whether it is meant to be rigid or descriptive. As he sees it, in the absence of an explicit semantic clause disclosing that intention, Kripke’s Modal Argument is not conclusive and the name could be taken one way or another. These considerations are independent from the thesis that names are rigid designators: what is under discussion is whether names introduced by descriptive referencefixing are necessarily rigid in virtue of this mechanism. They are not, although they may be, but for other reasons, as Donnellan would also agree.10 He has in mind something like Let ‘Aristotle’ be a rigid designator with its referent fixed by ‘the author of the Nicomachean Ethics’ But the fact is that something of the sort hardly occurs in ordinary linguistic practice.11 Hence, that names introduced by descriptive reference-fixing are rigid is, at best, only a theoretical possibility and the putative cases of contingent a priori truths are to be taken, for Donnellan, as just hypothetical cases. They are still worth discussing but only for showing the limits of our notions of knowledge and apriority. If Dummett is right, a proper name like ‘Neptune’ can always be seen as equivalent to a definite description, and a distinctive feature of names qua descriptions is that they always take primary scope (differently from ordinary descriptions, for which there might be ambiguity of scope). This feature of the corresponding descriptions would simulate the rigidity of proper names, although they are in fact equivalent to descriptions. Above we saw that the three following claims are jointly inconsistent: that names are rigid designators, that (A) is contingently true, and (D). While Kripke throws (D) away, Dummett explains away our intuition that names are rigid as an illusion caused by a syntactic feature of proper names qua descriptions.

10 See

Donnellan (1972) for his own reasons for taking names as non-equivalent to definite descriptions. 11 Donnellan is thinking primarily in terms of speaker’s intentions to use a name as rigid or as non-rigid. But the determination might be implicit in the context of the name introduction, or even given by presuppositions: in some contexts the name is supposed to be an abbreviation of a description, and in other contexts it is meant to be rigid in Kripke’s sense, i.e., to hold on to the same object independently of the description originally used in the reference-fixing. After Dummett and Donnellan, some authors expressed the view that whether or not a name is rigid might depend on the specific context in which it is introduced. See, e.g., Lewis (1986).

1.4 Two Classical Objections

11

There are two conflicting perspectives concerning Dummett’s suggestion that descriptivism can resist the Modal Argument by appealing to the scope ambiguity of descriptions.12 One might take it as an artificial move, without a principled justification or independent evidence for the rule that proper names are disguised descriptions that always take wide scope.13 Or one might see it as a convenient way of combining descriptivism (for which there are independent motivations, especially from the epistemic side) with the rigidity of proper names (for which there is the kind of evidence brought up by Kripke).14 So, the existence of a rule that descriptions replacing names always take wide scope would be the best way to reconcile two theses for which we have independent motivations. Kripke addresses the scope ambiguity objection mainly in the preface of Naming and Necessity. His main point is that Dummett’s suggestion misdescribes our intuition concerning rigidity: they are not about scope of proper names or descriptions embedded in modal operators, but about the truth conditions of the propositions expressed by sentences containing proper names. He presents at least two related arguments for this claim. The first is that the intuition persists even in simple sentences with no modal operators such as ‘Aristotle was fond of dogs’: here there is no issue of scope but the intuition persists that the corresponding proposition is true in those worlds in which one particular individual, i.e., “our” Aristotle, is fond of dogs. So, Kripke concludes, it is not that definite descriptions taken with wide scope cannot simulate rigidity, but the fact is that this does not explain our basic intuitions concerning the truth-conditions of sentences containing names.15 The second argument is that in constructions such as • Aristotle is not the author of the Nichomachean Ethics. That might have been the case. • Aristotle is not Aristotle. That could not be the case. the first sentence in each of them has no modal operator, so there is no question of scope ambiguity. Both constructions are true. How do we interpret the referent of ‘that’ in the second part of each construction? If it refers to the proposition expressed by the first part, then the fact that both constructions are true show that the content of 12 Thanks

to an anonymous referee for making this point salient to me. is how I think Kripke sees Dummett’s suggestion. He says in (what I take to be) a sarcastic remark that the tendency to give ‘Aristotle’ a wide scope reading, while other descriptions are given narrow scope reading would be “unaccountable!” (1980, p. 13). 14 This perspective is embraced, e.g., by Sosa (2001). 15 Sosa (2001) disagrees. He claims that, even in simple sentences with no modal operators we have “implicit semantic ambiguity” (p. 4) which might come to light when we outline the corresponding truth-conditions in counterfactual situations. So, even in the case of a simple sentence like ‘The author of the Nichomachean Ethics is fond of dogs’ we can spell out the truth conditions in a counterfactual situation S taking the description either with narrow scope (in which case the sentence is true if and only if the author of the Nichomachean Ethics in S is fond of dogs in S) or with wide scope (in which case the sentence is true if and only if the author of the Nichomachean Ethics, namely, Aristotle, is fond of dogs in S). According to Sosa, Kripke’s considerations about simple sentences with no modal operators show nothing about the relation between names and descriptions. 13 This

12

1 The Starting Point: Kripke’s “Magic”

‘Aristotle’ and ‘the author of the Nichomachean Ethics’ cannot be the same, because the first proposition could be true while the second could not. (If the contents of ‘Aristotle’ and ‘the author of the Nichomachean Ethics’ were the same, we should have the same proposition expressed in the first part of each construction and, hence, they could not be both true.)16 If ‘that’ refers to the sentence of the first part, then what we have in the second part is the first part within quotation marks, and there is no way to apply the scope ambiguity strategy within that context. So, Kripke concludes, Dummett’s objection ultimately misdescribes the intuitive semantics of natural language (of English, at any rate). Soames (1998) has a different (though similar in spirit) reply to the scope ambiguity objection. As he argues, under the scope ambiguity view, some inferences that are normally valid in English would have to be considered as invalid. E.g., take a proposition expressed by ‘n is F’ (where ‘n’ is a name), and assume that it is equivalent to a proposition expressed by ‘the G is F’ (for some description ‘the G’). Now consider the argument: (1) The proposition that if n is F, then something is both F and G is the same as the proposition that if the G is F, then something is both F and G. (2) The proposition that if the G is F, then something is both F and G is a necessary truth. (C) The proposition that if n is F, then something is both F and G is a necessary truth. (1) corresponds to the descriptivist assumption. (2) is a logical truth. (C) should follow from (1) and (2), but under the scope ambiguity view it does not, because (C) is equivalent to (C’) The G is such that the proposition that if it is F, then something is both F and G is a necessary truth which is false if F and G are not logically related, and G does not express a necessary property of the object named by ‘n’. So, for the scope ambiguity view, (1) and (2) are both true, while (C) may be false. But the argument has the following grammatical form: (i) α = β (ii) α is a necessary truth (iii) β is a necessary truth Any argument with this grammatical form in English is valid. Hence, Soames concludes, the scope ambiguity view does not work because it is primarily an attempt to describe the meaning of English sentences containing proper names, definite descriptions and modal operators, but it turns out to characterize valid arguments in English as invalid. 16 Another way to put it is that Dummett’s account in terms of scope ambiguity cannot explain why there is a difference in the first part of each construction (one is possibly true, but the other is not) because there is no possibility of scope ambiguity.

1.4 Two Classical Objections

13

1.4.2 Rigidified Descriptivism The second classical objection to the Modal Argument is based on the idea that names can be treated as synonymous not with ordinary definite descriptions, but with rigidified ones.17 These are descriptions that contain a rigidifier, so that the reference is selected as the individual satisfying the descriptive content in the actual world, and in every other possible world they back to that same individual selected in the actual world. Typically, adding ‘actually’ or ‘in the actual world’ to a description has this effect. (As we shall see in Chap. 2, ‘actual’ is an indexical that refers, in a context C, to the world of that context; moreover, it refers rigidly to that world.) Hence, according to this proposal, the name ‘Aristotle’ might be considered as equivalent to ‘the author of the Nichomachean Ethics in the actual world’ (or ‘the author of the Nichomachean Ethics in w@ ’, for short). The latter expression refers to Aristotle in w@ , but also, in any other possible world w, it refers not to the author of the Nichomachean Ethics in w, but to the author of the Nichomachean Ethics in w@ , being therefore rigid. This alternative seems to preserve the epistemic advantages of descriptivism and, at the same time, to avoid the conclusion of the Modal Argument, because sentences containing proper names and the corresponding rigidified descriptions do express propositions with the same modal behavior. This combination of descriptivism with rigidity is sometimes called in the literature Rigidified Descriptivism. There is also another position that combines Rigidified Descriptivism with a particular view of the relevant rigidified definite descriptions as something of the form (D) the actual entity standing in the appropriate causal-historical relation to this (my) token of the name ‘n’. This is sometimes called Rigidified Causal Descriptivism.18 One additional motivation for this version of descriptivism is that it allegedly has the advantage of escaping not only the Modal Argument (because the description is rigid), but also another famous argument employed by Kripke and known in the literature as the

17 This 18 This

view is embraced, e.g., by Plantinga (1974, 1978), Stanley (1997), and Jubien (1993). is how Lewis states the basic idea of this version of descriptivism:

Did not Kripke and his allies refute the description theory of reference, at least for names of people and places? Then why should we expect descriptivism to work any better for names of colours and colour experiences?. . . I disagree. What was well and truly refuted was a version of descriptivism in which the descriptive senses were supposed to be a matter of famous deeds and other distinctive peculiarities. A better version survives the attack: causal descriptivism. The descriptive sense associated with a name might for instance be ‘the place I have heard of under the name “Taromeo”’ or maybe ‘the causal source of this token: Taromeo’, and for an account of the relation being invoked here, just consult the writings of the causal theorists of reference. (1997, footnote 22)

14

1 The Starting Point: Kripke’s “Magic”

Semantic Argument.19 The basic conclusion of this latter argument is that there is no description (or set of descriptions) that every competent speaker associates with a name such that the latter refers to the unique entity satisfying that description (or set of descriptions). But for Rigidified Causal Descriptivism there is such a description backing every use of a name by a competent speaker, namely, (D). Soames (1998) argues against Rigidified Descriptivism as well. His main point is that, if ‘Aristotle’ means the same as ‘the author of the Nichomachean Ethics in w@ ’, then the belief attribution ‘A believes that Aristotle was a great philosopher’ would be equivalent to ‘A believes that the author of the Nichomachean Ethics in w@ was a great philosopher’. Now, assuming that the belief relation is one between the subject A and the proposition expressed by ‘the author of the Nichomachean Ethics in w@ was a great philosopher’, we would then be attributing to A a belief that has w@ as part of its content.20 But this seems absurd. It seems (or so Soames argues) perfectly possible to think of subjects in other possible worlds having the belief that Aristotle was a great philosopher and this belief does not involve w@ : we can imagine that A holds that belief while being completely oblivious to the existence of other worlds (and to the existence of w@ in particular). The same sort of argument can be applied, mutatis mutandis, against Rigidified Causal Descriptivism: we can think of subjects in other possible worlds having beliefs about Aristotle and these beliefs do not involve either w@ or the actual causal chain connecting that object with my particular token of the name.21

1.5 Two Gaps Quite apart from Dummett’s and Donnellan’s considerations concerning ambiguity of scope and whether or not this affects the rigidity of names (and, hence, the cogency of Kripke’s examples of contingent a priori truths) there are, in my view, two major gaps in Kripke’s account of his own examples of contingent a priori

19 This

argument is famously presented by Kripke (1980, pp. 83–91) in order to show that, as a matter of fact, descriptions are, in most cases, neither sufficient nor necessary for determining the reference of names. Kripke’s two famous examples are of the names ‘Gödel’ and ‘Feynman’. Most people associate with ‘Gödel’ the description ‘the one who proved the incompleteness of arithmetic’. But imagine that we come to discover that Gödel himself is not really responsible for that proof, and that he stole it from someone else (Schmidt); our intuition is that the name ‘Gödel’ would still refers to the same person (“our” Gödel) and not to Schmidt. On the other hand, most people associate with the name ‘Feynman’ only an indefinite description such as ‘a famous physicist’, but this does not prevent the name from referring to Feynman. 20 This must be so in virtue of the semantics of the indexical ‘actual’; as we shall better see in Chap. 2, ‘actual’ is a directly referential term, and its contribution to the proposition expressed by an utterance of a sentence containing it in a context is the possible world of the context. 21 For a critical evaluation of Soames’s argument, see Nelson (2002).

1.5 Two Gaps

15

truths. These gaps, if left open, tend to undermine the credibility of the whole idea of contingent a priori: • First, Kripke is unclear about the nature of the truth-makers of contingent a priori truths. Sometimes he gives an impression, without being explicit, that these are ordinary empirical (i.e., astronomical or physical) facts. This, I think, has been the source of some confusion in the literature concerned with the issue. Some critics point out that ordinary empirical facts cannot be the truth-makers of contingent a priori truths since it is impossible (for beings like us) to know such facts without resorting to some empirical experience. Such truth-makers would have to be facts generated (or first made accessible) by stipulations, but we find no proper explanation in Kripke of how exactly stipulations create (or give access to) such facts or what the latter are like. • Second, there is no explanation for the transmission of a priori knowledge from one speaker to another, and even from one speaker to herself at different times. Kripke himself notices that, in his account, the possibility of contingent a priori knowledge exists primarily for the stipulator (e.g., the meter-baptizer or Le Verrier).22 But what about the knowledge that people other than the stipulator might have that the stick is one meter long (based on the stipulation)? It seems fair to say that they have some kind of a priori knowledge (if they accept the rule created by the stipulator) because they do not need to actually measure the stick to know its length; but how can they know this a priori if they must have seen the stipulator perform the stipulation (thereby relying on information coming from their senses) or must have known of the stipulation from testimony? Or what about the stipulator herself at a time later than the moment of the baptism? It seems that she has to rely on memory of the baptism ceremony and, because of that, the a priori nature of her knowledge might be questioned. Can we give an account of the apriority of this kind of (later) knowledge and, if so, how is it connected with the knowledge that the stipulator has in virtue of her privileged position? Kripke does not address these questions (at least not directly and explicitly). This book is, in part, an attempt to fill these gaps. This is, in my view, crucial to evaluate the plausibility of the very notion of contingent a priori truths, and to show that they are not mere abstract and artificial possibilities conceived only for the sake of philosophical discussion, as Donnellan, Evans and Stalnaker seem to think.23 I think that the notion is very plausible and present in our cognitive lives. But we must, on the one hand, elaborate a view concerning the nature of the facts described by the sentences expressing these truths, for it seems that they are not natural facts (which 22 Remember that, in the passage quoted, before Kripke asks what is the epistemological status of (M), he specifies “for someone who has fixed the metric system by reference to stick S” (1980, p. 56, my emphasis). 23 Again, I stress the contrast between the skepticism and impression of artificiality in the reception of contingent a priori truths among many philosophers with the overall approval and impression of naturalness in the reception of necessary a posteriori truths.

16

1 The Starting Point: Kripke’s “Magic”

can only be learned through experience), but facts depending on language. On the other hand, we must account for the possibility of a priori knowledge not just for the one who stipulates but, in principle, for everyone. The perspective that I am about to present can provide, or so I shall argue, an answer compatible with our intuition that the stipulator retains a priori knowledge at later times, and that people other than the stipulator (and who accept her authority) also have a priori knowledge.24

References Burge, N. (1993). Content preservation. The Philosophical Review, 102(4), 457–488. Donnellan, K. (1972). Proper names and identifying descriptions. In Semantics of natural language (pp. 356–379). Berlin: Springer. Donnellan, K. (1977). The contingent a priori and rigid designators. Midwest Studies in Philosophy, 2(1), 12–27. Dummett, M. (1973). Frege: Philosophy of language. Cambridge, MA: Harvard University Press. Evans, G. (1979b). Reference and contingency. The Monist, 62(2), 161–189. Frege, G. (1892). Über Sinn und Bedeutung, Zeitschrift für Philosophie und philosophische Kritik, 100. Reprinted and translated by P. Geach, & M. Black. In M. Beaney (Ed.), The Frege reader (pp. 151–171, pp. 25–50). Oxford: Blackwell, 1997. Jubien, M. (1993). Proper names. Philosophical Perspectives, 7, 487–504. Kripke, S. (1980). Naming and necessity. Cambridge, MA: Harvard University Press. Lewis, D. (1986). On the plurality of worlds. Oxford: Blackwell. Lewis, D. (1997). Naming the colours. Australasian Journal of Philosophy, 75(3), 325–342. Nelson, M. (2002). Descriptivism defended. Noûs, 36(3), 408–435. Plantinga, A. (1974). The nature of necessity. New York: Oxford University Press. Plantinga, A. (1978). The Boethian compromise. American Philosophical Quarterly, 15(2), 129– 138. Russell, B. (1905). On denoting. Mind, 14(56), 479–493. Salmon, N. (1981). Reference and essence. Princeton, NJ: Princeton University Press. Searle, J. (1958). Proper names. Mind, 67(266), 166–173. Soames, S. (1998). The modal argument: Wide scope and rigidified descriptions. Noûs, 32(1), 1– 22. Sosa, D. (2001). Rigidity in the scope of Russell’s theory. Noûs, 35(1), 1–38. Stanley, J. (1997). Names and rigid designation. In B. Hale, & C. Wright (Eds.), A companion to the philosophy of language (pp. 555–585). Oxford: Oxford University Press. Wittgenstein, L. (1953). Philosophical investigations. London: Macmillan.

24 I do not claim, of course, that this is the only possible answer to the problem of transmission of a priori knowledge through memory or testimony, but only that this is a viable and so far unexplored alternative. For a different approach, see Burge (1993).

Chapter 2

Indexicals and Kaplan’s Cases

The central task of David Kaplan’s classical essay “Demonstratives” (1977) is to develop a formal semantics for indexicals having, as his starting point, some ordinary intuitions concerning the use of these terms. One of the most interesting aspects of Kaplan’s semantics is that it has as consequence a notion of validity that does not coincide with traditional logical necessity. The difference appears more saliently for sentences containing indexicals. Thus, according to Kaplan’s semantics, we can have an analogue of Kripke’s examples of necessary a posteriori and contingent a priori truths, although this time the effect is not due to the properties of proper names, but to the semantic features of indexicals. As we shall later see, some authors such as Evans (1979b) believe that Kripke’s examples are just particular instances of Kaplan’s cases. To understand Kaplan’s account of these semantic features, and how they generate the effect in question, we have to first review the basics of his theory.

2.1 Demonstratives and Pure Indexicals Kaplan’s theory starts with a distinction of indexical expressions between what he calls demonstratives and pure indexicals. Demonstratives are expressions such as ‘this’, ‘that’, ‘there’, etc., that need, as the name suggests, an accompanying demonstration (i.e., a non-linguistic element)1 in order to be semantically complete

1 Typically, a demonstration is a pointing gesture, and this is how the early Kaplan tends to see them. But there are other theoretical alternatives. In later work (1989), he tends to see the demonstrative intention as responsible for semantically completing the demonstratives. For some complications of this alternative, see Perry (2009). Wettstein (1984) considers a set of contextual factors as responsible for making a reference salient. Braun (1996) treats demonstrations in an abstract way such that they can take any of these forms.

© Springer Nature Switzerland AG 2022 M. Ruffino, Contingent A Priori Truths, Synthese Library 443, https://doi.org/10.1007/978-3-030-86622-8_2

17

18

2 Indexicals and Kaplan’s Cases

(i.e., in order to have a reference in a context). We can say that demonstratives have their content (reference or intension, depending on how we see it) fixed by the accompanying demonstration. Demonstratives without demonstrations (e.g., ‘that’ with no pointing in a context in which there are several objects as possible demonstrata, or ‘he’ in a context in which there are several male persons) are semantically incomplete (1977, p. 490). Kaplan reserves the designation pure indexicals for expressions such as ‘I’, ‘here’, ‘now’, ‘today’, etc. which, differently from demonstratives, do not need the completion of demonstrations: the mere occurrence2 of these expressions, in a context, is enough to fix their reference (e.g., ‘I’ said by me, or ‘now’ said at a certain time).3 The demonstration is, abstractly speaking, a non-linguistic element that accompanies the linguistic demonstrative expression. In some special contexts, a demonstration does not actually require any gesture or special action on the agent’s part since it is implicit,4 but this is not the typical situation. Normally some non-linguistic gesture or indication is needed. So, a pure indexical is itself semantically complete, while a semantically complete demonstrative is something of the form D[α] in which D is the demonstrative expression (‘this’, ‘he’, etc.), and α is a demonstration (typically, but not necessarily, a pointing towards the intended object or person). The combination of both, demonstrative expression plus demonstration, when considered in a context, refers to the demonstratum of the demonstration (a person, an object, a location, etc., depending on the nature of the demonstrative expression). A pure indexical, on the contrary, does not have its reference fixed as a

2 Kaplan

makes a subtle distinction between utterances and occurrences of indexicals in contexts (1977, Section XIII, 1989, p. 584). An utterance is a concrete event, which involves the production of a token expression in a context, and takes time, while an occurrence is an abstract notion, it is a combination of an expression with a context. The reason for this distinction is that we need to be able to consider several sentences in a context, but utterances take time, so that it is physically impossible to have several of them in one single context. On the other hand, occurrences are abstract combinations of expressions with contexts, so that we can consider several occurrences in one single context. Given the limited purposes of this chapter, the distinction will not affect our discussion and thus I shall ignore it, speaking interchangeably of utterances and occurrences. 3 “All this is by way of contrasting true demonstratives with pure indexicals. For the latter, no associated demonstration is required, and any demonstration supplied is either for emphasis or is irrelevant. [Footnote:] I have in mind such cases as pointing at oneself while saying ‘I’ (emphasis) or pointing at someone else while saying ‘I’ (irrelevance or madness or what?).” (1977, p. 491) 4 However, a demonstration may also be opportune and require no special action on the speaker’s part, as when someone shouts “Stop that man” while only one man is rushing toward the door. My notion of demonstration is a theoretical concept. I do not, in the present work, undertake a detailed ‘operational’ analysis of this notion although there are scattered remarks relevant to the issue. (1977, footnote 9) The question of what exactly constitutes a demonstration is not a trivial one, and Kaplan does not attempt to provide a full account (neither in this essay nor in its sequel Kaplan (1989)).

2.2 Character and Content

19

demonstratum, but rather automatically as some constitutive element of the context selected by the semantic rule that governs its use.

2.2 Character and Content A second fundamental ingredient of Kaplan’s theory is the distinction between two kinds of meaning that indexicals (both demonstratives and pure indexicals) have. An indexical (such as ‘I’ or ‘today’) is an expression that changes its reference from context to context, where the change is not arbitrary but governed by a semantic rule. That is to say, the phenomenon of indexicality is not the same as ambiguity, where a typographical item might change its referent from context to context without a systematic rule regulating the change. Although the reference changes from context to context, the semantic rule associating a contextual reference to the indexical is one and the same for every context. E.g., the rule governing ‘I’ associates, to each context of utterance, the speaker in that context; the rule governing ‘now’ associates the time of the utterance; the rule governing ‘here’ associates the location of the utterance, etc. Kaplan famously calls the rule governing the reference in each context the character of an indexical. Each indexical has a character that is the same in all contexts, and can be seen as a rule (or, in more formal terms, a function) that associates a reference to each context of utterance. The reference of an indexical in each context might be seen as its extension, but it also might be seen as its intension, depending on how we look at it. This is so because an indexical expression is directly referential (this is one of the main thesis in “Demonstratives”), and, therefore, its semantic contribution for the intension of a sentence containing it is the reference itself.5 In the Carnapian model of intension that is implicit in Kaplan’s approach, an intension is a function that associates an extension to possible worlds. Some intensions might be a constant function in the sense that they associate the same object to every possible world, and so an object might be seen as an extension or as an intension (a constant function).6 The character of a sentence is a composition of the character of its parts. Kaplan calls the intension expressed by a sentence in each context its content, itself being a composition of the contents expressed by its parts. If a sentence S is a composition of indexical and non-indexical expressions, the proposition expressed by S in a context c includes the reference(s) of its indexical part(s) and the intensions of its non-

5 This presupposes the Russellian picture of singular propositions, a conception that Kaplan is happy to endorse for sentences containing indexicals (1977, p. 496). 6 The same holds for proper names in Kripke’s theory: their references might be seen as extensions or as intensions, since they are directly referential. This does not happen in the Fregean picture of proper names: their intensions are senses, always distinct from their references (except in special contexts discussed by Frege in which the sense becomes the reference, but this is quite another matter).

20

2 Indexicals and Kaplan’s Cases

indexical parts, which is the same across every context. Thus, sentences containing an indexical such as • I am a philosopher • Napoleon was defeated here • The meeting starts now will express different contents (i.e., different propositions) in different contexts. These contents are determined by the character of ‘I’, ‘here’ and ‘now’, respectively, and the character of the remaining non-indexical expressions (‘philosopher’, ‘defeated’, ‘meeting’, etc.). The character of a pure indexical is normally a simple rule, known by any speaker, and can be found in the dictionary as the ordinary meaning of the term (e.g., ‘I’ means the subject of the utterance, ‘today’ means the day of the utterance, etc.).7 In the case of demonstratives, however, the character is given not only by the intuitive meaning of the term, but also by the accompanying demonstration.8 Hence, for expressions such as There[α] He[ψ] where α and ψ are demonstrations, the characters are only partially determined by ‘there’ (whose meaning requires a location as value) and ‘he’ (whose meaning requires a male individual as value), and have to be completed by the rules more or less expressed as ‘the demonstratum of α’ and ‘the demonstratum of ψ’. That is to say, the demonstrations α and ψ, which are non-linguistic entities, will be responsible for fixing the reference of the demonstratives (though not, as we will see, for giving their meaning), and when placed in different contexts, will yield different values (locations and persons). The binomial character-content might appear like an extension of the Fregean semantics of sense and reference to indexicals, so that the character could be seen as the Fregean sense of an indexical, and the content as its reference. However, the similarity is illusory for several reasons brought up by Kaplan in his essay.9 One important reason is that one and the same Fregean sense cannot have more than one reference, while in the case of indexicals we have that one character may select different references in different contexts. (E.g., ‘today’ has only one character, with a different and unrepeatable reference each day.) If we wanted to preserve the Fregean analogy here we would have to say that the indexical has a different Fregean sense in each context, i.e., different senses with different references in 7 This

is the view in Kaplan (1977, e.g., in p. 505, p. 520, p. 523). But later (1989, p. 577, footnote 26) he becomes skeptical about the character being completely transparent to competent speakers. For some complications of the identification of character and meaning in the case of demonstratives, see Braun (1996). 8 Kaplan (1977, p. 527). 9 For a similar extensive discussion of the difficulties in reducing the semantic dimensions of indexicals to Fregean sense and reference, see Perry (1977, 1979).

2.3 Context of Utterance and Circumstance of Evaluation

21

each context.10 This would, however, make it hard to identify sense and linguistic meaning of indexicals, as Frege presumably intended to do,11 since there would be as many senses as there are contexts (that differ in relevant ways), while the meaning of an indexical seems to be something unique and quite simple. Moreover, although the Fregean sense of a non-indexical expression is inseparable from it in counterfactual reasoning, this does not happen with the character of indexicals, as Kaplan’s Modal Argument discussed below shows. So, despite its Fregean and Carnapian inspiration, the binomial character/content diverges from the Fregean picture. The character of indexicals cannot be seen as a form of Fregean sense (although the character of non-indexical terms can be identified with such senses); and it diverges from the Carnapian model because the intension and the extension of an indexical in a context coincide (although they remain distinct for non-indexicals).

2.3 Context of Utterance and Circumstance of Evaluation Another fundamental distinction in Kaplan’s semantics of indexicals (and crucial to his cases of contingent a priori) is that between context of utterance and circumstance of evaluation. An occurrence of a sentence containing an indexical in a context expresses a content, i.e., a proposition. This proposition will depend on the context of utterance in a fundamental way, since the value taken by the indexical is an element (speaker, time, location, etc.) of it.12 Once the proposition is produced in a context, it might be evaluated as true or false with respect to the circumstances of evaluation. Although it is widely open to discussion what belongs to the context of utterance or occurrence (i.e., which elements are relevant to determine what is said), Kaplan operates with a simplified and highly technical notion of context that includes only

10 Frege

(1918) suggests something of this sort. is the standard interpretation of Frege. For a dissenting interpretation, see Burge (1979). 12 Kaplan endorses two closely related hypotheses concerning indexicals and contexts of utterances: first, that indexicals always take their values from the context of utterance, and never from another context. E.g., if one employs ‘today’ indexically, the value is always the day of utterance, and never another day from another context; if one employs ‘I’, the value is always the speaker, and never another person from another context. The second hypotheses concerns the technical idea of operators that forces an indexical to take its values from a different context; Kaplan calls such operators “monsters”. The second hypothesis is that, in natural language, there are no monsters. Both hypotheses have been challenged in view of more recent empirical evidence that there actually are monsters in natural language and also that there is a phenomenon of indexical shifts, i.e., linguistic situations in which some indexicals get their value from a contexts different from the one of utterance. See, e.g., Schlenker (2003), Rabern (2013), Deal (2020). 11 This

22

2 Indexicals and Kaplan’s Cases

an agent (the speaker), a location, a time, and a possible world.13 This is so because he wants to incorporate the notion of context in a formal semantics for a formal logic of demonstratives, thus many otherwise relevant pragmatic aspects of real contexts of utterances are left out. The notion of circumstance of evaluation derives basically from the classic idea of using possible worlds as points of evaluation of propositions as true or false, with the accretion of a time. The latter is just a technical option, however. Times are part of the circumstances of evaluation (and not of the content) because Kaplan does not want to include a temporal specification in the proposition expressed by sentences that do not include temporal indexicals (such as, e.g., ‘It is raining’). The reason for this is that he wants to preserve the meaningfulness of temporal operators. Here is his full justification: If we build the time of evaluation into the contents (thus removing time from the circumstances leaving only, say, a possible world history, and making contents specific as to time), it would make no sense to have temporal operators. To put the point another way, if what is said is thought of as incorporating reference to a specific time, or state of the world, or whatever, it is otiose to ask whether what is said would have been true at another time, in another state of the world, or whatever. Temporal operators applied to eternal sentences [. . . ] are redundant. (1977, p. 503)

So, for all theoretical purposes, we can represent a context of utterance as < s, t, l, w > where s is a speaker (the agent of the utterance), t is a time (the time of the utterance), l is a location (the location of the utterance), and w is the possible world in which the utterance takes place (which includes a total world history). A circumstance of evaluation can be represented as < w, t > where w is a possible world, and t is a time. Utterances containing indexicals have a double dependence on possible worlds. First, since a possible world is part of the context, what is said (the content) depends on the possible world of the utterance. Depending on the world of the context, different things might be said. Second, the content has to be evaluated with respect to a circumstance of evaluation, i.e., a possible world together with a time. According to the world of evaluation, what is said is evaluated as true or false.14

13 Later

(1989, p. 585), Kaplan toys with the idea of including other elements in the context such as the addressee and the demonstrata of demonstrations, which would play a similar role as the speaker’s demonstrative intentions. No serious development is given to these ideas in this essay. 14 There is no similar double dependence on time, although a time is an element both of the context of utterance and of the circumstance of evaluation. If the utterance does not include a temporal indexical, the time of the context of utterance will be irrelevant for the content expressed. And if the utterance does involve a temporal indexical, the time of the context will become part of the

2.4 Kaplan’s Modal Argument

23

2.4 Kaplan’s Modal Argument A crucial part of Kaplan’s account in “Demonstratives” is the presentation of a simple argument for the claim that both demonstratives plus demonstrations and pure indexicals, when used in a context, become rigid designators and, moreover, directly referential terms. The argument resembles Kripke’s Modal Argument in some ways, and for this reason it is sometimes referred to as Kaplan’s Modal Argument. Let us first review the argument for pure indexicals, e.g., for ‘I’.15 The character of ‘I’ can be given by a description such as ‘the speaker uttering the word ‘I”. Suppose that a teacher says (C) Imagine that Frege, and not me, were here giving this lecture right now; then he would be speaking and I would be silent just paying attention. What is ‘I’ referring to in (C)? Intuitively, it refers to the speaker (the teacher), and not to Frege, although in the situation described it is Frege the one who speaks (and possibly uses the word ‘I’ in his lecture). This seems to show that it is not the character associated with ‘I’ that is loaded into the proposition that the teacher is projecting into another possible world, but only the content that ‘I’ gets in the context of utterance (i.e., in our example, the teacher). The same kind of argument applies, mutatis mutandis, to other pure indexicals. E.g., consider someone saying (L) Suppose we were not living here, but in Paris; then someone else would be living here. The second occurrences of ‘here’ in (L) seems to refer to the place of utterance, and not to Paris, i.e., it takes its value from the context of utterance, and not from the described counterfactual situation. Again, it is not the character of ‘here’ that enters into proposition expressed in the context of utterance, but the location selected in that context. A similar argument can be given for demonstratives. Suppose that a teacher points (let the gesture be demonstration α) at a student sitting in the first row of chairs in the room who happens to be John, while Paul, his fellow student, is sitting in the last row. The teacher says: (D) Imagine that, instead of him[α], Paul had taken the chair first, then he[α] would have to be sitting elsewhere. Intuitively, the second occurrences of ‘he’ in (D) refers to John although, in the counterfactual situation described, α would have Paul as demonstratum. This seems to shows that what is loaded into the proposition is not the character of ‘he[α]’ (for this would have picked Paul in the counterfactual situation described), but the person

content; this will be an “eternal” content, so that the time of the circumstance of evaluation will now be irrelevant. 15 This is not exactly the way Kaplan presents the argument, but a way that I find instructive. The same holds for the argument for demonstratives.

24

2 Indexicals and Kaplan’s Cases

demonstrated in the context of utterance. If the character is not loaded as semantic value into the proposition expressed by sentences containing pure indexicals or demonstratives, then it must be considered as a mere auxiliary reference-fixing device that is discharged, for semantic purposes, after the reference is fixed. Like Kripke’s, Kaplan’s Modal Argument does not seem to appeal to any sophisticated theory of indexicals nor to any controversial claim, but only to the ordinary speaker’s intuitions. It seems to be a basic feature of our use of pure indexicals and demonstratives that, in modal contexts, they hold on rigidly to the object referred to in the context of utterance, and hence the descriptive apparatus that governs its use is no longer semantically relevant after the reference-fixing. The situation is subtly but relevantly different from that of a rigid definite description such as ‘the successor of the number four’: this description refers to the number five in every possible world in virtue of a mathematical law that is true in all possible worlds. In this case, the description expresses a content in a context (actually, the same content in all contexts, since it contains no indexicals), and this content selects the same extension in all possible worlds. For pure indexicals and demonstratives, differently, once the reference is fixed in the context of utterance, the character becomes irrelevant in other possible worlds. The morals of Kaplan’s Modal Argument is that both pure indexicals and demonstratives alike are directly referential. But since their corresponding characters do not explicitly contain this specification, these characters are, in a certain sense, basically incomplete (1977, p. 505).

2.5 Proper Contexts and ‘I am Here Now’ As we saw, Kaplan’s formal treatment assumes a simplified and very abstract notion of context < s, t, l, w > composed only by a speaker s, a time t, a location l, and a possible world w. In principle, any combination (any quadruple) of such elements would be a context from a technical point of view. However, there seems to be combinations that would be prima facie very odd as contexts for possible utterances, since they are such that, in the corresponding world w, the speaker s is not located at l in t. This seems to be an inappropriate context since an utterance, being a thing that is produced, can only occur if the speaker is at the location and at the time in which it occurs. This motivates Kaplan’s restriction of contexts of utterances to what he calls proper contexts, i.e., contexts in which the speaker s is at the location l and time t in the possible world w. As said before, one basic assumption in Kaplan’s semantics is that indexicals and demonstratives always take their semantic values from the context of utterance, and

2.5 Proper Contexts and ‘I am Here Now’

25

never from a different context. As a result both of the restriction and of this basic assumption, any utterance of (I) I am here now will be true in all contexts because in any (proper) context the speaker must be at the location and time in the corresponding possible world. More precisely, in any context < s, t, l, w > (I) will produce a content that is true in w. This is an almost trivial consequence of the restriction to proper contexts. Similarly with (E) I exist since, in any proper context, there must be a speaker (i.e., the agent) in the corresponding possible world. Nevertheless, in both cases, the content expressed (respectively, that M.R. is in his office at 9 am on August 26, 2020 and that M.R. exists) are contingent truths because, in other possible wolds, I could be elsewhere on August 26, 2020, and it could also be the case that I did not exist. Therefore, we have cases of sentences that must express true contents (propositions) in any context of utterance, but these contents are only contingently true.16 The contingent a priori nature of (I) and (E) is due to the semantic features of the pure indexical ‘I’. We should expect similar effects for other pure indexicals, such as ‘actual’ (which refers, in any context, to its possible world) and ‘elsewhere’ (which selects, in any context, all locations distinct from the one that is constitutive of it). Here are some candidates: I am making an utterance here now I am not making an utterance elsewhere now P if and only if actually P (where P is any true sentence) They all express a true content whenever uttered (i.e., in any context), and in each context the content is a different one, but always contingent. Notice that, interestingly, although (I) expresses a contingent proposition in any context, (I@ )

I am actually here now

expresses a necessary one: for in any possible world it will be true that, in the actual world, the speaker selected by ‘I’ in the context is in the place selected by ‘here’ at the time selected by ‘now’. 16 Sentence (E) plays, of course, a crucial role in Descarte’s famous Cogito argument. It must be true whenever uttered (or, in Descartes’ version, whenever thought). For different speakers, the utterance of (E) generates different contents, but the different contents are all true because the utterance takes place in a (proper) context, and the context requires the existence of a speaker. However, these contents are all contingent, since the speaker’s existence is contingent (unless God is the one thinking (E)).

26

2 Indexicals and Kaplan’s Cases

Although the cases of contingent a priori truths presented above involve indexicals, later in his writings Kaplan identifies a deeper source of these truths that is independent of the presence of indexicals, and has to do only with the relation between a possible world and a context: Any feature of a possible world which flows from the fact that it contains the context of use may yield validity without necessity. (1989, p. 596)

So, things like There is a speaker which is indexical-free, must be true in any possible world containing a context, so it expresses a contingent a priori truth as well.17

2.6 Dthat Cases According to Kaplan, there is a strong analogy between demonstrating an individual and describing the individual from a certain perspective, so that a demonstration can be seen as a kind of description and, vice-versa, a description can be seen as a kind of demonstration (1977, p. 514, 1978, p. 299). In the same way that we can consider a definite description in different counterfactual circumstances, we can also consider a demonstration in counterfactual circumstances. E.g., the same pointing at Venus early in the morning can be considered in a counterfactual situation in which Saturn is the body occupying that same position at that time. (The same demonstration selects distinct objects in distinct contexts.) The pointing would be something analogous to a description like ‘the brightest heavenly body in the direction showed by my finger’. Motivated by this analogy between demonstrations and definite descriptions, Kaplan introduces the semi-theoretical term ‘dthat’ which is meant to work as the term ‘that’ from natural language used demonstratively, i.e., not as in It was predictable that Trump would escape the impeachment but as in Bring me that book which operates on definite descriptions rigidifying them.18 Hence, dthat [the brightest heavenly body visible in the early evening]

17 This case might be seen as similar to Williamson’s (1986) “There is at least one believer”, which I shall discuss in Chap. 8. 18 The linguistic choice of this operator makes sense in English where ‘that’ can be either a demonstrative or a conjunction; in some other languages (e.g., Spanish, German and Portuguese), with different terms for each function, there is no corresponding ambiguous word.

2.6 Dthat Cases

27

works as a rigid designator of Venus, in the same way that dthat[α] (where α is a gesture of pointing at Venus) does.19 For any definite description α, the way ‘dthat[α]’ operates is the following: α selects an object o as reference (i.e., the one that satisfies its descriptive content), and ‘dthat’ rigidifies the reference to o. If α is non-rigid, however, we have a remarkable phenomenon: while ‘dthat[α]’ and α refer to the same object in any context of utterance (since, in that context, the former takes its reference from the latter), they will have different references in other contexts (since the former, but not the latter, is rigid). Hence, (DD)

dthat[α] = α

is true in every context of utterance (if there is an object referred to by α), but the proposition expressed is contingent, since the left-hand is rigid, while the right-hand is not rigid. To take a more concrete example, dthat [the president of the U.S. in 2020] = [the president of the U.S. in 2020] is something that we know to express an a priori truth in any context since both sides of the identity refer to the same individual (in any context having the actual world as constitutive world, both sides refer to Trump, but in a context having as world one in which Hillary Clinton won the 2016 presidential election, both sides refer to Hillary Clinton, etc.), while the truth that it expresses is contingent because whoever is the president of the U.S. in 2020 might not have been the president of the U.S. in 2020. The phenomenon only happens if α is a non-rigid description. On the contrary, if α is rigid, then (DD) expresses a necessary truth, since both sides of the identity refer rigidly to the same object. Examples with this general structure are, in several ways, analogous to Kripke’s meter stick case.20

19 Kaplan later (1989, pp. 579–81) questions this move. ‘Dthat’ is originally conceived as a paradigmatic demonstrative to be used together with a non-linguistic element. But in ‘dthat[α]’, where ‘dthat’ attaches to a definite description α, what we have is something different: ‘dthat’ is a rigidifying linguistic operator and, therefore, the whole complex ‘dthat[α]’ is rigid, but not exactly directly referential, as Kaplan meant in his first formulation. What he had in mind in introducing ‘dthat’ is that it should be seen as a demonstrative surrogate, analogous to a free variable. 20 Something important for Kaplan’s cases are the assumption that logical truth and apriority coincide in his system, i.e., the bearers of logical truth and of apriority are the same entities, and that a demonstrative, when combined with a demonstration in a context, becomes a rigid designator. As we saw, Kaplan provides some independent arguments for the latter claim in Kaplan (1977). I shall discuss Soames’ (2005) criticism of the first assumption in Chap. 4.

28

2 Indexicals and Kaplan’s Cases

2.7 Explanation: The Two Dimensions of Meaning How can there be sentences that express something true whenever used, but always contingent truths? (A similar question must have occurred to any reader of Descartes’ Meditations: how can ‘I exist’ be true whenever believed or just thought of and, nevertheless, express a contingent truth?) Kaplan’s explanation is that apriority and necessity have to do with different semantic dimensions of (sentences containing) indexicals, i.e., character and content. The character of an expression is, as we saw, a general rule that associates an extension to each context of use and, thereby, a content to the sentence containing it. The content, on the other hand, is the proposition associated with the sentence in a context (in the case of an indexical, the content is its reference), and it is the content that is evaluated in different circumstances of evaluation. The character of an indexical remains the same in all contexts, although the content might change from context to context. Kaplan’s idea is that the character is what we access a priori or a posteriori, while the content is what has the property of being contingent or necessary. Therefore, the bearers of apriority/aposteriority and of necessity/contingency are different entities (1977, p. 539). In the particular case of contingent a priori cases involving indexicals, this interpretation is in line with Kaplan’s separation between two epistemic roles that, in the Fregean theory, are played by the same entity (i.e., the thoughts): What we must do is disentangle two epistemological notions: the objects of thought (what Frege called “Thoughts”) and the cognitive significance of an object of thought. As has been noted above, a character may be likened to a manner of presentation of a content. This suggests that we identify objects of thought with contents and the cognitive significance of such objects with characters. (1977, p. 530)

That is to say, we have something parallel, though not identical, to Kripke’s separation between epistemic and alethic modalities. Kaplan’s distinction between cognitive significance and objects of thought is already situated at the epistemic level. Why is it the character, and not the content, that has cognitive significance? This is so because it is the character that is most relevant in individuating cognitive states insofar as these states are responsible for motivating action.21 For example, if Paul is the chair of a meeting in which I take part, he and I may be thinking exactly the same proposition. But I think it as ‘he should speak and open the meeting’, and he thinks it as ‘I should speak and open the meeting’ (i.e., I think of Paul under the character of

21 In this particular point, Kaplan’s considerations and conclusions are very close to Perry’s (1977, 1979). He quotes (p. 532) and endorses the following passage from Perry:

We use the manner of presentation, the character, to individuate psychological states, in explaining and predicting action. It is the manner of presentation, the character, and not the thought apprehended, that is tied to human action. (1979, p. 494)

2.7 Explanation: The Two Dimensions of Meaning

29

‘he’, and Paul thinks of himself under the character of ‘I’). But, although thinking the same proposition, Paul and I act in different ways: Paul starts speaking, and I remain silent, waiting for my turn to talk. The same proposition, captured under distinct characters, leads to two different actions (and, as far as actions are motivated by cognitive states, we have two distinct cognitive states). If, on the contrary, Paul and I are both thinking ‘I am thirsty’, then we both reach for a glass of water. Here we are thinking two distinct propositions (I am thinking one about myself and he is thinking one about Paul), but both captured under the same character (‘I am thirsty’), and thus the same action is performed (backed by the same cognitive state for me and for Paul). This sort of consideration is enough for Kaplan to identify the character as the bearer of cognitive significance. But it is less clear which epistemic role is then left for contents as objects of thought. One would expect that entertaining an object of thought would presuppose some sort of familiarity with the entire proposition, but Kaplan does not think that this is necessary. Rather, he endorses a highly abstract and liberal notion of objects of thought according to which a subject might entertain a singular proposition without knowing which objects are its components. In this conception, it suffices for having as object of thought a proposition containing an object o that a subject employs an indexical referring to o, and this is so even if the subject does not know which particular object o is nor that the indexical refers to o. I.e., this conception dispenses with the requirement of acquaintance with the component of singular propositions. It is known in the literature as semantic instrumentalism.22 The following example is meant to illustrate the possibility of having thoughts without knowing what the thought is about: A kidnapped heiress, locked in the trunk of a car, knowing neither the time nor where she is, may think ‘It is quiet here now’ and the indexical will remain directly referential.[Footnote:] Can the heiress plead that she could not have believed a singular proposition involving the place p since when thinking ‘here’ she didn’t know she was at p, that she was, in fact, unacquainted with the place p? No! Ignorance of the referent is no excuse. (1977, p. 536)

But this very example also makes us feel the counterintuitiveness of this position. Imagine that the kidnapped heiress had to go to the police office afterwards and report every belief she had during the kidnapping, so the police could have some clue to investigate the case. She would be unable to report the one that she had while

22 The

label “Instrumental” was coined by Kaplan himself for the thesis that

[W]e succeed in thinking about things in the world not only through the mental residue of that which we ourselves experience, but also vicariously, through the symbolic resources that come to us through our language. (1989, p. 604) I.e., the thesis comprises not just indexicals, but any device of direct reference (such as proper names). By employing such devices, we can have thoughts about their referent without being acquainted with them. See, e.g., Jeshion (2010) for a discussion of semantic instrumentalism. See Martone (2016) for a separation between conditions of reference and conditions of singular thought.

30

2 Indexicals and Kaplan’s Cases

locked in the trunk at the moment she thought ‘It is quiet here now’. Nevertheless, the corresponding proposition is an object of thought for Kaplan (and this is so because “Ignorance of the referent does not defeat the directly referential character of indexicals” (ibid.)). This seems to indicate that Kaplan’s notion of objects of thought is primarily a semantic one, related to the reference of one’s words rather than to the psychological states associated with those words. For this reason, it is harder to see why this notion has to be one of the epistemic dimensions of indexicals.

2.8 Necessary A Posteriori Truths There is also the dual phenomenon, i.e., of necessary truths that can only be known a posteriori. Consider the following sentence: dthat[α] = dthat[β] (where α and β are non-rigid definite descriptions to which we attach the ‘dthat’operator). As we know, ‘dthat[α]’ has, in any context of use, the same reference as α, and similarly for ‘dthat[β]‘ and β. Hence, the following is a logical truth: (dthat[α] = dthat[β]) ↔ (α=β) On the left side of the biconditional we have an identity between two rigid designators that, if true, is necessarily true; on the right side we have an identity between non-rigid definite descriptions. Now dthat[α] = dthat[β] is a priori if and only if α=β is also a priori, but the latter cannot be a priori, since some empirical research is needed to find out whether the object selected by α in a possible world is the same as the object selected by β in that world; hence, dthat[α] = dthat[β] is both a posteriori, and necessary. The example can be changed using ordinary demonstratives, e.g., he[pointing at someone in an old picture] is he[pointing at someone in a recent picture]. The example is in several aspects similar to Kripke’s ‘Hesperus=Phosphorus’, since here we also have an identity between rigid designators whose truth cannot be known a priori (in Kaplan’s case, we cannot know a priori that the demonstrata of two different demonstrations is the same).

2.9 Some Partial Conclusions

31

2.9 Some Partial Conclusions Although there are some symmetries between Kaplan’s and Kripke’s cases of contingent a priori truths (e.g., they both involve the combination of rigid and nonrigid referring devices, with the non-rigid being used to fix the reference of the rigid one) there are also some important differences (besides the triviality that Kaplan’s cases involve indexicals, while Kripke’s cases only involve proper names). First, in Kripke’s cases we have one single entity (i.e., the proposition expressed by sentences such as (M) and (N)) that is both contingent and a priori. But in Kaplan’s cases, we have two distinct entities: the character (which is not a proposition) as the bearer of apriority, and the content (which is a proposition) as the bearer of contingency. A second relevant difference is that, in Kripke’s cases, the proposition is known to the speaker because of the act of stipulative referencefixing. That is to say, the proposition becomes true by an act on the speaker’s part, while in Kaplan’s cases there is no such act involved (except perhaps the act of producing the utterance itself). Actually, the role of the stipulative act is even more fundamental in Kripke’s cases: a sentence such as (M) containing a new name only comes to express a proposition after the stipulation is made; if no specific stipulation is made concerning the proper name ‘meter’, then (M) expresses no proposition at all since, in that case, the name would be meaningless. But in Kaplan’s cases, there is no stipulation properly speaking because the character of the indexicals is not created anew by the speaker in the occasion of use (except, perhaps, in a situations in which a speaker introduces a new indexical, thereby stipulating a new character). No special act of reference-fixing is required on the speaker’s part to generate a contingent a priori truth; this is just a product of the way indexicals work. As we saw before, the contingent a priori in Kaplan’s semantics depends on two fundamental assumptions. The first is the restriction of contexts to proper contexts, i.e., those in which the speaker is at the location and time of the utterance in the corresponding possible world. If there is no such restriction, then things like (I) I am here now are not a priori true, because there is always at least one improper context in which the content is false in the corresponding world (a world in which the speaker is not at the location or time of the utterance). But Kaplan’s main motivation for this restriction is exactly to have sentences such as (I) coming out true when uttered in any context: Intuitively, [(I)] is deeply, and in some sense, which we will shortly make precise, universally, true. One needs only understand the meaning of [(I)] to know that it cannot be uttered falsely. No such guarantee applies to [‘David Kaplan is in Portland on 26 March 1977’]. A Logic of Indexicals which does not reflect this intuitive difference between [both sentences] has bypassed something essential to the logic of indexicals. (1977, p. 509)

On the face of it, the decision to restrict contexts to proper ones is meant as a way of preserving the special status of sentences like (I) and, hence, it should be no surprise

32

2 Indexicals and Kaplan’s Cases

that they come out true in any context of utterance, despite the fact that they express a contingent proposition in each context. We find no other independent motivation in Kaplan’s work for placing the aforementioned restriction. In fact, there seems to be more motivation for the opposite, that is, abolishing it and thus admitting improper contexts. Kaplan himself, in a rather brief footnote (1977, p. 491, footnote 12), recognizes that there might be some exceptional uses of pure indexicals in messages that are meant for future reproduction (e.g., a recording of ‘I am not here now‘ left in an answering machine, in which the ‘now’ is not meant to capture the time of the recording, but an indeterminate time of reproduction). No serious development is given to this remark in “Demonstratives”. However, some critics took cases such as these (e.g. delayed messages) as presenting strong reasons for abandoning the restriction to proper contexts.23 The second fundamental assumption is that pure indexicals and demonstratives always take their semantic values from the context of utterance, and never from another context. E.g., I cannot say ‘I am getting bored’ and have you, the reader, as semantic value of ‘I’. This assumption is related to Kaplan’s thesis that there are no “monsters” in natural languages and has also been challenged in the literature. Some critics take uses of indexicals in fiction as evidence that sometimes the relevant context of interpretation of indexicals is not the context of utterance, but an intended context.24 The differences between Kaplan’s and Kripke’s cases raise a suspicion that what has been treated in the literature under the label “contingent a priori truths” are actually two distinct phenomena.

References Braun, D. (1996). Demonstratives and their linguistic meanings. Noûs, 30(2), 145–173. Burge, T. (1979). Sinning against Frege. The Philosophical Review, 88(3), 398–432. Deal, A. R. (2020). A theory of indexical shift: Meaning, grammar, and crosslinguistic variation. Cambridge: MIT Press. Evans, G. (1979b). Reference and contingency. The Monist, 62(2), 161–189. Frege, G. (1918). Der Gedanke. Eine logische Untersuchung. Beiträge zur Philosophie des deutschen Idealismus, 2, Reprinted and translated by P. Geach, & M. Black, in M. Beaney (Ed.), The Frege reader, Oxford: Blackwell, 1997, pp. 325–344, pp. 58–77.

23 For a discussion of delayed messages and whether they can be accommodated in Kaplan’s original semantics, see Vision (1985); Sidelle (1991); Predelli (2005); Ruffino (2012). 24 Predelli (2005, pp. 40–75) refers to the basic Kaplanian semantics plus this assumption about semantic values always coming from the context of utterance as the Simple Minded View. As his terminology indicates, he thinks that the assumption does not reflect the complexity of some uses of indexicals.

References

33

Jeshion, R. (2010). Singular thought: Acquaintance, semantic instrumentalism and cognitivism. In R. Jeshion (Eds.), New essays on singular thought (pp. 105–149). New York: Oxford University Press. Kaplan, D. (1977). Demonstratives. An essay on the semantics, logic, metaphysics and epistemology of demonstratives and other indexicals. In J. Almog, J. Perry, & H. Wettstein (Eds.), Themes from Kaplan (pp. 481–563). Oxford: Oxford University Press, 1989. Kaplan, D. (1978). Dthat. In P. Cole (Ed.), Syntax and semantics. Reprinted in Martinich (ed.): The philosophy of language. New York: Oxford University Press, 1996, pp. 292–305, Academic Press, pp. 221–243. Kaplan, D. (1989). Afterhoughts. In J. Almog, J. Perry, & H. Wettstein (Eds.), Themes from Kaplan (pp. 565–614). Oxford: Oxford University Press. Martone, F. (2016). Singular reference without singular thought. Manuscrito, 39(1), 33–60. Perry, J. (1977). Frege on demonstratives. The Philosophical Review, 86(4), 474–497. Perry, J. (1979). The problem of the essential indexical. Noûs, 13(1), 3–21. Perry, J. (2009). Directing intentions. In The philosophy of David Kaplan (pp. 187–201). New York: Oxford University Press New York. Predelli, S. (2005). Contexts: Meaning, truth, and the use of language. Oxford: Oxford University Press. Rabern, B. (2013). Monsters in Kaplan’s Logic of Demonstratives. Philosophical Studies, 164(2), 393–404. Ruffino, M. (2012). Deferred utterances and proper contexts. Disputatio, 4(34), 807–822. Schlenker, P. (2003). A plea for monsters. Linguistics and Philosophy, 26(1), 29–120. Sidelle, A. (1991). The answering machine paradox. Canadian Journal of Philosophy, 21(4), 525– 539. Soames, S. (2005). Reference and description: The case against two-dimensionalism. Princeton University Press. Vision, G. (1985). I am here now. Analysis, 45(4), 198–199. Wettstein, H. (1984). How to bridge the gap between meaning and reference. Synthese, 58(1), 63–84. Williamson, T. (1986). The contingent a priori: Has it anything to do with indexicals? Analysis, 46(3), 113–117.

Chapter 3

Donnellan and the Acquaintance Requirement

It is somewhat ironic that one of the most influential objections to Kripke’s thesis about the existence of contingent a priori truths based on facts about rigidity comes from Keith Donnellan who is, along with Kripke, one of the most important defenders of the direct reference theory for proper names. In a classical paper on the issue, “The Contingent A Priori and Rigid Designators” (1977), he articulates a view that was later followed, if not in the letter, at least in the spirit, by many other critics.1 Donnellan’s most general point is that semantic facts (in particular, facts about rigid designation) alone are not sufficient to guarantee the sort of knowledge concerning particular objects that Kripke was interested in. For there to be any genuine knowledge of propositions about an object something else is necessary besides stipulative reference-fixing: some sort of direct epistemic contact (i.e., some sort of acquaintance) between the stipulator and the object. Donnellan’s criticism was followed by a counterattack from Jeshion (2001) based on the complaint that Donnellan has not considered (and, therefore, not excluded) an alternative explanation for the problem that he detects in Kripke’s cases, an alternative that is consistent with descriptive reference-fixing and with the presence of genuine de re knowledge. In this chapter I shall review Donnellan’s arguments against Kripke and express some misgivings about Jeshion’s counterattack. This is not because I think that Donnellan’s argument is completely successful, but because there is a different way of reading some of his remarks that, on the one hand, avoids her main objections and, on the other, actually makes possible a more positive appreciation of Kripke’s examples.2

Part of this chapter appeared in my “Descriptive reference fixing and epistemic privileges”, Aufklärung 8:123–132 (2021). 1 Donnellan 2 This

credits Levin (1975) with the basic line of thought followed in his article. alternative reading anticipates in some ways the perspective presented in Chap. 10.

© Springer Nature Switzerland AG 2022 M. Ruffino, Contingent A Priori Truths, Synthese Library 443, https://doi.org/10.1007/978-3-030-86622-8_3

35

36

3 Donnellan

3.1 Rigid Names and De Re Knowledge As mentioned in Chap. 1, Donnellan is sensitive to Dummett’s misgivings about the rigidity of proper names introduced by descriptive reference-fixing, i.e., to the claim that the intuition that names are rigid can be explained away by considering them as equivalent to descriptions that always take wide scope in modal contexts. However, he still thinks that the idea that proper names are rigid designators remains as a theoretical possibility worth discussing, since it could, in principle, be the case that such names were introduced with the explicit clause that they are to be taken as rigid designators. So, the kind of contingent a priori truth described by Kripke is also a theoretical possibility. But given that rigid proper names explicitly introduced by definite descriptions are a theoretical possibility, does it follow that contingent a priori truths in Kripke’s sense are also a theoretical possibility? Donnellan argues for the negative answer. Central to his argument is a distinction between (i) knowing that a sentence is true and (ii) knowing the truth that a sentence expresses. Although normally, when we know that a sentence is true this is in part because we know what it means, in some cases this might not be the case. I.e., it can happen that we know that a sentence is true without knowing which truth (i.e., which content) it expresses. To illustrate, we can think of a situation in which, being completely ignorant of Mandarin, I am informed by a competent speaker that a certain sentence appearing in an old Chinese manuscript is true. I do not have the slightest idea of which truth it expresses, but relying on the competent speaker, I may know that that particular sentence, whatever it means, is true. The kind of knowledge that I have in these cases is purely metalinguistic, i.e., only related to the properties of words and sentences, but not to non-linguistic facts. In order to have more than purely metalinguistic knowledge I must somehow understand the sentences and, so Donnellan thinks, have some sort of cognitive contact with the elements of the proposition that they express. In the case of a sentence containing a proper name, since the name is supposed to be directly referential, the putative knowledge must be de re.3 Donnellan avoids giving a more detailed account of de re knowledge, but briefly characterizes it as knowledge such that a belief report would allow both for substitution salva-veritate and for existential generalization in the position occupied by the proper name. This is so because de re knowledge relates the knowing subject directly to an object and not to concepts or descriptions of it. Although avoiding giving a full detailed account of de re knowledge, Donnellan outlines what he describes as two “loose principles” for counting an item of knowledge as de re (1977, p. 22). Later in the paper he argues that Kripke’s cases fail to satisfy these principles. (The reason why these principles are called “loose” is that Donnellan later in his text adds a curious footnote to the effect that these principles might not work in some cases of genuine de re knowledge. We will come 3 This is classically presented by contrast with de dicto knowledge, which requires a relation not with an object, but only with a proposition about that object.

3.2 The Acquaintance Requirement

37

to this footnote later.) The principles are not directly concerned with the relation between a knowing subject and the objects of knowledge, but are rather related to some constraints on knowledge reports. The principles are: (I) A true report ‘S knows that n is φ’ (formulated in an idiolect in which ‘n’ is a name of an object or person and ‘φ’ names a property) is of de re knowledge only if, in any other idiolect that contains a name ‘m’ for the same object or person and a translation ‘ψ’ for ‘φ’, the report ‘S knows that m is ψ’ is also true. (In other words, the report is of de re knowledge with respect to some object or person only if its truth does not depend on the particular name for that object or person.) (II) A true report ‘S knows that n is φ’ (formulated in an idiolect in which ‘n’ is a name of an object or person and ‘φ’ names a property) is of de re knowledge only if one can substitute ‘that’ (demonstrating the same object) or ‘you’ (in the presence of the person) for ‘n’ and the new report is also true. The second principle is actually a version of the first for indexicals. Taken together, they imply that a necessary condition for de re knowledge is that a true report of it must be insensitive to the particular name or indexical used to designate the object of knowledge.4 Donnellan formulates these principles for knowledge, but they could as well, mutatis mutandis, be formulated for weaker propositional attitudes such as belief or for just entertaining a singular thought: a belief or a thought concerning a particular object is de re only if a true report of that belief or thought does not become false under replacement of the name in the original report with another name (or indexical) for the same object.5

3.2 The Acquaintance Requirement Donnellan claims that Kripke’s Neptune case does not satisfy the first principle. He imagines a scenario in which there are inhabitants in Neptune who have, in their own idiolect, a different name (‘Enutpen’) for their planet. These inhabitants know that their planet is responsible for the perturbations in Uranus’ orbits. From there, they observe Le Verrier through a powerful telescope (before Le Verrier was able to actually observe Neptune) and see that he makes the stipulation that ‘Neptune’ refers to the planet causing the perturbations in Uranus’ orbits (if there is one). Now the way we would report Le Verrier’s knowledge is (L) Le Verrier knows that Neptune is the cause of the perturbations in

4 Donnellan explicitly characterizes these requirements as necessary; he is silent about whether they are sufficient as well. 5 See, e.g., Schiffer (1977, p. 29) for a more general formulation.

38

3 Donnellan

Uranus’s orbits but the translation of this report among Enupteneans in their idiolect is (L∗ ) Le Verrier knows that Enutpen is the cause of the perturbations in Uranus’ orbits. While (L) seems true, its translation (L*) seems intuitively false. (Remember, the report is made before Le Verrier observed the planet and had any experience besides naming it). This is a sign, or so Donnellan thinks, that the truth or falsity of the report depends on the particular name involved instead of depending solely on the object named and, hence, according to Principle (I), it cannot be considered as a report of genuine de re knowledge. Somewhat curiously, Donnellan does not apply the test to Kripke’s meter case, but to a case adapted from Kaplan (1978) in which we have the reference of a name fixed by applying the operator ‘dthat’ to a description without knowing its reference. Suppose that someone, call him John, takes the definite description ‘the first child born in the twenty-second century’.6 John has no idea of who this child will be, but decides (assuming there will be one with this characteristic) to call it ‘Newman I’ (if it exists). Hence, the definite description is used to fix the reference of ‘Newman I’, although the latter, being rigid, directly refers to that future child (whoever it is).7 Now, let us assume that in our idiolect we can truly say (J) John knows that Newman I is the first child born in the twenty-second century Now imagine that, about a 100 years from now, John’s grandson comes across the first person born in the twenty-second century, whose real name is ‘Pamela’, and tells her (J∗ ) My grandfather knew that you would be the first person born in the twenty-second century or (J∗∗ ) My grandfather knew that your name is ‘Newman I’.

6 In this case, not only we have no idea concerning its reference, but the reference does not even exist yet. 7 Actually, there is an additional assumption here that, even if one ignores the referent of a definite description α, the complex ‘dthat[α]’ is a rigid designator, although the object referred is ignored. As we saw in Chap. 2, Kaplan endorses this assumption as a consequence of the Instrumental Thesis.

3.2 The Acquaintance Requirement

39

Both (J*) and (J**) seem intuitively false (and this is not because we replaced the name ‘John’ with ‘my grandfather’), although (J) might perhaps be true. Therefore, or so Donnellan concludes, the first report cannot be taken as showing that John has genuine de re knowledge about Newman I (who turns out to be Pamela).8 Despite the negative results in the de re requirement test, we have the impression that both Le Verrier and John do have some sort of knowledge in virtue of the stipulation; after all, in both cases, the original report (formulated in the baptizer’s idiolect) seems to be intuitively true. This is the point in Donnellan’s discussion where he uses the previously introduced distinction between knowing that a sentence is true and knowing which truth it expresses: Le Verrier knows that the sentence Neptune is the cause of the perturbations in Uranus’ orbits is true solely in virtue of a successful linguistic stipulation, but he has no knowledge concerning the truth expressed by this sentence. Similarly for John concerning Newman I is the first child born in the twenty-second century. In both cases, according to Donnellan, what the baptizers have is the metalinguistic knowledge that a certain sentence is true, but not the knowledge of the truth that it expresses. Donnellan concludes: In the absence of any other explanation of why these principles should fail in these cases I suggest that the reason is that the stipulations have not given rise to any knowledge (other than of linguistic matters). And so not to any knowledge a priori. (1977, p. 22)9

Later we will see that it is incorrect that there is “absence of any other explanation” for the failure in these cases. What else would be necessary to warrant knowledge of the content in these cases? Donnellan suggests that some sort of epistemic contact between Le Verrier and Neptune is required. Since the propositions expressed is singular, this contact has to be somehow direct, and not mediated by descriptive contents, coming close to what Russell called acquaintance. Donnellan himself avoids using this term, perhaps because of the obscurity surrounding Russell’s original conception of acquaintance. But the literature after Donnellan (e.g., Jeshion, 2010) has referred to his position as belonging to the family of Acquaintance Theories. 8 Despite Donnellan’s claim, this is not completely clear. A possible perspective that avoids his conclusion is that the Newman I case involves a double baptism. We could know a priori that Newman I is the first person born in the twenty-second century, and a posteriori that Pamela is the first person born in the twenty second century, and afterwards discover that Newman I = Pamela, in the same way that the same mountain can receive different names when seen from different perspectives. The fact that someone was re-baptized as ‘Pamela’ does not eliminate the fact that she was also baptized ‘Newman I’ before that, and the knowledge that comes with this baptism. 9 For a similar point, see Blackburn (1984, pp. 333–6).

40

3 Donnellan

Curiously, Donnellan considers no similar example of test-failure for the meter case. Why is that? Let us try to construct such a case and see how it would work. Suppose there is one baptizer (call her Julia) of the meter using the same meter stick S and uttering (M). It seems correct to formulate the report Julia knows that the length of S at t0 is one meter. Suppose now that there is another community that uses the same stick S as standard of measurement, but they baptized its length as ‘one retem’. Consider now their report Julia knows that the length of S at t0 is one retem. Would this report be false? If we assume that Julia had contact with S, then the second report does not seem to be false. This is so because, by having perceptual contact with S, she also had contact, or so we seem to be allowed to assume, with something more abstract, namely, the length of the stick, which is not identical with the stick, but is given to perception together with the stick, and it is to this abstract length that both ‘meter’ and ‘retem’ refer. If, on the other hand, we think of a situation in which Julia does not perceive the stick, but has only a description of a platinum-iridium stick in Paris with such and such properties, than, as it seems, Donnellan would take the second report to be false for the same reason that the second reports in the Le Verrier and the Newman I cases were false. As I shall argue in the next chapter, however, this conclusion is not granted.

3.3 Jeshion on Donnellan on Neptune Donnellan’s paper raises some fundamental concerns about Kripke’s claims regarding contingent a priori truths based on stipulative reference-fixing. In her paper “Donnellan on Neptune” (2001), Robin Jeshion expresses some doubts about the real effectiveness of Donnellan’s argument. In this section, I examine Jeshion’s account and criticism of Donnellan’s argument. Her point is that Donnellan overlooks a loophole that makes his argument inconclusive. I will argue that Jeshion herself also overlooks a second loophole in Donnellan’s paper that leaves her own criticism inconclusive as well. But before discussing Jeshion’s criticism, it is worth mentioning another important point made by her that is, as far as I can see, independent from her overall evaluation of Donnellan.10 As she points out, although most of the perplexity caused by Kripke’s cases was originally derived from the fact that we have, as a result of stipulations, apriority combined with contingency, the same perplexity must be

10 Jeshion

(2000, pp. 301–2, 2001, pp. 113–4).

3.3 Jeshion on Donnellan on Neptune

41

present in some cases where we have, as a result of stipulations, apriority combined with necessity. Let us imagine, for example, that before the discovery of any sample of the element with atomic number 121 a researcher stipulates that ‘Angelesium’ is to refer to the element with atomic number 121. Then, assuming the essentialist perspective that having atomic number 121 is a necessary property of this element, the researcher can have a priori knowledge of the necessary truth that Angelesium is the element with atomic number 121. This is a metaphysically necessarily true singular proposition, the knowledge of which is normally a posteriori but for the baptizer it comes, so to speak, “for free” just in virtue of a stipulation. This should be just as perplexing as the case of a priori knowledge of contingent truths. Of course this possibility only represents a problem if we assume the Millian perspective of natural kind terms (i.e., that ‘Angelesium’ is directly referential), for otherwise the necessary truth under consideration would be something like a tautology expressed by ‘the element with atomic number 121 is the element with atomic number 121’. The Angelesium example involves the assumption of essentialism about natural kinds, and it is a case of a priori knowledge of necessary truths that are typically known in an a posteriori way; but other examples can be produced of a priori knowledge of necessary truths that are also typically known in an a priori way without assuming essentialism about natural kinds. Suppose, e.g., that we decide to 1000 baptize as ‘Jupe’ the first prime number greater than 10001000 . We have no idea 11 concerning which number in particular Jupe is; but if ‘Jupe’ is a rigid designator, as the Millian thinks, then ‘Jupe is a prime number’ expresses a necessary singular proposition. We can have thoughts with this proposition as content, although we have no idea about which particular number Jupe is, and that proposition is only made available to us because of the stipulation.12 As Jeshion says, the perplexity should not come from finding cases in which we have contingency combined with aprioricity, but from the fact that the stipulator might, merely as a result of the stipulation, have knowledge of some previously unknown propositions. (Again, Donnellan’s argument is directed at cases of putative knowledge, but his worry could be rephrased, mutatis mutandis, for other weaker propositional attitudes such as belief or as simply entertaining thoughts inaccessible before the stipulation.) Cases of necessary a priori truths such as those mentioned above seem to raise the same kind of perplexity as in the cases of contingent a priori truths, which suggests that the problematic aspect of such truths is not the particular combination of modalities but the fact that a stipulation, by itself, can open epistemic access to propositions otherwise inaccessible to our thought.13 Or 11 The largest known prime number as of January 2020 is 282,589,933 −1, which has only 24,862,048

digits. Jupe, whatever it is, is much, much larger. 12 Notice that we do not have here any existential worries concerning the existence of the reference

(as in the Neptune case) since Jupe necessarily exists. 13 Sutton (2001, p. 257) argues that what is really paradoxical in the situation is not the a priori nature of the known proposition, but that the stipulator gains epistemic access “much more easily than others do”, even if the outcome is a posteriori knowledge. Hawthorne and Manley (2010, p. 58) present a variation of the Neptune case that could be seen as an example of contingent

42

3 Donnellan

perhaps they are accessible under extensive empirical or mathematical investigation, but the stipulation provides a considerable shortcut.

3.3.1 The De Re Principle and Donnellan’s Problem The Newman I and the Enutpen situations are meant to illustrate cases in which the stipulator apparently has de re knowledge concerning the object named in the reference-fixing, but this appearance is illusory because the knowledge report may be true using one name for the object of belief, and false using another name (or demonstrative) for the same object. If there were genuine de re knowledge in these situations, Donnellan thinks, the reports should have been insensitive to the particular term employed. In both situations we have a conflict with the “loose” principles that he presents as a necessary condition for de re knowledge. Remember that Principle (II) is a version of Principle (I) for indexicals, so we can combine both in one single principle dealing at once with names and indexicals (i.e., the report must allow substitution salva veritate of any directly referential singular term for the original term). I follow Jeshion and call this combination the De Re Principle. And I shall call Donnellan’s Problem the fact that there is a strain of the De Re Principle in cases like Newman I and Neptune. Donnellan’s own explanation for Donnellan’s Problem is that there is no genuine de re knowledge in these cases. As we saw, the impression that there is some sort of knowledge involved comes, for him, from the fact that we have some purely metalinguistic knowledge that ‘Newman I is the first child born in the twenty-second century’ expresses a true proposition, and that Le Verrier has purely metalinguistic knowledge that ‘Neptune is the cause of the perturbations in the orbits of Uranus’ is true. But knowledge that a sentence is true does not imply knowledge of the truth that it expresses.

3.3.2 Donnellan’s First Loophole As we saw in the exposition of Donnellan’s argument, he characterizes the De Re Principle (or the two principles that we combined in one) as “loose”. And this is not, as one could expect, because of the vagueness of notions such as knowledge or de re but because, as he admits (and Jeshion bases her main argument on this), there are exceptions, i.e., cases presenting Donnellan’s Problem that are, nevertheless, cases of genuine de re knowledge. In a footnote he makes a brief remark concerning such exceptional cases:

(astronomical) a posteriori knowledge based on stipulation and that would, according to them, be no less puzzling basically for the same reasons.

3.3 Jeshion on Donnellan on Neptune

43

I have in mind the “Hesperus” and “Phosphorus” kind of cases. According to the usual story the ancient Babylonians used these as names for what they took to be two distinct heavenly bodies but which was in fact one, Venus. They believed—in fact, I think we can say, knew—that Hesperus was the first heavenly body to appear in the evening, but they would have denied that they believed this of Phosphorus. Did they nevertheless know this about Phosphorus, which, after all, is Hesperus? And should we say about them that they knew about Venus, though they did not call it by that name, that it was the first to appear in the evening? The structure of this sort of case that gives rise to the problems does not seem to me to be present in the “Neptune” and “Newman I” examples and so there seems no reason why the “loose” principles that I give should not apply to them if they do in fact present examples of knowledge. Also, I do have some temptation, at any rate, to say of the Babylonians that they knew about Venus, etc., but I have no corresponding temptation in the “Neptune” and “Newman I” examples. (1977, Footnote 22)

As we perceive from the passage, Donnellan sees a fundamental difference between cases involving descriptive reference-fixing (such as Kripke’s cases) and cases having a Frege’s Puzzle structure, although both kinds of cases exhibit Donnellan’s Problem. But he does not explain the difference, and there is no further development of the idea in the rest of his paper. Actually, there are two gaps in Donnellan’s account suggested in the footnote: (1) He does not explain the exceptionality of cases with a Frege’s Puzzle structure, i.e., why they might constitute cases of genuine de re knowledge despite the fact that they exhibit Donnellan’s Problem; (2) He offers no reasons for not counting cases involving descriptive referencefixing as having the same Frege’s Puzzle structure (being, therefore, also exceptional). That is to say, Donnellan leaves open, without explanation, a loophole for cases exibiting Donnellan’s Problem that are, nevertheless, cases of genuine de re knowledge. There is also no explanation for his claim that Kripke’s cases do not pass through the loophole. Jeshion criticizes Donnellan for both gaps. But the first one is easier to close by supplying an account compatible with Donnellan’s framework. She herself proposes such an account in terms of perspectives or guises that a subject might have towards the object of knowledge (or belief) and which, furthermore, she believes that Donnellan could endorse as well. Once she has this account, she moves on to the second gap (which is more problematic), and argues that there is no reason for keeping cases of descriptive reference-fixing apart from Frege’s Puzzle cases. This means that, contra Donnellan, the strain of the De Re Principle observed in the Neptune and Newman I cases are not unequivocal signs that there is no de re knowledge (or belief) involved in them. Here is a little more detailed exposition of Jeshion’s account of each gap. She uses as illustrations two situations in which someone identifies a person under one name, but not under another: In the first situation, someone sees a person on a TV show being called by the name ‘Smith’, and does not recognize that person as the same one that she knows

44

3 Donnellan

personally under the name ‘Grandpa Joe’. On the TV show Smith is presented as the first person born in the twentieth century. In the second situation, the same subject proposes to name ‘Oldman I’ the first person born in the twentieth century without knowing that the person who has this property is the same one that she knows personally under the name ‘Grandpa Joe’. The second situation is slightly (but very relevantly) different. In both situations we have Donnellan’s Problem, i.e., in both there seems to be a strain of the De Re Principle. The first case exhibits a Frege’s Puzzle structure, because the subject does not know that Smith = Grandpa Joe, although both names refer to the same person. And the second, but not the first, is a case of descriptive reference-fixing (i.e., ‘Oldman I’ is stipulated as being co-referential with ‘the first person born in the twentieth century’). Regarding the first gap in Donnellan’s account, Jeshion uses the first situation (in her example, ‘Petunia’ is the name of the protagonist in both situations) and proposes the following explanation: Petunia has two ways of taking the object of her belief, the one, Wtv , arising from watching α on the TV program, the other, Wg , arising from interactions she had with α at family gatherings, holidays, and weekend visits to his house. When she addresses α, saying ‘I believed that you are the first person born in the twentieth century’, she thinks of α under the Grandpa guise Wg , and fails to identify the object of Wtv and the object of Wg . (2010, p. 121)

She suggests that Donnellan would agree with this explanation.14 Regarding the second gap in Donnellan’s account, Jeshion claims that the explanation for the strain in the second situation is the same as for the strain in the first situation. It seems correct, at least from a Millian perspective, to attribute a de re belief in the first situation (after watching the show) because ‘Smith’ and ‘Grandpa Joe’ are both directly referential, despite the fact that this situation is a case of Donnellan’s Problem. The explanation for Donnellan’s Problem is that both cases involve two different guises. And Donnellan, according to her, is not justified in seeing an asymmetry between them: Donnellan offers us no supplementary argument for ruling out this alternative explanation. He says that the structure exhibited by Hesperus-Phosphorus cases does not seem to be present in the Neptune-type case, but why should we believe that? Or rather, there is no reason why it should fail to seem to be present unless one is antecedently committed to the idea that via the stipulative act, the stipulator attains no non-meta-linguistic de re belief at all. (2010, p. 121)

Jeshion not only claims that there is no reason for excluding Kripke’s cases as instances of situations with a Frege’s Puzzle structure, but she also proposes an argument to the effect that all descriptive reference-fixing must generate cases of this kind: 14 It is not quite clear what exactly these different guises are, and how they differ from Fregean senses; I take it that Jeshion is relying on some traditional Millian account of guises such as Salmon’s (1986, Chap. 8).

3.3 Jeshion on Donnellan on Neptune

45

Now, if we assume that the stipulation gives rise to the non-meta-linguistic de re belief about α that α is F, we have a special case of Frege’s Puzzle—a single belief content is uninformative (and directly a priori justified) to the stipulator but informative (and justified empirically or by proof) to the rest of us. In other words, and this is just the point I have been pressing, if the stipulator can have the relevant non-meta-linguistic de re belief, there will always be a Frege’s Puzzle structure in place due to the alternative guises under which stipulators and non-stipulators grasp the relevant proposition. (2001, pp. 122–3)

In other words, she takes the stipulation that underlies the descriptive referencefixing as creating one additional perspective under which one can think of the named object, namely, the one of the stipulator, and this generates a Frege’s Puzzle structure with other alternative guises (e.g., the one available to non-stipulators), and the identities between these perspectives is always informative. That is to say, we could have de re knowledge in the second situation as well, even in the presence of Donnellan’s Problem. In summary, Donnellan’s perspective is that in cases of descriptive referencefixing we have Donnellan’s Problem and, therefore, no genuine de re knowledge. Cases that exhibit a Frege’s Puzzle structure (like the Hesperus-Phosphorus case) escape this diagnosis, and we have Donnellan’s Problem coexisting with de re knowledge. Jeshion’s perspective is that there is no justification for keeping descriptive reference-fixing apart from Frege’s Puzzle cases. She thus concludes that Donnellan’s argument is at best inconclusive regarding the prospect of de re contingent a priori knowledge in Neptune and Newman I cases. In the next subsection, I speculate about Donnellan’s reasons for keeping descriptive referencefixing cases and cases with Frege’s Puzzle structure apart.

3.3.3 Donnellan’s Second Loophole: Postulating Contingencies Donnellan sees something really crucial in the fact that, in the Neptune and similar cases, and contrary to the cases exhibiting a Frege’s Puzzle structure, one of the names is being fixed by a stipulation. This implies that the Babylonians cannot be taken to be ignorant of the fact that Phosphorus is the first heavenly body to appear in the evening sky because they have stipulated that the name ‘Phosphorus’ refers to the first heavenly body to appear in the evening sky. There is an important passage (which was briefly considered, but quickly dismissed by Jeshion as an ill-considered remark) in which we can see a second loophole in his account: [B]ecause I think it somewhat illuminating to do it this way, I am going to propose instead that we think of the introduction as consisting of stipulating that a certain sentence shall express a contingent truth. If we want to introduce the name “N” by means of the description of “the φ” then the formula we would use would be: (a) Provided that the φ exists, let “N is the φ ” express a contingent truth.

46

3 Donnellan It is a condition on the stipulation that the φ exists and should it turn out that it does not, the stipulation, we might say, has been an unhappy one and not to be taken as being in effect. (1977, p. 19)

How can that be so? We normally take stipulative descriptive reference-fixing as being something of the form (N) Let ‘N’ refer to the φ but Donnellan suggests that it can also be taken as (N∗ ) Let ‘N is the φ’ express a contingent truth and he considers “somewhat illuminating” to do it so. We should notice that (N) is a stipulation concerning a name and, hence, purely metalinguistic, while (N*) does not seem to be metalinguistic, but is a stipulation concerning the status of a proposition.15 Jeshion protests: One thing [Donnellan] does tell us is misguided. In Kripke’s cases, the individual who introduces the name always makes a stipulation (either explicitly or implicitly) of the following form: Kripke Stipulation [KS]: Let ‘N’ refer to the F. Though Donnellan acknowledges that this is the standard form for the relevant set of stipulations, he claims that the following sort of stipulation may be substituted for the standard one: Donnellan Stipulation [DS]: Let ‘N is the F’ express a contingent truth. Donnellan claims that, though there may be problems with substituting [DS] for [KS], [DS] serves as a useful heuristic device for recognizing his thesis that stipulative descriptive reference-fixing engenders only a priori knowledge of the meta-linguistic truth: ‘N is the F’ expresses a contingent truth. [. . . ] This will not do. By stipulation one cannot make it true that ‘N is an F’ expresses a contingent truth. The modal status of the proposition is not something that avails itself to stipulation. (2001, Footnote 8).

Jeshion is certainly correct that the stipulator cannot alter the modal status of a proposition. And she is also correct that the content of (N) (which is metalinguistic) is not the same as the content of (N*) (which is not purely metalinguistic). But this is not, I think, Donnellan’s main point in the passage. Or, at least, not the only way to read it, and certainly not the most charitable one. As I see it, the emphasis of Let ‘N is the F’ express a contingent truth 15 At some points in the text, Donnellan seems to think that these two kinds of stipulations are equivalent; but they are not clearly so.

3.3 Jeshion on Donnellan on Neptune

47

should not be read in contingent but in truth, and the truth of the propositional content can be achieved if the stipulator performs a successful declarative16 illocutionary act. Such acts are not ordinary assertions, and they have as illocutionary point that of making a propositional content true by means of the very illocution. One cannot fix by stipulation that a proposition is contingent, but one can fix by stipulation that a contingent proposition is true.17 The point is that, in the Neptune case, we do not have an ordinary assertion of an identity that is true or false independently of the assertion; as the passage on contingent truth by stipulation suggests, Donnellan sees the Neptune case as one in which the identity is made true by the very linguistic act, hence, it must be a declaration from the perspective of speech act theory. If the act is unsuccessful, then, as Donnellan says in the last sentence of the quoted passage, “the stipulation, we might say, has been an unhappy one and not to be taken as being in effect”.18 That could explain why the Neptune case is not a case like Hesperus-Phosphorus (which would be the content of an ordinary assertion), and has to be treated separately.19 Another way of presenting the same point would be the following: Jeshion assumes (and takes Donnellan to do so as well) that there are distinct ways (or “guises”) of taking the objects of belief corresponding to the names ‘Hesperus’ and ‘Phosphorus’, and these perspectives explain the apparent strain of the De Re Principle, for there should presumably always be a strain in such principle whenever there is more than one perspective associated with the same object of belief. (There is always the possibility that an epistemic agent is inclined to assent to ‘a is φ’, but not inclined to assent to ‘b is φ’ even if ‘a’ and ‘b’ co-refer and are directly referential.) However, it is doubtful that things should be that simple if the two perspectives are brought together by an act of stipulation. In this case, we must consider an identity of the form a = the φ as made true by a successful act of stipulation. Since ‘the φ’ is a non-rigid designator, the identity above is a contingent one, and its content is first made true by the stipulation as a declarative act. I.e., it is not an ordinary identity that is true or false independently from the assertion. Regarding the latter, one might

16 I am using

the terminology of Searle and Vanderveken (1985) and Searle (1979b), which will be better explained in Chap. 9. 17 As we shall discuss in Chap. 9, there are contingent propositional contents that can be made true by the very utterance. E.g., ‘you are my lawyer’ can be made true if I decide to nominate you my lawyer, or ‘you are fired’ can be made true if I have the required authority over you. If the content is first made true by the utterance, it must be contingent, for otherwise it would be true anyway and the utterance would be simply irrelevant. 18 Donnellan employs here something resembling Austin’s vocabulary for stipulations as illocutionary acts. 19 In Chap. 10 we shall also discuss whether the declaration can ever be successful in the Neptune case. But this is a separate matter.

48

3 Donnellan

conclude with Jeshion that there is always the question of multiple perspectives and, hence, that there is a Frege’s Puzzle structure that explains the strain of the De Re Principle. But regarding the former, we have a special illocutionary act with some effects different from mere assertions.20 If this perspective is correct, we can see why Donnellan refuses to subsume cases like Neptune under the cases with Frege’s Puzzle structure of the first loophole: the latter are cases in which we have ordinary assertions of identities; but the former are cases in which the truth of an identity is a matter of stipulation by means of a different speech act. We can also see that Donnellan is wrong in saying “I am not in any way suggesting that this represents some sort of analysis of any practice that we have ever engaged in” (1977, p. 19), for there is a practice that we engage in of making stipulations by means of declarations, thereby making some propositional contents true. (He might as well have considered the act “unhappy” or without any effect if the stipulator lacks the corresponding authority to establish the truth of the relevant propositional content.) If we look at (N*) having in mind its main feature as an illocutionary act, it must be seen as a declarative act. One thing to be noticed is that the propositional content of such acts is typically contingent (for, otherwise, the proposition would not be first made true by the act itself, since it would, as a necessary truth, already have been true anyway). So, ‘contingent’ might be redundant in Donnellan’s formulation of (N*) above. Actually, ‘true’ is also redundant, and Donnellan could simply have formulated it as (N∗∗ ) Let N be the φ with ‘provided that the φ exists’ as a felicity-condition of the act.21 If this is so, then Donnellan can resist Jeshion’s claim that the exact same Frege’s Puzzle structure is present in cases of stipulative reference-fixing as it is in cases like Hesperus-Phosphorus. I do not claim that he articulates this view, but his paper has the elements for doing so.

3.4 Some Partial Conclusions My conclusion is that Jeshion is partly wrong: Donnellan cannot see the same structure in the Neptune case as in the Hesperus-Phosphorus case because, in the former, but not in the latter, we have a proposition made true by a stipulation, which

20 It

is curious that, in this paper, she quotes her own Jeshion (2002) in which she does recognize that stipulative reference-fixing must be a performative. But she does not develop this insight any further except in discussing the felicity condition of existence of the reference. 21 Again, it is quite a different question whether such an act can ever be successful in the Neptune case.

References

49

involves a declarative act (and not an assertion). So, she offers the wrong diagnosis of what is wrong with Donnellan’s argument. My point is that she fails to fully appreciate Donnellan’s second loophole, which is the suggestion that the stipulation is not an ordinary case of identities having a Frege’s Puzzle structure. But Donnellan is also partly wrong, since he thinks that making a contingent propositional content true by stipulation is absurd: Surely only God, if even He, could perform the miracle of stipulating how the world shall be. (1977, p. 19)

But it is not. This view is based on a narrow view of contingency,22 that leaves out an important class of contingent facts, namely, those generated by successful performances of declarative illocutionary acts. As Searle puts it, If somebody says, “The meeting is adjourned,” “I pronounce you husband and wife,” “War is declared,” or “You’re fired,” he may succeed in changing the world in the ways specified in these utterances just by performing the relevant speech acts. (1989, p. 548)

Some contingent contents might be the subject of a stipulation in the sense that they can be made true by the very stipulation (provided some preparatory conditions are in place), and there is nothing mysterious or miraculous about this. Of course not all contingent propositional contents can be made true by stipulation; we cannot, e.g., successfully stipulate that the sky is red, or that there is a sample of gold in my pocket, but some contents can.23 The standard meter case is a good example of a contingent content that can be made true by stipulation. But, of course, one cannot stipulate an astronomical fact. This brings us to a crucial difference between the Neptune case and the meter case, which I shall discuss in greater detail in Chap. 10. Acknowledgements I thank the editors of Aufklärung for their kind permission to use some of the printed material in this chapter.

References Blackburn, S. (1984). Spreading the word. Groundings in the philosophy of language. Oxford: Oxford University Press.

22 At the very beginning of the paper Donnellan makes clear the notion of contingency that he has in mind:

If a truth is a contingent one then it is made true, so to speak, by some actual state of affairs in the world that, at least in the sorts of examples we are interested in, exists independently of our language and our linguistic conventions. How can we become aware of such a truth, come to know the existence of such a state of affairs, merely by performing an act of linguistic stipulation? (1977, p. 13) 23 This

might require a special sort of authority of the stipulator as a preparatory condition.

50

3 Donnellan

Donnellan, K. (1977). The contingent a priori and rigid designators. Midwest Studies in Philosophy, 2(1), 12–27. Hawthorne, J., & Manley, D. (2010). The reference book. New York: Oxford University Press. Jeshion, R. (2000). Ways of taking a meter. Philosophical Studies, 99(3), 297–318. Jeshion, R. (2001). Donnellan on Neptune. Philosophy and Phenomenological Research, 63(1), 111–135. Jeshion, R. (2002). Acquaintanceless De Re Belief. In J. Campbell, M. O’Rourke, & D. Shier (Eds.), Meaning and truth. Investigations in philosophical semantics (pp. 53–78). Oxford: Oxford University Press. Jeshion, R. (2010). Singular thought: Acquaintance, semantic instrumentalism and cognitivism. In R. Jeshion (Ed.), New essays on singular thought (pp. 105–149). New York: Oxford University Press. Kaplan, D. (1978). Dthat. In Syntax and Semantics, ed. by P. Cole, Reprinted in Martinich (ed.): The Philosophy of Language, New York: Oxford University Press, 1996, pp. 292–305, Academic Press, pp. 221–243. Levin, M. (1975). Kripke’s argument against the identity thesis. The Journal of Philosophy, 72(6), 149–167. Salmon, N. (1986). Frege’s puzzle. Atascadero, CA: Ridgeview Publishing Company. Schiffer, S. (1977). Naming and Knowing. Midwest Studies in Philosophy, 2(1), 28–41. Searle, J. (1979b). Speech acts and recent linguistics. In J. Searle (Ed.), Expression and meaning: Studies in the theory of speech acts (pp. 162–180). Cambridge: Cambridge University Press. Searle, J. (1989). How performatives work. Linguistics and Philosophy, 12(5), 535–558. Searle, J., & Vanderveken, D. (1985). Foundations of illocutionary logic. Cambridge: Cambridge University Press. Sutton, J. (2001). The contingent a priori and implicit knowledge. Philosophy and Phenomenological Research, 63(2), 251–277.

Chapter 4

The Experience Requirement

This chapter discusses a family of objections against the idea of contingent a priori truths based on the claim that, even in cases of stipulative reference-fixing, there must be some perceptual contact between the stipulator and the object referred for knowledge of propositions about this object to be possible. In a way, these objections are subsumed under Donnellan’s criticism.1 But Donnellan himself is not explicit about what else is required (besides descriptive reference-fixing) for de re propositional attitudes. The family of objections here discussed is one particular elaboration of such requirement.

4.1 A Dialogue Between Plantinga and Philonous Plantinga (1974) was the first to formulate a certain kind of skepticism concerning cases where there is no perceptual contact with the referent of a name introduced by stipulative reference-fixing. He concentrates on the meter case: if one has never seen the stick and has absolutely no notion of its length then, by stipulating that ‘one meter’ is to be a rigid designator of the length of S at t0 , one can only have knowledge that the sentence ‘The length of S at t0 is one meter’ is true in the stipulator’s language, but not knowledge of the truth expressed by it. The objection, at least as Plantinga formulates it, considers a hypothetical situation in which the stipulator has no idea whatsoever about the length of S, i.e., he or she has “never

Parts of this chapter appeared in my “The Contingent A Priori and De Re Knowledge” in Carlo Penco, Massimiliano Vignolo, Valeria Ottonelli, and Cristina Amoretti (eds), Proceedings of the 4th Latin Meeting in Analytic Philosophy. Genoa: CEUR-WS.org, 2007, pp. 45–58. 1 This

is so despite the fact that one of them (Plantinga’s) appeared chronologically before Donnellan’s paper and is mentioned by Donnellan (footnote 13) as one of his influences, but the point that Donnellan is making seems to be much more general and elaborated than Plantinga’s. © Springer Nature Switzerland AG 2022 M. Ruffino, Contingent A Priori Truths, Synthese Library 443, https://doi.org/10.1007/978-3-030-86622-8_4

51

52

4 The Experience Requirement

seen S and hold[s] no views as to its length” (1974, p. 8, Footnote 1). But it says nothing about a situation in which, despite never having seen it, the stipulator does have some views (i.e., some notion) concerning the length of S. Is it really necessary to have visual contact with S in order to have an idea of the length of S that is operative in the sense that it allows the stipulator to distinguish that length from other lengths? E.g., imagine someone receiving S as a gift in a closed wooden box of 1.2 yards2 but never opens the box and never actually sees S. Nevertheless, this person wants to fix ‘one meter’ as the length of that never seen stick inside the box. Then he or she has an approximate idea of the relevant length and, although it is an imprecise one, there are several things that he or she knows about it, e.g., that it is larger than one yard, and shorter than two yards, etc. We can perhaps say that, for all practical purposes and within a certain margin of precision, he or she knows almost everything in terms of comparing the length of S with other lengths. Isn’t this enough to guarantee what Plantinga thinks is missing in the meter case? This seems to suggest that actually seeing S might not be necessary for being able to manipulate the unit of length within some limit of precision. But is it sufficient? Could one get a good intuitive idea (in the sense above, i.e., an idea that is operative in distinguishing one meter from other lengths) of its length by just seeing S? There is an old kind of argument employed by Berkeley in the first of the Three Dialogues Between Hylas and Philonous against the conception that there is a (primary) idea of length that one can get from the visual3 contact with an object: Phil. But as we approach to or recede from an Object, the visible Extension varies, being at one Distance ten or a hundred times greater than at another. Doth it not therefore follow from hence likewise, that it is not really inherent in the Object? (1713, p. 33)

Berkeley, as we know, is interested in showing that there is no such thing as the real extension of bodies. We do not need to follow him on this, but we can take his point about the shiftness of the idea of length acquired by visual contact with an object. I shall call this the Philonous’ Objection. Suppose that one sees S from a certain distance, but is under a wrong impression as to how far away S is, so that what appears visually to be of a certain length might actually be much larger (as when, e.g., we see a very tall building from a distance and it looks quite small). Similarly, if S is placed very close to the eye it might seem very large. In other words, there are many different impressions that one might get from seeing S depending on the perspective, on the angle, on how far from the eye S is, etc. So just seeing S does not give the stipulator a good (or unique) idea of what the length of one meter should be like. It does not help if S is seen together with other objects to get a better sense of S by comparison as when, e.g., in a picture taken of a captured fish one places it next to some familiar object like a pencil or a beer can to give an idea of how big it was, since, again, the idea of length of these other objects is subject to the same 2 This

is approximately 1.09 meters. argument is directed not only to the visual idea of length, but also to any form of length perception. 3 The

4.2 Salmon’s Theories

53

Philonous’ Objection. On the other hand, it might also be that the stipulator, seeing the same stick S in these two situations (very close and very far), is unable to tell that it has the same length. In more general terms, this leaves open the possibility that, even in the absence of direct (visual or any other form of) perceptual contact between the stipulator and the object referred, there might be some conception of the referred object that allows the stipulator, within some limits, to distinguish that object from others. On the other hand, having perceptual contact with the corresponding object might not be sufficient, in some circumstances, to get an idea that enables the stipulator to make such a distinction.4 This might require some experience (e.g., in our example above, seeing the wooden box in which S is kept), but not necessarily seeing S itself.

4.2 Salmon’s Theories One of the philosophers that follow the general lines of Plantinga’s criticism is Nathan Salmon, another champion of the direct reference theory of proper names. As I see it, he actually offers (at least) three slightly distinct theories concerning contingent a priori truths in his works explicitly devoted to the topic (1986, 1987), all of them vulnerable to Philonous’ Objection above. The first one, presented in Appendix B of (1986), I shall call the A Posteriori Only theory. The second one, presented in Section II of (1987), I shall call the Almost A Priori theory. And the third one, in Section IV of the same (1987), I shall call the Context Shift theory. The A Posteriori Only theory denies that there is any knowledge of lengths without measurement, which is a paradigm of an empirical experience. The Almost A Priori theory contains a concession to the special epistemic status of the standard meter stick S, although it does not give everything that Kripke wants. And the Context Shift theory amounts to an explanation, in the style of the second Wittgenstein, of our temptation to see the knowledge that S is one meter long as both trivial and contingent. I discuss each of them in turn.

4.2.1 Salmon I: The A Posteriori Only Theory In Frege’s Puzzle (1986), Salmon famously elaborates and defends a Millian perspective of proper names combined with an epistemology of de re beliefs, thus expanding on many aspects left unexplored by Kripke, and taking some of the latter’s views to their utmost consequences. Salmon’s theory is part of a rebellion against the Fregean-inspired methodology of subordinating the semantics

4 Kripke will explore something similar in his later lectures on the contingent a priori, as we will discuss in the next chapter.

54

4 The Experience Requirement

of proper names (and other expressions) to their epistemology. One of the pillars of Salmon’s approach is the distinction between what he calls the semantically encoded information of a sentence (i.e., the proposition strictly expressed by it and which, apart from indexicality, does not change from context to context) and the pragmatically imparted information which, as the name indicates, might be reached by all sorts of pragmatic inferences, and is relative to the audience. As Salmon explains, An utterance of any sentence typically imparts more information to the audience than merely the information semantically encoded. What information is imparted depends, in part, on the idiosyncratic conceptual association made by the listener or reader. The sentence ‘Aristotle wrote The Metaphysics’ may impart some information to one reader and different (perhaps overlapping) information to another, but it encodes a single piece of information, and, if all goes well, that encoded information is part (though only part) of the information imparted. (1986, p. 69)

This distinction serves as basis, e.g., to Salmon’s simple and radical solution to Frege’s Puzzle (which gives the title to his book) by simply denying one of its premises, i.e., that cognitive differences in sentences must be reflected by their semantic content. Salmon believes, contra Frege, that the proposition expressed by the following three sentences Superman is a hero Clark Kent is a hero I am a hero [said by Clark Kent] is one and the same, namely, the singular Russellian proposition that has as elements the individual Clark Kent and the property of being a hero. If someone, say, Lois Lane, believes the proposition expressed by the first sentence, she perforce also believes the proposition expressed by the second and third, despite the fact that she would probably express disagreement with Clark Kent’s utterance of the last one. Her different attitudes toward utterances of the sentences is not explainable, for Salmon, in terms of different semantic contents (which is one and the same, and is believed by her), but in terms of the information pragmatically imparted by each sentence. At the very end of Salmon’s book, in Appendix B, he addresses Kripke’s examples. He concentrates on the meter case or, as he prefers to call it, the meter sentence, i.e., S, if it exists, is one meter long at t and evaluates the plausibility of Kripke’s position in light of the previous distinction. This is how he expresses his view concerning the status of the meter sentence: I agree entirely with Kripke’s assessment of the modal status of this sentence, but I disagree sharply concerning the epistemological status. And here again, I would caution against confusing semantically encoded information—the information about the particular length that the stick S is just that long at t—from pragmatically imparted but not semantically encoded information, such as the information that the sentence ‘Stick S, if it exists, is

4.2 Salmon’s Theories

55

exactly one meter long at t’ is true. The latter information may indeed be a priori for the agent in Kripke’s example, and this apparent fact seems to be the source of Kripke’s contention that the (content of) the sentence is itself a priori. But even if the former information is a priori for the agent, it does not follow that the latter, semantically encoded information is. In fact, it is not. (1986, p.141)

His argument is, as we see in the passage, based on the distinction mentioned above between the semantically encoded and pragmatically imparted information of a sentence. It is similar to Plantinga’s and Donnellan’s objections in spirit, but differs in the details. A speaker may grasp the semantically encoded information (i.e., the proposition strictly expressed by the utterance of a sentence) and also grasp some of its pragmatically imparted information. This seems in accordance with a Gricean-like picture in which there is a primary proposition expressed, and another proposition reached by means of pragmatic inferences based, e.g., on conventional or conversational implicatures. E.g., consider the situation in which you are planning to leave home for an appointment at 9 am and I tell you that there is a lot of traffic around 9 a.m.; based on the understanding of that proposition (i.e., that there is a lot of traffic around 9 a.m.), you are able to infer one or more different implicated propositions (e.g., that it is prudent to leave home much earlier than 9 a.m., etc.). This inference goes from the semantically encoded to the pragmatically imparted proposition. A speaker or hearer might grasp the semantically encoded information without also grasping the pragmatically imparted one (e.g., if you understand my proposition about the traffic at 9 a.m. but do not infer that it is prudent to leave home much earlier than 9 a.m.). But, according to Salmon (and this is crucial to his position about contingent a priori truths), the opposite might happen as well, i.e., someone might grasp the information pragmatically imparted by a sentence without grasping the semantically encoded information. This situation deviates from the standard Gricean picture. Here we do not have the primary proposition (the semantically encoded) as a basis for the inference of the pragmatically imparted information as implicature.5 But, according to Salmon, this is precisely what is going on in Kripke’s meter sentence. In the case of the meter stick, a speaker may know a priori the pragmatically imparted (metalinguistic) information that the meter sentence is true without thereby knowing the semantically encoded information. This is a case in which knowing a piece of metalinguistic information requires less than knowing the semantically encoded information. In order to know the semantically encoded information expressed by the meter sentence, one needs something stronger. For all the stipulator knows, there are several Russellian propositions that might be

5 Actually, Salmon allows for the even more radical possibility of pragmatically imparted information with no semantically encoded information (1986, p. 127). This happens in those cases of sentences including empty names, which express no Russellian proposition at all, but which might have pragmatic implications. One might wonder what the pragmatic inference is based on, and maybe “pragmatically imparted” is not the best designation for this kind of information if this means information pragmatically inferred from some other information. But this terminological question is not really fundamental here.

56

4 The Experience Requirement

expressed by ‘one meter is the length of S (if it exists)’, and that we could represent as l1 , P , l2 , P , l3 , P , . . . where l1 , l2 , etc. are different lengths (we have to assume here that abstract lengths are particulars and, hence, possible elements of singular Russellian propositions), and P is the abstract property of being the exact length of S if it exists. In order to know and understand which proposition is expressed, the speaker must know which of the particulars l1 , l2 , etc. is expressed by ‘one meter’. But in order to understand the expressed proposition, the speaker must have cognitive contact with the selected particular, which can only be the result, in Salmon’s account, of seeing and measuring S.6 Presumably, Salmon’s point is that, in order to know (or simply to consider) the proposition that has the length of stick S as one of its elements, the speaker must, at some point, have perceptual contact with that stick, which means that he or she cannot know (or even think) that Russellian proposition a priori. The point is similar to Donnellan’s original complaint:7 in order to grasp the length of the stick, the speaker must have some sort of contact (acquaintance or something of the sort) with the stick’s length, and this might require some sort of acquaintance with the stick itself. This implies, at least in accordance with Salmon’s first theory, measuring the stick at some point. But this disqualifies this kind of knowledge as a priori. Thus, a speaker can only have a posteriori knowledge of the proposition expressed by ‘one meter is the length of S (if it exists)’ which is, for Salmon, contingent. As it seems, Salmon’s view is as vulnerable to Philonous’s Objection as Plantinga’s: the kind of cognitive contact with the abstract length that Salmon takes to be necessary for understanding a singular proposition containing it can hardly be guaranteed by any visual experience of stick S.8 Salmon’s main concern regarding the possibility of a priori knowledge of lengths is better explained in the very last sentence of Appendix B (1986, p. 142). It resembles somehow the kind of argument known as “slippery slope”. Knowledge concerning the length of objects is the result of measurement and, as such, a paradigm of a posteriori knowledge. Thus, Salmon thinks, if there is a case in which someone can know the length of some object a priori then it should be possible for everyone to know the length of just any object a priori, presumably, by performing an analogous stipulation and, hence, no such knowledge needs to be a posteriori for anyone. This is, presumably, what he wants to say with 6 “Knowledge concerning a particular length that a certain stick (if it exists) is exactly that long would seem to be the paradigm of a posteriori knowledge.” (1986, p. 142). 7 Salmon credits (footnote 18) Donnellan as one of the influences for his argument. As I said at the beginning of this chapter, the accounts here reviewed all resemble Donnellan’s in spirit, although they differ in the details. 8 For a discussion of some complications in Salmon’s idea of apprehending the elements of a singular proposition in the right sort of way, see Oppy (1994).

4.2 Salmon’s Theories

57

It would seem that if it is not a posteriori for everyone, then nothing is a posteriori for anyone. (1986, p. 142)

This reasoning is partially correct insofar as fixing one’s own standard of measurement by picking any object and stipulating that it is the standard of measurement is always technically possible. But it is incorrect on two accounts. First, if we consider that measurement is, almost always, a practice grounded in public standards, then we can see that one public standard of measurement excludes another (e.g., there cannot be two conflicting definitions of meter). In the same way that, as far as personal idiolects are concerned, one is free to fix his or her own meaning for words, but engaging in language as a social practice excludes multiple standards of meaning. Measurement is normally carried out as a social practice, using public standards, and this precludes as inappropriate the multiplication of standards of measurement that Salmon considers as a possible (absurd) consequence of there being contingent a priori knowledge of some length. As I will discuss in Chap. 10, stipulating a standard of measurement (and, thereby, coming to know a priori some truth concerning lengths) may depend on an institutional authority not granted to everyone. Second, the experience of measuring and, by means of it, gaining a posteriori knowledge cannot happen unless there is some previously established standard of measurement and, hence, some truth about length that one knows without measurement. Therefore, there can be no a posteriori knowledge unless there is some a priori knowledge of a contingent truth regarding some standard of measurements.9

4.2.2 Salmon II: The Almost A Priori Theory The core of Salmon’s argument in Frege’s Puzzle is repeated in his later paper on the subject (1987), although with some slight (but important) modifications. One of the few changes is the following: while in the original version Salmon says that, in order to know the proposition expressed by the meter sentence, it is essential that the stipulator actually measures S (thereby having only a posteriori knowledge that S is one meter long at t0 ), in the later version he no longer considers the act of measuring as necessary, although he still thinks that having visual contact with S is necessary.10 This is so because, as Salmon thinks, by establishing visual contact with S, one is also derivatively establishing a cognitive relation to the abstract length of S. Indeed, Salmon wants to draw a subtle distinction between:

9 Kripke

makes a similar point in his later lectures on contingent a priori truths. focuses on visual contact; but what is essential for him is not so much that vision is involved, but that some form of sense experience is involved. We can imagine a blind person going through a stipulation ceremony in which ‘one meter’ is stipulated as the length of a stick that he or she can touch and scroll with the fingers. I take it that everything Salmon says about visual experience with S also holds, mutatis mutandis, for other forms of sensory experiences. 10 Salmon

58

4 The Experience Requirement

(i) Knowing of a certain length that an object has that length and (ii) Knowing exactly which length an object has. (ii) implies (i), but not vice-versa. As he sees it, by having visual contact with a stick, one also establishes a cognitive relation to a certain particular length and knows, of that length, that the stick has exactly that length; but one might not know which length that is, since this would require measuring it using a standard unit. The distinction only makes sense assuming that there is a rival unit of length in use independently of the one introduced by means of S. Otherwise, if there is no other independent unit of length, then by just seeing S (and thereby entering into a cognitive relation to its length), one would also know exactly which length that was, for seeing a length and measuring it would amount to the same thing.11

11 For

this reason, Salmon rejects, in footnote 12, an argument that Kennedy (1987) formulates on his behalf against Kripke. The argument presupposes a Millian theory of proper names plus a semantics of belief reports according to which substitutions of co-referential names in such reports are truth-preserving. It also presupposes that one might have visual experience of an object without knowing its length. Or, to be more precise, it presupposes that one can have visual experience of an object and hold false beliefs about its length. Kennedy considers a situation to the effect that a subject (he calls her ‘Claudia’) has a visual experience of stick S at t0 which has, as usual, exactly one meter in length. But, since she is not sure about the length that she is perceiving, she does not know its length and does not believe it to be exactly one meter. She then decides to baptize the length of S at t0 as ‘schmoo’. Since it is assumed that both ‘schmoo’ and ‘meter’ are Millian names, it follows that, if she knows a priori that the length of S at t0 is one schmoo, then she also knows a priori that the length of S at t0 is one meter, even though she acted (before the stipulation) as if she did not know that the length is one meter. The difference in Claudia’s cognitive state is the result only of the stipulation. But it seems absurd to think that the stipulation alone can change her belief that the stick is not one meter long into a belief that it is one meter long. The reason for Salmon’s rejection of the argument is that Kennedy presupposes, as we saw, that by looking at S, Claudia does not know of its length (i.e., one meter) that it has exactly that length. On the contrary, for Salmon, just by looking at S, Claudia knows of a particular length that S has that length, although she is unable to know which length that is with respect to any standard of measurement. There is another reason to be unsatisfied with Kennedy’s argument, at least from the perspective of Salmon’s Millian semantics and epistemology. According to the latter, the objects of propositional attitudes are propositions and, since ‘one meter’ and ‘schmoo’ are non-descriptive singular terms, the propositions in question must be singular Russellian propositions. As a consequence of the distinction between semantically encoded and pragmatically imparted information, Salmon holds that one might believe a singular Russellian proposition without being aware of such belief. E.g., assuming that ‘Superman’ and ‘Clark Kent’ are Millian names, Lois Lane believes that Superman is a hero and, thereby, she also believes that Clark Kent is a hero, even though she does not know that she believes that, and even if she acts and talks as if she does not believe it. Now, in Kennedy’s example, Claudia acts and talks as if she does not know that S is one meter long at t0 . But she (or at least the example does not explicitly rules this out) had contact with something that is one meter long (otherwise she would have no reason at all for judging, even mistakenly, that S is not one meter long at t0 ). By performing a baptism ceremony she just introduced a new name (‘schmoo’) for something with which she did have acquaintance before, so there was not, after all, an epistemic upgrade solely in virtue of the baptism, as Kennedy claims against Kripke. In other words, Kennedy’s argument hardly works if one assumes, as Salmon does, that we can have propositional attitudes towards a Russellian proposition without being aware that one of its components is referred to by a particular name.

4.2 Salmon’s Theories

59

At any rate, establishing the cognitive relation to the relevant length requires actually seeing S and, therefore, requires some empirical experience: No physical measurement is required beyond merely perceiving the object (taking it in lengthwise in one fell swoop, etc.). But some sensory experience in which S plays a crucial role seems to be required. The metre sentence is apparently a posteriori, even if physical measurement is not required for verification. (1987, p. 205)

The problem with this piece of reasoning is that it is equally vulnerable to Philonous’ Objection: the abstract length with which one establishes a cognitive relation changes according to the perspective under which one sees S. Salmon does show some awareness of this threat when he places a qualification regarding the perception of the exact length of S at t0 (he calls this length ‘Leonard’) through visual perception of S: Of course, merely perceiving an object will not always result in such empirical knowledge. Perhaps in order to see an object’s length one must be able to take in the object lengthwise, from end to end, in one fell swoop. Perhaps the visual presentation cannot be under circumstances that create optical illusions (such as might be created by surrounding the object with miniature artifacts, each reduced to the same scale, etc.). Perhaps not. In any case, if the reference-fixer does indeed see S under the required circumstances, he can thereby know of its present length, Leonard, that S is presently exactly that long. (ibid., my emphasis)

But Salmon does not take the threat seriously enough. The passage says “under the required circumstances”, but there is no specification of what would be required except excluding perspectives that would clearly give the wrong impression (such as viewing S laterally, perceiving only its thickness). But seeing S “lengthwise, from end to end, in one fell swoop”, as Berkeley says through Philonous’ Objection, might still give many different impressions of length depending on one’s distance from S. As we saw in the discussion of Plantinga’s objection, no sensory experience of S will be enough to enter into the appropriate cognitive relation to its length.12 Within the framework of his new theory, Salmon concedes that S is epistemically unique in that it is the only object (assuming that there is no other object that works as an alternative standard of measurement) of which one can know the length just by seeing it, without having to measure it against some other object: However, we still get a rather curious result, not unlike Kripke’s claim that the referencefixer knows S’s length a priori: if the reference-fixer knows without measuring and just by looking that S is precisely one metre long, then he knows precisely how long S is without measuring and just by looking. (ibid., p. 208)

But, again, this is not enough to make knowledge of S’s length a priori according to Salmon because one has to look and see the stick in order to know it. We could say that Salmon admits that knowledge of S’s length is almost a priori in the sense that

12 Again, we do not have to go as far as Berkeley and conclude that there is no such thing as an objective length; but the point is that it is highly questionable that seeing S will lead to knowledge of that length, at least the way Plantinga and Salmon describe the experience.

60

4 The Experience Requirement

it can be known just by looking at it; but not a priori in an absolute sense. This is why I call this variation of his view the Almost A Priori theory. Although he yields to Kripke’s view that S is epistemically exceptionable, Salmon thinks that this exceptionality leads to a kind of a paradox that ultimately results in skepticism: in the language game of measurement, S is the only object whose length can be known just by looking at it, but only because S was arbitrarily chosen as the standard. However, or so he seems to think, an arbitrary choice cannot, just by itself, turn an object that is knowable in the ordinary universal way into an object that is knowable in a unique way.13 This invites the skeptical conclusion that, pace Kripke, one does not really know the length of S. But, since S is the standard by comparison with which every other length is known, it seems that we have to extend the skepticism to every other object and conclude that no length at all can be known. This motivates Salmon’s third theory, which I shall discuss below. Also in the framework of his second theory, Salmon repeats Donnellan’s general point against Kripke for not distinguishing between knowledge of the propositional content of the meter sentence and the a priori statement that the exact length of S at t0 is referred to (in the idiolect of the stipulator) by ‘one meter’ or, alternatively, the a priori statement that ‘One meter is the exact length of S at t0 ’ is true in the idiolect of the stipulator. This is how his argument goes: assume that what is to be known is the Russellian proposition (he calls it ‘Peter’) having as elements Leonard (to remember, he calls ‘Leonard’ the exact length of S at t0 ) and the abstract property of being the exact length of S at t0 if S exists. Salmon says: I conjecture that Kripke, in his discussion, either failed to distinguish properly between the a posteriori content of the metre sentence, i.e. Peter, and the arguably a priori truth that the length at t0 of S is referred to (in the reference-fixer’s present idiolect) as ‘one metre’ (or something similar, such as the proposition that the metre sentence is true in the referencefixer’s idiolect), or else he failed to appreciate that the reference-fixer’s visual experience of S in the very introduction of the term ‘metre’ is a crucial part of the justification for the reference-fixer’s belief of Peter. (1987, p. 203)

As it is clear, Salmon thinks that one could not know a priori (by means of a stipulation) of Leonard that it has the property of being the exact length of S at t0 if S exists. No matter that the property P figuring as element in this proposition involves complexities such as an existential clause concerning S. The problem comes from Leonard as a particular. Salmon’s point seems to rely on a general assumption that is also crucial to Donnellan’s line of criticism: a subject can only have de re

13 I think Salmon is wrong here: one can, as a result of a decision, and provided one has the required authority, turn an ordinary object into one that has a special epistemic status if that object is part of an institutional fact. He says:

Despite its ‘peculiar role in the language-game’, it is still a stick, a physical object subject to the same natural laws and knowable in the same way as any other. (1987, p. 209) This seems to beg the question. Not all epistemic aspects of an object derive from its being subject to natural laws. Some are derived exactly from the fact that they are surrounded by (and eventually part of) institutions. This will be better developed in Chaps. 9 and 10.

4.2 Salmon’s Theories

61

knowledge (or, more broadly, de re epistemic attitudes) about a particular x if the subject is in some sort of epistemic contact with it that can only be provided by sense experience.14 And, therefore, there can be no a priori knowledge of Leonard, but only of the truth of the metalinguistic sentence described above. Moreover, Salmon also seems to assume that the only way in which one could have acquaintance with Leonard is by means of contact with some concrete object that has that exact length, e.g., S at t0 . In other words, we not only have the acquaintance requirement (as in Donnellan’s complaint), but something stronger: in cases of singular propositions that include an abstract entity such as Leonard, acquaintance must be mediated by acquaintance with a concrete object that has it as length.15 Even if someone could entertain or believe the proposition containing Leonard without having visual contact with S, the justification of that belief would require such contact: Even if the reference-fixer came to believe of Leonard (so conceived) that S, if it exists, is also exactly that long at t0 , but did so somehow solely though contemplation and reflection on his concepts without experiential justification (i.e., not by estimating S’s length from its appearance etc.), he still could not properly be said to know this of Leonard. At best, it seems more like extremely lucky guesswork. It is only by seeing S and its length that the reference-fixer comes to know that S (if it exists) is just that long. (1987, p. 202).

There are some reasons to consider this line of thought less than satisfying. First, once more, Philonous’ Objection applies here: having visual contact with S is not enough to get into the required epistemic contact with Leonard that Salmon takes to be fundamental. Moreover, it is doubtful that it is necessary either. Remember the case, presented in the section on Plantinga, of someone having contact with an object of approximately the same size as S, and this enabling the person to have an almost perfect manipulation of Leonard within some practical limits of precision. Second, we could also think of a situation in which someone is never presented with anything that has Leonard as its exact length, but is presented with two other sticks: one that has exactly 1.10 meters (call this length ‘Big Leonard’) and another that has exactly 0.10 (call this length ‘Little Leonard’) and, based on that, comes to the idea of Leonard by mentally subtracting the lengths (Big Leonard minus Little Leonard) without ever having seen anything that has Leonard as exact length. Finally, this position places too much weight on the perceptual requirements for a speaker to be able to understand a singular proposition. If this requirement is taken seriously, it 14 There is an obvious difference between Donnellan’s and Salmon’s complaints in that the former focuses on the Neptune case and, hence, on knowledge of a singular proposition containing a concrete object (Neptune), while the latter focuses on the meter case and, hence, on knowledge of a similar proposition but with an abstract particular (a length). Salmon’s version of the complaint must include some additional refinements to account for acquaintance with lengths. 15 It is not clear, to me at least, whether Salmon would extend this requirement to other abstract objects such as numbers. Suppose someone wants to stipulate that ‘Seribi’ is to rigidly designate the number of fish alive in the Amazon rivers at t0 . Then, following Donnellan’s requirement, in order to understand the true proposition that Seribi is the number of fish alive in the Amazon rivers at t0 , one must have acquaintance with the number designated by ‘Seribi’. Does this mean that, at some point, one has to be presented with a group of concrete objects having exactly that number of elements?

62

4 The Experience Requirement

might imply that there can be no a priori knowledge at all regarding abstract entities because, at some point, some sort of experience will be necessary for a speaker to even understand any proposition. E.g., understanding the proposition expressed by All red apples are red might require some familiarity with the concept of apple and the color red, thereby requiring some experience. However, we are not inclined to consider knowledge of this proposition as a posteriori just because of that, since it is a logical proposition and, as such, a paradigm of an a priori proposition. Similarly, to understand 2+2=4 one must at least know that ‘2’ refers to the number two, ‘+’ to the addition operation, that the second occurrence of ‘2’ refers to the same number as the first, etc., and this might, at the bottom, require some experience.16 But, again, this seems to be a paradigm of a proposition that does not require any experience to be known. The understanding of any such proposition involves some experience. Hence, lest we want to concede to a radical form of empiricism for which that there are no a priori propositions at all, we should better follow, e.g., BonJour (1998, p. 10) and Kant (1787, B1) in considering a proposition a priori when its justification appeals to no further experience once it is understood (although understanding might require some experience). This is a way out that Salmon actually does consider, but very quickly dismisses: One might maintain that the reference-fixer’s visual experience of S in the introduction of ‘metre’ likewise enables the reference-fixer to apprehend Peter but plays no further role in justifying that belief. The case for a-priority along these lines, however, is far from clear. The reference-fixer’s visual experience of S can play an important role in enabling him to apprehend propositions directly concerning S, but it does play a crucial role in justifying his belief of Peter. (ibid., p. 202)

No further argument is given for the very last claim and I see no ultimate reason for taking it as a universal rule. As the argument in the passage is presented, it looks more like a petitio. As I shall argue in Chap. 10, the ultimate justification that the stipulator has for the belief in Peter is not the visual experience, but the fact that there is a stipulation. As said before, stipulations are, from the perspective of speech act theory, special in that they, if successful, make a propositional content true by the very illocution and, hence, are the ultimate justification for this truth. The visual experience might have an important role, but only as a way of checking whether one of the conditions of success of the stipulative speech act is satisfied. It might happen that knowledge of the preparatory conditions of a speech act is a posteriori, while knowledge resulting from the same act is a priori. The clearest

16 Children typically learn elementary arithmetic by adding one group of objects to another and then counting the resulting group.

4.2 Salmon’s Theories

63

case is perhaps that of definitions in mathematics: in order to be successful, there are some conditions that a definition must satisfy, e.g., that it introduces a new term, that no other (different) definition for that same term occurs previously in the same theory, that the definiens contains only terms that are primitive in the system or that have been defined before, etc., and these might all correspond to a posteriori information. But when this information is in place, the result of the new definition is something known a priori by the mathematician (and the readers).17

4.2.3 Salmon III: The Context Shift Theory Salmon’s third theory,18 as said before, is conceived in the characteristic style of the second Wittgenstein of detecting (and recommending a “therapy” for) an apparently irresistible tendency that we might have of advancing philosophical theses that generate puzzles. In this case, the problematic thesis is that statements such as that S is exactly one meter long at t0 are true. (Wittgenstein, contrary to Kripke, thinks that it is neither true nor false.) The leading idea is to reconcile the thesis that the stipulator trivially knows the length of S (the statement must hence be true) without measuring it with the view that this is the kind of knowledge that can only be reached by measurement. Following the Wittgensteinean inspiration, Salmon focuses on the genesis of measurement as a kind of language game which starts when someone, “a very clever caveman”, decides to take a particular stick S as the standard of measurement, something that nobody had ever thought before, and “his fellow tribesmen agree to his scheme” (1987, p. 206). It is intrinsic to the institution created by any such game that there is a universal criterion for knowing how long any arbitrary object is, i.e., one must compare the object with another object adopted as the standard in order find out its exact length. Suppose we assume that the standard stick S has a very peculiar property, namely, that the stipulator can know its exact length, although there is no other standard of measurement that can be used to measure it. (Let us also assume the “genesis” scenario in which there is no other concurrent standard of measurement such as a different stick with a different length, and, hence, the stipulator does not know the length of S by comparing the two sticks and doing the numerical conversion.) If the stipulator can say that S is one meter long without comparing it with another standard, then there are only two alternatives. The first is that the stipulator knows that completely a priori, depending on no empirical experience whatsoever. As we saw, Salmon rejects this alternative since he thinks that some visual experience of the stick is essential. (As we also saw, this

17 See

Ruffino et al. (2020) for a more detailed account of mathematical definitions as speech acts. himself does not mean it as a third theory but as an addendum to the second one. However, it clearly introduces elements that are not required by the latter, so they can be taken independently as a third approach to the meter case. 18 Salmon

64

4 The Experience Requirement

is vulnerable to Philonous’s Objection, but let us leave this aside for now.) So, the second and only remaining alternative is that the stipulator knows the exact length of S just by looking at it. But this is something that, according to Salmon, cannot occur in a regular language game of measurement, for such a game requires comparison with a standard (and there is no other standard besides S itself). Hence, on the one hand, we would like to say that the stipulator does know the exact length of S. On the other hand, the skeptical conclusion seems very tempting that the same stipulator does not know the length of S because it cannot be measured against anything, as is generally required by the game of measurement. The situation is even worse in the sense that we might be sliding into a deeper form of skepticism: given that we do not know the length of S, and given that S is the parameter with respect to which the length of everything else is known, it follows that we do not know the length of anything.19 Salmon suggests that there is a way of reconciling both claims, i.e., that one does know the exact length of S just by looking at it and that there is no way of knowing its length without measurement. This is based on the observation that ‘knowing which F ’ is an indexical locution, expressing different epistemic relations in different contexts (p. 156). First, there is one (everyday, non-philosophical) standard according to which one might trivially be said to know the exact length of S at t0 simply by looking at it (and, therefore, to know that S is one meter long at t0 ). But there is a second (more technical, philosophical) sense of ‘knowing’ which requires more than simply seeing S: But (in part for that very reason) merely uttering the sentence ‘The Standard Metre is exactly one metre long’ tends to raise the ante to a level at which its utterance becomes epistemically unjustified—and threatens to invoke the skeptic’s favourite level, at which its utterance is in principle unjustifiable. (1987, p. 216)

The impression of a puzzle arises because, on the one hand, we have the feeling that someone trivially knows the length of S simply by looking at it but, on the other hand, switching the epistemic standards without noticing it, we also think that there is no way one could know that length simply by looking at it.20 However, this seems unconvincing. There is no strong reason for saying that we must unwarnedly pass from one (everyday and naive) context to another (philosophical and skeptical) in order to feel the peculiarity of the meter case. It is not difficult to distinguish two epistemic standards such that, according to one of them, we might consider that someone knows that the Earth is round, that a river contains mostly water, etc., whereas, according to the other one, we might be

19 Salmon sees this as one particular case of a general form of skepticism regarding any unit of measurement such as, e.g., 1 h as a fraction of the period of rotation of the Earth on its axis, etc. He calls this the “Does-anybody-really-know-what-time-it-is skepticism” (1987, p. 211). As he sees it, skepticism concerning our knowledge regarding S’s length opens the door for this more radical form of skepticism. 20 A similar diagnosis of contingent a priori truths as the outcome of shifting contexts with different epistemic standards is also suggested in Stojanovi´c (2004).

4.3 Soames and “The Serpert’s Egg”

65

more reluctant to attribute knowledge in these cases. When doing philosophy we can, at least in principle, keep these two standards apart in discussing instances of potential knowledge. So, there is no compelling reason for saying that we tend to slide unwarnedly from one standard to another, and even less for using this transition to explain away a contingent a priori proposition like the meter case. As a general strategy to avoid sliding from one context to another, Salmon recommends (very much in the spirit of a Wittgensteinean “therapy”) that one should not say that it is true that the stick S is one meter long and that this is known to the stipulator: If saying something that is trivially true leads us to say further things that sound much more alarming than they really are, it may be better to say nothing. (ibid.)

That is to say, Salmon yields to Wittgenstein’s position that we should not say that the meter stick is not one meter long (because this would be trivially false), but neither should we say that it is one meter long because this leads to a situation in which we raise the standards of epistemic justification without being aware of that. But there seems to be no ultimate reason for this claim either. This is so especially in view of Salmon’s remark, in the first paragraphs of the paper: Frankly, I suspect that Wittgenstein is ultimately completely wrong regarding the Standard Metre. (1987, p. 195)

4.3 Soames and “The Serpent’s Egg” of Two-Dimensional Semantics Soames’ (2005) is conceived as a vigorous attack against what he takes to be a descriptivist “counterrevolutionary” reaction against the direct reference “revolution” inaugurated with the works of Kripke, Kaplan and Putnam, among others. The descriptivist “counterrevolution” is represented by the so-called two-dimensional semantics (which will be the subject of Chap. 8) and their accompanying epistemology. Although we do not need to discuss Soames’ criticism of these semantics here, it is interesting for our purposes to discuss what he considers to be some important mistakes in Kripke’s and Kaplan’s account. As he sees it, these mistakes inspired and stimulated the “counterrevolution”: I will show how certain parts of their [Kaplan’s and Kripke’s] discussions have encouraged descriptive two-dimensionalists. In all such cases, I will argue that the passages by Kaplan and Kripke contain errors, slips, or misleading suggestions that are all too easily interpreted as pointing away from the guiding insights and lasting lessons of their brilliant works. (2005, p. 43)

The “errors, slips and misleading suggestions” that Soames refers to in the passage are mostly related to his appreciation of Kaplan’s and Kripke’s account of contingent a priori truths, so we can take interest in Soames’ points even independently of his global case against two-dimensional semantics. Soames builds a criticism against the alleged cases of contingent a priori truths in Kaplan and in Kripke; some of these

66

4 The Experience Requirement

criticisms are meant to apply equally to both philosophers, and some are directed at particularities either of Kripke’s or of Kaplan’s account. One thing that is clear from the outset is that Soames, like many others, takes the phenomenon of contingent a priori truths to be one single phenomenon, with one single root, of which Kaplan’s and Kripke’s cases are like garden varieties. I will focus on the discussion presented in Chap. 4 of Soames’ (2005), for it is the one more directly related to our present interests. Soames raises three main points against Kaplan’s version of the contingent a priori. The last point is repeated, mutatis mutandis, for Kripke’s versions of it. He also has other critical points to Kripke’s overall project on direct reference and reference-fixing (see the complete list of all his critical points in p. 55), but these other points do not relate directly to the contingent a priori. Hence, I shall only deal with the first criticism of Kripke, which is basically the same as the third criticism of Kaplan’s view.

4.3.1 Characters and Logical Truths In Section XIX of “Demonstratives”, Kaplan defines validity or logical truth in his language LD (“Logic of Demonstratives”) as truth in all contexts of all LDstructures, and an LD-structure U as something of the following kind: U = C, W, U, P , T , I where C is a non-empty set of contexts, W a non-empty set of possible worlds, U a non-empty set of individuals, P a non-empty set of spatial positions, T is the set of positive integers (thought as temporal instants), and I the interpretation-function associating to each predicate and each functional symbol of LD an appropriate intension. Each context c∈C consists in a possible world w∈W, an agent a∈U a location p∈P, and a time t∈T. As we saw in Sect. 2.7, Kaplan takes the bearers of logical truth and of contingency to be different entities, namely, character and content, and this is how he goes on to explain the possibility of contingent a priori and necessary a posteriori truths. However, in other parts of his text, Kaplan also identifies the character of an expression with its meaning, i.e., the part that remains fixed from context to context, and which any speaker understands just by being linguistically competent. In other words, the meaning (i.e., character) is the bearer of logical truth, and, hence, of apriority. Soames’ first critical point is directed at the identification of character as the bearer of logical truth. For, as Soames points out, character corresponds to the intuitive notion of meaning, and meaning is something that changes from model to model: The inclusion of the interpretation function—which is indispensable in model theoretic semantics—ensures that the interpretations, and hence, meanings of the nonlogical vocabulary will vary from model to model. This is significant, because the nonlogical vocabulary

4.3 Soames and “The Serpert’s Egg”

67

includes all primitives of the language—all names, predicates, and function signs—other than the standard logical symbols, the modal and the tense operators, and the special indexical terms and operators introduced by Kaplan in his formal system. (2005, p. 48)

That is to say, if character is meaning, character cannot be the bearer of logical truth, because there might be a different meaning for the non-logical vocabulary in each LD-structure, while logical truth requires that we have the same meaning true in all LD-structures. Soames’ point would carry weight if, in Kaplan’s system, we had names in the logical vocabulary, whose meaning could change from structure to structure. However, an important feature of this system (at least in Kaplan’s original formulation of Section XVIII) is that it does not include names in the non-logical vocabulary, but only predicates and functional symbols. Is this a failure of Kaplan’s system? It depends on how one looks at it. The technical apparatus of his formal system was introduced for a particular theoretical purpose, i.e., for studying a particular semantic phenomenon. This is how he justifies the introduction of LD: Just to be sure we have not overlooked anything, here is a machine against which we can test our intuitions. (1977, p. 541)

But what are the relevant intuitions here? Those related to indexicals which, in Kaplan’s system, do not belong to the non-logical, but to the logical vocabulary. (Kaplan calls “primitive symbols” those representing ‘now’, ‘actual’, ‘yesterday’, ‘dthat’, ‘I’, and ‘here’.) That is to say, there is no change in the meaning of these terms from one LD-structure to another. Soames points out that there are two ways of understanding Kaplan’s notion of character (and, apparently, Kaplan oscillates between both). On one interpretation, characters are, in Soames’ words, “full-blown meanings—characters in the intended model, if you will”. On the other interpretation, they are “unrelativized to models— what I have called schematic, or unrelativized, characters” (2005, p. 49). But it turns out that, for the indexicals that are relevant in Kaplan’s system, these two kinds of character are one and the same, since they are part of the logical vocabulary, and their meaning is the same in all LD-structures. Consequently, for the portion of natural language that Kaplan’s system is supposed to model, it is unproblematic to identify meaning with character, and hence logical truth with apriority. What would happen if Kaplan had included names as part of the non-logical vocabulary? Then, since he accepts Kripke’s idea that names are rigid designators, some more or less obvious adjustments would have to be made in his notion of validity (or in the interpretation-function I).

4.3.2 Characters and the Objects of Thought Soames’ second critical point is against Kaplan’s thesis (crucial for his examples) that the bearers of logical truth are the same as the bearers of apriority. Since apriority is an epistemic notion, we would expect that items of apriority are items

68

4 The Experience Requirement

of knowledge (and, hence, objects of propositional attitudes). Soames thinks that taking logical truths as bearers of apriority is not consistent with the distinction that Kaplan introduces between “objects of thought” and “cognitive significance” in a passage of section XVII, part of it already quoted in Chap. 2: Is character, then, the object of thought? If you and I both say to ourselves, [. . . ] “I am getting bored” have we thought the same thing? We could not have, because what you thought was true, while what I thought was false. What we must do is disentangle two epistemological notions: the objects of thought (what Frege called “Thoughts”) and the cognitive significance of an object of thought. As has been noted above, a character may be likened to a manner of presentation of a content. This suggests that we identify objects of thought with contents and the cognitive significance of such objects with characters [. . . ] According to this view, the thought associated with dthat[α] = dthat[β] and dthat[α] = dthat[α] are the same, but the thought (not the denotation, mind you, but the thought) is presented differently. (1977, p. 530)

That is to say, Kaplan wants to call “objects of thought” the content expressed by sentences. But the content in itself has no cognitive significance: the latter is given by the character, i.e., the form in which the content is accessed. The same content can be given in different ways and, hence, with different cognitive significances. For instance, I might have thought yesterday that I should give a lecture on September 20th under the form ‘I should give a lecture tomorrow’, and today I think under the form ‘I should give a lecture today’. The same content is presented under two different characters, and with two different cognitive significances (yesterday I prepared for the lecture, today I start talking). On the other hand, I might think in two different days ‘today I should give a lecture’; here we have two different contents under the same character and, therefore, the same cognitive significance (on both days I start talking). Soames takes Kaplan’s terminology literally here, and claims that, if the latter calls contents the objects of thought, then characters cannot be the bearers of apriority, since only things that are objects of thought can be a priori or a posteriori. Soames uses this conclusion, i.e., that contents are the objects of thought, to argue against the following two theses: (LE)

(LV)

Any sentence logically equivalent to an a priori truth is also a priori. (Alternatively, any sentence equivalent to an a posteriori truth is also a posteriori.) Any valid (i.e., logically true) sentence is a priori.

The objection is important because (LE) and (LV) are fundamental steps in the construction of Kaplan’s cases of necessary a posteriori and contingent a priori truths, e.g., in his Remark 10 on the formal system (1977, p. 550). As we saw in Chap. 2, Kaplan explains some of his crucial examples in the following way: take the non-rigid descriptions α and β. Because they are non-rigid, ‘α = β’ is a posteriori. Now, we have as a consequence of the semantics of ‘dthat’ that | dthat[α] = dthat[β] ↔ α = β

4.3 Soames and “The Serpert’s Egg”

69

i.e., that ‘dthat[α] = dthat[β]’ and ‘α = β’ are logically equivalent. We also have as a consequence of the semantics of ‘dthat’ that | α = dthat[α]. i.e., that ‘α = dthat[α]’ is valid. Now: If ‘dthat[α] = dthat[β]’ and ‘α = β’ are logically equivalent and ‘α = β’ is a posteriori, from (LE) it follows that ‘dthat[α] = dthat[β]’ is also a posteriori; If α = dthat[α] is valid, from (LV) it follows that it is a priori. Soames thinks that, if contents are the objects of thought (as Kaplan says), both (LE) and (LV) are incorrect and, therefore, Kaplan’s classical examples involving ‘dthat’ do not get off the ground. (LE) is incorrect because being logically equivalent cannot mean having the same epistemic status, since the latter has to do with the objects of thought, i.e., with content; two sentences can be logically equivalent without having the same content. He adds (p. 50) the following kind of reductio of (LE): Kaplan’s explanation in Remark 10 comes to the conclusion that, as a consequence of (LE), ‘dthat[α]=dthat[β]’ is a posteriori. But if contents are the objects of thought, and if α and β are co-referential, it follows that ‘dthat[α]=dthat[α]’ would have to be considered a posteriori as well because it has exactly the same content. But this is absurd; the content of the latter sentence is a paradigm of something that can be known a priori. Hence, if contents are the objects of thought, (LE) must be false. This affects primarily Kaplan’s examples of necessary a posteriori truths, but his examples of contingent a priori are also undermined if Soames’ point is sound. According to him, (LV) is incorrect basically for the same reason: even if a sentence is a logical truth, it does not mean that it has a content that is knowable a priori. (Additional reason will be brought against (LV) in his third criticism, discussed in the next section.) However, Soames’ conclusion is not inevitable; his point is entirely based on a questionable interpretation of Kaplan’s terminology. In the passage quoted by Soames, Kaplan indeed uses the term “objects of thought” but since he never goes on to explain what exactly he means by it and what epistemic role it is supposed to play, I think that “objects of thought” should rather be taken as a technical term, peculiar to Kaplan’s philosophy of language, without the epistemic import that Soames reads into it. It is well known that Kaplan follows Perry in considering the cognitive states as individuated by characters, and not by contents. Therefore, characters individuate beliefs understood as psychological states that might motivate action. Soames is perhaps right that Kaplan’s terminology is somewhat unfortunate here, since he wants to distinguish the bearers of two functions that Frege attributes to one single entity. Frege originally uses the name ‘thought’ for this entity and, hence, ‘objects of thought’ seems to inherit the same ambiguity. However, although Kaplan’s terminology is misleading, he certainly does not mean that contents (or, at least, contents alone) are the object of epistemic attitudes. Kaplan says in the same Section XVII (in a passage also quoted before in Chap. 2) that one might have

70

4 The Experience Requirement

propositional attitudes even in a situation of ignorance of the content, and this is so because What allows us to take various propositional attitudes towards singular propositions is not the form of our acquaintance with the objects but is rather our ability to manipulate the conceptual apparatus of direct reference. (1977, p. 536)

So, although contents are the “objects of thought”, there are no propositional attitudes towards them alone. Propositional attitudes are always mediated (and individuated) by a character. Summarizing: instead of taking Kaplan’s view that characters (i.e., the beares of logical truth) are the bearers of apriority as incompatible with calling contents the objects of thought, as Soames does, we can (and should) have a more charitable reading. When Kaplan uses the term ‘objects of thought’ he most likely means it in a non-standard way, without implying that they are the object of our propositional attitudes.

4.3.3 De Re Propositional Attitudes So far, the critical points discussed are concerned with specific details of Kaplan’s treatment. But Soames’ third critical point concerns not only Kaplan’s example of contingent a priori truths involving ‘dthat’, but also Kripke’s meter case, as we shall see.21 Soames criticism is directed at the following thesis, which is fundamental both to Kaplan and to Kripke: If a sentence is valid (logical truth) then it is a priori.22 As we know, this thesis is behind examples like dthat[α] = α which is a logical truth in Kaplan’s formal system (because in any context, the object referred to by ‘dthat[α]’ is the object selected by the description α). According to the thesis in question, if anything is a logical truth, it is a priori. But if α is non-rigid, then the proposition expressed is contingent. Soames considers the following example (adapted from Quine’s classical example): (1) Dthat[the youngest Chinese spy] is the youngest Chinese spy.23

21 His

point is basically a repetition of the main argument in Chapter 16 of Soames (2003). as Soames formulates it, logical truth is a species of the a priori. 23 For the sake of brevity, I omit the clause ‘if anyone is’ from this and the following examples. 22 Or,

4.3 Soames and “The Serpert’s Egg”

71

According to Kaplan’s characterization, the truth of this statement can be known simply on the basis of the truth of (2) The youngest Chinese spy is the youngest Chinese spy (which is certainly a priori) and the semantic properties of the ‘dthat’-operator. (A relevant element in this example is that, presumably, nobody knows who the youngest Chinese spy is, although we know that there must be one, if the set of Chinese spies is non-empty.) In other words, (1) can be known a priori. According to Soames, however, the example seems to require something impossible, namely, that solely on the basis of (2), which is a tautology, we could come to know of a certain individual that he or she is the youngest Chinese spy. That is to say, Soames is skeptical about the possibility that, just by manipulating a conceptual apparatus, we can know of the individual referred by ‘dthat[the youngest Chinese spy]’ that he or she is the youngest Chinese spy, and this is required to know the truth of (1). The same objection is repeated against Kripke, but with some minor changes so as to make the example dependent not on the rigidity of ‘dthat[the youngest Chinese spy]’ but on the rigidity of a proper name ‘Lee’ that has its reference fixed by means of the description ‘the youngest Chinese spy’: (3) Lee is the youngest Chinese spy. Soames repeats the point: in order to know (3), one must know of the individual that is the youngest Chinese spy that he or she has this property, and this is something that cannot be known a priori, but only by means of empirical research (that, ultimately, will involve some perceptual experience). To press this point, he imagines two possible scenarios. In the first one, there is no epistemic contact with the individual referred to by ‘Lee’, i.e., the relevant subject has no idea at all about who could be the youngest Chinese spy. According to Soames, in this circumstance there can be no knowledge of the truth of (3), and a mere linguistic stipulation (or a “baptism ceremony”) cannot change that. In the second scenario, the subject has, on the basis of perceptual experience and collected information, some kind of knowledge or belief of a particular person that he or she is the youngest Chinese spy. In this case, the subject can know the truth of (3), but this can hardly be called a priori knowledge. That is to say, either we have some sort of knowledge about a certain person that he or she is the one corresponding to a description (and in this case we have only a posteriori knowledge), or we have no previous knowledge at all, and in this case, Soames concludes, we cannot say that there is knowledge of (3).24

24 The same argument was formulated, mutatis mutandis, for Kripke’s meter case in Chapter 16 of Soames’ (2003). One difference between the two cases is that the named object in the meter case is an abstract length, with which we can (or, in Soames’ understanding, must) have acquaintance through some physical object, while in the Chinese spy case the named object is concrete (a person). For this reason, it is not clear that Philonous’ Objection would apply here.

72

4 The Experience Requirement

One reason for this perspective comes from the consideration not exactly of statements like (3) above, but of statements that report some sort of a priori knowledge, in which (3) appears within the scope of epistemic verbs, as in (4) Dirk knows a priori that dthat[the youngest Chinese spy] is the youngest Chinese spy That is to say, the attribution of a priori knowledge seems to require, or so Soames thinks, that we are capable of reporting it in indirect discourse. But the truth of (4) seems to require that Dirk knows the proposition embedded in the epistemic context. This is the result of an interesting semantic phenomenon, first noticed by Perry (1977, p. 19) and by Kaplan (1977, p. 557). In order to understand this phenomenon, we must have present some important features of Kaplan’s theory (that are basically also features of Perry’s theory, despite some differences in the terminology). As we saw, Kaplan distinguishes the content expressed by a sentence in a context (which is a Russellian proposition) from its cognitive significance, which is not a proposition but its character: it is the character that has relevance for our beliefs, at least for those beliefs related to our location in the world, those which Perry calls “selflocating beliefs”. For this reason, Kaplan and Perry think that beliefs (insofar as they are relevant for action) are basically individuated by their character and not by their content. Although this is so, whenever we want to report a belief in indirect speech, the content gains priority over the character. Consider the situation in which Dirk thinks on September 20th (5) I should give a lecture today According to this picture, Dirk’s basic cognitive relation is with the character associated with ‘I’ and ‘today’. However, if we want to report the next day in indirect speech what Dirk believed the day before, it would not do simply to place (5) within the scope of the epistemic verb, as in (6) Dirk believed that I should give a lecture today since this clearly gives the wrong report. To give the right report we would have to say something like (7) Dirk believed that he should have given a lecture yesterday In (7), what is of relevance in the embedded sentence is not the character (since it changed from (5) to (7)), but the content. This is probably the reason why Soames points out that the truth of (4) requires that Dirk knows of the person that is the youngest Chinese spy that he or she is the youngest Chinese spy. I do not think that Soames’ critical conclusion follows from his consideration of de re belief reports such as (4). The change in perspective by means of which character loses preponderance to content is pointed out by Kaplan and Perry as a semantic phenomenon that arises within indirect belief-reports, but that does not necessarily imply something about the nature of belief (and, hence, about the nature of a priori de re knowledge). Although we are, when reporting a third-person a

4.3 Soames and “The Serpert’s Egg”

73

priori belief, forced by our semantic apparatus to privilege the content instead of the character, this does not necessarily mean that a priori belief is a relation between a subject and a proposition. There are several examples illustrating the fact that one can have a priori knowledge of the character, without knowing the corresponding proposition. Perry (1977) presents the example of crazy Heimson who believes himself to be David Hume. However mistaken he might be about the proposition expressed by his utterance of ‘I am here’, he nevertheless has a priori knowledge corresponding to the character of ‘I am here’. Dirk might suffer from amnesia and have no clue about his identity and about the fact that he is in Rio de Janeiro, but nevertheless he has a priori knowledge corresponding to ‘I am here’. If we try to report what he believes, we will have to present the proposition that Dirk is in Rio de Janeiro, but that does not imply that there is a cognitive relation between him and this proposition.25 Be that as it may, the main reason for Soames’ rejection of Kaplan’s and Kripke’s cases of contingent a priori is that he endorses a view according to which knowing of a certain object that it has a certain property always requires some sort of direct cognitive contact with that object and, hence, in a typical situation, something that cannot be achieved without some experience. It follows from this view that neither Kaplan’s examples of contingent a priori involving the ‘dthat’operator nor Kripke’s cases involving descriptive reference-fixing of rigid names are genuine cases of a priori knowledge.26 More generally, this view is that having a singular thought requires having some special, direct contact with the particular that constitutes it. This is in sharp contrast to Kaplan’s position in the crucial passage from “Epistemological Remarks” (p. 536) (quoted in the previous section) that “There is nothing inaccessible to the mind about the semantics of direct reference, even when the reference is to that which we know only by description”. Soames disagrees: As I have indicated, I do not find this claim credible. (2005, p. 53)

He offers no special argument for this view on de re propositional attitudes. On the contrary, he shows awareness that it is not consensual, and that many hold the

25 Actually, there is a stronger and a weaker version of this claim about the epistemic preponderance of the character. The weaker is that the subject of our propositional attitudes is the character taken together with a content. The stronger is that character alone is the subject of propositional attitudes. Perry seems closer to the weaker claim. But Kaplan seems closer to the stronger claim, when he says, at the end of “Epistemological Remarks”, that

This doesn’t prove that the cognitive content of, say, a single sentence or even a word is to be identified with its character, but it strongly suggests it. (1977, p. 532) 26 Soames

extends this negative result for cases involving natural kind terms (2005, p. 63).

74

4 The Experience Requirement

opposite view (i.e., no acquaintance and no perceptual experience are necessary to have singular thoughts).27 So, as it stands, we do not find a principled reason for his exclusion of Kaplan’s cases of contingent a priori truths involving ‘dthat’ and Kripke’s involving descriptive reference-fixing. As it seems, Soames’ only motivation is to discredit those elements in Kaplan’s and Kripke’s thought that could motivate and support ambitious two-dimensionalism because It is all too easy for ambitious two-dimensionalists of different stripes to look back at Kaplan’s text and find passages that seem to support their particular two-dimensionalist theses. (2005, p. 54)

However, as I see it, in his strife for depriving his philosophical rivals of some inspiration, Soames seems to mutilate Kaplan’s theory in some fundamental ways and leave out some of the riches and interesting aspects of it, thereby throwing out the baby with the bathwater.

4.4 Some Partial Conclusions As we saw throughout this chapter, despite appearances to the contrary, it is quite hard to get a satisfactory articulation of what I called the experience requirement (here represented by Platinga’s, Salmon’s and Soames’ positions, but embraced, explicitly or implicitly, by many critics of the very notion of the contingent a priori). On the one hand, Philonous’ Objection seems to threat the requirement of perceptual contact with the standard of measurement. Due to the possibility of multiple impressions from multiple perspectives, such contact is not sufficient to place a stipulator in epistemic contact with an abstract length. (As we saw, Salmon considers this threat, but does not take it seriously enough.) On the other hand, some of Salmon’s and Soames’ objections seem to be based on controversial principles. In Salmon’s case, that we can only know the length of anything as a result of measuring it. This is controversial because measuring already presupposes that there is a standard of measurement in use, but this only seems possible if some sort of stipulation was already previously accepted (this involves knowing a priori the length of something used as standard). Soames raises three objections, the first two based, as I argued, either in something that Kaplan does not really assume, or in taking Kaplan’s terminology too literally. The third and most important objection depends on an alternative view on singular thoughts for which he ultimately offers no additional argument except the claim that, if we embrace Kaplan’s and Kripke’s instrumental view, we open the door for ambitious two-dimensionalism. This objection is clearly not well grounded. 27 In

Chapter 16, Footnote 13 of (2003) he mentions Harman (1977), Jeshion (2002), besides Kaplan (1977) himself as holding the opposite view. More recently Jeshion (2010) and Hawthorne and Manley (2010) also hold it.

References

75

Acknowledgements I thank the editors of the Proceedings of the 4th Latin Meeting in Analytic Philosophy for the kind permission to use some of the printed material in this chapter.

References Berkeley, G. (1713). Three dialogues between Hylas and Philonous. Reprint Edition. Chicago: The Open Court Publishing Company, 1906. BonJour, L. (1998). In defense of pure reason: A rationalist account of a priori justification. Cambridge: Cambridge University Press. Harman, G. (1977). How to use propositions. American Philosophical Quarterly, 14(2), 173–176. Hawthorne, J., & Manley, D. (2010). The reference book. New York: Oxford University Press. Jeshion, R. (2002). Acquaintanceless De Re Belief. In J. Campbell, M. O’Rourke, & D. Shier (Eds.), Meaning and truth. Investigations in philosophical semantics (pp. 53–78). Oxford: Oxford University Press. Jeshion, R. (2010). Singular Thought: acquaintance, semantic instrumentalism and cognitivism. In R. Jeshion (Ed.), New essays on singular thought (pp. 105–149). New York: Oxford University Press. Kant, I. (1787). Kritik der reinen Vernunft, Edited and Translated by Guyer, P, and Wood, A. Critique of pure reason. Cambridge: Cambridge University Press, 1998., Riga: Verlag Johann Friedrich Hartknoch. Kaplan, D. (1977). Demonstratives. An essay on the semantics, logic, metaphysics and epistemology of demonstratives and other indexicals. In J. Almog, J. Perry, & H. Wettstein (Eds.), Themes from Kaplan (pp. 481–563). Oxford: Oxford University Press, 1989. Kennedy, R. (1987). Salmon versus Kripke on the A Priori. Analysis, 47(3), 158–161. Oppy, G. (1994). Salmon on the contingent A Priori and the necessary A Posteriori. Philosophical Studies, 73(1), 5–33. Perry, J. (1977). Frege on demonstratives. The Philosophical Review, 86(4), 474–497. Plantinga, A. (1974). The nature of necessity. New York: Oxford University Press. Ruffino, M., Venturi, G., & San Mauro, L. (2020). Speech acts in mathematics. Synthese, 198, pp. 10063–10087. Salmon, N. (1986). Frege’s puzzle. Atascadero, CA: Ridgeview Publishing Company. Salmon, N. (1987). How to measure the standard metre. In Proceedings of the Aristotelian Society, 88, 193–217. Soames, S. (2003). Philosophical analysis in the twentieth century. Volume I, The dawn of analysis. Princeton University Press. Soames, S. (2005). Reference and description: The case against two-dimensionalism. Princeton University Press. Stojanovi´c, I. (2004). The contingent a priori: Much Ado about nothing. Croatian Journal of Philosophy, 4(11), 291–300.

Chapter 5

Kripke’s Reformulation of the Contingent A Priori

In a series of lectures given some years after the publication of Naming and Necessity, Kripke (1986) introduces some slight but relevant changes in his views concerning descriptive reference-fixing with impacts for the notion of the contingent a priori. These changes are, in part, a reaction to Donnellan’s and Salmon’s objections discussed in Chaps. 3 and 4, respectively. Although Kripke makes some concessions to both Donnellan and Salmon, the main point of these lectures is to show that the objections raised leave untouched the core of the original position from Naming and Necessity concerning this kind of truths. It is worth going through some of the changes and assess their plausibility. This is the subject of this chapter.

5.1 Elaborating On Donnellan’s Distinction As we saw in Chap. 3, the distinction between knowing that a sentence is true and knowing which truth it expresses is central to Donnellan’s evaluation of Kripke’s cases; according to Donnellan, in some situations one might know a priori that a sentence expresses a true proposition without knowing which proposition it expresses, and this is, according to him, the case in Kripke’s examples of contingent a priori truths deriving from descriptive reference-fixing of directly referential names. One of the things that Kripke does in these lectures is to make Donnellan’s distinction as clear and plausible as possible, and then evaluate how far it really impacts his view on the contingent a priori. Suppose someone wanted to resist Donnellan’s distinction by saying the following: given that the name ‘Neptune’ has its reference fixed by ‘the planet causing the perturbations in Uranus’ orbits’, then we can know a priori that the sentence Neptune is the planet causing the perturbations in Uranus’ orbits (if there is one) is true. Now, appealing to something like Tarski’s T-schema, one could say that it is also known a priori that © Springer Nature Switzerland AG 2022 M. Ruffino, Contingent A Priori Truths, Synthese Library 443, https://doi.org/10.1007/978-3-030-86622-8_5

77

78

5 Kripke’s Reformulation

(TN)

‘Neptune is the planet causing the perturbations in Uranus’ orbits (if there is one)’ is true iff Neptune is the planet causing the perturbations in Uranus’ orbits (if there is one).

Since we know a priori the left-hand side of the biconditional (because of the stipulation), then, by implication, we know a priori its right-hand side as well, i.e., we know a priori that Neptune is the planet causing the perturbations in Uranus’ orbits. That is to say, we know a priori the singular proposition containing Neptune and corresponding to an astronomical fact. Kripke argues that a priori knowledge of things like (TN) is unwarranted because we might not know which proposition is expressed on the right-hand side of the biconditional. Indeed, one might know only a posteriori which proposition that is. An example illustrating this situation is the following: suppose we introduce a symbol ‘’ to express the proposition (1) That Aristotle’s tallest great-grandfather was left-handed if it is true that Aristotle’s tallest great-grandfather was left-handed, or (2) That 2+2=4 if it is false that Aristotle’s tallest great-grandfather was lefthanded. Now, no matter whether Aristotle’s tallest great-grandfather was left-handed or not, ‘’ must certainly express a true proposition, for if he was left-handed, then ‘’ expresses the true (contingent) proposition that he was left-handed, and if he was not left-handed, then ‘’ expresses the true (necessary) proposition that 2 + 2 = 4. Assuming that we do not know whether Aristotle’s tallest great-grandfather was lefthanded, we do not know which proposition ‘’ actually expresses either. Hence, we cannot say that we know that ; but we do know a priori that ‘’ expresses a true proposition. Indeed, the situation is that we do not know a priori whether the proposition expressed by ‘’ is knowable a priori or only a posteriori. One can cast Donnellan’s criticism of Kripke in the following terms: by fixing the reference of a name and stipulating that it is rigid we may know a priori that a sentence containing this name expresses a true proposition, in the same way that we know that ‘’ expresses a true proposition; but we do not know a priori which proposition this sentence expresses, as we do not know a priori which proposition ‘’ expresses. This is so because which proposition is actually expressed depends on an empirical fact after all (i.e., it depends on whether Aristotle’s tallest greatgrandfather was left-handed or not). This contrasts with knowing that ‘Aristotle’s great-grandfather was left-handed’ expresses the proposition that Aristotle’s great-grandfather was left-handed In this case one knows a priori which proposition is expressed because the proposition is determined by semantics alone (and, let us assume, one knows a

5.2 Three Models of Descriptive Reference-Fixing

79

priori everything that depends on semantics alone) . In the ‘’ case, on the contrary, semantics alone is not enough to determine the proposition expressed.1 How does the ‘’ example apply to the Neptune-case? Here we have the rigid proper name ‘Neptune’ and, therefore, the proposition expressed by a sentence containing it is a singular proposition containing the object Neptune. Now, if Le Verrier does not know which object Neptune is (before observing it), he cannot know the proposition that includes the object Neptune.2 It would be different if we were dealing with a Fregean proposition, or if the proposition only involved properties such as the one expressed by ‘every star older than 13 billion years is hot’, which does not include any particular object (only the properties of being a star, being older than 13 billion years, and of being hot); one can have perfect understanding of this proposition without necessarily knowing a single instance of a star that has this property.

5.2 Three Models of Descriptive Reference-Fixing In Naming and Necessity, we are presented with a simple picture of the mechanism according to which a definite description of the form the φ is used to fix the reference of a name ‘N’: one stipulates that ‘N’ is to refer to the only object that has the property φ (if there is one). In a footnote, Kripke suggests that baptism by ostension

1 As

Kripke says, knowing a priori that ‘’ expresses a truth does not mean knowing a priori the proposition expressed by ‘’, since the latter can only be known a posteriori. 2 This seems to be an elaboration of the main point made against Kripke by Levin (1975), and that inspired Donnellan’s similar criticism. Salmon (2020) formulates a similar point in more general terms. The point is that, central to Kripke’s claim that (N) is a priori, is the assumption that (TN) has this same status. However, this is illusory. As Salmon points out (employing the Church-Langford translation test), a T -sentence such as ‘Snow is white’ is true in English if and only if snow is white is entirely equivalent in content to its French translation ‘Snow is white’ est vrai en anglais si et seulement si la neige est blanch which lays down the conditions for ‘Snow is white’ to be true in terms of a non-linguistic necessary and sufficient condition. However, this latter sentence is contingent and a posteriori in the sense that a competent speaker of French but with no knowledge of English cannot know it without experience. There is an interesting passage in Russell (1912, p. 57) in which he says that, by using the name ‘Bismarck’ in ‘Bismarck was an astute diplomat’, we cannot know the proposition that contains Bismarck himself since, for us, the name ‘Bismarck’ is an abbreviation of a description; only Bismarck himself could know the singular proposition that has himself and the property of being an astute diplomat as elements (call it proposition (B)). The most that we can do, according to Russell, is to describe that proposition (B); i.e., although we can talk about (B) and know many things about it, we do not have acquaintance with it since we do not have acquaintance with Bismarck.

80

5 Kripke’s Reformulation

(i.e., a situation in which one points at an object and names it) can also be understood under this simple schema: The case of baptism by ostension can perhaps be subsumed under the description concept also. [. . . ] Usually a baptizer is acquainted in some sense with the object he names and is able to name it ostensively. (1980, p. 96, footnote 42)

This is so, presumably, because ostension involves a demonstration and a demonstration can be seen as a sort of description of the demonstratum. But in his later lectures, and in view of Donnellan’s point regarding acquaintance with the named object, we have a different perspective. Kripke now sees three different ways in which a definite description might be used to fix a reference, and they result from an application of Donnellan’s (1966) famous distinction between referential and attributive use to the reference-fixing description: (1) The description employed in the reference-fixing might be used to present an object (concrete such as a chair or abstract such as a color, a length, etc.) that the speaker is already (and independently) acquainted with. In these cases, the description has a demonstrative element, being therefore rigidified. This can be represented, using Kaplan’s ‘dthat’ operator, as ‘dthat[the color of. . . etc.]’. Kripke calls cases like this “acquaintance guiding rigid fixing of a reference”, i.e., the description fixes the reference by directing the speaker’s attention to an object that he or she is already (and independently) acquainted with. In these cases, one makes a referential use (in Donnellan’s sense) of the definite description. The descriptive content is, so to speak, not essential; what is essential is, as Kripke explains, the object that the description is “pointing at” and with which the speaker is already familiar with in another way.3 (2) A description might be employed with no presentation at all of the reference (in which case no acquaintance is involved). In these cases, the description has a purely attributive use (in Donnellan’s sense), and it is essential that the object referred satisfies its descriptive content.4 (3) There might be hybrid cases in which the reference-fixing description is employed attributively, but also presents an object of acquaintance, and it is essential for the understanding of the description that the speaker has a particular object in mind as the one corresponding to it. These are hybrid cases in which the description has both a descriptive and a referential use (in Donnellan’s sense). The first kind of case can be illustrated by a situation in which one fixes the reference of a name ‘G’ by using ‘the color of this book here’ or the reference of ‘Harry’ as ‘the man sitting over there’. It is not hard to imagine a Donnellan-like case in which one uses the description to capture an object that does not correspond to it (such as ‘the man drinking Martini’ to indicate a man drinking only water). E.g., if I decide 3 1986,

p. 35.

4 This is, perhaps, the most common understanding of the role of definite descriptions in reference-

fixing in Naming and Necessity.

5.3 The Reply to Donnellan’s Acquaintance Requirement

81

to introduce the name ‘G’ using ‘dthat[the color of book B]’ for the referencefixing, the book being actually green, but for some reason (e.g., some effect of a particular illumination) I am under the impression that it is red, then the description is calling my attention to the color red (and not to the color green), and it is this color that is named ‘G’. Here the description is rather similar to just pointing to a referent (in this case, pointing to the color red) and the descriptive content, even if it is rigidified, is not really essential because I can misapprehend or misremember it lately. The second kind is illustrated by the Neptune case, i.e., the descriptive content is essential and the speaker, supposedly, has no acquaintance at all with the object captured by the description. And the third kind is illustrated by the meter case: by being presented to the meter stick, I have my attention called to a certain length that I am already independently acquainted with. But it is also crucial for the referent that it is the length of this particular stick taken as standard of measurement, i.e., it has to satisfy the description. Thus, we have a hybrid case of use of a description, in which both the descriptive element and the object aimed at are essential. Using the distinction above, Kripke advances the two most important points in these lectures that represent a revision of his original view from Naming and Necessity concerning contingent a priori truths. Let us discuss these revisions in turn.

5.3 The Reply to Donnellan’s Acquaintance Requirement Donnellan’s criticism that there is no real knowledge involved in the Neptune case (apart from metalinguistic knowledge) rests on two assumptions: (I) If we are dealing with rigid designators such as ‘Neptune’ (and assuming that it was intentionally and explicitly introduced as such by Le Verrier), then de dicto knowledge reported using this term implies de re knowledge, i.e., knowing that Neptune is the planet responsible for the perturbations in Uranus’ orbits implies knowing of Neptune that it is responsible for the perturbations in Uranus’ orbits. (II) We do not know (or Le Verrier does not know) of Neptune that it is the cause of perturbations in Uranus’ orbits if there is no way of identifying Neptune and distinguishing it from all other objects. The first assumption seems reasonable given the fact (granted by Donnellan as a theoretical possibility, as we saw in Chap. 3) that ‘Neptune’ is directly referential (i.e., de jure rigid). For, if this is the case, then the contribution of the name ‘Neptune’ to the relevant proposition (which must be a Russellian singular proposition) is not any descriptive content or property, but the planet Neptune itself, and thus knowing this proposition requires knowing of Neptune that it has the property of being the cause of the perturbations of Uranus’ orbits. But this is precisely the kind of knowledge that one (i.e., Le Verrier) cannot have if the second assumption is not satisfied: by fixing the reference of ‘Neptune’ with the description in question,

82

5 Kripke’s Reformulation

and saying ‘Neptune is the planet causing the perturbations in Uranus’ orbits’, Le Verrier expresses a singular proposition (if there is such a planet), but he is not in a position to know the proposition that is expressed; what is missing to bridge the gap between knowing that a sentence expresses a true proposition and knowing that proposition is some way of identifying the object that is part of the proposition. The second assumption can be seen as a restriction: one can introduce a name and consider it as part of one’s language only if the name is accompanied by an identifiability requirement.5 If there is no way of distinguishing who or what the name refers to from any other object, then the name is not fully part of one’s language, despite the fact that it was introduced by a reference-fixing description. E.g., if I introduce the name ‘Lee’ as referring to the person with the highest level of hemoglobin in China during the entire Ming dynasty (without having the slightest idea of who that person could be), and if I explicitly mean ‘Lee’ to be a de jure rigid name, then, in a way, ‘Lee’ is not really part of my language because I do not fully understand it. The identifiability requirement might have different degrees of strength: it might be liberal in the sense that just being able to provide some identifying description would be enough, or it might be very strict, in the sense that only in the presence of acquaintance with the referred object there would be real identification. Kripke raises three fundamental questions concerning the second assumption. First, whether it is justified or even reasonable. Second, whether it is satisfied in the Neptune and meter cases. Third, whether the examples of contingent a priori disappear if such requirement is satisfied (and there are identifiabity criteria for the reference of names introduced by descriptive reference-fixing) or whether they come up in a different form. In particular, do they disappear if the requirement is satisfied in its strongest form, i.e., acquaintance? With respect to the first question, in an apparent concession to Donnellan and somewhat in the opposite direction of the one taken in Naming and Necessity (in which no acquaintance or even identification criteria seem to be required or to play any role for reference or thought), Kripke now thinks that some identity criterion is indeed necessary for reference. He concedes that even acquaintance with the reference might sometimes be needed in descriptive reference-fixing. This happens in part in the meter case; one cannot be said to know the reference of ‘one meter’ based solely on the description ‘the length of stick S at t0 ’ if one does not have some form of independent acquaintance with the abstract length of one meter. And sometimes it is fully satisfied, as in the case in which I name a color sensation, or (this is a slightly more complicated case) when I name a proposition such as ‘the proposition expressed by this utterance. . . [with a demonstration of an utterance expressing something that I am fully acquainted with]’. Concerning the second question, let us remember that Donnellan’s point about the Neptune case (i.e., that one does not know that Neptune is the cause of the

5 This is the term that Kripke uses meaning by it, or so I take it, a criterion for distinguishing the reference from any other object.

5.3 The Reply to Donnellan’s Acquaintance Requirement

83

perturbations in Uranus’ orbits, although one might know that ‘Neptune is the cause of the perturbations in Uranus’ orbits’ is true) comes from the fact that knowing which proposition the sentence expresses requires knowing which object constitutes the proposition, and this seems to require some sort of acquaintance with it. However, as Kripke sees it, Donnellan goes too far: we do not need to slide from a reasonable identifiability requirement into an acquaintance requirement as a general rule. We have (or Le Verrier has) actually established, through the description that fixes the reference of ‘Neptune’, a criterion of identity of Neptune that distinguishes it from every other object. This is so because there is not just one single simple description backing the introduction of the name, but a complex of descriptions that, taken together, provide enough information so that one can distinguish the object named by ‘Neptune’ from all other objects (if it exists). In the case of the name ‘Neptune’, the description is ‘the cause of the perturbations in the orbits according to Newtonian mechanics’, and the last part (in italics) brings into the description a whole cluster of descriptions such as ‘it can be seen at such and such position in such and such a time’, etc. Hence, one can know exactly where and when to look for the object (if it exists); if the object is not located at the time and place predicted by the complex description, then we would probably have to say that the description does not have a reference after all. Thus, by introducing ‘Neptune’ as a rigid name and presenting that complex description, Le Verrier does have a criterion of identification of Neptune basically because he can predict its location in space and time (if it exists), although he has no acquaintance at all with Neptune.6 Concluding: the Neptune case is one in which the definite description employed in referencefixing has a purely attributive use, there being no acquaintance involved. However, the identifiability criterion is satisfied because the description is actually a cluster of descriptions that includes enough resources to distinguish the referred object from any other; e.g., it contains the ‘according to Newtonian mechanics’ clause, which brings in the entire resources of Newtonian mechanics to predict its position in space and time. How does the meter case fare in view of the new identifiability requirement? Differently from the position defended in Naming and Necessity, Kripke now sees the meter case as intermediary between the Neptune and the color cases (i.e., a case in which some sort of acquaintance is necessary). This is a concession to critics such as Plantinga and Salmon (as we saw in Chap. 4): if someone has no visual contact whatsoever with the stick, then one cannot be said to know that it is one meter long because one might have a completely wrong idea of what the abstract length of one meter looks like.7 This seems plausible. Imagine the following situation: some

6 In suggesting this alternative, Kripke also seems to make a concession to the so-called cluster theory of proper names, a theory that was under attack in Naming and Necessity. 7 “Sometimes I thought one could swallow the idea that there was no acquaintance involved, but I think probably not with our ordinary concept of a meter. Someone who thought that a meter was a mile long or, say, ten miles long—even though he remembered it was the length of a certain stick but somehow got a fantastic idea of how long that stick was or had never seen the stick [. . . ] even though he remembered the definition ‘that, the length of this stick’ (and so on) probably shouldn’t

84

5 Kripke’s Reformulation

scientist reads an ancient script about the life of an eccentric pharaoh who wanted to be buried in the desert together with a platinum stick that he thought to have the power of helping him in the journey after death. There is good evidence that the script is accurate and that the stick really existed, but the script does not give details such as how long it was nor the exact location where the pharaoh was buried. For all that the scientist knows, the stick is buried somewhere in the middle of a large desert, and can be very small like a pin, or very large, made with all the platinum available at the pharaoh’s time. The scientist, fascinated by the story of this stick, decides to fix the meter as (MP) One meter is the length of the pharaoh’s stick at the time of the burial. So the scientist has some idea about the pharaoh’s stick and good evidence that it really exists, but has no idea concerning its length. The scientist knows that (MP) expresses a true proposition, but given that ‘one meter’ is a directly referential device for an abstract length, and given that the scientist has no idea concerning which length that could be, it seems (or so Kripke now plausibly thinks) that the scientist does not know which proposition it expresses. In the meter case, in order to fully understand the notion introduced by descriptive reference-fixing in terms of the length of the standard stick, it may be important to actually see the stick. If I am in possession only of the definition of a meter as the length of the stick, I might be under a wrong impression about how long it is (or have no idea at all, like the scientist in the pharaoh’s stick case) and, hence, whenever using the term, I may be using it incorrectly. The meter stick is useful for calling my attention to a length, but this is a length with which I am already independently acquainted. Of course, the stick is less than absolutely precise as an acquaintance guide, and this is so for two reasons. First, ‘the length of stick S’ has some residual intrinsic vagueness since the stick is a physical object and, as any physical object, there is always some vagueness regarding its boundaries. Second, there is also some vagueness in my perception since I cannot, e.g., visually distinguish a stick that is exactly one meter long from another one that is between 0.999 and 1.001 meter long. But using a stick and the description ‘the length of this stick’ is a better way to drive my attention to the precise abstract length of one meter than my own visual capacity or imagination. (I can certainly imagine the abstract length, but I can hardly distinguish it in my imagination from other lengths very close to it.) This is a mixed case because here the descriptive content of ‘the length of stick S’ is crucial (since no other way of getting the object would be precise enough, i.e., without the stick to guide my attention I could never get clear about the exact length of one meter). But, at the same time, the point of the description is to drive my attention to a specific object (i.e., an abstract length) with which I be said to be using the term ‘meter’ properly. ‘Oh, look, you’re asking me to walk a meter just to get to this place. I have to go by car’. Such a person probably doesn’t know what a meter is and isn’t using the term correctly even though he remembers the stick but somehow has gotten this wild idea of how very long it was. So, to that extent, this is not like the Neptune case.” (1986, p. 37).

5.4 Contingent A Priori Truths as the Basis of Other Contingencies

85

am already independently acquainted. Hence, in this particular case the use of the description is both attributive and referential. Concerning the third question, Kripke makes, contra Donnellan, the more general point that, even if one places the strongest form of the identifiability requirement, some form of contingent a priori truths still emerges. Suppose that there is an object o with which I am acquainted, and for which I have the name ‘N’. Now, that name was introduced using a definite description ‘the φ’. Even if the reference-fixing was by ostension (a case in which I name an object of my acquaintance), this can be subsumed under the general case of reference-fixing using a description (in this case, something like ‘dthat[the φ]’), as we saw in footnote 42 from Naming and Necessity quoted above.8 Now, even in this case we have N is the φ as something that can be known a priori although it expresses a contingent proposition. This comes from the fact that, although ‘dthat[the φ]’ is rigid, ‘the φ’ is not rigid since it could have picked a distinct object that is qualitatively like o (i.e., that would appear to me in exactly the same way that o does).9

5.4 Contingent A Priori Truths as the Basis of Other Contingencies Quite apart from the reformulations of the original view on contingent a priori truths accommodating acquaintance and identifiability criteria, Kripke also makes, in these lectures, an important point against the sort of criticism raised, e.g., by Salmon that a priori knowledge of the length of the stick is impossible since it can only be discovered a posteriori as a result of measurement. His general reply to this view is that, if we do not know a priori that S is one meter long, we cannot know its length a posteriori either and, more broadly speaking, we could not know the length of anything a posteriori.10 While the claim might seem strange at first sight, it actually indicates something quite deep that goes straight to what we could call a transcendental condition of the experience of measuring: there cannot be any such

8 This is so unless we hold a view similar to Russell’s (1912) according to which I can have names of objects of my acquaintance without any guiding description at all. Indeed, this is the essence of acquaintance for him; if there is any guiding description, no matter how close it is to bare perception, this is not Russellian acquaintance anymore. 9 “Place the strictest requirements that you want on the introduction of a name (and this is as strict as I think someone could reasonably get—unless he was Russell), then you will still have an example of the contingent a priori of the very type that I was talking about in Naming and Necessity because, here, this will be a proposition fully in my language and it will have exactly the properties of the proposition about Neptune.” (1986, p. 57). 10 Kripke (1986, p. 63).

86

5 Kripke’s Reformulation

experience unless some standard of measurement is fixed. In this aspect, Kripke’s view is similar to Wittgenstein’s position that he criticized earlier in Naming and Necessity (for other reasons). There is no such thing as measuring in abstract without a standard of measurement. But in fixing the standard of measurement say, by stipulating that stick S is one meter long, one thereby knows a priori that stick S is one meter long. A standard of measurement must be fixed somehow so as to make it possible to measure anything. If there is no meter, one cannot measure anything to find out that it is 2.75 meters or 0.25 cm long, etc. It is of no help to say that we could use other systems (yards, inches, etc.) instead, because some other measurement system would have to be originally fixed as well, and other contingent a priori truths would then result in a similar way. Fixing the meter by reference to the standard stick implies being able to say, without measuring it, that it is one meter long. The same applies to any other standard of measurement. If I decide, for some crazy reason, to reject the metric and any other existing system and create my own system, defining ‘one Tico’ to be the exact shortest distance between my computer and the cup of coffee next to it (in this exact moment), then I do not have to measure this distance in order to know that the cup is exactly one Tico away from the computer in this exact moment. I may have to measure other distances now, thereby discovering a posteriori that, e.g., the printer is two Ticos away, and the door seven Ticos away, but the possibility of these items of a posteriori knowledge depend on knowing a priori that the distance between the computer and the cup is one Tico.

5.5 Some Partial Conclusions In these later lectures, we find the essence of Kripke’s original position preserved. He makes some few adjustments to accommodate an identifiability requirement for the object whose name is introduced by description. The requirement may take different forms, ranging from a more liberal version–one in which identification is given by other descriptions, possibly derived from the original cluster of descriptions introducing the name and in which acquaintance is not required–to a more strict version, in which acquaintance is required. But even in the strictest version, we do not have to take the Russellian conception of acquaintance, but can take acquaintance as given by a rigidified description dthat[the φ] where ‘the φ’ is equivalent to ‘the object having such and such appearance’. The Neptune case is one in which Le Verrier has the more liberal identification criterion (possibility of determination of Neptune’s location embodied in the complexity of the original description), while in the meter case, the acquaintance needed is not with the stick itself, but with an abstract length, and the stick can help only as long

References

87

as it works as an acquaintance guide to something with which we are independently acquainted.11 In synthesis, this later and more liberal version of Kripke’s contingent a priori truths can get around some of the classic objections advanced against the original formulation. Globally considered, however, it is hard to avoid the impression that Kripke’s suggestion that the meter stick is an acquaintance guide is somewhat artificial and has the sole purpose of accommodating or neutralizing Salmon’s and Plantinga’s complaint that some visual contact with the meter stick is necessary for the stipulation to go through. Apart from its artificiality, it is hard to see how it could work for other cases. For example, for a long time the definition of ‘one kilogram’ was given as the weight of a particular platinum cylinder taken as the international prototype kilogram (called Le Grand K). So, presumably one could in principle also have contingent a priori knowledge of ‘one kilogram is the weight of Le Grand K’. But it seems odd to say that ‘dthat[the weight of Le Grand K]’ is a guide to acquaintance with a weight that we are independently acquainted with. For what could this “abstract weight” be, and how could we be independently acquainted with it? On the other hand, even in the meter case, we nowadays have a different definition: one meter is now defined as the distance travelled by the light 1 in vacuum in 299,792,458 seconds. It is quite doubtful that this description works as an acquaintance guide in this case, for there cannot be acquaintance (at least for beings like us) with such a small fraction. So, what happened here? Did we previously have a definition in need of an acquaintance guide (the meter stick), but now we do not need an acquaintance guide anymore? This gives the impression that Kripke’s suggestion that the meter stick in the original (old) definition is a sort of acquaintance guide lacks a more substantial motivation. On the other hand, by claiming that no a posteriori knowledge of length can be achieved without some a priori knowledge of some length, Kripke makes the forceful point that such truths are not just exotic cases generated by funny uses of language, but are fundamental to the establishment of standards of measurement, without which no measuring (and, ultimately, no science) is possible. The two fundamental gaps mentioned in Chap. 1 still remain, however.

References Donnellan, K. (1966). Reference and definite descriptions. The Philosophical Review, 75(3), 281– 304. Kripke, S. (1980). Naming and necessity. Cambridge, MA: Harvard University Press.

11 We could compare the role that the meter stick has in Kripke’s reformulation to the role that Frege attributes, in the preface of Begriffsschrift, to the written signs of a formal language, i.e., the role of fixing our attention at something (logical relations and inferences) that we already know from some other source.

88

5 Kripke’s Reformulation

Kripke, S. (1986). Rigid designation and the contingent a priori: The meter stick revisited. Exxon distinguished lectures at Notre Dame University. Unpublished. Levin, M. (1975). Kripke’s argument against the identity thesis. The Journal of Philosophy, 72(6), 149–167. Russell, B. (1912). The problems of philosophy. Reprint Edition, With Introduction by John Perry. New York: Oxford University Press, 1997. Salmon, N. (2020). Naming and non-necessity. In A. Bianchi (Ed.), Language and reality from a naturalistic perspective (pp. 237–248). Springer.

Chapter 6

Evans and the Varieties of Contingency

Evans (1979b) presents an almost entirely different approach to contingent a priori truths, which he considers, following Dummett, to be an “intolerable paradox” (p. 161) engendered by Kripke’s framework (if taken at face value). He develops two strategies to deal with what he describes as the “puzzle” represented by such truths, one particular and one more general. His particular strategy focuses on the specific version discussed by Kripke and Donnellan, and centers on descriptive referencefixing. It presupposes Evans’ own theory of what he calls descriptive names, i.e., names whose reference is fixed by a definite description; Evans’ favorite example is ‘Julius’, which he introduces as referring to the inventor of the zip, whoever that is. Although he thinks (p. 162) that such names are very rare in ordinary language (and this is basically an empirical claim for which he presents no evidence), the few cases that do exist are enough to generate the puzzle anyway. But, contrary to Kripke and Donnellan, he thinks that these names do keep their descriptive content after their reference has been fixed. The general strategy, on the other hand, is supposed to deal with all forms of contingent a priori truths, even those that do not involve singular terms. As we know, the main obstacle to this claim is Kripke’s Modal Argument, which is based on the fact that ‘Julius is φ’ and ‘the inventor of the zip is φ’ exhibit different modal behavior, since ‘Julius’ is rigid, while ‘the inventor of the zip’ is non-rigid. The Modal Argument presupposes that two sentences correspond to two different propositions if they differ in their modal profile. When this is so, they have two different contents. In other words, this argument presupposes that content and proposition are one and the same thing and, hence, if two sentences correspond to two distinct propositions, they cannot have the same cognitive content. But this is something that Evans wants to challenge.

This chapter appeared in part in my “Superficially and Deeply Contingent A Priori Truths”, Croatian Journal of Philosophy 16 (2):247–266 (2016). © Springer Nature Switzerland AG 2022 M. Ruffino, Contingent A Priori Truths, Synthese Library 443, https://doi.org/10.1007/978-3-030-86622-8_6

89

90

6 Evans

As we shall see, an important part of Evans’ account depends on the claim that the same content can correspond to two distinct propositions, one of them a necessary, and another one a contingent truth. He takes inspiration in examples like the following pair of sentences: (i) John is as tall as John (ii) John is as tall as himself in which (i) attributes to John a property that some, but not all, objects have, namely, to be as tall as John, while (ii) apparently attributes to John a different property, namely, that of being as tall as himself (which every object has). In symbols, the properties attributed to John in (i) and (ii) can be represented as, respectively, (P) λx(x is as tall as John) (P’) λx(x is as tall as x) Thinking of intensions as functions from possible worlds to extensions, these properties correspond to different intensions, since (P) will select, in each world, the class of objects that are as tall as John, while (P’) will select, in each world, the class that includes the totality of objects in that world. Anyway, Evans’ strategy presupposes dissociating differences in the modal profile of expressions such as ‘Julius is F’ and ‘the inventor of the zip is F’ from differences in their cognitive content: they might behave differently in modal terms but, nevertheless, have the same cognitive content. In what follows I will discuss Evans’ particular and general strategies in greater detail.

6.1 Descriptive Names The phenomenon of contingent a priori truths discovered by Kripke and illustrated in cases such as the Neptune one are derived from the possibility of descriptive reference-fixing of rigid names. Evans understands these examples as dependent on what he calls descriptive names, of which ‘Neptune’ would be an instance. But his understanding of such names is quite peculiar, and does not exactly correspond either to the way that Kripke understands rigid names with descriptively fixed reference, or to Donnellan’s understanding (or Frege’s or Russell’s for that matter). Indeed, on the face of it, Evans’ descriptive names are a kind of monster, similar to those mythological creatures that result from a forbidden relationship between humans and gods. For they result from the identification of expressions belonging to two entirely different semantic kinds: we have a proper name like ‘Julius’, with the essential properties of names such as referentiality and rigidity, having

6.1 Descriptive Names

91

its content (or Fregean sense, as Evans calls it)1 given by a definite description. But definite descriptions, in Evans’ conception, are not only non-rigid, they are not even referential expressions. According to him, definite descriptions are binary quantifiers. E.g., (Z) The inventor of the zip is American is, according to this conception, entirely parallel to All inventors of the zip are Americans Some inventors of the zip are Americans Most inventors of the zip are Americans . . . etc. These can be formally treated as second-order operators that associate pairs of conceptual expressions to truth-values, similar to All(I , A) Some(I , A) . . . etc. (where ‘I ’ and ‘A’ abbreviate ‘inventor of the zip’ and ‘American’, respectively). By parity, (Z) has the form The(I , A) It follows that ‘Julius’ is actually equivalent to The(I , X) where X is a variable ranging over concepts. So, the Fregean sense of a descriptive name in Evans’ conception is not given by another referential expression (as we would expect), but is the sense of a second order operator. The characterization above of ‘Julius’ as a descriptive name involves two claims: first, that ‘Julius’ is a referential expression and, second, that ‘the inventor of the zip’ (which gives the former its Fregean sense) is not a referential expression. So, we must ask: 1. Why is ‘Julius’ a referential expression? Evans offers two reasons. The first, more theoretical and important one, is that we can easily make ‘Julius’ fit into a minimal and general theory of reference. The same does not hold for ‘the inventor of the zip’, as we shall see. A minimal theory of reference (in the sense that it contains all that is necessary and sufficient

1 His understanding of Fregean senses is also quite different from the way that Frege himself understood it.

92

6 Evans

to characterize what is essential in reference) for names (such as ‘Max Freund’) can be given by the homophonic clause: ‘Max Freund’ refers to Max Freund Why is this clause enough? Because it gives everything that, according to Evans, one should expect from any theory of reference, i.e., it can be combined with the notion of truth and satisfaction in order to make the following principle true: If ‘a’ refers to o, and ‘Fa’ is atomic, then ‘Fa’ is true iff o satisfies ‘F’ In our case, If ‘Max Freund’ refers to Max Freund, and ‘P(Max Freund)’ is atomic, then ‘P(Max Freund)’ is true iff Max Freund satisfies ‘P’. In other words, no matter what the foundational reasons are for saying that Max Freund is the reference of ‘Max Freund’ (e.g., that it corresponds to the intention of someone using the name, or that the name is causally tied to Max Freund, or whichever theory one might have that connects ‘Max Freund’ to Max Freund), and for saying that Max Freund satisfies ‘P’, all that matters as far as the concept of reference is concerned, according to Evans, is that it combines properly with the concept of truth and with the concept of satisfaction. The second reason is simply an appeal to the intuition that ‘Julius’ is used rigidly, i.e., that we normally consider sentences like the following If you had invented the zip, you would have been Julius If Julius had not invented the zip, he would not have been Julius. as false. 2. Why are descriptions not referential expressions? Here, Evans offers a meta-theoretical argument quite different from Russell’s reasons for not considering descriptions as referential. For Russell, an expression is not referential if a sentence containing it has truth conditions even under the assumption that the expression lacks a reference, which means that a sentence containing a genuinely referential expression has no truth conditions if the expression has no reference. Evans’ justification is that assimilating descriptions into the category of referential expressions would require making adjustments in the theory of reference in order to leave room for some phenomena that are typical of descriptions. For example, non-rigidity and ambiguity of scope when embedded in sentences containing negation (such as ‘The current king of France is not bald’) or modal operators (such as ‘The first man in space could have been an American’) or epistemic operators (such as ‘George IV wants to know whether Scott is the author of Waverley’). But this would introduce a great discontinuity in the theory of reference, besides making it artificial. Paradigmatic referential expressions (like pronouns and ordinary names) never really take up the possibilities left open by this new (modified) theory.

6.2 Descriptive Names, Free Logic and Evans’ Particular Strategy

93

6.2 Descriptive Names, Free Logic and Evans’ Particular Strategy Consider Kripke’s sentence (N): (N)

If there is one and only one planet causing the perturbations in Uranus’ orbits, then Neptune is that planet.

which we could so represent in predicate logic (N) ∃!xφx → (Neptune = the x φx). ‘Neptune’ is supposed to be rigid, while ‘the planet causing the perturbations in Uranus’ orbits’ is non-rigid, and this is why the truth expressed by the sentence is contingent: there is a possible world in which the antecedent is true (i.e., there is one unique planet causing the perturbations in Uranus’ orbits) and the consequent is false (i.e., Neptune, in that world, is not the planet causing the perturbations). If ‘Neptune’ were equivalent to a description, (N) would be a necessary truth because the consequent would be a tautology. Evans thinks that (N) only presents an interesting case (or “puzzle”, as he calls it) if there is a guarantee that either the name ‘Neptune’ has a reference (in case there is a planet corresponding to the description) or, if the name has no reference (in case there is no such planet), there is at least a descriptive content backing it. This is so because there is always the risk that ‘Neptune’ could turn out to be something like the name ‘Vulcan’.2 If ‘Neptune’ is empty (like ‘Vulcan’) and there is no descriptive content backing it, there is no proposition to be understood as the one expressed by (N) and, hence, no clear candidate for contingent a priori truth. This requires that a name like ‘Neptune’ is descriptive in Evans’ sense (i.e., similar to his example ‘Julius’). This seem to be the reasoning behind his conclusion: for ordinary (i.e., non-descriptive) rigid names we have a semantic clause stipulating their reference as (R) ‘John’ refers to John so that, from (R), we can derive the truth conditions for any sentence φ(J ohn) as ‘φ(J ohn)’ is true iff John is φ But if ‘John’ is an empty name, we cannot derive any informative truth-condition since the right-hand side of the above biconditional would also be empty and, hence, express no proposition. Hence, if ‘John’ is empty, there is no proposition to be understood under φ(J ohn). For descriptive names such as ‘Julius’, on the contrary, since their reference is given by a clause such as (J) ‘Julius’ refers to the one who uniquely invented the zip

2 See

Chap. 1, Footnote 6.

94

6 Evans

we can always derive ‘φ(Julius)’ is true iff the man who uniquely invented the zip is φ which does outline truth-conditions of a sentence of the form φ(J ulius) because the right-hand side expresses a proposition independently of the existence of a man who uniquely invented the zip. Hence, there is a proposition to be understood even in case there is no individual who uniquely invented the zip. So, taking ‘Julius’ to be a descriptive name guarantees the meaningfulness of sentences of the form φ(J ulius) (whereas taking ‘Julius’ to be an ordinary name makes the meaningfulness of φ(J ulius) depend on the existence of its reference). The same would apply for Kripke’s ‘Neptune’, according to Evans. Now, if descriptive names are admitted, we must take some care since some otherwise valid inferences are no longer valid. For example, there might be a situation in which φ(J ulius) is true, despite the fact that there is no Julius. (E.g., the negative existential ‘There is no Julius’ would be equivalent to ‘There is no individual who uniquely invented the zip’, and would be true if there is no such unique individual.) So, classical existential generalization from φ(Julius) to ∃xφ(x) would also be invalid in this situation. Similarly, universal instantiation from ∀x(x = x) to Julius = Julius would be invalid. Hence, some restrictions have to be placed in these logical rules if we are to admit descriptive names.3 In other words, we must adopt some form of free logic if such names are admitted. Evans actually suggests something stronger, in the form of two theses: (i) The acceptance of descriptive names requires the acceptance of free logic (1979a, p. 166); (ii) The acceptance of free logic requires the acceptance of descriptive names (ibid., p. 173).

3 One formal alternative is to allow existential generalization only over portions of the sentence marked with a scope indicator; analogously for universal instantiation.

6.3 Contingency: Deep and Superficial

95

Taken together, these two theses imply that the adoption of free logic and the acceptance of descriptive names go together. Either (N) is formulated using a free logic (i.e., without the assumption that ‘Neptune’ refers), or there is no clear candidate for the corresponding contingent a priori truth: [T]here simply is no puzzle unless the use of free logic is accepted. Unless a sentence concerning the name ‘Julius’ can be formulated which is free of existential commitment, there is not even a candidate for the status of the contingent a priori, but within a classical framework, there are no such sentences. (1979b, p. 172)4

From these considerations comes Evans’ first (particular) strategy to deal with the “puzzle”: according to him, without the adoption of free logic there is not even a puzzle in the first place for, as explained above, there is no guarantee that (N) expresses a proposition to be known. Evans’ paper was published shortly after Donnellan’s (1977), and he has Donnellan in mind as one of his main interlocutors. One of his goals is to reject Donnellan’s approach. Donnellan would not accept the existence of descriptive names in Evans’ sense for two reasons. First, even if a name has its reference fixed by a description and there is an explicit convention that it is rigid, Donnellan thinks (along with Kripke) that the description does not remain attached to the name after fixing its reference. In other words, if a name has no reference, there is nothing to be understood as expressed by a sentence containing it. Second, and more importantly, understanding a name requires, for Donnellan, some sort of contact (acquaintance) with its referent, thus there can be no such understanding if the name does not refer. Hence, Evans’ point against Donnellan is, as he himself calls it, ad hominen: if (N) is to be a candidate for contingent a priori knowledge, then it must not allow for existential quantification in the position occupied by the name. But if this is so, one must accept free logic. But free logic (at least Evans’ version of it) requires descriptive names, i.e., names such that sentences containing them do not depend on the existence of referents to have meaning. Donnellan does not accept such names and hence, as Evans sees it, should not recognize a puzzle in the first place.

6.3 Contingency: Deep and Superficial Whether or not Evans’ reflections on descriptive names and free logic yield a convincing solution to the “puzzle”, the fact is that they turn out to be not strictly necessary, since there is a second, simpler and more general strategy that does

4 Hawthorne and Manley (2010, p. 58) raise a number of objections to this thesis, the most important of which is that there may be formulations of the “puzzle’ in which there is no existential worry involved, e.g., if we name ‘Alpha’ the number of planets, then no matter how the universe turns out to be (e.g., if there are no planets), ‘Alpha’ always has a reference.

96

6 Evans

not concern our theory of reference, but our theory of contingency.5 The strategy is based on Evans’ distinction between what he calls superficially and deeply contingent truths. A sentence P is superficially contingent if and only if it is false at some possible world, i.e., ♦¬P is true. And it is deeply contingent if its semantics alone does not provide a “guarantee that there exists a verifying state of affairs” (1979a, p. 185). If we can talk of contingency as a sort of requirement for truths, superficial and deep contingency require different things. Superficial contingency requires a certain minimal modal profile, while deep contingency requires something from its truthmaker, i.e., that it is not generated by semantics alone. Evans formulates the distinction between superficially and deeply contingent truths primarily for sentences (and not for propositions). This formulation needs some explanation for it might be a bit misleading. Take the example of ‘I am here now’. In any context in which this sentence might be uttered (in any possible world), it takes the agent, the place and the time of the utterance as values, respectively, of ‘I’, ‘here’ and ‘now’, and expresses the propositional content that the agent of the utterance was at the place of the utterance at the time of the utterance. If contexts are restricted to what Kaplan calls proper (see Chap. 2), then in each context the proposition expressed must be true. Hence, if we allow only proper contexts, it might seem that ‘I am here now’ can never express a falsity and, hence, that its negation can never be true. That is to say, ‘♦¬I am here now’ is false. This is certainly not what Evans has in mind, for truths that might be known a priori in virtue of indexicality are, for him, the paradigm of truths that are only superficially contingent. So, Evans’ formulation has to be read together with a qualification concerning the behavior of indexical expressions within modal contexts. Kaplan (1989, p. 499) pointed out that pure indexicals tend to assume wide scope when embedded in modal, temporal and locational operators, i.e., they take as semantic values elements from the context of utterance (instead of values from the context intended with the operator). Hence, in It is possible that in Pakistan, in 5 years, only those who are actually here now are envied (to use an example given by Kaplan), the values taken by ‘here’, ‘now’ and ‘actually’ are, respectively, the place of the utterance (and not Pakistan), the time of the utterance (and not 5 years after it) and the world of the utterance (and not some other possible worlds). This is so, according to Kaplan, because pure indexicals tend to “cling rigidly” to the referents determined in the context of use. Although Evans is not explicit about this, it seems plausible to assume that he presupposes

5 This is also the strategy that is most discussed in the literature on contingent a priori truths. There is hardly any discussion of the first, more particular one.

6.4 Propositions, Cognitive Content and Evans’ General Strategy

97

both Kaplanian theses, i.e., that there are only proper contexts (otherwise ‘I am here now’ would not be an a priori truth) and that pure indexicals always take their values from the context of utterance (otherwise ‘I am here now’ could never be false in any possible world, and hence ‘♦¬I am here now’ would also be false, which means that ‘I am here now’ would not be a paradigmatic case of superficially contingent truth as Evans thinks). Hence, we have to understand Evans’ characterization of superficially contingent sentences in the following way: S is superficially contingent iff ‘♦¬S’ is true provided that all indexicals in S take wide scope and provided that only proper contexts are allowed. Evans claims that the cases presented by Kripke are, at best, only superficially contingent: [T]here is nothing particularly perplexing about a statement which is both knowable a priori and superficially contingent, which is the most that the problematical statements may be claimed to be. (1979b, pp. 161–2)

Hence, he thinks that the semantic stipulations behind Kripke’s cases (the “problematic statements”) guarantee their verification. Presumably, this also goes for Kaplan’s cases since, in them, the truth of the statements in any context is guaranteed by the semantics of indexicals alone. Thus, all cases of superficially contingent a priori truths have the same source. Although Evans does not explain the distinction in further detail in his paper, the classification of Kripke’s cases as “superficially contingent” (and its reduction to indexical cases) seems to bring some relief to those who, like Evans himself (and Dummett before him), take it to be “intolerable” for there to be true sentences that are deeply contingent and knowable a priori.

6.4 Propositions, Cognitive Content and Evans’ General Strategy Since propositions are true or false in possible worlds, we might also think of them as properties that hold or fail to hold of possible worlds, or as requirements for being true in each possible world. Now, a property or requirement of this kind might be based on a contingent feature of the actual world. In this case, on the one hand, it trivially applies to the actual world (because it was extracted from it) and, on the other hand, it is not a property that all possible worlds have (because it is based on a contingency) and hence does not apply trivially to any world. Let C be any contingent fact of the actual world (e.g., that the sky is blue), and consider the property (P) of possible worlds expressed by (P) λw(w includes C iff @ includes C)

98

6 Evans

(where @ is the actual world). ‘@ includes C’ expresses a necessary truth, since in any possible world it will be true that @ includes C. This is a property that not all possible worlds have, but @ certainly has. Now consider a different property (P∗ ) λw(w includes C iff w includes C) (P) and (P*) correspond to two distinct properties, since the first is false of some possible worlds, but the second is true of all of them. However, Evans claims that they yield cognitively equivalent propositions when applied to @, since both result in saying that @ includes C iff @ includes C. What is operative here is a distinction that Evans wants to draw between proposition and content. Proposition is understood in the standard Carnapian way, i.e., as a function from possible worlds to truth values expressed by a sentence, while content is understood in somewhat vaguer terms as “what is believed by one who understands and accepts” a sentence (p. 176). Two sentences might express the same proposition (e.g., two logical truths, which are the same function from possible worlds to truth values) but have different contents. For Evans, content is what is epistemically relevant, since he calls epistemically equivalent two sentences having the same content (ibid.). We saw that Evans’ first strategy, trading on specific features of descriptive names, was actually not necessary, since there was a second broader and more fundamental one to solve the problem of contingent a priori truths in its greater generality. The second strategy explores this distinction by reducing contingent a priori truths to things that do not represent substantial but only trivial knowledge about the actual world. Better said: it reduces them to attributions of nontrivial properties to the actual world that are epistemically equivalent to attributions of trivial properties. How does this general strategy apply to Neptune-like cases (i.e., cases involving descriptive names)? Evans deals with a slightly modified version of Kripke’s (N) sentence,6 but basically with the same effect: (N) If there is one and only one planet causing the perturbations in Uranus’ orbits, then Neptune is a planet causing the perturbations in Uranus’ orbits. In Evans’ formal notation: (N) ∃!xφ(x) → [Neptune]φ(Neptune) Here, as before,‘φ’ abbreviates the property of being a planet causing the perturbations in Uranus’ orbits, while ‘[Neptune]’ is a scope indicator of the name ‘Neptune’, i.e., it indicates that existential quantification is allowed only in the portion of the formula that follows it. In this case, only locally in the ‘Neptune’ position in ‘φ(Neptune)’, but not over the whole formula, i.e., only ∃!xφ(x) → ∃yφ(y) 6 The original version has an identity (‘Neptune = the φ’) as consequent, but Evans’ version has a predication (‘φ(Neptune)’).

6.4 Propositions, Cognitive Content and Evans’ General Strategy

99

can be validly inferred from (N). This is, according to Evans, a necessary condition for there being a puzzle of contingent a priori truth in the first place.7 Now if we consider that ‘Neptune’ is a descriptive name equivalent to ‘the xφ(x)’, and that ‘φx’ is true or false under an assignment of x only in relation to a possible world w, we might rewrite ‘φx’ as a binary relation ‘φ ∗ (x, w)’ between objects and possible worlds, and (N) as ∃!xφ ∗ (x, w) → φ ∗ (the x φ ∗ (x, @), w) Using lambda-notation to make it explicit that this is a property of possible worlds we get (λw)(∃!xφ ∗ (x, w) → φ ∗ (the x φ ∗ (x, @), w)) This property requires of a possible world w that, if there is one and only one object that is φ in w, then the one and only object that is φ in @ is φ in w. This property is not satisfied by all possible worlds: it is false in those worlds in which there is one and only one φ, but the only object that is φ in @ is not φ in w. (E.g., we can think of a world in which there is one and only one planet causing the perturbations in Uranus’ orbits, but that is not Neptune.) If this property is applied to the actual world it yields ∃!xφ ∗ (x, @) → φ ∗ (the x φ ∗ (x, @), @)

7 Suppose

that we had [Neptune](∃!xφ(x) → φ(Neptune))

instead, i.e., the name scopes over the whole formula. We would have to allow for existential generalization on the whole formula, i.e., ∃y(∃!xφ(x) → φ(y)), would be a logical consequence. But this is not knowable a priori. To see this, consider that the antecedent ∃!xφ(x) can be true or false: if it is true (i.e., there is one unique φ), then the global sentence is also true because trivially there is an object that has the property φ under the assumption that there is a unique φ; but if the antecedent is false, any existing object makes ∃!xφ(x) → φ(y) true, but the global sentence is only true in those possible worlds in which there is at least one object, and this is something that cannot be known a priori, or so Evans seems to think. Chap. 8 discusses a perspective under which we can actually know a prioi that there is at least one object as an instance of what I call LDO-valid sentences.

100

6 Evans

which says of @ that, if there is one and only one φ in @, then the one and only one φ in @ is a φ in @. This has the same content as the application of the property (λw)(∃!x φ ∗ (x, w) → (φ ∗ (the x φ ∗ (x, w), w))) to @, which is a property that absolutely every possible world has (i.e., if there is one and only one φ in w, then the φ in w is a φ in w). We can see the intended effect leaving the definite description untouched. But if we want to follow Evans in the minimal details and rewrite any sentence (the x φ ∗ (x, w), w) as I x(φ ∗ (x, w); (x, w)) (where ‘I x’ is the binary operator corresponding to the definite article), then the first property above becomes (λw)(∃!x φ ∗ (x, w) → I x(φ ∗ (x, @); φ ∗ (x, w))) which, applied to the actual world, yields ∃!x φ ∗ (x, @) → I x(φ ∗ (x, @); φ ∗ (x, @)) This is, Evans claims, epistemically equivalent to the attribution to @ of the trivial property (λw)(∃!x φ ∗ (x, w) → I x(φ ∗ (x, w); φ ∗ (x, w))) In other words, attributing to the actual world the property corresponding to Le Verrier’s sentence, which is not a trivial property (in the sense that not all possible worlds have it), is cognitively equivalent to attributing a trivial property to it, although the properties are derived from propositions with different modal profiles. Evans does not further develop the notion of epistemic equivalence; what is clear is only that he wants to detach the epistemic relevance of a proposition from its modal profile (i.e., from its role as an intension in the Carnapian sense).

6.5 Contingency and Existence An important aspect of Evans’ general strategy is that it does not have to appeal to a claim like the following: If a statement is contingent, it must be existentially committing and, hence, cannot be a priori, since matters of existence of ordinary objects are not knowable a priori.

6.5 Contingency and Existence

101

I.e., contingency is not necessarily related to the existence of objects with such and such features. There is a trivial and a non-trivial reason for this. The trivial is that there might be contingent a priori truths that do not involve ordinary singular terms and, a fortiori, do not require the existence of anything. E.g., the following sentence • France won the 2018 FIFA World Cup if and only if France won the 2018 FIFA World Cup in @. The biconditional is a priori true because, in any world treated as actual (i.e., the referent of ‘@’), it is true that France won the 2018 FIFA World Cup if and only if France won the 2018 FIFA World Cup in that world, and is contingent because the left-hand side is true in the world considered as actual but false in other possible worlds, while the right-hand side is true in every other possible world, for in every such world it is true that France won the 2018 FIFA World Cup in @. Hence, its truth does not require the existence of any object except the world of utterance. The non-trivial reason is encapsulated in an enigmatic passage: [I]t does not follow from the fact that a sentence is contingent because it is formulated with the use of a referring expression, that its contingency is due to the contingent possession of a certain property by the object to which that expression refers. (1979b, p. 175)

We should try to unveil the passage because it seems to contain another fundamental aspect of Evans’ view on superficial contingencies. Apparently he has in mind sentences like our (N) ∃!xφ(x) → [Neptune]φ(Neptune). The contingency of (N) requires that it could be the case that ∃!xφ(x) is true, while φ(Neptune) is false. If ‘Neptune’ were not a rigid expression, the latter would be a necessary truth since it would be equivalent to φ(thexφ(x)) and, hence, (N) would be also necessary. ‘Neptune’ is a descriptive name and, because it is taken with narrow scope, we are not allowed to infer (Ne ) ∃y(∃!xφ(x) → (φ(y)))

102

6 Evans

which would be contingent in virtue of the fact that its truth demands the existence of an object with a specific property, i.e., the property corresponding to λy(∃!x φ(x) → φ(y)). (Ne ) could only be derived if the name in (N) had wide scope. The only existential claim allowed by free logic to follow from (N) is ∃!x φ(x) → ∃x φ(x) which is knowable a priori. Summing up, there are some crucial features of (N) that explain Evans’ enigmatic passage (i.e., the possibility of a sentence being contingent because it contains a referential expression, but not in virtue of the existence of the object referred with a specific property): (i) It contains a name (“referring expression”); if (N) contained a description instead of a name, then it would express a necessary (and not a contingent) truth. This is why (N) is contingent because it contains a referring expression. (ii) The name in (N) is descriptive (in Evans’ sense); if the name were not descriptive, the existence or not of a proposition to be known would be dependent on a further contingency. (iii) The name is taken as having narrow scope;8 if the name were taken as having wide scope, it would be open to existential quantification and, hence, the proposition could not be a priori, but only a posteriori. Although (i)–(iii) are necessary and sufficient conditions for (N) being contingent, Evans subtle point is that it is not contingent in virtue of the existence of an object corresponding to ‘Neptune’. (N)’s contingency is a superficial one: it is due to the modal profile of (N) and not to the existence of an object with a particular property in the world since, to our surprise, the truth of (N) does not require such existence.

6.6 Some Partial Conclusions Evans’ terminology is misleading in several ways. It naturally invites one to think of deep contingency as a kind of “real” modal status, as opposed to a “merely apparent” modal status represented by the superficial contingency. Alternatively, one might be induced to think that a deep contingency is, at the same time, also a superficial contingency. Neither association is correct. As we shall see in the

8 “I have stressed that the puzzle does not even get off the ground unless we are allowed to formulate (S) [(N)] in a free logic, and in a way which gives the name ‘Julius’ [‘Neptune’] narrow scope.” (1979b, p. 174).

6.6 Some Partial Conclusions

103

next chapter, the distinction and its motivation become a lot clearer in Davies and Humberstone’s (1980) formal elaboration. It will become evident that, as a result of rigidity and descriptive reference-fixing, we can have all sorts of combinations: there are truths that are both deeply and superficially contingent, truths that are neither deeply nor superficially contingent (being therefore deeply and superficially necessary),9 truths that are not deeply contingent (being therefore deeply necessary) but superficially contingent, and also truths that are not superficially contingent (being therefore superficially necessary) but deeply contingent. Evans’ terminology is also misleading in that what he calls a Fregean name (such as ‘Julius’) is not something that Frege himself (or a standard Fregean theory) would have recognized as a proper name. First because it is rigid and, as such, if it has a reference, its contribution to the proposition expressed is its reference and, second, if there is no reference, it contributes with an entirely different content to the proposition. (This second content is not even a descriptive content of a referential expression since, as we saw, Evans does not take definite descriptions to be of this kind; their content is that of a binary second-order operator.) So, the semantic value of these names changes radically depending on the contingent existence of an object satisfying the definite description. Frege would never recognize these expressions as proper names. The same goes for the designation “descripive names”. They are descriptive, but not in the sense that worried Kripke and Donnellan. As it seems, a descriptive name in Evans’ conception is a highly idealized kind of expressions that might, sometimes, turn out to be a rigid referential term. Consequently, it is unclear how far he can be taken as really in dialogue with Frege, Kripke or Donnellan, and how far his ideas are actually idiosyncratic. All cases covered by Evans’ general strategy involve, in one way or another, an actuality operator (‘@’) and, thereby, reference to a world treated as actual. This makes possible the conception of a property that is trivially satisfied in that world (because it uses some empirical fact that obtains in the world treated as actual) but not satisfied in some other worlds. In the same way that ‘the actual φ’ is a rigidified description, ‘actually P ’ is necessarily true (if P is true in the actual world) or necessarily false (if P is false in the actual world). But reference must be made to a world treated as actual in the shaping of such properties; otherwise we do not get the same effect. The strategy was meant by Evans to be completely general and as not involving aspects belonging to a theory of reference. But the actuality operator is an indexical (indeed, a pure indexical, according to Kaplan’s terminology), with all the properties that indexicals have, including rigidity and direct reference. So, the strategy is not, after all, completely independent from aspects of the theory of reference, and not quite as general as Evans might have thought.

9 Evans himself does not talk of deep and superficial necessity in his classic paper, but the literature has taken these as the natural dual notions of the two kinds of contingencies. The common understanding is that α is superficially necessary iff it is not superficially contingent, and deeply necessary iff it is not deeply contingent. The same misleading association is suggested for deep and superficial necessity.

104

6 Evans

A more serious concern is that Evans’ strategy relies heavily on the distinction between cognitive content and proposition. But this distinction might be controversial. For one thing, it implies that the cognitive content of a sentence is independent from its modal behavior. This is not clearly reasonable, unless one understands it in terms of Kaplan’s distinction between character and content. As we saw, in Kaplan’s framework we can explain away the strangeness of contingent a priori truths by attributing distinct roles to each of these dimensions. The character is a priori (in case all propositional contents generated by occurrences of that sentence in any context are true propositions) or a posteriori (in case some, but not all, such propositional contents are true). The content is the proposition generated by the character in that context, and it is the content that is contingent or necessary, according to its modal profile. If this is so, Evans can be understood as holding that all cases of contingent a priori truths are due to the semantic properties of indexicals, and also that they are only superficially contingent. Hence, Evans can be seen as concluding that there are only superficially contingent a priori truths.10 But what if there are indexical-free cases? It seems that they would have to be treated as deeply contingent in Evans’ sense. In Chap. 8, we will see some cases that are meant to be of this kind (i.e., are meant to show that Evans’ restriction is fundamentally incorrect). We can go further and question whether the distinction between superficial and deep contingencies makes sense in the first place. (And, if it does, we can also ask whether it can be motivated independently from the fact that it provides a way of discrediting contingent a priori truths or whether it is simply ad hoc.) Evans and some of his followers (e.g., Hawthorne, 2002) seem to think that the distinction is sufficiently clear, and that it helps showing that there can be no coherent account of cases of genuine knowledge of deeply contingent a priori truths. But there seems to be some obscurity surrounding the distinction, particularly in the way that Evans means to apply it. From the examples given by him, it seems that all cases of superficial contingency follow a certain pattern. We can know some superficially contingent truth about an object if we consider a property by reference to this same object. Hence, a has the same size as a a has the same color as a a is as old as a all state contingent properties of a that are not possessed by every object. This is not different when the actuality operator is used, since it applied in a possible world means as it is in this world. Hence, reflexivity seems to be the source of superficial contingencies in Evans’ sense, while no reflexivity is involved in deeply contingent truths. But there are different ways of stating a property of an object by reference to itself. Some of them yield trivial properties of the object, while others

10 Contrary to Evans, Kaplan does not treat characters that always yield true contingent propositions in virtue of indexicals as less interesting (or superficial) than those that do not always yield true propositions. Kaplan draws no distinction between superficial and deep contingency.

References

105

yield properties that are not trivial. Evans seems to base his general approach on properties of a of the form a has the same P as a. But now consider: Hesperus is as large as Hesperus Hesperus is as large as Phosphorus Are they both sources of the same property? Or two distinct properties, one of them holding trivially of Hesperus? To take another example from Kripke’s famous discussion,11 consider Paderewski has the same DNA sequence as Paderewski Does it state a trivial property of Paderewski or something that one might be surprised in discovering (e.g., if one thinks that there are two persons named ‘Paderewski’, as in Kripke’s example)? This brings us to the vicinity of Frege’s Puzzle, i.e., the problem of explaining informativeness when reference to the same object is made repeatedly. There is no guarantee that, by defining a property of an object by reference to this same object, we thereby get something that is trivially known of this object, even if reference is done using the same name-type. This might depend on there being coordination (in the sense developed in Fine, 2007) in language or thought between different tokens of names. Acknowledgements I thank the editors of the Croatian Journal of Philosophy for the kind permission to use some printed material in this chapter.

References Davies, M., & Humberstone, L. (1980). Two notions of necessity. Philosophical Studies: An International Journal for Philosophy in the Analytic Tradition, 38(1), 1–30. Evans, G. (1979a). Comments on “Two Notions of Necessity”. In M. García-Carpintero, & J. Macià (Eds.), Two-dimensional semantics (pp. 176–180). Oxford: Oxford University Press, 2006. Evans, G. (1979b). Reference and contingency. The Monist, 62(2), 161–189. Fine, K. (2007). Semantic relationism. Oxford: Blackwell Publishing. Hawthorne, J. (2002). Deeply contingent a priori knowledge. Philosophy and Phenomenological Research, 65(2), 247–269. Hawthorne, J., & Manley, D. (2010). The reference book. New York: Oxford University Press. Kripke, S. (1979). A puzzle about belief. In A. Margalit (Ed.), Meaning and use (pp. 239–283). Springer.

11 Kripke

(1979).

Chapter 7

Two-Dimensionalism

Starting in the late 1970s, there was the development of the so-called TwoDimensional Semantics, a theoretical framework that treats in a unified way the phenomena of contingent a priori and necessary a posteriori truths detected by Kripke and Kaplan. Some of the main proponents of this framework were Stalnaker (1978), Davies and Humberstone (1980), Chalmers (1996, 2006b, 2010), and Jackson (1998). In this chapter we will see the general features of some of these proposals, and how they handle the phenomenon that is our focus in this book. All varieties of two-dimensionalism have in common the following view: when we consider an expression (e.g., ‘water’ or ‘water is H2 O’) and ask about the extension that it determines in a possible world w (e.g., the substance corresponding to ‘water’ or the truth-value of ‘water is H2 O’ in that world), there are two ways in which the answer might depend on how w is, and these two ways are derived from the different perspectives that we might have towards w. In one way, the expression has an extension under the assumption that w is the actual world. We could, in Carnapian terms, talk of an intension of a first kind as a function that associates possible worlds taken as actual to extensions. In the other way, the expression has an extension under the assumption that w is a counterfactual situation of evaluation of what was said in the world treated as actual (which might be different from w). Again, in Carnapian terms, we could talk of an intension of a second kind as a function that associates possible worlds taken as counterfactual situations to extensions. If we consider the meaning of an expression as given by its intension, this picture suggests considering the global meaning of any expression as bifurcated between a dimension that interacts with a possible world taken as actual, and a dimension that interacts with a possible world taken as a counterfactual situation. Alternatively, we could look at the global meaning of an expression as a one-place function from possible worlds into intensions of the second kind, which are themselves one-place functions from possible worlds into extensions of the appropriate kind.

© Springer Nature Switzerland AG 2022 M. Ruffino, Contingent A Priori Truths, Synthese Library 443, https://doi.org/10.1007/978-3-030-86622-8_7

107

108

7 Two-Dimensionalism

Kaplan’s distinction between character and content, with the corresponding distinction between context of utterance and circumstance of evaluation, is perhaps the most obvious inspiration for the two-dimensional framework and, according to one interpretation, the whole two-dimensional proposal can be seen as a generalization of the Kaplanian framework (which was conceived primarily for indexicals) to other kinds of expressions.1 Since the first explicit formulation of two-dimensionalism in the pioneering works of Stalnaker and Chalmers, there have been many changes and reinterpretations of this framework, including some by Stalnaker and Chalmers themselves.2 In this chapter I shall not deal exhaustively with all these many versions and the varieties of issues that they are meant to address, but shall only look at the general proposal in the grounding works, particularly in the way they are meant to deal with contingent a priori truths.

7.1 Stalnaker’s Assertion The first and most schematic formulation of the two-dimensionalist framework can be found in Stalnaker’s classic paper “Assertion” (1978). His primary motivation was to give an account of necessary a posteriori truths or, as he would later put it, to provide an answer to the question [H]ow is it that a necessarily true statement [such as ‘Hesperus is Phosphorus’] could be used to convey contingent information? (2004, p. 296)

Another motivation was to reconcile the phenomenon of indexical propositional attitudes with the classical Fregean perspective of belief as a binary relation between a thinker and a single abstract object by changing the conception of the object of the attitude. Stalnaker’s reflections on the two-dimensional nature of the meaning comes combined with (although it can be seen as independent from) a Gricean-Lewisian dynamic picture of assertions as the exclusion of possibilities within a conversation. Let us first have a sketch of this picture. 1 This

is so despite the fact that Stalnaker (2004, pp. 298–9) sees some fundamental differences between the explanatory role of Kaplan’s character and of his notion of propositional concepts. Chalmers (1996, p. 365, footnote 25) also calls attention to some subtle differences between what he calls primary intentions and Kaplan’s character. Despite these differences in details, it is clear that Kaplan’s ideas are the roots of the two-dimensionalist proposal. Soames (2005, Chapter 4) calls Kaplan’s semantics in terms of character and content a “benign two-dimenionsionalism”, to distinguish it from the “ambitious two-dimensionalism” (embraced by Chalmers and Jackson), which is, according to him, misguided. 2 See, e.g., Chalmers (2010, 541–68) and also Schroeter (2019) for a survey of the many varieties that two-dimensionalism has taken since its origins.

7.1 Stalnaker’s Assertion

109

7.1.1 The Point of Making an Assertion Following a general model suggested by Carnap (1947), we can consider a proposition in an abstract way as a function that associates possible worlds to truthvalues, i.e., associates the value true to the worlds in which the proposition is true, and the false to those worlds in which the proposition is false. Alternatively, we might consider a proposition as a set of possible worlds (i.e., those worlds in which it is true). If we have a proposition as the set P1 and another proposition as the set P2 , then the conjunction of two propositions might be seen as P1 ∩ P2 , i.e., the set of worlds in which both are true, while the disjunction might be seen as P1 ∪ P2 , i.e., the set of worlds in which at least one of them is true, etc. A conversation can be seen as a dynamic process in which, at each stage, there are many propositions presupposed (i.e., taken to be true) both by the speaker and the audience. Let C be the set of presuppositions. If we apply the model above, we have that C can be seen as the set {Pi /Pi ∈ P } of possible worlds corresponding to the generalized intersection of these many presupposed propositions. Stalnaker calls this the context set.3 The elements (possible worlds) of the context set can be seen as possible alternatives (“live options”, as he puts it) compatible with everything that is accepted so far in the conversation. Stalnaker sees an assertion made at a certain point in a conversation as a proposal to exclude some of these alternatives (formally, as a proposal to include another element Pj in the set C and, hence, to shrink the previous intersection); if accepted by the interlocutors, the context set is reduced (i.e., fewer possible worlds are now seen as “live options”). E.g., suppose you and I are engaged in a conversation at a party, and the only propositions that we assume as background are that we are at a party and that the current day of the conversation is Friday. There is a set of possible worlds corresponding to this common assumption that includes both worlds in which my nephew is married, and worlds in which my nephew is not married (since both kinds of worlds are compatible with everything assumed so far). Now I tell you ‘my nephew just got married’; if you accept that, then we are excluding from our common assumptions all worlds in which my nephew is not married, and start to operate with a new (more restricted) set of possible worlds. So, each assertion might be taken as a proposal to shrink the set of alternative possible worlds; each step in the conversation, i.e., each assertion is, abstractly speaking, a proposal to reduce the context set.4 3 This model can be a little more complicated if, in a conversation, the speaker and the audience are working with different sets of presuppositions, i.e., different sets of possible worlds or of propositions taken as true. If there is cooperation between speaker and audience, these sets are adjusted along the conversation so as to converge or to become “close enough to being nondefective if the divergences do not affect the issues that actually arise in the course of the conversation” (1978, p. 85). I shall overlook this possible complication here and assume that speaker and audience share the same set of presuppositions. 4 Presumably, in an ideal conversation in which all assertions that are informative are made without inconsistency, we would end up with a singleton context set {w1 }, where w1 is the world that

110

7 Two-Dimensionalism

7.1.2 The Two-Dimensional Semantics The model of assertion just sketched is highly abstract and compatible with multiple theories. If we were to use Searle and Vanderveken’s terminology, we could see this model as an explication of the illocutionary point of assertions, but not necessarily of their propositional content. But Stalnaker also proposes a semantic theory for the content of assertions. His leading idea is that there are two distinct ways in which this content can be sensitive to the context. The first way is that the context might determine the proposition expressed. The second way is that the truth-value of a proposition depends on the possible world taken as argument (i.e., the world in which we evaluate it). For this reason, Stalnaker finds it useful to replace the classical notion of a proposition determined by an assertion with the notion of a propositional concept. Formally, a propositional concept is a function from possible worlds to propositions or, alternatively, a function from ordered pairs of possible worlds (the first element of the pair being the world considered as actual, and the second element being the world taken as argument of the proposition) to truthvalues. A propositional concept, hence, is basically a two-dimensional intension that yields truth-values to pairs of possible words. Let us put this in a more schematic way. As said already, propositions can be seen as functions from possible worlds to truth-values, so we can represent a particular proposition P like this w1 T

w2 F

w3 T

i.e., P is true at w1 , false at w2 and true at w3 (let us assume that these are the only relevant possible worlds). Now, we might consider the possibility that, in different possible worlds, an utterance generates different propositions; so we might introduce other lines corresponding to these other propositions generated, and the result would be something resembling a matrix, such as

w1 w2 w3

w1 T F T

w2 F T T

w3 T F T

The horizontal and the vertical axes are the same sequence, i.e., w1 , w2 and w3 , but they display the three possible worlds with two distinct roles. The horizontal axis represents the possible worlds taken as arguments of propositions as functions (or what we might call points of evaluation), while the vertical axis represents the same possible worlds taken as contexts of assertion (i.e., as determining what is said, the propositions expressed). The matrix displays the double dependence of the assertion speaker and audience take to be the actual one. If there is any inconsistency along the way, the context set becomes the empty set, since there is no “live option” in which everything that was accepted so far can be true.

7.1 Stalnaker’s Assertion

111

on possible worlds and can be seen as a representation of the propositional concept. The first line of the matrix represents the proposition produced by the assertion in the world w1 taken as context; the second line represents the proposition produced by the same assertion in the world w2 taken as context, and the third line represents the corresponding proposition for w3 as context. Let us build a slightly more concrete example. Suppose that I say ‘You are French’, and let w1 be a world exactly like the actual world (as far as nationalities are concerned) in which I am talking to David Kaplan, w2 a world exactly like the actual one but in which I am talking to François Recanati disguised as David Kaplan, and w3 a world exactly like the actual world except for the fact that David Kaplan and François Recanati have switched nationalities, i.e., the former is French while the latter is American, and I am talking to David Kaplan. (Again, let us assume that these are the only three possible worlds.) Then we would have the following matrix F representing the propositional concept: F w1 w2 w3

w1 F T F

w2 F T F

w3 T F T

The first line represents the singular proposition that David Kaplan is French (which is the one that I generate by saying ‘You are French’ in w1 ), the second line represents the singular proposition that François Recanati is French (which is the one that I generate by saying the same thing in w2 ), and the third line represents, again, the proposition that David Kaplan is French (which is the one that I generate by saying the same thing in w3 ). The same phenomenon observed for sentences, i.e., the double dependence on possible words for truth values, is present in other kinds of expressions. Although a matrix is most clearly understood when representing a propositional content, it can be generalized to any kind of intension.5 Take, for example, the expression ‘water’ meant to designate the clean and drinkable liquid filling lakes and rivers, and let w1 be the actual world in which, as we know, this liquid has the underlying structure H2 O, w2 the world in which this liquid has the underlying structure XY Z, and w3 a world in which the liquid has the underlying structure MNT . Then, given the fact that the term ‘water’ is rigid, if it has its extension fixed in one of these worlds, it carries the same extension to other possible worlds. We can represent the twodimensional intension of ‘water’ by means of the matrix H:

5 Stalnaker

(1978, footnote 7).

112

7 Two-Dimensionalism

H w1 w2 w3

w1 H2 O XY Z MNT

w2 H2 O XY Z MNT

w3 H2 O XY Z MNT

In H, the lines, which represent the extension of ‘water’ in each possible world given that one of them is taken as actual, are constant functions, and this reflects the fact that ‘water’ is a rigid designator. On the other hand, the matrix can also be useful in representing the intension corresponding to a non-rigid term such as the one expressed by ‘Obama’s favorite drink’. Let w1 be a world in which Obama’s favorite drink is beer, w2 a world in which his favorite drink is wine, and w3 a world in which his favorite drink is tea. Since ‘Obama’s favorite drink’ is non-rigid, it is irrelevant for its extension in any possible world how the extension was fixed in the world taken as actual, i.e., the extension at each world will not be committed to the extension in the actual world. Hence, the matrix representing this two-dimensional intension would be something like D: D w1 w2 w3

w1 beer beer beer

w2 wine wine wine

w3 tea tea tea

Here, the lines are identical, so the values at each possible world vary equally. Correspondingly, the columns are constant functions. This also reflects the fact that ‘Obama’s favorite drink’ is non-rigid, so in each possible world, the intension picks Obama’s favorite drink in that world independently of the extension picked in the world taken as actual.

7.1.3 The Diagonal Proposition and Kripke’s Contingent A Priori In the example above of matrix F, it is intuitively clear which proposition is expressed in each possible world, and they are represented as the lines of F. But, again abstractly speaking, any function that associates each possible world to a truthvalue is also a proposition, independently of being the one intuitively meant by the speaker or the one expressed in one of the possible situations. So, if we look at the diagonal from the upper left to the lower right of this matrix, we find the following function w1 w2 w3 F T T

7.1 Stalnaker’s Assertion

113

that is itself a proposition. Although this is not what is said in any of the three possible situations of assertions, it can be abstracted from the matrix. The diagonal proposition (as Stalnaker calls it) corresponds, intuitively speaking, to the value assumed in each possible world by the proposition that is generated in that same world taken as actual. Another way to put it is that the diagonal proposition is a function that associates to each possible world the value for that world of the proposition generated in the same world.6 Consider the propositional concept corresponding to ‘Obama’s favorite drink is alcoholic’, which could be represented by the following matrix DA (in which the worlds w1 , w2 and w3 are, as far as Obama’s drinking preferences are concerned, as described above): DA w1 w2 w3

w1 T T T

w2 T T T

w3 F F F

Here, the lines are exactly the same, and they are also identical to the diagonal proposition. This reflects the fact that, on the one hand, in all possible wolds we have the same intension evaluated, no matter which world is taken as actual, because the expression ‘Obama’s favorite drink’ is non-rigid.7 On the other hand, for this propositional content, there is no difference in treating a world as actual or as a point of evaluation. Each line of a matrix such as F or DA above determine a proposition. The diagonal proposition might be identical to some (or all) of them, as in DA, but it might not be identical to any of them, as in F. In any case, the diagonal proposition is, as we saw, an abstraction from the matrix (and it is easier to see this in the case of F, since the diagonal is not identical to any proposition intuitively expressed). But it has an important explanatory role in Stalnaker’s two-dimensional semantics, especially in dealing with Kripke’s cases of contingent a priori and necessary a posteriori truths. A contingent a priori truth in Kripke’s sense can be represented as

6 There are other propositions that can be abstracted from the matrix, e.g., what we could perhaps call the “cross-diagonal”, which is the diagonal from the upper right to the lower left, and, in the case of F , it would be the proposition

w1 T

w2 T

w3 F

This proposition, however, plays no clear explanatory role, quite differently from the diagonal proposition. 7 The expression contains the name ‘Obama’, which is rigid; but the description ‘Obama’s favorite drink’ is non-rigid.

114

7 Two-Dimensionalism

a propositional concept that has, in each line, a contingent proposition, but whose diagonal is necessary (i.e., true in every world).8 Let us take Kripke’s example of the meter, fixed by reference to the length of a standard stick S: in each possible world wi , S has a different length, so in each possible world the proposition expressed by ‘one meter is the length of stick S’ expresses a different proposition, and this proposition is true only in the world in which the proposition is produced. So, the propositional concept corresponding to ‘one meter is the length of S’ could be represented by the matrix M M w1 w2 w3

w1 T F F

w2 F T F

w3 F F T

Each line of M is a contingent proposition that is true in only one possible world. Each cell of the diagonal represents the fact that we are taking the possible world both as the actual world and as a point of evaluation. In this particular case, the diagonal is a necessary truth, and this represents the fact that, in each world, the proposition generated taking that word as actual is true in that same world taken as counterfactual. On the other hand, we can also represent a necessary a posteriori truth in a matrix in which the lines are necessary truths (or falsities), and the diagonal is a contingent proposition. Indeed, accommodating this kind of truth is the primary motivation for the whole two-dimensionalist apparatus. E.g., consider the propositional concept expressed by ‘water is H2 O’ and, as before, w1 as a world in which water has the underlying structure H2 O, w2 a world in which water has the underlying structure XY Z, and w3 a world in which water has the underlying structure MNT . We could represent this propositional concept by the matrix W W w1 w2 w3

w1 T F F

w2 T F F

w3 T F F

Given that ‘water’ is a rigid term and that, in w1 , water is H2 O, the proposition expressed in this world taken as actual is that H2 O is H2 O, which is true in all worlds, and the first line is a necessary truth. But, given that water is XY Z in w2 , the proposition expressed in this world is that XY Z is H2 O, which is false in every possible world, hence the second line is a necessary falsity. And for a similar reason, the proposition expressed in w3 is that MNT is H2 O, which is also false in every possible world. So, each line of the matrix is either a necessary truth or a necessary

8

Stalnaker (1978, pp. 83–4, 2001, p. 146).

7.1 Stalnaker’s Assertion

115

falsity. But the diagonal is a contingent proposition, and this represents the fact that the necessary truth or falsity cannot be known a priori, but only a posteriori. Something to be noticed about Stalnaker’s account of contingent a priori and necessary a posteriori truths in terms of propositional concepts and the diagonal proposition is that it is not a foundational approach in the sense that it does not explain why we have these particular combinations of modalities. It is just a way of representing the meaning of expressions that are in a double way sensitive to context. On the one hand, it is a convenient way to represent these combinations of modalities. On the other hand, it is a way to dispel a possible discomfort with Kripke’s cases, because it does not see both contingency and apriority as belonging to the same (one-dimensional) proposition; it looks at the bearers of contingency and apriority as a two-dimensional intension (the propositional concept), which is something sensitive in two distinct ways to possible worlds.

7.1.4 Rational Communication and the Necessary A Posteriori Stalnaker employs the diagonal proposition as an explanatory device for the dual phenomenon, i.e., Kripke’s cases of necessary a posteriori truths. And the way he does this is by showing that Kripke’s cases apparently violate what we could consider three principles of rational communication. The violation of these principles leads us to reinterpret these cases as meaning something special, very much in the same spirit that some violations of Grice’s conversational maxims can be taken as an indication that we need to reinterpret what is meant by the speaker. Let us first review these principles. Given the double sensitivity to possible worlds, and the model of assertion as a proposal to reduce the context set, it follows more or less naturally that there are some conditions on rational communication, similar in spirit to Grice’s well known conversational maxims. Stalnaker formulates these conditions in the form of three principles: 1. A proposition asserted is always true in some but not in all of the possible worlds in the context set. 2. Any assertive utterance should express a proposition, relative to each possible world in the context set, and that proposition should have a truth-value in each possible world in the context set. 3. The same proposition is expressed relative to each possible world in the context set. (1978, p. 88)

Although Stalnaker emphasizes the Gricean inspiration for these principles, perhaps a more helpful way of understanding them comes from the consideration of the illocutionary point of an assertion as a proposal to reduce the context set. As it will be better explained in Chap. 9, some conditions of success of an illocutionary act follow from its illocutionary point. If we see things this way, Stalnaker’s Principles above turn out to be natural consequences of the notion of assertion: Principle 1 is, by far, the most obvious, and holds because, if an assertion is a proposal to reduce the context set, there can only be such a reduction if the

116

7 Two-Dimensionalism

proposition asserted is false in some possible worlds (those that are meant to be excluded from the context set). If it is true in all of them, there is no reduction.9 But it cannot be false in all possible worlds either (i.e., it must be true in some of them), for the purpose of an assertion is to exclude some possible worlds but not all of them10 (a situation in which the conversation is inconsistent and leads to no “live option”; if this happens, it is a sign that the conversation turned out to be irrational at some point). Principle 2 holds because, again, if we understand assertions as proposals to exclude some possible worlds from the context set, no such exclusion is made if, in a possible world, no proposition is expressed. (The proposition is what excludes from the context set those worlds in which it is false.) On the other hand, if a proposition is only a partial function from possible worlds into truth-values, it does not complete the task of reducing the context set, since some possible worlds would be neither in nor out. Stalnaker fleshes out this principle in terms of the notions of semantic presupposition and pragmatic presupposition (in the sense presented in Stalnaker, 1999), i.e., if a sentence semantically presupposes a proposition, then the speaker presupposes that proposition, which means that it is part of the context set. E.g., I can only make the utterance (in a rational and informative way) that John struggled to write his Ph.D. dissertation if it is part of the context set that John wrote a Ph.D. dissertation; therefore, every possible world in the context set must be such that the proposition that John wrote a Ph.D. dissertation is true and, hence an assertive utterance that John struggled to write his Ph.D. dissertation expresses a proposition in all of them. Among the worlds in the context set, there are those in which John struggled to write his Ph.D. dissertation, and those in which he did not struggle; hence, the expressed proposition has a truth-value in all such worlds. Principle 3 also follows from the explication of assertions as having the illocutionary point of excluding some possible worlds from the context set. If there is no unique instruction to exclude some worlds, then it is not clear which reduction is meant by the speaker. This principle has its roots in the idea (that Stalnaker traces back to Wittgenstein’s Tractatus (Prop. 2.0211)) that a sentence expresses the same proposition in every possible world, so which proposition is expressed cannot depend on how each world is (otherwise we have different propositions in each world). Leaving some obvious complications aside (e.g., that words might mean different things in different possible worlds), Principle 3 seems initially plausible if we consider sentences that do not include indexicals or other context-sensitive expressions such as ‘water’. E.g., suppose that we are having a conversation that, as far as David Kaplan is concerned, only assumes as presupposition that he is a famous philosopher; so there are in the context set at this stage both worlds in which he is 9 Some qualifications seems to be required to account for assertions in logic, mathematics and metaphysics. We shall skip this complication. 10 Again, some qualification seems to be required for assertions of necessary falsities.

7.1 Stalnaker’s Assertion

117

American and worlds in which he is not American. Now I tell you ‘David Kaplan is American’, and let w1 and w2 be worlds in which David Kaplan is American, and w3 a world in which he is not American. Then we would have the following matrix corresponding to the propositional concept K: K w1 w2 w3

w1 T T T

w2 T T T

w3 F F F

I.e., the same proposition is expressed in all three worlds w1 , w2 and w3 , with a clear and unambiguous purpose: to exclude world w3 from the context set. But Principle 3 will be violated if we consider the utterance of sentences that include indexicals or terms like ‘water’. E.g., suppose that I say ‘He is David Kaplan’, and w1 is a world in which I am pointing at David Kaplan, w2 a world in which I am pointing at François Recanati disguised as David Kaplan, and w3 a world in which I am pointing at John Perry mistakenly thinking that he is David Kaplan. Then we have the following matrix DK DK w1 w2 w3

w1 T F F

w2 T F F

w3 T F F

Assuming that ‘He’ is a rigid designator of the demonstratum, the proposition expressed in w1 is that David Kaplan is David Kaplan (which is necessarily true), the proposition expressed in w2 is that François Recanati is David Kaplan (which is necessarily false) and the proposition expressed in w3 is that John Perry is David Kaplan (which is also necessarily false). We have different propositions expressed in different possible worlds; hence, the reduction of possible worlds, which was supposed to be the point of making the utterance, is ambiguous, thus defeating its illocutionary point of determining a clear exclusion of some worlds. We would have to know the world in which the utterance takes place in order to know which of the reductions is meant. As it looks like, situations like this are such that we find a systematic violation of Principle 3 because which proposition is expressed depends on the facts in each world. If we try to make the utterance conform to Principle 3 by excluding w2 and w3 (or, alternatively, w1 ) from the context set, then we would have a violation of Principle 1, because the only proposition expressed would be either a necessary truth or a necessary falsehood. So, there is a tension here, in which Principles 1 and 3 cannot be both obeyed in many cases of interest. But the fact is that, intuitively, the utterance of ‘He is David Kaplan’ seems to be informative and to display a clear intention on the speaker’s part, despite the ambiguity in the proposition expressed: the speaker has in mind the exclusion of worlds w2 and w3 from the context set.

118

7 Two-Dimensionalism

Stalnaker’s explanation is that, in the same way that sometimes a clear violation of Gricean conversational maxims can be seen as a sign that the speaker means something else by an utterance (e.g., in cases of irony), the violation of Principle 3 in some cases (like the one in our example) can be taken as a sign that the speaker does not mean to convey any of the lines, but the diagonal proposition. That is to say, the propositional concept expressed by an utterance of ‘He is David Kaplan’ is not meant by the speaker to express the propositional concept represented in matrix DK above, but the one represented by taking the diagonal of DK and projecting it in all three lines of the matrix, i.e., the matrix †DK11 †DK w1 w2 w3

w1 T T T

w2 F F F

w3 F F F

thereby conforming both with Principle 3 and Principle 1. In our case, the speaker means to exclude only worlds w2 and w3 from the context set, although this would not be the proposition expressed in any of the possible worlds as context. This explanation generalizes to many cases of apparent violation of Stalnaker’s principles, in particular to Kripke’s cases of necessary a posteriori truths. As Stalnaker would later say The proposal made in “Assertion” was that in special cases, where there was a prima facie violation of certain conversational rules, utterances should be reinterpreted to express the diagonal proposition, rather than the proposition expressed according to the standard semantic rules. [. . . ] Our problematic example [‘Hesperus is Phosphorus’], and all cases of necessary truths that would be informative (in the sense that the addressee does not already know that they are true) will be prima facie violations of this maxim, and so will require reinterpretation. Reinterpreting by taking the diagonal proposition to be the one the speaker intends to communicate brings the statement into conformity with the rule, and seems to give the intuitively correct result. (2004, p. 297)

It is worth noticing that, in Stalnaker’s framework, there is a difference between the role of the diagonal proposition in cases of contingent a priori and in cases of necessary a posteriori truths. In the latter cases, the diagonal proposition, although it is an abstraction from the matrix, is the one that we understand as actually meant 11 Dagger is a two-dimensional operator, which means that it takes propositional concepts as arguments and yields propositional concepts as values. There is a relevant contrast between this kind of operator and one-dimensional operators such as , which takes a proposition as argument, and yields another proposition as value, e.g., takes P as argument and yields P as value. † takes a propositional concept and projects its diagonal in every line, generating another propositional concept. In some cases, e.g., the propositional concept represented by the matrix L expressed by ‘The oldest living person in 2020 is a female’ (in which there is no context-sensitive expression and Principle 3 is already respected), †L is the same as L, because the diagonal is identical to every line of the matrix.

7.1 Stalnaker’s Assertion

119

by the speaker, for otherwise there would be a violation of some of the principles of rational communication. But in the case of contingent a priori truths, we do not understand the speaker as expressing it; the diagonal is only an indication that we have apriority combined with contingency (if the proposition expressed in the corresponding possible world is contingent).12 Another thing to be noticed is that, as Stalnaker stresses in his account,13 there is no proposition that is both contingent and a priori or necessary and a posteriori: it is a propositional concept that has each of these combinations of properties when analyzed from different points of view (i.e., according to the lines, or according to the diagonal). Hence, the two-dimensional representation of the phenomenon discovered by Kripke in a way distorts its essence since, in Kripke’s view, we do have one single entity (propositions) that presents these properties.14 Moreover, as we shall discuss later in this chapter, depending on the way the formal apparatus is interpreted, it leads to the disappearance of any prospect of a priori knowledge in such cases.

7.1.5 The Context-Relativity of Propositional Concepts Another relevant aspect of Stalnaker’s interpretation15 of propositional concepts (and one that undermines the possibility of representing a priori knowledge) is his idea that not only the proposition expressed by an utterance might be contextdependent (and we have the matrix representing the distinct propositions in distinct worlds), but also the propositional concept itself might be context-dependent: Given these facts, it follows that when someone makes an assertion, his words determine not only a proposition, but a propositional concept, relative the possible worlds compatible with the speaker’s presumed background information. (1987, p. 121) Propositional concepts are defined, for an utterance token, only relative to possible worlds in which the utterance event takes place, and the diagonal proposition determined by propositional concepts are local and context-dependent. (2004, p. 301)

Given that the propositional concept can be seen as a function that associates propositions to each possible world, what these passages seem to suggest is that not only we have different propositions in each possible world, but also different functions generated in different possible worlds. But how can the propositional

12 In Chalmers’ framework, as we shall see below, there is no asymmetry: both the primary and the secondary intensions are expressed by the speaker in each possible world. 13 E.g., in Stalnaker (2001, p. 146). 14 This is so, although Kripke avoids talking explicitly about propositions. 15 Later in this chapter it will become clearer how Stalnaker’s interpretation (which he characterizes as metasemantic) of the two-dimensional framework differs from other alternative interpretations.

120

7 Two-Dimensionalism

concept itself be also world-relative? The explanation is that the possible worlds that are taken as relevant for building a matrix might depend on the particular explanatory interests, e.g., on whether one is interested in considering worlds that are relevant (or “live options”) to represent the current state of a conversation with an interlocutor, or worlds that are relevant for representing one’s visual experience of a planet.16 Another way of understanding these remarks is by noticing that the description of the logical space (i.e., what counts as “live options”) might change in accordance with the language that is used to describe it. In order to have a unique description of all relevant possible worlds that enter in the matrix of an utterance we would need a “neutral” language; but Stalnaker is skeptical regarding the possibility of such a language (2004, p. 308). There might always be disagreement concerning the basis of the description, and there might be more than one way of describing possible worlds, hence more than one set of “live options” and so more than one propositional concept associated with an utterance. Finally, another way of understanding these remarks is that the way we partition the logical space of possibilities is dependent on the context, and there is no most fundamental way of doing this (or, as Stalnaker puts it, a “maximally fine partition”).17

7.1.6 Problems With Belief Attributions Stalnaker applies the diagonalization strategy to the semantics of belief attribution. Suppose you and I are in a conversation and I want to attribute to a third subject x (that is ignorant of some astronomical facts) the belief that Hesperus is Mars. If we accept the hypothesis that ‘Hesperus’ and ‘Mars’ are rigid designators, we face the same apparent problem that we had before with simple assertions, i.e., that the belief apparently attributed to x is a necessarily false proposition, which is not what I wanted to do. Intuitively, I want to attribute to x a belief in a false contingent proposition. Stalnaker’s hypothesis is that we must interpret the content of the belief attribution as the diagonal corresponding to ‘Hesperus is Mars’, i.e., the proposition that is true in a possible world in which Mars appears in exactly the

16 For a more detailed discussion of distinct theoretical interests generating distinct propositional concepts, see Schroeter (2019). 17 “If the alternative possibilities there are vary with context, then so do the propositions which are, according to the conception of content I am sketching, just ways of distinguishing between the alternative possibilities. One can make sense of questions about the identity and difference of the propositions expressed in different utterances or acts of thought only given a common context— a common set of possibilities that the propositions are understood to distinguish between. This yields a conception of proposition which is less stable than, and very different from, the traditional conception, but it is, I think, more adequate to the phenomena of speech and thought.” (1981, p. 136)

7.1 Stalnaker’s Assertion

121

same positions as Venus (and so the name ‘Hesperus’ was given to it), and false in the actual world. This is basically an application of the diagonalization strategy to the assertion of the complement of the belief attribution ‘x believes that Hesperus is Mars’. But an important detail here is that, in our conversation, since you and I know elementary astronomy, we both know that Hesperus is Venus, and a world in which Venus appears in the same position as Mars is not in our context set, i.e., there is no possible world compatible with all our assumptions in which Hesperus is Mars. How can we then have a diagonal that includes a world in which the proposition that Hesperus is Mars is true? We must calculate what Stalnaker calls the derived context, i.e., the set of possible worlds compatible not with the speaker’s beliefs (i.e., with the totality of presuppositions that you and I share in our conversation), but the set of possible worlds compatible with the totality of x’s beliefs.18 It is against the latter context set that the propositional concept relevant to my belief attribution is calculated and from which we abstract the diagonal proposition that is, in fact, the content of the belief that I mean to attribute to x.19 One difficulty with this idea is that it is not clear how far we can or must go in calculating the derived context. For suppose that x, the subject of our attribution, has an incomplete understanding of English; hence, there is a possible world compatible with x’s beliefs in which the semantic rules are different, so that even the content of an ordinary belief such as that the sky is blue will turn out to be false. E.g., it could be that, in this world, ‘the sky’ means the Moon and ‘blue’ means made predominantly of water. When we abstract the diagonal, we get a proposition that is false in that world, although it is a world in which the sky is blue. So we can no longer truthfully attribute such a simple belief to x, and the same problem would appear for far too many simple belief attributions.20 Another closely related difficulty is the following: there are several possible propositional concepts associated with a sentence; which one is the correct one in the case of belief reports? In the case of an assertion, as 18 Stalnaker

(1987, p. 126). slightly more technical aspect of Stalnaker’s account of belief attribution is that the worlds in the derived context set (i.e., the worlds that are epistemically available to the subject of the attribution) must include an utterance of the complement of the belief attribution (in our case, an utterance of ‘Hesperus is Mars’). This is related to a fundamental feature of asserting the complement of a belief attribution, i.e., that it is taken for granted that the act of assertion takes place in every world in the context set (1987, p. 121, 1981, p. 138). But this requires some subtleties. First, we need an account of the identification of counterparts of the same utterance in the distinct worlds of the derived context set. This involves some sort of identity criteria of utterances across possible worlds, but also raises some questions such as, e.g., whether the counterpart utterance must occur in a counterpart context. Second, there is the possibility that the language of the speaker might be different from the language of the subject of the attribution; so in which language must the utterance occurrences in the worlds in the derived set be? Stalnaker’s suggestion concerning this latter point is that the propositional concept relevant in the belief attribution is the one for the sentences in the speaker’s language, and not for the sentences that the subject would use in the relevant possible worlds. This brings further complications to Stalnaker’s account, which I shall skip here. 20 See Taschek (2003) for a discussion of this and related problems concerning belief attribution in Stalnaker’s framework. 19 A

122

7 Two-Dimensionalism

we saw, the propositional concept depends on the speaker’s explanatory purposes. But in the case of belief reports, the speaker’s purposes are not the only relevant factor (since what is relevant is not the context set but the derived context set) and, hence, there must be some guessing in order to know which propositional concept is operative for the subject of the report. Stalnaker suggests that there are other fruitful applications of the diagonalization strategy (e.g., to negative existential claims).21 In this chapter, our interest is limited to the application of the strategy to contingent a priori truths, thus we shall not review all of them here.

7.2 Davies and Humberstone’s Alternative Notion of Necessity In an influential paper, Davies and Humberstone (1980) introduce two new operators in the language of modal logic, with the corresponding semantic clauses, that are meant to formalize ordinary intuitions concerning how the truth value and modal status of some sentences in a possible world can depend on which world is being treated as actual. These operators are interesting because they may represent, in a formal way, Evans’ notions of deep and superficial necessity (or, alternatively, deep and superficial contingency). As we saw in Chap. 6, the way in which Evans introduces his terminology tends to obscure its motivation. Davies and Humberstone’s formal rendering of it, employing the new operators together with the corresponding semantics, makes the distinction considerably clearer and natural. Let us first see the intuitions behind the introduction of these operators. Suppose that we have a formula α and ask whether it is necessary or contingent in a possible world w; this is equivalent to asking whether ‘α‘ is true in w. This is, so to speak, a first level of evaluation. There are two possible answers, each of them yielding a superficial status of α: α is true in w, then α is superficially necessary; α is false in w, then α is superficially contingent. ‘Superficial’ here does not mean that it is less interesting or less thoughtful as, perhaps, Evans’ terminology might suggest, but only that it is the product of one first evaluative step taking only the possible world w into consideration: we are asking only how ‘α’ behaves in w taken as a world of evaluation. This gives us one dimension of the evaluation of the modal profile of α. Of course evaluating ‘α’ in w involves asking how α behaves in all other worlds accessible to w, without discriminating one of them as the actual world. But we might also ask about the status of α in w when α is evaluated in a world w∗ (that might be different from w) taken as actual. I.e., we ask, in w, how α behaves in w∗ considered as actual. We 21 See,

e.g., Stalnaker (1978, pp. 92–4).

7.2 Davies and Humberstone’s Alternative Notion of Necessity

123

might further ask what is, in w, the status of α in the world taken as actual when the latter changes; we might even ask how α behaves, in w, when we consider all other possible worlds taken as actual. This gives us a second, deep level of evaluation. In the case of some necessary truths such as ‘P ∨ ¬P ’ things should be indifferent: ‘(P ∨ ¬P )’ is true at w simpliciter, and it remains true if the world w ∗ taken as actual changes. In the case of some contingent truths such as ‘The sky is blue’, things should also be indifferent: ‘(The sky is blue)’ is false at w, and it also happens that ‘The sky is blue’ will be true in some w∗ s taken as actual and false in others. We can say that ‘P ∨ ¬P ’ is superficially necessary (when considered in w only) and also deeply necessary (i.e., remains true for all choices of w∗ as actual). We can also say that ‘The sky is blue’ is superficially contingent (when considered in w only), and also deeply contingent (i.e., it is true for some choices of w∗ and false for others). But there are some interesting cases where there is a difference between the modal status considered only locally (in w) and considered remotely (making the world w∗ taken as actual change). Consider M as abbreviating Kripke’s sentence ‘One meter is the length of stick S at t0 ’. We can have a world w in which ‘M’ is false (e.g., a world in which S is not a meter long in some of the worlds accessible to it) and, therefore, M is superficially contingent. But now we may ask, in w, about which evaluation M has in a world w∗ taken as actual; since w ∗ is the world in which the meter convention was established, it is true in w that M is true in w∗ . We can further consider a variation of w ∗ (i.e., consider what would happen if other worlds were taken as actual), and it turns out that, in any such world, since it is the world in which the convention is established (possibly we have different conventions for different worlds taken as actual) and the meter is thereby introduced, we have that in w, M is true in any world w∗ taken as actual. Hence, although at the first level of evaluation M was superficially contingent, at the second level of evaluation (the remote or “deep” one), M is necessary. So, M is both superficially contingent and deeply necessary. This is in contrast with ‘The sky is blue’, which is contingent in the first level, and remains contingent in the second level. Now consider a second case of divergence, e.g., let W abbreviate ‘Water is H2 O’. If we assume that ‘water’ has its meaning fixed in our actual world (i.e., the world in which water is H2 O), then W expresses a metaphysical truth. We will have that in a world w, ‘W’ is true since W is equivalent to ‘H2 O is H2 O’. So, at the first level of evaluation, W is necessary, being therefore superficially necessary. But, again, we might want to consider, in w, what happens with W when the world taken as actual changes; one of these worlds taken as actual might be one in which the liquid called ‘water’ is XYZ and, hence, in that world, W is false. So, in the second, deeper level of evaluation, W turns out not to be necessary; hence, although W is superficially necessary, it is deeply contingent.22 We could represent these cases in a little table:

22 This shows that a natural suggestion induced by Evans’ terminology might be misleading; one could expect that if something is deeply contingent, then it is thereby also superficially contingent.

124

7 Two-Dimensionalism

Sentence P ∨ ¬P

First (local) level of evaluation Necessary

Second (with respect to different choices of w ∗ as actual) level of evaluation Necessary

The sky is blue

Contingent

Contingent

M

Contingent

Necessary

W

Necessary

Contingent

Combination Superficially and Deeply Necessary Superficially and Deeply contingent Superficially Contingent and Deeply Necessary Superficially Necessary and Deeply Contingent

These intuitive considerations help to understand the motivation behind Davies and Humberstone’s introduction of two new operators in the standard language of modal logic and the corresponding adjustments in the semantics.23 The leading idea in the semantics for this language is to consider models with a world designated as actual, i.e., models as structures such as W =< W, w∗ , V > where W is a set of possible worlds, w∗ is the world designated as actual, and V is a function assigning truth-values to each pair (P is a propositional variable and w ∈ W ). They also introduce an equivalence relation ≈ that holds between two models W and W if and only if they differ only regarding the world designated as actual. The first new operator (originally introduced earlier in Crossley and Humberstone (1977)), is the Actuality operator ‘A’.24 Intuitively, ‘Aα’ reads α in the world But we see that for sentences like W this is not the case. Similarly, M shows that we can have a deeply necessary truth that is not superficially necessary. 23 For a further, retrospective explanation of the motivations, see Davies (2004). 24 There is another, technical reason, for introducing the Actuality operator. In the classical language of modal logic, in which we only have ‘’ and ‘♦’, we cannot express things like ‘Every student that actually failed in the class could have passed’. The two possible candidates ∀x(Fx → ♦Px) ♦∀x(Fx → Px) do not capture the intuitive reading. The first does not express that all students that actually failed could have passed altogether (in the same world), and the second envisages a situation in which whoever failed also passed with no connection with those students that actually failed. By introducing the Actuality operator we can write ♦∀x(AFx → Px) which gives the correct reading.

7.2 Davies and Humberstone’s Alternative Notion of Necessity

125

considered as actual. Formally, for any model W = and any x ∈ W , we have W | x Aα iff W | w∗ α They also introduce the “fixedly‘’ operator ‘F’. Intuitively, ‘Fα’ reads α for any world taken as actual. It is a rough approximation to usual necessity, in that a sentence of the form Fα is true in a model if it is true for any world designated as actual. Formally, W | x Fα iff for any W ≈ W, W | x α The most interesting outcome of the introduction of these operators is the combination of both in ‘FA’ (the fixedly actual operator) which gives an alternative notion of necessity that contrasts with the classical notion (represented by ‘’). Intuitively ‘FAα’ reads α is true in any possible world taken as actual. The semantic clause for formulas with this operator would be: W | x FAα iff for any W ≈ W, if y is the designated actual world in W , then W | y α.25 This combinations of operators can formalize Evans’ notion of deep necessity as opposed to classical superficial necessity. (Similarly for deep contingency, as opposed to superficial contingency.) We can have the following translation, in Davies and Humberstone’s theory, of Evans’ terminology: α is superficially necessary iff ‘α’ is true; α is superficially contingent iff ‘α’ is false; α is deeply necessary iff ‘FAα’ is true; α is deeply contingent iff ‘FAα’ is false. The main interest of Davies and Humberstone’s paper is that they provide a formal semantics for two-dimensionalism in which we might represent some examples of contingent a priori truths as superficially contingent and deeply necessary, as well as of necessary a posteriori truths as superficially necessary and deeply contingent. This is so even though Davies and Humberstone are not quite explicit about how exactly the truth of ‘FAα’ relates to the fact that α is knowable a priori. In their framework, identities between names are considered as

25 In a posthumous comment on a draft on Davies and Humberstone’s paper, Evans suggests that the idea of the paper would be clearer if ‘F A’ were taken as one single operator ‘’ with the same semantic clause since

[T]his is closer to a necessity operator right from the start. (1979a, p. 176) However, as Evans himself admits in the same note, this would perhaps make it harder to see that this notion of necessity depends on a fundamental way on treating a world as actual (as opposed to merely possible).

126

7 Two-Dimensionalism

necessary a posteriori only if at least one of the names is descriptive in Evans’ sense. Otherwise, if, e.g., both ‘Hesperus’ and ‘Phosphorus’ are rigid names without a descriptive content, then the identity ‘Hesperus is Phosphorus’ is both superficially necessary and deeply necessary (because ‘(Hesperus is Phosphorus)’ is true, but ‘FA(Hesperus is Phosphorus)’ is also true). This semantics provides formal renderings of two perspectives that we might have towards a possible world: we might regard it as just an ordinary world of evaluation of a formula (in which case the world is simply an argument of the function V of the model), or we might treat it as the designated actual world (in which case the world is one of the constitutive elements of the model). The semantics itself is neutral with respect to the two-dimensionalist interpretation, in the sense that multiple interpretations can be represented in this framework. It is also neutral in the sense that, although it provides a formal framework to represent deeply necessary sentences, it does not force the view that all contingent a priori truths are of this kind. Indeed, the framework does not seem committed to any particular view on a priori knowledge.

7.3 Chalmers’ Primary and Secondary Intensions Chalmers (1996) develops a version of the two-dimensionalist approach as an accessory to his dualist theory of the mind.26 (Here we are not concerned with his theory of mind, but only with the accessory.) His overall aim is to deal with the situation presented by Kripke that some necessary truths might be discovered only a posteriori, since this can bring some problems for his methodology of a priori analysis of the mind.27 Hence, the driving force behind Chalmers’ development of two-dimensionalism is to accommodate Kripke’s a posteriori necessity within his favored link between conceptual truths, necessity and the a priori. Later, in Chalmers (2006b), he uses the metaphor of a “golden triangle” connecting necessity, apriority and meaning (mentioning Frege, Carnap and Kant as each “side” of the triangle in the sense that they are responsible for connecting the “vertices”). The “golden triangle” was broken by Kripke’s discovery, and Chalmers sees the main task of the two-dimensionalist semantics as being that of restoring it. As Chalmers sees it, the main lesson from Kripke’s discovery of necessary a posteriori truths is that “no single intension can do all the work that a meaning needs to do” (1996, p. 56) as it was earlier supposed to be the case with the Fregean senses

26 Basically,

the same approach is presented in Jackson (1998) with minor technical and terminological differences, so that sometimes people talk about the Chalmers-Jackson two-dimensional approach. I shall here take notice that the general outline of both approaches is basically the same; however, since Chalmers’ account goes into more details than Jackson’s and is more explicit concerning the fundamental assumptions of both accounts, I shall here concentrate on the former. 27 “At various points in this book, I use a priori methods to gain insight into necessity; this is the sort of thing that Kripke’s account is often taken to challenge”. Chalmers (1996, p. 56)

7.3 Chalmers’ Primary and Secondary Intensions

127

(and thoughts). And since there is an apparent divorce between conceptual truths and necessity, we need to recognize that there are in fact two intensions associated with expressions, i.e., “two quite distinct patterns of dependence of the referent of a concept on the state of the world” (ibid., p. 57). Chalmers’ idea, inspired in Kaplan’s distinction between character and content, is that there actually are two intensions associated with any expression: the primary intension (meant as a generalization of the Kaplanian character) which, when considered in a possible world taken as actual, generates an extension (which, on its turn, can be understood as the Kaplanian content, despite some differences that will be mentioned below), and the secondary intension (which, in each possible world taken as counterfactual, also generates an extension).28 The global intension is the product of both the primary and the secondary. It can be seen as a binary function taking pairs of possible worlds as arguments and yielding truth-values (or, more generally, extensions) as values. Since there might be indexicals or other terms that have a context-sensitive extension (such as ‘water’) in the expression, the first element of the pair is more appropriately seen as a centered possible world, i.e., a possible world together with a “marked” speaker and a time.29 Hence, the global intension can be seen as a binary function F : W∗ × W → R where W ∗ is the set of centered worlds, W is the set of possible worlds, and R the set of appropriate extensions. As Chalmers sees it, the function F is fully determined a priori, provided that all a posteriori factors are included in the arguments. To use a familiar illustration, take the expression ‘water’, and let w1 be the actual world, w2 be a world in which water is XY Z, and w3 a world in which water is MNT ; let w1 *, w2 * and w3 * be the corresponding centered worlds (i.e., w1 , w2 and w3 with a marked speaker and a time). Then we could represent both intensions (and, hence, the global intension) F associated with ‘water’ with the matrix: w1 w1 ∗ H2 O w2 ∗ XY Z w3 ∗ MNT

w2 H2 O XY Z MNT

w3 H2 O XY Z MNT

The primary intension can be represented by the diagonal, while each line represents a secondary intension. Alternatively, the binary function can be fully represented by the following sequence of values: F(w1 ∗ , w1 ) = H2 O, F(w1 ∗ , w2 ) = H2 O, F(w1 ∗ , w3 ) = H2 O, F(w2 ∗ , w1 ) = XY Z, F(w2 ∗ , w2 ) = XY Z, F(w3 ∗ , w3 ) = XY Z, 28 Jackson

(1998, p. 48) calls these, respectively, the A-intension (‘A’ for actual) and the Cintension (‘C’ for counterfactual). 29 The notion of centered possible worlds was first introduced by Quine (1969) and later elaborated by Lewis (1979). It bears close resemblance to Kaplan’s contexts of utterances, at least in Chalmers’ first formulations.

128

7 Two-Dimensionalism

F(w3 ∗ , w1 ) = MNT , F(w3 ∗ , w2 ) = MNT , F(w3 ∗ , w3 ) = MNT The primary intension is, in this case, the function f such that f (w1 ) = H2 O, f (w2 ) = XY Z, f (w3 ) = MNT . Some important facts about F : (i) As already noticed, F can be fully known a priori, in the sense that the function is completely specified by the set of values (or ordered triples w1 *, w1 , H2 O, etc.) above. This assumes that all relevant empirical facts are included in the description of each possible world.30 (ii) We can obtain the primary intension f from F by taking f (w) = F (w ∗ , w) (where w∗ is the centered world with w as possible world). This means that the primary intension is the result of evaluating in each world, the intension generated in that same world taken as actual. Formally speaking, f is the analogue of Stalnaker’s diagonal proposition. (iii) The primary intension is unique, i.e., it does not depend on which possible world we are in. But the secondary intension does depend on the possible world, and each line of the matrix above represents one secondary intension. How can we get the secondary intension g? It is a one-place function that picks an extension in a world w provided the primary intension has already “acted” in the centered world taken as actual. Hence, there might be different secondary intensions: gwn∗ (w) = F(wn∗ , w) Which g we are dealing with depends on which world is taken as actual, i.e., we can only know this a posteriori once we discover which world we are in. (iv) A consequence from the above definitions is that f (@) = F (@∗ , @) = g@∗ (@)

30 As we shall later see, Stalnaker is skeptical about this assumption since it presupposes the existence of a neutral (i.e., a non-“twin-earthable”) language in which all relevant facts in all possible worlds could be described. (See also Schroeter (2004).) Moreover, it assumes that there is one fundamental partition of the logical space, another assumption about which Stalnaker is skeptical.

7.3 Chalmers’ Primary and Secondary Intensions

129

i.e., primary and secondary intensions coincide in their application to the actual world (here represented as @). (v) Another consequence of the above definitions is that we can derive the binary function F and, thereby, also the secondary intensions, from the primary intension f , provided there is a rule about how the secondary intension depends on the primary intension (e.g., whether we have rigidity or not). As we saw above, from F we can derive (a priori) f (the primary intension) and gwn (the secondary intensions). Now, from f we can also derive (a priori) g provided there is a rule of dependence of g on f , and F (which is the product of f and g).31 On balance, if we can give an a priori description of the logical space (i.e., the relevant possible worlds), the function F is also determined a priori and, hence, the same goes for the primary and the secondary intensions. And, given the same description of the logical space and the primary intension, we can also derive the secondary intensions as well as the function F a priori. In later writings (e.g., 2006b), Chalmers himself recognizes that his original notion of primary intension is ambiguous between two distinct conceptions: what he calls the contextual interpretation (in which we have contextual intensions) and what he calls the epistemic interpretation (in which we have epistemic intensions). The key notions of the first conception are contexts (or, as he would later call them, scenarios), which are conceptually closer to contexts of utterances in Stalnaker and Kaplan’s sense, and primary intension, which conceptually resemble Stalnaker’s diagonal proposition and Kaplan’s character. This conception is primarily concerned with the semantic behavior of utterances (or, in Kaplan’s version, occurrences) of sentences or subsentential tokens in contexts represented by centered possible worlds. The second conception, as the name suggests, has its contexts (scenarios) given not properly by possible worlds but by epistemic possibilities, which include any possible state of the world not excluded a priori by a rational subject. In this second conception, primary intension is much closer to Fregean senses, defined in terms of epistemic necessity in view of a complete description of a world state. As Chalmers would later admit (2006b, p. 129), his version of twodimensional semantics only has the originally intended explanatory role if taken under the epistemic interpretation, since the contextual interpretation cannot provide a satisfactory connection between primary intensions and apriority. 31 This framework also has the implicit assumption that the way the primary intension determines the secondary intension (i.e., whether we are dealing with a rigid or a non-rigid term) is known a priori as well. Jackson makes this assumption explicit:

[W]hether or not a term is a two-dimensional term is a priori in that the answer to it does not depend on the nature of the actual world. So there are two a priori parts to the conceptual analysis story: the part concerned with the A-intensions of various terms, and the part concerned with whether the A-intensions and C-intensions of various terms differ. (1998, p. 52)

130

7 Two-Dimensionalism

In Chalmers’ conceptual apparatus, we can understand contingent a priori truths as those that have a necessary primary and a contingent secondary intensions. E.g., consider ‘Water is watery stuff’: since the primary intension of ‘water’ and of ‘watery stuff’ is the same, the primary intension of the sentence is necessary (i.e., in each world taken as actual, the extension of the word ‘water’ will be the same as the extension of ‘watery stuff’). But the secondary intensions will be contingent because ‘water’ is rigid while ‘watery stuff’ is not. On the other hand, we can also understand necessary a posteriori truths as those in which the secondary intension is necessary, but the primary intension is contingent. E.g., ‘Water is H2 O’: given that, in the actual world, water is H2 O and that the terms ‘water’ and ‘H2 O’ are both rigid, we have that the statement expresses a true identity and, hence, is true in all possible worlds (necessary secondary intension). But since the primary intension associated with ‘water’ and with ‘H2 O’ are not the same, there will be worlds taken as actual in which the proposition expressed will be a false identity (e.g., a world in which water is XY Z). It is the primary intension that represents the speaker’s epistemic perspective. In the case of terms such as ‘water’ or ‘tiger’, the primary intension represents the intuitive content associated with the term. In the case of statements, such as ‘Water is H2 O’ or ‘Tigers are mammals’, the primary intension represents a kind of a priori perspective that the subject has concerning the world that he or she believes to be in: The primary proposition, more than the secondary proposition, captures how things seem from the point of view of the subject: it delivers the set of centered worlds which the subject, in having the belief, is endorsing as potential environments in which he or she might be living [. . . ] It is also fairly easy to argue that the primary proposition, rather than the secondary proposition, governs the cognitive and rational relations between thoughts. For this reason it is natural to think of the primary proposition as the cognitive content of a thought. (1996, p. 65)

Chalmers expresses a kind of optimist regarding the fact that primary intensions can be known a priori by the speaker. He even talks about “the true intension”, indicating that there might be one particular correct primary intension.32 But how is the “true” intension determined? Chalmers himself sees a space for decision and adjustments here, depending on the way the world turns out to be. He also sees a space for some vagueness (“a little looseness around the edges of a primary intension is entirely compatible with my applications of the framework”).33 So he seems to think that there is some margin for shaping a concept (primary intension), but once that is done, we can derive the primary intension (hence, the secondary as well) entirely a priori. Perhaps his perspective is a little too optimistic as an account

32 Chalmers

(1996, p. 57). p. 365, footnote 23. In this sense, Chalmers position reminds us of Kant’s remarks (1787, B 755–6) about the definition of ‘gold’ and what, ultimately, grounds statements such as ‘gold is yellow’ (which Kant takes to be analytic): whether a statement (or “judgment”, as Kant would say) is analytic or synthetic is an objective matter, and depends only on the internal structure of the concepts involved in it; but the internal structure of the concepts might be open to adjustments dictated by the pragmatics of empirical science.

33 Ibid.,

7.4 Primary/Secondary Intensions and Character/Content

131

of ordinary language and knowledge. It might be very hard for ordinary speakers to determine the correct primary intension of, e.g., ‘mineral’, ‘living creature’, and ‘physical body’. As it seems, some primary intensions can only be correctly determined by experts.

7.4 Primary/Secondary Intensions and Character/Content As already said, in his first formulation of the two-dimensional framework, Chalmers sees the notions of primary and secondary intensions as a generalization of Kaplan’s character and content, respectively. But he also sees some few relevant differences between the two pairs of notions (although he does not place a great weight on them): Some differences: (1) Kaplan’s content corresponds very closely to a secondary intension, but he presents character as a function from context to content, whereas a primary intension is a function from context to extension. Given rigidification, however, a primary intension is straightforwardly derivable from a character and vice versa. I use the former for reasons of symmetry and simplicity. (2) Kaplan uses his account to deal with indexical and demonstrative terms like “I” and “that,” but does not extend it to deal with natural-kind terms such as “water,” as he takes “water” to pick out H2 O in all contexts (the soundalike word on Twin Earth is simply a different word), and he takes the process of reference fixation here to be part of “metasemantics” rather than semantics. As before, whether it is part of metasemantics or semantics makes little difference for my purposes; all that matters is that reference fixation depends in some way on how the actual world turns out. (1996, pp. 365–6, footnote 25)34

How should we understand the first difference? Kaplan presents character as a function that yields, for each context, a content, and that is supposed to play a role similar to that of an intension (i.e., a function that associates an extension to each possible world), while Chalmers’ primary intension is a function that yields an extension in each context. Kaplan’s contents and intensions are very close, but not exactly the same thing. Indeed, for Kaplan, the representation of contents as intensions is useful only up to a certain point but, if generalized, it obscures some fundamental aspects of the metaphysics of content that he wants to embrace. He understands contents of sentences as Russellian propositions, i.e., as structured entities composed of universals and particulars (Kaplan, 1977, p. 494). So, it might

34 In

2002, p. 167, he will point out yet another important difference both with Kaplan’s and Stalnaker’s frameworks, namely, that in these the worlds on the first dimension of evaluation are not epistemic possibilities, but contexts of utterances, i.e., they must contain the token of the expression. The differences between primary intension and Kaplan’s character are also highlighted in Chalmers (2006a). In Chalmers (2010), in reply to the criticism that names and natural-kind terms should not be regarded as indexicals, he will say: However, primary intensions are not Kaplanian characters or contextual intensions. (p. 562)

132

7 Two-Dimensionalism

well be the case that two distinct contents correspond to the same intension. The content of a directly referential expression, and the content of a description that turns out to select that same object in every circumstance of evaluation correspond to the same intension, but are not the same content, since one of them is an extension, and the other a descriptive content (ibid., p. 502). The same applies for any two necessarily true singular sentences such as ‘2 is even’ and ‘3 is odd’: what they express in any context yield the same extension in any possible world (the True), being therefore the same intension, but they have different contents (one content has the number 2 and the property even as components, while the other has the number 3 and the property odd as components). For indexicals (and directly referential expressions in general), however, content and extension turn out to be the same thing, since their extension (i.e., the reference) is their only contribution to the proposition expressed in any context of utterance. Hence, as far as indexicals are concerned, we have the curious situation in which an extension (a particular object) can, at the same time, be seen as a constant function, i.e., one that associates that same extension to every possible world. This is what Kaplan calls fixed content. But for non-rigid expressions, there is a difference. The character of, e.g., ‘the author of Nicomachean Ethics’ yields, in a context placed in the actual world, a descriptive content, while Chalmers’ primary intension would have Aristotle himself as value. The output of a primary intension is always an extension (because this is what intensions are supposed to do: they select extensions), and not a descriptive content or an intension. How should we understand the second difference? Kaplan introduces the distinction between character and content for indexicals. For non-indexicals, the character coincides with the Fregean sense. He does not extend, however, the difference between character and content to proper names or natural kind terms because he does not see the mechanisms by means of which the latter get an extension as part of their semantics. Kaplan classifies as metasemantic any process or fact that is the basis for an expression to have a particular meaning, but is not part of the meaning itself. Any mechanism such as a description that ties a proper name to its reference by a causal chain is part of the metasemantics of the name, whereas only the reference is part of the semantics of the name. This is in contrast to the character of an indexical: although the character is not part of the semantic contribution of an indexical to the relevant content, it is more appropriately seen as part of the semantics of the indexical. Chalmers says that it does not matter for his purposes whether the primary intension is part of the semantics or metasemantics of a term. But it does seem to matter, since the primary intension is conceived by him as part of the meaning of a term (e.g., ‘water’ having the same primary intension as ‘watery stuff’ seems to say something about their meaning, and not about the metasemantics of each expression). In later work, Chalmers’ characterization of primary intensions is given, differently from Kaplan’s notion of character or Stalnaker’s notion of diagonal proposition, in purely epistemological terms, appealing to the following formula:

7.5 Stalnaker’s Later Retraction

133

[T]he primary intension of a sentence S is true at a scenario w iff the hypothesis that w is actual should lead us to rationally endorse S. Somewhat more carefully, we can say that the primary intension of S is true at a scenario w iff D epistemically necessitates S, where D is a canonical specification of w. (2010, pp. 550–1)

A “scenario” can be understood as a maximally specific description of a possible world state together with a center (e.g., a location and a time). The problematic notion in this characterization is the one of canonical specification. For this specification is supposed to be one of the uncentered worlds made in a “semantically neutral” vocabulary, i.e., one that is free of terms (such as names and natural kind terms) that give rise to Kripkean a posteriori necessities and a priori contingencies. (ibid.)35

As mentioned before, some are skeptic about the availability of such a semantically neutral vocabulary and the corresponding canonical specification.36 At a certain point, Chalmers mentions the possibility that we might have descriptions in which all potentially problematic (i.e., “twin-earthable”) terms are eliminated and substituted by existentially quantified variables as in Ramsey sentences, compounding a maximally specific hypothesis corresponding to a scenario.37 This could be a last technical option for an epistemic two-dimensionalist to avoid the charge that a semantically neutral vocabulary is unavailable, but it introduces a great deal of artificiality in the two-dimensional picture of primary intensions. Ramsey’s sentences are a highly idealized form of description of an epistemic possibility. But, from a practical point of view, it looks implausible that a rational subject goes over a maximally specific set of existentially generalized sentences as epistemic possibilities in order to calculate the values given by the primary intensions. The primary intension is supposed to capture the subject’s epistemic perspective but, in this model, it seems dependent on an interaction with something that looks far removed from a subject’s concrete epistemic reality.

7.5 Stalnaker’s Later Retraction This chapter started with the observation that Kaplan’s semantics in terms of character and content was perhaps the most prototypical kind of two-dimensional semantics and that, in a way, the whole two-dimensional project could be seen as a generalization of it. This is indeed the impression that one gets from Stalnaker’s very first formulation in “Assertion”, and from the early developments given by Chalmers and Jackson. However, in his later writings, Stalnaker strives to 35 For more on neutral vocabulary and “neutrally specified information”, and how they are supposed to guide one’s reason in knowing primary intensions, see Chalmers (2002, pp. 144–9). 36 E.g., Schroeter (2004), Stalnaker (2004). 37 Chalmers (2010, p. 567).

134

7 Two-Dimensionalism

distinguish his notion of propositional concept from Kaplan’s semantics.38 He correspondingly also changes his view on the prospects of representing the a priori status of contingent a priori truths within his formal apparatus. Indeed, he later advocates a kind of skepticism concerning the representation of a priori truths in general (and of contingent a priori truths in particular) by means of the twodimensional framework, although he still holds the view (e.g., in 2001 and 2004) that his notion of propositional concept provides a good strategy to handle cases of necessary a posteriori by means of diagonalization. This comes as the result of a contrast between his own interpretation of the two-dimensionalist framework and interpretations such as Chalmers’ and Jackson’s that give a broader epistemic role to the diagonal proposition. Stalnaker’s later perspective is based on Kaplan’s distinction (1977, pp. 573– 4) between semantic and metasemantic theories. The distinction provides a helpful way to organize the theoretical battlefield between Fregeans (broadly conceived) and direct reference theorists concerning singular terms. A semantic theory is, strictly speaking, concerned only with the meanings that expressions have, while a metasemantic theory is concerned with everything that might count as the basis for ascribing meanings to expressions. That an expression has such and such a meaning is a semantic fact; if we are Fregeans, it is a semantic fact about ‘Homer’ that it refers to a particular person and that it has the sense of, say, ‘the author of Odyssey’; if we are direct reference theorists, only the fact that ‘Homer’ refers to a particular person is semantic. On the other hand, metasemantic facts are those (historical, sociological, etc.) that explain why an expression has such and such particular meaning(s). In the previous example, direct reference theorists consider the fact that ‘Homer’ has its reference fixed by ‘the author of Odyssey’ (or perhaps by different descriptions at different times) as metasemantic. If there is a mechanism of reference-fixing for a singular term, this mechanism may be taken as part of the semantics of this term (i.e., as its contribution to a propositional content), or as part of its metasemantics. Kaplan initially employs this distinction to characterize two ways of understanding a causal theory of reference. On the one hand, one could see the description ‘the individual who lies at the other end of the historical chain that brought this token to me’ as giving the descriptive meaning of a proper name. On the other, one could see it as the basis on which a proper name has its only meaning, namely, the referred object (1977, p. 574). The first interpretation gives a semantic role to the description, whereas the second gives it only a metasemantic one. Since he follows the Kripkean view of proper names, Kaplan himself believes that the description is only part of the metasemantics of these expressions, and not of their meaning. But he also raises the analogous question for indexicals, i.e., whether the character of an indexical belongs to its semantics or to its metasemantics. As we know, indexicals (pure and

38 This

is so despite the fact that he sometimes (e.g., in 2007, p. 256, footnote 4) classifies Kaplan’s notion of character as the analogue of his own notion of propositional concept. The formal apparatus is the same, but the spirit and the uses are quite different, according to him.

7.5 Stalnaker’s Later Retraction

135

demonstratives alike) are directly referential, which means that their contents are only the referred objects. Therefore, it is tempting to regard their character not as part of their semantics, but as part of their metasemantics. But this is implausible because it would imply that indexicals are systematically ambiguous instead of being governed by a rule associating a reference to each context. Hence, Kaplan, concludes, the character must be part of the semantics of an indexical. This means that to be part of the semantics and to be the semantic contribution to a proposition are not the same thing. Now, a similar question can be posed regarding Stalnaker’s propositional concept: is it part of the semantics or of the metasemantics of an expression? Stalnaker characterizes his own interpretation as metasemantic. This is so because the propositional concept is not foundational in the sense of coming first in the order of explanation of the objects of thought (i.e., the propositions generated in each context). On the contrary, the propositional concept is explained in terms of those propositions. This means that we must first have the propositions expressed by a token expression in all possible contexts and, only then, extract the propositional concept (and, therefore, the diagonal) as an abstraction. On the other hand, Stalnaker characterizes interpretations such as Chalmers’ and Jackson’s as “semantic”, because they see both the propositional concept and the diagonal (primary intension, in Chalmers’ terminology, or A-intension, in Jackson’s terminology) as semantic values. In Chalmers’ and Jackson’s view, the diagonal reflects the a priori perspective of the speaker, and the lines or secondary intensions are derived from it. The diagonal is, in this case, a kind of generalization of the Kaplanian character for every expression. This is why Stalnaker refers to this perspective as the “generalized Kaplan interpretation” (keeping in mind that, for Kaplan, the character is, as we saw, part of the semantics and not of the metasemantics of indexicals). This latter perspective assumes that speakers have some sort of a priori access to the truthconditions of their utterances. The metasemantic interpretation is more harmonious with an externalist perspective (which Stalnaker endorses), i.e., it can hardly be said beforehand which propositions will be expressed in different counterfactual situations if the former will ultimately depend on facts. Those facts determine, in the first place, the secondary propositions, and these propositions determine the primary proposition (diagonal) by abstraction. The semantic interpretation, on the other hand, seems to presuppose an internalist perspective regarding the primary proposition, i.e., speakers are able, by introspection, to see what the meaning of utterances are and, hence, to calculate the secondary intension in each possible world.39 In the metasemantic approach, a greater share of the job of generating the propositional concept and, hence, the primary intension is given to facts (so one can hardly say a priori how a primary intension will turn out to be), while in the semantic approach,

39 This is the spirit of Chalmers’ original characterization of the concept associated with an expression as a binary function from centered possible worlds and possible worlds into extensions as completely determined a priori (1996, p. 61).

136

7 Two-Dimensionalism

almost the entire job is given to semantics alone. The metasemantic interpretation expects considerably less from the semantics, while the semantic interpretation expects almost everything.40 But if the secondary intensions are prior to the propositional concept (and, hence, prior to the diagonal proposition), there will be situations in which an utterance occurs and the audience understands a proposition quite different from the one meant by the speaker. One illustrative example (Stalnaker, 2001, p. 296) is the following: we have our word ‘tiger’ used to designate a natural species of animals, but suppose that there is a possible world w in which the word ‘tiger’ is used to designate sofas. So, in the actual world, ‘tigers are not pieces of furniture’ expresses a necessary truth (provided that we accept that ‘tiger’, as a natural kind term, is rigid, as Kripke, Putnam, etc., advocate). On the other hand, ‘tigers are not pieces of furniture’ expresses a falsehood in w. Suppose now that we are in a conversation with someone that does not have a complete understanding of the English language and, for all that this person knows, we can be in a world in which ‘tiger’ means some natural species or in a world in which ‘tiger’ means sofa. If we now say ‘tigers are not pieces of furniture’ we are indeed communicating something informative, i.e., the diagonal proposition that is true in the actual world and false in w (thereby, we propose to exclude w). In cases like this, the diagonal represents the intuitive understanding of an assertion like ‘tigers are not pieces of furniture’ even in situations in which there is linguistic misunderstanding. But given the fact that the utterance can have a completely different understanding in some contexts, it follows that we can hardly expect a promising account of a priori knowledge of a contingent proposition, or so Stalnaker sees it: Even paradigm cases of truths knowable a priori (for example simple mathematical truths) will have contingent diagonals in some contexts, on the metasemantic account. Consider a context in which a person is uncertain about whether the intended meaning of a certain token of “7 + 5 = 12” is the usual one, or one that uses a base 8 notation, with the same numerals for one through seven. In some possible worlds compatible with the beliefs of this person, the token expresses the falsehood that seven plus five is ten, and so the diagonal will be contingent. More generally, any utterance, no matter how trivial the proposition that it in fact is used to express, might have been used to say something false, and a person might have misunderstood it to say something false. So the metasemantic interpretation yields no account of representation of a priori truth or knowledge, and does not depend on any notion of the a priori. (2004, pp. 302–3)

There is also a second (related) reason for Stalnaker’s preferring the metasemantic interpretation: he wants the matrix to be able to represent imperfect contexts of communication in which a speaker’s utterance is understood with a different meaning by the audience. E.g., a situation in which the metasemantic and the semantic primary intension might diverge. (This is perhaps a more realistic way of representing communication.) And this is so even if we are considering indexical expressions with a Kaplanian character, e.g., ‘I’. For ‘I’ has a certain character (i.e.,

40 As already said, in later work (e.g., in 2002, 2006a and 2006b) Chalmers’ greatest expectation is

not related to the semantics properly speaking, but to the epistemology of the primary intensions.

7.6 Some Partial Conclusions

137

it selects, in each context of utterance, the speaker), but its having this particular character and not another one rests ultimately on empirical facts. Suppose there is a Twin Earth scenario where the language is exactly like our English except for the fact that ‘I’ and ‘you’ have switched characters, i.e., ‘you’ refers to the speaker, and ‘I’ refers to the addressee.41 In this case, the semantic value of ‘You are charming’ will be one according to the semantic interpretation (that works with the original character of ‘you’), but a different one according to the metasemantic interpretation.42 We could perhaps summarize Stalnaker’s later misgivings as the recognition that every utterance has a contingent diagonal proposition (contrary to what he says in “Assertion”). That is to say, in his later writings, Stalnaker no longer sees the two-dimensional framework as a way of representing contingent a priori truths, although it still can be used to represent necessary a posteriori truths.43 However, as Chalmers points out (2006b, p.114), by claiming that every diagonal proposition is contingent (i.e., that there is some cell for which it is false), Stalnaker would also lose a way of distinguishing necessary a posteriori truths such as that expressed by ‘water is H2 O’ from ordinary necessary truths such as that 2 + 2 = 4. According to the standard two-dimensionalist strategy (endorsed by Stalnaker in “Assertion”), we distinguish these two kinds of necessary truth by saying that the latter, but not the former, would have a necessary diagonal proposition. Under the metasemantic interpretation, however, both have a contingent diagonal proposition. So, there is no clear way of distinguishing the two sorts of necessity anymore.

7.6 Some Partial Conclusions We conclude this chapter with the question: what are the prospects for the twodimensional framework to deal with the phenomenon of contingent a priori truths? It depends on how we interpret the framework. The label ‘two-dimensional semantics’ covers a great variety of philosophical theories, with different outcomes and prospects for dealing with the phenomenon. Some of them treat these truths as “puzzles” that need to be explained away (e.g., Evans and Davies and Humberstone). Others treat them as normal epistemic or semantic phenomena, to be accommodated

41 This

example is originally due to Almog (1984). e.g., Stalnaker (2001, pp. 151–2). It is also doubtful that, in this perspective, we could represent sentences like ‘I am here now’ as expressing something a priori. 43 He would later make the following comment about his early view: 42 See,

I did say in “Assertion” that a certain two-dimensional modal operator, which says that the diagonal proposition is necessary, could be understood as the a priori truth operator [. . . ]. I now think that this was an ill-considered remark. The notion of a priori truth that this identification yields is at best a very local and context-dependent one. (2004, Footnote 12)

138

7 Two-Dimensionalism

in a more general theory of meaning (e.g., the early Stalnaker) or cognitive content (e.g., Chalmers and Jackson).44 As we saw, although Chalmers and Jackson’s interpretation is primarily designed to deal with necessary a posteriori truths, we can have, in their view, full access to the primary intension, and the latter might be such that it is true in all epistemically accessible wolds, despite the fact that the secondary intensions might be contingent propositions. Although Chalmers’ account undergoes some changes in his multiple works, we can say that, broadly speaking, his overall proposal depends on some highly idealized factors such as the availability of a canonical description of epistemic possibilities, as well as of an ideal and exhaustive listing of epistemic possibilities based on the subject’s rational access to the primary proposition (which must, accordingly, be fully scrutable). Moreover, the ideal explanation and division of the space of epistemic possibilities, which are supposed to reveal the propositions that are a priori as those true in any world taken as actual, already seem to depend on the rational agent’s full a priori access to the epistemic content of the primary intension. On the other hand, in Stalnaker’s (later) interpretation, there is no room for representing contingent a priori truths (and perhaps not even apriority at all) as the diagonal of a matrix because the latter might have the lines largely depending on empirical facts and on different interpretations of language, so that no expression will have a necessary diagonal proposition. Concluding, the ability of two-dimensional semantics to represent the phenomenon of contingent a priori truths depends on the way we interpret its framework, and on the kind of assumptions that are made. At any rate, this approach is not foundational in the sense that it is only a means of representation, but provides no explanation of why there are cases of sentences with necessary primary intensions and contingent secondary intensions (or vice-versa).

References Almog, J. (1984). Semantical anthropology 1. Midwest Studies in Philosophy, 9(1), 478–489. Carnap, R. (1947). Meaning and necessity: A study in semantics and modal logic. Chicago: University of Chicago Press. Chalmers, D. (1996). The conscious mind: In search of a fundamental theory. Oxford: Oxford University Press. Chalmers, D. (2002). On sense and intension. Philosophical Perspectives, 16, 135–182. Chalmers, D. (2006a). Scott Soames’ two-dimensionalism. In Author-meets-critics session on Scott Soames’ reference and description: The case against two-dimensionalism. Chicago: The Central Meeting of the American Philosophical Association.

44 A more complete survey of the many interpretations of the two-dimensionalist framework of intensions and worlds can be found in Chalmers (2006b).

References

139

Chalmers, D. (2006b). The foundations of two-dimensional semantics. In Reprinted in M. GarcíaCarpintero, & J. Macià (Eds.), Two-dimensional semantics (pp. 55–140). Oxford: Oxford University Press. Chalmers, D. (2010). The character of consciousness. Oxford: Oxford University Press. Crossley, J., & Humberstone, L. (1977). The logic of “actually”. Reports on Mathematical Logic, 8(1), 1–29. Davies, M. (2004). Reference, contingency, and the two-dimensional framework. Philosophical Studies, 118, 1–2. Reprinted in M. García-Carpintero, & J. Macià (Eds.), Two-dimensional semantics. Oxford: Oxford University Press, 2006, pp. 41–175, pp. 83–131. Davies, M., & Humberstone, L. (1980). Two notions of necessity. Philosophical Studies: An International Journal for Philosophy in the Analytic Tradition, 38(1), 1–30. Evans, G. (1979a). Comments on “Two Notions of Necessity”. In M. García-Carpintero, & J. Macià (Eds.), Two-dimensional semantics (pp. 176–180). Oxford: Oxford University Press, 2006. Jackson, F. (1998). From metaphysics to ethics: A defence of conceptual analysis. Oxford: Oxford University Press. Kant, I. (1787). Kritik der reinen Vernunft, Edited and Translated by Guyer, P, and Wood, A. Critique of pure reason. Cambridge: Cambridge University Press, 1998. Riga: Verlag Johann Friedrich Hartknoch. Kaplan, D. (1977). Demonstratives. An essay on the semantics, logic, metaphysics and epistemology of demonstratives and other indexicals. In J. Almog, J. Perry, & H. Wettstein (Eds.), Themes from Kaplan (pp. 481–563). Oxford: Oxford University Press, 1989. Lewis, D. (1979). Attitudes De Dicto and De Se. The Philosophical Review, 88(4), 513–543. Quine, W. (1969). Propositional objects. In Ontological relativity and other essays (pp. 147–155). Columbia University Press. Schroeter, L. (2004). The rationalist foundations of Chalmers’s 2-d semantics. Philosophical Studies, 118(1–2), 227–255. Schroeter, L. (2019). Two-dimensional semantics. In E. N. Zalta (Ed.), The Stanford encyclopedia of philosophy. Winter 2019, Metaphysics Research Lab, Stanford University. Soames, S. (2005). Reference and description: The case against two-dimensionalism. Princeton University Press. Stalnaker, R. (1978). Assertion. In R. Stalnaker (Ed.), Context and content (pp. 78–95). New York: Oxford University Press, 1999. Stalnaker, R. (1981). Indexical belief. Synthese 129–151. Reprinted in Context and content: Essays on intentionality in speech and thought. Oxford University Press, 1999. Stalnaker, R. (1987). Semantics for belief. Philosophical Topics, 15(1), 177–190. Reprinted in Context and content: Essays on intentionality in speech and thought. Oxford University Press, 1999. Stalnaker, R. (1999). Pragmatic presupposition. In R. Stalnaker (Ed.), Context and content (pp. 47–62). New York: Oxford University Press, 1999. Stalnaker, R. (2001). On considering a possible world as actual. In Aristotelian Society supplementary volume, 1(75), 141–156. Stalnaker, R. (2004). Assertion revisited: On the interpretation of two-dimensional modal semantics. Reprinted in M. García-Carpintero, & J. Macià (Eds.), Two-dimensional semantics (pp. 293–309). Oxford: Oxford University Press, 2006. Stalnaker, R. (2007). Critical notice of Scott Soames’ case against two-dimensionalism. The Philosophical Review, 116(2), 251–266. Taschek, W. (2003). Book review of Context and Content. The Journal of Philosophy, 100(2), 98–108.

Chapter 8

Some Other Cases

8.1 LDO Sentences Evans’ distinction between deeply and superficially contingent truths, and the way he applies it to Kripke’s examples of contingent a priori truths, leaves a question in the air. As we discussed in Chap. 6, Evans believes that all alleged cases of contingent a priori truths derive from the presence, explicit or implicit, of indexicals, being therefore only superficially contingent. Although not explicitly, he seems to suggest that if there were indexical-free cases of such truths, they would be deeply contingent in his sense. But he is skeptic concerning the existence of such cases. (If there was a deeply contingent a priori truth, it would present, he thinks, an “intolerable” paradox (1979b, p. 161).) Still, we might ask whether Evans is justified in his skepticism. (We might also try to explain away the air of paradox, but this will be the task of Chap. 10.) Perhaps there are indexical-free cases of such truths. Kaplan’s Logic of Demonstratives (LD) and its semantics were primarily designed to model some interesting properties of sentences containing demonstratives and pure indexicals, particularly regarding the notion of validity. It is well known that, in LD, we have things like | (I am here now) but ¬ | (I am here now) given Kaplan’s definition of validity as truth in every context c of every LD-structure U (1977, p. 547). But it is less frequently noticed that the semantics of LD also has, as side-effect, an impact on the notion of validity for some indexical-free sentences. Indeed, sentences such as

© Springer Nature Switzerland AG 2022 M. Ruffino, Contingent A Priori Truths, Synthese Library 443, https://doi.org/10.1007/978-3-030-86622-8_8

141

142

8 Some Other Cases

∃x∃p Located (x, p) or ∃x Exists x 1 must be valid because, in any proper context, there will be an agent (speaker) in some location; but these are not necessary truths. In the same Section XIX, Remark 9, Kaplan says that The logic of demonstratives determines a sublogic of those formulas of LD which contains no demonstratives. These formulas (and their equivalents which contain inessential occurrences of demonstratives) are exactly the formulas with a Stable Character. The logic of demonstratives brings a new perspective even to formulas such as these. The sublogic of LD which concerns only formulas of Stable Character is not identical with traditional logic. Even for such formulas, the familiar Principle of Necessitation (if | φ then | φ) fails [. . . ] From the perspective of LD, validity is truth in every possible context. For traditional logic, validity is truth in every possible circumstance. (1977, pp. 548–9)

I will call LDO-valid those sentences that are indexical-free, but that are valid only in LD. Something interesting is that, although they are indexical-free (or, as Kaplan says, have a Stable Character), they are valid only because the semantics for LD (with the formal notion of LD-structures U) was designed to capture the peculiarities of the semantics of indexicals. So, in this sense, the fact that we have LDO-valid sentences can be seen as a side effect of such a semantics. As Kaplan would later say, Any feature of a possible world which flows from the fact that it contains the context of use may yield validity without necessity. (1989, p. 596)

Hence, since proper contexts must include a speaker, a location, etc., it is only natural to expect some valid existential truths. Hence, one easy reply to Evans’ skepticism would be to present any of the LDOvalid sentences as deeply contingent a priori, since they are valid and indexical-free. But, on the other hand, as we saw in Chap. 2, something central to Kaplan’s semantics is the restriction of contexts to proper contexts. This restriction is embodied in items 10 and 11 of the definition of LD Structures in Section XVIII of “Demonstratives” (p. 544), which say that, if c is a context with the agent ca , location cl , time ct and possible world cw then The pair ca ,cl  belongs to the extension of the two-place predicate ‘Located’ in the circumstance of evaluation (cw , ct ); ca belongs to the set of existing individuals in the circumstance of evaluation (cw , ct ).

1 In LD, the one-place predicate ‘Exists’ is a primitive symbol interpreted in each circumstance of evaluation, and it is not the same as the quantifier ‘∃’.

8.2 Williamson and The Believer

143

In more intuitive terms, the restrictions requires that any admissible context is such that, in its possible world, the agent exists and is at the location and time of the context. One might be reluctant to concede that LDO-valid sentences are deeply contingent in Evans’ sense, although they do not contain indexicals, because they are the product of a semantics designed to capture semantic features of indexicals.2 We could, perhaps in the same spirit as Evans’ original distinction, introduce a refinement recognizing a third category: besides the superficially contingent a priori truths, which explicitly or implicitly contain an indexical, and the deeply contingent a priori truths, which do not contain indexicals, we could also have superficially deep contingent a priori truths, which, like the LDO-valid sentences above, are not valid because of the presence of indexicals (either explicit or implicit), but are valid as a side-effect of the semantics for LD.

8.2 Williamson and The Believer Williamson (1986) also challenges Evans’ skepticism, and presents a purported instance of indexical-free contingent a priori truth. If there is such an indexical-free truth, he concludes, there is at least one deeply contingent a priori truth in Evans’ sense. At first sight this seems to be an easy task, since sentences like Julius invented the zip (if anyone did) (with Julius having its reference fixed by ‘the one who invented the zip’ and the verb taken with a tenseless reading) contain the proper name ‘Julius’ but appear to contain no indexicals. But this easy alternative would not be indexical-free because ‘Julius’ allegedly has its reference fixed with the help of a definite description taken in the actual world (i.e., something like ‘the one who invented the zip in @’) and, hence, somehow the actuality operator (which is an indexical) is behind the name. Williamson’s example results from his consideration of a method for forming beliefs that are guaranteed to be true and a priori: (M) If there is a valid deduction from the premise that someone believes that P to the conclusion that P , believe P .3 He makes three related but distinct claims about (M):

2 There is no possibility of hidden indexicals in LD, since what counts as a formula is strictly controlled by the formation rules; if there is an indexical in a formula of LD, it can only be explicit. 3 In a later paper, Williamson (1988, p. 219) says that (M) should be understood as an “abstract mechanism for generating beliefs, not as inference rule”. It is not clear what he means by “mechanism” as opposed to an inference rule.

144

8 Some Other Cases

(i) Since the method requires the existence of a valid deduction of P from the mere belief that P , then necessarily any belief formed by this method is true. (ii) (M) is absolutely reliable since it is impossible to have a false belief obtained by means of it. (iii) We can know a priori that (M) is absolutely reliable. (i) guarantees that any belief formed by means of (M) is true. In order to guarantee that this true belief counts as knowledge, we need a warrant that it is also justified. (ii) guarantees that such beliefs are justified in an absolute way.4 So, any belief formed by means of (M) counts as knowledge. Now, in order to confer the status of a priori knowledge, we must assure that the justification provided by (M) is a priori. This is what (iii) grants, based on the fact that (M) appeals only to a valid deduction.5 Conclusion: any belief obtained by applying (M) must count as a priori knowledge. The most interesting thing about (M) is that we can show that (1) There is at least one believer is knowable a priori, since there is a valid deduction of it from the premise (2) Someone believes that there is at least one believer. We could perhaps so represent this valid deduction in first-order logic with existential generalization, where ‘Bxy’ is the binary predicate ‘x believes that y’: ∃xBx(∃x∃yBxy) ∴ ∃x∃yBxy 6 There are two crucial features of (1). First, it is a contingent truth in the sense that there might be possible worlds with no believers at all. Second, it can be derived independently of anything containing indexicals such as (3) I am a believer

4 This is not completely free of controversy. Hawthorne (2002) raises some doubts about knowledge obtained through absolutely reliable methods. 5 The classic conception of knowledge as justified true beliefs seems to be behind Williamson’s reasoning here. He famously champions a different conception in later work (2000), known as the knowledge first conception. But in the present text, he seems to be operating with the classical view. 6 We can here skip some minor complications such as the fact that the first order variables have both believers and propositions in their domain, and that the existential generalization would, strictly speaking, require the previous transformation of the existential sentence in the corresponding proposition by means of a that-operator.

8.2 Williamson and The Believer

145

So, (1) is not only indexical-free, but also can be derived from indexical-free premises by using method (M), and is a contingent a priori truth, being therefore deeply contingent in Evans’ sense.7 In order to deal with the objection that (1) might actually be a necessary truth (in case one thinks of God as a believer that necessarily exists), Williamson proposes the following modified version of it: (1 ) There is at least one fallible believer. Consider the premise that someone believes that there is one fallible believer. The one who believes this is either fallible or infallible. If the believer is infallible (such as God), then the content believed must be true, and it follows that there is at least one fallible believer. If the believer is fallible, then it follows by existential generalization, that there is at least one fallible believer. Although Williamson’s The Believer example seems like a convincing case of a contingent a priori truth, there has been some skepticism as to the claim that it involves no indexicality. One might, e.g., wonder whether the imperative form of (M) (“believe P ”) requires an indexical so that the reasoning following (M)’s prescription would involve, in one way or another, something like I must believe that there is at least one believer However, this seems implausible since, if there is a hidden indexical here because of the imperative, there must be one in every reasoning made in accordance with any rule of inference (such as Modus Ponens), for any such rule can be understood as an imperative.8 Oppy (1987) objects that there is a hidden indexical element in the conclusion that there is at least one believer. If any believer is drawing the conclusion that there is at least one believer, then the conclusion must not be about some empty world in which there are no believers (in which case it would be false). Hence, it is more appropriately seen as an inference to the claim that there is at least one believer in the world that is actual relative to the inference. If this is so, then Williamson’s

7 Presumably,

there are other related truths derivable by means of (M) that are contingent. E.g., Something is an object of belief

or, in second-order logic, There is a non-empty epistemic relation can be known a priori and are contingent. 8 Any rule of inference can be understood as a directive speech act, to use the terminology of Searle and Vanderveken (1985), with some order (or invitation, permission, etc.) explicitly or implicitly directed at the interlocutor.

146

8 Some Other Cases

claim that the believer sentence represents an indexical-free contingent a priori truth is wrong, for it must include the indexical ‘actual’. Oppy understands Williamson as employing the following rule of inference: (APK) BxP → P ∴P And he sees this rule as analogous to (MP) Q → P Q ∴P except that, differently from (MP), (APK) is truth-preserving only when used. Now, one problem with this representation is that, since the arrow in (APK) seems to stand for material implication (because Oppy uses the same symbol to represent (MP)), it does not seem to capture exactly the way Williamson conceives of his rule (M), for in the latter there must be a valid deduction from ‘Someone believes that P ’ to P (and not just a true conditional with the former as antecedent and the latter as consequent). Moreover, (APK) leads to false beliefs. E.g., if P is false and no one believes it, BxP → P is true and (APK) would recommend the belief in P . In his reply to Oppy, Williamson (1988) first points out this inadequacy. He suggests that, instead of (APK), Oppy’s argument should have used the following principle of beliefformation, which is closer to M: (MAPK)

Given a belief that if someone believes that p then p, then believe p

There is something puzzling about Williamson’s suggestions: as it stands, (MAPK) does not seem to be absolutely reliable, for sometimes it might lead to the formation of false beliefs. E.g., there might be a (false) belief that if someone believes that unicorns exist then unicorns exist, but that unicorns exist remains a falsehood. Anyway, the core of Williamson’s reply is not focused on the correctness of (MAPK) but on whether (1) is indexical-free. Oppy claims that it must contain an actuality operator (hence, an indexical), no matter the exact form of the principle that is followed. Williamson has two replies, both of them unconvincing, in my view. The first reply is that, even if the immediate conclusion that one can reach a priori by means of (M) includes something like the actuality operator, i.e., There is at least one believer (in @) one also has a priori knowledge of the following conditional:

8.2 Williamson and The Believer

147

If there is at least one believer (in @) then there is at least one believer.9 Now, since by applying (M) we already have a priori knowledge of the antecedent, we can, by Modus Ponens, reach a priori knowledge of the consequent, i.e., that there is at least one believer. The route leading to it involved indexicality, but the conclusion does not. Hence, it is a deeply contingent truth, according to Williamson. This reply seems to deviate from the standards settled in Williamson’s first paper on the issue. For it detaches the content of the a priori belief from the route that has to be taken in order to reach it. In the original paper, Williamson seems to consider that an inference to (1) by means of existential generalization over (2) (which contains an indexical) would disqualify the former as deeply contingent. That is why he formulates (M) in the first place (which supposedly leads to (1) without employing indexicals). So, this reply indicates a certain shift in his standards. The second reply is: (APK) should be replaced by (MAPK), which has no need of indexical qualifications such as ‘(in this world)’; to believe that P is to believe it about one’s own world (which is not to say that all beliefs are indexical). Similarly, (M) stands in no need of indexical qualifications. (1988, p. 220)

This passage seems awkward. Williamson concedes that a belief in P obtained through the two methods is about ones’s own world, but sees no need of indexical qualification in either of them. It is unclear how this could be possible. As I see it, the dispute between Williamson and Oppy in terms of methods of a priori belief formation and their material adequacy tends to obscure the main point of Williamson’s The Believer example and how far it proves his point. The comparison with Kaplan’s LDO-valid sentences can clarify the issue and show that both Williamson and Oppy have an important point. To remember, LDO-valid sentences are indexical-free, but their validity comes from considerations of proper contexts, which are thought of as capturing something essential about the semantics and epistemology of indexicals (the conditions of possibility of occurrences of indexical sentences). Despite the inadequacy of his representation of Williamson’s method, Oppy does detect a crucial aspect of Williamson’s suggestion, and its dependence, somehow, on indexicality, for the belief in (1) must somehow be tied to the context in which the belief takes place, otherwise there is no guarantee of its truth. Oppy thinks that there must be a disguised indexical ‘actual’ in the content of the belief that there is at least one believer tying that belief to the world of the believer, for otherwise it could be a false belief. But the point can be better put in the following terms. The strategy behind the example can be seen as a variation of Kaplan’s LDO-valid sentences with the difference that, while Kaplan explores the notion of proper contexts of utterances, Williamson explores an

9 Presumably

this comes from the fact that the biconditional

There is at least one believer (in @) if and only if there is at least one believer is knowable a priori, i.e., is true in any circumstance of use.

148

8 Some Other Cases

analogous notion of proper contexts of beliefs. Like Kaplan’s LDO-valid sentences, the contingent a priori truths produced by Williamson’s (M) do not need to contain hidden indexicals, but they are valid in virtue of a crucial aspect of proper contexts of beliefs (e.g., that there must be a subject of the belief, and an empty world contains no such adequate contexts), in the same way that Kaplan’s LDO-valid sentences are so in virtue of the fact that only contexts with some features are proper for utterances. Hence, Williamson’s example is a case of contingent a priori truth not in virtue of the presence of a hidden indexical (as Oppy claims), but in virtue of the semantics designed to capture adequately the phenomenon of indexicality (as far as it is relevant for the occurrence of beliefs as well). I conclude that we could regard Williamson’s example as belonging to what I call superficially deep contingent truths.

8.3 Best Explanations Laurence BonJour (1998, 2005, 2009) famously makes a defense of a priori knowledge as one that is obtained solely in virtue of understanding a proposition by means of a rational insight (or rational intuition), by which he means that [I]ts very content provides, for one who grasps it properly, an immediately accessible reason for thinking that it is true. (1998, p. 102)

Although the most paradigmatic cases are knowledge of necessary truths such as 2+2=4 and P ∨ ¬P , BonJour seems to admit that the kind of justification provided by rational insight extends to some non-necessary propositions.10 In particular, it can provide an a priori justification for a general principle of induction.11 Section 7.7 of BonJour’s (1998) is an outline of how to provide such a justification. The reasoning is more or less the following: suppose there is an observed regular association between occurrences of events of kind A and events of kind B (e.g., 95% of occurrences of As are accompanied by occurrences of Bs). This is the inductive premise and coming to know it obviously requires some experience. Now if we think of explanations for the inductive premise, there are only two candidates: either the regularity is due just to chance or an extremely unlikely coincidence, or there must be some objective regularity due to the nature of A and B and the way they relate to each other. The former is a bad explanation (basically a non-explanation). Hence,

10 Despite

sometimes suggesting that anything known a priori is necessary, such as in

[W]hen I carefully and reflectively consider the proposition (or inference) in question, I am able simply to see or grasp or apprehend that the proposition is necessary, that it must be true in any possible world or situation. (1998, p. 106) 11 In this, BonJour is similar to Russell (1912, p. 68), for whom a general principle of induction can be directly intuited, being therefore known a priori.

8.3 Best Explanations

149

[A]n objective regularity of a sort that would make the conclusion of a standard inductive argument true provides the best explanation for the truth of the premise of such an argument. (1998, p. 207)

and, because of this, there would be a justification of a general inductive principle, although such principle is not a necessary truth. This reasoning is entirely a priori. We do need experience to know that there is an observed regularity, but not to know that, given such a regularity, the existence of an objective connection is its best explanation. But the reasoning does not exclude the possibility of there being worlds in which the inductive premise is true by pure chance. And this is the reason why the inductive principle is not a necessary truth. BonJour’s approach depends on some fundamental claims. First, if the inductive premise is true, and supposing that there is a question of finding an explanation for it,12 then most likely there is such an explanation (other than chance or sheer coincidence), although there might be possible worlds in which the inductive premise is true just by chance. Second, the best explanation is most likely the standard inductive one (compared to other possible, non-inductive explanations). Actually, BonJour speaks a couple of times of induction as being the best explanation (besides pure chance), but in a little exercise (carried out in (1998, pp. 209–12) and repeated in (2009, pp. 67–8)), he seems to conclude that it is the only explanation, since there is no other real alternative to standard induction. This is so basically because any such alternative explanations would have the inductive premise explained as true, but would not allow the inference to the standard inductive conclusion due to the presence of a third property or factor C behind the correlation between A and B. However, the alternative explanation appeals, in one way or another, to a correlation between A and C, so we are back to the dilemma between pure chance and standard induction again. Third, the claim that, if something is just likely or unlikely (as opposed to being definitely true or false), it can be justified a priori. One might perhaps resist all three claims. But BonJour sees no principled reason for that; in particular, he sees no principled reason to restrict a priori justification to propositions that are definitely true or false (1998, p. 208). There is a related example of deeply contingent a priori truth proposed by Hawthorne (2002). He imagines a situation in which an incorporeal being (he calls it “The Explainer”), before having any perceptual experience whatsoever, considers different possible experiential life histories, and possible explanations for each of them. So, The Explainer seems to know a priori that T is the best explanation for a particular possible experiential life history L, although this is not necessarily true (even if L turns out to be the case, the best explanation could be false). Although Hawthorne’s example is less credible than BonJour’s, it has the same spirit, exploring the possibility of reasoning that concerns best explanations and that is entirely a priori.

12 He

describes this as “the overwhelmingly obvious” question (1998, p. 207).

150

8 Some Other Cases

8.4 The “Most Unlikely Event” Turri (2011) endorses BonJour’s characterization of a priori justification, and stresses that there is nothing in it implying that a priori propositions must be necessary. There might be propositions known by “rational insight” that are contingent in a broad sense. Strictly related to this remark is a second one, that the notion of self-evidence is not restricted to necessary truths either.13 But Turri thinks that BonJour’s and Hawthorne’s cases of a priori justification of induction in terms of best explanation are flawed basically for the same reason: If we suppose, along with BonJour and most other epistemologists, that a good epistemic reason for believing Q must at least make Q more likely than not, then we shouldn’t conclude that the mere fact that a hypothesis best explains the data constitutes a good epistemic reason for believing the hypothesis. Accordingly we shouldn’t accept that it could, by itself, provide a priori justification or knowledge. (2011, p. 337)

I.e., Turri believes that justification in terms of explanatory power provides a weak connection, because even if T is the best explanation for L, and L is verified, still there normally are plenty of possible worlds in which L is true but T is false: Knowledge clearly requires a stronger truth-connection than this, or so you might reasonably maintain. (2011, p. 336)

It is hard to see a reason for denying that, if T is the best explanation for L, we could conclude that T is most likely in the presence of L. In my perspective, this does make T most likely if anything does (in the absence of a logical implication between L and T ), although, of course, it is another matter whether it provides a priori justification for T . One thing that might be said in favor of Turri’s point is that the notion of best explanation might be, in several ways, of greater complexity than it is presupposed in BonJour’s and Hawthorne’s examples. It might, e.g., involve some pragmatic elements and be dependent on choices made by the one providing the explanation. So, T might be the best explanation for L only given some parameter P that is a matter of a previous decision or convention. But then, the claim that T is the best explanation of L according to parameter P might seem necessary, and not contingent. An additional reason brought by Turri against Hawthorne’s The Explainer case is that it can hardly work as an example of human knowledge (which is, after all, what we are after), for it involves a super-human power of projecting entire experiential life histories and projecting explanations for them. Anyway, Turri proposes an example that keeps the same spirit as the former two, although differing in the details. The starting point is BonJour’s semi-classical view of a priori truths as those for which there is a reason to believe that they are true simply by understanding their content. The fact that such propositions are known immediately to anyone who just grasps them by means of rational insight is 13 This

view is similar in many ways to the one expressed in Audi (1999, pp. 211–2).

8.4 The “Most Unlikely Event”

151

not incompatible with the requirement that some experience might be required for understanding these propositions. Turri’s example which, according to him, avoids the deficiency pointed out in BonJour’s and Hawthorne’s case is the following: one can know, at any time tn , and simply by reflection (i.e., just by understanding) that (Most Unlikely) The most unlikely event is not occurring now where an event is considered as the most unlikely at tn according to a projection made, under deliberation, at the time tn−1 immediately before tn . Notice that there is not one proposition (Most Unlikely) that is known but, at each moment in time, there is one different proposition, since the most unlikely event is the result of a deliberation at tn−1 . So, at each moment tn−1 there might be one event that is, under deliberation, the most unlikely to occur at tn . However, Turri’s example seems to suffer from deficiencies similar to those that he pointed out in BonJour’s and Hawthorne’s cases. First, the example should be one of knowledge that is accessible to humans, but this seems hardly the case here, for it involves instant reasoning at tn−1 and instant comparison, at tn , with a deliberation made at the moment immediately before, something that hardly any human being could do: it requires a capacity that only God-like intellects could have of making instant reasoning. Second, Turri says, against the Best Explanation examples, that making them plausible requires providing a suitable account of explanation, and what it is for a hypothesis to best explain some data—a monumental undertaking, to say the least. (2011, p. 336)

However, the same seems to apply to the notion of most unlikely event. The latter seems, like best explanation, also in need of a deeper explanation. One could perhaps say that it is also relative to some parameters P , so that the most unlikely event according to P is not occurring now might turn out to be necessary. Moreover, depending on what one includes in the notion of event and how events compare in terms of likelihood with one another, there might well not be a most unlikely event as a matter of necessity (since for any unlikely event, there must be another less likely). I.e., it could be that, given any event E with probability k, there is always an event E  with probability k  such that 0 < k < k If there is no most unlikely event, then knowledge of (Most Unlikely), though a priori, would be necessary. Turri considers this objection, but the only possibility that he discusses is that of E  being the conjunction of E with some other event E ∗ , so that the likelihood of E  would be the product of the likelihoods of E and E ∗ . He tries to meet the objection by stipulating that, in reflecting on (Most Unlikely), we consider only atomic events (p. 339). This seems to make a priori knowledge of (Most Unlikely) even less plausible as an example of humanly accessible knowledge, for it excludes some aspects that are usually employed in discriminating events. If intensity of features such as color or temperature is taken

152

8 Some Other Cases

into account in discriminating events, then an event E  can be less likely than E without necessarily being the conjunction of E with some other event.14

8.5 Proper Contexts of Explanation Although BonJour considers the possibility of a priori justification of induction using the idea of best explanation (being, therefore, vulnerable to the kind of criticism raised by Turri), we could reformulate it in a simplified way in terms of explanation simpliciter, in the sense that, in the absence of induction, there is no explanation at all of observed regularities.15 We could think along the same lines that lead to the recognition of validity without necessity in the case of indexical sentences in Kaplan’s LD. As we remember, the leading idea is the restriction of contexts to proper contexts of utterances, i.e., to contexts such that, in the corresponding worlds, the speaker is at the time and location of the utterance, so that things like I am here now come out true. The restriction is motivated by what seems to be minimally required by utterances, i.e., an utterance cannot take place in a possible world in which the speaker is not at its time and location. The idea can be transposed to what we could call a proper context of explanation, which is a further elaboration of the notion of proper context of utterance by imposing one additional plausible requirement on the corresponding possible world: only contexts having as possible world one in which observed regularities are backed by objective regularities are proper for explanations. Worlds in which this does not happen are worlds in which explanations of observed regularities cannot take place, being in a specific sense improper. We are thus assuming that chance or pure coincidence is not an explanation at all. As we saw, BonJour considers alternative, rival explanations to induction besides chance or coincidence, and comes to the conclusion that they are not real alternatives after all, but disguised forms of induction. In other words, worlds in which the induction principle does not hold are metaphysically possible, but cannot contain proper contexts of explanation. BonJour’s idea is that we can have a priori justification for

14 E.g., an explosion with a higher intensity of light and heat can be seen as a distinct event from an explosion with a lower intensity. 15 This goes in the same spirit as Russell’s justification in The Problems of Philosophy

[W]e must either accept the inductive principle on the ground of its intrinsic evidence, or forgo all justification of our expectations about the future. (1912, p. 68)

8.5 Proper Contexts of Explanation

153

(Induction) The observed regular connection between As and Bs is due to something objective in the nature of A and B and the way they relate to each other. If induction is the only real alternative to chance or coincidence, then this claim can be a priori justified by The actual world is one in which regularities do not occur only by chance or coincidence. This is a consequence of (PCE)

The actual world can support proper contexts of explanation

or, alternatively, of (PCE*) The actual world is not located in that portion of the logical space in which regular connections occur by chance or coincidence. Now, (PCE) is valid under the restriction of contexts to proper contexts of explanation, in the same way that I am here now or Someone is located somewhere are valid as an effect of the restriction of contexts to proper contexts of utterances. Both (PCE) and (PCE*) are valid without being necessary. If we were to apply Evans’ terminology here, we would have to recognize (PCE) as superficially contingent. Hence, insofar as the a priori justification of induction is based on it, induction would also be superficially contingent. However, there are truths such as If there is an observed regularity between events of kind x and of kind y, then there is an inductive explanation for it with no indexicals, but which have the status of LDO-valid sentences. BonJour’s defense can be understood as the claim that, as long as explanation is a concern (and assuming that chance or coincidence is not really an explanation), (PCE) (or (PCE*)) is a contingent a priori truth. It is not a priori only if explanation is not a concern (in the same way that I am here now or Someone is located somewhere is not a priori only if utterances and the conditions under which they are possible are not a concern). Under these lights, BonJour’s defense of induction as an a priori principle appears far more plausible than Turri’s interpretation would have allowed (to whom the

154

8 Some Other Cases

notion of best explanation is problematic). We can also avoid the possible skeptical doubt concerning a priori justification based on something that is only likely (and not definitely true), as BonJour mentions in his discussion (1998, p. 208). It is only likely not in the sense that there is a high probability that it is true, but in the sense that not all possible worlds are worlds in which explanations can occur; but in every world in which explanations are possible, induction must hold.

8.6 Some Partial Conclusions Turri’s “Most Unlikely Event” case seems to involve some super-human capacities (such as instantly calculating the probability of possible events from one moment to the immediately next one). It also imposes a questionable restrictions to atomic events. Hence, it seems to have basically the same deficiency that Turri sees in Hawthorne’s The Explainer case, namely, it can hardly be seen as an example of human knowledge. Although Williamson’s The Believer example was received with some skepticism concerning its alleged indexical-free structure, there is a way of placing it under more favorable lights. We can extend the principle behind LDO-valid sentences to this example, by turning the notion of appropriate context of utterance into a notion of appropriate context of belief (i.e., by some natural restrictions on the possible worlds of admissible contexts). And, presumably, LDO-valid sentences include no indexicals. Similarly for BonJour’s defense of a general principle of induction (although his notion of “Best Explanation” is vulnerable to objections such as Turri’s). In this case we have to forge a notion of appropriate context of explanation (assuming, with Russell, that without induction there simply is no explanation at all for a class of phenomena or, with BonJour, that the only alternative to an inductive explanation is some other disguised form of inductive explanation). But the whole strategy depends on a more or less arbitrary restriction of the admissible contexts according to the relevant concept involved (utterance, belief, explanation). As noticed before, some people are skeptical about the legitimacy of such restriction.16

References Audi, R. (1999). Self-evidence. Philosophical Perspectives, 13, 205–228. BonJour, L. (1998). In defense of pure reason: A rationalist account of a priori justification. Cambridge: Cambridge University Press. BonJour, L. (2005). In defense of the a priori. In E. Sosa, M. Steup, & J. Turri (Eds.), Contemporary debates in epistemology (2nd ed.) Wiley-Blackwell.

16 E.g.,

Predelli (2005, p. 66).

References

155

BonJour, L. (2009). Epistemology: Classic problems and contemporary responses. Row man & Littlefield Publishers. Evans, G. (1979b). Reference and contingency. The Monist, 62(2), 161–189. Hawthorne, J. (2002). Deeply contingent a priori knowledge. Philosophy and Phenomenological Research, 65(2), 247–269. Kaplan, D. (1977). Demonstratives. An essay on the semantics, logic, metaphysics and epistemology of demonstratives and other indexicals. In J. Almog, J. Perry, & H. Wettstein (Eds.), Themes from Kaplan (pp. 481–563). Oxford: Oxford University Press, 1989. Kaplan, D. (1989). Afterhoughts. In J. Almog, J. Perry, & H. Wettstein (Eds.), Themes from Kaplan (pp. 565–614). Oxford: Oxford University Press. Oppy, G. (1987). Williamson and the contingent a priori. Analysis, 47(4), 188–193. Predelli, S. (2005). Contexts: meaning, truth, and the use of language. Oxford: Oxford University Press. Russell, B. (1912). The problems of philosophy. Reprint Edition, With Introduction by John Perry. New York: Oxford University Press, 1997. Searle, J., & Vanderveken, D. (1985). Foundations of illocutionary logic. Cambridge: Cambridge University Press. Turri, J. (2011). Contingent a priori knowledge. Philosophy and Phenomenological Research, 83(2), 327–344. Williamson, T. (1986). The contingent a priori: Has it anything to do with indexicals? Analysis, 46(3), 113–117. Williamson, T. (1988). The contingent a priori: A reply. Analysis, 48(4), 218–221. Williamson, T. (2000). Knowledge and its limits. Oxford: Oxford University Press.

Chapter 9

Basic Tools: Elements of a Theory of Speech Acts

A substantial part of this book’s contribution to a new (and, so far, unexplored) perspective on contingent a priori truths is derived from the properties of some illocutionary acts, namely, those involved in the generation of Kripke’s examples. In some cases, saying something changes the world in the sense of creating a fact corresponding to what is said (i.e., over and above just producing sounds or written words). So, it might be a good idea to have an outline of some essential aspects of the standard contemporary speech act theory, and to get clear about which acts are apt to have this effect and how. The systematic study of speech acts started with the work of Austin (1962) on performative utterances and performative verbs, and received its most widely accepted elaboration and generalization in the works of Searle (especially 1969, 1979a, 1985), Searle and Vanderveken (1985) and Vanderveken (1990a, 1990b). So, I will base my exposition on this theory.1 A reader familiar with speech act theory might prefer to skip this chapter and go straight to the next one where the theory is applied to Kripke’s examples.

9.1 The Dimensions of a Speech Act We can begin by noticing that almost all illocutionary acts (e.g., ordering, asking, asserting, suggesting, etc.) can be seen as the application of some illocutionary force

1 This is not to deny that there are other approaches with alternative taxonomies, e.g., the one more

recently presented in Roberts (2018). For a survey of alternative approaches and taxonomies to the one here favored, see Green (2017). © Springer Nature Switzerland AG 2022 M. Ruffino, Contingent A Priori Truths, Synthese Library 443, https://doi.org/10.1007/978-3-030-86622-8_9

157

158

9 Basic Tools

F (e.g., order, question, assertion, suggestion, etc.) to a propositional content P, so that the whole act can be represented as F(P) This is, in a way, an elaboration of Austin’s (1962, Chap. VIII) original idea that all acts in communication are a composition of a locutionary act (i.e., the production of a meaningful utterance) with an illocutionary act, i.e., the attachment of some illocutionary force to a meaningful expression.2 Despite the enormous variety of different ways in which one might use language as a tool and, hence, the enormous varieties of acts that one can linguistically perform, Searle (1979a) and later Searle and Vanderveken (1985) propose a method for developing a systematic classification of them into some few categories.3 This classification is based on an analysis of all illocutionary forces as a composition of seven basic elements:4 (1) Illocutionary Point: This is the general point or purpose of performing an illocutionary act. It might be, e.g., saying how the world is (describing, guessing, etc.), or saying how one wants the world to be (i.e., trying to get the audience to do some sort of action), or creating a new rule, or just expressing a feeling, etc. (2) Degree of Strength of the Illocutionary Point: E.g., a mere sketch of a certain area and a precise map have the same illocutionary point of describing that area, but differ in strength. (3) Preparatory Conditions: Most illocutionary acts require some things to be in place in order to be successful. E.g., saying ‘I hereby declare the defendant guilty’ (said at the end of a trial) can only be successful if the speaker is a judge in charge of that case; ‘I bequeath this land to my son’ can only be successful if the land belongs to the speaker, etc. Similarly, if one promises something, it must be clear that the thing or action promised is in the interlocutor’s interest.

2 Talking about a “composition” here does not mean that these two acts are produced separately and then combined; on the contrary, Austin conceives locutionary and illocutionary as occurring originally combined in an illocutionary act, and the locutionary act is conceivable only as an abstraction from the latter. 3 “There are not, as Wittgenstein (on one possible interpretation) and many others have claimed, an infinite or indefinite number of language games or uses of language. Rather, the illusion of limitless uses of language is engendered by an enormous unclarity about what constitutes the criteria for delimiting one language game or use of language from another. If we adopt illocutionary point as the basic notion on which to classify uses of language, then there are a rather limited number of basic things we do with language: we tell people how things are, we try to get them to do things, we commit ourselves to doing things, we express our feelings and attitudes, and we bring about changes through our utterances.” (1979a, p. 29) 4 Actually, Searle’s view on the basic elements changes over time. There are four elements in Searle (1969), twelve in Searle (1979a) (though many of them do not have a clear relevance in the determination of different forces), and seven in Searle and Vanderveken (1985). The latter work has a fuller formal development.

9.2 The Classes of Illocutionary Acts

159

(4) Propositional Content Conditions: Some (but not all) illocutionary acts require some restrictions concerning the appropriate kind of propositional contents for that kind of act. E.g., a promise requires that the propositional content concerns an action by the speaker and that is supposed to occur in the future (it is nonsensical to make a promise about the past). Similarly, an apology can only have a propositional content concerning an action done by the speaker (since one cannot successfully apologize, e.g., for the weather or for a mathematical law). (5) Mode of Achievement: Some (but not all) illocutionary acts are performed in some characteristic ways. E.g., an order is usually performed by invoking some sort of authority of the speaker.5 (6) Sincerity Conditions: Most (but not all) illocutionary acts disclose a psychological state on the speaker’s part. E.g., an assertion discloses a belief, while an order discloses a desire, etc. An act performed without the corresponding psychological state can be successful, but is insincere. (7) Degree of Strenght of the Psychological Condition: The sincerity conditions might require different strengths in different acts; e.g., strongly asserting something requires a stronger belief than merely guessing; begging requires a stronger desire than simply inviting. The illocutionary point is mainly determined by what Searle calls the direction of fit between word and world: it might be word to world (i.e., what is said has to fit the world, such as in assertions), or the opposite world to word (i.e., the world has to fit what is said, such as in orders and promises). It might also be both word to world and world to word (i.e., what is said creates a verifying fact and, hence, there is an automatic fit between word and world). Finally, it might also be null in the sense that no fit is required (such as in greeting or expressing joy).

9.2 The Classes of Illocutionary Acts Searle and Vanderveken’s thesis is that every illocutionary force might be understood as a combination of these seven elements or dimensions. The illocutionary point is, by far, the most important but not the only one in the determination of illocutionary force.6 Based on this identification of the basic elements, they propose the following famous classification of all illocutionary acts: 5 As

Searle and Vanderveken remark (1985, p. 16), sometimes the mode of achievement and the strength of the illocutionary point are interdependent, e.g., an order (issued by invoking the speaker’s authority) has a stronger degree of strength of the illocutionary point (getting the interlocutor to do something) than requesting. 6 Searle (1979a) notices that it would be desirable to build a taxonomy solely in terms of the direction of fit, but sees no way of doing this. That is to say, his taxonomy needs to use the other dimensions in order to do justice to the variety of illocutionary acts. For an attempt to build an alternative taxonomy based only on the direction of fit, see Roberts (2018).

160

9 Basic Tools

(1) Assertives: These are acts such as asserting, describing, guessing, testifying, etc., that have as illocutionary point giving a description of an independently existing reality (in the sense that the fact described is not created by the utterance itself). These acts can be represented as having the following structure: D ↓ B(P ) where D is the illocutionary point (describing a fact), ‘↓’ indicates that the direction of fit is from word to world (i.e., what is said has to fit the way the world is), B is the sincerity condition (which, in this case, is belief), and P is the propositional content (what is asserted, etc.). There is no general restriction on the propositional content, but in some fields there might be local restrictions on the kind of content that might be asserted. E.g., in assertions within mathematics, the propositional content cannot include any indexical or non-mathematical term.7 (2) Commissives: These are acts such as promising, vowing, accepting invitations, etc., whose illocutionary point is to make the speaker take responsibility for a future action.8 These acts have the structure C ↑ I (P ) where C is the illocutionary point (as described above), ‘↑’ indicates that the direction of fit is world to word (i.e., the world has to fit what is said), I indicates the sincerity condition (which is, in this case, intention, i.e., the speaker must have the intention to perform the action). Commissive acts have a basic propositional content condition, namely, the proposition must concern a future action on the speaker’s part. (3) Directives: These are acts such as ordering, asking, begging, suggesting, etc., whose illocutionary point is to try to get the audience to engage in some sort of action. They can be seen as having the following structure O ↑ D(P ) where O is the illocutionary point (as described above), ‘↑’ indicates the direction of fit (the same as in the commissive acts since, in both cases, the point is to try to get some action done so as to make the world fit what is said), D is the sincerity condition, which is, in this case, desire (i.e., the speaker must have a desire to see the relevant future action performed). As in the case of commissive acts, there is a basic propositional content condition for directive

7 For

the peculiarities of assertions within mathematics, see Ruffino et al. (2020). characterize the illocutionary point as taking responsibility for a future action instead of performing a future action because there might be unfulfilled promises that are, nevertheless, promises. 8 We

9.3 The Creative Aspect of Declarative Acts

161

acts, which is that the proposition must concern a future action by the audience. (Again, there would be no sense in, e.g., asking for the audience to perform something in the past, or asking for something that is not an action under the audience’s reach.) (4) Expressives: These are acts such as greeting, apologizing, congratulating, etc. in which the speaker does not attempt to describe anything, nor to take responsibility for or engaging anybody in a future course of action, but simply expresses a psychological state (regret, joy, etc.) in view of a propositional content that is taken to be true.9 They can be seen as having the structure EØψ(P ) in which E is the expressive illocutionary point, Ø is the null direction of fit. There is no particular sincerity condition here, and ψ is whatever is required for each particular expressive (e.g., apologizing requires regret, congratulating requires joy, etc.). Not all expressive acts require a propositional content; some of them (like greeting someone) do not seem to have one. (5) Declaratives: These are acts such as defining, nominating, baptizing, excommunicating, etc., This is a very special kind of act (and the one that most interests us here) in which the speaker tries to make the propositional content true by means of the utterance itself. If successful, it creates an automatic alignment between word and world because it creates the corresponding fact by means of the very illocution. They can be seen as having the structure D  ψ(P ) where D is the illocutionary point, ‘’ is the double direction of fit (since the utterance creates a fact that, on its turn, makes the utterance true), ψ is the sincerity condition, which might be required for some acts.10

9.3 The Creative Aspect of Declarative Acts An assertive illocutionary act does not, by itself, bring about the truth of the propositional content; the latter is true or false independently of the act (e.g., there might be a wrong prediction of the outcome of an election). Neither do commissive or directive acts: by themselves, they cannot make true the propositional

9 Many such acts correspond to what Austin called “behabitive”, i.e., they express an appreciation in view of other people’s (or the speaker’s own) actions. 10 In Searle’s original formulation (1979a) there is no sincerity conditions, but in (1985) Searle and Vanderveken recognize that some declarative acts might require a psychological state such as, e.g., belief.

162

9 Basic Tools

content concerning future actions by the speaker or by the audience. There might be successful promises that are not fulfilled or successful orders that are not followed. Their illocutionary point is not making that content true, but rather to count as a commitment or as an attempt to get someone to make that content true; the promise or the order themselves are normally not the relevant action. Likewise, expressive acts do not make any propositional content true (and, as said, in some cases, have no propositional content at all). Declarative acts, differently from the others, have as illocutionary point bringing about the truth of their propositional content by representing the way the world shall turn out to be. As Searle and Vanderveken put it, By definition, a declarative illocution is successful only if the speaker brings about the state of affairs represented by its propositional content in the world of the utterance [. . . ] All successful declarations have a true propositional content and in this respect declarations are peculiar among speech acts in that they are the only speech acts whose successful performance is by itself sufficient to bring about a word-world fit. In such cases, “saying makes it so”. (1985, p. 57)

E.g., when the chair in a meeting successfully says The meeting is open then the meeting is open as an effect of the declaration; similarly if the chair successfully says The meeting is adjourned at the end. To take another example, if a judge says at the end of a trial The defendant is guilty then, under this circumstance, and provided that the trial occurred with all formalities predicted by the law (preparatory conditions), the propositional content becomes true as an effect of the judge’s illocution (i.e., the defendant now has a certain legal status). The passage quoted from Searle and Vanderveken says that “All successful declarations have a true propositional content”. But it is important to have it clear that the content cannot already be true before the declarative act was made. In each case, the propositional content is first made true by the linguistic act itself. Since declarative acts are those that, if successful, make the propositional content true, the latter must be contingent since, if it were a necessary truth, it would be true independently of the illocutionary act.11 Although there is nothing grammatically wrong with saying I hereby declare that P or not-P,

11 Vanderveken

(1990a, p. 140).

9.4 Performatives

163

the act cannot be successful because the propositional content, which is a tautology, is already true anyway, and does not become true as a conventional result of the declaration. It might at first look like a magical effect of language, i.e., to create facts besides the utterances themselves. How can it be that some illocutionary acts make their propositional contents true just in virtue of their performance? The fact is that most languages have conventional instruments to perform this task, provided that the speaker satisfies some conditions. Some declarative acts require some sort of authority on the speaker’s part. More broadly speaking, many such acts presuppose the existence of an institution backing that act so that it can have the intended effect of creating another institution. E.g., if someone other than the chair says ‘the meeting is open’ or ‘the meeting is adjourned’, it is not thereby true that the meeting is open or adjourned, for the person performing the act lacks the required authority. A meeting being open or being over is an institutional condition, first created by the declarative act, and that can only be made true by a person having a certain institutional role. Searle (1989, p. 548) lists four general conditions for most declarative acts to be successful: (1) There must exist an extra-linguistic institution; (2) The speaker must have a special position within that structure; (3) A linguistic convention to the effect that uttering a certain sentence constitutes the performance of declarations within that institution must be in place; (4) There must be an intention on the speaker’s part that the utterance of that sentence has the declarative effect of creating the fact corresponding to the propositional content. The first two conditions correspond to what we were discussing, the institution and the role of the speaker within it, and the last two are an echo of Searle’s theory of meaning (1969, pp. 43–5), which combines a Gricean ingredient (speaker’s intention) with the requirement that this intention must be expressed using conventional means (words with rules of use).

9.4 Performatives Austin (1962) notices that some utterances of sentences are not, despite appearances, descriptions of facts in the world, but rather constitute actions of a special kind, and these actions are normally the ones denoted by the corresponding verb. E.g., (i) I promise to pay you back. (ii) I declare you husband and wife. [Said by a priest at the end of a wedding ceremony] (iii) I order you to leave.

164

9 Basic Tools

Saying (i) is not just describing an existing promise, but it is an act of promising; saying (ii) by the priest is not describing an existing marital status, but it is the creation of a new status (the Catholic marriage); saying (iii) is not just describing an order, but it is an act of ordering. Austin famously calls utterances like these performatives.12 This is by way of distinguishing them from utterances such as (iv) I play soccer. (v) I fry an egg. (vi) I climb the Everest. in which, by just saying, I am not performing the corresponding action: I do not (and cannot) play soccer just by saying (vi), I do not (and cannot) get an egg fried just by saying (v), and I do not (and cannot) climb the Everest just by saying (vi). Austin calls utterances of the latter kind constatives, and they correspond more closely to our ordinary, classic notion of utterance (of sentences) that are meant to be descriptions, being thereby true or false in virtue of how the world is. In the first chapters of his essay, Austin explores the idea that the performative effect of some utterances (like (i)–(iii) above) is ultimately due to the occurrence of a performative verb, i.e., verbs such as ‘promise’, ‘declare’, ‘order’, ‘baptize’, etc. He also seeks for some grammatical criteria to distinguish such utterances,13

12 “In

these examples it seems clear that to utter the sentence (in, of course, the appropriate circumstances) is not to describe my doing of what I should be said in so uttering to be doing or to state that I am doing it: it is to do it. None of the utterances cited is either true or false: I assert this as obvious and do not argue it. It needs argument no more than that ‘damn’ is not true or false: it may be that the utterance ‘serves to inform you’—but that is quite different. To name the ship is to say (in the appropriate circumstances) ‘I name, etc.’. When I say, before the registrar or altar, etc., ‘I do’, I am not reporting on a marriage: I am indulging on it.[. . . ]I propose to call it performative sentence or a performative utterance, or, for short, ‘a performative’.[. . . ] The name is derived, of course, from ‘perform’, the usual verb with the noun ‘action’: it indicates that the issuing of the utterance is the performing of an action—it is not normally thought of as just saying something.”(1962, pp. 6–7) Though many people agree with Austin that performative utterances are actions denoted by the corresponding verbs, fewer agree that such utterances are neither true nor false. 13 E.g., in most (but not all) cases, the verb is in the first person, present tense and indicative mood. E.g., ‘I promise. . . ’ is performative, but ‘He promises. . . ’ is not; ‘I promised. . . ’ is a report, but not a performative; similarly with ‘Do I promise. . . ?’. Most performative utterances admit the accretion of ‘hereby’ without changing their intuitive meaning. E.g., I order you to leave remains unchanged in meaning in I hereby order you to leave. The same does not happen with I play soccer and

9.4 Performatives

165

but ultimately becomes convinced that the performative effect is not necessarily dependent on grammatical features, neither is it dependent on the presence of performative verbs, since some utterances can have a performative effect without the presence of any such verb. E.g., Dangerous dog! can have the same effect as I warn you that the dog is dangerous without the presence of any performative verb. Austin also becomes progressively convinced, towards the end of his essay, that the distinction between performative and constative utterances is artificial, since all utterances effectively employed in communication have one or another performative effect, independently of the presence of a performative verbs. For that reason he moves the focus from the notion of performative verbs and utterances to the more general notion of illocutionary force, which is the element to be added to what he calls the locutionary acts (i.e., the production of meaningful utterances) in order to produce characteristically linguistic actions that he calls illocutionary acts (assertions, questions, warnings, orders, etc.).14 Austin sees all utterances employed in communication as a form of performative, with or without the occurrence of a performative verb. I.e., in a way, they are all special kinds of actions. Performative verbs are, so to speak, just a conventional instrument for making explicit the intended illocutionary force.15 But the notions of performative verbs and performative utterances remain important for Austin, and he undertakes an attempt to build a general taxonomy of illocutionary forces based on a classification of the illocutionary verbs into five big classes: Verdictives (which express “the delivering of a finding, official or unofficial, upon evidence or reason” (p. 152), and include verbs such as ‘acquit’, ‘convict’, ‘describe’, etc.);

I hereby play soccer. The former might be true, while the latter is plainly false. said before, the distinction between locutionary and illocutionary act is, for Austin, also artificial and made only for theoretical purposes since, in a speech situation, a locutionary act only occurs as part or aspect of an illocutionary act. 15 This claim inspired what would be later called the Performative Hypothesis (or Performative Deletion Analysis), i.e., the hypothesis that any utterance of a sentence in ordinary communication has a performative verb in its deep structure that is not necessarily evident in the surface structure (e.g., Ross, 1970, and Sadock, 1974). The Hypothesis was severely criticized both by philosophers (e.g., Searle, 1979b) and linguists (e.g., Levinson, 1983) based on overwhelming empirical counterevidence. 14 As

166

9 Basic Tools

Exercitives (“giving of a decision in favour of or against a certain course of action” (p. 154), e.g., ‘command’, ‘order’, ‘recommend’, ‘advise’, etc.); Commissives (“The whole point of a commissive is to commit the speaker to a certain course of action” (p. 156), e.g., ‘promise’, ‘undertake’, ‘give my word’, etc.); Behabitives (“include the notion of reaction to other people’s behaviour and fortunes, and of attitudes and expressions of attitudes to someone else’s past conduct or imminent conduct” (p. 159), e.g., ‘apologize’, ‘thank’, ‘deplore’, etc.); Expositives (“used in acts of exposition involving the expounding of views, the conducting of arguments, and the clarifying of usages and of references” (p. 160), e.g., ‘distinguish’, ‘postulate’, ‘deduce’, etc.). Austin’s original taxonomy of illocutionary acts is cumbersome, vague and ambiguous at many points, but it was the starting point for Searle and Vanderveken’s later taxonomy, which became a kind of received view nowadays, and is not based on a classification of performative verbs, but focuses directly on illocutionary forces and their components.16

9.5 Performatives and Truth Can we say that performative utterances such as I promise to pay you back describe something, e.g., that an illocutionary act (like promising) is taking place (being, therefore, true or false)? There are two conflicting answers to this question, and Harris (1978) called them the descriptivist and the anti-descriptivist interpretations of performatives.17 Austin himself held the anti-descriptivist view, i.e., performative utterances do not describe anything and, hence, are neither true nor false (1962, p. 6), but only felicitous (to use his terminology) if all preparatory conditions are in place, or misfires, in case some of these conditions are not satisfied.18 Austin’s main reason for thinking along these lines is that, in uttering a performative under the appropriate circumstances, a speaker is not so much describing an action as actually performing that action, and actions are not usually seen (or so Austin thought) as the kind of things that are true or false.

16 For a critical view of the latter taxonomy, and a study of its roots in Austin’s, see Sadock (2002). 17 This binomial here is not to be confounded with the descriptivist/anti-descriptivist theories of reference of singular terms, which label something quite different. 18 Austin includes yet another category of “infelicities” which he calls abuses, i.e., those cases in which some appropriate psychological state is missing like, e.g., a promise without the intention of fulfilling it (p. 18).

9.5 Performatives and Truth

167

Austin’s perspective was very influential and found echo among many philosophers and linguists.19 However, it is not consensual, and a number of other philosophers and linguists hold the opposite view, i.e., that performative utterances do have truth-value (being almost always true), and that they are made true by the facts that they describe, whereas these facts are created by the very utterances. Which fact is described by a performative utterance? The standard descriptivist answer is: it describes what the speaker does in uttering those words ‘I apologize’, ‘I promise’, ‘I declare’, etc. in the appropriate circumstances. Lewis (1970, p. 59), for instance, claims that, because performative utterances can rarely be anything but true (and are made true exactly by uttering them), it is easy to ignore (as Austin did) that they have a truth-value. But they could also be false if uttered under some non-appropriate circumstances (e.g., in play-acting, or in absurd situations like a sergeant issuing an order to a general). Lemmon (1962) argues that performatives are part of a broader class of utterances that are made true just by being uttered. Hedenius proposes the following definition: “S” is a performative ≡ “S” is true if and only if the utterance of “S” causes the state of affairs which makes “S” true and “S”’s social function is to be uttered in those circumstances where the utterance of “S” causes “S”’s truth. (1963, p. 119)

and argues that the main reasons advanced by the opponents of descriptivism are, in general, unjustified.20 Some have argued that descriptivism is not only misguided, but also incoherent. Most notably, Harris (1978) proposes what he regards as a fatal dilemma for this interpretation. His argument is based on the fact that, in many situations, one might correctly describe an action making a non-performative use of a verb, but a performative use of the same verb in that situation would be unsuccessful. Here are some of Harris’ examples: • S apologized to the lamppost, and even the police sergeant laughed. • S apologized on T ’s behalf, although T himself refused to do it. In both situations, the performative use of the verb (i.e., if S says ‘I apologize’ in those circumstances) would be defective.21 Now, suppose that, as the descriptive interpretation advocates, the performative expresses a description of the corresponding action. Is the description true or false in those situations? If we consider it false (or truth-valueless), we are adopting, for performatives, a standard of truth that is different from (and stricter than) the one adopted for ordinary descriptions since, in both situations, an ordinary description (with non-performative uses of the verb)

19 E.g.,

Hartnack (1963), Black (1963), Harris (1978), Taylor and Wolf (1981), Recanati (1987), Jary (2007). 20 The list of defenders of the descriptivist interpretation also include Kempson (1975), Edmondson (1979, 1983), Leech (1976), Spielmann (1980) and Wiggins (1971). 21 In the first situation, it is defective because S could only apologize to another person, but not to an object. In the second situation, it is presumably defective because, strictly speaking, S’s apology only makes sense regarding S’s own actions, but not someone else’s actions.

168

9 Basic Tools

is true. Hence, descriptivism is self-refuting. If, on the other hand, we consider it true, this is in conflict with the fact that the performative use in each situation is defective. Hence Harris’ dilemma: descriptivism is either self-refuting or materially inadequate.22

9.6 Performative Utterances as Declaratives Searle and Vanderveken (1985) consider successful performative utterances true because they are instances of declarative illocutionary acts. This is so because they treat the phenomenon of performativity, discovered by Austin, as a particular case of a phenomenon peculiar to successful declarative utterances, i.e., they can make their propositional content true.23 They also have a more restricted view on what counts as a performative utterance: although any communicative use of language is an illocutionary act, only those utterances containing an explicit performative verb

22 Harris

describes the situation in the following way:

There is no way out of this dilemma for the descriptivist: his interpretation automatically commits him to one or other of two self-defeating explanations of how to assign a truth value to what S says. No parallel problem in such cases emerges for the non-descriptivist, for the simple reason that for him the question of whether a performative utterance is true or false does not arise. (1978, p. 310) Harris’ dilemma gave rise to an interesting discussion (mainly among linguists) concerning the prospects of the descriptivist interpretation. Edmondson (1979, 1983) seeks to escape the dilemma (and, hence, to save descriptivism) by introducing a distinction between the semantics and the pragmatics of performatives, i.e., on some occasions (like those illustrating Harris’ dilemma) a statement like ‘I apologize’ might be true, although pragmatically the act of apologizing was not successful. So, meaning and truth of performative utterances are, in his view, independent from success in achieving the illocutionary point. Wachtel (1980) goes in the same direction by distinguishing the “going through the motions of performing a speech act” (which gives the semantic content of a performative utterance and its truth-conditions) from the satisfaction of its felicity conditions (which is neither contained nor implied by “going through the motions of performing a speech act”). Taylor and Wolf (1981) complained that Edmondson’s (and, by extension, also Wachtel’s) attempt to escape the dilemma makes the descriptivist’s position even worse because it relies on the assumption that one might correctly describe an illocutionary act as ‘he φed’ even though the felicity conditions of φing are not fulfilled (and, hence, the act was not carried on). Finally, Rajagopalan (1984) sees both positions as compatible insofar as they talk about distinct aspects of an illocutionary act: the semantic aspect (descriptive interpretation) and the act itself (anti-descriptive interpretation). 23 Declarative acts are such that We thus achieve the world-to-word direction of fit, but we achieve that direction of fit by way of representing the world as having been changed, that is, by way of word-to-world direction of fit. (Searle, 2008, p. 451)

9.6 Performative Utterances as Declaratives

169

are properly performatives. That is to say, performative utterances are a restricted class of illocutionary acts. In one aspect, this deviates from Austin’s original view since he thought that performative utterances are neither true nor false, being rather a special kind of action. But declaratives are, in Searle and Vanderveken’s approach, true if successful, and they are made true by the very utterance. In another aspect, their approach is close to Austin’s view, since it regards each performative (orders, questions, etc.) as a distinct speech act. That is to say, orders are orders, promises are promises, requests are requests, etc., and not indirect speech acts derived from a more basic and primitive act such as assertion.24 The main alternative view treats all performatives as assertions from which orders, promises, etc., can be inferred by means of Gricean-like processes.25 According to this approach, most performatives are indirect speech acts. Searle (1989) forcefully criticized this approach, basically insisting that in an adequate treatment, an order should be regarded as an order, a promise as a promise, etc., although his own account would have these performative utterances as being, at the same time and derivatively, assertions (i.e., an order such as ‘I order you to do A’ is at the same time an assertion that I order you to do A).26 Although some declarative acts require, as we saw, a special authority on the speaker’s part (e.g., being a judge, a priest, a chair, a legislator, etc.), some others require no such authority. This happens when the utterance involves performative verbs. If I say, e.g., (vii) I ask what time it is (viii) I promise to pay you back (ix) I declare that the Earth is flat then it is true that I ask what time it is in (vii), and that I promise to pay you back in (viii), and that I declare that the Earth is flat in (ix).27 In these cases, just by producing the utterance I make the content true, differently from (x) You are fired (xi) The meeting is adjourned in which something more than just saying is required, i.e., that I am the employer in (x) and that I am the chair in (xi).

24 Recanati

(1987) has a similar account, but with some differences. He divides performatives into three big classes: directives, commissives and declarations. The main feature of directives and commissives is that bringing about the state of affairs represented in the propositional content is intended to be the responsibility of the speaker and hearer, respectively, while in the case of a declaration, bringing about this state of affairs is meant to be nobody’s responsibility, but simply an immediate consequence of the declaration itself. 25 E.g., Hedenius (1963), Lewis (1970), Bach and Harnish (1979), Ginet (1979) and GarcíaCarpintero (2013). 26 Pagin (2004) and Jary (2007) follow Searle in this criticism. 27 Of course my declaration will be unsuccessful (despite the flat-earthers’ unshakable faith), but the fact is that a declaration comes into existence by the use of a performative.

170

9 Basic Tools

If we subsume performative utterances under declarative acts then, as a rule, the propositional content of any utterance such as I VP (where V is a performative verb, and P is any embedded propositional content) is that the speaker Vs P , and this is made true by the utterance itself. Performative utterances do not require any special authority on the speaker’s part to make their propositional content true. Or, to state it the way Searle sometimes does, language itself is an institution that backs performative utterances and is enough to enable the speaker to bring about the truth of such utterances. The facts created by performative utterances are just linguistic facts, i.e., the fact that the speaker Vs P . On the other hand, some declarative utterances that do not involve performatives, such as (x) and (xi) above, do not make the propositional content automatically true, since its conditions of success include some other form of authority (i.e., simply being a competent speaker is not enough).

9.7 Institutional and Linguistic Facts The outcome of a successful declarative illocutionary act cannot be a fact of an ordinary kind such as, e.g., that the sky is blue or that 2 + 2 = 4.28 For these are facts (or true propositions) that exist independently of any illocutionary act, whereas the point of a declarative act is to make a propositional content true in virtue of the act itself. Therefore, the kind of fact created by the declarative act must be of a special kind. Consider utterances involving a performative verb such as (vii), (viii) or (ix) above. The facts created are, respectively, that a question, a promise and a declaration exist. These are, in a way, purely linguistic facts (that are, nevertheless, not merely sounds or written symbols). But in cases like (x) and (xi), we have something different, i.e., that you have a new status (being unemployed) and that the meeting is officially adjourned. Searle (following Anscombe, 1958) calls the second kind of facts institutional facts. As the name says, these facts are institutions that are created by language through declarative acts. Performative utterances, on the other hand, are also declarative acts, but they only generate ordinary linguistic facts such as questions, promises, declarations, etc. For this reason, we can say that performative utterances are purely linguistic declarative acts and only generate linguistic facts, while acts of the same kind that do not involve performatives (such

28 Although many people see the latter fact as extraordinary because it involves abstract entities (numbers and operations). I mean here by ‘ordinary’ facts that exist independently of a linguistic act.

9.7 Institutional and Linguistic Facts

171

as (x) and (xi)) are extra-linguistic declarative acts that generate, if successful, institutional facts. Typically, a non-performative declarative act presupposes, as a preparatory condition, the existence of a previous institutional fact (e.g., some form of hierarchy and a particular position or authority of the speaker in it) and generates, if successful, another institutional fact. E.g., suppose that at the end of a trial a judge condemns a defendant by saying (C) The defendant is guilty. This declaration requires, as a preparatory condition, that the speaker is a judge, i.e., has a particular institutional status, that it comes at the end of a trial (another institution) following such and such rituals, etc.; if any other person utters the same sentence, the declaration is not successful, i.e., does not make its propositional content true. And, if successful, it creates a new institutional fact, i.e., that the defendant now has a specific legal status. If instead of (C) the judge says (C*)

I declare that the defendant is guilty,

since the utterance (C*) contains a performative verb, the propositional content that is made true is that the judge declares the defendant guilty, i.e., that a declaration exists. This is the linguistic fact that is directly created by the declaration of (C*). Anyone can say (C*) and, thereby, create a linguistic fact (that a declaration exists), since the only institutional fact needed to produce a declaration is language itself. But the linguistic fact, on its turn, only has as consequence the institutional fact that the defendant is guilty if the one producing the declaration is a judge. So, by uttering both (C) and (C*) the judge makes it true that the defendant is guilty (directly or indirectly). Any other person that is not the judge can utter (C*) and produce the fact that a declaration is made, but not the further fact that the defendant is guilty. In later works, Searle (1995, 2009) gives a more detailed account of the genesis and structure of institutional facts (as part of a global account of social reality) based on three conceptual tools. First, the notion of status function; second, the distinction between regulative and constitutive rules; third, the notion of collective intentionality. Let’s sketch each of them and see how they articulate to generate institutional facts in his view. We can impose on some objects functions that are derived just from their physical properties. E.g., we can impose on a bamboo stick the function of reaching out fruits on tall trees or impose on a knife the function of cutting. But one of the capacities that distinguish us from many other animal species is that of imposing on some entities functions that are not exclusively derived from their physical properties. E.g., we can impose on certain pieces of paper the function of being a universal medium of exchange (money), or to some words said in such and such a way the function of being a promise. In these cases, we are imposing what Searle calls a status on some entity, and the new function can only be displayed in virtue of that imposition. Sometimes the entity on which a status function is imposed is something material, such as a signed check, and sometimes it is an institutional fact itself, such as imposing to a promise the status function of creating duties and rights. The status

172

9 Basic Tools

function of an object is not derived only from the physical (or natural) properties it has, but depends on an act that usually involves some performative verb.29 The distinction between regulative and constititive rules is basically the classical one formulated in Searle (1969). Regulative rules, as the name says, regulate an activity that exists previously to the creation of the rule itself. E.g., the rule that professional football players of the same team must play using numbered T-shirts with the same color is a way of regulating a game that exists previously to (and independently from) the introduction of such rule. Or the rule (in some countries) that one should drive on the right side of the road. But constitutive rules are those that make a new kind of activity possible in the first place. E.g., the rule that one can touch a ball with any body part except one’s hands and arms, or that making the ball completely pass the line under the crossbar and between the goal posts counts as a goal are the kind of rules that make it possible to play football. Constitutive rules have, according to Searle, the form ‘X counts as Y in context C’. Another prototypical example is money: pieces of paper of such and such form and design (the X element) count as money (the Y element) in such and such country (the context C). Collective intentionality is required to collectively impose a status function on some entity. This involves collective acceptance of a constitutive rule, and an institutional fact continues to exist as long as that rule continues to be accepted. Searle sees collective intentionality as a primitive aspect of human nature, and it is not reducible to the individual intentionality of agents engaged in a collective task. Here is a passage showing how, in Searle’s account, the three elements (status function, constitutive rule and collective intentionality) come together to generate institutional facts: But the truly radical break with other forms of life comes when humans, through collective intentionality, impose functions on phenomena where the function cannot be achieved solely in virtue of physics and chemistry but requires continued human cooperation in the specific forms of recognition, acceptance, and acknowledgement of a new status to which a function is assigned. This is the beginning form of all institutional forms of human culture, and it must always have the structure X counts as Y in C. (1995, p. 40)

There are some interesting consequences of Searle’s view on institutional facts. One is that every institutional fact is a social fact (i.e., a fact involving the agency of several members of a community), but not every social fact is an institutional one because the latter requires more than just collective agency: it requires an institution (1995, p. 38). Social facts require just any form of collective behavior

29 In

Searle (2009) he presses the point and formulates what he describes as a “very strong theoretical claim”: The claim that I will be expounding and defending in this book is that all of human institutional reality is created and maintained in existence by (representations that have the same form as) S[tatus] F[unction] Declarations, including the cases that are not speech acts in the explicit form of Declarations. (p. 13)

9.7 Institutional and Linguistic Facts

173

(e.g., going for a walk or having dinner together), even if people do not do things together but synchronically do the same thing (e.g., if I am in my office quietly watching a game instead of doing my job, without knowing that my colleagues next door are doing exactly the same). But institutional facts require the imposition of status functions on entities through collective intentionality. When we come to the discussion in Chap. 10 of Kripke’s meter case as an institutional fact, the fact that there is collective intentionality behind the stipulation will help to explain why there can be contingent a priori knowledge not only for the stipulator, but also for anyone taking part in the collective acceptance of the meter standard.30 Another consequence is what Searle calls “the logical priority of brute facts over institutional facts” (1995, p. 34). The creation of an institutional fact requires a constitutive rule of the form ‘X counts as Y in C’, which attributes a status function to the X term. As we saw, the X term might itself be an institutional fact, on which we impose a new status function (e.g., passing a bar exam, which is an institutional fact, counts as being a certified lawyer in such and such context), or it might be a brute fact (e.g., being a piece of gold of a particular shape and weight counted as money in ancient times). So, there might be a chain of institutional facts being transformed into new institutional facts by means of the imposition of a new status function, but at the origin of that chain there must always be a brute fact to start with because a chain of institutions cannot be created simply out of thin air; they must start from some brute facts on which status functions are collectively imposed. As we shall also see in Chap. 10, in my view it is a brute fact that the meter stick has a certain length. And this brute fact is the basis for the institutional fact that it is one meter long (i.e., the unit of length). In this case we have a status function attributed to the stick by a declarative act. One last interesting consequence of Searle’s view that is relevant to our discussion is that language itself is an institutional fact. Saying things about the world, even describing brute facts, requires the existence of language. But sometimes language alone might not be enough to describe some brute facts, in that other institutions might be required for that description. E.g., Thus the statement that the sun is ninety-three million miles from the earth requires an institution of language and an institution of measuring distances in miles, but the fact stated, the fact that there is a certain distance between the earth and the sun, exists independently of any institutions. (1995, p. 27)

The distance between the Sun and the Earth is an objective feature of reality, but the fact that it is reported as being ninety-three million miles requires an institution besides language. As I shall argue in Chap. 10, something similar happens to the

30 Sometimes Searle talks of institutions instead of institutional facts as a “system of constitutive rules, and such a system automatically creates the possibility of institutional facts” (2009, p. 10). But the difference between institutions (such as universities) and institutional facts (such as academic degrees) seems to be just a matter of complexity: institutions are complex institutional facts that can be explored to generate, in a systematic way, new institutional facts. As I interpret him, they are, at the bottom, the same kind of entity.

174

9 Basic Tools

meter stick: it has the same length before and after a stipulation is made, but after the stipulation there is an institutional fact that makes it possible to say that it is one meter long.

9.8 Illocutionary Commitment and Inconsistency It will be very helpful to outline Searle and Vanderveken’s notion of illocutionary commitment, which is a relation that holds between two illocutionary acts α and β when, by performing α, a speaker is committed (in some special way) to perform β. This does not happen primarily because of logical entailment between propositional contents, but rather because of a relation at the level of illocutionary forces in the sense that one cannot consistently perform one of them but not the other. There are, according to them, two forms of such commitment. First, it might be impossible to perform α in any context without effectively and explicitly performing β. This is what they call strong commitment. Second, it might occur that, by performing α, the speaker is committed to perform β, although he or she might never come to actually perform β. This is because performing an illocutionary act incompatible with β would generate a special kind of inconsistency not explainable in terms of the propositional content, but in terms of the felicity conditions of α and β. They call this weak commitment. E.g., a report strongly commits the speaker to an assertion because reporting is just a special case of asserting. By asserting that P implies Q and asserting P , a speaker is weakly committed to assert Q, although he or she might never actually do so; by ordering that P a speaker is weakly committed to a permission that P . Some forms of illocutionary commitment hold simply because performing an act with a stronger illocutionary point entails a commitment to acts with a weaker illocutionary point. But some forms of illocutionary commitment occur between acts with different illocutionary points. E.g., a declarative act at t0 commits the speaker to assertions at later times performed in specific ways. If a speaker successfully nominates L his or her legal representative at a certain time, the nomination commits him or her to assert, at later times, that L is his or her legal representative. Moreover, the later assertion has to be done having as its main justification the former declarative illocutionary act. Closely related to illocutionary commitment is the notion of illocutionary inconsistency, which occurs when two acts cannot both be successfully performed at the same time, e.g., utterances of I order you to leave and I forbid you to leave. The conditions of success of one of them are incompatible with the conditions of success of the other one (in Austin’s terms, if one of them is felicitous, the other one must be a misfire), and the incompatibility is not necessarily at the propositional

References

175

level.31 Austin points at something along these lines when he talks about “selfstultifying procedures” that are parallel (but not identical) to contradictions: ‘I promise but I ought not’ is parallel to ‘it is and it is not’. [. . . ] Just as the purpose of assertion is defeated by an internal contradiction (in which we assimilate and contrast at once and so stultify the whole procedure), the purpose of a contract is defeated if we say ‘I promise and I ought not’. This commits you to it and refuses to commit you to it. It is a self-stultifying procedure. (1962, p. 51)

The notions of illocutionary commitment and illocutionary consistency, as suggested by Austin, and later outlined by Searle and Vanderveken, can be helpful in closing one of the gaps left open in Kripke’s account of contingent a priori knowledge, namely, the transmission of such knowledge from time to time, and from person to person.32 The next chapter, which contains the core of our proposal in this book, is basically an application of Searle and Vanderveken’s approach (without arguing for it, since this would go beyond the limited goals of this book) to Kripke’s cases of contingent a priori truths. In particular, we explore the creative aspect of declarative acts, and the notion of illocutionary commitment.

References Anscombe, G. E. M. (1958). On brute facts. Analysis, 18(3), 69–72. Austin, J. (1962). How to do things with words. Oxford: Clarendon Press. Bach, K., & Harnish, R. (1979). Linguistic communication and speech acts. Cambridge, MA: The MIT Press. Black, M. (1963). Austin on performatives. Philosophy, 38(145), 217–226. Edmondson, W. (1979). Harris on performatives. Journal of Linguistics, 17(2), 331–334.

31 See

Vanderveken (1990, p. 29, 152). the less known work of Adolf Reinach (1913), from the beginning of the twentieth century on the foundations of the civil law, we find passages like the following that seem to anticipate, in some ways, the idea that the performance of one illocutionary act may, as a matter of necessity, give rise to the existence (actual or just potential) of other illocutionary acts: 32 In

There is no doubt that the causal relation is no necessary “relation of ideas.” But it would be a mistake to extend this principle to every relationship obtaining between temporally existing things. The case which is now before us is the best proof of this. A “cause” which can generate claim and obligation is the act of promising. From this act, as we shall show more exactly, proceed claim and obligation; we can bring this to evidence when we consider clearly what a promise is, and achieve the intuition (erschauen) that it lies in the essence of such an act to generate claim and obligation under certain conditions. And so it is by no means experience in the sense of observation (Erfahrung) which instructs us, not even indirectly, about the existential connection of these legal entities; we have rather to do here with a self-evident and necessary relation of essence. (1913, p. 15)

176

9 Basic Tools

Edmondson, W. (1983). The descriptivist and performatives (again). Journal of Linguistics, 19(1), 183–185. García-Carpintero, M. (2013). Explicit Performatives revisited. Journal of Pragmatics, 49(1), 1– 17. Ginet, C. (1979). Performativity. Linguistics and Philosophy, 3(2), 245–265. Green, M. (2017). Speech acts. In E. Zalta (Ed.), The Stanford Encyclopedia of Philosophy, Winter 2017. Metaphysics Research Lab, Stanford University. Harris, R. (1978). The descriptive interpretation of performative utterances. Journal of Linguistics, 14(2), 309–310. Hartnack, J. (1963). The performatory use of sentences. Theoria, 29(2), 137–146. Hedenius, I. (1963). Performatives 1. Theoria, 29(2), 115–136. Jary, M. (2007). Are explicit performatives assertions? Linguistics and Philosophy, 30(2), 207–234. Kempson, R. (1975). Presupposition and the delimitation of semantics. Cambridge: Cambridge University Press. Leech, G. (1976). Metalanguage, pragmatics and performatives. In C. Rameh (Ed.), Semantics: Theory and application (pp. 81–98). Georgetown University Press. Lemmon, E. (1962). On sentences verifiable by their use. Analysis, 22(4), 86–89. Levinson, S. (1983). Pragmatics. Cambridge: Cambridge University Press, New York. Lewis, D. (1970). General semantics. Synthese, 22(1/2), 18–67. Pagin, P. (2004). Is assertion social? Journal of Pragmatics, 36(5), 833–859. Rajagopalan, K. (1984). The Harris–Edmondson dispute: Identifying the strawmen. Journal of Linguistics, 20(2), 251–256. Recanati, F. (1987). Meaning and force: The pragmatics of performative utterances. Cambridge: Cambridge University Press. Reinach, A. (1913). Die apriorischen Grundlagen des bürgerlichen Rechtes. Reprinted and translated by Crosby, J. as “The A Priori Foundations of the Civil Law”, Aletheia, III, 1983, pp. 2–142. Roberts, C. (2018). Speech acts in discourse context. In D. Fogal, D. Harris, & M. Moss (Eds.), New work in speech acts (pp. 317–359). New York: Oxford University Press. Ross, J. (1970). On declarative sentences. Readings in English Transformational Grammar, 222, 272. Ruffino, M., Venturi, G., & San Mauro, L. (2020). Speech acts in mathematics. Synthese, 198, ˝ 10063U-10087. Sadock, J. (1974). Toward a linguistic theory of speech acts. Academic Press. Sadock, J. (2002). Toward a grammatically realistic typology of speech acts. In S. Tsohatzidis (Ed.), Foundations of speech act theory (pp. 401–414). Routledge. Searle, J. (1969). Speech acts: An essay in the philosophy of language. Cambridge: Cambridge University Press. Searle, J. (1979a). A taxonomy of illocutionary acts. In J. Searle (Ed.), Expression and meaning: Studies in the theory of speech acts (pp. 1–29). Cambridge University Press. Searle, J. (1979b). Speech acts and recent linguistics. In J. Searle (Ed.), Expression and meaning: Studies in the theory of speech acts (pp. 162–180). Cambridge: Cambridge University Press. Searle, J. (1985). Expression and meaning: Studies in the theory of speech acts. Cambridge: Cambridge University Press. Searle, J. (1989). How performatives work. Linguistics and Philosophy, 12(5), 535–558. Searle, J. (1995). The construction of the social world. New York: The Free Press. Searle, J. (2008). Language and social ontology. Theory and Society, 37(5), 443–459. Searle, J. (2009). Making the social world. New York: Oxford University Press. Searle, J., & Vanderveken, D. (1985). Foundations of illocutionary logic. Cambridge: Cambridge University Press. Spielmann, R. (1980). Performative utterances as indexical expressions. Comment on Harris. Journal of Linguistics, 16(1), 89–93. Taylor, T., & Wolf, G. (1981). Performatives and the descriptivist’s dilemmas. Journal of Linguistics, 17(2), 329–332.

References

177

Vanderveken, D. (1990a). Meaning and speech acts, Volume 1: Principles of language use (Vol. 1). Cambridge: Cambridge University Press. Vanderveken, D. (1990b). Meaning and speech acts, Volume 2: Formal semantics of success and satisfaction (Vol. 2). Cambridge: Cambridge University Press. Wachtel, T. (1980). Going through the motions. Journal of Linguistics, 16(1), 85–88. Wiggins, D. (1971). A reply to Mr. Alston. In D. Steinberg & L. Jakobovits (Eds.), Semantics. An interdisciplinary reader in philosophy, linguistics and psychology (pp. 48–52). London: Cambridge University Press.

Chapter 10

Stipulations as Performatives

10.1 Two Gaps As we saw before, there seems to be two fundamental gaps in Kripke’s discussion of the classic cases. First, the nature of the truth-makers of contingent a priori truths is left unclear. Sometimes Kripke hints, without being completely explicit, that these are ordinary empirical (i.e., astronomical or physical) facts.1 This, I think, has been the source of some confusion in the literature concerned with the issue. Some critics point out that ordinary empirical facts cannot be the truth-makers of contingent a priori truths since it is impossible (for beings like us) to know such facts without resorting to some empirical experience. Such truth-makers would have to be facts generated (or first made accessible) by stipulations, but we find no further explanation in Kripke of how exactly stipulations create (or give access to) such facts or what they are like. Second, there is no explanation for the transmission of a priori knowledge from one speaker to another, and even from one speaker to herself at different times. Remember the way Kripke phrases a crucial aspect of the meter case:

This chapter is a slightly modified version of my “Performatives and Contingent A Priori Truths”, Synthese 198 (Suppl 22):S5593–S5613 (2021) 1 This

is so although in at least one passage he says that no new ordinary fact is known through the stipulation: But, merely by fixing a system of measurement, has he thereby learned some (contingent) information about the world, some new fact that he did not know before? It seems plausible that in some sense he did not, even though it is undeniably a contingent fact that S is one meter long. (1980, p. 63, footnote 26)

© Springer Nature Switzerland AG 2022 M. Ruffino, Contingent A Priori Truths, Synthese Library 443, https://doi.org/10.1007/978-3-030-86622-8_10

179

180

10 Stipulations as Performatives

What then is the epistemological status of the statement ‘Stick S is one meter long at t0 ’, for someone who has fixed the metric system by reference to stick S? It would seem that he knows it a priori. (1980, p. 56, my emphasis)

I.e., the alleged possibility exists, in his account, primarily for the baptizer (e.g., the meter-baptizer or Le Verrier). But what about the knowledge that people other than the baptizer might have that the stick is one meter long (based on the baptism)? It seems fair to say that they have some kind of a priori knowledge (if they accept the rule created by the baptizer) because they do not have to actually measure the standard stick in order to know its length; but how can they know this a priori if they must have seen the baptizer perform the baptism (thereby relying on information coming from their senses) or must have heard about it from someone else? Or what about the baptizer himself or herself at a time later than the moment of the baptism? It seems that he or she has to rely on memory, and because of that the a priori nature of his or her knowledge might be questioned.2 Can we give an account of the a priori nature of this kind of (later) knowledge and, if so, how is it connected with the knowledge that the baptizer has in virtue of its privileged position as baptizer? One might consider that, for the non-stipulators, knowledge must have come from testimony. And the status of knowledge transmitted by testimony is a controversial matter. Many authors (such as, e.g., Biro, 1995 and Malmgren, 2006) have it that the outcome of testimony can only be a posteriori knowledge. Others (such as Kant, 1787 and Burge, 1993) argue that, under some restrictions, testimony and memory do not change the original status of an item of knowledge, so its a priori nature is preserved. Kripke does not address these questions. The perspective that I am about to present can provide an answer compatible with the intuitions that the baptizer retains a priori knowledge after the performance of the baptism, and that people other than the baptizer (and who accept the latter’s authority) also have a priori knowledge. I shall not enter a general discussion of the many issues involving a priori preservation through testimony (and whether or not there is a strong parallel with the memory case, as Burge suggests). My aim here is to present a different perspective (one that explores the illocutionary aspects of stipulating using the theoretical apparatus from the last chapter) under which we might do justice to the above intuitions.

10.2 Stipulations and Illocutionary Acts Kripke considers the truth corresponding to (M) and (N) in abstraction from the circumstances under which they are produced. But a more realistic approach to this sort of sentences and the truths that their utterances generate should pay attention to what is involved in the linguistic act of their employment. Kripke says a couple

2 Chisholm (1989), for instance, argues that, if at any point in a reasoning one has to rely on memory, there is no a priori justification of the conclusion.

10.2 Stipulations and Illocutionary Acts

181

of times that the special contingent truths knowable a priori are the result of stipulations: We could make the definition more precise by stipulating that one meter is to be the length of S at a fixed time t0 . (1980, p. 54) [W]e have determined the reference of the phrase ‘one meter’ by stipulating that ‘one meter’ is to be a rigid designator of the length which is in fact the length of S at t0 . (ibid., p. 56) But that is obtained a priori solely in terms of a stipulation. (1986, p. 67)

If this is so, we should give a closer look at what is involved in stipulations. For they do not occur all by themselves and in abstraction from speakers and circumstances, but are essentially the result of a special linguistic act or, to use Austin’s terminology, the result of a special illocutionary act under appropriate circumstances, and their successful uptake usually requires some subsequent course of events and behavior of speakers and audience. The linguistic act involved in creating a contingent a priori truth is not simply the description of a pre-existing fact, but generates itself a non-purely linguistic fact. What is meant by “non-purely linguistic” in this context is that the fact is not reducible to an utterance (or to what Austin would call the locutionary act), but is something that comes as a conventional consequence of that utterance. Of course there are cases (e.g., a successful baptism) in which the fact created is just a linguistic convention and, in this sense, is also linguistic. But the existence of the convention is not just the locutionary act itself. Neptune with all its physical properties is not a bit different than it was before Le Verrier baptizing the cause of the perturbations in Uranus’ orbits as ‘Neptune’ (except perhaps that it now has the property of being called ‘Neptune’). But Le Verrier’s utterance in itself cannot be true or, better said, has no effect at all unless it is taken as a stipulation.3 As seen in the previous chapter, the most explicit forms of stipulations that produce an effect besides the mere linguistic act of uttering are those that Austin calls performative utterances such as I baptize this ship ‘Titanic’ [said by someone entitled to baptize ships] I declare you husband and wife [said by a priest at the end of a catholic wedding ceremony]. In these cases, by just saying or uttering (in the appropriate circumstances and with the appropriate intentions), the speaker is performing the corresponding action (of baptizing or of declaring a couple husband and wife before the Catholic Church). As we saw in the last chapter, Austin originally introduces the term performative in order to distinguish this kind of utterance from those that aim at describing independently existing facts and do not by themselves create the corresponding action (or the corresponding truth-maker).

3 One might hold a stronger view that, in some cases, the content of the utterance is neither true nor false before the utterance takes place (if, e.g., one is introducing a new word in a language by means of it).

182

10 Stipulations as Performatives

(M) and (N) themselves are not true and have no effect at all unless uttered by the meter-baptizer or Le Verrier in a particular way (and under some appropriate circumstances) that could be made explicit by the presence of a performative verb such as in (M*) (N*)

I stipulate (define, declare, etc.) that [‘One Meter’ refers to the length of S at t0 ]. I stipulate (define, declare, etc.) that [‘Neptune’ refers to the planet causing the perturbations in Uranus’ orbits].4

What appears before the brackets in the above statements is a performative verb that does not show up in the surface of Kripke’s original formulation. As Austin and Searle insist, the presence, explicit or implicit, of the performative verb is not really crucial. What is important is that the stipulation has a special illocutionary force (which might be made explicit by the performative verb) that is different from, e.g., the force of an ordinary assertion, or of a question, etc. Kripke and almost everyone discussing his examples of contingent a priori truths pay no due attention to the performative nature of stipulations and of reference-fixing.5 If this is so, we can look at Kripke’s examples from an entirely different perspective, i.e., as cases of performative utterances or, more broadly speaking, of declarative illocutionary acts.6 The most important feature of such acts is the double direction of fit, i.e., the act itself, if successful, creates automatic alignment between word and world (differently from other kinds of illocutionary acts) because the act itself brings about a new fact that verifies it.

4 From

this point on, it seems more adequate to suppress the existential clause from the propositional content since, from the perspective of illocutionary acts that I want to motivate, existence fits better as a felicity condition than as part of the propositional content. 5 An exception is Jeshion (2002) who points out that Within this debate, philosophers have tended to forget that all acts of naming—ostensive and descriptive alike—are genuine performatives, often explicit performatives. (p. 63) Kripke’s examples are based on the stipulative introduction of rigid names and, hence, depend on performatives. She goes on proposing a list of felicity conditions peculiar to the act of naming, and concludes that many (or most) examples of naming in the literature concerned with Kripke’s thesis are actually infelicitous (i.e., some of the conditions are not met). Her main goal is actually to defend the possibility of acquaintanceless de re beliefs, and her felicity conditions (which deviate, e.g., from Searle’s, 1969) include some sincerity and psychological clauses that supposedly make this kind of belief more palatable to a skeptic. Incidentally, I find her list of felicity conditions not quite convincing because they include the speaker’s interest in introducing a name, something that appears to me quite vague and not clearly relevant. But, at any rate, she focuses only the primary performatives (M*) and (N*) and their credentials as possible sources of acquaintanceless de re beliefs, but leaves aside what I will call (M**) and (N**) as possible sources of contingent a priori truths. Another exception is Horowitz (1983), whose view on the topic we shall discuss later in this chapter. 6 Remember that we are adopting Searle and Vanderveken’s account of performatives as declarative acts.

10.2 Stipulations and Illocutionary Acts

183

But what exactly is the propositional content made true by the successful performative (declaration) implicit in Kripke’s cases? Is it just the proposition that an object (a length or a planet) has a particular name? Or something else? (Or maybe both?) To get a better perspective, let us consider for a moment a more contemporary situation illustrating the spirit of Kripke’s examples. I have in mind the new definition of kilogram introduced by the General Conference of Weights and Measures (GCWM, held in Versailles in November 2018). As it is widely known, the GCWM changed the old definition of ‘One kilogram’ as the weight of a particular piece of platinum-iridium (known as “Le Grand K”) to a new and more precise one in terms of Planck’s Constant (h). (For the sake of simplicity, I express the definite description used in the new definition as the (h).) Now one might consider what the relation is between the following two distinct declarative illocutionary acts: (IA1 ) (IA2 )

I (or we, the GCWM) stipulate (define, declare, etc.) that [‘One kilogram’ refers to the (h)] I (or we, the GCWM) stipulate (define, declare, etc.) [One kilogram is the (h)].

The first illocutionary act (IA1 ), which is the one originally performed by the GCWM, is a baptism or linguistic stipulation, a performative act having a metalinguistic proposition as content and, if successful, brings about its truth.7 But the second act (IA2 ) has as content a different kind of proposition and, if successful, brings about not just a metalinguistic fact, but also something else. Indeed, if successfully performed by the appropriate authority, (IA2 ) has an impact not just on language and the way we speak but also in the practice of measurement worldwide, particularly in science and industry. Now, there is a crucial aspect of this situation: by performing (IA1 ), the GCWM was also committed to a performance (implicit or explicit) of (IA2 ).8 It seems clear that (IA2 ) presupposes (IA1 ), because if there were no (IA1 ) there would not even be a proposition expressed by the portion in brackets in (IA2 ) (since 7 One could perhaps object here that there is no real baptism being performed by the GCWM since ‘One kilogram’ already had a reference (fixed in terms of ‘the weight of Le Grand K’). One way to articulate this criticism is that we do not have another entity (one kilogram) getting the name ‘One kilogram’, but just the old entity being selected by a different description. This question belongs, I believe, to the philosophy of science, and I shall forgo a deeper discussion of it here. I shall only say that there are some reasons for thinking that this is not the case. One of them is that there was a margin of error and vagueness of ‘the weight of Le Grand K’ that made the old stipulation imprecise. Hence, what was baptized was not a precise weight (or a single entity), but a fuzzy range of weights (a fuzzy range of entities), and the elimination of vagueness is what motivated the GCWM’s act in the first place. I shall assume that the result of the GCWM’s utterance in IA1 is an entirely new baptism, giving a new meaning to (or fixing a new reference for) ‘One kilogram’, even though the new meaning resembles in some ways the old one. 8 Remember that I am using the notion of illocutionary commitment discussed in last chapter, which is a relation holding between two illocutionary acts if, by performing one of them, a speaker is committed (in some special way) to perform the other one. The claim here is that a performance of (IA1 ) seems to weakly commit the speaker (the GCWM) to a performance of (IA2 ).

184

10 Stipulations as Performatives

in that case the name ’One kilogram’ would not have a reference). But given that (IA1 ) was (successfully) performed, and given the GCWM’s authority not just in linguistic matters but also in matters of measurement, we can say that the GCWM is committed to (IA2 ) as well. Now back to Kripke’s examples. As we saw, (M) and (N) must be seen as part of performative illocutionary acts, otherwise they have no effect at all. The acts are represented by (M*) and (N*), in which a performative verb is explicit (or implicit). Let us focus on (M*) first, since the effect that I want to highlight is clearer in this case. (Later I will come back to (N*) and argue that, despite appearances, it has some fundamental differences with (M*).) Now, this is not the only act performed by the baptizers since, as we saw in the GCWM example, the baptism or stipulation has not only a linguistic effect, but also an extra-linguistic (institutional) effect. The additional act performed as an illocutionary consequence of the original one can be expressed by (M**)

I stipulate (define, declare, etc.) that [One meter is the length of S at t0 ].

Something is crucial about (M**): as said, it is not an assertion, but a declarative speech act, and in order to know whether any speech act is successful, we need to know whether its preparatory conditions are satisfied. This most of the time involves empirical (a posteriori) knowledge. But once we know that a declaration is successful, there is no need for further empirical justification of its propositional content: it must be taken as true, because the illocutionary point of a declaration is precisely to make a propositional content true by means of the very utterance. Hence, if we know that the preparatory conditions of (M**) are in place, we can take its propositional content as true without any need of the relevant experience (in our case, measuring the standard stick to find out its length). In that sense, we have a priori knowledge of the propositional content of (M**). Here we have a close parallel with a priori knowledge of a proposition expressed by an utterance (e.g., of ‘2 + 2 = 4’). First, we need to know that the utterance expresses a particular proposition, and that might depend on empirical knowledge: it involves, e.g., knowing that the first occurrence of ‘2’ stands for the number two instead of something else, that ‘+’ stands for addition, that the second occurrence of ‘2’ stands for the same thing as the first one, etc.; and if the meaning depends on the speaker’s intentions, it also involves finding out what these intentions are, and also finding out whether the utterance came out right in the sense that it reflects the speaker’s intention, etc. This knowledge certainly involves some sort of experience, being therefore a posteriori. But once it is known which proposition is expressed by that utterance, there is no further appeal to experience in order to justify it and, in that sense, we can say that we know a priori the proposition expressed. This is not, however, an uncontroversial view. One might be tempted to think that the need for empirical knowledge in order to simply understand a proposition destroys the credential of any a priori knowledge of that proposition. I have no space for a deeper discussion on the nature of a priori knowledge in relation with experience here. I shall just say that I follow the spirit of BonJour’s proposal that

10.2 Stipulations and Illocutionary Acts

185

allows for empirical experience in the understanding of a proposition that can, nevertheless, be known a priori: I choose to follow Kant and the overall tradition by stipulating that a proposition will count as being justified a priori as long as no appeal to experience is needed for the proposition to be justified once it is understood, where it is allowed that experience may have been needed to achieve such an understanding. (BonJour 1998, p. 10)

One could raise the following objection, based on the fact that knowledge of the satisfaction of the preparatory conditions of a speech act must be a posteriori: suppose that we have a right triangle, and by measuring the two shortest sides we conclude that they are 3 and 4 inches in length; then, applying the Pythagorean theorem we reach the conclusion that the largest side is 5 inches in length. But even if we are employing an a priori form of reasoning (the Pythagorean theorem), knowledge concerning the longest side was based on a posteriori knowledge of the premises concerning the shorter sides. Hence, our knowledge concerning the longest side can at best be a posteriori. The objection is that the same would apply to knowledge of the outcome of a successful declaration: in order to know that a declaration is successful we need to know that the preparatory conditions are satisfied, and this is typically a posteriori. Then, one can (normally) only know a posteriori that a declaration is successful and, hence, knowledge of its effect (the fact created by it, e.g., the conviction of someone in a trial) can at best only be a posteriori. Notice that this objection, if effective, would show that there could never be any a priori knowledge of any outcome of a declarative speech act, since every possible such act has some preparatory condition whose satisfaction can only be known a posteriori. Hence, there could not be, e.g., a priori knowledge of the outcome of a mathematical definition, because even such definitions, being declarative acts, have preparatory conditions (e.g., that the terms occurring in the definiens are either primitive or have been previously defined in the theory, and that the definition has come out according to the mathematician’s intentions, etc., which are instances of a posteriori knowledge). No such knowledge would be possible for mathematical postulates either, since a postulate, as the name says, must have been postulated at some point, and postulating is a declarative act, with similar preparatory conditions (e.g., that the postulate comes out according to the author’s intentions, that the formulation does not involve extra-mathematical terms, etc.). So, knowing that something is a postulate would depend on knowing that the act of postulation was successful. But in the same way that we can, following BonJour (and Kant), distinguish between what is necessary to understand a proposition (which might include empirical experience) from what is necessary to justify that proposition once it is understood (which might come out simply from rationally understanding it), we can distinguish between what is required to know which speech act was intended and whether or not it was successful (which might involve experience) from what we can infer from the fact that it was successful (which is an outcome just of our linguistic competence).

186

10 Stipulations as Performatives

The analogy with the triangle case also breaks down at another point: in the latter, there is a necessary falsehood if we assert that the shorter sides are 3 and 4 inches in length and the longer side is not 5 inches in length; but in the declarative case, there is no necessary falsehood if we take a declaration to be successful but deny that its propositional content is true (although there seems to be some kind of Moorean paradox here). The connection between the successful declaration and the assertion of its propositional content does not come from propositional logic, but from what Searle and Vanderveken call illocutionary logic, i.e., a connection between illocutionary acts. One might also raise another (related) objection that we can be more confident that there is a priori knowledge of the content of mathematical definitions because we are in the realm of mathematics, while things are less clearly so in stipulations such as (M*) and (M**): here we must know that the stipulation was successful and this also requires empirical knowledge such as, e.g., that the one fixing the meter standard has the required authority, etc. To deal with this objection we must first get clear that, when we talk about the content (or knowledge) of mathematical definitions, we are not talking about the content (or knowledge) of ordinary mathematical assertions (such as theorems, lemmas, etc.). The content of a mathematical definition is neither true nor false before the definition is actually made.9 It is only after the definition is made that its content can be asserted (and after the definition it becomes true). So, when we consider knowledge of the content of a mathematical definition, we are not talking about knowledge of an ordinary mathematical proposition (which, as a necessary truth, is true independently of any speech act). Instead, we are talking about knowledge of a content that is first made true by the illocutionary act characteristic of definitions. Now, the objection assumes that we have to know that preparatory conditions are successful in the case of a stipulation (and this involves empirical knowledge) and this would mark a difference with mathematical discourse; but this contrast is artificial, since the same requirement is present in the case of mathematical definitions as well. From the illocutionary point of view, both a mathematical definition and the stipulation of a standard of measurement have the same structure, i.e., are declarative illocutionary acts (and, therefore, both are performative utterances)10 performed by speakers (the 9 This relates to a classical point made by Frege (1879 §24, 1914, p. 211), i.e., that the content of a definition is not the object of assertion because it introduces a new term. The content of a definition is made true by the very act of the definition (for which Frege has a special sign, distinct from the assertion sign). In the mathematical practice, however, one might find exceptions to this rule since sometimes one takes a true mathematical content (e.g., that an infinite set can be brought into a bijection with a proper part of itself) and decides that this is to count as a definition (in this case, of infinite sets). I shall leave the discussion of these exceptions for another occasion. Anyway, the nature of the content of a mathematical definition is quite a delicate matter. I shall discuss Frege’s early view on definitions in the next chapter. 10 Any definition in mathematics contains, explicitly or implicitly, a performative verb such as ‘define’, ‘call’, ‘name’, etc. Austin (1962, p. 163) classifies ‘define’ as an “expositive” performative, while Searle and Vanderveken (1985, p. 205) classify ‘call’, ‘name’, etc. as “declarative” illocutionary verbs.

10.2 Stipulations and Illocutionary Acts

187

mathematician and the stipulator); both have conditions of success, the knowledge of which most likely involve some empirical information.11 And both are successful if and only if they first make their propositional content true by means of the utterance itself.12 The ultimate justification for knowledge of the content of a mathematical definition is that there is such a definition in the system, and that it was successful as a declarative act. With mathematical definitions we have to separate two distinct aspects: one thing is knowledge that the preparatory conditions are satisfied (and this might almost always involve some empirical, a posteriori knowledge), and another thing is the justification of its content (which appeals to no further empirical evidence but simply to the fact that a successful definition was made). The presence of some unavoidable residual empirical knowledge concerning the satisfaction of the preparatory conditions is not normally seen (at least according to the BonJour-Kant epistemology that we are assuming) as destroying the credentials of mathematical definitions as generating a priori knowledge. My point here is that something similar holds for declarative speech acts such as (M**) since they are not different from mathematical definitions with respect to their illocutionary structure and also with respect to the fact that knowledge of the preparatory conditions must involve some empirical information. If the act depends somehow on the speaker’s intention, we need to recognize what these intentions are, and also that the utterance comes out according to them (instead of being gabbled or a spoonerism, etc.). All these items seem to qualify as a posteriori knowledge. But once we know which kind of act was intended, and that the intended act was successfully performed, there is, in the case of a declaration, no need of further empirical justification for taking the propositional content as true.13 As Searle (1989,

11 In the case of a mathematical definition, I have in mind information such as: that the defined term (or notion) was not previously defined in a different way within the same theory (in which case the new definition could perhaps be seen as a misfire since its purpose is defeated); that the definiens does not contain any term that is not either primitive or has been previously defined; that the definition came out corresponding to the author’s intentions; that its author was using language properly, etc. See Belnap (1962) for a discussion of some criteria of admissibility of definitions such as consistency with antecedent assumptions, conservativeness and uniqueness. Of course, after a mathematical definition is successfully made, one might have to provide yet another sort of justification such as that it is fruitful (i.e., yields interesting results), or materially adequate, or that it captures an intuitive pre-theoretical idea. But this is a different level of justification in terms of the methodological advantages brought by the new definition. Its content is always true anyway if it was the subject of a successful declaration; whether it is satisfactory and useful for scientific purposes is quite another matter. 12 This is not to say, of course, that there are no differences in the preparatory conditions of mathematical definitions and stipulations such as (M*) or (M**). E.g., the latter might require a special kind of authority of the stipulator, while the former perhaps requires no special authority except being a working mathematician, etc. 13 As we saw in the previous chapter, this holds for declarative acts, but not necessarily for other illocutionary acts. If, e.g., an order (or a promise) is successfully performed, this does not mean that its propositional content is thereby true, since it might not be obeyed (or fulfilled). I do not mean that any successful speech act generates a fact that one can know just because the act is successful.

188

10 Stipulations as Performatives

p. 547) puts it, a successful declaration is in itself sufficient to bring about the fit between words and world. Hence, there is a priori knowledge of the propositional content of (M**) in the following sense: no other justification is necessary except the successful utterance itself.

10.3 Horowitz The driving question in this chapter is the same as Horowitz’s (1983), i.e., what is involved in the act of stipulation that serves as basis of Kripke’s cases of contingent a priori truths. But the conclusion that I am suggesting is quite different from the one that she takes as unavoidable. She expresses an overall skepticism concerning the prospects of what she describes as an “epistemological privilege” (i.e., the possibility of a priori knowledge) resulting from stipulative referencefixing, in the same spirit as the Quine-Putnam attack on the empiricist philosophy of language. Horowitz’s main point is based on the observation that there are preparatory conditions to be met for a successful stipulation (such as, e.g., that the stipulator has the required authority to introduce a name, that the stipulation came out according to the stipulator’s intentions, that other participants in a public language will follow the same use of that name, etc.) and, since knowing all these conditions involve in one way or another empirical information, there can be no a priori knowledge at all in Kripke’s cases. In my view, her overall skepticism is exaggerated because she does not fully appreciate the fact that the relevant speech act in such cases is a declarative one, and that the illocutionary point of such acts is to make a propositional content true by the very act. Typically, knowing a contingent proposition of the form φa (in which a contingent property φ is predicated of an object a) requires some characteristic empirical experience (e.g., measuring a). But if φa is the content of a successful declarative act, then this particular experience is not required anymore in the case of a, and the ultimate justification for attributing φ to it is that φa was the subject of a successful declaration. (There might be the need for further justification of this content qua stipulation in terms of methodological advantages or adequacy of the same kind that was mentioned earlier (see footnote 11) for mathematical definitions, but this is another matter, and does not have to do with the justification of the truth of the content.) This is in contrast with any other proposition φx (for x other than a), because the latter does require the characteristic experience. Of course, knowledge of the preparatory conditions of the act almost invariably requires some empirical information. This is, I suppose, the reason for Kripke’s careful note in his later lectures on contingent a priori truths:

My claim here is restricted to declarative speech acts: these are acts whose illocutionary point is to make a proposition true by means of the very utterance.

10.4 Contingency

189

But then a startling—or to some people startling, anyway—conclusion seem to follow: the agent who has introduced this knows, apparently a priori (or anyway very close to it) just by making this stipulation that if stick S exists, S is one meter long at t0 . (1986, p. 7, my emphasis.)

The qualification “or anyway very close to it” is meant, as I see it, to leave room for the fact that knowledge of the surrounding conditions of success might be empirical. Horowitz seems to think that, because of this, the whole thing is dependent on empirical information, being therefore a posteriori. But, as already said, if she is right her argument precludes the possibility of any a priori knowledge of any definition or even assertion in mathematics (or logic), since in every such situation there is some relevant residual empirical information of the preparatory conditions of the corresponding act. Moreover, some such a priori knowledge is required, in the Kripkean perspective, to ground, e.g., a measurement system, without which there is no empirical experience concerning measurement (and, hence, no a posteriori knowledge either). Kripke makes the point (in the later lectures) in reply to Salmon that if no one can know a priori the length of S, than no one can know a posteriori any length at all either. The same point could be made in reply to Horowitz’s skepticism as well. If we apply this perspective to the example of the new definition of a kilogram made at the GCWM in terms of the (h) we can see that some empirical information is necessary to understand the description, since it involves (h) (i.e., Planck’s constant, relating the energy of a photon to its frequency). But this still leaves open a margin of decision for the GCWM as to the exact amount of Planck’s constant to be taken as the kilogram, and the particular decision taken (i.e., the exact amount) and proclaimed as the new definition was not mandatory by any law of nature: its ultimate justification does not rest on empirical information (otherwise there would be no need for a voting at the end of the conference, as it in fact happened, but just the need for more accurate research to find out how nature is), but on the fact that a decision was made by the competent authority. Again, it might well be that the new definition has to be justified in terms of methodological adequacy, but this does not concern its truth.

10.4 Contingency As Vanderveken (1990a, p. 140) points out, a declarative illocutionary act must have a contingent propositional content, for otherwise, i.e., if the content were a necessary truth, it would not be made true by the utterance itself because necessary truths are already true independently from the utterance.14 I.e., a successful declarative illocutionary act combines both features: apriority (in the sense that we know that its propositional content must come out true just in virtue of the act itself) and

14 There

might be exceptions, as mentioned in Footnote 9.

190

10 Stipulations as Performatives

contingency (since the propositional content of a successful declaration must be contingent).15 Do we have the same effect in the case of (N*), i.e., can we say that by performing (N*) the speaker is also (weakly) committed to (N**)

I stipulate (define, declare, etc..) that [Neptune is the planet causing the perturbations in Uranus’ orbits]?

Things are less clear here since, after all, that Neptune is the planet that causes the perturbations in Uranus’ orbits is an astronomical fact that cannot be stipulated (at least not by human beings); this seems to be true independently of any linguistic act. This point is at the heart of Donnellan’s criticism. Kripke indeed sometimes speaks as if Le Verrier had a priori knowledge of an astronomical fact as a result of stipulation.16 And, as we saw, Donnellan complains that it seems preposterous to expect that a mere act of stipulation could bring about a contingent fact. To quote a crucial passage again: Surely only God, if even He, could perform the miracle of stipulating how the world shall be. (1977, p. 19)

But we can think about the issue from the perspective of performatives along the lines already explained. We must start by distinguishing two questions concerning (N**): first, is it the case that, by performing (N*), Le Verrier was also committed (in the sense explained in Sect. 9.8) to (N**)? Second, can a performance of (N**) ever be successful as a declaration? The answer to the first question is less clear than in the meter case, and I do not see a decisive reason for an affirmative answer (or for a negative either). But the fact is that both Kripke and Donnellan think of the Neptune case as if Le Verrier were performing something resembling (N**)

15 Kripke

himself seems to oscillate between (M), (M*) and (M**) and to think that utterances of all three kinds are equivalent (although, of course, he does not formulate the discussion in terms of illocutionary acts). Although most of the time he says that the relevant stipulation is the act of naming (i.e., something like (M*)), in some passages he thinks of the stipulation as something like (M**): We could make the definition more precise by stipulating that one meter is to be the length of S at a fixed time t0 . Is it then a necessary truth that stick S is one meter long at time t0 ? Someone who thinks that everything one knows a priori is necessary might think: ‘This is the definition of a meter. By definition, stick S is one meter long at t0 . That’s a necessary truth.’ (1980, pp. 54–5; my emphasis.)

16 This is the impression that one might get, e.g., from Kripke’s description of Le Verrier’s epistemic situation at the time of the baptism:

At that time he was unable to see the planet even through a telescope. (1980, footnote 33) This sentence seems to have the literary effect of setting the stage for an extraordinary and somewhat heroic epistemic deed of Le Verrier, i.e., knowing an astronomical fact before observing Neptune just by means of a clever stipulation.

10.4 Contingency

191

(besides, of course, performing (N*)).17 Regarding the second question, it is helpful to look at Searle’s (1989) suggestion about performative utterances: there are some things that can be achieved by means of them, and some things that cannot, but this is not due to the semantics of performatives, but to the way things are in the world. As he puts it, If God decides to fry an egg by saying, “I hereby fry an egg,” or to fix the roof by saying, “I hereby fix the roof,” He is not misusing English. It is just a fact about how the world works, and not part of the semantics of English verbs, that we humans are unable to perform these acts by declaration. But there is nothing in the semantics of such verbs that prevents us from intending them performatively; it is just a fact of nature that it won’t work.[. . . ] There is nothing linguistically wrong with the utterance, “I hereby make it the case that all swans are purple.” The limitation, to repeat, is not in the semantics, it is in the world. (1985, p. 554)

This means that performative formulas attempting to achieve, by means of language, something impossible to human beings are doomed to failure, but are not absurd. We can follow both Kripke and Donnellan and see (N**) as something that Le Verrier (explicitly or implicitly) tries to perform, but as an act that can never achieve its illocutionary point (i.e., bringing about an astronomical fact) because this is not the kind of thing that human beings can do by means of utterances, although, as Searle says, there is nothing grammatically wrong or nonsensical with Le Verrier’s attempting to do it performatively, and maybe even as an illocutionary consequence of performing (N*).18 Now, Kripke thought of all cases of contingent a priori truths as similar to the meter case (M**), i.e., to a case in which the declarative act can be successful and a contingent fact is indeed created by the declaration. Donnellan, on the contrary, focuses only on the Neptune case and seems to think that all cases of contingent a priori truths are doomed to failure because the declaration cannot achieve the illocutionary point of making a contingent proposition true by means

17 Donnellan actually proposes to understand the stipulation made by Le Verrier in terms of something like (N**) in a passage already quoted in Chap. 3:

[B]ecause I think it somewhat illuminating to do it this way, I am going to propose instead that we think of the introduction as consisting of stipulating that a certain sentence shall express a contingent truth. If we want to introduce the name “N” by means of the description of “the φ” then the formula we would use would be: (a) Provided that the φ exists, let “N is the φ” express a contingent truth. It is a condition on the stipulation that the φ exists and should it turn out that it does not, the stipulation, we might say, has been an unhappy one and not to be taken as being in effect. (1977, p. 19) In the last sentence of the passage, he seems to come very close to understanding the stipulation as a special illocutionary act (although, unfortunately, he does not fully appreciate the consequences of doing so). 18 Kripke seems to think that Le Verrier can succeed in performing (N**) (if some preparatory conditions are in place), while Donnellan thinks that he must fail.

192

10 Stipulations as Performatives

of the illocution. The perspective that I am proposing suggests that there is an asymmetry between the two cases and, hence, both Kripke and Donnellan are partially right and partially wrong. In one of the cases, the declaration does succeed in achieving the illocutionary point (provided that the one performing (M**) has the required institutional authority, etc.), it indeed creates a fact, and there is a priori knowledge of this fact. In the other case, (N**) fails to achieve the point (which would be to make true that Neptune is the planet that causes the perturbations in Uranus’ orbits by means of the very utterance), although the related declaration (N*) (which is presupposed by (N**)) does achieves its illocutionary point and creates a fact (i.e., the metalinguistic fact that a particular planet has the name ‘Neptune’). We can say that the first declaration (M**) can be successful, but the second ((N**)) cannot, and this is not due to semantics, but to the limitations concerning what we can produce linguistically. And this is so despite the fact that (M*) and (N*) can both be successful.

10.5 Closing the Two Gaps As said before, Kripke’s discussion concerning contingent a priori truths leaves out two important questions. First, the question regarding the nature of their truthmakers. And, second, of how there can be transmission from the speaker to the hearer (or to the speaker herself at a different time) of a priori knowledge. What follows is, admittedly, only a sketch of a solution (based on what was said in the previous sections), since a lot more remains to be said about institutional facts as objects of cognition and illocutionary commitments. But it indicates that the perspective here explored is promising. With respect to the first gap, if we accept the claim that Kripke’s examples of contingent a priori truths must involve a hidden performative (or require the appropriate illocutionary force that could be made explicit by a performative), we have an interesting way of explaining how these truths are generated and known by exploring the approach to performatives as declaratives. As we saw, a declarative is a special kind of illocutionary act because, if successful, its propositional content is made true by the act itself. In this sense, the act creates the state of affairs represented in the propositional content, thereby making the latter true by the very utterance. These states of affairs are neither natural nor necessary, but are contingent and generated by the successful utterance. Therefore, they seem to belong to the category that we could call (following Searle) institutional facts, as opposed to brute or natural facts. Typical examples of institutional facts are contracts, nominations, laws, etc.: they depend on a declaration that makes true the propositional content and creates a fact not reducible to the utterance itself. Hence, we have an explanation of the truth-makers of Kripke’s examples in terms of institutional facts (broadly conceived). As we saw, some authors (e.g., Evans, 1979b and Hawthorne, 2002) claim that all contingent a priori truths (at least those considered by Kripke) are only superficially

10.5 Closing the Two Gaps

193

contingent because there is a semantic guarantee of their truth. And some (e.g., Evans, 1979b and Soames, 2005) argue that the existence of contingent a priori truths is part of the broader phenomenon of indexicality, which generates sentences like I am here now The sky is blue if and only if actually the sky is blue which must be true in any context of use due to the character of the pure indexicals ‘I’, ‘here’, ‘now’ and ‘actually’. But, according to the approach here proposed, both claims are incorrect: although the truth-makers of contingent a priori truths are generated by the use of language, there is no absolute semantic guarantee that they will be generated since there is no guarantee that the performative will be successful (instead of a “misfire” in Austin’s terminology). Moreover, the cases here considered are an effect of performativity rather than indexicality. With respect to the second gap, we might remark that placing Kripke’s examples under the perspective of declarative illocutionary acts may give us an interesting way of connecting the knowledge of the baptizer at time t0 both with his or her knowledge and with the knowledge of other people at a later time. For there are internal connections between illocutionary acts, so that one act may commit the speaker (or the interlocutor, if he or she recognizes the speaker’s authority) to another. Or, better said, the conditions of success of one illocutionary act might involve the successful performance of another illocutionary act. So, we have a variety of implications that are not via the propositional contents, but via connections among illocutionary acts and their conditions of success. Austin already suggests something along these lines when he remarks that some illocutionary acts invite subsequent effects for their successful achievement. In that sense, we have a kind of implication between illocutionary acts, and this might provide a clue to the understanding of the transference of a priori knowledge from speaker to speaker or from one speaker at t0 to that same speaker at a later time. That is to say, if someone defines X as Y and in a future occasion does not follow this same definition, this implies that the definition was not successful as an illocutionary act. A successful definition (which is a declarative illocutionary act) requires the speaker to engage in some assertions and not others at later times.19

19 In

some relevant passages in which he tries to isolate special features of illocutionary acts (as opposed to perlocutionary acts), Austin mentions the connection between some such acts and subsequent acts: The illocutionary act ‘takes effect’ in certain ways, as distinguished from producing consequences in the sense of bringing about states of affairs in the ‘normal’ way, i.e., changes in the natural course of events. Thus ‘I name this ship the Queen Elizabeth’ has the effect of naming or christening the ship; then certain subsequent acts such as referring to it as the Generalissimo Stalin will be out of order. (1962, p. 116)

194

10 Stipulations as Performatives

Now, taking the clue from Austin and from Searle and Vanderveken’s notion of illocutionary commitment (see last chapter), the thesis here embraced is that (M**) and (N**)20 at t0 involve declarative illocutionary acts which include, as part of their felicity condition, other illocutionary acts at later times: they commit the speaker to assertions at ti (for 0  i) so that if at t0 the speaker has a priori knowledge of the propositional content of (M**) or (N**) (meaning by this that the speaker’s justification for believing the propositional content is based only on the successful illocutions of (M**) and (N**) themselves), that is only so if at later times the same speaker is committed to the assertion of that content, without appealing to anything else as ultimate justification but only to the early declaration. This would offer a model of how a priori knowledge of a fact created by a declaration at a certain time continues to be a priori knowledge at later times, i.e., the fact that the speaker has to rely on memories at later times does not make that kind of knowledge a posteriori. Something similar applies if we are considering not just one speaker at different times, but two speakers, the one who performs the declarative utterance (the baptizer or stipulator, in Kripke’s perspective) and one or more interlocutors (who are present at the baptism or who rely on testimony about it). A declaration, especially if it involves some form of authority on the speaker’s part, might be successful only if it commits the interlocutors to do (or say) something so that, if they fail to do so, the declaration was unsuccessful. In that sense, the declarations performed with (M**) and (N**) require, for their success, that other speakers assert the propositional content at later times and have as main justification for their assertions only the original declarations. Another element that might help to explain how other people (besides the baptizer) can also have a priori knowledge concerning the length of the stick is Searle’s notion of collective intentionality (as explained in Sect. 9.7). As we saw, an institutional fact requires imposing a status function on some entity (a function that is not derived only from the physical properties of that entity) through collective intentionality. The attribution of the status function of unit of measurement to a particular stick might be just an individual act if the stick is employed only by the baptizer in his or her private activities that somehow involve measuring things. But if the stick S is to serve as the basis of a measurement system for an entire community, and if that involves the collective imposition of a status function, then although there might be one official stipulator in charge (the authority producing and presenting S as the standard of measurement or, in the case of the kilogram, the GCWM), the content of the stipulation is known by all members of this community not as a piece

[T]he performative ‘I define X as Y’ (in the fiat sense say) commits me to using those terms in special ways in future discourse, and we can see how this is connected with such acts as promising. (ibid., p. 136) 20 As we saw, things are less straightforward in the case of (N**) because there are doubts regarding the possibility that the act can ever achieve its illocutionary point. But the point I am trying to make here holds of (N**) anyway in the hypothetical (and somewhat fictitious) situation that the illocutionary point could be achieved. (It certainly holds of (N*).)

10.5 Closing the Two Gaps

195

of empirical information, but as the result of the collective imposition of a status function through the stipulator’s declaration. In the same way that we say that the Congress as a whole passed a law even if passing the law requires that one particular individual, e.g., the chair of the Congress, officially declares that law valid at the end of a section. Following Searle, we can say that each member of the Congress, at each stage of the meeting, has a particular individual intention, but this intention can be understood only by reference to a collective intention of the Congress. In the meter stick case, although there is one baptizer, since the purpose of the act of stipulation is not just creating a private use of the stick but rather creating the foundation of the entire practice of measuring in science and industry, the individual stipulation might be seen as part of a collective act. We might also have a better understanding of the persistence through time of a priori knowledge related to the length of the meter stick by considering how institutional facts persist through collective intentionality. Taking (again) the clue from Searle, The secret of understanding the continued existence of institutional facts is simply that the individuals directly involved and a sufficient number of members of the relevant community must continue to recognize and accept the existence of such facts. (1995, p. 117)

As long as the members of a community continue to impose a status function through collective intentionality to some entity, the corresponding institutional fact continues to exist and, hence, contingent a priori knowledge of those facts can persist. Another aspect of Kripke’s examples that can be illuminated (but perhaps not in the way that Kripke himself would have thought) is the role of empirical information involved in getting to know the truth of the propositional content of (M**) and (N**). As I said before, some scholars (e.g., Plantinga, 1974; Horowitz, 1983; Salmon, 1986, and Soames, 2005) claim that at least some minimal empirical information is required on the baptizer’s part, e.g., that he or she really sees a stick that will be taken as the meter standard. Thereby, there would always be some a posteriori element behind the knowledge of the length of the meter stick. Similar concerns have been raised about Neptune:21 some knowledge is required to now that the name ‘Neptune’ is not empty, thereby destroying the credentials of the propositional content of (N**) as a priori. But, if we take seriously that (M**) and (N**) are introduced by special illocutionary acts that create a non-linguistic fact F , we can explore the distinction between knowing F (which might be granted simply by knowing that the act took place successfully) and knowing the preparatory conditions of F (which might require some empirical knowledge). Knowledge of the surrounding conditions both of the meter stick and of the success of Le Verrier’s baptism is knowledge of the preparatory conditions, and it is certainly empirical (a posteriori). But once the preparatory conditions are in place, there is automatic knowledge of the truth of

21 E.g., Carter (1976). For discussion of the “existential complaint” against Kripke’s Neptune case,

see Cowles (1994) and Ray (1994).

196

10 Stipulations as Performatives

the propositional content because these are made true by the very utterances of (M*) and (N*) (and, consequently, of (M**) and (N**)) under appropriate circumstances.

10.6 Some Partial Conclusions My primary aim in this chapter was to defend the plausibility of Kripke’s thesis that there are contingent a priori truths. But I have done so in a way that Kripke would probably not endorse. The strategy depends on what I take to be an obvious fact: Kripke’s cases like the ones expressed by (M) and (N) only make sense if they are introduced by performative utterances such as (M*) and (N*). Moreover, (M*) and (N*) might have (M**) and (N**) as illocutionary consequences in the sense that a performance of the former commits the speaker to a performance of the latter (at least in principle). We also saw that here there may be a difference between the two cases: while (M**) can be successful, it is doubtful that (N**) can (for human beings at least) because its propositional content is independently true, and represents something that we, human beings, cannot make true by means of illocutions. Most philosophers who worked on the issue overlooked this difference. Kripke seems to think that all cases can be successful like the meter case, while Donnellan thinks that all cases are doomed to failure such as the Neptune case. Under the perspective of how performatives work and of what can be achieved by means of performative utterances, both have a point but both are wrong in generalizing it. If this is correct, however, we have a natural explanation of the kind of contingent a priori knowledge generated (at least in the easier case of (M**)): it is contingent because the propositional content of declarative illocutionary acts must be contingent, and it is a priori because no other evidence is necessary to know the propositional content other than the successful declaration itself.22 This perspective helps us close the two gaps in Kripke’s original discussion: first we may account for the kind of truth-makers of contingent a priori sentences: these are institutional facts broadly conceived. Second, we may account for the transmissibility of contingent a priori knowledge from speaker to hearer (or from speaker to speaker at different

22 The position here defended is different from the one criticized by Quine in “Carnap on Logical truth” (1954). (Incidentally, Quine wrote his criticism before Austin’s theory of performatives and the development of contemporary speech act theory. Carnap was also not thinking of anything like the illocutionary force behind meaning postulates as generator of logical truths.) The point is not concerning logical truths created by linguistic convention (a conception that Quine takes to be quite void and uninformative), but rather the implementation of contingent truths (e.g., that a meeting is adjourned or that someone is nominated my legal representative). This can be achieved by performing declarative illocutionary acts (of which performatives are a particular class according to Searle and Vanderveken’s approach). It is a distinguished feature of such acts that the truth of the propositional content is first made true by the very act itself; hence, the propositional content cannot be (or become, in Quine’s sense) a logical truth, since a logical truth must be true independently of any illocutionary act.

References

197

times): the connection goes through relations of illocutionary commitment in the sense that declarations require, for their success, further acts on the speaker’s and the hearer’s part. Kripke’s point in Naming and Necessity is originally presented (e.g., on p. 54) as a debate with Wittgenstein, who claims that sentences like ‘the length of the stick at t0 is one meter’ are neither true nor false, and this is so, as he argues, not because the stick has any extraordinary properties, but because it has a special role in the language game of measuring in the metric system. Kripke, on the other hand, claims that this sentence is true. If the hypothesis of this book is correct, however, both are partially correct, but for the wrong reasons. (M) and (N), conceived as sentences in abstraction from any illocutionary act, cannot be true (and, depending on how we understand the proposition expressed, maybe cannot be false either). They must first be seen as (M*) or (N*) (so, Kripke is not entirely correct) and, derivatively, engender (M**) and, perhaps, (N**). But now, in order to explain the special status of the sentences expressing the propositional part of (M**) (and (N**) if it were successful), we do not need to employ the conceptual apparatus of language games, but only to appeal to the special semantic features of performatives (and, hence, Wittgenstein is not entirely correct either). Acknowledgements I thank the editors of Synthese for the kind permission to use some printed material in this chapter.

References Austin, J. (1962). How to do things with words. Oxford: Clarendon Press. Belnap, N. (1962). Tonk, Plonk and Plink. Analysis, 22(6), 130–134. Biro, J. (1995). Testimony and “a priori” knowledge. Philosophical Issues, 6, 301–310. Burge, T. (1993). Content preservation. The Philosophical Review, 102(4), 457–488. Carter, W. (1976). On a priori contingent truths. Analysis, 36(2), 105–106. Chisholm, R. (1989). Truths of reason. In R. Chisholm (Ed.), Theory of knowledge. Englewood Cliffs, NJ: Prentice-Hall. Cowles, D. (1994). The contingent a priori: An example free of existential worries. Philosophical Studies, 74, 137–141. Donnellan, K. (1977). The contingent a priori and rigid designators. Midwest Studies in Philosophy, 2(1), 12–27. Evans, G. (1979b). Reference and contingency. The Monist, 62(2), 161–189. Hawthorne, J. (2002). Deeply contingent a priori knowledge. Philosophy and Phenomenological Research, 65(2), 247–269. Horowitz, T. (1983). Stipulation and epistemological privilege. Philosophical Studies: An International Journal for Philosophy in the Analytic Tradition, 44(3), 305–318. Jeshion, R. (2002). Acquaintanceless De Re belief. In J. Campbell, M. O’Rourke, & D. Shier (Eds.), Meaning and truth. Investigations in philosophical semantics (pp. 53–78). Oxford: Oxford University Press. Kant, I. (1787). Kritik der reinen Vernunft, Edited and Translated by Guyer, P, and Wood, A. Critique of pure reason. Cambridge: Cambridge University Press, 1998., Riga: Verlag Johann Friedrich Hartknoch. Kripke, S. (1980). Naming and necessity, Cambridge, MA.: Harvard University Press.

198

10 Stipulations as Performatives

Kripke, S. (1986). Rigid designation and the contingent a priori: The meter stick revisited. Exxon distinguished lectures at Notre Dame University. Unpublished. Malmgren, A. (2006). Is there a priori knowledge by testimony? The Philosophical Review, 115(2), 199–241. Plantinga, A. (1974). The nature of necessity. New York: Oxford University Press. Quine, W. (1954). Carnap and logical truth. In The ways of paradox and other essays (pp. 00–126). Random House, 1966. Ray, G. (1994). Kripke and the existential complaint. Philosophical Studies, 74, 121–135. Searle, J. (1969). Speech acts: An essay in the philosophy of language. Cambridge: Cambridge University Press. Searle, J. (1985). Expression and meaning: Studies in the theory of speech acts. Cambridge: Cambridge University Press. Searle, J. (1989). How performatives work. Linguistics and Philosophy, 12(5), 535–558. Searle, J., & Vanderveken, D. (1985). Foundations of illocutionary logic. Cambridge: Cambridge University Press. Salmon, N. (1986). Frege’s puzzle. Atascadero, CA: Ridgeview Publishing Company. Soames, S. (2005). Reference and description: The case against two-dimensionalism. Princeton University Press. Vanderveken, D. (1990a). Meaning and speech acts, Volume 1: Principles of language use. Cambridge: Cambridge University Press.

Chapter 11

One Ancestor: The Early Frege on Definitions

How un-Fregean is Kripke’s conception of contingent a priori truths and, in particular, under the perspective proposed in this book? The standard interpretation is that it goes against a long established tradition that includes Frege (among many other classic philosophers). It indeed might look strongly un-Fregean if we compare with what became Frege’s standard doctrine on sense, reference and identities after “On Sense and Reference” (1892). But less so if we look at the way Frege deals with definitions in his first work on logic and philosophy of language, i.e., the Begriffsschrift (1879), which is, in some respects, quite different from his later approach. This chapter is meant to be a historical digression about some ideas put forward by Frege in the design of his formal language and that anticipate, in a way, some aspects of the perspective that we are exploring in this book. He saw a special illocutionary force as “turning” synthetic into analytic truths within his formal system, and these truths serve as platforms for the discovery of other analytic truths. This is a bit surprising since Frege’s thought is normally seen as a paradigm of Platonic realism in mathematics and logic, a perspective according to which mathematical and logical realities are objective and completely independent of our thought or linguistic practice. But he is also perhaps the first great logician to call attention to the need for distinct kinds of illocutionary acts (and to the necessity of illocutionary force indicators) within the formal language of logic.

11.1 Assertion and the Assertion Sign In the Beggriffsscrift (1879), Frege introduces a famous distinction between two elements of any judgement, namely, the propositional content or, as he calls it, judgeable content (“beurteilbare Inhalt”), which can, in principle, be simply considered without being asserted, and the assertive force properly speaking. He © Springer Nature Switzerland AG 2022 M. Ruffino, Contingent A Priori Truths, Synthese Library 443, https://doi.org/10.1007/978-3-030-86622-8_11

199

200

11 One Ancestor

also introduces special symbols to indicate this distinction within the formal system. The symbol ‘ ’, which precedes every propositional content, is a composition of two simpler symbols, the content (horizontal) stroke ‘ ’ and the assertion (vertical) stroke ‘ ’. In Frege’s original notation, the horizontal stroke can only precede a judgeable content,1 and the vertical one indicates the assertive act properly speaking. It is interesting to notice that, although Frege is normally considered as having adopted a logical language that leaves out, as much as possible, pragmatic elements (such as presuppositions, implicatures and anything that suggests subjective elements), he is also the first to advocate the inclusion of a symbol in this same language which is, as we would now call it, an illocutionary force indicator. Frege thought it necessary to introduce such a symbol in order to properly distinguish between asserting a thought and just entertaining it (e.g., presenting a thought that one knows to be false as the antecedent of a conditional without asserting it). In natural language, the usual indicator of assertion is the indicative mood of the verb, but this has no place in a logical language. He saw the confusion between just presenting and asserting a thought as pervasive in the formal languages of some of his contemporary logicians. The issue comes up, e.g., when he compares his Begriffsschrift with Peano’s formal language: Herr Peano has no such [assertion] sign: he, on the contrary, uses his relation signs now with and now without assertoric force, and in fact the principal relation sign invariably carries assertoric force. From this it follows that for Herr Peano it is impossible to write down a sentence which does not occur as part of another sentence, without putting it forward as true. (1896, p. 12)

Something similar is said about Schröder’s formalism in Frege’s notes on an article by Philip Jourdain that contains a section explaining Frege’s ideas on logic and foundations of arithmetic: On Schröder’s proposals, I make the following remarks. The asserting force, which with my notation is in the line of judgement, must here be found in the sign of equality. Hence it follows that theorems like ‘f (a) = 1’,‘f1 (a)=1’ may never occur without being asserted, and therefore never as conditional theorems or deductions, and this is a great defect. (1912, p. 193)

One should be able, as he says in the posthumous notes “A Brief Survey of my Logical Doctrines” (from 1906), to disentangle these two things: We need to be able to express a thought without putting it forward as true. (1976b, p. 198)

It is important to stress that the illocutionary force indicator is, for him, not part of the metalanguage, but of the object language of logic and, as the passages above show, an essential part, and not just a detachable ornament.2

1 In

his later system (1893), the horizontal stroke can be placed in front of any singular term, not just a judgeable content. 2 As Vanderveken notices, Modern logicians and philosophers of language did not recognize fully the philosophical importance of Frege’s idea of admitting illocutionary force markers in the ideal object-

11.2 Definitions and the Definition Sign

201

11.2 Definitions and the Definition Sign The assertion sign is not the only illocutionary force indicator that Frege incorporates in the formal system of the Begriffsschrift. In §24, when the first definition is introduced (formula 69), he explains that, in cases like this, we do not have an assertion properly speaking because a new piece of vocabulary is being introduced: This sentence [69] is different from those considered previously since symbols occur in it which have not been defined before; it itself gives the definition. It does not say, “The right side of the equation has the same content as the left side.”; but, “They are to have the same content.” This sentence is, therefore, not a judgement [. . . ] Although originally (69) is not a judgement, still it is readily converted into one; for once the meaning of the new symbol is specified, it remains fixed from then on; and therefore formula (69) holds also as a judgement, but as an analytic one, since we can only get out what was put into the new symbols [in the first place]. This dual role of the formula is indicated by the doubling of the judgement stroke. Thus, with respect to the derivation which follows, (69) can be treated as an ordinary judgement.

The act of the definition is not the assertion of a previously existing content. On the other hand, it stipulates the truth of a content. So, only after the definition takes place one can assert the content (and it is analytically true, as the passage says). This is the reason why Frege finds it necessary to introduce the new symbol ‘ ’ indicating something different from assertion. Although Frege does not call it so, we can recognize in this symbol another illocutionary force indicator. Like the assertion sign, it is not just part of the metalanguage, but of the object language itself. From that point in the text on, all definitions in Begriffsschrift and in Frege’s later system of Grundgesetze der Arithmetik (1893) are preceded by this symbol; all other formulas that are not definitions are preceded by the assertion-sign.3 Since Frege himself recognizes that definitions (at least the way he understands them) are not originally assertions, we might ask which kind of speech acts they are. And here, again, we can make use of the taxonomy of Searle and Vanderveken reviewed in Chap. 9. To remember, there are, according to this taxonomy, five big categories of speech acts, namely: assertives, expressives, comissives, directives, and declaratives. Definitions are not assertive acts, as Frege indicates, so we have four remaining categories. We can, from the outset, say that they are not expressive acts either, for a definition is not the expression of a speaker’s feelings or reactions

languages of logic. Rapidly Frege’s assertion sign was eliminated from the object-languages of logic, and became a meta-linguistic sign used only to identify the sentences of objectlanguages which are provable in axiomatic systems. I believe that this failure to recognize the indispensability of illocutionary force markers in language is responsible for the failure of contemporary logical semantics to interpret adequately performative and non-declarative sentences. (1990a, p. 68) 3‘

’and ‘ ’ are the only two illocutionary force indicators that Frege explicitly recognizes in his system. There are, however, other such indicators in his system. See Ruffino et al. (2020) for details.

202

11 One Ancestor

towards a propositional content taken as true. First, because the speaker’s feelings or reactions would be, in Frege’s eyes, completely irrelevant to definitions and, second, because it is doubtful that there is a true content prior to the definition in the first place. It would be odd to classify definitions as commissive acts (such as promises), because such acts only make sense with a propositional content that involves an action on the speaker’s part that takes place in a future time. But contents in Frege’s logical system are not actions, and even less located in time (unless one wants to read a definition as a promise to use a newly introduced term in such and such a way, but this seems out of place when we are dealing with a formal language). Another reason is that a successful commissive speech act does not automatically make its propositional content true (e.g., I can successfully make a promise that P without thereby P becoming true). But a successful definition always has a true propositional content (or at least Frege thinks so). For the same reasons, it seems inadequate to classify a definition as a directive speech act (such as an order or invitation to use a term in the specified way). The only remaining alternative is to classify definitions as declarative speech acts, i.e., as those acts that bring about the truth of a propositional content by means of the mere utterance. (As said before, this might require some preparatory conditions and, in some cases, some form of authority on the speaker’s part.) This classification fits nicely the usual way of introducing definitions since these normally take the form of performative utterances (such as ‘I (we) define’, ‘I (we) stipulate’, ‘I (we) call’, etc.), and performative utterances are, according to the theoretical apparatus that we are assuming, declarative speech acts. The double assertion-sign then indicates that something different from an assertion is being made with a propositional (or conceptual) content. Let us give a closer look to what goes on inside the definition (i.e., in the structure of the appropriate propositional content) according to Frege.

11.3 Identities as Synthetic Judgements All definitions introduced in Frege’s Begriffsschrift have as content (i.e., what follows the symbol ‘ ’) statements of the form ‘A ≡ B’, where ‘≡’ is the symbol introduced in §8 for a relation that he calls identity of content. So, in order to understand how definitions work in the Begriffsschrift, we should take a closer look at Frege’s treatment of identity of content, which turns out to be something quite different from identity as it is classically understood. As Perry (2019, especially in Chapter 3) reminds us in his recent study of Frege’s Begriffsschrift, although later in his writings Frege famously refers back to his view on identity in the Begriffsschrift (e.g., 1892), the symbol ‘≡’ in fact does not represent identity properly speaking (i.e., the content of the symbol ‘=’, which occurs only twice in the Begriffsschrift, and only in marginal examples of mathematical assertions formulated in an informal

11.3 Identities as Synthetic Judgements

203

way).4 Identity, as it is classically understood, holds between objects (or contents); but this is not how Frege understands the relation of identity of contents. He introduces the symbol for identity of content in the following passage: Identity of content differs from conditionality and negation by relating to names, not to contents. Although symbols are usually only representatives of their contents—so that each combination [of symbols usually] expresses only a relation between their contents—they at once appear in propria persona as soon as they are combined by the symbol of identity of content, for this signifies the circumstance that the two names have the same content. (1879, §8)

Just a few lines later comes an example taken from geometry in which we have two points A and B of a circumference determined by a diametrical line, and then allow the line to rotate around A (i.e., A is fixed but B will change as we rotate the line). When the line is perpendicular to the original diametrical line (i.e., when the rotating line becomes the tangent line touching the circumference in A), the points A and B must coincide, but we have the same point determined in two different ways (i.e., as the fixed point A and as the rotating point B of intersection between the line and the circumference). Concerning this example, Frege says: Thus the need of a symbol for identity of content rests upon the following fact: the same content can be fully determined in different ways [. . . ] It follows from this that different names for the same content are not always merely an indifferent matter of form; but rather, if they are associated with different modes of determination, they concern the very heart of the matter. In this case, the judgement as to identity of content is, in Kant’s sense synthetic. A more superficial reason for the introduction of a symbol for identity of content is that it is occasionally convenient to introduce an abbreviation for a lengthy expression. We must then express identity of content between the abbreviation and the original form. Now, let (A ≡ B) mean: the symbol A and the symbol B have the same conceptual content, so that we can always replace A by B and vice-versa. (ibid.)

At least three things emerge from these passages that are of great importance to understand Frege’s idea of identity of content and, indirectly, his conception of definitions in the Begriffsschrift: (1) He understands a statement of identity of content as strictly metalinguistic, i.e., as being about the names flanking the identity sign, where these names stand for themselves (instead of standing for their content, as it happens in normal combinations of symbols). (2) Frege qualifies the geometrical example of identity of content as synthetic in Kant’s sense. Hence, he has Kant’s conceptual framework in the background.

4 The

symbol ‘=’ also occurs in Frege’s “On The Aim of The Conceptual Notation” (1883), which was a lecture delivered few years later and in which he explains some applications of his formalism. But in the examples presented, ‘=’ means ordinary identity between numbers, and not identity of content.

204

11 One Ancestor

As we know, for Kant, such judgements can be either a priori or a posteriori. The latter is true in virtue of an empirical fact, while the former has a different source of its truth (we do not need to go deeper into Kant’s view here). Frege might have meant that identities of content such as ‘A ≡ B’ are synthetic a posteriori in the sense that it is ultimately a matter of decision whether a particular sign stands for a particular object, as we could perhaps say that ‘2 + 2 ≡ 4’ (which, we must notice, is not the same as ‘2 + 2 = 4’) is synthetic a posteriori in the sense that it is ultimately an empirical matter that the symbol ‘2’ has the number two as content, that ‘+’ stands for the operation of addition, and that the symbol ‘4’ has the number four as content. But in the same section, he adds the following comment about the geometrical example: What point corresponds to the position of the straight line when it is perpendicular to the diameter? The answer will be: The point A. Thus, in this case, the name B has the same content as the name A; and yet we could not have used only one name from the beginning since the justification for doing so is first given by our answer.

That is to say, Frege seems to regard the epistemic status of a statement of identity of content as dependent on its justification; in this particular example, the justification comes from a law of geometry, and therefore ‘A ≡ B’, although expressing a metalinguistic relation, is synthetic a priori. If this is correct, then in cases of non-mathematical content identity statements (such as ‘Hesperus ≡ Phosphorus’) we apparently would have a synthetic judgement in Kant’s sense, but this time an a posteriori one because there is no synthetic a priori law connecting the content of ‘Hesperus’ and of ‘Phosphorus’. Frege is not explicit about this, but this seems to be the most plausible interpretation of what he says. (3) Frege contrasts the geometrical example (in which there is a geometrical law as justification) with cases in which there is only “a more superficial reason” for the symbol of identity of content, and this is just the convenience of an abbreviation of a lengthy expression, without any a priori law backing them as in the geometrical case. Therefore, such judgements seem to be synthetic a posteriori for him, unless we have the same symbol flanking the identity of content sign (i.e., something of the form ‘A ≡ A’). But a curious thing is that there is no empirical law backing these statements either, since here the matter is but one of practicality of arbitrary conventions regarding the content of those names. Hence, in Frege’s view, ordinary identities of content have a synthetic content, this content is metalinguistic, and it can be the object of a judgement (i.e., it can be asserted). But with definitions, things are a little more complicated, as the passage from §24 quoted above suggests. Because a new symbol that was not previously defined (and that is not a primitive either) occurs in it. Let us consider the general form of a definition as Frege sees it: (A ≡ B)

11.3 Identities as Synthetic Judgements

205

What follows the definition sign ‘ ’ is a formula expressing identity of content but in which a new symbol B appears. According to the explanation of ‘≡’ given in §8, ‘A ≡ B’ means that the symbol A and the symbol B have the same content, but before the definition is made B has no content at all since the point of the definition is to introduce it and fix its content in the first place. What should we then say about ‘A ≡ B’ before it becomes a definition? Is it true or false? Or is it a truth-valueless expression? There are two ways in which we can understand this. First, we may consider ‘A ≡ B’ as just a claim that the two symbols have the same content, and that is true or false regardless of the fact that one of them (B) does not belong to the language in which the sentence is formulated (and, hence, does not have a content yet). In the same way that we could  take something that is not a name (or even an expression) in English such as ‘ ’ and ask whether Donald Trump ≡



 is true or false (we would be asking whether ‘Donald Trump’ and ‘ ’ have the same content and the answer is, of course, that they do not and, hence, the sentence is false). The impression that we might have that the sentence is syntactically defective would then be illusory, since any symbol flanking ‘≡’ is the name of itself, hence, any sentence of identity of content is meaningful. Under this understanding, the definition would transform a false metalinguistic content into a true one by means of a stipulation. Alternatively, we may consider that if a symbol does not have a content, or if it is not previously recognized as belonging to the language in which the identity statement is made, then it has no proper content (or, alternatively, it is not properly a symbol yet), and the statement, if literally understood, is neither true nor false. E.g., since ‘ ’ is not an English term, then the above identity, literally understood, is neither true nor false, and the identity seems to express no content at all. If one adopts this view, it seems more appropriate to understand the content not as the literal one, but as the one corresponding to the claims that  – ‘ ’ is a symbol and – it has the same content as ‘Donald Trump’. The conjunction would be synthetic a posteriori in Kant’s terms, and false under this interpretation (since both conjuncts are false before the definition is made), but it would be a complete content that is made true by the definition in any case. The first interpretation seems more compatible with Frege’s view in §8 that the symbols flanking the identity of content represent nothing but themselves and, hence, that the identity is completely about the symbols, whether or not they have a content and whether or not they belong to the same language. But both interpretations imply that the content on which the definition operates is a false synthetic a posteriori metalinguistic content.

206

11 One Ancestor

Frege says in §24 that formula 69 is not a judgement, but not that its content is not a judgeable content; from the perspective that we are here considering, the content is indeed a judgeable content about the symbols flanking the identity of content, and what it says is originally false because it is false that both symbols have the same content (since one of them has no content at all). But as soon as the definition is made, as he also remarks, the content becomes true and also analytic. In other words, if we were to take the original content as the content of a judgement, we would have to say that it is false, because it is false that both symbols have the same content, or false that one of them is really a symbol. But, as Frege says, a definition is not a judgement, and what it does is to turn the false content into a true one.

11.4 Definitions: Turning the Synthetic Propositions Analytic The content of a definition is always a formula stating some identity of content. As we saw in the last section, Frege classifies in §8 of Begriffsschrift the example coming from geometry as synthetic, and he most likely has synthetic a priori in mind because its justification appeals to a law of geometry. Therefore, the justification is the key as to the status as synthetic or analytic of an identity of content statement. If an identity of content is true in virtue of a logical law, it must then be an analytic truth. If, on the other hand, we need to appeal to empirical information to justify the identity of content (such as ‘Hesperus ≡ Phosphorus’), then it is synthetic a posteriori. It might as well be false if the two symbols do not have the same content (which subsumes the situation in which one of them has no content at all). This is in sharp contrast to ordinary identities such as ‘A=B’ for, here, if A or B lacks a content, then the identity is neither true nor false since no proposition is expressed. Hence, we must conclude that definitions have a synthetic content, for if there were a logical law connecting the content of both sides flanking ‘≡’, the definition would be pointless. However, in the passage quoted from §24 Frege says that, immediately after the definition, “formula (69) holds also as a judgement, but as an analytic one, since we can only get out what was put into the new symbols”. It then seems that, in the Begriffsschrift, Frege holds a view of definitions according to which they turn a false synthetic content into an analytic truth by the special act involved (which, as he remarks, is not the act of assertion). The propositional content of a definition is not something that is true “by nature” (or a logical truth), but is made so by the very act of the definition. In other realms outside logic or mathematics this act might require, as a preparatory condition, that the speaker has some kind of authority, e.g., in issuing a legal norm. But in logic and mathematics no special authority is required, except perhaps that of being a working logician or a mathematician. And right after the stipulation, one might take it as an analytic truth which is instrumental in the proof of other analytic truths that are not themselves the result of stipulation but of discovery. In this sense, there seems to be an interesting parallel between Frege’s view of definitions and Searle and Vanderveken’s view of declarative acts: a contingent

11.4 Definitions: Turning the Synthetic Propositions Analytic

207

content is made true by a linguistic act, and the content made true can serve as a basis for building the rest of the theory. The fact that the content of a definition is made true by an act does not mean that no further justification is necessary, at least for some definitions.5 One might need to justify the definition in terms of the methodological advantages brought by its introduction, e.g., fruitfulness (i.e., whether it yields interesting new results), material adequacy (i.e., whether it correctly captures an intuitive pre-theoretical notion), practicality, etc. But this further justification has nothing to do with the truth of its content properly speaking, since this is an almost inevitable result of the definition in an appropriate way. The fact is that the content is always made true if the definition is successful. We can conclude that, despite appearances, there is something resembling contingent a priori truths in Frege’s early thought, and this comes from his view on definitions as requiring an act that is not an assertion.6 In his early writings Frege distinguishes two sorts of definitions. On the one side, there are those that are mere abbreviations, introduced just for the sake of brevity. On the other side, there are those that are “fruitful” (as he calls them) in the sense that they lead to the proof of non-trivial results and, hence, as he says in comparing his own definitions with Boole’s (in a text from 1881), [W]hat we may discover in them has a far higher claim on our attention than anything that our everyday trains of thought might offer. For fruitfulness is the acid test of concepts, and scientific workshops the true field of study for logic. (1976b, p. 33)

Fruitful concepts result not just from ordinary abbreviation of expressions, but from the partition of propositional (or judgeable) contents in non-trivial ways.7 We do not need a deeper look at the notion of fruitful definitions in Frege at this point; what interests us now is his opinion regarding two distinct kinds of definitions. In his early writings, Frege claims that all definitions presented in his original system of the Begriffsschrift are fruitful, and these are definitions that are meant to capture intuitive concepts such the property F is hereditary in the sequence f and y follows x in the f-sequence. These are not just abbreviations, but attempts to capture some intuitive notions that are (or so Frege thought) on the basis of arithmetic. The latter kind of definition is justified not just by its practicality, but also by its fruitfulness.8

5 It does not mean that there are no preparatory conditions either. One such condition is that the defined symbol has to be new in the language, i.e., it cannot already have a content previously to the definition, and it cannot have been defined already in some other way. 6 As said, Frege later (1893) changes his mind concerning identity and the content of definitions. He goes back to the classic conception of identity and also employs the classic symbol ‘=’ instead of ‘≡’, so we have a different picture in this later formulation. But the spirit is similar: we have a proto-content (i.e., something that is not a propositional content) being transformed (for which he uses the same symbol ‘ ’), into an analytic truth by the act of definition. 7 For the mechanism of generation of fruitful concepts in Frege, see Schirn (1990), Ruffino (1991) and Dummett (1991, Chapter 4). 8 For a comparison of contentual with abbreviative definitions in Frege from an illocutionary point of view, see Ruffino et al. (2020).

208

11 One Ancestor

In his later writings, however, Frege changed his mind and held that, inside logical or mathematical theories, definitions only play the role of arbitrary abbreviations. The other aspects of a definition (its fruitfulness, material adequacy, etc.) belong to pre-theoretical work.9

11.5 Some Partial Conclusions As we saw, Frege is perhaps the first logician to advocate the introduction of illocutionary force indicators in the formal language of logic. He also recognizes that defining is not the same as asserting, and introduces a second, different illocutionary force indicator for these acts. As he sees it, some propositional contents are made true by the particular act involved in definitions, and these truths serve as basis, within his formal system, both for discovering and proving other truths (theorems). In this sense, we have an analogue, in Frege’s early treatment of definitions, of Kripke’s contingent a priori truths. His entire formal system can be seen as a combination of propositions that are asserted (i.e., axioms and theorems) and propositions that are defined (stipulated) to be true.10 But Frege is also considered as the paradigm of a realist philosopher about logic and mathematics (more precisely, in arithmetic), i.e., he holds the view that there is a realm of mathematical entities that are independent of our thought and practice, entities that are not created but merely discovered and described. How can such a realism be compatible with the claim that speech acts are central to a mathematical theory, to the point of “creating” some truths by means of definitions? There are different degrees in which one can claim that illocutionary acts interfere with mathematical reality. Under some (more radical) constructivist perspective, illocutionary acts create mathematical objects in the same way that they create institutions and social ontology (laws, contracts, authority, governments, etc.). Under this perspective, illocutionary acts, particularly declaratives acts, entirely create mathematical reality.11 This is certainly not Frege’s perspective. On the other hand, one can hold the view that speech acts do not create mathematical reality, but are nevertheless necessary to generate the appropriate background against which mathematical objects and relations are to be studied (or “observed”). This seems to be Frege’s perspective. Although the mathematical and logical reality is objective and exists independently of our knowledge and language, we create, by means of conceptual networks, forms of apprehension of this reality. Hardy

9 See, 10 He

e.g., Frege (1976b, pp. 210–11). says in a letter to Hilbert:

I should like to divide up the totality of mathematical propositions into definitions and all the remaining propositions (axioms, fundamental laws, theorems). (1976a, p. 36) 11 This

perspective is defended, e.g., in Cole (2008) and Cole (2013).

References

209

(1940) metaphorically compares the task of mathematicians with that of observing and describing distant mountains. For realists like Hardy and Frege, mathematical reality, like the mountains, is not created by us, but can only be observed from afar and described. Using the same metaphor, we could compare the role of definitions (regarded as creative illocutionary acts) with that of producing observation platforms and scaffoldings around these mountains, so that observation can take place. The platforms and scaffoldings are themselves not part of the mountains (which are the real objects of study), but are essential for the work of observing and describing them. Some platforms and scaffoldings are better than others in that they provide a better perspective. Something similar happens with definitions: they provide a perspective for discovering and describing a mathematical reality that exists independently. Some definitions are better than others (although they are all true). But there is no observation, comparison, ordering, inference, etc., from a completely neutral point of view: there always are some definitions, some choices of primitive entities, relations, etc. The distinction between brute and institutional facts can be helpful in this connection: brute facts are those that exist independently, while institutional facts are those created by illocutionary acts. Reality as a whole is composed by brute facts surrounded by institutional facts (in a broad sense), and these institutional facts are ways of dealing with (and studying) brute facts. This distinction can be applied to describe Frege’s realist perspective in mathematics in the following way: mathematical objects and their relations are brute mathematical facts which we try to understand and describe in doing mathematics, but they are surrounded by mathematical institutional facts created by illocutionary acts (definitions, postulates, etc.), without which mathematical reality as such cannot be studied.12

References Cole, J. (2008). Mathematical domains: Social constructs. In Proof and other dilemmas: Mathematics and philosophy (pp. 109–128). Cole, J. (2013). Towards an institutional account of the objectivity, necessity, and atemporality of mathematics. Philosophia Mathematica, 21(1), 9–36. Dummett, M. (1991). Frege’s philosophy of mathematics. Cambridge, MA: Harvard University Press. Frege, G. (1879). Begriffsschrift. Eine der arithmetischen nachgebildete Formelsprache des reinen Denkens. Reprinted and translated in Bynum, T. (Ed.), Frege. Conceptual notation and related articles (pp. 101–202). Oxford: Oxford University Press, 1972. Halle a. S.: Louis Nebert. Frege, G. (1883). Über den Zweck der Begriffsschrift. Reprinted and translated by Bynum, T. (Ed.), Conceptual notation and related articles (pp. 90–100). Oxford: Oxford University Press, 2013. Jena: Verlag von G. Fischer.

12 In this particular aspect, mathematics is not different from many other sciences that have brute facts as their primary object of study, but with a multitude of institutional facts (definitions, postulates, methodologies, etc.) mediating the research.

210

11 One Ancestor

Frege, G. (1892). Über Sinn und Bedeutung , Zeitschrift für Philosophie und philosophische Kritik, 100. Reprinted and translated by P. Geach, & M. Black. In M. Beaney (Ed.), The Frege reader (pp. 151–171, pp. 25–50). Oxford: Blackwell, 1997. Frege, G. (1893). Grundgesetze der Aritmetik. Reprinted and translated by P. Ebert, & M. Rossberg (Ed.), Basic laws of arithmetic. Oxford: Oxford University Press, 2013, Verlag Hermann Pohle. Frege, G. (1896). Über die Begrifffsschrift des Herrn Peano und meine eigene. Vortrag, gehalten in der ausserordenlichen Sitzungvom 6 Juli 1896, “Berichte über die Verhandlungen der KöniglichSächsischen Gesellschaft der Wissenschaften zu Leipzig. Mathematisch-Physische Klasse” 48, 1896, 48. Reprinted and translated by Dudman, H. In Australasian Journal of Philosophy, 47(1), 1969, pp. 1–14, pp. 361–378. Frege, G. (1912). Notes to Philip E. B. Jourdain, ‘The development of the theories of mathematical logic and the principles of mathematics’. The Quarterly Journal of Pure and Applied Mathematics, 43. Reprinted in P. Long, & R. White. Gottlob Frege. Philosophical and mathematical correspondence (pp. 179, pp. 237–69). Chicago: The University of Chicago Press, 1980. Frege, G. (1976a). Gottlob Frege. Wissenschaftlicher Briefwechsel, Edited by Gabriel, G., Hermes, H., Kambartel, F., Thiel, C., Veraart, A. Translated by Long, P., White, R. Gottlob Frege. Philosophical and Mathematical Correspondece, Chicago: The University of Chicago Press, 1980., Hamburg, Felix Meiner Verlag. Frege, G. (1976b), Nachgelassene Schriften, Edited by Hermes, H., Kambartel, F., and Kaulbach, F. Translated by Long, P, and White, R., Posthumous writings. Oxford: Basil Blackwell, 1979., Felix Meiner Verlag. Hardy, G. (1940). A mathematician’s apology. Cambridge: Cambridge University Press. Perry, J. (2019), Frege’s Detour: An essay on meaning, reference, and truth. New York: Oxford University Press. Ruffino, M. (1991). Context principle, fruitfulness of logic and the cognitive value of arithmetic in Frege. History and Philosophy of Logic, 12(2), 185–194. Ruffino, M., Venturi, G., & San Mauro, L. (2020). Speech acts in mathematics. Synthese, 198, 10063–10087. Schirn, M. (1990). Frege on the purpose and fruitfulness of definitions. Manuscrito, 13(1), 7–23. Vanderveken, D. (1990a). Meaning and speech acts, Volume 1: Principles of language use. Cambridge: Cambridge University Press.

Chapter 12

Global Conclusions: The Varieties of Contingent A Priori Truths

The primary aim of this book was, besides reviewing a multitude of different perspectives and strategies for dealing with the intriguing phenomenon of contingent a priori truths, to defend the plausibility of Kripke’s thesis that there are indeed such truths, and that they are not just artificial cases introduced only for the sake of philosophical discussion concerning aspects of singular reference and rigidity. But I have done so (mainly in Chap. 10) in my own terms and in a way that Kripke would probably not endorse. We could so synthesize the trajectory of this book up to Chap. 8. I present, along the way, reasons for seeing most of the treatments and objections proposed in the literature to deal with Kripke’s classical examples as, in one way or another, less than satisfying. I indicate that even Kripke’s own formulation leaves two important gaps open: we are left in the dark concerning the nature of the known facts in these examples, and there is no clear prospect of contingent a priori knowledge for anyone other than the stipulator and for any time other than the exact moment of the stipulation. This seems much less than we would have expected for such an interesting and fundamental kind of knowledge. Donnellan’s skepticism is not fully justified because, as we saw, he leaves open two loopholes in his account, one of them explored by Jeshion, but the other overlooked by her (and the second hints at something similar to my treatment in Chap. 10). I presented reasons to be less than fully convinced by some of the most prominent formulations of the experience requirement, such as Platinga’s, Salmon’s and Soames’. Among these reasons is the fact that these formulations seem to expect something unreasonable from the perceptual experience with a physical object, i.e., that it is necessary and sufficient to grant epistemic access to an abstract length. Contra these authors, I argued that perceptual experience with an object is neither necessary nor sufficient for that. Salmon briefly considers the risk of some shiftness in the perception of length, but too quickly dismisses the potential problems for his account. Evans’ particular strategy is meant as an ad hominen argument directed especially at Donnellan, but it depends on the acceptance of the very peculiar category of descriptive names © Springer Nature Switzerland AG 2022 M. Ruffino, Contingent A Priori Truths, Synthese Library 443, https://doi.org/10.1007/978-3-030-86622-8_12

211

212

12 Global Conclusions

(which are expressions semantically different from proper names in any of the classic frameworks). Donnellan would not be willing to accept these names (neither, I suspect, would Kripke). Evans’ general strategy depends on the interesting distinction between superficially and deeply contingent truths, but its motivation and application remain less than clear in Evans’ own work. His strategy also uses a less than fully clear distinction between epistemic content and proposition, which seems to detach the epistemic import of a sentence from its modal behavior. Evans’ distinction becomes considerably clearer in Davies and Humberstone’s technical formulation. Notwithstanding the limitations of Evans’ own discussion, several attempts have been made to show that there are deeply contingent a priori truths in Evans’ sense (and contrary to his own hypothesis that all examples of this kind contains implicit uses of indexicals). Some of them (e.g., Williamson’s The Believer example) were received with skepticism concerning its indexical-free status. But even if one is suspicious of these cases, there are many others (such as what I call LDO-valid sentences) that are much more plausibly seen as indexical-free (at least according to Kaplan, who “discovered” them). The same reasoning behind LDOvalidity could be extended to save Williamson’s example (provided we extend the conception of adequate contexts of utterances, in some natural way, to accommodate adequate contexts of beliefs). Similarly, the same reasoning could be hiding behind BonJour’s justification of a general principle of induction (which would be a clear example of a deeply contingent a priori truth in Evans’ sense). One could perhaps be less than fully satisfied with the general LDO-validity strategy, at least if it is meant to establish the existence of deeply contingent a priori truths, if one takes into account that Kaplan’s semantics is meant to capture the semantic properties of indexicals, and this semantics has LDO-valid sentences as side effect. Moreover, the semantics depends on the restriction to proper contexts. The declared motivation for this restriction is to preserve the intuitive truth of sentences like ‘I am here now’ and ‘I exist’ in any context. Hence, contingent a priori truths in Kaplan’s framework are, in a way, the product of a methodological decision. Concerning two-dimensional semantics, we do not need to share Stalnaker’s more radical skepticism concerning the possibility of representing a priori truths in general. (As we saw, for him, even standard cases of truths such as ‘2 + 2 = 4’ come out false in some cell of the diagonal.) But the capacity to represent such truths in the epistemic interpretation (in Chalmers’ most elaborated version) depends on idealizations such as a subject’s capacity of operating a complete division of the epistemic space based on a clear and previous understanding of what is a priori excluded or not. It also depends on the assumption of the existence of a neutral description of epistemic possibilities, which ultimately ends up in Ramsey-like sentences with complete world descriptions. Many people have expressed skepticism about the latter assumption. Given the way that contingent a priori and necessary a posteriori truths were introduced in the agenda of analytic philosophy in the twentieth century, i.e., through Kripke’s discussion of the consequences of rigidity of proper names and natural kind terms, one often has the feeling that the two phenomena are two faces of the same coin, and that any systematic account of one of them must, at the same time, lead to an account of the other. This feeling was reinforced by Evans’

12 Global Conclusions

213

treatment of deep and superficial contingency on the same level as (and as the dual phenomena of) superficial and deep necessity. The same spirit is intrinsic to the two-dimensional approaches, in which we find an attempt of a unified account of both sorts of phenomena (although, as we saw, Stalnaker gives up any hope of modeling a priori truths in the two-dimensional framework in his later writings on the subject, while retaining the goal of modeling necessary a posteriori truths). If, however, we consider both phenomena from the illocutionary point of view, we can have a more accurate perspective and treat them as belonging to two distinct classes. This point of view, which I take to be the most substantive contribution of the present book, is developed in Chaps. 9 and 10. The approach here outlined does not regard as imperative that an adequate treatment of contingent a priori truths must, at the same time, yield a treatment of necessary a posteriori truths. This is so because contingent a priori truths, at least of the kind isolated by Kripke, involve stipulations, a special kind of speech act with special consequences, while necessary a posteriori truths are usually not the product of this kind of speech acts, but are the basis of ordinary assertions, and assertions have no creative effects. In the case of necessary a posteriori truths, what makes them both necessary and a posteriori are particular features of their propositional content, and the illocutionary force plays no fundamental role in the sense that an unasserted necessary a posteriori truth is still necessary and a posteriori. But in the case of contingent a priori truths, at least in those envisaged by Kripke, the illocutionary force plays a foundational role in the sense that a contingent proposition that is not stipulated is not a priori either. There is, as we saw, a relatively vast literature discussing issues under the label “contingent a priori”. But the approach developed in this book does not recognize a uniformity, a “natural kind” of phenomena that could be unified under the category of contingent a priori truths (as there is, e.g., a natural kind of electric phenomena or a natural kind of processes called “combustion”). One of the points of the present study is that the source of contingency and apriority in cases like Kripke’s is very different from the source in the case of indexicals, to the point that they have to be distinguished as quite different phenomena. (It was perhaps misleading for Kaplan to discuss the special nature of ‘I am here now’ in parallel terms to Kripke’s treatment the (M) sentence, for in the former we do not have the operation of any performative verb, while in the latter we must have, otherwise the example does not get off the ground.) Another moral to be drawn from the discussion in this book is that the phenomenon of contingent a priori truths is much more widespread than Kripke perhaps thought, and is not restricted to some few exotic cases. In fact, many successful declarative illocutionary acts generate contingent facts that are knowable a priori by speakers. Indeed, our social and political institutions are replete of facts thus generated and that are known on the basis of the declarations themselves. They are also not reducible, as some have claimed, to the phenomenon of indexicality. I am not suggesting a reductive treatment, i.e., that all cases of contingent a priori truths considered in the literature (many of them discussed throughout the book) would be reducible to cases explainable via declarative illocutionary acts. This would be a much bolder claim. We would have to, e.g., subsume cases

214

12 Global Conclusions

involving indexicality under cases involving performativity.1 Nor am I saying that all successful performative utterance generate contingent a priori truths in the interesting sense that Kripke was after. As Searle (1989, p. 549) remarks, many performative utterances such as I order that P I promise that P only generate the linguistic fact that an order, a promise, etc. exists. Although these are truths created by performative utterances, they are not the sort of truths that Kripke is concerned with. This is in contrast to some other special performative utterances which, if backed by the appropriate authority on the speaker’s part (and possibly other social conventions), may generate institutional facts, some of which are at the basis of our cognitive lives (such as a measurement system, which makes the scientific practice possible) and social lives (such as a system of laws).2

References Hintikka, J. (1962). Cogito, ergo sum: Inference or performance? The Philosophical Review, 71(1), 3–32. Reinach, A. (1913). Die apriorischen Grundlagen des bürgerlichen Rechtes. Reprinted and translated by Crosby, J. as “The A Priori Foundations of the Civil Law”, Aletheia, III, 1983, pp. 2–142.

1 I shall not enter this discussion in the present book, for it would require a much deeper study on the epistemology and semantics of indexicals than the one offered here. But I shall notice that some suggestion has been made in this direction (e.g., Hintikka, 1962). 2 See, e.g., Reinach (1913) for an early attempt to connect the notion of civil law with a priori knowledge based on performative utterances.

Index

A Abstract, 4, 15, 17, 18, 24, 29, 40, 56, 57, 59–62, 71, 74, 80, 82–84, 86, 87, 108–110, 121, 143, 170, 211 Abstraction, 113, 118, 135, 158, 180, 181, 197 Abuse(s), 166 Acquaintance, xii, 29, 35–49, 56, 58, 61, 70, 71, 74, 79–87, 95 Actual, ix, 2, 4, 5, 7, 9, 13, 14, 25, 27, 49, 67, 97–101, 103, 107, 110–114, 121–133, 136, 138, 143, 145–147, 153, 175 Actuality operator, 103, 104, 124, 143, 146 Almog, J., 137 Almost A Priori, 53, 57–63 Analytic, 130, 199, 201, 206–208, 212 Angelesium, 41 Anscombe, G.E.M., 170 Anti-descriptivist, 166 A posteriori, ix–xi, 1–3, 15, 17, 28, 39, 41, 42, 53–57, 59, 60, 62, 63, 66, 68, 69, 71, 78, 79, 85–87, 102, 104, 107, 108, 113–119, 125–128, 130, 133, 134, 137, 138, 180, 184, 185, 187, 189, 194, 195, 204–206, 212, 213 Aristotle, 3, 7–14, 54, 78, 132 Assertion, xi, xii, 47–49, 108–122, 133, 136, 137, 158–160, 165, 169, 174, 175, 182, 184, 186, 189, 193, 194, 199–202, 206, 207, 213 Assertive(s), 115, 116, 160, 161, 199–201 Atomic, x, 41, 92, 151, 154 Attributive use, 80, 83 Austin, J., xii, 47, 157, 158, 161, 163–169, 174, 175, 181, 182, 186, 193, 194, 196

B Bach, K., 169 Baptizer, xi, xii, 39–41, 80, 180, 184, 193–195 Begriffsschrift, 87, 199–203, 206, 207 Behabitive(s), 161, 166 Belief, xii, 1, 14, 29, 36, 37, 41–45, 47, 53, 58, 60–62, 69–73, 108, 120–122, 130, 136, 143–148, 154, 159–161, 182, 212 Belief attribution(s), 14, 120–122 Believer, The, 143–148, 154, 212 Belnap, N., 187 Berkeley, G., 52, 59 Best explanation, 148–152, 154 Bismarck, 79 Blackburn, S., 39 Black, M., 167 BonJour, L., 62, 148–154, 184, 185, 212 Braun, D., 17, 20 Brute fact(s), 173, 209 Burge, T., 16, 21

C Canonical specification, 133 Carnap, R., 109, 126, 196 Caveman, 63 Chalmers, D., ix, 107, 108, 119, 126–138, 212 Character, x, 19–21, 23, 24, 28–31, 66–70, 72, 73, 104, 108, 127, 129, 131–137, 142, 193 Chinese spy, 70–72 Chisholm, R., 180 Church-Langford, 79 Circumstance of evaluation, 21–23, 108, 132, 142

© Springer Nature Switzerland AG 2022 M. Ruffino, Contingent A Priori Truths, Synthese Library 443, https://doi.org/10.1007/978-3-030-86622-8

215

216 Clinton, H., 27 Cluster, 3, 8, 83, 86 Cognitive content, 73, 89, 90, 97–100, 104, 130, 138 Cognitive significance, 28, 29, 68, 72 Cole, J., 208 Collective intentionality, 171–173, 194, 195 Commissive(s), 160, 161, 166, 169, 202 Constitutive rule(s), 171–173 Context of utterance, x, 19, 21–25, 27, 32, 96, 97, 108, 127, 128, 131, 132, 137, 147, 152–154, 212 Context set, 109, 110, 115–118, 121, 122 Context shift, 53, 63–65 Contextual interpretation, 129 Contingency(ies), x–xii, 28, 31, 40, 41, 45–49, 66, 85–86, 89–105, 115, 119, 122, 125, 133, 189–192, 213 Contingent truth(s), xi, 1, 9, 25, 28, 41, 45–47, 57, 90, 96, 97, 102–104, 123, 141, 144, 147, 148, 181, 191, 196, 212 Convention, 4, 9, 49, 95, 123, 150, 163, 181, 196, 204, 214 Counterfactual, 4, 7, 11, 21, 23, 26, 107, 114, 127, 135 Creative, 161–163, 175, 209, 213 Crossley, J., 124

D Dagger operator, 118 Davies, M., x, 103, 107, 122–126, 137, 212 Deal, A.R., 21 Declaration(s), 47, 48, 162, 163, 169–172, 183–188, 190–192, 194–197, 213 Declarative, xi, 47–49, 161–163, 168–171, 173–175, 182–189, 191–194, 196, 201, 202, 206, 208, 213 Deeply contingent, 96, 97, 103, 104, 123–125, 141–143, 145, 147, 149, 212 Deeply necessary, 103, 123–126 Definite description(s), xi, 2–4, 6–14, 24, 26, 27, 30, 36, 38, 79, 80, 83, 85, 89, 91, 100, 103, 143, 183 Degree of strength, 158, 159 Demonstration(s), 17, 18, 20, 22, 23, 26, 27, 30, 80, 82 Demonstrative(s), x, xi, 17–20, 22–24, 27, 30, 32, 42, 66, 80, 131, 135, 141–143 De re, x, xii, 35–39, 42–45, 47, 48, 51, 53, 60, 61, 70–74, 81, 182 Derived context, 121, 122 Descartes, R., 25, 28

Index Descriptive, 9, 10, 13, 24, 35, 36, 43–46, 51, 65, 73, 74, 77, 79–82, 84, 89, 90, 93, 102, 103, 126, 134, 167, 168, 182 Descriptive content(s), 13, 27, 39, 80, 81, 84, 89, 93, 103, 126, 132 Descriptive name(s), x, 89–95, 98, 99, 101, 103, 211 Descriptivist, 10, 12, 65, 166–168 Diagonal, 112–115, 118–121, 127–129, 132, 134–138, 212 Diagonalization, 120–122, 134 Direction of fit, 159–161, 168, 182 Directive(s), 145, 160, 161, 169, 201, 202 Donnellan, K., x, xii, 8–10, 14, 15, 35–49, 51, 55, 56, 60, 61, 77–85, 89, 90, 95, 103, 190–192, 196, 211, 212 Donnellan’s Problem, 42–45 Dthat, 26–28, 30, 38, 67–74, 80, 81, 85–87 Dummett, M., 8–12, 14, 36, 89, 97, 207

E Edmondson, W., 167, 168 Empirical, ix, 2, 5, 15, 21, 30, 42, 53, 59, 63, 71, 78, 89, 103, 128, 130, 137, 138, 165, 179, 184–189, 195, 204, 206 Enutpen, 37, 38, 42 Epistemic interpretation, 129, 212 Epistemic possibilities, 129, 131, 133, 138, 212 Evans, G., x, xii, 4, 6, 15, 17, 89–105, 122, 123, 125, 126, 137, 141–143, 145, 153, 192, 193, 211, 212 Exercitive(s), 166 Experience, 4, 13, 15, 16, 29, 38, 51–74, 79, 85, 86, 120, 148, 149, 151, 175, 179, 184, 185, 188, 189, 211 Explainer, The, 149, 150, 154 Explanation(s), 2, 15, 28–30, 35, 39, 42–44, 53, 69, 96, 118, 120, 124, 135, 138, 148–154, 168, 179, 192, 196, 205 Expositive(s), 166, 186 Expressive(s), 161, 162, 201 Extension, 19–21, 24, 28, 52, 90, 107, 111, 112, 127, 128, 130–132, 135, 142, 168

F Felicitous, 166, 174, 182 Felicity conditions, 48, 168, 174, 182, 194 Fine, K., 105 Fixedly operator, 125 Free logic, 6, 93–95, 102

Index Fregean, 3, 8, 19–21, 28, 44, 53, 79, 91, 103, 108, 126, 129, 132, 134 Frege’s Puzzle, 43–45, 48, 49, 53, 54, 57, 105 Freund, M., 92 G García-Carpintero, M., 169 General Conference of Weights and Measures (GCWM), 183, 184, 189, 194 Ginet, C., 169 God, 25, 49, 90, 145, 151, 190, 191 Goldbach’s conjecture, 1–3 Golden triangle, 126 Grandpa Joe, 44 Green, M., 157 Gricean, 55, 108, 115, 118, 163, 169 Grundgesetze der Arithmetik, 201 Guise(s), 43–45, 47 H Hardy, G., 208, 209 Harman, G., 74 Harnish, R., 169 Harris’ dilemma, 168 Harris, R., 166–168 Hartnack, J., 167 Hawthorne, J., 41, 74, 95, 104, 144, 149–151, 154, 192 Hedenius, I., 167, 169 Hereby, 158, 162, 164, 165, 191 Hesperus, ix, x, 2, 30, 43, 47, 105, 108, 118, 120, 121, 126, 204, 206 Hilbert, D., 208 Hintikka, J., 214 H2 O, ix, 107, 111, 112, 114, 123, 127, 128, 130, 131, 137 Horizontal, 110, 200 Horowitz, T., xii, 182, 188–189, 195 Humberstone, L., 103, 107, 122–126, 137, 212 I Identifiability requirement, 82, 83, 85, 86 Identity, ix–xi, 2, 4, 5, 7, 27, 30, 45, 47–49, 73, 82, 83, 98, 120, 121, 125, 126, 130, 199, 202–207 Identity of content, 202–206 Illocutionary act(s), xi, xii, 47–49, 115, 157–163, 165, 166, 168–170, 174, 175, 180–197, 199, 208, 209 Illocutionary commitment, 174–175, 183, 192, 194, 197 Illocutionary force, 157–159, 165, 166, 174, 182, 192, 196, 199–201, 208, 213

217 Illocutionary inconsistency, 174–175 Illocutionary point, 47, 110, 115–117, 158–162, 168, 174, 184, 186, 188, 191, 192, 194, 207, 213 Induction, 148–150, 152–154, 212 Inductive premise, 148, 149 Inductive principle, 149, 152 Infelicities, 166, 182 Information, 15, 54–55, 58, 63, 71, 83, 108, 119, 133, 179, 180, 187–189, 195, 206 Institutional fact(s), 60, 170–174, 192, 194–196, 209, 214 Instrumental, 29, 38, 74, 206 Instrumentalism, 29 Intension, 18, 19, 21, 66, 90, 100, 107, 111–113, 115, 119, 126–133, 135, 136, 138 Intention, 10, 17, 22, 92, 108, 117, 160, 163, 166, 181, 184, 185, 187, 188, 195 Intuition(s), 2, 7, 8, 10, 11, 14, 16, 17, 24, 36, 67, 92, 122, 148, 175, 180 J Jackson, F., ix, 107, 108, 126, 127, 129, 133–135, 138 Jary, M., 167, 169 Jeshion, R., xii, 29, 35, 39–48, 74, 182, 211 Jubien, M., 13 Judgeable content, 199, 200, 206, 207 Julius, 89–95, 102, 103, 143 Jupe, 41 K Kant, I., 62, 126, 130, 180, 185, 203–205 Kaplan, D., x, xi, 17–32, 38, 65–74, 80, 96, 97, 103, 104, 107, 108, 111, 116–118, 127, 129, 131–135, 141, 142, 147, 148, 152, 212, 213 Kempson, R., 167 Kennedy, R., 58 Kent, C., 54, 58 Kilogram, 87, 183, 184, 189, 194 Knowledge first, 144 L Lane, L., 54, 58 Language, x, xi, 1, 3, 12, 16, 21, 26, 29, 32, 49, 51, 57, 60, 63, 64, 66, 67, 69, 82, 85, 87, 89, 105, 120–122, 124, 128, 131, 136–138, 158, 163, 170, 171, 173, 181, 183, 187, 188, 191, 193, 197, 199–202, 205, 207, 208

218 Language game, 3, 60, 63, 158, 197 LDO-valid, 99, 142, 143, 147, 148, 153, 154, 212 Lee, 71, 82 Leech, G., 167 Le Grand K, 87, 183 Lemmon, E., 167 Length(s), 3, 4, 15, 40, 51–54, 56–61, 63, 64, 71, 74, 80–87, 114, 123, 173, 174, 180–186, 189, 190, 194, 195, 197, 211 Le Verrier, U., 5, 6, 15, 37–40, 42, 79, 81–83, 86, 100, 180–182, 190, 191, 195 Levin, M., 35, 79 Levinson, S., 165 Lewis, D., 10, 13, 127, 167, 169 Lewisian, 108 Linguistic fact(s), 170–174, 181, 214 Locutionary act(s), 158, 165, 181 Logical space, 120, 128, 129, 153 Logical truth(s), 12, 27, 30, 66–70, 98, 196, 206 Logic of demonstratives (LD), 22, 66, 67, 141–143, 152 Logic of indexicals, 31 Loophole, 40, 42–49, 211

M Manley, D., 41, 74, 95 Martone, F., 29 Mathematical definition(s), 63, 185–188 Matrix, 110–114, 117–120, 127, 128, 136, 138 Meaning, 3, 5, 7, 12, 19–21, 28–31, 57, 66, 67, 82, 95, 107, 108, 115, 123, 126, 132, 134–136, 138, 163–165, 168, 183, 184, 194, 196, 201 Measurement, xi, 3, 4, 40, 53, 56–60, 63, 64, 74, 81, 85, 87, 179, 183, 184, 186, 189, 194, 214 Metaphysical, 1, 4, 7, 123 Metasemantic(s), 119, 131, 132, 134–137 Millian, 41, 44, 53, 58 MNT, 111, 112, 114, 127, 128 Modal argument, 7–11, 13, 21, 23–24, 89 Modality, x, 1, 9, 28, 41, 115 Mode of achievement, 159 Monsters, 21, 32, 90 Most unlikely, 150–152, 154

Index N Naming and Necessity, ix, x, xii, 1–3, 7, 11, 77, 79–83, 85, 86, 197 Narrow scope, 9, 11, 101, 102 Necessary truth(s), ix, 1–3, 5, 12, 27, 30, 41, 48, 93, 98, 101, 114, 115, 117, 118, 123, 124, 126, 136, 137, 142, 145, 148–150, 162, 186, 189, 190 Necessity, ix–xii, 1–3, 7, 11, 17, 26, 28, 41, 77, 79–83, 85, 86, 103, 122–127, 129, 133, 137, 142, 151, 152, 175, 197, 199, 213 Nelson, M., 14 Neptune, xi, 5, 6, 10, 37, 39–49, 61, 77–79, 81–86, 90, 93–95, 98, 99, 101, 102, 181, 182, 190–192, 195, 196 Neutral language, 120 Newman I, 38–40, 42, 43, 45 Nicomachean ethics, 3, 7–10, 132

O Obama, 112, 113 Object(s) of thought, 28–30, 67–70, 135 Oldman I, 44 Oppy, G., 56, 145–148

P Paderewski, 105 Pagin, P., 169 Peano, G., 200 Performative(s), 48, 157, 163–172, 179–197, 201, 202, 213, 214 Performative hypothesis, 165 Performative verb(s), 157, 164–166, 168–172, 182, 184, 186, 213 Perry, J., 17, 20, 28, 69, 72, 73, 117, 202 Pharaoh, 84 Philonous, 51–53, 56, 59, 61, 64, 71, 74 Phosphorus, ix, x, 2, 30, 43–45, 47, 48, 105, 108, 118, 126, 204, 206 Plantinga, A., xii, 13, 51–53, 55, 56, 59, 61, 83, 87, 195 Possible world(s), ix, 2, 4, 6, 7, 13, 14, 19, 22–26, 30, 31, 66, 90, 93, 96–101, 104, 107, 109–133, 135, 136, 142–144, 148–150, 152, 154 Pragmatically imparted, 54, 55, 58 Predelli, S., 32, 154 Preparatory condition(s), 49, 62, 158, 162, 166, 171, 184–189, 191, 195, 202, 206, 207 Presupposition(s), 10, 109, 116, 121, 200 Primary intension, 127–133, 135, 136, 138

Index Proper context(s), 24–26, 31, 32, 36, 97, 142, 147, 148, 152–154, 212 Proper name(s), x, xi, 2–4, 7, 9–13, 17, 19, 29, 35, 36, 53, 54, 58, 71, 79, 83, 90, 103, 132, 134, 143, 212 Propositional attitude, 37, 41, 51, 58, 68, 70–74, 108 Propositional concept, 108, 110, 111, 113–115, 117–122, 134–136 Propositional content, 47–49, 60, 62, 96, 104, 110, 111, 113, 134, 158–163, 168–171, 174, 182–184, 186–190, 192–196, 199, 200, 202, 206–208, 213 Psychological condition, 159 Pure indexical(s), x, xi, 17–20, 23–25, 32, 96, 97, 103, 141, 193 Putnam, 65, 136, 188 Puzzle, x, xii, 43–45, 48, 49, 53, 54, 57, 63, 64, 89, 93, 95, 99, 102, 105, 137 Pythagorean theorem, 185

Q Quine, W., 70, 127, 196

R Rabern, B., 21 Rajagopalan, K., 168 Ramsey sentences, 133, 212 Rational communication, 115–119 Rational insight, 148, 150 Recanati, F., 111, 117, 167, 169 Reference-fixing, x, 5, 7, 9, 10, 24, 31, 35, 36, 40, 42–46, 48, 51, 66, 73, 74, 77, 79–85, 89, 90, 103, 134, 182, 188 Referential use, 80 Reflexivity, 104 Regulative rule(s), 171, 172 Reinach, A., 175, 214 Rigid, x, 1–5, 7–10, 13, 23, 24, 27, 30, 31, 35–38, 41, 51, 67, 73, 78–83, 85, 89, 90, 93, 95, 101, 103, 111–114, 117, 120, 126, 129, 130, 136, 181, 182 Rigidity, ix, x, xi, 6–11, 13, 14, 35, 36, 71, 90, 103, 129, 211, 212 Rigid name(s), 7, 36–37, 73, 82, 83, 90, 93, 126, 182 Roberts, C., xii, 157, 159 Ross, J., 165 Russell, B., 3, 9, 39, 79, 85, 92, 148, 154 Russellian proposition(s), 19, 54–56, 58, 60, 72, 81, 131

219 S Sadock, J., 165, 166 Salmon, N., xii, 8, 44, 53–65, 74, 77, 79, 83, 85, 87, 189, 195, 211 San Mauro, L., 63, 160, 201, 207 Scenario, 37, 63, 71, 129, 133, 137 Schirn, M., 207 Schlenker, P., 21 Schröder, E., 200 Schroeter, L., 108, 120, 128, 133 Scope, 8–12, 14, 36, 72, 92, 94, 96–99, 101, 102 Scope ambiguity, 8–12 Searle, J., xi, xii, 8, 47, 49, 110, 145, 157–159, 161–163, 165, 166, 168–175, 182, 186, 187, 191, 192, 194–196, 201, 206, 214 Secondary intension, 119, 126–133, 135, 136, 138 Semantically encoded, 54, 55, 58 Semantically neutral, 133 Semantic content, 9, 54, 168 Sense, 3, 7, 9, 10, 13, 15, 19–22, 24, 26, 31, 36, 44, 49, 52, 57–61, 64, 79, 80, 82, 91, 93, 95, 100, 102–105, 113, 115, 116, 118, 120, 126, 128–130, 132, 134, 135, 138, 141–145, 150, 152, 154, 157, 159–161, 167, 174, 175, 179–181, 184, 188–190, 192–194, 196, 197, 199, 202–204, 206–209, 212–214 Seribi, 61 Sincerity conditions, 159–161 Soames, S., xii, 8, 12, 14, 27, 65–74, 108, 193, 195, 211 Sosa, D., 11 Speech act(s), xi, 47–49, 62, 63, 145, 157–175, 184–188, 196, 201, 202, 208, 213 Spielmann, R., 167 Stalnaker, R., ix, x, 15, 107–122, 128, 129, 131–138, 212, 213 Stanley, J., 13 Status function, 171–173, 194, 195 Stipulation(s), ix, xi, 4, 7, 15, 31, 37, 39–42, 45–49, 56–58, 60, 62, 71, 74, 78, 87, 97, 173, 174, 179–197, 205, 206, 211, 213 Stipulator, xi, 4, 15, 16, 35, 41, 42, 44–49, 51–53, 55, 57, 60, 62–65, 74, 173, 187, 188, 194, 195, 211 Stojanovi´c, I., 64 Strong commitment, 174 Superficially contingent, 96, 97, 103, 104, 122–125, 141, 143, 153 Superficially necessary, 103, 122–126

220 Sutton, J., 41 Synthetic, 130, 199, 202–208 T Taschek, W., 121 Taxonomy, xii, 157, 159, 165, 166, 201 Taylor, T., 167, 168 Thought(s), x, 6, 22, 25, 28–30, 35, 37, 41, 61, 63, 66–70, 73, 74, 82–84, 103, 105, 120, 127, 130, 135, 147, 164, 166, 169, 191, 195, 199, 200, 207, 208, 213 Trump, D., 26, 27, 205 Truth-maker(s), 15, 179, 181, 192, 193, 196 Turri, J, 150–154 Twin-earthable, 128, 133 Two-dimensionalism, 74, 107–138 Two-dimensional semantics, ix, xii, 65–74, 107, 110–113, 129, 133, 137, 138, 212 V Vagueness, 42, 84, 130, 183 Valid, 12, 68–70, 94, 142–144, 146, 148, 153, 195 Validity, 17, 26, 66, 67, 141, 142, 147, 152 Vanderveken, D., xi, xii, 47, 110, 145, 157–159, 161, 162, 166, 168, 169, 174,

Index 175, 182, 186, 189, 194, 196, 200, 201, 206 Venturi, G., 63, 160, 201, 207 Verdictive(s), 165 Vertical, 110, 200 Vision, G., 32l Visual, 52, 53, 56–63, 83, 84, 87, 120 Vulcan, 6, 93

W Water, ix, 29, 64, 80, 107, 111, 114, 116, 117, 121, 123, 127, 130–132, 137 Weak commitment, 174 Wettstein, H., 17 Wide scope, 9, 11, 36, 96, 97, 102 Wiggins, D., 167 Williamson, T., 26, 143–148, 154, 212 Wittgenstein, L., 3, 53, 63, 65, 86, 116, 158, 197 Wolf, G., 167, 168

X XYZ, 111, 112, 114, 123, 127, 128, 130